Handbook of Human Motion [1 ed.] 9783319144177, 9783319144184

Citation preview

Bertram Müller Sebastian I. Wolf Editors-in-Chief Gert-Peter Brüggemann · Zhigang Deng Andrew S. McIntosh · Freeman Miller W. Scott Selbie  Section Editors

Handbook of Human Motion

Handbook of Human Motion

Bertram Müller • Sebastian I. Wolf Editors-in-Chief

Gert-Peter Brüggemann • Zhigang Deng Andrew S. McIntosh • Freeman Miller W. Scott Selbie Section Editors

Handbook of Human Motion With 522 Figures and 68 Tables

Editors-in-chief Bertram Müller Motion and More Barcelona, Spain

Section Editors Gert-Peter Brüggemann Institute for Biomechanics und Orthopedics German Sport University Cologne Cologne, Germany

Sebastian I. Wolf Clinic for Orthopedics and Trauma Surgery Center for Orthopedics Trauma Surgery and Spinal Cord Injury Heidelberg University Hospital Heidelberg, Germany Zhigang Deng Department of Computer Science University of Houston Houston, TX, USA

Freeman Miller Andrew S. McIntosh Wilmington, Delaware, USA McIntosh Consultancy and Research Sydney, NSW, Australia Australian Collaboration for Research into Injury in Sport and its Prevention (ACRISP) Federation University Australia Ballarat, VIC, Australia Monash University Accident Research Centre Monash University Melbourne, VIC, Australia W. Scott Selbie HAS-Motion Inc. Kingston, ON, Canada C-Motion Inc. Germantown, MD, USA ISBN 978-3-319-14417-7 ISBN 978-3-319-14418-4 (eBook) ISBN 978-3-319-14419-1 (print and electronic bundle) https://doi.org/10.1007/978-3-319-14418-4 Library of Congress Control Number: 2017957051 # Springer International Publishing AG, part of Springer Nature 2018 This work is subject to copyright. All rights are reserved by the Publisher, whether the whole or part of the material is concerned, specifically the rights of translation, reprinting, reuse of illustrations, recitation, broadcasting, reproduction on microfilms or in any other physical way, and transmission or information storage and retrieval, electronic adaptation, computer software, or by similar or dissimilar methodology now known or hereafter developed. The use of general descriptive names, registered names, trademarks, service marks, etc. in this publication does not imply, even in the absence of a specific statement, that such names are exempt from the relevant protective laws and regulations and therefore free for general use. The publisher, the authors and the editors are safe to assume that the advice and information in this book are believed to be true and accurate at the date of publication. Neither the publisher nor the authors or the editors give a warranty, express or implied, with respect to the material contained herein or for any errors or omissions that may have been made. The publisher remains neutral with regard to jurisdictional claims in published maps and institutional affiliations. Printed on acid-free paper This Springer imprint is published by the registered company Springer International Publishing AG, part of Springer Nature. The registered company address is: Gewerbestrasse 11, 6330 Cham, Switzerland

Foreword

The field of human locomotion biomechanics has developed drastically in the last 50 years. I remember when I first began in biomechanics in the early 1970’s how underdeveloped the field of human motion analysis and the related methodologies were. There were very few places in the world that were involved in this field of study, and we had to develop appropriate methodologies often from scratch. I remember a discussion with experts as whether or not the impact forces measured with force plates were artifacts or real. I also remember the high-speed film cameras and the complicated and time-consuming film analysis. Today, human movement analysis is a highly developed field with many facets, excellent high-technology equipment, highly sophisticated mathematical methods of data analysis, and thousands of experts working all over the world in many different fields of applications. Additionally, human movement analysis is instructed in almost all universities in many different faculties (e.g., kinesiology, engineering, medicine, injury prevention, etc.). The Handbook of Human Motion reflects these developments and provides an introduction to the subject, as well as a description of the current state-of-the-art technologies and a vision for the future of the field. The various sections provide a broad representation of the different fields of application of movement analysis and allow cross-fertilization between these fields. Furthermore, the Handbook of Human Motion provides a wealth of information from outstanding experts in the field that can be used to explore this innovative field of study. The study of human movement is exciting and will develop in the next few decades even further using wearable sensors and chips that provide real-time information to the athlete and patient. The principal work of the human movement analyst is to provide appropriate interpretation of the wealth of generated results. This handbook helps in contributing to this development. Calgary, Summer 2017

Benno Nigg

v

Preface

In the beginning, the science of human movement was represented by just a few outstanding personalities. Nowadays, this is no longer the case as the number of experts in the field has increased exponentially. With this increase, the field has diversified into many different scientific areas and specialities. Working within different fields but meeting up regularly at conferences related to human motion capture, we both realized that there may be a huge amount of knowledge that we were not even aware of. Specialized textbooks are available for all of the different disciplines, but to our knowledge there is, as yet, no compendium that serves to overarch them all. Therefore, in conjunction with five other specialists in different fields and a publishing house that shared our philosophy, we began compiling this handbook. With well over 100 chapters now realized, and with more than 200 authors involved, the handbook has advanced to this first printed edition. This handbook is intended not only for beginners in the field, providing an overview of the different approaches, but also for experts in a particular area needing information for a related one. Some fields have developed along with the technology, such as animations in the motion picture industry. Others, such as medicine, are introducing it to improve understanding of human movement for diagnosis and treatment. Its application in the field of sports science, for instance, might enhance human performance while diminishing the risk of injury. When legal aspects are involved or when dealing with occupational medicine, the focus of human motion might shift toward other related issues and the subject of forensics. New fields are continually emerging, guided by the creativity of people working with such systems. In each chapter of this book, information is provided on state-of-the-art science, as well as insight into future directions, providing detailed information about each subject. Our hope is that this handbook will not only increase the level of knowledge of the individual reader, but also facilitate understanding between different scientific fields and related areas. The inclusion of authors from around the world not only reflects different approaches but also a wide variation in methodological and linguistic styles. The latter, for instance, is responsible for the transmission of information, which is the basis for communication and understanding in a multicultural society, and can involve imparting many ambiguous terms and concepts. While standardization

vii

viii

Preface

might not be an attainable goal, knowledge of the differences may well enhance interaction and communication between professionals. And finally, the variety of subject matter within this compact handbook might also produce new ideas for one’s own field of study. Interpolating knowledge with information from fields not previously believed to be related might just lead to novel approaches. What it does already provide is evidence of diversity in measurements and their interpretation, which will doubtless lead to improvements in results in the quest for objective data. What we, the editors, have also noted is the rapid development and expansion in the discovery of human motion and its application. While this first physical edition provides extensive information, it is also recommended to look up the online version, where frequent updates can be seen. We are thankful to all of the authors for their contribution to this edition and to their continuing interest in keeping the information up to date. We would also like to invite all of those who find any omissions to contribute to this work in the future. Barcelona/Heidelberg October 2017

Bertram Müller Sebastian I. Wolf

Acknowledgments

Editing this book was for both editors an adventure. We did not know in advance what result to expect since neither of our scientific backgrounds and networks overarched the disciplines presented in this book. We therefore give thanks to our section editors Scott Selbie, Freeman Miller, Gert-Peter Brüggemann, Zhigang Deng, and Andrew S. McIntosh for extending our network and helping in collecting these chapters. Our special thanks go to Manfred Nusseck and Benita Kuni for making the sections of human movement in music and dance possible. Furthermore, we wish to thank all contributing authors for their idealism in sharing their knowledge and dedicating their time to writing a contribution to this book. Finally, we would also like to extend our particular gratitude to the team of Springer with Tom, Lydia, Johanna, and Barbara for supporting us in realizing this overview on human motion throughout the past years.

ix

Contents

Volume 1 Section I Part I

Methods and Models

Rigid Body Modeling . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

1

Observing and Revealing the Hidden Structure of the Human Form in Motion Throughout the Centuries . . . . . . . . . . . . . . . . . . . . . . Aurelio Cappozzo

3

Three-Dimensional Reconstruction of the Human Skeleton in Motion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Valentina Camomilla, Aurelio Cappozzo, and Giuseppe Vannozzi

17

Estimation of the Body Segment Inertial Parameters for the Rigid Body Biomechanical Models Used in Motion Analysis . . . . . . . . . . . . . Raphaël Dumas and Janis Wojtusch

47

Part II

.....

79

3D Dynamic Pose Estimation from Marker-Based Optical Data . . . . . . W. Scott Selbie and Marcus J. Brown

81

Discriminative Methods in Dynamic Pose Estimation

Measurement of 3D Dynamic Joint Motion Using Biplane Videoradiography . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Hans Gray, Shanyuanye Guan, Peter Loan, and Marcus Pandy 3D Musculoskeletal Kinematics Using Dynamic MRI . . . . . . . . . . . . . . Frances T. Sheehan and Richard M. Smith

101 117

Cross-Platform Comparison of Imaging Technologies for Measuring Musculoskeletal Motion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Richard M. Smith and Frances T. Sheehan

135

Ultrasound Technology for Examining the Mechanics of the Muscle, Tendon, and Ligament . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Glen Lichtwark

157 xi

xii

Part III

Contents

Generative Methods in Dynamic Pose Estimation . . . . . . .

3D Dynamic Probabilistic Pose Estimation from Data Collected Using Cameras and Reflective Markers . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Thomas M. Kepple and Alan R. De Asha 3D Dynamic Pose Estimation from Markerless Optical Data . . . . . . . . Steven Cadavid and W. Scott Selbie Three-Dimensional Human Kinematic Estimation Using MagnetoInertial Measurement Units . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Andrea Cereatti, Ugo Della Croce, and Angelo M. Sabatini

177

179 197

221

Gait Parameters Estimated Using Inertial Measurement Units . . . . . . . Ugo Della Croce, Andrea Cereatti, and Martina Mancini

245

Physics-Based Models for Human Gait Analysis . . . . . . . . . . . . . . . . . . Petrissa Zell, Bastian Wandt, and Bodo Rosenhahn

267

Part IV

293

Body Scaling . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

Scaling and Normalization . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . At L. Hof

295

Part V

307

Extended Modeling Techniques . . . . . . . . . . . . . . . . . . . . . . .......

309

Optimal Control Modeling of Human Movement . . . . . . . . . . . . . . . . . Brian R. Umberger and Ross H. Miller

327

Time Series Analysis in Biomechanics . . . . . . . . . . . . . . . . . . . . . . . . . . W. Brent Edwards, Timothy R. Derrick, and Joseph Hamill

349

Hill-Based Muscle Modeling . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Ross H. Miller

373

Induced Acceleration and Power Analyses of Human Motion Anne K. Silverman

Simulation of Soft Tissue Loading from Observed Movement Dynamics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Scott C. E. Brandon, Colin R. Smith, and Darryl G. Thelen

395

...

429

Gait Symmetry Measures and Their Relevance to Gait Retraining Silvia Cabral Part VI

Dynamic Electromyography . . . . . . . . . . . . . . . . . . . . . . . .

Surface Electromyography to Study Muscle Coordination . . . . . . . . . . François Hug and Kylie Tucker

449 451

Contents

xiii

Section II Part VII

Medical Application

Gait Assessment in Clinical Context

.................

471

Clinical Gait Assessment by Video Observation and 2D Techniques . . . Andreas Kranzl

473

The Conventional Gait Model - Success and Limitations . . . . . . . . . . . Richard Baker, Fabien Leboeuf, Julie Reay, and Morgan Sangeux

489

Variations of Marker Sets and Models for Standard Gait Analysis Felix Stief

...

509

Next-Generation Models Using Optimized Joint Center Location Ayman Assi, Wafa Skalli, and Ismat Ghanem

....

527

Kinematic Foot Models for Instrumented Gait Analysis . . . . . . . . . . . . Alberto Leardini and Paolo Caravaggi

547

Trunk and Spine Models for Instrumented Gait Analysis . . . . . . . . . . . Robert Needham, Aoife Healy, and Nachiappan Chockalingam

571

..........

583

Upper Extremity Models for Clinical Movement Analysis Andrea Giovanni Cutti, Ilaria Parel, and Andrea Kotanxis Part VIII

Interpreting Kinetics and EMG in Gait . . . . . . . . . . . . . . .

607

Interpreting Ground Reaction Forces in Gait . . . . . . . . . . . . . . . . . . . . Nachiappan Chockalingam, Aoife Healy, and Robert Needham

609

Interpreting Joint Moments and Powers in Gait . . . . . . . . . . . . . . . . . . L. H. Sloot and M. M. van der Krogt

625

EMG Activity in Gait: The Influence of Motor Disorders . . . . . . . . . . . Dimitrios A. Patikas

645

Part IX

671

Scores and Spatiotemporal Parameters . . . . . . . . . . . . . . .

Gait Scores: Interpretations and Limitations . . . . . . . . . . . . . . . . . . . . . Veronica Cimolin and Manuela Galli

673

Interpreting Spatiotemporal Parameters, Symmetry, and Variability in Clinical Gait Analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Arnaud Gouelle and Fabrice Mégrot

689

Part X

709

Pedobarography . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

Assessing Pediatric Foot Deformities by Pedobarography . . . . . . . . . . . Dieter Rosenbaum

711

xiv

Contents

Assessing Clubfoot and Cerebral Palsy by Pedobarography . . . . . . . . . Julie A. Stebbins

727

Low Density Pedoboragraphy as a Gait Analysis Tool . . . . . . . . . . . . . Ruopeng Sun, Tyler A. Wood, and Jacob J. Sosnoff

741

The Importance of Foot Pressure in Diabetes . . . . . . . . . . . . . . . . . . . . Malindu E. Fernando, Robert G. Crowther, and Scott Wearing

759

Integration of Foot Pressure and Foot Kinematics Measurements for Medical Applications . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Claudia Giacomozzi, Paolo Caravaggi, Julie A. Stebbins, and Alberto Leardini

789

Volume 2 Part XI

Energy Consumption During Gait . . . . . . . . . . . . . . . . . . . .

811

Assessing the Impact of Aerobic Fitness on Gait . . . . . . . . . . . . . . . . . . Annet Dallmeijer, Astrid Balemans, and Eline Bolster

813

Oxygen Consumption in Cerebral Palsy . . . . . . . . . . . . . . . . . . . . . . . . Hank White, J. J. Wallace, and Sam Augsburger

825

The Use of Kinematics for Pulmonary Volume Assessment . . . . . . . . . . Carlo Massaroni

847

Aerobic Capacity and Aerobic Load of Activities of Daily Living After Stroke . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . I. J. Blokland, T. IJmker, and H. Houdijk

863

Part XII

885

Gait and Balance . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

Measures to Determine Dynamic Balance . . . . . . . . . . . . . . . . . . . . . . . Timothy A. Niiler

887

Slip and Fall Risk Assessment . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Feng Yang

915

Detecting and Measuring Ataxia in Gait . . . . . . . . . . . . . . . . . . . . . . . . Mariano Serrao and Carmela Conte

937

Gait During Real-World Challenges: Gait Initiation, Gait Termination, Acceleration, Deceleration, Turning, Slopes, and Stairs . . . . . . . . . . . . Michael Orendurff Gait Retraining for Balance Improvement . . . . . . . . . . . . . . . . . . . . . . . Robert G. Crowther and Jessica May Pohlmann

955 977

Contents

Part XIII

xv

Pathoanatomy and Diagnostics in Cerebral Palsy . . . . . .

Diagnostic Gait Analysis Use in the Treatment Protocol for Cerebral Palsy . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Freeman Miller and John Henley

987

989

Walking and Physical Activity Monitoring in Children with Cerebral Palsy . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1005 Kristie F. Bjornson and Nancy Lennon Spasticity Effect in Cerebral Palsy Gait . . . . . . . . . . . . . . . . . . . . . . . . . 1037 Marlene Cristina Neves Rosa and André Gonçalo Gomes Roque Natural History of Cerebral Palsy and Outcome Assessment . . . . . . . . 1053 Erich Rutz and Pam Thomason Skeletal Muscle Structure in Spastic Cerebral Palsy . . . . . . . . . . . . . . . 1075 Adam Shortland Part XIV

Movement Deviations in Cerebral Palsy

.............

1091

Swing Phase Problems in Cerebral Palsy . . . . . . . . . . . . . . . . . . . . . . . . 1093 Ana Presedo Strength Related Stance Phase Problems in Cerebral Palsy . . . . . . . . . 1109 Justin Connor and Mutlu Cobanoglu Foot and Ankle Motion in Cerebral Palsy . . . . . . . . . . . . . . . . . . . . . . . 1121 Jon R. Davids and Sean A. Tabaie The Arm Pendulum in Gait . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1151 Jaques Riad Upper Extremity Movement Pathology in Functional Tasks . . . . . . . . . 1167 Lisa Mailleux, Cristina Simon-Martinez, Hilde Feys, and Ellen Jaspers Part XV

Other Neurologic Gait Disorders . . . . . . . . . . . . . . . . . . . .

1185

Idiopathic Toe Walking . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1187 Karen Davies, Lise Leveille, and Christine Alvarez Gait Disorders in Persons After Stroke . . . . . . . . . . . . . . . . . . . . . . . . . 1205 Johanna Jonsdottir and Maurizio Ferrarin Hereditary Motor Sensory Neuropathy: Understanding Function Using Motion Analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1217 Sylvia Õunpuu and Kristan Pierz Motor Patterns Recognition in Parkinson’s Disease . . . . . . . . . . . . . . . 1237 Pierpaolo Sorrentino, Valeria Agosti, and Giuseppe Sorrentino

xvi

Contents

Gait and Multiple Sclerosis James McLoughlin

. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1253

Functional Dystonias . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1267 Jessica Pruente and Deborah Gaebler-Spira Part XVI

Traumatic and Orthopedic Gait Disorders . . . . . . . . . . . .

1281

Gait Changes in Skeletal Dysplasia . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1283 William G. Mackenzie and Oussama Abousamra Impact of Scoliosis on Gait . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1289 Elizabeth A. Rapp and Peter G. Gabos Concussion Assessment During Gait . . . . . . . . . . . . . . . . . . . . . . . . . . . 1307 Robert D. Catena and Kasee J. Hildenbrand Functional Effects of Ankle Sprain Ilona M. Punt and Lara Allet Part XVII

. . . . . . . . . . . . . . . . . . . . . . . . . . . . 1325

Orthotics and Prosthetics in Gait . . . . . . . . . . . . . . . . . . .

1341

Prosthetic Foot Principles and Their Influence on Gait . . . . . . . . . . . . . 1343 Andrew Hansen and Felix Starker The Influence of Prosthetic Knee Joints on Gait . . . . . . . . . . . . . . . . . . 1359 Steven A. Gard Influence of Prosthetic Socket Design and Fitting on Gait . . . . . . . . . . . 1383 Arezoo Eshraghi and Jan Andrysek Functional Effects of Foot Orthoses . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1407 Christopher Nester Functional Effects of Shoes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1423 Benedicte Vanwanseele Gait Rehabilitation with Exoskeletons . . . . . . . . . . . . . . . . . . . . . . . . . . 1433 Stefano Federici, Fabio Meloni, and Marco Bracalenti Brain-Computer Interfaces for Motor Rehabilitation . . . . . . . . . . . . . . 1471 Rüdiger Rupp Part XVIII

Gait After Joint Replacement . . . . . . . . . . . . . . . . . . . . .

1503

Effects of Total Hip Arthroplasty on Gait . . . . . . . . . . . . . . . . . . . . . . . 1505 Swati Chopra and Kenton R. Kaufman

Contents

xvii

Effects of Knee Osteoarthritis and Joint Replacement Surgery on Gait . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1521 Cheryl L. Hubley-Kozey and Janie Astephen Wilson The Effects of Ankle Joint Replacement on Gait . . . . . . . . . . . . . . . . . . 1551 Justin Michael Kane, Scott Coleman, and James White Brodsky Shoulder Joint Replacement and Upper Extremity Activities of Daily Living . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1563 Hendrik Bruttel, David M. Spranz, Jan M. Eckerle, and Michael W. Maier

Section III Part XIX

Arts and Human Performance

Sports

.........................................

Sprint Running: Running at Maximum Speed Michiyoshi Ae

1581

. . . . . . . . . . . . . . . . . . . 1583

Running Shoes: Injury Protection and Performance Enhancement . . . 1613 Steffen Willwacher Landings: Implications for Performance . . . . . . . . . . . . . . . . . . . . . . . . 1629 Laura A. Held, Henryk Flashner, and Jill L. McNitt-Gray Airborne Movements: Somersaults and Twists . . . . . . . . . . . . . . . . . . . 1661 Maurice R. Yeadon Ski Jumping: Aerodynamics and Kinematics of Take-Off and Flight . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1681 Mikko Virmavirta The Segmental Movements in Front Crawl Swimming . . . . . . . . . . . . . 1703 Ross H. Sanders, Jordan T. Andersen, and Hideki Takagi Movement Analysis of Scull and Oar Rowing . . . . . . . . . . . . . . . . . . . . 1719 Patria A. Hume Segmental Movements in Cycling . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1741 Rodrigo R. Bini and Felipe P. Carpes Movement Analysis of the Golf Swing . . . . . . . . . . . . . . . . . . . . . . . . . . 1755 Patria A. Hume and J. Keogh The Motor Solutions of Throws in Sports . . . . . . . . . . . . . . . . . . . . . . . 1773 Bing Yu

xviii

Contents

Volume 3 Part XX

Motion in Music . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

1787

Body Movements in Music Performances: The Example of Clarinet Players . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1789 Manfred Nusseck, Marcelo M. Wanderley, and Claudia Spahn Investigating Aspects of Movement in Violin Performance . . . . . . . . . . 1803 Gongbing Shan, Peter Visentin, Manfred Nusseck, and Claudia Spahn Movement and Touch in Piano Performance . . . . . . . . . . . . . . . . . . . . . 1821 Werner Goebl Movements, Timing, and Precision of Drummers . . . . . . . . . . . . . . . . . 1839 Sofia Dahl Observing and Learning Complex Actions: On the Example of Guitar Playing . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1859 Tom Gardner and Emily S. Cross Part XXI

Dance

.........................................

1873

Functional Movement Analysis in Dance . . . . . . . . . . . . . . . . . . . . . . . . 1875 Andrea Schärli Motion Analysis as Pedagogic Tool in Dance . . . . . . . . . . . . . . . . . . . . . 1889 Martin Puttke and Dimitri Volchenkov Somatic Practices: How Motion Analysis and Mind Images Work Hand in Hand in Dance . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1911 Sarah Whatley

Section IV Part XXII

Animation

Human Animation Control . . . . . . . . . . . . . . . . . . . . . . . .

1927

Video-Based Performance Driven Facial Animation . . . . . . . . . . . . . . . 1929 Fuhao Shi Depth Sensor-Based Facial and Body Animation Control . . . . . . . . . . . 1943 Yijun Shen, Jingtian Zhang, Longzhi Yang, and Hubert P. H. Shum Real-Time Full-Body Pose Synthesis and Editing Edmond S. L. Ho and Pong C. Yuen Real-Time Full Body Motion Control John Collomosse and Adrian Hilton

. . . . . . . . . . . . . . . . . 1959

. . . . . . . . . . . . . . . . . . . . . . . . . . 1975

Contents

Part XXIII

xix

Human Animation Generation . . . . . . . . . . . . . . . . . . . .

2001

Data-Driven Character Animation Synthesis . . . . . . . . . . . . . . . . . . . . . 2003 Taku Komura, Ikhsanul Habibie, Jonathan Schwarz, and Daniel Holden Physically Based Character Animation Synthesis . . . . . . . . . . . . . . . . . 2033 Jie Tan Biped Controller for Character Animation . . . . . . . . . . . . . . . . . . . . . . 2055 KangKang Yin, Stelian Coros, and Michiel van de Panne Data-Driven Hand Animation Synthesis Sophie Jörg

. . . . . . . . . . . . . . . . . . . . . . . . 2079

Example-Based Skinning Animation . . . . . . . . . . . . . . . . . . . . . . . . . . . 2093 Tomohiko Mukai Part XXIV

Facial Animation and Gestures . . . . . . . . . . . . . . . . . . . .

2113

Visual Speech Animation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2115 Lei Xie, Lijuan Wang, and Shan Yang Blendshape Facial Animation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2145 Ken Anjyo Eye Animation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2157 Andrew T. Duchowski and Sophie Jörg Head Motion Generation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2177 Najmeh Sadoughi and Carlos Busso Hand Gesture Synthesis for Conversational Characters . . . . . . . . . . . . 2201 Michael Neff Laughter Animation Generation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2213 Yu Ding, Thierry Artières, and Catherine Pelachaud Part XXV

Crowd Simulation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

2231

Functional Crowds . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2233 Jan M. Allbeck Crowd Formation Generation and Control . . . . . . . . . . . . . . . . . . . . . . 2243 Jiaping Ren, Xiaogang Jin, and Zhigang Deng Crowd Evacuation Simulation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2255 Tomoichi Takahashi

xx

Part XXVI

Contents

Facial Expressions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

2269

Perceptual Study on Facial Expressions . . . . . . . . . . . . . . . . . . . . . . . . . 2271 Eva G. Krumhuber and Lina Skora Part XXVII

Human to Virtual-Human Interaction . . . . . . . . . . . . . .

2287

Utilizing Unsupervised Crowdsourcing to Develop a Machine Learning Model for Virtual Human Animation Prediction . . . . . . . . . . 2289 Michael Borish and Benjamin Lok

Section V

Forensics and Legal Application

Part XXVIII Functional Capacity Evaluation

..................

2307

Functional Capacity Evaluation and Preemployment Screening . . . . . . 2309 Elizabeth Chapman, Anne M. Felts, and Matthew Klinker Functional Capacity Evaluation and Quantitative Gait Analysis: Lower Limb Disorders . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2331 Isabella T. Klöpfer-Krämer and Peter Augat Preparticipation Physical Evaluation in Sport . . . . . . . . . . . . . . . . . . . . 2349 James A. Onate and Daniel R. Clifton Part XXIX

Forensics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

2361

Injury Mechanisms in Traffic Accidents . . . . . . . . . . . . . . . . . . . . . . . . 2363 Brian D. Goodwin, Sajal Chirvi, and Frank A. Pintar Vehicle Occupants in Traffic Accidents Garrett A. Mattos

. . . . . . . . . . . . . . . . . . . . . . . . . 2399

Slips, Trips, and Falls . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2417 Andrew Short and Len Cubitt Biomechanical Forensics in Pediatric Head Trauma . . . . . . . . . . . . . . . 2447 Brittany Coats and Susan Margulies Head Impact Biomechanics of “King Hit” Assaults . . . . . . . . . . . . . . . . 2463 Declan A. Patton and Andrew S. McIntosh Expert Opinion and Legal Considerations . . . . . . . . . . . . . . . . . . . . . . . 2475 Henry M. Silvester Applications in Forensic Biomechanics Andrew S. McIntosh

. . . . . . . . . . . . . . . . . . . . . . . . . 2495

Index . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2509

About the Editors

Bertram Müller started his professional life with two degrees in precision mechanics and electronics, he worked in the Institute of Experimental Physics at the Martin Luther University in Halle/Saale. A third engineering degree and an exchange program in a center for children with special needs in Australia awoke his interest in bridging Engineering and Medicine. Consequently, he undertook a Ph.D. in Bioengineering at Strathclyde University in Glasgow and graduated in Medical Science at University Rey Juan Carlos in Madrid. In 2000, he moved to Barcelona, where he managed the Biomechanics Laboratory of Egarsat until 2012. He presently divides his professional life between teaching and applied biomechanics. He is Associate Professor of Physics and Biomechanics at the University of Girona (EUSES), the University of Manresa, and the National Institute of Physical Education in Barcelona. He also works as a Consultant for applied biomechanics, including clinical and occupational biomechanics, as well as sports applications, having collaborated with the High-Performance Centre (CAR) in St. Cugat. His current research interest is functional capacity evaluation in various disciplines, including legal matters. He is Editor-in-Chief of the Journal of Forensic Biomechanics, as well as reviewer for several other scientific journals. Sebastian I. Wolf with a Ph.D. in Physics, Sebastian Wolf spent several years in pure physics research before he moved into the field of human movement analysis in 2001. For 10 years, he was Technical Director of the Gait Analysis Laboratory in the Department of Orthopedic Surgery at Heidelberg University. Since 2010, he has been the Director of the Division of Human Movement Analysis and is responsible for both the clinical gait analysis service as well as overseeing clinical research in this field. In 2015, he became Associate

xxi

xxii

About the Editors

Professor for Orthopedic Biomechanics at the Medical Faculty of Heidelberg University. In the same year, he was elected President of the European Society for Movement Analysis in Adults and Children (ESMAC). Sebastian Wolf has published more than 80 peer-reviewed articles relating to clinical movement analysis and is a reviewer for numerous journals in this field with continuing scientific interest in advancing knowledge on neurologic and orthopedic gait disorders and mobility-related medical healthcare.

Section Editors

Gert-Peter Brüggemann Institute for Biomechanics und Orthopedics, German Sport University Cologne, Cologne, Germany Zhigang Deng Department of Computer Science, University of Houston, Houston, TX, USA Andrew S. McIntosh McIntosh Consultancy and Research, Sydney, NSW, Australia Australian Collaboration for Research into Injury in Sport and its Prevention (ACRISP), Federation University Australia, Ballarat, VIC, Australia Monash University Accident Research Centre, Monash University, Melbourne, VIC, Australia Freeman Miller Wilmington, Delaware, USA Sebastian I. Wolf Clinic for Orthopedics and Trauma Surgery, Center for Orthopedics, Trauma Surgery and Spinal Cord Injury, Heidelberg University Hospital, Heidelberg, Germany W. Scott Selbie HAS-Motion Inc., Kingston, ON, Canada C-Motion, Inc., Germantown, MD, USA

xxiii

Advisory Panel

Benita Kuni Clinic for Orthopedics and Trauma Surgery, Heidelberg University Hospital, Heidelberg, Germany Benno Nigg Faculties of Kinesiology, Engineering and Medicine, University of Calgary, Calgary, AB, Canada Manfred Nusseck University of Music and University Clinic Freiburg, Freiburg, Germany Mark Stringer Clarivate, Barcelona, Spain

xxv

Contributors

Oussama Abousamra Nemours Alfred I. duPont Hospital for Children, Wilmington, DE, USA Michiyoshi Ae Faculty of Sport Science, Nippon Sport Science University, Tokyo, Japan Valeria Agosti Department of Motor Sciences and Wellness, University of Naples Parthenope, Naples, Italy Institute Hermitage-Capodimonte, Naples, Italy Jan M. Allbeck George Mason University, Fairfax, VA, USA Lara Allet Department of Physical Therapy, University of Applied Sciences of Western Switzerland, Carouge, Switzerland Department of Community Medicine, Geneva University Hospitals and University of Geneva, Geneva, Switzerland Christine Alvarez Shriners Gait Lab, Sunny Hill Health Centre for Children, Vancouver, BC, Canada British Columbia Children’s Hospital, Vancouver, BC, Canada Jordan T. Andersen Exercise and Sport Science, Faculty of Health Sciences, The University of Sydney, Sydney, NSW, Australia Jan Andrysek Holland Bloorview Kids Rehabilitation Hospital, Bloorview Research Institute, Toronto, ON, Canada Institute of Biomaterial and Biomedical Engineering, University of Toronto, Toronto, ON, Canada Ken Anjyo OLM Digital, Setagaya, Tokyo, Japan Thierry Artières Ecole Centrale Marseille, Marseille, France Laboratoire d’Informatique Fondamentale (LIF), UMR CNRS 7279, Université AixMarseille, Paris, France Ayman Assi Laboratory of Biomechanics and Medical Imaging, Faculty of Medicine, University of Saint-Joseph, Mar Mikhael, Beirut, Lebanon xxvii

xxviii

Contributors

Institut de Biomécanique Humaine Georges Charpak, Arts et Métiers ParisTech, Paris, France Janie Astephen Wilson School of Biomedical Engineering, Faculty of Medicine, Dalhousie University, Halifax, NS, Canada Nova Scotia Health Authority, Halifax, NS, Canada Peter Augat Berufsgenossenschaftliche Unfallklinik Murnau und Paracelsus Medizinische Privatuniversität Salzburg, Institut für Biomechanik, Murnau am Staffelsee, Germany Sam Augsburger Motion Analysis Center, Shriners Hospitals for Children Medical Center, Lexington, KY, USA Richard Baker University of Salford, Salford, UK Astrid Balemans Department of Rehabilitation Medicine, MOVE Research Institute Amsterdam, EMGO+ Institute for Health and Care Research, VU University Medical Center, Amsterdam, The Netherlands Brain Center Rudolf Magnus and Center of Excellence for Rehabilitation Medicine University Medical Center, Utrecht, The Netherlands De Hoogstraat Rehabilitation, Utrecht, The Netherlands Rodrigo R. Bini La Trobe Rural Health School, College of Science, Health and Engineering, La Trobe University, Bendigo, VIC, Australia Kristie F. Bjornson Seattle Children’s Research Institute, University of Washington, Seattle, WA, USA I. J. Blokland Faculty of Behavioral and Movement Sciences, Vrije Universiteit Amsterdam, Amsterdam Movement Sciences, Amsterdam, The Netherlands Heliomare Rehabilitation, Research and Development, Wijk aan Zee, The Netherlands Eline Bolster Department of Rehabilitation Medicine, MOVE Research Institute Amsterdam, EMGO+ Institute for Health and Care Research, VU University Medical Center, Amsterdam, The Netherlands Michael Borish Computer and Information Sciences and Engineering Department, University of Florida, Gainesville, FL, USA Marco Bracalenti Department of Philosophy, Social and Human Sciences and Education, University of Perugia, Perugia, Italy Scott C. E. Brandon Department of Mechanical Engineering, University of Wisconsin-Madison, Madison, WI, USA

Contributors

xxix

James White Brodsky Faculty, Foot and Ankle Fellowship Program, Baylor University Medical Center, Dallas, TX, USA University of Texas Southwestern Medical School, Dallas, TX, USA Texas A&M HSC College of Medicine, Bryan, TX, USA Marcus J. Brown HAS-Motion, Inc., Kingston, ON, Canada Hendrik Bruttel Clinic for Orthopedics and Trauma Surgery, Heidelberg University Hospital, Heidelberg, Germany Carlos Busso Multimodal Signal Processing Lab, University of Texas at Dallas, Dallas, TX, USA Silvia Cabral Laboratório de Biomec^anica e Morfologia Funcional, CIPER, Faculdade de Motricidade Humana, Universidade de Lisboa, Cruz Quebrada, Dafundo, Portugal Steven Cadavid KinaTrax Inc., Palm Beach, FL, USA Valentina Camomilla Interuniversity Centre of Bioengineering of the Human Neuromusculoskeletal System, Department of Movement, Human and Health Sciences, University of Rome Foro Italico, Rome, Italy Aurelio Cappozzo Interuniversity Centre of Bioengineering of the Human Neuromusculoskeletal System, Department of Movement, Human and Health Sciences, University of Rome “Foro Italico”, Rome, Italy Paolo Caravaggi Movement Analysis Laboratory and Functional-Clinical Evaluation of Prostheses, Istituto Ortopedico Rizzoli, Bologna, Italy Felipe P. Carpes Applied Neuromechanics Research Group, Faculty of Health Sciences, Federal University of Pampa, Uruguaiana, RS, Brazil Robert D. Catena Gait and Posture Biomechanics Lab, Washington State University, Pullman, WA, USA Andrea Cereatti Department POLCOMING, University of Sassari, Sassari, Italy Interuniversity Centre of Bioengineering of the Human Neuromusculoskeletal System, University of Sassari, Sassari, Italy Department of Electronics and Telecommunications, Politecnico di Torino, Turin, Italy Elizabeth Chapman Workforce Solutions, BTE Technologies, Greenwood Village, CO, USA Sajal Chirvi Neuroscience Research Labs – Research 151, Medical College of Wisconsin, Zablocki VA Medical Center, Milwaukee, WI, USA

xxx

Contributors

Nachiappan Chockalingam Life Sciences and Education, Staffordshire University, Stoke On Trent, UK Swati Chopra Motion Analysis Laboratory, Mayo Clinic, Rochester, MN, USA Veronica Cimolin Department of Electronics, Information and Bioengineering, Politecnico di Milano, Milano, Italy Daniel R. Clifton School of Health and Rehabilitation Sciences, The Ohio State University, Columbus, OH, USA Jameson Crane Sports Medicine Institute, The Ohio State University Wexner Medical Center, The Ohio State University, Columbus, OH, USA Brittany Coats Mechanical Engineering, University of Utah, Salt Lake City, UT, USA Mutlu Cobanoglu Department of Orthopedics and Traumatology, Adnan Menderes University Faculty of Medicine, Aydın, Turkey Scott Coleman Department of Orthopaedics, Baylor University Medical Center, Dallas, TX, USA Department of Orthopedics, Baylor Scott and White, Dallas, TX, USA John Collomosse Centre for Vision Speech and Signal Processing (CVSSP), University of Surrey, Surrey, UK Justin Connor Nemours A.I. duPont Hospital for Children, Wilmington, DE, USA Carmela Conte Movement Analysis LAB, Rehabilitation Centre Policlinico Italia, Rome, Italy Stelian Coros Carnegie Mellon University, Pittsburgh, PA, USA Emily S. Cross Bangor University, Bangor, North Wales, UK Robert G. Crowther Sport and Exercise, School of Health and Wellbeing, University of Southern Queensland, Ipswich, QLD, Australia Smart Movement, Brisbane, QLD, Australia Len Cubitt Tullamarine, VIC, Australia Andrea Giovanni Cutti Applied Research, INAIL Prosthetic Center, Vigorso di Budrio, BO, Italy Sofia Dahl Department of Architecture, Design and Media Technology, Aalborg University Copenhagen, Copenhagen, Denmark Annet Dallmeijer Department of Rehabilitation Medicine, MOVE Research Institute Amsterdam, EMGO+ Institute for Health and Care Research, VU University Medical Center, Amsterdam, The Netherlands

Contributors

xxxi

Jon R. Davids Northern California Shriner’s Hospital for Children, Sacramento, Sacramento, CA, USA Karen Davies Shriners Gait Lab, Sunny Hill Health Centre for Children, Vancouver, BC, Canada Alan R. De Asha C-Motion Inc., Germantown, MD, USA Ugo Della Croce Department POLCOMING, University of Sassari, Sassari, Italy Interuniversity Centre of Bioengineering of the Human Neuromusculoskeletal System, University of Sassari, Sassari, Italy Zhigang Deng Department of Computer Science, University of Houston, Houston, TX, USA Timothy R. Derrick Department of Kinesiology, Iowa State University, Ames, IA, USA Yu Ding University of Houston, Houston, TX, USA Andrew T. Duchowski Clemenson University, Clemson, SC, USA Raphaël Dumas LBMC UMR_T9406, Univ Lyon, Université Claude Bernard Lyon 1, IFSTTAR, Lyon, France Jan M. Eckerle Clinic for Orthopedics and Trauma Surgery, Heidelberg University Hospital, Heidelberg, Germany W. Brent Edwards Human Performance Laboratory, Faculty of Kinesiology, University of Calgary, Calgary, AB, Canada Arezoo Eshraghi Holland Bloorview Kids Rehabilitation Hospital, Bloorview Research Institute, Toronto, ON, Canada Stefano Federici Department of Philosophy, Social and Human Sciences and Education, University of Perugia, Perugia, Italy Anne M. Felts Workforce Solutions, BTE Technologies, Greenwood Village, CO, USA Malindu E. Fernando Podiatry Service, Kirwan Community Health Campus, Townsville, QLD, Australia College of Medicine, James Cook University, Townsville, QLD, Australia Maurizio Ferrarin Biomedical Technology Department, IRCCS Fondazione Don Carlo Gnocchi Onlus, Milan, MI, Italy Hilde Feys Research Group for Neuromotor Rehabilitation, KU Leuven, Leuven, Belgium

xxxii

Contributors

Henryk Flashner Department of Aerospace and Mechanical Engineering, University of Southern California, Los Angeles, CA, USA Peter G. Gabos Nemours A.I. duPont Hospital for Children, Wilmington, DE, USA Deborah Gaebler-Spira Shirley Ryan Ability Lab, Chicago, IL, USA Manuela Galli Department of Electronics, Information and Bioengineering, Politecnico di Milano, Milano, Italy IRCCS “San Raffaele Pisana” Tosinvest Sanità, Roma, Italy Steven A. Gard Department of Physical Medicine and Rehabilitation, Feinberg School of Medicine, Northwestern University Prosthetics-Orthotics Center (NUPOC), Chicago, IL, USA Department of Physical Medicine and Rehabilitation, Feinberg School of Medicine, Northwestern University, Chicago, IL, USA Jesse Brown VA Medical Center, Department of Veterans Affairs, Chicago, IL, USA Tom Gardner Bangor University, Bangor, North Wales, UK Ismat Ghanem Laboratory of Biomechanics and Medical Imaging, Faculty of Medicine, University of Saint-Joseph, Mar Mikhael, Beirut, Lebanon Hôtel-Dieu de France Hospital, University of Saint-Joseph, Beirut, Lebanon Claudia Giacomozzi Department of Cardiovascular Diseases, Dysmetabolic Diseases and Ageing, Italian National Institute of Health, Rome, Italy Werner Goebl Department of Music Acoustics – Wiener Klangstil (IWK), University of Music and Performing Arts Vienna, Vienna, Austria Brian D Goodwin Neuroscience Research Labs – Research 151, Medical College of Wisconsin, Zablocki VA Medical Center, Milwaukee, WI, USA Arnaud Gouelle Gait and Balance Academy, ProtoKinetics, Havertown, PA, USA Hans Gray Department of Mechanical Engineering, The University of Melbourne, Parkville, VIC, Australia Shanyuanye Guan Department of Mechanical Engineering, The University of Melbourne, Parkville, VIC, Australia Ikhsanul Habibie School of Informatics, University of Edinburgh, Edinburgh, UK Joseph Hamill Department of Kinesiology, University of Massachusetts, Amherst, MA, USA Andrew Hansen Minneapolis VA Health Care System, Minneapolis, MN, USA University of Minnesota, Minneapolis, MN, USA

Contributors

xxxiii

Aoife Healy Life Sciences and Education, Staffordshire University, Stoke On Trent, UK Laura A. Held Department of Biological Sciences, University of Southern California, Los Angeles, CA, USA John Henley Nemours A.I. duPont Hospital for Children, Wilmington, DE, USA Kasee J. Hildenbrand Athletic Training Program, Washington State University, Pullman, WA, USA Adrian Hilton Centre for Vision Speech and Signal Processing (CVSSP), University of Surrey, Surrey, UK Edmond S. L. Ho Department of Computer and Information Sciences, Northumbria University, Newcastle upon Tyne, UK At L. Hof Center for Human Movement Sciences and Laboratory of Human Movement Analysis, Department of Rehabilitation, University Medical Center Groningen, Groningen, The Netherlands Daniel Holden School of Informatics, University of Edinburgh, Edinburgh, UK H. Houdijk Faculty of Behavioral and Movement Sciences, Vrije Universiteit Amsterdam, Amsterdam Movement Sciences, Amsterdam, The Netherlands Heliomare Rehabilitation, Research and Development, Wijk aan Zee, The Netherlands Cheryl L. Hubley-Kozey School of Physiotherapy, Faculty of Health Professions, Dalhousie University, Halifax, NS, Canada School of Biomedical Engineering, Faculty of Medicine, Dalhousie University, Halifax, NS, Canada Nova Scotia Health Authority, Halifax, NS, Canada François Hug Laboratory “Movement, Interaction, Performance” (EA4334), University of Nantes, Nantes, France NHMRC Centre of Clinical Research Excellence in Spinal Pain, Injury and Health, School of Health and Rehabilitation Sciences, The University of Queensland, Brisbane, QLD, Australia Patria A. Hume Auckland University of Technology, Auckland, New Zealand T. IJmker Faculty of Behavioral and Movement Sciences, Vrije Universiteit Amsterdam, Amsterdam Movement Sciences, Amsterdam, The Netherlands Ellen Jaspers Neural Control of Movement Lab, ETH Zurich, Zurich, Switzerland Sophie Jörg School of Computing, Clemson University, Clemson, SC, USA

xxxiv

Contributors

Xiaogang Jin State Key Lab of CAD&CG, Zhejiang University, Hangzhou, China Johanna Jonsdottir LaRiCE, Department of Neurorehabilitation, IRCCS Fondazione Don Carlo Gnocchi Onlus, Milan, Italy Justin Michael Kane Baylor University Medical Center, McKinney, TX, USA Faculty, Foot and Ankle Fellowship Program, Baylor University Medical Center, Dallas, TX, USA Orthopedic Associates of Dallas, Dallas, TX, USA Kenton R. Kaufman Motion Analysis Laboratory, Mayo Clinic, Rochester, MN, USA J. Keogh Bond University Australia, Robina, QLD, Australia Thomas M. Kepple C-Motion Inc., Germantown, MD, USA Matthew Klinker Workforce Solutions, BTE Technologies, Greenwood Village, CO, USA Isabella T. Klöpfer-Krämer Berufsgenossenschaftliche Unfallklinik Murnau und Paracelsus Medizinische Privatuniversität Salzburg, Institut für Biomechanik, Murnau am Staffelsee, Germany Taku Komura School of Informatics, University of Edinburgh, Edinburgh, UK Andrea Kotanxis Leon Root Motion Analysis Laboratory, Hospital for Special Surgery, New York, NY, USA Andreas Kranzl Laboratory for Gait and Human Motion Analysis, Orthopedic Hospital Speising, Vienna, Austria Eva G. Krumhuber University College London, London, UK Alberto Leardini Movement Analysis Laboratory and Functional-Clinical Evaluation of Prostheses, Istituto Ortopedico Rizzoli, Bologna, Italy Fabien Leboeuf School of Health Sciences, University of Salford, Salford, UK Nancy Lennon Nemours A.I. duPont Hospital for Children, Wilmington, DE, USA Lise Leveille Shriners Gait Lab, Sunny Hill Health Centre for Children, Vancouver, BC, Canada British Columbia Children’s Hospital, Vancouver, BC, Canada Glen Lichtwark Centre for Sensorimotor Performance, School of Human Movement and Nutrition Sciences, The University of Queensland, St Lucia, QLD, Australia Peter Loan C-Motion, Inc., Germantown, MD, USA Benjamin Lok Computer and Information Sciences and Engineering Department, University of Florida, Gainesville, FL, USA

Contributors

xxxv

William G. Mackenzie Nemours A.I. duPont Hospital for Children, Wilmington, DE, USA Michael W. Maier Clinic for Orthopedics and Trauma Surgery, Heidelberg University Hospital, Heidelberg, Germany Lisa Mailleux Research Group for Neuromotor Rehabilitation, KU Leuven, Leuven, Belgium Martina Mancini Department of Neurology, Oregon Health and Science University, Portland, OR, USA Susan Margulies Bioengineering, University of Pennsylvania, Philadelphia, PA, USA Carlo Massaroni Unit of Measurements and Biomedical Instrumentation, Campus Bio-Medico di Roma University, Rome, Italy Garrett A. Mattos Transport and Road Safety (TARS) Research Centre, University of New South Wales, Sydney, NSW, Australia Andrew S. McIntosh Australian Collaboration for Research into Injury in Sport and its Prevention (ACRISP), Federation University Australia, Ballarat, VIC, Australia Monash University Accident Research Centre, Monash University, Melbourne, VIC, Australia McIntosh Consultancy and Research, Sydney, NSW, Australia James McLoughlin Flinders University, Adelaide, Australia Jill L. McNitt-Gray Departments of Biological Sciences and Biomedical Engineering, University of Southern California, Los Angeles, CA, USA Fabio Meloni Department of Philosophy, Social and Human Sciences and Education, University of Perugia, Perugia, Italy Fabrice Mégrot Unité Clinique d’Analyse de la Marche et du Mouvement, Centre de Médecine Physique et de Réadaptation pour Enfants de Bois-Larris – CroixRouge Française, Lamorlaye, France UMR CNRS 7338: Biomécanique et Bioingénierie, Sorbonne Universités, Université de Technologie de Compiègne, Compiègne, France Freeman Miller Nemours A.I. duPont Hospital for Children, Wilmington, DE, USA Ross H. Miller Department of Kinesiology, University of Maryland, College Park, MD, USA Tomohiko Mukai Tokai University, Tokyo, Japan

xxxvi

Contributors

Robert Needham Life Sciences and Education, Staffordshire University, Stoke On Trent, UK Michael Neff Department of Computer Science and Program for Cinema and Digital Media, University of California – Davis, Davis, CA, USA Christopher Nester School of Health Sciences, University of Salford, Salford, UK Timothy A. Niiler Gait Laboratory, Nemours A.I. duPont Hospital for Children, Wilmington, DE, USA Manfred Nusseck Freiburg Institute for Musicians’ Medicine, University of Music Freiburg, Medical Center – University of Freiburg, Faculty of Medicine, University of Freiburg, Freiburg, Germany Sylvia Õunpuu Center for Motion Analysis, Division of Orthopaedics, Connecticut Children’s Medical Center, Farmington, CT, USA James A. Onate School of Health and Rehabilitation Sciences, The Ohio State University, Columbus, OH, USA Jameson Crane Sports Medicine Institute, The Ohio State University Wexner Medical Center, The Ohio State University, Columbus, OH, USA Michael Orendurff Motion and Sports Performance Laboratory, Lucile Packard Children’s Hospital Stanford, Sunnyvale, CA, USA Marcus Pandy Department of Mechanical Engineering, The University of Melbourne, Parkville, VIC, Australia Ilaria Parel Unit of Shoulder and Elbow Surgery, Cervesi Hospital, Cattolica, RN, Italy Dimitrios A. Patikas School of Physical Education and Sport Science, Aristotle University of Thessaloniki, Thessaloniki, Greece Declan A. Patton Australian Collaboration for Research into Injury in Sport and its Prevention (ACRISP), Federation University Australia, Ballarat, VIC, Australia Oslo Sports Trauma Research Centre (OSTRC), Norwegian School of Sport Sciences, Oslo, Norway Sport Injury Prevention Research Centre (SIPRC), University of Calgary, Calgary, AB, Canada Catherine Pelachaud CNRS - ISIR, Université Pierre et Marie Curie, Paris, France Kristan Pierz Center for Motion Analysis, Division of Orthopaedics, Connecticut Children’s Medical Center, Farmington, CT, USA Frank A. Pintar Neuroscience Research Labs – Research 151, Medical College of Wisconsin, Zablocki VA Medical Center, Milwaukee, WI, USA

Contributors

xxxvii

Jessica May Pohlmann Sport and Exercise, University of Southern Queensland, Ipswich, QLD, Australia Smart Movement, Brisbane, Australia Ana Presedo Pediatric Orthopaedics Department, Robert Debré University Hospital, Paris, France Jessica Pruente Shirley Ryan Ability Lab, Chicago, IL, USA Ilona M. Punt Department of Epidemiology, Maastricht University, CAPHRI, Maastricht, The Netherlands Department of Physical Therapy, University of Applied Sciences of Western Switzerland, Carouge, Switzerland Martin Puttke Board of German Federal Association of Dance, Berlin, Germany Elizabeth A. Rapp University of Delaware, Newark, DE, USA Julie Reay School of Health Sciences, University of Salford, Salford, UK Jiaping Ren State Key Lab of CAD&CG, Zhejiang University, Hangzhou, China Jaques Riad Skaraborg Hospital Skövde, Skövde, Sweden André Gonçalo Gomes Roque Physiotherapy, University of Averio, Aveiro, Portugal Marlene Cristina Neves Rosa Piaget Institute, Viseu, Portugal Dieter Rosenbaum Funktionsbereich Bewegungsanalytik, Institut für Experimentelle Muskuloskelettale Medizin, Zentrum für Muskuloskelettale Medizin, Universitätsklinikum Münster, Münster, Germany Bodo Rosenhahn Institut für Informationsverarbeitung, Leibniz Universität Hannover, Hannover, Germany Rüdiger Rupp Spinal Cord Injury Center – Experimental Neurorehabilitation, Heidelberg University Hospital, Heidelberg, Germany Erich Rutz Pediatric Orthopaedic Department, University Children’s Hospital Basel, Basel, Switzerland Angelo M. Sabatini The BioRobotics Institute, Scuola Superiore Sant’Anna, Pisa, Italy Najmeh Sadoughi Multimodal Signal Processing Lab, University of Texas at Dallas, Dallas, TX, USA Ross H. Sanders Exercise and Sport Science, Faculty of Health Sciences, The University of Sydney, Sydney, NSW, Australia Morgan Sangeux Hugh Williamson Gait Analysis Laboratory, The Royal Children’s Hospital, Parkville/Melbourne, VIC, Australia

xxxviii

Contributors

Gait laboratory and Orthopaedics, The Murdoch Childrens Research Institute, Parkville/Melbourne, VIC, Australia Andrea Schärli Institute of Sport Science, University of Bern, Bern, Switzerland Jonathan Schwarz School of Informatics, University of Edinburgh, Edinburgh, UK W. Scott Selbie HAS-Motion Inc., Kingston, ON, Canada C-Motion Inc., Germantown, MD, USA Mariano Serrao Department of Medical and Surgical Sciences and Biotechnologies, Sapienza University of Rome, Latina, Italy Movement Analysis LAB, Rehabilitation Centre Policlinico Italia, Rome, Italy Gongbing Shan Department of Kinesiology, Faculty of Arts and Science, University of Lethbridge, Lethbridge, AB, Canada Frances T. Sheehan Rehabilitation Medicine Department, Functional and Applied Biomechanics Section, National Institutes of Health, Bethesda, MD, USA Yijun Shen Northumbria University, Newcastle upon Tyne, UK Fuhao Shi Texas A&M University, College Station, TX, USA Andrew Short University of Melbourne, Melbourne, Australia Adam Shortland One Small Step Gait Laboratory, Evelina Children’s Hospital, Guy’s and St Thomas’ NHS Foundation Trust, London, UK Hubert P. H. Shum Northumbria University, Newcastle upon Tyne, UK Anne K. Silverman Functional Biomechanics Laboratory, Department of Mechanical Engineering, Colorado School of Mines, Golden, CO, USA Henry M. Silvester Barry Nilsson Lawyers, Sydney, NSW, Australia Cristina Simon-Martinez Research Group for Neuromotor Rehabilitation, KU Leuven, Leuven, Belgium Wafa Skalli Institut de Biomécanique Humaine Georges Charpak, Arts et Métiers ParisTech, Paris, France Lina Skora University College London, London, UK L. H. Sloot Department of Rehabilitation Medicine, MOVE Research Institute Amsterdam, VU University Medical Center, Amsterdam, The Netherlands Colin R. Smith Department of Mechanical Engineering, University of WisconsinMadison, Madison, WI, USA Richard M. Smith Rehabilitation Medicine Department, Functional and Applied Biomechanics Section, National Institutes of Health, Bethesda, MD, USA

Contributors

xxxix

Pierpaolo Sorrentino Department of Engineering, University of Naples Parthenope, Naples, Italy Giuseppe Sorrentino Department of Motor Sciences and Wellness, University of Naples Parthenope, Naples, Italy Institute Hermitage-Capodimonte, Naples, Italy Jacob J. Sosnoff Department of Kinesiology and Community Health, University of Illinois at Urbana-Champaign, Urbana, IL, USA Claudia Spahn Freiburg Institute for Musicians’ Medicine, University of Music Freiburg, Medical Center – University of Freiburg, Faculty of Medicine, University of Freiburg, Freiburg, Germany David M. Spranz Clinic for Orthopedics and Trauma Surgery, Heidelberg University Hospital, Heidelberg, Germany Felix Starker Biomechatronic Systems, Fraunhofer Institute for Manufacturing Engineering and Automation, Stuttgart, Germany Julie A. Stebbins Oxford Gait Laboratory, Oxford University Hospitals NHS Foundation Trust, Oxford, UK Nuffield Department of Orthopaedics, Rheumatology and Musculoskeletal Sciences, University of Oxford, Oxford, UK Felix Stief Movement Analysis Lab, Orthopedic Friedrichsheim gGmbH, Frankfurt/Main, Germany

University

Hospital

Ruopeng Sun Department of Kinesiology and Community Health, University of Illinois at Urbana-Champaign, Urbana, IL, USA Sean A. Tabaie Northern California Shriner’s Hospital for Children, Sacramento, Sacramento, CA, USA Hideki Takagi Faculty of Health and Sport Sciences, University of Tsukuba, Ibaraki, Japan Tomoichi Takahashi Department of Information Engineering, Meijo University, Nagoya, Japan Jie Tan Georgia Institute of Technology, Atlanta, GA, USA Darryl G. Thelen Department of Mechanical Engineering, University of Wisconsin-Madison, Madison, WI, USA Pam Thomason Hugh Williamson Gait Analysis Laboratory, Royal Children’s Hospital, Melbourne, VIC, Australia Kylie Tucker NHMRC Centre of Clinical Research Excellence in Spinal Pain, Injury and Health, School of Health and Rehabilitation Sciences, The University of Queensland, Brisbane, QLD, Australia

xl

Contributors

School of Biomedical Sciences, The University of Queensland, Brisbane, QLD, Australia Brian R. Umberger Department of Kinesiology, University of Massachusetts, Amherst, MA, USA Michiel van de Panne University of British Columbia, Vancouver, BC, Canada M. M. van der Krogt Department of Rehabilitation Medicine, MOVE Research Institute Amsterdam, VU University Medical Center, Amsterdam, The Netherlands Giuseppe Vannozzi Interuniversity Centre of Bioengineering of the Human Neuromusculoskeletal System, Department of Movement, Human and Health Sciences, University of Rome Foro Italico, Rome, Italy Benedicte Vanwanseele Department of Movement Sciences, KU Leuven, Leuven, Belgium Fontys University of Applied Sciences, Eindhoven, The Netherlands Mikko Virmavirta Biology of Physical Activity, The Faculty of Sport and Health Sciences, University of Jyväskylä, Jyväskylä, Finland Peter Visentin Department of Music, Faculty of Fine Arts, University of Lethbridge, Lethbridge, AB, Canada Dimitri Volchenkov Mathematics and Statistics, Texas Tech University, Lubbock, TX, USA Center for Nonlinear Physics, Sichuan University of Science and Engineering, Sichuan, China J. J. Wallace Motion Analysis Center, Shriners Hospitals for Children Medical Center, Lexington, KY, USA Marcelo M. Wanderley Input Devices and Music Interaction Laboratory (IDMIL), CIRMMT, McGill University, Montreal, QC, Canada Bastian Wandt Institut für Informationsverarbeitung, Leibniz Universität Hannover, Hannover, Germany Lijuan Wang Microsoft Research, Redmond, WA, USA Scott Wearing Institute of Health and Biomedical Innovation, Queensland University of Technology, Brisbane, QLD, Australia Faculty for Sport and Health, Technische Universität München, Munich, Bavaria, Germany Sarah Whatley Centre for Dance Research, Coventry University, Coventry, UK Hank White Motion Analysis Center, Shriners Hospitals for Children Medical Center, Lexington, KY, USA

Contributors

xli

Steffen Willwacher Institute of Biomechanics and Orthopedics, German Sport University, Cologne, Germany Institute of Functional Diagnostics, Cologne, Germany Janis Wojtusch Department of Computer Science, Simulation, Systems Optimization and Robotics Group, TU Darmstadt, Darmstadt, Germany Tyler A. Wood Department of Kinesiology and Community Health, University of Illinois at Urbana-Champaign, Urbana, IL, USA Lei Xie School of Computer Science, Northwestern Polytechnical University (NWPU), Xi’an, P. R. China Shan Yang School of Computer Science, Northwestern Polytechnical University, Xi’an, China Feng Yang Department of Kinesiology, The University of Texas at El Paso, El Paso, TX, USA Longzhi Yang Northumbria University, Newcastle upon Tyne, UK Maurice R. Yeadon Loughborough University, Loughborough, UK KangKang Yin Simon Fraser University, Burnaby, BC, Canada Department of Computer Science, Singapore, Singapore Bing Yu Division of Physical Therapy, Department of Allied Health Science, School of Medicine, The University of North Carolina at Chapel Hill, Chapel Hill, NC, USA Pong C. Yuen Department of Computer Science, Hong Kong Baptist University, Kowloon Tong, Hong Kong Petrissa Zell Institut für Informationsverarbeitung, Leibniz Universität Hannover, Hannover, Germany Jingtian Zhang Northumbria University, Newcastle upon Tyne, UK

Part I Rigid Body Modeling

Observing and Revealing the Hidden Structure of the Human Form in Motion Throughout the Centuries Aurelio Cappozzo

Abstract

Observing, revealing the hidden structure, and understanding the human locomotor system have been a goal for artists alone at first and for both artists and scientists later. How and why this goal was achieved is illustrated in this chapter. This is done following the fil rouge of history, believing that the understanding of the phylogenesis of knowledge effectively accompanies its ontogenesis. The realistic representation of the human form, as opposed to its metaphorical depiction, started in ancient Greece, but reached its apex in the Renaissance when artists understood that they needed a deeper understanding of reality in order to create an illusion of it. This was the premise for the scientific revolution in general and with regard to human motion in particular. During the nineteenth century, thermodynamics and the introduction of novel measurement and recording technologies gave renewed impulse to the study of the human locomotor system as if it were a machine designed either for fighting or for working. At the beginning of the twentieth century, avant-garde artists cooperated with the science of human movement by adding stronger human emotions and feelings to the scientific narration and therefore establishing a deeper perception of the natural phenomenon. The development of reconstructive orthopedic surgery in the second half of the twentieth century made the acquisition of deeper knowledge about the hidden movement of bones and the mechanics of human joints urgent. Since then, ever more sophisticated and accurate mathematical models of the neuromusculoskeletal system have been developed taking advantage of digital technology for measurement, computing, and virtual animation.

A. Cappozzo (*) Interuniversity Centre of Bioengineering of the Human Neuromusculoskeletal System, Department of Movement, Human and Health Sciences, University of Rome “Foro Italico”, Rome, Italy e-mail: [email protected] # Springer International Publishing AG, part of Springer Nature 2018 B. Müller, S.I. Wolf (eds.), Handbook of Human Motion, https://doi.org/10.1007/978-3-319-14418-4_199

3

4

A. Cappozzo

Keywords

History • Human movement • Locomotor apparatus • Biomechanics

Contents Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Throughout the Centuries . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Until the Middle Ages . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . The Renaissance . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . From the Age of Enlightenment to the Industrial Revolution or the Machine Age . . . . . . . . . Toward the Digital Age . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . The Contemporary Age . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

4 4 4 6 7 10 13 14 14

Introduction This chapter tackles a very basic problem in the biomechanics of the human locomotor apparatus. It’s about the observation – in the Galilean sense of course – of the human form while it moves, the observation of what is visible and, above all, of what is not visible, that is, the hidden structure of the locomotor system. To this purpose the fil rouge of history will be followed, believing that the understanding of the phylogenesis of knowledge effectively accompanies the ontogenesis of it. The reader will therefore be accompanied through the different methods and purposes that characterized the observation of human motion in the course of the past centuries and as far as the threshold of the current state of the art presented in this handbook.

Throughout the Centuries Until the Middle Ages For several millennia, the observation of a natural phenomenon, in general, and of the human form in motion, in particular, has been the prerogative of artists and philosophers in terms of the unveiling of its hidden structure, its interpretation, and its reproduction either in a symbolic or in a realistic fashion. As paradigmatic examples in this respect, we may examine two masterpieces that we have inherited from the ancient Greeks. One is represented by a Kouros (seventh century before the Christian era, Fig. 1). This is a statue depicting a young man who may seem to walk, but obviously goes nowhere. This was not due to inability of the artist to depict reality, quite the contrary, since the objective was not conveying a message of mobility, of life, but it was portraying a timeless figure, an image that transcends time. This was appropriate because this figure was a grave marker. The second

Observing and Revealing the Hidden Structure of the Human Form in Motion. . .

5

Fig. 1 Statue of a Kouros (young man) (580 BCE). New York, The Metropolitan Museum of Art

example is the Doryphoros by Polykleitos (440 before the Christian era, Fig. 2). Here the artist observed the outer form of the body, guessed the muscular and bony structure beneath the skin, and produced a figure that exists in the real world and is moving in it. This figure displays the correct pelvic attitude and the correct hip, knee, and ankle angles. All six determinants of gait, as described by Saunders and colleagues in 1953 are there! The Kouros and the Doryphoros represent two different ways of modelling reality. Just like our mathematical models, which may be metaphorical or naturalistic. Ancient Rome was a very pragmatic interpreter of Greek as well as Etruscan, native Italic, and even Egyptian visual culture, but did not bring any novelty in the way the human form, stationary or in motion, was represented. Medieval images of the human body were characterized by the fact that they did not reflect close observation from real life. Again, they aimed at representing an idea, and they also were metaphorical representations in nature. In those times, in fact, at least in the western part of the world, there was very little interest in the human body and in its inner workings. The body was seen as a mere temporary receptacle of the soul and cause of temptation.

6

A. Cappozzo

Fig. 2 A Roman copy of The Doryphoros (spear bearer) by Polykleitos (440 BCE). Naples, The National Archaeological Museum

The Renaissance Naturalistic representation is retrieved and reached its apex in the Renaissance (fourteenth to the seventeenth century) when artists realized that it was not sufficient to observe only the outer surface of a human body, as the ancient Greeks did, but it was necessary to go inside it and study anatomy through dissection. This approach allowed them to paint or sculpt the outer appearance of the body in many different static and dynamic postures accounting for muscle shape variations due to contraction, tension in the tendons, bony prominences, etc. This entailed revealing the outer human form starting from the hidden structure underneath, what we may name an “inside-outward” approach. Adam’s body as depicted in the Sistine Chapel in Rome is the mighty result of Michelangelo’s dissections and anatomical observations. Renaissance artists understood that they needed a deep understanding of reality in order to create an illusion of it. This awareness shows that times were ripe for an artist or philosopher to slowly and progressively start changing their skin and undergo a mutation and eventually become a scientist or rather a “natural philosopher.” At this time, art and science started a common itinerary and became travelling companions toward the ultimate goal of understanding, possessing, and controlling

Observing and Revealing the Hidden Structure of the Human Form in Motion. . .

7

natural phenomena, among which the human body, stationary or in motion, made no exception. Another important event characterized the Renaissance: mathematics sneaked into the artists’ work and became as important as observation. Linear perspective and the mathematics associated with it were invented. The three-dimensional world could be effectively portrayed on a planar canvas or wall. This is the same mathematics we use today in analytical stereophotogrammetry, although the other way round, using planar views, we reconstruct the object of interest in its three dimensions. The anthropometric measurements of Leonardo da Vinci (1510) as well as of Albrecht Durer (1528) may also be framed in this mathematical context. Of course, the ancient Greeks had already touched upon this last endeavor, as shown by the golden ratio or divine proportion that has intrigued artists, architects, philosophers, and scientists from Euclid onward. The Roman architect Vitruvius Polonius, who inspired the famous Leonardo’s drawing, named the Vitruvian Man (Stemp 2006), should also be mentioned in this context. In the seventeenth century, quantitative experimental observation, coupled with the analytical power of mathematics, became the normal approach for the purpose of describing and understanding the natural phenomenon and foreseeing or forecasting its evolution in time and space. This is the so-named scientific revolution. At that time, Galileo’s message struck a young medical doctor whose name was Giovanni Alfonso Borelli. He was the first one to apply the scientific method to the study of human motion, but with two important limitations: first, he was not aware of the inertia principle, thus his analyses were strictly static, and, second, he had no adequate instrumentation available. In addition, as he himself admitted, he was unable to carry out anatomical dissections. Despite this, his book, the De Motu Animalium, remains an unparalleled masterpiece (Borelli 1680, 1681). On purely speculative grounds, Borelli described human walking using the paradigm of the compass familiar to all human movement scientists. He also carried out the first estimate of internal loads as depicted in Fig. 3. Borelli added up the magnitudes of the forces exerted by single muscles and concluded that an individual must exert a total muscular force that is almost 50 times larger than the transported load. A rather naïve procedure, of course, but the message is strong and effective (Cappozzo and Marchetti 1992).

From the Age of Enlightenment to the Industrial Revolution or the Machine Age The abovementioned scientific revolution finds its natural continuation in the Age of Enlightenment that characterized the eighteenth century. Although this was an outstanding and absolutely crucial fragment of the history of mankind, no notable scientific contribution to the understanding of human motion can be recorded. It is, however, worth mentioning what Diderot and d’Alembert wrote in their encyclopedia (1751–1772) about the notion of “movement” as applied to a human being: “Movement is the action of a living body which is necessary for the conservation of

8

A. Cappozzo

Fig. 3 Frontispiece of Borelli’s De Motu Animalium first volume. External and internal forces acting on the human body while carrying a load

its health; the lack of movement as well as the excess of it are extremely prejudicial to the body.” During the nineteenth century, the formalization of thermodynamics gave renewed impulse to the study of the human locomotor system looked upon as if it were a machine designed either for fighting, in the rather unsettled Europe of that time, or for working, in the Second Industrial Revolution framework. It is not by chance that books reporting studies on human movement were given titles such as The Animal Machine (Marey 1873) or The Human Motor (Amar 1914). Of course technology helped this endeavor: Marey’s sensors and recording devices (Marey 1885, 1894), Muybridge’s highly sensitive photographic material (Muybridge 1887, 1899, 1901; Stillman 1882), Braune and Fischer’s stereophotogrammetric model (Braune and Fischer 1895–1904), and, later on, the high sampling rates used by Bernstein to record movement, just to mention the most important achievements. For a review of the work of these scientists, the reader may refer to Bouisset (1992), Tosi (1992), Maquet (1992), and Jansons (1992), respectively. At the beginning of the twentieth century, art outshined the science of human movement adding stronger human emotions and feelings to the scientific narration and therefore a deeper perception of the natural phenomenon. This, in turn, means a more thorough knowledge, a humanistic knowledge of the phenomenon that goes well beyond what may be stored in the memory of a computer. The “Nude Descending a Staircase”, painted by Marcel Duchamp in 1912, is a vivid example in this respect. Another example worth mentioning is the Italian artistic and philosophical movement named Futurism that celebrated the modern world of industry and technology, conjugated physics and aesthetics, and very effectively portrayed motion and speed both on canvas and in bronze through an evident influence of the abovementioned biomotion studies (Poore 1913; Fig. 4).

Observing and Revealing the Hidden Structure of the Human Form in Motion. . .

9

Fig. 4 Movement and dynamism of Futurism: (a) “Girl Running on a Balcony” by Giacomo Balla (1912), (b) “Unique Forms of Continuity in Space“ by Umberto Boccioni (1913)

In the course of the first half of the twentieth century, movement scientists could reconstruct the movement of a stick model of the human locomotor system as projected on a 2-D space or, in rare cases, in the 3-D space. These models were activated by the trajectories of target points located on the skin surface – most of the time on points approximating the joint centers – so that a sort of virtual monodimensional exoskeleton in motion could be reconstructed. However, no observation of any inner structure was attempted. These models, despite the obvious limitations and inaccuracies and the fact that they allowed more narrative than quantitative descriptions of the phenomena involved, sufficed the purpose of describing the human machine for pursuing several practical objectives. The impact that these studies had on society appears evident, for instance, when we look at the development of industrial production methodology. Just think of the Taylorism and Fordism and the implementation of these theories in America, where they were formulated, but also in Europe and even in the Soviet five-year plans. Human movement science, together with Pavlov’s learning theory, had a tremendous impact on performing arts as well. This is particularly evident in the work of the Russian theatrical actor and director Vsevolod Meyerhold who promoted an acting style, named Biomechanics, according to which “. . . it is when the actor has found the correct positions that he can pronounce the words, and only then these will sound meaningful . . ..” This acting style resulted to be in opposition with the more, so to speak, naturalistic Stanislavsky system (Law and Gordon 1995).

10

A. Cappozzo

Toward the Digital Age At the end of the Second World War, a very thorough study on human locomotion was carried out at the University of California with the aim of designing advanced lower limb substitutes (Eberhart 1947; Paul 1992). This study provided a piece of information regarding the description of human locomotion not available before. Specifically, the rotation of the pelvis, femur, and tibia about their longitudinal axes during walking was measured. For the first time we may legitimately talk of bone movement as opposed to segmental movement. But to this purpose pins had to be inserted into the volunteer’s bones acknowledging the fact that targets located on the skin surface would have not been able to track the rotation of the underlying bone reliably (Fig. 5). The development of endo- and arthro-prostheses and of reconstructive orthopedic surgery in general in the 1950s and 1960s made both the acquisition of deeper knowledge concerning the hidden movement of bones and the mechanics of human joints urgent. John Paul, at the University of Glasgow, responded to this urgency by initiating pioneering work unveiling the hidden structure around 1965. This author

Fig. 5 Test subject with pins and markers attached. Pelvis, femur, and tibia rotation during walking (Modified from Eberhart 1947)

Observing and Revealing the Hidden Structure of the Human Form in Motion. . .

11

used an outside-inward approach. The hidden structure was estimated using information collected on the body surface using cinecameras (Fig. 6). He was able to provide very useful data concerning the muscular forces and hip loading (Paul 1966, 1967, 1969), definitely a milestone in the history of biomechanics. At this point in time, we may say that two methodological approaches to the study of human movement are possible. While analyzing a motor task, one may aim at describing segmental movement using low to medium resolution (chapters ▶ “3D Dynamic Pose Estimation from Marker-Based Optical Data” and ▶ “3D Dynamic Pose Estimation from Markerless Optical Data”), and this may be adequate for a number of clinical applications such as motor function limitation assessment or in ergonomics and sports. The other approach leads to a high-resolution reconstruction of the movement of the inner invisible structures for the sake of musculoskeletal system modelling (chapters ▶ “Ultrasound Technology for Examining the Mechanics of the Muscle, Tendon, and Ligament,” ▶ “Physics-Based Models for Human Gait Analysis,” and ▶ “Optimal Control Modeling of Human Movement”). When we talk of high resolution of bone pose reconstruction, we mean to be able to resolve millimeters and degrees, as paradigmatically illustrated in Iwaki et al. (2000) with respect to the knee joint. From the mid-1970s onward, methods have been proposed that in fact allow the direct observation of the bones and thus an adequate resolution for the purpose

Fig. 6 (a) Positions of markers on the test subject. (b) Typical curve of hip joint force to body weight ration with time (Modified from Paul 1967)

12

A. Cappozzo

Fig. 7 A set of Cartesian set of axes is associated with the 3-D marker (bone-embedded frame). (a) Estimating the instantaneous pose of a bone using a 3-D marker assumed to be rigidly associated with the bone; (b) the 3-D marker may be represented by a cluster of landmarks or (c) a magnetoinertial measurement unit

mentioned previously. One is roentgen-stereophotogrammetry, which is of course invasive and suitable only to assess micromotion of orthopedic implants or in ex vivo joint mechanics investigations (Selvik 1974, 1990). Another powerful method able to allow the direct observation of bones during motion and less prone to criticisms from the ethical point of view is digital fluoroscopy which started to be used in the present context at the end of the 1990s, thanks to the work of Scott Banks and Hodge (1996), and has come to full maturation in the course of the last decade (chapter ▶ “Measurement of 3D Dynamic Joint Motion Using Biplane Videoradiography”). However, it exhibits these limitations: it is invasive (ionizing radiation), the measurement volume is limited, and the procedure is operator time-consuming. In any case these limitations make the direct observation of bones in motion inapplicable when the objective is the analysis of large portions of the musculoskeletal system moving in large volumes and/or characterized by high accelerations. Under these circumstances other methods must be devised. These consist in the use of observable 3-D markers, that is, three-dimensional objects – external to the body, for this reason directly observable using noninvasive methods – which are assumed to be rigidly associated with the underlying bone (Fig. 7a) (chapter ▶ “Three-Dimensional Human Kinematic Estimation Using Magneto-Inertial Measurement Units”). The abovementioned 3-D marker may be defined by three or more landmarks (Fig. 7b). The position in space of these landmarks may be recorded using retroreflective or light-emitting point markers located on the surface of the body segment

Observing and Revealing the Hidden Structure of the Human Form in Motion. . .

13

and stereophotogrammetry. As it occurred just about 100 years before with photography, the availability of optoelectronic sensors and high-performance digital technology at an accessible cost revolutionized experimental movement analysis. In the 1970s, time became ripe for the development of optoelectronic stereophotogrammetric systems. John Paul, by now at the Bioengineering Unit of Strathclyde University, understood it and invested the Ph.D. students Mick Jarrett and Brian Andrews with the responsibility of tackling this bold endeavor (Jarret et al. 1976). A possible alternative is using, as 3-D marker, a magneto-inertial measurement unit mounted on the body segment of interest (Fig. 7c) (chapter ▶ “Three-Dimensional Human Kinematic Estimation Using Magneto-Inertial Measurement Units”). This technique was first explored in Oxford (UK) by the graduate student Julian Morris, under the supervision of John O’Connor (Morris 1973), using only accelerometers. Magneto-inertial measurement units have recently known a remarkable technological development in terms of miniaturization and performance. The signals provided by these devices, however, allow the estimate of the orientation of the bone-embedded frame, but not of its position. This circumstance of course limits, but does not exclude, their use in the present context. Due to the interposed soft tissues, the abovementioned assumption that the 3-D markers are rigid with the underlying bone is of course disputable, and this causes remarkable inaccuracies that must be taken care of. This is the so-named soft tissue artefact issue, the solution of which does not seem to be just round the corner (chapter ▶ “3D Dynamic Probabilistic Pose Estimation from Data Collected Using Cameras and Reflective Markers”).

The Contemporary Age At the beginning of the 1980s, the above-illustrated optoelectronic stereophotogrammetric systems started to be marketed together with the relevant software. This was preceded by the appearance on the market of the first professional grade six-component force plate as a result of a joint venture between Jürg Wartenweiler of the ETH Zurich and a private company. At this point in time, we may say that the modern human movement analysis laboratory was finally available and the contemporary history of this discipline began. During the last decade the abovementioned laboratory has undergone a remarkable technological development and price reduction. Micro-Electro-Mechanical Systems and wireless technology have made wearable magneto-inertial sensors available almost ubiquitously, video-based markerless technology has become feasible, and medical imaging instrumentation is evolving rapidly, enriching the capability of observing both the morphology and the motion of inner structures while minimizing invasiveness. Lastly, methodologies developed in robotic engineering have lately shown that a cooperative synergy between this discipline and biomechanics is a very productive endeavor.

14

A. Cappozzo

Conclusion The evolution of human motion studies, throughout the different cultural ages of mankind, and the theoretical and applied motivations that drove it have been illustrated. This was intended to contribute to the understanding of the scenario within which the results presented in the subsequent chapters have been obtained and that forms the basis for future directions.

References Amar J (1914) Le Moteur Humainet les Bases Scientifiques du Travail Professional. Published by H. Dunod and E. Pinat, Paris Banks SA, Hodge WA (1996) Accurate measurement of three-dimensional knee replacement kinematics using single-plane fluoroscopy. IEEE Trans Biomed Eng 43(6):638–649 Borelli AJ (1680–1681) De Motu Animalium. Pars prima and Pars altera. Pub. by A. Bernabò, Roma. Translation by P. Maquet: On the movement of animals. Springer, Berlin/Heidelberg, 1989 Bouisset S (1992) Etienne-Jules Marey, or when motion biomechanics emerged as a science. In: Cappozzo A, Marchetti M, Tosi V (eds) Biolocomotion: a century of research using moving pictures. Promograph, Roma, pp 71–88 Braune W, Fischer O (1895–1904) Der Gang des Menschen. Published by B.G. Teubner. Translation by P. Maquet and R: Furlon: The human gait. Published by Springer, Berlin/Heidelberg, New York, London, Paris, Tokyo, 1987 Cappozzo A, Marchetti M (1992) Borelli’s heritage. In: Cappozzo A, Marchetti M, Tosi V (eds) Biolocomotion: a century of research using moving pictures. Promograph, Roma, pp 33–47 Leonardo da Vinci (1510) Human proportions. In: Anatomical drawings, Royal Library – Windsor Castle, pp 143–147. http://www.metmuseum.org/art/metpublications/leonardo_da_vinci_ana tomical_drawings_from_the_royal_library_windsor_castle# Dürer A (1528) VierBücher von Menschlicher Proportion (Four Books on Human Proportion). Published by Hieronymus Formschneyder, Nuremberg Eberhart HD (1947) Fundamental studies of human locomotion and other information relating to design of artificial limbs. Prosthetic Devices Research Report. University of California, Berkeley Iwaki H, Pinskerova V, Freeman MA (2000) Tibiofemoral movement: the shapes and relative movements of the femur and tibia in the unloaded cadaver knee. J Bone Joint Surg Br 82 (8):1189–1195 Jansons H (1992) Bernstein: the microscopy of movement. In: Cappozzo A, Marchetti M, Tosi V (eds) Biolocomotion: a century of research using moving pictures. Promograph, Roma, pp 137–174 Jarret MO, Andrews BJ, Paul JP (1976) A television/computer system for the analysis of human locomotion. In: IERE golden jubilee conference on the applications of electronics in medicine. IERE Conference Proceedings No. 34 JBdecM S, Inman VT, Eberhart HD (1953) The major determinants in normal and pathological gait. J Bone Joint Surg Am 35:543–558 Law AH, Gordon M (1995) Meyerhold, Eisenstein and biomechanics: actor training in revolutionary Russia. Published by McFarland & co, Jefferson Maquet P (1992) “The human gait” by Braune and Fischer. In: Cappozzo A, Marchetti M, Tosi V (eds) Biolocomotion: a century of research using moving pictures. Promograph, Roma, pp 115–126

Observing and Revealing the Hidden Structure of the Human Form in Motion. . .

15

Marey E-J (1873) La Machine Animale, Locomotion TerrestreetAérienne. Published by G. Baillie, Paris Marey E-J (1885) La Méthode Graphique dans les Sciences Expérimentales. Published by G. Masson, Paris Marey E-J (1894) La Méthode Graphique dans les Sciences Expérimentales. Second édition avec supplément. Le Développement de la Méthode Graphique par la Photographie. Published by G. Masson, Paris Morris JRW (1973) Accelerometry – a technique for the measurement of human body movements. J Biomech 6(6):729–732 Muybridge E (1887) Animal Locomotion: an electro-photographic investigation of consecutive phases of animal movements, commenced 1872 – completed 1885. Published under the auspices of the University of Pennsylvania by J.B. Lippincott Co Muybridge E (1899) Animals in Motion. Published by Chapman and Hall (reprinted by Dover, New York, 1957) Muybridge E (1901) The human figure in motion. Published by Chapman and Hall (reprinted by Dover, New York, 1955) Paul JP (1966) Biomechanics. The biomechanics of the hip-joint and its clinical relevance. Proc R Soc Med 59(10):943–948 Paul JP (1967) Forces at the human hip joint. PhD thesis. http://theses.gla.ac.uk/3913/ Paul JP (1969) Loading on the head of the femur. J Anat 105:187–188 Paul JP (1992) The Californian contribution. In: Cappozzo A, Marchetti M, Tosi V (eds) Biolocomotion: a century of research using moving pictures. Promograph, Roma, pp 176–195 Poore, H.R., 1913. The new tendency in art: post impressionism, cubism, futurism. Published by Cornell University Library's print collections and scanned on an APT BookScan and converted to JPG 2000 format by Kirtas Technologies in 2009 Selvik GA (1974) Roentgen stereophotogrammetric method for the study of the kinematics of the skeletal system. Ph.D. thesis, University of Lund, Sweden Selvik GA (1990) Roentgen stereophotogrammetric analysis. Acta Radiologica 31(2):113–126 Stemp, R. (2006) The secret language of the Renaissance: decoding the hidden symbolism of Italian Art. Published by Duncan Baird, London Stillman JDB (1882) The horse in motion as shown by instantaneous photography, with a study on animal mechanics founded on anatomy and the revelations of the camera, in which is demonstrated the theory of quadrupedal locomotion. Published by J. R. Osgood & Co., Boston Tosi H (1992) Marey and Muybridge: how modern biolocomotion analysis started. In: Cappozzo A, Marchetti M, Tosi V (eds) Biolocomotion: a century of research using moving pictures. Promograph, Roma, pp 51–70

Three-Dimensional Reconstruction of the Human Skeleton in Motion Valentina Camomilla, Aurelio Cappozzo, and Giuseppe Vannozzi

Abstract

This chapter illustrates the conceptual background underlying the in silico reconstruction of the human skeletal motion. A specific focus is given to the experimental and analytical methods that allow acquiring information related to both bone movement and morphology in vivo in the framework of rigid body mechanics. This process involves the definition of global and local frames of reference. Common anatomical and mathematical conventions that are used to describe global bone pose and joint kinematics are illustrated. Issues concerning accuracy and reliability of the estimated quantities when using skin markers and stereophotogrammetry and magneto-inertial measurement units are also dealt with. Keywords

Rigid body mechanics • Human movement analysis • Bone pose estimation • Anatomical calibration • Joint kinematics

Contents Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . State of the Art . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Global and Local Frames: Terminology . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Global Frames . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Local Frames . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

18 18 22 22 24

V. Camomilla (*) • A. Cappozzo Interuniversity Centre of Bioengineering of the Human Neuromusculoskeletal System, Department of Movement, Human and Health Sciences, University of Rome “Foro Italico”, Rome, Italy e-mail: [email protected]; [email protected] G. Vannozzi Interuniversity Centre of Bioengineering of the Human Neuromusculoskeletal System, Department of Movement, Human and Health Sciences, University of Rome Foro Italico, Rome, Italy e-mail: [email protected] # Springer International Publishing AG, part of Springer Nature 2018 B. Müller, S.I. Wolf (eds.), Handbook of Human Motion, https://doi.org/10.1007/978-3-319-14418-4_146

17

18

V. Camomilla et al.

Estimate of the Bone-Embedded Frame Pose . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Stereophotogrammetry . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Magneto-Inertial Measurement Units . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Anatomical Calibration . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Subject-Specific Morphology Data . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Movement-Morphology Data Registration . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Construction of the Anatomical Frame . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Joint Kinematics Estimation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Translational Degrees of Freedom . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Rotational Degrees of Freedom . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Precision and Accuracy of Joint Kinematics Estimates . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Future Directions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

25 26 28 28 28 31 33 33 36 36 38 39 41

Introduction Protecting and enhancing human motor function is an important strategic aim of modern society dwelling within the grand challenges concerning health and wellbeing (personalized care, prevention and early diagnosis, integrative and holistic approaches, healthy aging, sustainability of chronic diseases). Efforts must be made to strengthen our multi-scale understanding of both structure and function of the human locomotor system and to develop investigative and operational methods that can be translated into professional practice. In this context, quantitative human movement analysis is of paramount importance, and, for many applications, it must be carried out in the three-dimensional space. This analysis is based on measurements (motion capture, dynamometry, electromyography, calorimetry, medical imaging, etc.) and computational modeling (mathematical models of the anatomy and physiology of the tissues, organs, and systems involved). It provides information on the functions of the locomotor subsystems and on the overall strategy of motor activity. These outcomes contribute to the understanding of the key factors that affect joint motion and internal loading and, thus, injury, tissue degeneration or regeneration, as well as motor control and its adaptation, energy consumption, and fatigue. Quantification of subjectspecific variables can be effectively used in prevention, early diagnosis (e.g., monitoring of functional status in the elderly, specific workers, or athletes), intervention (e.g., prognosis, therapeutic programming, workplace optimization, training), and quantifying relevant outcomes. Prospectively, advanced movement analysis technology may be used for the implementation of real-time biofeedback (virtual and augmented reality) both in rehabilitation and training (institutionalized or not).

State of the Art A prerequisite for many of the above-mentioned endeavors is the accurate threedimensional reconstruction in silico of the portion of the locomotor system of interest while performing a motor task. The solution of this problem is based on the following fundamental considerations.

Three-Dimensional Reconstruction of the Human Skeleton in Motion

19

A body may be thought to be made of P particles. The larger the number P, the more details of the body shape are portrayed. The motion of this body may be described by providing information about the motion of each of those particles in the form, for instance, of three Cartesian coordinates (or a position vector), relative to a given set of orthogonal axes (global frame), and their time derivatives, in each sampled instant of time (1. . .N ). The resulting set of numbers describing position is made of 3
P
N numbers. This data set may be split into two parts: one describing movement and the other describing shape. This is a profitable exercise because, if we are not interested in shape variation, i.e., in deformation, the second data subset may be considered to be time invariant, and this results in a remarkable reduction of the data set dimension with the obvious experimental and computational advantage. This is achieved through the following procedure. A set of orthogonal axes, generally named local frame (l), is defined rigidly associated with the body while assuming its current shape (Fig. 1). The location in space of this frame relative to the global frame (g), in a given instant of time, is described using two separate bits of information: one deals with its position and the other with its orientation (also named attitude). The former information is provided by the position vector of the origin of the local frame (Fig. 1): g

o¼


g

ox g oy g oz



(1)

The orientation of the local frame may be described by providing the orientation of each of its axes. This is done through the components of the three unit vectors aligned with these axes, relative to the global frame axes. Through obvious geometrical considerations, it can be seen that these components are equal to the cosines (named direction cosines) of the angles formed by each unit vector with the global frame axes: g

h i uxl ¼ cos θxl xg cos θxl yg cos θxl zg

(2)

Fig. 1 Global (gx, gy, gz) and local (lx, ly, lz) frames. P represents a particle of the body of interest and gp and lp its position vectors in the two frames, respectively. Left superscripts denote the frame with respect to which a position vector or an orientation of a frame is represented. Subscripts denote the entity the vector or the matrix describes

20

V. Camomilla et al. g

h i uyl ¼ cos θyl xg cos θyl yg cos θyl zg

(3)

h i uzl ¼ cos θzl xg cos θzl yg cos θzl zg

(4)

g

Equations 2, 3, and 4 are normally represented in a matrix named orientation matrix (also rotation or attitude matrix): 2

cos θxl xg g Rl ¼ 4 cos θxl yg cos θxl zg

cos θyl xg cos θyl yg cos θyl zg

3 cos θzl xg cos θzl yg 5 cos θzl zg

(5)

It is useful to mention here that given the position vector and the orientation matrix of the local frame, relative to a global frame, and the position of a point (or of an ensemble of points, that is, a body) in the local frame (Fig. 1), it is possible to represent this position in the global frame through the following operator (coordinate transformation): g

p ¼ g Rl l p þ g ol

(6)

Of course the use of this operator may be generalized, since it allows representing the position of a point given in a reference frame to any other referred frame. It is important to note that the nine direction cosines are not independent; in fact the sum of the squares of each triplet equals one (the amplitude of the unit vector), and the vector product of each pair of unit vectors, being orthogonal, equals zero. The resulting six relationships leave only three of the nine direction cosines independent. In summary, the orientation of the local frame relative to the global frame is fully described by three scalars, which, added to the three scalars necessary to describe position (Eq. 1), makes a total of six scalars (the degrees of freedom of the body). The ensemble of position and orientation of a local frame and, thus, of a body, is named pose. Another and more compact way of representing the orientation of a body, i.e., of a local frame, is to exploit the Euler’s theorem and use the kinematic variables that describe the fictitious rotation that takes the local frame from being parallel to the global frame to the target orientation. We can achieve this by decomposing this rotation into three finite successive rotations about body-fixed axes in their current orientation. The corresponding three rotation angles, called Euler angles, completely describe the given rotation. There are twelve possible sets of Euler angles. Six imply the first and third rotation about the same axis (symmetric sets), and six imply rotations about the three different axes, in all possible combinations (asymmetric sets). The latter angle sets are also referred to as Cardan or Bryant angles. A second approach explicitly identifies the axis of rotation n (unit vector) and the angle of rotation θ that realize the target rotation and uses the following orientation (or rotation) vector: θ ¼ θn

(7)

Three-Dimensional Reconstruction of the Human Skeleton in Motion

21

Both the orientation vector and the Euler angles are illustrated in standard kinematics and dynamics texts and are not discussed here (Shuster 1993). However, a few remarks will be made in a subsequent section with regard to the use of these quantities to describe joint kinematics. After the above-illustrated representation of the instantaneous pose of a body, it is seen that, if the body can be hypothesized to keep its shape unaltered during the analyzed movement, the number of scalars necessary to describe it is reduced from 3
P
N to 6
N + 3
P. Under these very favorable circumstances, we talk of a rigid body and use the part of classical mechanics that operates under the mentioned hypothesis. The question arises now whether the rigid body hypothesis is applicable to the constituents of the locomotor apparatus in motion. In this respect we have two orders of problems: one regards the mathematical modeling of the locomotor apparatus and the other the acquisition of experimental data. It is commonly accepted that if the focus of the investigation is skeletal motion, then bones may be considered to be rigid bodies without having a significant impact on the end results. When aiming at the estimation of internal loads or mechanical energy, then entire body segments are involved in the modeling exercise and may be disputable to consider them rigid bodies due to the fact that deformation and displacements occurring during movement of muscular and visceral masses may significantly change some inertia parameters with consequent inertial effects (mass moments of inertia and location of the center of mass; Clark and Hawkins 2010; Pain and Challis 2001; Zelik and Kuo 2010) (chapter ▶ “Estimation of the Body Segment Inertial Parameters for the Rigid Body Biomechanical Models Used in Motion Analysis”). It is evident that this issue becomes critical when high accelerations are involved (Challis and Pain 2008; Gruber et al. 1998; Liu and Nigg 2000; Riddick and Kuo 2016). An experimental issue consists on the fact that bone movement can be accurately recorded only using invasive techniques (intracortical pins or medical imaging involving the use of ionizing radiations). As better illustrated later, under normal circumstances, we can only capture the movement of 3-D markers (through stereophotogrammetry or magneto-inertial measurement units) attached to the skin surface above the bony segment of interest. When using stereophotogrammetry, the 3-D marker is constructed using three or more point markers (marker cluster). A 3-D marker is unavoidably mobile with respect to the underlying bone due to the interposed soft tissue, and, thus, its global motion defers from that of the bone. In general, this local displacement of the marker is caused by skin sliding associated joint movement, soft tissue volumetric deformation due to muscular contraction, gravity, and inertial effects on relevant masses (wobbling). In this context, the relative motion between marker and underlying bone is to be regarded as an artifact, the so-named soft tissue artifact, which, if not properly dealt with, has very serious consequences on the reliability of the results of the analysis (Cappozzo 1991; Garling et al. 2007; Lamberto et al. 2016; Leardini et al. 2005; Li et al. 2012). If the objective is a kinetic or energy analysis, then quantitative information concerning the abovementioned soft tissue relative motion, which is concerning the instantaneous body segment shape and mass distribution during movement and the estimate of the

22

V. Camomilla et al.

consequent time-varying inertia parameters, may be required posing an experimental problem very difficult to be dealt with using the present state of the art of measurement technology (Wakeling and Nigg 2001) (chapter ▶ “Simulation of Soft Tissue Loading From Observed Movement Dynamics”). In the general framework illustrated above and accepting the related limitations, in human and animal movement analysis, rigid body mechanics is normally used to model the locomotor apparatus. The objective of this chapter is to illustrate the way this is done. From now on we will refer to the estimate of bone pose being confident of the fact that, to all practical purposes, bones may be considered rigid bodies and that models of the relevant soft tissues may be associated to the bone in its current pose. Note that most analyses demand not only the estimate of the instantaneous pose of the bones involved but also the full reconstruction of their outer surface in the 3-D in silico space. It is important to acknowledge the fact that the quantitative description of human motion involves the use of kinematic and kinetic vector components, the values of which depend on the Cartesian reference axes with respect to which they are defined. Thus, for the sake of their repeatability, equally repeatable set of axes must be made available. As will be clearer later, it is easy to accomplish this requirement for the global frames, while, by cause of technical difficulties and morphological complexity, it is more difficult for the local frames associated with the bones (Della Croce et al. 1999, 2005). Furthermore, these sets of axes must be defined so that the resulting kinematic and kinetic scalar quantities have a clear functional meaning and effectively and consistently describe the anatomical and functional entities used in the medical-biological literature.

Global and Local Frames: Terminology As reported above, describing the skeletal-system movement involves the use of sets of coordinate system axes that define the global and the local frames. The definition and the methods for identifying these coordinate systems vary with the objective of the analysis and the measuring instrument involved.

Global Frames In a movement analysis laboratory, the following inertial, global frames can be defined (Fig. 2; Cappozzo et al. 1995, 1997b). Motion capture system frame: it is composed of the set of axes used by either the stereophotogrammetric system or the magneto-inertial system to represent pointmarker position or 3-D marker orientation, respectively. This is arbitrarily defined during the motion capture system calibration procedure.

Three-Dimensional Reconstruction of the Human Skeleton in Motion

23

Fig. 2 Movement analysis laboratory equipped with a stereophotogrammetric system and two force plates. Global frames are depicted. If a locomotor act is analyzed, the motor task frame may coincide with the frame of one of the two force plates

Motor task frame: this frame is consistent with the analyzed motor task and sometimes describes its basic features. For instance, when locomotor acts are investigated, one axis of the frame indicates the mean direction of progression. According to the general recommendations from the International Society of Biomechanics (Wu and Cavanagh 1995), in human locomotion analysis, right-handed orthogonal coordinate systems should have the x-axis pointing in the direction of progression, y pointing vertically upward, and z pointing to the right. Dynamometer frame: this is the frame in which force and moment components are given by the instrument and is defined during its calibration. Plumb line: this is a single axis and represents the orientation of the gravity line, usually assumed to point downward. Therefore, within the same experiment, different mechanical quantities can be measured with respect to different global frames. However, normally, their interpretation, or their use as input to musculoskeletal models allowing the estimation of further non-measurable quantities, requires that all of them be represented in the same frame (primary global frame). The latter role is usually assumed by the motor

24

V. Camomilla et al.

task frame. This procedure involves the determination of the position vector and the orientation matrix of all secondary global frames involved relative to the primary frame, allowing to obtain any vector quantity in the primary frame using Eq. 6. In order to achieve this result, the dynamometers and the pathway, staircase, or any other implement used to perform the motor task must be equipped with a 3-D marker, the pose of which may be detected by the motion capture system being used (Rabuffetti et al. 2003).

Local Frames We deal here with the frames associated with a bony segment (Fig. 3). Anatomical frame. This frame is made of orthogonal axes that exhibit the following properties: (i) compatibility with the anatomical axes and planes defined in the anatomy literature and (ii) intra- and inter-subject repeatability. The concept of repeatability must, of course, be extended to the portability of the results of movement analysis among different laboratories (Benedetti et al. 2013). This entails the standardization of the definition of the anatomical frames (Cappozzo et al. 1995; Wu and Cavanagh 1995; Wu et al. 2002). The following local frames depend on the specific technique used for the determination of their pose. For this reason they are named technical frames. Bone-embedded frame. This is a frame rigidly connected with the bone, but with no repeatable relationship with its anatomy. Motion technical frame. This is the frame the pose of which is provided by the motion capture system and related data processing. It, normally, represents an estimate of the bone-embedded frame. Fig. 3 Local frames

Three-Dimensional Reconstruction of the Human Skeleton in Motion

25

Morphology technical frame. This is the frame used by the measurement instrument (i.e., imaging apparatus, such as MRI or CT) that provides a digital model of the bone and, eventually, of the relevant soft tissues.

Estimate of the Bone-Embedded Frame Pose Bone movement may be directly tracked using medical imaging techniques. X-ray fluoroscopy, either planar or biplanar, is mostly used to record a single joint movement (Banks and Hodge 1996). This technique, however, exhibits the following limitations: it uses ionizing radiation, the measurement volume and the sampling frequency are limited, and the procedure is operator time-consuming. Recently systems have been developed that move alongside the subject, thus overcoming the limitation of the measurement volume, at least in one dimension (Guan et al. 2016) (chapter ▶ “Measurement of 3D Dynamic Joint Motion Using Biplane Videoradiography”). Other techniques are under development that exploit magnetic resonance and ultrasonic technologies (chapter ▶ “3D Musculoskeletal Kinematics Using Dynamic MRI”). In this chapter we deal only with the motion capture techniques mostly used in human movement analysis that track bone motion indirectly, but display the advantages of being noninvasive, having large measurement volumes and high sampling rates: stereophotogrammetry associated with skin markers, or magneto-inertial measurement units (Fig. 4). The fundamental concepts illustrated regarding the use of

Fig. 4 Input and output data of bone pose estimators using skin markers and stereophotogrammetry (a) or magneto-inertial measurement units (b)

26

V. Camomilla et al.

rigid body mechanics for the modeling of the human locomotor apparatus in motion are, however, valid whatever technique is used to monitor this motion.

Stereophotogrammetry Optoelectronic stereophotogrammetry is, to date, the most widely used solution for measuring skeletal kinematics. It is made of a number of video cameras, connected to a computer, the fields of view of which intersect defining the measurement volume. This system provides the position (Cartesian coordinates) of point markers, either emitting or retro-reflecting light, located on the skin surface relative to a global frame. This is done through mathematical operators that receive the 2-D coordinates of the point-marker images, measured in the image plane of at least two cameras at any given instant of time, and parameters that describe the location in space and optical features of the cameras. These parameters are obtained through the calibration of the stereophotogrammetric system. The reconstructed positions in the global frame of three or more nonaligned skin markers (marker cluster or 3-D marker) located above the bone of interest and a mathematical estimator are used to construct a motion technical frame and determine its pose relative to the global frame (Fig. 4) (chapters ▶ “3D Dynamic Pose Estimation from Markerless Optical Data” and ▶ “3D Dynamic Probabilistic Pose Estimation from Data Collected Using Cameras and Reflective Markers”). Markers are located in such a way to comply with technical requirements such as visibility to a sufficient number of cameras, maximal relative distance between markers (Cappozzo et al. 1997a), and minimal relative movement between them and the underlying bone. Virtual markers may also be used to construct a marker cluster. A virtual marker is a point, the position of which is calculated using the positions of the available physical markers and a geometric or statistical rule. A typical example of virtual marker is a point approximating a joint center, which is in common between two adjacent bones. If the pose of one bony segment is successfully reconstructed, the global position time history of the virtual marker, the position of which in this segment is known, can be obtained using Eq. 6 and used to reconstruct the pose of the second body segment as if it were a physical marker. This procedure allows reducing the number of physical markers to be tracked. Note that, while this economical attitude could be justified years ago, this is not the case with current optical motion capture technology that can deal with a very high number of markers simultaneously. A marker cluster observed, for instance, in a selected instant of time or subject’s posture, is taken as a rigid cluster model. Using some geometric rule, a set of orthogonal axes is associated to it (model frame). If the skin markers were rigidly attached to the underlying bone, the rigid cluster model could be superimposed onto the current marker cluster and the model frame taken as motion technical frame. Unfortunately, this is not the case: during movement, the reconstructed

Three-Dimensional Reconstruction of the Human Skeleton in Motion

27

Fig. 5 The artifact displacements of the skin markers (arrows), occurring in a given interval of time, result in a rigid transformation (a translation plus a rotation) plus a nonrigid transformation (a change in size and shape) of the marker cluster (Andersen et al. 2012; Barré et al. 2013; Benoit et al. 2015; De Rosario et al. 2012; Dumas et al. 2014; Grimpampi et al. 2014). The former transformation has been proved to be the dominant part (Andersen et al. 2012; Barré et al. 2013; Benoit et al. 2015; Bonci et al. 2015; Dumas et al. 2015)

marker local positions undergo variations due to stereophotogrammetric errors (Chiari et al. 2005) and, as already mentioned, the soft tissue deformation (Leardini et al. 2005). Thus, the marker cluster changes its shape and moves relative to the bone (Fig. 5). As a consequence, the above-mentioned superimposition exercise must be carried out using an approximation approach (“single-body optimization”). This consists in a least-squares method, the most popular of which minimizes the sum of the squared distances between corresponding points of the cluster model and of the marker cluster, the so-named Procrustes distance (Soderkvist and Wedin 1993). In this way the instantaneous pose of the motion technical frame can be determined and taken as an estimate of the bone-embedded frame. Instead of dealing with one marker cluster, and thus one bone, at a time, it is possible to carry out the least-squares superimposition procedure illustrated above using a multi-body system made of a chain of rigid cluster models connected by kinematic models of the joints involved (“multi-body kinematics optimization”). The joint models proposed in the literature typically involve major simplifications with respect to real and subject-specific joints and have less than six independent degrees of freedom (Charlton et al. 2004; Duprey et al. 2010; Li et al. 2012; Lu and O’Connor 1999; Reinbolt et al. 2005; Richard et al. 2016; Scheys et al. 2011). It should be emphasized that while instrumental errors that are random in nature have no significant effect on the end result, the illustrated procedures (single-body and multi-body kinematics optimization) do not attenuate the propagation of the largest portion of the soft tissue artifact (Andersen et al. 2010; Bonci et al. 2015; Clément et al. 2015, 2017; Gasparutto et al. 2015; Li et al. 2012). The latter issue remains the greatest obstacle to the accurate reconstruction of skeletal movement and deserves further attention by the human movement analyst community.

28

V. Camomilla et al.

Magneto-Inertial Measurement Units Miniature magneto-inertial measurement units, embedding a microprocessor and often endowed with wireless communication technology, are an increasingly popular alternative to stereophotogrammetry for 3-D human movement analysis (chapter ▶ “Three-Dimensional Human Kinematic Estimation Using Magneto-Inertial Measurement Units”). These units are attached to the body segment of interest. They comprise three-axis linear accelerometer and angular rate sensors and a three-axis magnetometer. The physical quantities provided by each sensor are measured with respect to the axes of a unit-embedded frame generally aligned with the edges of the unit case. Through algorithms able to fuse the redundant information available and compensate for sensor noise and drift, the 3-D orientation of the unit-embedded frame relative to a global frame is provided. The unit-embedded frame is assumed to be the motion technical frame and, again, despite the soft tissue artifact, is assumed to be an estimate of the bone-embedded frame (Fig. 4). As opposed to stereophotogrammetry, magneto-inertial measurement units do not supply reliable positional information. The advantage of this technique is that it does not impose limits to the measurement volume.

Anatomical Calibration So far we have seen how a bone-embedded frame, as observed during the execution of a motor task, may be reconstructed in silico in its current pose. The next step is the representation of a digital model of the relevant bone in this frame. In order to comply with the repeatability issues discussed in a previous section, this digital model must carry an anatomical frame with it. The experimental acquisition of the parameters that allow the construction in silico of the bone model in its current pose and of the relevant anatomical axes is referred to as “anatomical calibration.”

Subject-Specific Morphology Data The 3-D digital model of the bone of interest may be defined at different levels of resolution that depend on the number of points used to describe the bone external surface. The minimum number of these points is 3, as imposed by the possibility of constructing an anatomical frame. A larger number of points allow a more realistic rendering of the bone in silico and a more accurate association of the soft tissue digital models to the bone. A subject-specific bone digital model may be measured or estimated using medical imaging. A full reconstruction of that model may be obtained using magnetic resonance, but this is rarely possible for logistic and economic reasons. A way of estimating a subject-specific bone digital model, with sufficient accuracy for most purposes, is through two planar X-rays of the bone and a relevant statistical model and shape recognition algorithm (Chaibi et al. 2012). This method is made applicable

Three-Dimensional Reconstruction of the Human Skeleton in Motion

29

by a low-dose X-ray imaging technology performed at a low dose and with an expanded dynamic range that allows for whole-body scanning while the subject is under normal weight-bearing conditions (Melhem et al. 2016). It is evident that with all methods that involve medical imaging, the bone digital model is represented in a morphology technical frame that is different from the motion technical frame. For a number of reasons that will be illustrated later, the subject-specific bone digital model must carry labeled anatomical landmarks (sites with recognizable anatomical features). Given a clear definition of these landmarks, this labeling may be carried out in silico through a virtual palpation (Van Sint Jan et al. 2003). An alternative approach is the estimate of the subject-specific bone model using subject-specific partial information collected in the movement analysis laboratory and a template bone model that is made to match the above-mentioned morphological information. This partial information may consist of the 3-D position of isolated points or clouds of adjacent points that lie on the bone surface and of lines oriented as anatomical axes or lie in anatomical planes. In this case, morphological information is available in the motion technical frames. The position of anatomical landmarks may be determined with an ad hoc stereophotogrammetric acquisition by either temporarily locating skin markers over them (Cappozzo 1984) or using a wand equipped with a marker cluster (Fig. 6a; Cappozzo et al. 1995), after identification by manual palpation (Van Sint Jan et al. 2003). In some experimental protocols, skin

Fig. 6 Anatomical calibration in the stereophotogrammetric laboratory. (a) Skin markers are located consistently with technical requirements; anatomical landmarks are identified by manual palpation and their position measured using a wand. (b) Some skin markers are located on anatomical landmarks or lying in an anatomical plane. (c) As in (a), but portions of the bone surfaces are digitized by moving the tip of the wand over them. The center of the acetabulum is estimated using mathematical models or identified through a functional approach

30

V. Camomilla et al.

markers used to track motion are located over anatomical landmarks, the position of which is therefore readily available, or in an anatomical plane (Fig. 6b; Davis et al. 1991; Frigo et al. 1998; Leardini et al. 2007) (chapters ▶ “The Conventional Gait Model - Success and Limitations” and ▶ “Variations of Marker Sets and Models for Standard Gait Analysis”). More detailed morphological information may be acquired by using the above-mentioned wand over clouds of points of the bone covered with a layer of soft tissue that allows for their palpation through the skin (Fig. 6c; Donati et al. 2007, 2008). To these points, internal anatomical landmarks may be added when noninvasively identifiable. The position of these landmarks may be determined as a function of the position of other accessible anatomical landmarks and/or readily available anthropometric measures using population statistical models. Typical examples in this respect are the regression equations used to determine the position of the center of the acetabulum (Bell et al. 1990; Davis et al. 1991; Hara et al. 2016; Harrington et al. 2007; Leardini et al. 1999; Seidel et al. 1995) or the position of the clavicle, scapula, and humeral bone (Sholukha et al. 2009). In some cases, an internal anatomical landmark may be considered to coincide with a joint center of rotation (▶ “NextGeneration Models Using Optimized Joint Center Location”). This allows determining the position of these landmarks using a “functional approach.” This entails performing an ad hoc experiment during which the subject, equipped with suitable markers (skin-marker clusters or magneto-inertial measurement units), is asked to execute a 3-D movement of the joint involved. The time history of the pose of one bone relative to the other forming the joint is estimated and used to determine the position of the joint center and, thus, of the relevant anatomical landmark (Crabolu et al. 2016; Halvorsen 2003; McGinnis and Perkins 2013). An example in this respect is the center of the acetabulum that can be assumed to coincide with the center of the femoral head and the center of rotation of the femur relative to the pelvic bone (Fig. 7; Camomilla et al. 2006; Cappozzo 1984; Cereatti et al. 2010; Kainz et al. 2015; Leardini et al. 1999; Piazza et al. 2004). Similarly, the center of the Fig. 7 Movement normally used for the estimate of the center of the acetabulum using the functional approach. The figure depicts the trajectory of a point located on the distal portion of the femur and represented in the transverse plane of the pelvis. Numbers represent the movement sequence (Camomilla et al. 2006)

Three-Dimensional Reconstruction of the Human Skeleton in Motion

31

head of the humerus may be identified as the rotation center of the glenohumeral joint (Campbell et al. 2009; Lempereur et al. 2010). Lines that lie in anatomical planes also carry useful morphological information. By suitably choosing a joint movement, the resulting rotation axis may be assumed to have this property. Such an axis may be determined through the already mentioned functional approach (Gamage and Lasenby 2002; Halvorsen et al. 1999). Examples in this respect are the knee (Colle et al. 2016; De Rosario et al. 2017; Ehrig et al. 2007), the elbow joint (Fraysse and Thewlis 2014), and the talocrural joint (Sheehan 2010; van den Bogert et al. 1994) moving in the sagittal plane. The estimated rotation axis is assumed to lie in the frontal plane and, together with at least an anatomical landmark, allows the determination of this plane. It is worth emphasizing that, as opposed to what in some cases may appear in the literature, an axis or line, obtained using this method, is not an anatomical axis per se, although sometime it may be considered to approximate it. When magneto-inertial measurement units are used, the position of anatomical landmarks cannot be reliably measured. The only morphological information that can be collected concerns the orientation of lines that lie in anatomical planes. This information can be obtained using either a functional approach (Bouvier et al. 2015; Cutti et al. 2008, 2010; Favre et al. 2009; Luinge et al. 2007; Seel et al. 2012) or a specifically designed calibration device consisting in a rod carrying a magneto-inertial measurement unit and two mobile pointers perpendicular to it (Fig. 8). This unit provides the orientation relative to the motion technical frame, made available by the unit mounted on the body segment of interest, of a line joining two palpable anatomical landmarks pointed by the calibration device (Picerno et al. 2008).

Movement-Morphology Data Registration After the subject-specific morphological information is collected, the entire subjectspecific digital model of the bone must be associated to the motion technical frame in each sampled instant of time during the analyzed movement and, therefore, represented in silico in its current pose relative to the global frame. When the entire subject-specific digital model is made available, either measured or estimated through medical imaging, and represented in a morphology technical frame different from the motion technical frame, a transformation of the position vectors given in the former frame into position vectors in the latter frame must be carried out (the procedure is known as movement-morphology data registration). To this purpose Eq. 6 can be used, provided that the anatomical calibration procedure includes the determination of the orientation matrix and origin position vector of one frame with respect to the other. This may be achieved by having the position of at least three points in both frames and applying a superimposition procedure similar to that described in section “Estimate of the Bone-Embedded Frame Pose.” These points may be anatomical landmarks or, when using stereophotogrammetry, the skin markers. In the latter circumstance, the medical imaging procedure must be

32

V. Camomilla et al.

Fig. 8 Anatomical calibration using magneto-inertial measurement units. The orientations of the two dashed lines shown in the figure are detected by the unit mounted on the calibration device and define the sagittal plane of the femur. The line joining the medial epicondyle (LE) and the greater trochanter (GT) may be used for the orientation of the y-axis. The orientation of the other two axes is, thus, available as well. If the center of the femoral head is identified using the functional approach (Crabolu et al. 2016), a digital model of the femur may be matched with the available subject-specific morphological information and the definition of the femoral anatomical frame illustrated in Fig. 9 may be used

carried out with the subject carrying the marker set that will be used for the motion tracking. While doing this with the subject lying on a horizontal surface, as normally occurs with magnetic resonance, we may incur in a problem. His/her soft tissues are deformed in a different fashion then when assuming the static or dynamic postures under analysis. The thus obtained pose of the motion technical frame relative to the morphological technical frame is, therefore, different from that occurring during the motor task of interest. If partial subject-specific morphology information is collected in the motion capture laboratory, then the subject-specific bone model in its current pose may be obtained by matching a suitably chosen template bone model, provided in a morphology technical frame, with the available morphological data available in the motion technical frame. To this purpose a superimposition procedure associated with non-isomorphic scaling and a reorientation of the template model may be used (Chaibi et al. 2012; Donati et al. 2007, 2008; Quijano et al. 2013). In this case, the bone model is already represented in the motion technical frame, and, therefore, no registration procedure is required.

Three-Dimensional Reconstruction of the Human Skeleton in Motion

33

The bone model thus obtained normally carries a number of labeled anatomical landmarks or lines, those that have been used for its construction and registration. If the position of all anatomical landmarks required for further processing is not provided by this procedure, then a virtual palpation or geometric construction is carried out on the bone model. An example of geometrical construction is that of a line that may be supposed to lie in the frontal plane of the femur. This is the so-named cylinder axis, which is the axis that joins the centers of the two spheres that fit the medial and lateral posterior condyles of the femur and that is supposed to lie in the femur frontal plane (Yin et al. 2015).

Construction of the Anatomical Frame Once the subject-specific digital model of a bone is available, the relevant anatomical frame must be constructed. As already mentioned, this is done using the position of anatomical landmarks, and/or points or lines that lie on an anatomical plane, and a geometric rule. The anatomical frame may be constructed in the morphology technical frame or in the motion technical frame and, then, represented in whatever other frame it is required through the transformation represented by Eq. 6. It is evident that, given the many anatomical landmarks and other morphological features of the bone that can be made available, many anatomical sets of axes may be defined and have, in fact, been defined and illustrated in the literature (Fig. 9; Cappozzo et al. 1995; Kadaba et al. 1990; Wu and Cavanagh 1995; Wu et al. 2002). As already mentioned, while defining the term “anatomical frame,” the issue here is the inter- and intra-subject repeatability of this set of axes and their portability. Critical factors are the following. Anatomical landmarks are areas and not points, as assumed to be in the present construction, and their definition may slightly change depending on the source that describes them. When an operator palpates an external anatomical landmark, these circumstances may lead to large inter- and intra-operator variability as quantified in Della Croce et al. (1999). The uncertainty that affects internal anatomical landmarks or lines, such as those assumed to coincide with a joint center or a joint rotation axis, may also impact very negatively on repeatability (Stagni et al. 2000). By merging the human movement and the ever-developing medical imaging laboratories, we may be able, in the next future, to find better solutions to this problem. In addition, these enhanced solutions should be shared within the human movement analyst community and possibly undergo a standardization process for the sake of data portability.

Joint Kinematics Estimation The assessment of the stability and mobility of a joint is based on the observation of the relative movement of the two adjacent bones involved, that is, of joint kinematics. For each bone of interest, the procedures described in the previous sections

34

V. Camomilla et al.

Fig. 9 Example of anatomical frame definition (Cappozzo et al. 1995). Right-handed set of axes with the following characteristics (only two axes are defined because the third axis is the vector product of them). (a) Pelvis; origin: midpoint between the right anterior superior iliac spine (RASIS) and left anterior superior iliac spine (LASIS); zp-axis: oriented as the line passing through the RASIS and the LASIS with positive direction to the right; xp-axis: lies in the plane defined by the RASIS and LASIS and the midpoint between RPSIS and LPSIS and with its positive direction forward. (b) Femur; origin: midpoint between the lateral epicondyle (LE) and medial epicondyle (ME); yf-axis: it joins the origin with the center of the femoral head (FH) and its positive direction is proximal; zf-axis: it lies in the plane defined by FH, ME, and LE, and its positive direction is from left to right. (c) Tibia and fibula; origin: midpoint between the lateral malleolus (LM) and medial malleolus (MM); ys-axis: it is the line of intersection between the plane defined by the head of the fibula (HF), LM and MM (frontal plane), and the plane orthogonal to it passing through the tibial tuberosity (TT), LM, and MM; its positive direction is proximal; zs-axis: it lies in the frontal plane with positive direction from left to right

provide an estimate of the pose of an anatomical frame (only orientation if magneto-inertial sensors are used) relative to a global frame of choice in each sampled instant of time during movement (Eqs. 1 and 5). As already mentioned in a previous section, the relative pose between two rigid bodies is described by six scalar quantities, the six degrees of freedom; three describing the mutual orientation and three the mutual position. The objective of this section is to review the possible descriptions of joint kinematics and the issues associated with them. Given the orientation matrices and the position vectors of the anatomical frames of the proximal and distal bones of the joint under analysis relative to a global frame in any given instant of time (Fig. 10) g

Rp g op and g Rd g od ,

(8)

Three-Dimensional Reconstruction of the Human Skeleton in Motion

35

Fig. 10 Proximal and distal bone anatomical frames, their pose relative to the global frame (gRp, gop, and gRd, god respectively) and the pose of the distal bone relative to the proximal bone (pRd, pod)

we may represent the position vector of any given point in the global frame as a function of its position vector expressed in both local frames: p ¼ g Rp p p þ g op

(9)

p ¼ g Rd d p þ g od:

(10)

g

and g

By eliminating gp in Eqs. 9 and 10 and pre-multiplying both left and right terms of the resulting equation by the transposed matrix g RpT , p p may be calculated as p

  p ¼ g RpT g Rd d p þ g RpT g od  g op :

(11)

From this equation, and keeping in mind Eq. 6, it results that the orientation and position of the distal bone relative to the proximal bone, are given by p

  Rd ¼ g RpT g Rd and p od ¼ g RpT g od  g op , respectively:

(12)

Although pRd and pod fully describe joint kinematics, the scalar quantities that appear in them do not necessarily comply with the requirements of consistency with the anatomical and physiological terminology and thus of an effective description of function. This issue is tackled as illustrated in the following.

36

V. Camomilla et al.

Translational Degrees of Freedom The relative position of two adjacent bones in a given instant of time is described making reference to a vector (t) joining the position of a point defined in the proximal (Kp) and that of a point defined in the distal (Kd) local frames. These two points are normally chosen so that they coincide while the subject assumes a reference posture (zero joint translation). The translational degrees of freedom are described by the way the three components of vector t vary in time during the movement. Since during function translation is coupled with rotation, vector t depends on the location of the above-mentioned points. This location must therefore be chosen with care and accounting for the characteristics of the joint dealt with and based on the objective of the analysis. For instance, when dealing with the hip joint, Kd and Kp may be defined as coinciding with the mean center of rotation of that joint and associated with both pelvic bone and femur in a selected instant of time during movement or during a selected posture (Cereatti et al. 2010). If the analysis regards a joint that is more complex from the mechanical point of view, such as the tibiofemoral joint, than the midpoint of the transepicondylar axis (Fig. 11; Grood and Suntay 1983), a contact point between the two bones involved may be of interest. The other choice that needs to be made regards the set of axes with respect to which vector t is represented, again in each sampled instant of time. This set of axes may be that of the proximal or distal anatomical frame. Normally, the axes used to describe the three rotational degrees of freedom, illustrated later, are used in their current orientation. It is worthwhile noticing that, due to the fact that joint linear displacements are normally smaller or in the order of the resolution of most motion capture systems, there is little literature dealing with them and no shared convention for their description.

Rotational Degrees of Freedom The quantification of the rotational degrees of freedom is typically based on the observation, in the sampled instants of time during movement, of the orientation of Fig. 11 A possible definition of knee translation vector (t) during flexion. Kp is made to coincide with the midpoint of the transepicondylar axis; Kd and Kp are rigid with the tibia and with the femur, respectively, and they coincide during a selected reference posture

Three-Dimensional Reconstruction of the Human Skeleton in Motion

37

the distal bone relative to the proximal bone using a sequence of three rotations about the distal bone anatomical axes. The rotation sequence proposed by Grood and Suntay (1983) and recommended by Wu and Cavanagh (1995) and Wu et al. (2002), that is, the Cardan sequence, yields three angles that best match the way that functional anatomy uses for most joints (Fig. 12): (i) angle γ around zd (coinciding with zp) for flexion-extension, (ii) angle α around the current orientation of xd for abduction-adduction, and (iii) angle β around the current orientation of yd for internal-external rotation. If, during the illustrated sequence, each of these three axes is frozen with the orientation it assumes during the related rotation, a set of non-orthogonal axes is obtained (Fig. 12). The axis (xd) around which the second rotation occurs is referred to as floating axis, and the three axes are named joint axes. The equations that allow the calculation of the Cardan angles from the orientation matrix pRd are α ¼ sin 1p Rd ð3, 2Þ β ¼ sin 1 ½p Rd ð3, 1Þ= cos α

(13)

γ ¼ sin 1 ½p Rd ð1, 2Þ= cos α The demonstration of these equations can be found in the standard mechanical literature. Note that, if the second rotation equals π/2 or a multiple of it, then we incur in a singularity condition (gimbal lock). This circumstance prevents the use of the Cardan convention for those joints which, during movement, may undergo an abduction-adduction greater than π/2.

Fig. 12 The Cardan angles that describe an orientation of the distal bone relative to the proximal bone are defined according to the following sequence of operations: (i) to begin with the anatomical frames of the proximal ( p, hip bone) and distal bone (d, femur) are parallel; (ii) the first rotation γ occurs around the dz-axis; (iii) the second rotation α around the dx-axis in the orientation, it assumes after the first rotation (d1x); (iv) the third rotation β around the dy-axis in the orientation, it assumes after the second rotation (d2y). Axes in red (dz, d1x, d2y) are the joint axes and d1x is the floating axis

38

V. Camomilla et al.

Moreover, using this method, the instantaneous orientation of the distal bone is represented starting from a reference orientation that is equal to that of the proximal bone. Since during upright posture, the two anatomical frames are not necessarily parallel; the corresponding joint angles are not zero as often assumed to be in the biomedical literature. It is also important to remark that the three angles thus obtained univocally describe an orientation, but not true rotations. In addition, the Cardan angles, as more in general the Euler angles, do not enjoy the additive property. This means that, in principle, we cannot subtract two values of a given Cardan angle, observed in two different instants of time during movement or different postures, and, indeed, the result cannot be interpreted as an angular displacement, nor can we calculate relevant arithmetical averages (Pierrynowski and Ball 2009). For the same reason, when plotting a Cardan angle versus time, a continuous line should not be made to join two sampled values because this would imply an interpolation between them. We may conclude that, paraphrasing what Churchill is told to have said with regard to democracy, the Cardan angles is the worst form of joint rotation description, except for all the others. The second approach, definitely less popular than and as abstract as the previous one, is based on the orientation vector pθd applied to the orientation, observed in the sampled instants of time during movement, of the distal bone relative to the proximal bone anatomical frame (Fioretti et al. 1997; Woltring 1994). This vector may be represented in any set of local axes of choice, the proximal bone or the distal bone anatomical axes, or the non-orthogonal joint axes. This choice shall be tailored upon the consistency of the three resulting angles with functional anatomy. A positive property of this method is that it does not suffer the gimbal lock issue, but displays all other limitations of the Cardan angles. A third method to describe joint kinematics uses the following geometric approach (Paul 1992). The flexion-extension rotation is measured through the angle formed by the y-axis of the distal bone and the projection onto the xy (sagittal) plane of this bone of the y-axis of the proximal bone. The abduction-adduction rotation is given by the angle formed by the y-axis of the distal bone and the projection of the y-axis of the proximal bone onto the yz (frontal) plane of the distal bone. The internal-external rotation angle is the angle formed by the x-axis of the distal bone and the projection of the x-axis of the proximal bone onto the xz (transverse) plane of the distal bone. This approach is very intuitive; in fact it is similar to the way used in functional anatomy to describe joint motion. It is as abstract as the other methods, but it may deserve some attention from the human movement analyst community.

Precision and Accuracy of Joint Kinematics Estimates Two major issues affect the estimate of joint kinematics. One impacts on accuracy and it is mainly associated with the soft tissue artifact. The other affects precision and depends on the variability with which anatomical landmarks and, therefore, anatomical frames are determined.

Three-Dimensional Reconstruction of the Human Skeleton in Motion

39

Fig. 13 The six degrees of freedom of the knee joint during the support phase of running of a lean young adult able-bodied male subject. Black lines represent reference values obtained using clusters of markers mounted on intracortical pins (Cereatti et al. 2017). Gray lines represent the values obtained using clusters of skin markers and a single-body least-squares bone pose estimator (Camomilla et al. 2015)

In a preceding section, we have illustrated the origin of the soft tissue artifact and how and in what measure it impacts the estimate of bone pose while using noninvasive motion capture techniques. Figure 13 shows an example of the propagation of this artifact to the six degrees of freedom of the knee during the support phase of running, when the single-body optimization method previously illustrated is used. Inaccuracies in the order of 10 and 10 mm for rotations and linear displacements, respectively, may be observed. As far as the precision issue is concerned, in Table 1 we report data that provide an idea of the impact that the variability in the identification of the orientation of the mediolateral anatomical axis of the femur may have on knee kinematics during level walking as described using for different conventions.

Future Directions Although the potential of quantitative movement analysis in research, professional decision-making, and intervention practice is fully recognized, its application to large portions of the locomotor apparatus that move in sizeable volumes is currently

40

V. Camomilla et al.

Table 1 Maximum variation of the knee joint angles (degrees) with respect to the reference values as a result of the variation of the orientation of the mediolateral anatomical axis of the femur over the range 15 (Fioretti et al. 1997). Method used for the joint kinematics description: (1) Cardan angles, (2) projection of the joint orientation vector onto the femoral anatomical axes, (3) projection of the joint orientation vector onto the knee joint axes, and (4) geometric approach (see previous section for relevant details) Method 1 2 3 4

Flexion-extension 1.0 3.9 2.5 1.6

Abduction-adduction 13.8 17.5 12.4 14.3

Internal-external rotation 9.0 0.0 6.5 21.0

limited by several problems. The experimental and analytical protocols, which have been described in this chapter, provide results regarding skeletal motion with precision and accuracy that are insufficient to answer many of the questions posed by scientists and professionals. At the moment, however, using optical or magnetoinertial technology, the reconstruction of skeletal movement may be carried out with errors that may be in the order of 10 mm and 10 or more, for position and orientation, respectively. This collides with the fact that some relative movements between adjacent joints, among those of great interest in several applications, are in the order of a few millimeters and degrees. Thus, future research must aim at this resolution by exploiting emerging technology (including other domains, e.g., medical imaging, robotics, animation) and innovative experimental and analytical methods. In particular it should focus on the following issues: (a) Development of algorithms able to process the data provided by the motion capture systems and make data describing the movement of the skeleton with the mentioned accuracy available. A specific challenge is the compensation for the artifact movement between the superficially tracked motion and the underlying structures. Two approaches are being attempted. One uses “intelligent markers” which, through an ultrasound technology, gathers trial-specific information about this artifact movement (Masum et al. 2016). Another approach uses bone pose estimators, either single-body or multi-body kinematics optimization, that embed a mathematical model of the artifact (Bonnet et al. 2017; Camomilla et al. 2015). (b) Devising non- or minimally invasive experimental techniques, possibly usable in the movement analysis laboratory, and statistical shape analysis able to provide accurate and detailed subject-specific bone and soft tissue morphology to be registered with motion data. (c) The definition and construction of anatomical frames that allow for an adequate repeatability of the results of the analysis and their standardization for the sake of data portability. (d) Implementation of advanced probabilistic approaches that accommodate for intraindividual variability and input data uncertainty.

Three-Dimensional Reconstruction of the Human Skeleton in Motion

41

In addition, other open issues are the following. Computational models of the neuromusculoskeletal system are not as integrative across space scales as demanded by a holistic approach and encounter difficulties in incorporating the characteristics of a specific subject. Finally, there is no consensus on the optimal metrics to use when assessing motor function.

References Andersen MS, Benoit DL, Damsgaard M, Ramsey DK, Rasmussen J (2010) Do kinematic models reduce the effects of soft tissue artefacts in skin marker-based motion analysis? An in vivo study of knee kinematics. J Biomech 43:268–273 Andersen MS, Damsgaard M, Rasmussen J, Ramsey DK, Benoit DL (2012) A linear soft tissue artefact model for human movement analysis: proof of concept using in vivo data. Gait Posture 35:606–611 Banks SA, Hodge WA (1996) Accurate measurement of three-dimensional knee replacement kinematics using single-plane fluoroscopy. IEEE Trans Biomed Eng 43:638–649 Barré A, Thiran JP, Jolles BM, Theumann N, Aminian K (2013) Soft tissue artifact assessment during treadmill walking in subjects with total knee arthroplasty. IEEE Trans Biomed Eng 60:3131–3140 Bell AL, Pedersen DR, Brand RA (1990) A comparison of the accuracy of several hip joint center location prediction methods. J Biomech 23:617–621 Benedetti MG, Merlo A, Leardini A (2013) Inter-laboratory consistency of gait analysis measurements. Gait Posture 38:934–939 Benoit DL, Damsgaard M, Andersen MS (2015) Surface marker cluster translation, rotation, scaling and deformation: their contribution to soft tissue artefact and impact on knee joint kinematics. J Biomech 48:2124–2129 van den Bogert AJ, Smith GD, Nigg BM (1994) In vivo determination of the anatomical axes of the ankle joint complex: an optimization approach. J Biomech 27:1477–1488 Bonci T, Camomilla V, Dumas R, Chèze L, Cappozzo A (2015) Rigid and non-rigid geometrical transformations of a marker-cluster and their impact on bone-pose estimation. J Biomech 48:4166–4172 Bonnet V, Richard V, Camomilla V, Venture G, Cappozzo A, Dumas R (2017) Joint kinematics estimation using a multi-body kinematics optimisation and an extended Kalman filter, and embedding a soft tissue artefact model. J Biomech (in press) Bouvier B, Duprey S, Claudon L, Dumas R, Savescu A (2015) Upper limb kinematics using inertial and magnetic sensors: comparison of sensor-to-segment calibrations. Sensors 15:18813–18833 Camomilla V, Cereatti A, Vannozzi G, Cappozzo A (2006) An optimized protocol for hip joint centre determination using the functional method. J Biomech 39:1096–1106 Camomilla V, Bonci T, Dumas R, Chèze L, Cappozzo A (2015) A model of the soft tissue artefact rigid component. J Biomech 48:1752–1759 Campbell AC, Lloyd DG, Alderson JA, Elliott BC (2009) MRI development and validation of two new predictive methods of glenohumeral joint center location identification and comparison with established techniques. J Biomech 42:1527–1532 Cappozzo A (1984) Gait analysis methodology. Hum Mov Sci 3:27–50 Cappozzo A (1991) Three-dimensional analysis of human locomotor acts: experimental methods and associated. Hum Mov Sci 10:589–602 Cappozzo A, Catani F, Della Croce U, Leardini A (1995) Position and orientation in space of bones during movement, anatomical frame definition and determination. Clin Biomech 10:171–178 Cappozzo A, Cappello A, Della Croce U, Pensalfini F (1997a) Surface-marker cluster design criteria for 3-D bone movement reconstruction. IEEE Trans Biomed Eng 44:1165–1174

42

V. Camomilla et al.

Cappozzo A, Della Croce U, Lucchetti L (1997b) Gait data, terminology and definition. In: Allard P, Cappozzo A, Lumberg A, Vaughan K (eds) Three-dimensional analysis of human locomotion. Wiley, New York, pp 129–132 Cereatti A, Margheritini F, Donati M, Cappozzo A (2010) Is the human acetabulofemoral joint spherical? J Bone Joint Surg (Br) 92:311–314 Cereatti A, Bonci T, Akbarshahi M, Aminian K, Barré A, Begon M, Benoit DL, Charbonnier C, Dal Maso F, Fantozzi S, Lin CC, Lu TW, Pandy MG, Stagni R, van den Bogert AJ, Camomilla V (2017) Standardization proposal of soft tissue artefact description for data sharing in human motion measurements. J Biomech. https://doi.org/10.1016/j.jbiomech.2017.02.004 Chaibi Y, Cresson T, Aubert B, Hausselle J, Neyret P, Hauger O, de Guise JA, Skalli W (2012) Fast 3D reconstruction of the lower limb using a parametric model and statistical inferences and clinical measurements calculation from biplanar X-rays. Comput Methods Biomech Biomed Eng 15:457–466 Challis JH, Pain MTG (2008) Soft tissue motion influences skeletal loads during impacts. Exerc Sport Sci Rev 36:71–75 Charlton IW, Tate P, Smyth P, Roren L (2004) Repeatability of an optimised lower body model. Gait Posture 20:213–221 Chiari L, Cappozzo A, Della Croce U, Leardini A (2005) Human movement analysis using stereophotogrammetry. Part 2: experimental errors. Gait Posture 21:197–211 Clark T, Hawkins D (2010) Are fixed limb inertial models valid for dynamic simulations of human movement? J Biomech 43:2695–2701 Clément J, Dumas R, Hagemesiter N, de Guise JA (2015) Soft tissue artifact compensation in knee kinematics by multi-body optimization: performance of subject-specific knee joint models. J Biomech 48:3796–3802 Clément J, Dumas R, Hagemeister N, de Guise JA (2017) Can generic knee joint models improve the measurement of osteoarthritic knee kinematics during squatting activity? Comput Methods Biomech Biomed Eng 20:94–103 Colle F, Lopomo N, Visani A, Zaffagnini S, Marcacci M (2016) Comparison of three formal methods used to estimate the functional axis of rotation: an extensive in-vivo analysis performed on the knee joint. Comput Methods Biomech Biomed Eng 19:484–492 Crabolu M, Pani D, Raffo L, Cereatti A (2016) Estimation of the center of rotation using wearable magneto-inertial sensors. J Biomech 16:3928–3933 Cutti AG, Giovanardi A, Rocchi L, Davalli A, Sacchetti R (2008) Ambulatory measurement of shoulder and elbow kinematics through inertial and magnetic sensors. Med Biol Eng Comput 46:169–178 Cutti AG, Ferrari A, Garofalo P, Raggi M, Cappello A, Ferrari A (2010) “Outwalk”: a protocol for clinical gait analysis based on inertial and magnetic sensors. Med Biol Eng Comput 48:17–25 Davis RB, Ounpuu S, Tyburskky D, Gage R (1991) A gait analysis data collection and reduction technique. Hum Mov Sci 10:575–587 De Rosario H, Page A, Besa A, Mata V, Conejero E (2012) Kinematic description of soft tissue artifacts: quantifying rigid versus deformation components and their relation with bone motion. Med Biol Eng Comput 50:1173–1181 De Rosario H, Page A, Besa A (2017) Analytical study of the effects of soft tissue artefacts on functional techniques to define axes of rotation. J Biomech. https://doi.org/10.1016/j. jbiomech.2017.01.046 Della Croce U, Cappozzo A, Kerrigan DC (1999) Pelvis and lower limb anatomical landmark calibration precision and its propagation to bone geometry and joint kinematics. Med Biol Eng Comput 37:155–161 Della Croce U, Leardini A, Chiari L, Cappozzo A (2005) Human movement analysis using stereophotogrammetry. Part 4: assessment of anatomical landmark mislocation and its effects on joint kinematics. Gait Posture 21:226–237

Three-Dimensional Reconstruction of the Human Skeleton in Motion

43

Donati M, Camomilla V, Vannozzi G, Cappozzo A (2007) Enhanced anatomical calibration in human movement analysis. Gait Posture 26:179–185 Donati M, Camomilla V, Vannozzi G, Cappozzo A (2008) Anatomical frame identification and reconstruction for repeatable lower limb joint kinematics estimates. J Biomech 41:2219–2226 Dumas R, Camomilla V, Bonci T, Chèze L, Cappozzo A (2014) Generalized mathematical representation of the soft tissue artefact. J Biomech 47:476–481 Dumas R, Camomilla V, Bonci T, Chèze L, Cappozzo A (2015) What portion of the soft tissue artefact requires compensation when estimating joint kinematics? J Biomech Eng 137:064502. https://doi.org/10.1115/1.4030363 Duprey S, Chèze L, Dumas R (2010) Influence of joint constraints on lower limb kinematics estimation from skin markers using global optimization. J Biomech 43:2858–2862 Ehrig RM, Taylor WR, Duda GN, Heller MO (2007) A survey of formal methods for determining functional joint axes. J Biomech 40:2150–2157 Favre J, Aissaoui R, Jolles BM, de Guise JA, Aminian K (2009) Functional calibration procedure for 3D knee joint angle description using inertial sensors. J Biomech 42:2330–2335 Fioretti S, Cappozzo A, Lucchetti L (1997) Joint kinematics. In: Allard P, Cappozzo A, Lumberg A, Vaughan K (eds) Three-dimensional analysis of human locomotion. Wiley, New York, pp 173–189 Fraysse F, Thewlis D (2014) Comparison of anatomical, functional and regression methods for estimating the rotation axes of the forearm. J Biomech 47:3488–3493 Frigo C, Rabuffetti M, Kerrigan DC, Deming LC, Pedotti A (1998) Functionally oriented and clinically feasible quantitative gait analysis method. Med Biol Eng Comput 36:179–185 Gamage SSHU, Lasenby J (2002) New least squares solutions for estimating the average Centre of rotation and the axis of rotation. J Biomech 35:87–93 Garling EH, Kapteina BL, Mertens B, Barendregt W, Veeger HEJ, Nelissen RGHH, Valstar ER (2007) Soft-tissue artefact assessment during step-up using fluoroscopy and skin-mounted markers. J Biomech 40:S18–S24 Gasparutto X, Sancisi N, Jacquelin E, Parenti-Castelli V, Dumas R (2015) Validation of a multibody optimization with knee kinematic models including ligament constraints. J Biomech 48:1141–1146 Grimpampi E, Camomilla V, Cereatti A, De Leva P, Cappozzo A (2014) Metrics for describing softtissue artefact and its effect on pose, size, and shape of marker clusters. IEEE Trans Biomed Eng 61:362–367 Grood ES, Suntay WJ (1983) A joint coordinate system for the clinical description of threedimensional motions: application to the knee. J Biomech Eng 105:136–144 Gruber K, Ruder H, Denoth J, Schneider K (1998) A comparative study of impact dynamics: wobbling mass model versus rigid body models. J Biomech 31:439–444 Guan S, Gray HA, Keynejad F, Pandy MG (2016) Mobile biplane X-ray imaging system for measuring 3D dynamic joint motion during overground gait. IEEE Trans Med Imaging 35:326–336 Halvorsen K (2003) Bias compensated least square estimate of the center of rotation. J Biomech 36:999–1008 Halvorsen K, Lesser M, Lundberg A (1999) A new method for estimating the axis of rotation and the center of rotation. J Biomech 32:1221–1227 Hara R, McGinley J, Briggs C, Baker R, Sangeux M (2016) Predicting the location of the hip joint centres, impact of age group and sex. Sci Rep 6:37707 Harrington ME, Zavatsky AB, Lawson SE, Yuan Z, Theologis TN (2007) Prediction of the hip joint centre in adults, children, and patients with cerebral palsy based on magnetic resonance imaging. J Biomech 40:595–602 Kadaba MP, Ramakrishnan HK, Wootten ME (1990) Measurement of lower extremity kinematics during level walking. J Orthop Res 8:383–392 Kainz H, Carty CP, Modenese L, Boyd RN, Lloyd DG (2015) Estimation of the hip joint centre in human motion analysis: a systematic review. Clin Biomech 30:319–329

44

V. Camomilla et al.

Lamberto G, Martelli S, Cappozzo A, Mazzà C (2016) To what extent is joint and muscle mechanics predicted by musculoskeletal models sensitive to soft tissue artefacts? J Biomech. https://doi. org/10.1016/j.jbiomech.2016.07.042i Leardini A, Cappozzo A, Catani F, Toksvig-Larsen S, Petitto A, Sforza V, Cassanelli G, Giannini S (1999) Validation of a functional method for the estimation of hip joint centre location. J Biomech 32:99–103 Leardini A, Chiari L, Cappozzo A, Della Croce U (2005) Human movement analysis using stereophotogrammetry. Part 3: soft tissue artifact assessment and compensation. Gait Posture 21:212–225 Leardini A, Sawacha Z, Paolini G, Ingrosso S, Nativo R, Benedetti MG (2007) A new anatomically based protocol for gait analysis in children. Gait Posture 26:560–571 Lempereur M, Leboeuf F, Brochard S, Rousset J, Burdin V, Rémy-Néris O (2010) In vivo estimation of the glenohumeral joint center by functional methods: accuracy and repeatability assessment. J Biomech 43:370–374 Li K, Zheng L, Tashman S, Zhang X (2012) The inaccuracy of surface-measured model-derived tibiofemoral kinematics. J Biomech 45:2719–2723 Liu W, Nigg BM (2000) A mechanical model to determine the influence of masses and mass distribution on the impact force during running. J Biomech 33:219–224 Lu TW, O’Connor JJ (1999) Bone position estimation from skin marker coordinates using global optimization with joint constraints. J Biomech 32:129–134 Luinge HJ, Veltink PH, Baten CTM (2007) Ambulatory measurement of arm orientation. J Biomech 40:78–85 Masum MA, Pickering MR, Lambert AJ, Scarvell JM, Smith PN (2016) Multi-slice ultrasound image calibration of an intelligent skin-marker for soft tissue artefact compensation. J Biomech. https://doi.org/10.1016/j.jbiomech.2016.12.030 McGinnis RS, Perkins NC (2013) Inertial sensor based method for identifying spherical joint center of rotation. J Biomech 46:2546–2549 Melhem E, Assi A, El Rachkidi R, Ghanem I (2016) EOS ® biplanar X-ray imaging: concept, developments, benefits, and limitations. J Child Orthop 10:1–14 Pain MTG, Challis JH (2001) High resolution determination of body segment inertial parameters and their variation due to soft tissue motion. J Appl Biomech 17:326–334 Paul JP (1992) Terminology and units. Deliverable n. 4, C.E.C. Program AIM, Project A-2002, CAMARC-II Piazza SJ, Erdemir A, Okita N, Cavanagh PR (2004) Assessment of the functional method of hip joint center location to reduced range of hip motion. J Biomech 37:349–356 Picerno P, Cereatti A, Cappozzo A (2008) Joint kinematics estimate using wearable inertial and magnetic sensing modules. Gait Posture 28:588–595 Pierrynowski MR, Ball KA (2009) Oppugning the assumptions of spatial averaging of segment and joint orientations. J Biomech 42:375–378 Quijano S, Serrurier A, Aubert B, Laporte S, Thoreux P, Skalli W (2013) Three-dimensional reconstruction of the lower limb from biplanar calibrated radiographs. Med Eng Phys 35:1703–1712 Rabuffetti M, Ferrarin M, Mazzoleni P, Benvenuti F, Pedotti A (2003) Optimised procedure for the calibration of the force platform location. Gait Posture 17:75–80 Reinbolt JA, Schutte JF, Fregly BJ, Koh B, Haftka RT, George AD, Mitchell KH (2005) Determination of patient-specific multi-joint kinematic models through two-level optimization. J Biomech 38:621–626 Richard V, Lamberto G, Lu TW, Cappozzo A, Dumas R (2016) Knees kinematics estimation using multi-body optimisation embedding a knee joint stiffness matrix: a feasibility study. PLoS ONE. https://doi.org/10.1371/journal.pone.01570 Riddick RC, Kuo AD (2016) Soft tissues store and return mechanical energy in human running. J Biomech 49:436–441

Three-Dimensional Reconstruction of the Human Skeleton in Motion

45

Scheys L, Desloovere K, Spaepen A, Suetens P, Jonkers I (2011) Calculating gait kinematics using MR-based kinematic models. Gait Posture 33:158–164 Seel T, Schauer T, Raisch J (2012) Joint axis and position estimation from inertial measurement data by exploiting kinematic constraints. In: Proceedings of IEEE international conference on control applications, Dubrovnik, Croatia, pp 45–49 Seidel GK, Marchinda DM, Dijkers M, Soutas-Little RW (1995) Hip joint center location from palpable bony landmarks – a cadaver study. J Biomech 28:995–998 Sheehan FT (2010) The instantaneous helical axis of the subtalar and talocrural joints: a non-invasive in vivo dynamic study. J Foot Ankle Res 3:13. https://doi.org/10.1186/17571146-3-13 Sholukha V, Van Sint JS, Snoeck O, Salvia P, Moiseev F, Rooze M (2009) Prediction of joint center location by customizable multiple regressions: application to clavicle, scapula and humerus. J Biomech 42:319–324 Shuster MD (1993) A survey of attitude representations. J Astronaut Sci 41:439–517 Soderkvist I, Wedin PA (1993) Determining the movements of the skeleton using well-configured markers. J Biomech 26:1473–1477 Stagni R, Leardini A, Cappozzo A, Benedetti MG, Cappello A (2000) Effects of hip joint centre mislocation on gait analysis results. J Biomech 33:1479–1487 Van Sint JS, Hilal I, Salvia P, Sholukha V, Poulet P, Kirokoya I, Rooze M (2003) Data representation for joint kinematics simulation of the lower limb within an educational context. Med Eng Phys 25:213–220 Wakeling JM, Nigg BM (2001) Soft-tissue vibrations in the quadriceps measured with skin mounted transducers. J Biomech 34:539–543 Woltring HJ (1994) 3-D attitude representation of human joints, a standardisation proposal. J Biomech 27:1399–1414 Wu G, Cavanagh PR (1995) ISB recommendations for standardization in the reporting of kinematic data. J Biomech 28:1257–1261 Wu G, Siegler S, Allard P, Kirtley C, Leardini A, Rosenbaum D, Whittle M, D’Lima DD, Cristofolini L, Witte H, Schmid O, Stokes I (2002) ISB recommendation on definitions of joint coordinate system of various joints for the reporting of human joint motion – part I, ankle, hip, and spine. J Biomech 35:543–548 Yin L, Chen K, Guo L, Cheng L, Wang F, Yang L (2015) Identifying the functional flexionextension axis of the knee: an in-vivo kinematics study. PLoS ONE 10:e0128877 Zelik KE, Kuo AD (2010) Human walking isn’t all hard work: evidence of soft tissue contributions to energy dissipation and return. J Exp Biol 213:4257–4264

Estimation of the Body Segment Inertial Parameters for the Rigid Body Biomechanical Models Used in Motion Analysis Raphaël Dumas and Janis Wojtusch

Abstract

Body segment inertial parameters (BSIPs) of the human body are key parameters in biomechanics to study the dynamics of human motion. BSIPs can be obtained in different ways including direct measurements on cadavers or photogrammetry and medical imaging on living humans, but they are more generally estimated by regression equations (based on those measurements). This chapter overviews three widely used regression equations reported by Winter (2009), de Leva (1996a), and Dumas et al. (2007a). These regression equations are presented for the head with neck, thorax, abdomen, pelvis, and right upper arm, forearm, hand, thigh, shank, and foot segments. The segment endpoints and segment reference frames defined at the time of the BSIPs assessment and regression computation are reviewed so that the reader can consider how they match with the construction of the rigid body biomechanical models they would like to use for motion analysis. The segment definitions and regression equations that remain undefined or unavailable are indicated, and some assumptions are proposed to amend them, where found applicable. The computation of the segment mass, position of center of mass, moments, and products of inertia from these regression equations are fully detailed, including the modification of the designation of the segment axes and the transformation from right to left segments.

R. Dumas (*) LBMC UMR_T9406, Univ Lyon, Université Claude Bernard Lyon 1, IFSTTAR, Lyon, France e-mail: [email protected] J. Wojtusch Department of Computer Science, Simulation, Systems Optimization and Robotics Group, TU Darmstadt, Darmstadt, Germany e-mail: [email protected] # Springer International Publishing AG, part of Springer Nature 2018 B. Müller, S.I. Wolf (eds.), Handbook of Human Motion, https://doi.org/10.1007/978-3-319-14418-4_147

47

48

R. Dumas and J. Wojtusch

Keywords

Segment mass • Center of mass • Moments of inertia • Regression equations • Segment length • Segment endpoints • Joint center • Segment reference frame

Contents Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . State of the Art . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Computation of the BSIPs from the Regression Equations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Segment Definition and Regression Equations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Future Directions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

48 49 50 52 74 75

Introduction Body segment inertial parameters (BSIPs) are required in biomechanics for the computation of intersegmental moments, angular momentum, mechanical work, and for the study of the whole body dynamic stability (see chapters ▶ “Induced Acceleration and Power Analyses of Human Motion”, ▶ “Optimal Control Modeling of Human Movement”, and ▶ “Physics-Based Models for Human Gait Analysis”). BSIPs refer to the segment mass, the position of the segment center of mass with respect to a segment reference frame, and the segment moments and products of inertia with respect to a segment point (typically the segment center of mass or a segment endpoint) and with respect to a segment reference frame (see chapter ▶ “Three-Dimensional Reconstruction of the Human Skeleton in Motion”). BSIPs of the human body can be obtained in different ways. The history and the description of the techniques used to assess BSIPs can be found in Pearsall and Reid 1994, Reid and Jensen 1990, and Drillis et al. 1964. BSIPs can be directly measured on cadavers (Dempster 1955; Clauser et al. 1969; Chandler et al. 1975) and indirectly measured on living subjects, typically through photogrammetry (Ackland et al. 1988; Jensen 1978; McConville et al. 1980; Young et al. 1983) or medical imaging (Bauer et al. 2007; Dumas et al. 2005; Durkin et al. 2002; Mungiole and Martin 1990; Pearsall et al. 1996; Zatsiorsky et al. 1990; Cheng et al. 2000). Based on these measurements, estimations of the BSIPs from regression equations are more classically used. Obviously, these regression equations are limited by the number and the nature of the subjects on which they have been established. In other words, regression equations for BSIPs are better established for older Caucasian non-pathological males than for females, children, and, as a matter of fact, for pathological subjects. Nevertheless, regression equations are widely used because of their expediency. Furthermore, when studying the dynamics of human motion with a 3D rigid body biomechanical model, another difficulty is to match the segment definition used for the BSIP assessment with the model construction. Indeed, BSIPs have been obtained

Estimation of the Body Segment Inertial Parameters for the Rigid Body. . .

49

with their own rationale for segment endpoints and segment reference frames, generally constrained by experimental and technical limitations. Conversely, rigid body biomechanical models and especially “conventional gait models” widely used in clinical motion analysis (Davis et al. 1991; Kadaba et al. 1990) (see chapter ▶ “The Conventional Gait Model - Success and Limitations”), are based on a chosen marker set and calibration protocol that aims at approximating the joint centers and axes. In this context, several adjustment procedures of existing regression equations have been proposed (de Leva 1996a; Dumas et al. 2007a; Hinrichs 1990) for a better correspondence between segment definition and model construction. Still, no specific adjustment for “conventional gait models” has been proposed, and it is commonly assumed that the segment definition used for the BSIP assessment and the model construction are consistent. Therefore, classically, the segment center of mass is expected to align with the axis linking the joint centers, and this axis plus two orthogonal axes are expected to be principal axes of inertia. This chapter overviews three widely used sets of regression equations for BSIPs, allowing for a 16-segment rigid body biomechanical model. The segment definitions (i.e., segment endpoints and segment axes or planes) used for the BSIP assessment, regression computation, and adjustment are reviewed so that the reader can consider how they match with the construction of the “conventional gait models” or any rigid body biomechanical models they would like to use. The segment definitions and regression equations that remain undefined or unavailable are specified, and some assumptions are proposed to amend them, where found applicable. Note that the planes of segmentation used for the BSIP assessment can be also an issue (leading to some differences between the regression equations, especially for the trunk segments), but, conversely to the segment endpoints, axes, and planes, this does not directly interfere with the construction of the biomechanical models used in motion analysis and this is not reviewed in this chapter.

State of the Art Three widely used regression equations for BSIP are the regression equations of Winter (2009) derived from the data of Dempster (1955), the regression equations of de Leva (1996a) adjusted from the data of Zatsiorsky et al. (1990), and the regression equations of Dumas et al. (2007a, b) adjusted from the data of McConville et al. (1980) and Young et al. (1983). Dempster (1955) directly measured the BSIPs of eight male cadavers (mean age 68.5 years old, mean weight 61.1 kg, mean stature 1.69 m) using equilibrium and pendulum methods. Zatsiorsky et al. (1990) indirectly measured the BSIPs by frontal gamma-ray scanner on 100 males (mean age 23.8 years old, mean weight 73.0 kg, mean stature 1.74 m) and 15 females (mean age 19.0 years old, mean weight 61.9 kg, mean stature 1.73 m). They obtained the surface density of the body from subjects lying supine. The foot segment was scanned in a lateral view

50

R. Dumas and J. Wojtusch

separately from the rest of the body. The whole human body was modeled as rectangular cuboids of 2 cm width and length and of different heights estimated from anthropometric measurements. The masses and the distances from the geometrical centers of the cuboids to reference anatomical landmarks were known from the surface density. The centers of the mass of the cuboids were assumed at their geometrical centers and the principal axes of inertia of the cuboids were assumed aligned with the axes of symmetry of the cuboids. Then, the BSIPs of the segments were computed by summing up the BSIPs of the cuboids (i.e., using weighted barycenter and parallel axis theorem). McConville et al. (1980) and Young et al. (1983) indirectly measured the BSIPs by photogrammetry on 31 males (mean age 27.5 years old, mean weight 77.3 kg, mean stature 1.77 m) and 46 females (mean age 31.2 years old, mean weight 63.9 kg, mean stature 1.61 m), respectively. The BSIPs are defined relative to skin anatomical landmarks assuming a homogenous density of 1 g/cm3. Based on these datasets and according to the aforementioned adjustments, linear regression equations have been proposed (de Leva 1996a; Dumas et al. 2007a; Winter 2009). The segment mass is computed as a percentage of the body mass. The position of the center of mass is computed as a percentage of the segment length, defined as the distance between the segment endpoints. The radii of gyration (i.e., the square roots of the moments of inertia divided by the segment mass) are computed as percentages of the segment length. In these regression equations, the number of segments is 16: the head with neck, thorax, abdomen, pelvis, right and left upper arms, forearms and hands, and thighs, shanks, and feet. In the literature, other segmentations and other regression equations (i.e., nonlinear (Zatsiorsky et al. 1990), involving anthropometric measurements such as segment circumferences (Yeadon and Morlock 1989; Zatsiorsky et al. 1990)) exist but appear hardly used, probably because they do not fit well with the motion analysis protocols (i.e., they involve not only skin markers but also calipers and tape measures).

Computation of the BSIPs from the Regression Equations As previously mentioned, the segment mass, ms, of segment s = 1,. . .,16 is estimated as: ms ¼ ps M

(1)

where ps is the percentage of the body mass M. The position of the center of mass with respect to the segment reference frame is estimated as: 0

1 cXs A rs ¼ Ls @ cY s Z cs

(2)

Estimation of the Body Segment Inertial Parameters for the Rigid Body. . .

51

Z where cXs , cY s , cs are coordinates of the center of mass expressed as percentages of the segment length Ls. The inertia matrix (moments and products of inertia in and out the diagonal, respectively) with respect to the center of mass and the segment reference frame is estimated as:

2 6 Is ¼ ms ðLs Þ2 4

r XX s

2

 XY 2 r  sYY 2 rs

sym:

 XZ 2 3 r  sYZ 2 7 5 r  sZZ 2 rs

(3)

YY ZZ where r XX s , r s , r s are radii of gyration (i.e., the square roots of the moments of inertia divided by the segment mass) expressed as percentages of the segment length Ls. The products of inertia are expressed in the same way (i is indicated in case of negative products of inertia). The designation of the segment axes can be different from one rigid body biomechanical model to another. The regression equations in the following tables will be labeled with respect to the anterior-posterior, superior-inferior, and mediallateral axes. These labels are intended for a subject in anatomic posture (standing upright with arms at the sides, palms facing forward, and feet parallel). Note that in the “conventional gait models,” I axis is anterior, J axis is lateral (to the left), and K axis is superior (Davis III et al. 1991; Kadaba et al. 1990). However, according to the standardization of the International Society of Biomechanics (ISB), X axis is anterior, Y axis is superior, and Z axis is lateral (to the right) (Wu et al. 2002, 2005). The sign convention used in the following tables complies with this standardization of the ISB (X, Y, and Z axes). Nevertheless, the position of the center of mass and the inertia matrix with respect to the I, J, and K axes can be easily computed using a permutation matrix P:

# 0 X1 cs 1 0 0 A rs ¼ 0 0 1 L s @ cY s 0 1 0 Z cs |fflfflfflfflfflfflfflfflfflffl{zfflfflfflfflfflfflfflfflfflffl} "

(4)

P

and 2 6 Is ¼ ms ðLs Þ2 ½P4

r XX s

2

sym:

 XY 2 r  sYY 2 rs

 XZ 2 3 r  sYZ 2 7 T 5 ½ P : r  sZZ 2 rs

(5)

In the same way, the regression equations in the following tables are for the right upper and lower limb segments. The position of the center of mass and the inertia matrix for the left segments can be computed (in the X, Y, and Z axes) using a symmetry matrix S (in place of P) in Eqs. 4 and 5:

52

R. Dumas and J. Wojtusch

2

1 S ¼ 40 0

3 0 0 1 0 5: 0 1

(6)

The inertia matrix (moments and products of inertia) in Eq. 3 is given with respect to the center of mass and the segment reference frame. The inertia matrix with respect to the origin of the segment reference frame can be computed with the parallel axis theorem: Is  ¼ Is þ m s

   ðrs ÞT rs E33  rs ðrs ÞT

(7)

where E3  3 is the identity matrix.

Segment Definition and Regression Equations For the regression equations of Winter (2009), the BSIPs for the segments of the trunk are revised based on the original data of Dempster (1955) and the additional estimations of Plagenhoef (1971): the center of mass of the abdomen and pelvis segments was estimated by equilibrium methods applied to a cardboard-lead cutout modeling the anthropometry and mass distribution of the trunk as reported by Dempster (1955). Note that, in the original data of Dempster (1955), only an abdominopelvic segment is reported. Its mass is 26.4% of body mass and its segment density is 1.01 g/cm3. The position of its center of mass about the superior-inferior axis is 59.9% of segment length (i.e., thoracic joint center to midpoint between hip joint centers). The abdomen and pelvis segments are presented in the following tables for consistency among the different regression equations. Nevertheless, this abdominopelvic segment may be preferred if this segmentation matches with the rigid body biomechanical model used for motion analysis. Moreover, the abdomen and pelvis segments were also available in the regression equations reported by Winter (2009), but the position of the center of mass of these segments, as well as the thorax segment, was given in percentage of the longitudinal distance between midpoints between glenohumeral and hip joint centers (i.e., 30.4% of total height (Plagenhoef 1971)). As for the regression equations based on the data of McConville et al. (1980) and Young et al. (1983), the regression equations provided in the following tables include updates for the thorax and abdomen segments (Dumas et al. 2015) and a revision made for this chapter (i.e., some typos and inconsistency in the regression equations for the joint centers previously used in the adjustment procedure were corrected). The segment endpoints and segment axes are presented in Fig. 1 with the related skin landmarks and joint centers estimated by different regression equations. The regression equations for the BSIPs of the different segments are given in (Tables 1–10).

3.2%

0.6%

Gleno-humeral joint centre

Hip joint centre

Tip of 2nd toe

1rst metatarsal head

Heel

Medial maleollus Shyrion

100%

Knee joint centre Tibiale

7.4% Medial femoral epicondyle

0.7%

3rd metacarpale

5th metacarpal heads

Ulnar styloid

Wrist joint centre

100%

4.3% Medial humeral epicondyle Elbow joint centre

10.4%

12th thoracic vertebra

8th thoracic vertebra

Acromion

7th cervical vertebra

Head vertex

Hip joint centre

Right anterior-superior iliac spine

3.2% 37.0% 36.1%

Midpoint between posterior-superior iliac spines Lumbar joint centre

Fig. 1 Skin landmarks, joint centers estimated by regression equations, and body segments

5th metatarsal head

Ankle joint centre

Lateral maleollus

Fibula head

Lateral femoral epicondyle

100%

Greater trochanter

2nd metacarpal heads

Radius styloid

Lateral humeral epicondyle Radiale

100%

Acromion

8°

94°

100%

33.5%

52%

33% 11°

55%

9.5%

100%

Left anterior-superior iliac spine

Thoracic joint centre

Xiphoid

Gleno-humeral joint centre

Suprasternal notch

Cervical joint centre

Sellion

Estimation of the Body Segment Inertial Parameters for the Rigid Body. . . 53

(Dempster 1955; Winter 2009) Head vertex skin landmarka Cervical joint center obtained by dissection (center of C7-T1 disc)

Superior-inferior axis from cervical joint center to head vertex skin landmark Sagittal plane set as plane of oscillation of the pendulum method Origin at head vertex skin landmarka

The head with neck Segment endpoints

Segment reference frame

Table 1 Head with neck

Superior-inferior axis: N/A Frontal plane parallel to the gamma-ray scanner acquisition plane Origin at head vertex skin landmark projected on superior-inferior axis

(de Leva 1996a; Zatsiorsky et al. 1990) Skin landmark on head vertex projected on superiorinferior axis Skin landmark on 7th cervical vertebra projected on superior-inferior axis

(Dumas et al. 2007b; McConville et al. 1980; Young et al. 1983) Skin landmark on the head vertex Cervical joint center estimated by the regression equations (Dumas et al. 2007a) Female: cervical joint center on a direction forming an angle of 14 in the sagittal plane with the vector from 7th cervical vertebra to suprasternal notch skin landmarks and at 53% of the thorax width (i.e., distance between 7th cervical vertebra and suprasternal notch skin landmarks) from 7th cervical vertebra skin landmark Male: angle of 8 and percentage of 55% Superior-inferior axis from cervical joint center to head vertex skin landmark Sagittal plane containing cervical joint center and skin landmarks on the head vertex and sellion Origin at cervical joint center

54 R. Dumas and J. Wojtusch

a

Female: 0.1 and male: 0.1 Female: 30 and male: 28 Female: 24 and male: 21

Female: 26.1 and male: 26.1 Female: 27.1 and male: 30.3 Assumed negligible Assumed negligible Assumed negligible

Assumed negligible Assumed equal to mediallateral axis N/A Female: N/A and male: 28.1a Assumed negligible Assumed negligible Assumed negligible

Value of Winter (2009) replaced by original value of Dempster (1955), radius of gyration worked out with a rule of 3

Product of inertia

Moment of inertia (radius of gyration in % of the segment length)

Female: 0 and male: 3

Female: 31 and male: 30 Female: 5(i) and male: 7(i) Female: 1 and male: 2(i)

Female: 55.9 and male: 53.4

Female: 48.41 and male: 50.02 Assumed negligible Female: 29.5 and male: 31.5

Female: N/A and male: 43.3a

Female: 0.8 and male: 2.0

Assumed to be 1 Female: 6.7 and male: 6.7

Female: 243 and male: 278

N/A

Female: 243.7 and male: 242.9 N/A Female: 6.68 and male: 6.94

Assumed negligible

1.11 Female: N/A and male: 8.10

Segment density (g/cm3) Segment mass (% of total body mass) Position of center of mass (% of the segment length)

Anterior-posterior axis (X) Superior-inferior axis (Y) Medial-lateral axis (Z) Anterior-posterior axis (X) Superior-inferior axis (Y) Medial-lateral axis (Z) Sagittal plane (X,Y) Transverse plane (X, Z) Frontal plane (Y,Z)

N/A

Segment length (mm)

Estimation of the Body Segment Inertial Parameters for the Rigid Body. . . 55

Segment reference frame

The thorax Segment endpoints

Table 2 Thorax

Superior-inferior axis from thoracic to cervical joint center Sagittal plane set as plane of oscillation of the pendulum method Origin at cervical joint center

(Dempster 1955; Winter 2009) Cervical joint center (see head with neck segment) Thoracic joint center obtained by dissection (center of T12-L1 disc)

Superior-inferior axis: N/A Frontal plane parallel to the gamma-ray scanner acquisition plane Origin at 7th cervical vertebra skin landmark projected on superior-inferior axis

(de Leva 1996a; Zatsiorsky et al. 1990) Skin landmark on the seventh cervical vertebra projected on superior-inferior axis Skin landmark on xiphoid projected on superior-inferior axis

(Dumas et al. 2015; McConville et al. 1980; Young et al. 1983) Cervical joint center (see head with neck segment) Thoracic joint center estimated by the regression equations (Dumas et al. 2015) Female: thoracic joint center on a direction forming an angle of 92 in the sagittal plane with the vector from 12th to 8th thoracic vertebra skin landmarks and at 50% of thorax width (see head with neck segment) from 12th thoracic vertebra skin landmark Male : angle of 94 and percentage of 52% Superior-inferior axis from thoracic to cervical joint center Sagittal plane containing skin landmarks on 7th cervical and 8th thoracic vertebra and suprasternal notch Origin at cervical joint center

56 R. Dumas and J. Wojtusch

a

Anterior-posterior axis (X) Superior-inferior axis (Y) Medial-lateral axis (Z) Anterior-posterior axis (X) Superior-inferior axis (Y) Medial-lateral axis (Z) Sagittal plane (X, Y) Transverse plane (X,Z) Frontal plane (Y, Z) Assumed negligible Assumed negligible

Assumed negligible Assumed negligible

Female: 46.6 and male: 50.5

N/A N/A

Female: 44.9 and male: 46.5

N/A

N/A

Female: 31.4 and male: 32.0

N/A

Female: 54.2 and male: 55.5

Female: 50.50 and male: 50.66 Assumed negligible

Female: N/A and male: 62.7a Assumed negligible

Female: 1 and male: 3

Female: 3(i) and male: 1

Female: 12(i) and male: 11(i)

Female: 34 and male: 36

Female: 32 and male: 33

Female: 38 and male: 42

Female: 0.1 and male: 0.4

Female: 1.5 and male: 0.0

Female: 322 and male: 334 Assumed to be 1 Female: 26.3 and male: 30.4

N/A

Female: 228.0 and male: 242.1 N/A Female: 15.45 and male: 15.96

Assumed negligible

N/A 0.92 Female: N/A and male: 21.60

Value of Winter (2009) replaced by original value of Dempster (1955)

Product of inertia

Moment of inertia (radius of gyration in % of the segment length)

Segment length (mm) Segment density (g/cm3) Segment mass (% of total body mass) Position of center of mass (% of the segment length)

Estimation of the Body Segment Inertial Parameters for the Rigid Body. . . 57

Segment reference frame

The abdomen Segment endpoints

Table 3 Abdomen

Superior-inferior axis: N/A Sagittal plane of the cardboard-lead cutouts modeling the anthropometry and mass distribution of the trunk Origin at thoracic joint centera

(Dempster 1955; Plagenhoef 1971; Winter 2009) Thoracic joint center (see thorax segment) Inferior endpoint: N/A

Superior-inferior axis: N/A Frontal plane parallel to the gamma-ray scanner acquisition plane Origin at xiphoid skin landmark projected on superior-inferior axis

(de Leva 1996a; Zatsiorsky et al. 1990) Skin landmark on xiphoid projected on superior-inferior axis Skin landmark on omphalion projected on superior-inferior axis

(Dumas et al. 2015; McConville et al. 1980; Young et al. 1983) Thoracic joint center (see thorax segment) Lumbar joint center estimated by the regression equations (Dumas et al. 2007a) Female: in the pelvis reference frame (see pelvis segment) with origin translated at midpoint between anterior-superior iliac spine skin landmarks, lumbar joint center at 34.0%, 4.9% and 0% of pelvis width (see pelvis segment) about the anteriorposterior, superior-inferior, and mediallateral axes, respectively Male: percentages of 33.5%%, 3.2%, and 0.0%, respectively Superior-inferior axis from thoracic to lumbar joint center Sagittal plane: N/A (no axial rotation at lumbar joint center assumed) Origin at thoracic joint center

58 R. Dumas and J. Wojtusch

N/A N/A

Superior-inferior axis (Y)

Medial-lateral axis (Z) N/A Assumed negligible Assumed negligible

Assumed negligible N/A

Medial-lateral axis (Z) Anterior-posterior axis (X)

Sagittal plane (X,Y) Transverse plane (X,Z) Frontal plane (Y,Z)

N/A Female: N/A and male: 34.6a

Anterior-posterior axis (X) Superior-inferior axis (Y)

N/A Female: 45.12 and male: 45.02 Assumed negligible Female: 35.4 and male: 38.3 Female: 41.5 and male: 46.8 Female: 43.3 and male: 48.2 N/A Assumed negligible Assumed negligible

Female: 14.65 and male: 16.33

Female: N/A and male: 13.90

N/A

Female: 205.3 and male: 215.5 N/A

N/A

Female: 25 and male: 11 Female: 3(i) and male: 6(i) Female: 5(i) and male: 5(i)

Female: 52 and male: 40

Female: 78 and male: 66

Female: 0.3 and male: 3.3 Female: 65 and male: 54

Female: 21.9 and male: 17.6 Female: 41.0 and male: 36.1

Female: 4.1 and male: 2.9

Assumed to be 1

Female: 125 and male: 151

a Adapted from Plagenhoef (1971), position of center of mass expressed as percentage of the length of the abdominopelvic segment (thoracic joint center to the midpoint between the hip joint centers)

Product of inertia

Moment of inertia (radius of gyration in % of the segment length)

Segment length (mm) Segment density (g/cm3) Segment mass (% of total body mass) Position of center of mass (% of the segment length)

Estimation of the Body Segment Inertial Parameters for the Rigid Body. . . 59

Female: 181.5 and male: 145.7 N/A

N/A

N/A

Segment length (mm) Segment density (g/cm3)

Superior-inferior axis: N/A Frontal plane parallel to the gammaray scanner acquisition plane Origin at omphalion skin landmark projected on superior-inferior axis

(de Leva 1996a; Zatsiorsky et al. 1990) Skin landmark on omphalion projected on superior-inferior axis Midpoint between hip joint centers estimated by the regression equations (de Leva 1996b) Female: N/A Male: on superior-inferior axis, hip joint center at 0.7% of thigh length (i.e., longitudinal distance between tibial and greater trochanter skin landmarks) from the greater trochanter skin landmark (percentages on anterior-posterior and media-lateral axes: N/A)

Superior-inferior axis: N/A Sagittal plane of the cardboard-lead cutouts modeling the anthropometry and mass distribution of the trunk Origin at midpoint between hip joint centers

(Dempster 1955; Plagenhoef 1971; Winter 2009) Superior endpoint: N/A Midpoint between hip joint centers obtained by dissection (acetabulum center)

Segment reference frame

The pelvis Segment endpoints

Table 4 Pelvis

Assumed to be 1

(Dumas et al. 2007a; McConville et al. 1980; Young et al. 1983) Lumbar joint center (see abdomen segment) Midpoint between hip joint centers estimated by the regression equations (Dumas et al. 2007a) Female: in the pelvis reference frame (see below) with origin translated at midpoint between anterior-superior iliac spine skin landmarks, right/left hip joint center at 13.9%, 33.6%, and +/37.2% of pelvis width (i.e., distance between left and right anterior-superior iliac spine skin landmarks) about the anteriorposterior, superior-inferior, and medial-lateral axes, respectively Male: percentages of 9.5%, 37.0%, and +/36.1%, respectively Medial-lateral axis from left to right anterior superior iliac spine skin landmarks Transverse plane containing skin landmarks on left and right anteriorsuperior iliac spines and midpoint between posterior-superior iliac spines Origin at lumbar joint center Female: 103 and male: 93

60 R. Dumas and J. Wojtusch

Anteriorposterior axis (X) Superiorinferior axis (Y) Mediallateral axis (Z) Anteriorposterior axis (X) Superiorinferior axis (Y) Mediallateral axis (Z) Sagittal plane (X,Y) Transverse plane (X,Z) Frontal plane (Y,Z) Assumed negligible Assumed negligible

Assumed negligible

Assumed negligible

Female: 43.3 and male: 61.5

N/A

N/A

Female: 44.4 and male: 58.7

N/A

N/A

Female: 40.2 and male: 55.1

Assumed negligible

Assumed negligible

N/A

Female: 22.8 and male: 28.2

Female: 49.20 and male: 61.15

Female: N/A and male: 15.6a

Female: 2(i) and male: 8(i)

Female: 3(i) and male: 12(i)

Female: 35(i) and male: 25(i)

Female: 82 and male: 96

Female: 105 and male: 106

Female: 95 and male: 102

Female: 0.2 and male: 0.6

Female: 7.2 and male: 0.2

Female: 14.7 and male: 14.2

N/A

Female: 12.47 and male: 11.17

N/A

Female: N/A and male: 14.20

a Adapted from Plagenhoef (1971), position of center of mass expressed as percentage of the length of the abdominopelvic segment (thoracic joint center to the midpoint between the hip joint centers)

Product of inertia

Moment of inertia (radius of gyration in % of the segment length)

Segment mass (% of total body mass) Position of center of mass (% of the segment length)

Estimation of the Body Segment Inertial Parameters for the Rigid Body. . . 61

Superior-inferior axis from elbow to glenohumeral joint center Frontal plane parallel to the gammaray scanner acquisition plane Origin at glenohumeral joint center

Female: 275.1 and male: 281.7 N/A

Female: N/A and male: 286a

1.07

Segment length (mm) Segment density (g/cm3)

(de Leva 1996a; Zatsiorsky et al. 1990) Glenohumeral and elbow joint centers estimated by the regression equations (de Leva 1996b) Female: N/A Male: on superior-inferior axis, glenohumeral joint center at 10.4% of upper arm length (i.e., longitudinal distance between acromion and radial skin landmarks) from acromion skin landmark and elbow joint center at 4.3% of upper arm length from radial skin landmark (percentages on anterior-posterior and media-lateral axes: N/A)

Superior-inferior axis from elbow to glenohumeral joint center Sagittal plane set as plane of oscillation of the pendulum method Origin at glenohumeral joint center

(Dempster 1955; Plagenhoef 1971; Winter 2009) Glenohumeral joint center obtained by dissection (center of curvature of humeral head) Elbow joint center obtained by dissection (axis of humeral trochlea at narrowest cross section of ulnar articulation)

Segment reference frame

The upper arm Segment endpoints

Table 5 Upper arm

Assumed to be 1

(Dumas et al. 2007a; McConville et al. 1980; Young et al. 1983) Glenohumeral joint center estimated by the regression equations (Dumas et al. 2007a) Elbow joint center estimated at midpoint between lateral and medial humeral epicondyle skin landmarks Female : glenohumeral joint center on a direction forming an angle of 5 in the sagittal plane with the vector from 7th cervical vertebra to suprasternal notch skin landmarks and at 36% of thorax width (see head with neck and thorax segments) from the acromion skin landmark Male: angle of 11 and percentage of 33% Superior-inferior axis from elbow to glenohumeral joint center Frontal plane containing glenohumeral joint center and skin landmarks on lateral and medial humeral epicondyles Origin at glenohumeral joint center Female: 251 and male: 277

62 R. Dumas and J. Wojtusch

a

Anteriorposterior axis (X) Superiorinferior axis (Y) Mediallateral axis (Z) Anteriorposterior axis (X) Superiorinferior axis (Y) Mediallateral axis (Z) Sagittal plane (X, Y) Transverse plane (X,Z) Frontal plane (Y,Z) Assumed negligible

Assumed negligible Assumed negligible

Assumed negligible

Assumed negligible

Female: 27.8 and male: 28.5

Female: N/A and male: 32.2

Assumed negligible

Female: 14.8 and male: 15.8

N/A

Assumed negligible

Assumed negligible

Female: 26.0 and male: 26.9

Female: 57.54 and male: 57.72

Female: N/A and male: 43.6

Assumed equal to medial-lateral axis

Assumed negligible

Female: 2.55 and male: 2.71

Assumed negligible

Female: N/A and male: 2.80

Value of Plagenhoef (1971), 16.9% of total height

Product of inertia

Moment of inertia (radius of gyration in % of the segment length)

Segment mass (% of total body mass) Position of center of mass (% of the segment length)

Female: 3 and male: 13(i)

Female: 5 and male: 3

Female: 3(i) and male: 5

Female: 30 and male: 30

Female: 15 and male: 13

Female: 30 and male: 29

Female: 3.3 and male: 3.1

Female: 50.0 and male: 48.2

Female: 5.5 and male: 1.8

Female: 2.3 and male: 2.4

Estimation of the Body Segment Inertial Parameters for the Rigid Body. . . 63

Segment length (mm) Segment density (g/cm3) Segment mass (% of total body mass)

Segment reference frame

The forearm Segment endpoints

Table 6 Forearm

Female: N/A and male: 1.60

(Dempster 1955; Plagenhoef 1971; Winter 2009) Elbow joint center (see upper arm segment) Wrist joint center obtained by dissection (center of curvature of proximal end of capitate bone) Superior-inferior axis from wrist to elbow joint center Sagittal plane set as plane of oscillation of the pendulum method Origin at elbow joint center Female: N/A and male: 269a 1.13 Assumed to be 1

N/A

Female: 1.4 and male: 1.7

Female: 247 and male: 283

Female: 264.3 and male: 268.9

Female: 1.38 and male: 1.62

Superior-inferior axis from wrist to elbow joint center Frontal plane containing elbow joint center and skin landmarks on radial and ulna styloids Origin at elbow joint center

(Dumas et al. 2007a; McConville et al. 1980; Young et al. 1983) Elbow joint center (see upper arm segment) Wrist joint center estimated at midpoint between radial and ulna styloid skin landmarks

(de Leva 1996a; Zatsiorsky et al. 1990) Elbow joint center (see upper arm segment) Wrist joint center estimated by the regression equations (de Leva 1996b) Female: N/A Male: on superior-inferior axis, wrist joint center at 0.6% of forearm length (i.e., longitudinal distance between radial and radius styloid skin landmarks) from radius styloid skin landmark (percentages on anterior-posterior and media-lateral axes: N/A) Superior-inferior axis from wrist to elbow joint center Frontal plane parallel to the gamma-ray scanner acquisition plane Origin at elbow joint center

64 R. Dumas and J. Wojtusch

a

Female: 10 and male: 8 Female: 3 and male: 1(i) Female: 13(i) and male: 2

Female: N/A and Female: 26.1 and male: 27.6 male: 30.3 Assumed negligible Assumed negligible

Assumed negligible Assumed negligible

Assumed negligible Assumed negligible

Female: 25 and male: 28

Female: 14 and male: 11

N/A

Female: 9.4 and male: 12.1

Female: 27 and male: 28

Assumed equal to Female: 25.7 and male: 26.5 medial-lateral axis

Female: 41.1 and male: 41.7

Female: 1.9 and male: 1.1

Female: 45.59 and male: 45.74

Female: 2.1 and male: 1.3

Assumed negligible Assumed negligible

Female: N/A and male: 43.0

Assumed negligible Assumed negligible

Value of Plagenhoef (1971), 15.9% of total height

Anteriorposterior axis (X) Superiorinferior axis (Y) Medial-lateral axis (Z) Moment of Anteriorinertia (radius posterior axis of gyration in % (X) of the segment Superiorlength) inferior axis (Y) Medial-lateral axis (Z) Product of Sagittal plane inertia (X,Y) Transverse plane (X,Z) Frontal plane (Y,Z)

Position of center of mass (% of the segment length)

Estimation of the Body Segment Inertial Parameters for the Rigid Body. . . 65

Female: 76.8 and male: 83.9

Female: 74.74 and male: 79.00 Assumed negligible

Female: N/A and male: 50.6 Assumed negligible

Female: 4.8 and male: 7.5

Female: 7.7 and male: 8.2

Superior-inferior axis from midpoint between 2nd and 5th metacarpal head skin landmarks to wrist joint center Frontal plane containing wrist joint center and skin landmarks on 2nd and 5th metacarpal heads Origin at wrist joint center Female: 71 and male: 80 Assumed to be 1 Female: 0.5 and male: 0.6

(Dumas et al. 2007a; McConville et al. 1980; Young et al. 1983) Wrist joint center (see forearm segment) Midpoint between 2nd and 5th metacarpal head skin landmarks

N/A

Female: 78.0 and male: 86.2 N/A Female: 0.56 and male: 0.61

(de Leva 1996a; Zatsiorsky et al. 1990) Wrist joint center (see forearm segment) 3rd metacarpal skin landmark projected on superior-inferior axis Superior-inferior axis: N/A Frontal plane parallel to the gamma-ray scanner acquisition plane Origin at wrist joint center

N/A

N/A 1.17 Female: N/A and male: 0.60

Segment length (mm) Segment density (g/cm3) Segment mass (% of total body mass) Position of center of mass (% of the segment length)

Anteriorposterior axis (X) Superiorinferior axis (Y) Mediallateral axis (Z)

Superior-inferior axis: N/A Sagittal plane set as plane of oscillation of the pendulum method Origin at wrist joint center

(Dempster 1955; Winter 2009) Wrist joint center (see forearm segment) Interphalangeal knuckle of 3rd finger skin landmark

Segment reference frame

The hand Segment endpoints

Table 7 Hand

66 R. Dumas and J. Wojtusch

Product of inertia

Moment of inertia (radius of gyration in % of the segment length)

Anteriorposterior axis (X) Superiorinferior axis (Y) Mediallateral axis (Z) Sagittal plane (X,Y) Transverse plane (X,Z) Frontal plane (Y,Z) Assumed negligible Assumed negligible

N/A N/A

Female: 53.1 and male: 62.8

Female: N/A and male: 29.7

Assumed negligible

Female: 33.5 and male: 40.1

N/A

N/A

Female: 45.4 and male: 51.3

N/A

Female: 28(i) and male: 20(i)

Female: 23 and male: 15

Female: 29 and male: 22

Female: 59 and male: 56

Female: 43 and male: 38

Female: 64 and male: 61

Estimation of the Body Segment Inertial Parameters for the Rigid Body. . . 67

Segment length (mm) Segment density (g/cm3) Segment mass (% of total body mass)

Segment reference frame

The thigh Segment endpoints

Table 8 Thigh

Superior-inferior axis from knee to hip joint center Sagittal plane set as plane of oscillation of the pendulum method Origin at hip joint center Female: N/A and male: 395a 1.05 Female: N/A and male: 10.00

(Dempster 1955; Plagenhoef 1971; Winter 2009) Hip joint center obtained by dissection (center of curvature of femoral head) Knee joint center obtained by dissection (middle of a line through the center of curvature of the posterior aspect of femoral condyles)

(de Leva 1996a; Zatsiorsky et al. 1990) Hip joint center (see pelvis segment) Knee joint center estimated by the regression equations (de Leva 1996b) Female: N/A Male: on superior-inferior axis, knee joint center at 7.4% of thigh length (i.e., longitudinal distance between greater trochanter and tibial skin landmarks) from tibial skin landmark (percentages on anterior-posterior and medialateral axes: N/A) Superior-inferior axis from knee to hip joint center Frontal plane parallel to the gamma-ray scanner acquisition plane Origin at hip joint center Female: 368.5 and male: 422.2 N/A Female: 14.78 and male: 14.16

Superior-inferior axis from knee to hip joint center Frontal plane containing hip joint center and skin landmarks on lateral and medial femoral epicondyles Female: 379 and male: 432 Assumed to be 1 Female: 14.6 and male: 12.3

(Dumas et al. 2007a; McConville et al. 1980; Young et al. 1983) Hip joint center (see pelvis segment) Knee joint center estimated at midpoint between midpoint between lateral and medial femoral epicondyle skin landmarks

68 R. Dumas and J. Wojtusch

a

Anteriorposterior axis (X) Superiorinferior axis (Y) Mediallateral axis (Z) Anteriorposterior axis (X) Superiorinferior axis (Y) Mediallateral axis (Z) Sagittal plane (X,Y) Transverse plane (X,Z) Frontal plane (Y,Z) Assumed negligible Assumed negligible Assumed negligible

Assumed negligible Assumed negligible

Female: 36.9 and male: 32.9

Female: N/A and male: 32.3

Assumed negligible

Female: 16.2 and male: 14.9

N/A

Assumed negligible

Assumed negligible

Female: 36.4 and male: 32.9

Female: 36.12 and male: 40.95

Female: N/A and male: 43.3

Assumed equal to medial-lateral axis

Assumed negligible

Assumed negligible

Value of Plagenhoef (1971), 23.4% of total height

Product of inertia

Moment of inertia (radius of gyration in % of the segment length)

Position of center of mass (% of the segment length)

Female: 7(i) and male: 7(i)

Female: 2 and male: 2(i)

Female: 7(i) and male: 7

Female: 32 and male: 30

Female: 19 and male: 15

Female: 31 and male: 29

Female: 0.8 and male: 3.3

Female: 37.7 and male: 42.9

Female: 7.7 and male: 4.1

Estimation of the Body Segment Inertial Parameters for the Rigid Body. . . 69

Female: 438.6 and male: 440.3 N/A Female: 4.81 and male: 4.33

Superior-inferior axis from ankle to knee joint center Sagittal plane set as plane of oscillation of the pendulum method Origin at knee joint center

Female: N/A and male: 428a

1.09

Female: N/A and male: 4.65

Segment length (mm) Segment density (g/cm3) Segment mass (% of total body mass)

(Dempster 1955; Plagenhoef 1971; Winter 2009) Knee joint center (see thigh segment) Ankle joint center obtained by dissection (center of the area of the cut body of the talus)

Segment reference frame

The shank Segment endpoints

(de Leva 1996a; Zatsiorsky et al. 1990) Knee joint center (see thigh segment) Ankle joint center estimated by the regression equations (de Leva 1996b) Male: on superior-inferior axis, ankle joint center at 3.2% of shank length (i.e., longitudinal distance between tibial and sphyrion skin landmarks) from sphyrion skin landmark (percentages on anterior-posterior and media-lateral axes: N/A) Superior-inferior axis from ankle to knee joint center Frontal plane parallel to the gammaray scanner acquisition plane Origin at knee joint center

Table 9 Shank

Female: 4.5 and male: 4.8

Assumed to be 1

Superior-inferior axis from ankle to knee joint center Frontal plane containing knee and ankle joint centers and the fibula head skin landmark Origin at knee joint center Female: 388 and male: 433

(Dumas et al. 2007a; McConville et al. 1980; Young et al. 1983) Knee joint center (see thigh segment) Ankle joint center estimated at midpoint between lateral and medial malleolus skin landmarks

70 R. Dumas and J. Wojtusch

a

Anteriorposterior axis (X) Superiorinferior axis (Y) Mediallateral axis (Z) Anteriorposterior axis (X) Superiorinferior axis (Y) Mediallateral axis (Z) Sagittal plane (X,Y) Transverse plane (X,Z) Frontal plane (Y,Z) Assumed negligible Assumed negligible Assumed negligible

Assumed negligible

Assumed negligible

Female: 26.7 and male: 25.1

Female: N/A and male: 30.2

Assumed negligible

Female: 9.2 and male: 10.2

N/A

Assumed negligible

Assumed negligible

Female: 26.3 and male: 24.6

Female: 43.52 and male: 43.95

Female: N/A and male: 43.3

Assumed equal to medial-lateral axis

Assumed negligible

Assumed negligible

Value of Plagenhoef (1971), 25.3% of total height

Product of inertia

Moment of inertia (radius of gyration in % of the segment length)

Position of center of mass (% of the segment length)

Female: 6 and male: 4

Female: 1 and male: 2(i)

Female: 2 and male: 4(i)

Female: 28 and male: 28

Female: 10 and male: 10

Female: 28 and male: 28

Female: 3.1 and male: 0.7

Female: 40.4 and male: 41.0

Female: 4.9 and male: 4.8

Estimation of the Body Segment Inertial Parameters for the Rigid Body. . . 71

Segment density (g/cm3) Segment mass (% of total body mass) Position of center of mass (% of the segment length)

Segment length (mm)

Segment reference frame

The foot Segment endpoints

Table 10 Foot

Anteriorposterior axis (X) Superiorinferior axis (Y) Mediallateral axis (Z)

Female: 40.14 and Male: 44.15 Assumed negligible

Assumed negligible

Assumed negligible

Assumed negligible

Female: 228.3 and male: 258.1 N/A Female: 1.29 and male: 1.37

(de Leva 1996a; Zatsiorsky et al. 1990) Heel skin landmark Tip of longest toe skin landmark Anterior-posterior axis from heel to tip of longest toe skin landmarks Frontal plane parallel to the gamma-ray scanner acquisition plane Origin at heel skin landmark

Female: N/A and male: 42.9a

1.09 Female: N/A and male: 1.45

(Dempster 1955; Winter 2009) Heel skin landmarksa Tip of 2nd toe skin landmarksa Anterior-posterior axis from heel to tip of 2nd toe skin landmarks Sagittal plane set as plane of oscillation of the pendulum method Origin at heel skin landmarka N/A

Female: 5.5 and male: 3.4

Female: 30.9 and male: 19.9

Female: 38.2 and male: 50.2

Assumed to be 1 Female: 1.0 and male: 1.2

Female: 117 and male: 139

(Dumas et al. 2007a; McConville et al. 1980; Young et al. 1983) Ankle joint center (see shank segment) Midpoint between 1st and 5th metatarsal head skin landmarks Anterior-posterior axis from heel to midpoint between 1st and 5th metatarsal head skin landmarks Transverse plane containing skin landmarks on heel and 1st and 5th metatarsal heads

72 R. Dumas and J. Wojtusch

a

Anteriorposterior axis (X) Superiorinferior axis (Y) Mediallateral axis (Z) Sagittal plane (X,Y) Transverse plane (X,Z) Frontal plane (Y,Z) N/A Assumed negligible

N/A Assumed negligible

Female: 29.9 and male: 25.7

Female: N/A and male: 40.7a

N/A

Female: 27.9 and male: 24.5

Assumed equal to mediallateral axis

N/A

Female: 13.9 and male: 12.4

N/A

Value of Winter (2009) replaced by original value of Dempster (1955), radius of gyration worked out with a rule of 3

Product of inertia

Moment of inertia (radius of gyration in % of the segment length)

Female: 5(i) and male: 0

Female: 9 and male: 11(i)

Female: 15(i) and male: 17

Female: 50 and male: 48

Female: 50 and male: 49

Female: 24 and male: 22

Estimation of the Body Segment Inertial Parameters for the Rigid Body. . . 73

74

R. Dumas and J. Wojtusch

Future Directions As previously mentioned in the introduction, regression equations (specifically regression equations based on body mass and segment lengths) are widely used because of their expediency but remain limited and sometimes difficult to harmonize with the rigid body biomechanical models used for motion analysis. The adaptation of the previously developed adjustment procedures to elderly subjects has been recently proposed (Ho Hoang and Mombaur 2015) and may be extended to other specific populations. There is an increasing interest in using subject-specific BSIPs of the lower limb indirectly measured by medical imaging (Bauer et al. 2007; Dao et al. 2012; Ganley and Powers 2004; Sreenivasa et al. 2016; Taddei et al. 2012; Valente et al. 2014) for gait analysis, especially in case of pathologic subjects (see chapter ▶ “Cross-Platform Comparison of Imaging Technologies for Measuring Musculoskeletal Motion”). Here again, the issue of matching segment definition used for the BSIP assessment with the biomechanical model used for motion analysis exists. The segment definition and model construction are generally assumed consistent or anatomical landmarks are virtually palpated on the MRI or CT scan reconstructions (Dao et al. 2012; Sreenivasa et al. 2016; Valente et al. 2014). Occasionally, skin markers are placed on the subject before CT scan and gait analysis, both performed consecutively (Taddei et al. 2012). Nevertheless, the main purpose of these cumbersome personalization methods was generally focussed on bone and muscle geometry. The introduction of skin markers used for motion analysis within the procedure of the BSIPs assessment was more widely adopted with photogrammetry (Davidson et al. 2008; Pillet et al. 2010; Verriest 2012). This means that calibrated photographs are taken during a static posture (just before or after motion analysis) allowing to build a 3D volume model of the subject, straightforwardly registered with the rigid body biomechanical model. Moreover, without additional calibrated photographs and not performed during a static posture but a dynamic movement, the BSIPs can be estimated by identification methods (Ayusawa et al. 2014; Jovic et al. 2016; Vaughan et al. 1982). In this case, a rigid body biomechanical model and the equations of motion are directly used to compute the BSIPs that minimize the errors between model-derived and measured ground reaction forces and moments. However, as for the functional calibration of the joint centers and axes, these methods may require dedicated movements. With that respect, a method that simultaneously identifies BSIPs, joint centers, and segment lengths has been recently proposed (Bonnet et al. 2017). This method, based on extended Kalman filters, minimizes the errors between model-derived and measured ground reaction forces and moments as well as skin marker trajectories. Other advanced methods for inverse dynamics (i.e., residual elimination/reduction algorithms) typically alter the generalized accelerations of the model but may also include either some (Delp et al. 2007) or all of the BSIPs (Jackson et al. 2015) as design variables of the minimization process.

Estimation of the Body Segment Inertial Parameters for the Rigid Body. . .

75

Most of the abovementioned methods for BSIPs assessment involve numerous skin markers, especially when all the segments of the human body (e.g., the 16 segments previously mentioned) are of interest. Therefore, another direction is to reduce the number of skin markers at its minimum while estimating the center of mass of the human body, typically for the analysis of the dynamic stability. Adapted segment definition, regression equations, and marker set have been dedicated for such specific applications (Tisserand et al. 2016; Yang and Pai 2014).

References Ackland TR, Blanksby BA, Bloomfield J (1988) Inertial characteristics of adolescent male body segments. J Biomech 21(4):319–327. https://doi.org/10.1016/0021-9290(88)90261-8 Ayusawa K, Venture G, Nakamura Y (2014) Identifiability and identification of inertial parameters using the underactuated base-link dynamics for legged multibody systems. Int J Robot Res 33(3):446–468. https://doi.org/10.1177/0278364913495932 Bauer JJ, Pavol MJ, Snow CM, Hayes WC (2007) MRI-derived body segment parameters of children differ from age-based estimates derived using photogrammetry. J Biomech 40(13):2904–2910. https://doi.org/10.1016/j.jbiomech.2007.03.006 Bonnet V, Dumas R, Cappozzo A, Joukov V, Daune G, Kulić D, Fraisse P, Andary S, Venture G (2017) A constrained extended Kalman filter for the optimal estimate of kinematics and kinetics of a sagittal symmetric exercise. J Biomech. https://doi.org/10.1016/j.jbiomech.2016.12.027 Chandler RF, Clauser CE, McConville JT, Reynolds HM, Young JW (1975) Investigation of inertial properties of the human body. Aerospace Medical Research Laboratory, Wright-Patterson Air Force Base, Dayton Cheng C-K, Chen H-H, Chen C-S, Lee C-L, Chen C-Y (2000) Segment inertial properties of Chinese adults determined from magnetic resonance imaging. Clin Biomech 15(8):559–566. https://doi.org/10.1016/S0268-0033(00)00016-4 Clauser CE, McConville JT, Young JW (1969) Weight, volume, and center of mass of segments of the human body. Aerospace Medical Research Laboratory, Wright–Patterson Air Force Base, Dayton Dao TT, Marin F, Pouletaut P, Charleux F, Aufaure P, Ho Ba Tho MC (2012) Estimation of accuracy of patient-specific musculoskeletal modelling: case study on a post polio residual paralysis subject. Comput Methods Biomech Biomed Eng 15(7):745–751. https://doi.org/10.1080/ 10255842.2011.558086 Davidson PL, Wilson SJ, Wilson BD, Chalmers DJ (2008) Estimating subject-specific body segment parameters using a 3-dimensional modeller program. J Biomech 41(16):3506–3510. https://doi.org/10.1016/j.jbiomech.2008.09.021 Davis RB III, Õunpuu S, Tyburski D, Gage JR (1991) A gait analysis data collection and reduction technique. Hum Mov Sci 10(5):575–587. https://doi.org/10.1016/0167-9457(91)90046-Z de Leva P (1996a) Adjustments to Zatsiorsky-Seluyanov’s segment inertia parameters. J Biomech 29(9):1223–1230. https://doi.org/10.1016/0021-9290(95)00178-6 de Leva P (1996b) Joint center longitudinal positions computed from a selected subset of Chandler’s data. J Biomech 29(9):1231–1233. https://doi.org/10.1016/0021-9290(96)00021-8 Delp SL, Anderson FC, Arnold AS, Loan P, Habib A, John CT, Guendelman E, Thelen DG (2007) OpenSim: Open-source software to create and analyze dynamic simulations of movement. IEEE Trans Biomed Eng 54(11):1940–1950. https://doi.org/10.1109/TBME.2007.901024 Dempster WT (1955) Space requirements for the seated operator. Wright Air Development Center, Wright-Patterson Air Force Base, Dayton

76

R. Dumas and J. Wojtusch

Drillis R, Contini R, Bluestein M (1964) Body segment parameters: a survey of measurement techniques. Artif Limbs 8(1):44–66 Dumas R, Aissaoui R, Mitton D, Skalli W, de Guise JA (2005) Personalized body segment parameters from biplanar low-dose radiography. IEEE Trans Biomed Eng 52(10):1756–1763. https://doi.org/10.1109/TBME.2005.855711 Dumas R, Cheze L, Verriest JP (2007a) Adjustments to McConville et al. and Young et al. body segment inertial parameters. J Biomech 40(3):543–553. https://doi.org/10.1016/j.jbiomech. 2006.02.013 Dumas R, Cheze L, Verriest JP (2007b) Corrigendum to “Adjustments to McConville et al. and Young et al. body segment inertial parameters”. J Biomech 40(7):1651–1652. https://doi.org/ 10.1016/j.jbiomech.2006.07.016 Dumas R, Robert T, Cheze L, Verriest J-P (2015) Thorax and abdomen body segment inertial parameters adjusted from McConville et al. and Young et al. Int Biomech 2(1):113–118. https:// doi.org/10.1080/23335432.2015.1112244 Durkin JL, Dowling JJ, Andrews DM (2002) The measurement of body segment inertial parameters using dual energy X-ray absorptiometry. J Biomech 35(12):1575–1580. https://doi.org/10.1016/ S0021-9290(02)00227-0 Ganley KJ, Powers CM (2004) Determination of lower extremity anthropometric parameters using dual energy X-ray absorptiometry: the influence on net joint moments during gait. Clin Biomech 19(1):50–56. https://doi.org/10.1016/j.clinbiomech.2003.08.002 Hinrichs RN (1990) Adjustments to the segment center of mass proportions of Clauser et al. (1969). J Biomech 23(9):949–951. https://doi.org/10.1016/0021-9290(90)90361-6 Ho Hoang K-L, Mombaur K (2015) Adjustments to de Leva-anthropometric regression data for the changes in body proportions in elderly humans. J Biomech 48(13):3732–3736. https://doi.org/ 10.1016/j.jbiomech.2015.08.018 Jackson JN, Hass CJ, Fregly BJ (2015) Residual elimination algorithm enhancements to improve foot motion tracking during forward dynamic simulations of Gait. J Biomech Eng 137(11):111002. https://doi.org/10.1115/1.4031418 Jensen RK (1978) Estimation of the biomechanical properties of three body types using a photogrammetric method. J Biomech 11(8-9):349–358. https://doi.org/10.1016/0021-9290(78) 90069-6 Jovic J, Escande A, Ayusawa K, Yoshida E, Kheddar A, Venture G (2016) Humanoid and human inertia parameter identification using hierarchical optimization. IEEE Trans Robot 32(3):726–735. https://doi.org/10.1109/TRO.2016.2558190 Kadaba MP, Ramakrishnan HK, Wootten ME (1990) Measurement of lower extremity kinematics during level walking. J Orthop Res 8(3):383–392. https://doi.org/10.1002/jor.1100080310 McConville JT, Churchill TD, Kaleps I, Clauser CE, Cuzzi J (1980) Anthropometric relationships of body and body segment moments of inertia. Aerospace Medical Research Laboratory, Wright-Patterson Air Force Base, Dayton Mungiole M, Martin PE (1990) Estimating segment inertial properties: comparison of magnetic resonance imaging with existing methods. J Biomech 23(10):1039–1046. https://doi.org/ 10.1016/0021-9290(90)90319-X Pearsall DJ, Reid G (1994) The study of human body segment parameters in biomechanics. Sports Med 18(2):126–140. https://doi.org/10.2165/00007256-199418020-00005 Pearsall DJ, Reid JG, Livingston LA (1996) Segmental inertial parameters of the human trunk as determined from computed tomography. Ann Biomed Eng 24(2):198–210. https://doi.org/ 10.1007/BF02667349 Pillet H, Bonnet X, Lavaste F, Skalli W (2010) Evaluation of force plate-less estimation of the trajectory of the centre of pressure during gait. Comparison of two anthropometric models. Gait Posture 31(2):147–152. https://doi.org/10.1016/j.gaitpost.2009.09.014 Plagenhoef S (1971) Patterns of human motion: a cinematographic analysis. Prentice-Hall, Englewood Cliffs

Estimation of the Body Segment Inertial Parameters for the Rigid Body. . .

77

Reid JG, Jensen RK (1990) Human body segment inertia parameters: a survey and status report. Exerc Sport Sci Rev 18(1) Sreenivasa M, Chamorro CJG, Gonzalez-Alvarado D, Rettig O, Wolf SI (2016) Patient-specific bone geometry and segment inertia from MRI images for model-based analysis of pathological gait. J Biomech 49(9):1918–1925. https://doi.org/10.1016/j.jbiomech.2016.05.001 Taddei F, Martelli S, Valente G, Leardini A, Benedetti MG, Manfrini M, Viceconti M (2012) Femoral loads during gait in a patient with massive skeletal reconstruction. Clin Biomech 27(3):273–280. https://doi.org/10.1016/j.clinbiomech.2011.09.006 Tisserand R, Robert T, Dumas R, Chèze L (2016) A simplified marker set to define the center of mass for stability analysis in dynamic situations. Gait Posture 48:64–67. https://doi.org/ 10.1016/j.gaitpost.2016.04.032 Valente G, Pitto L, Testi D, Seth A, Delp SL, Stagni R, Viceconti M, Taddei F (2014) Are subjectspecific musculoskeletal models robust to the uncertainties in parameter identification? PLoS One 9(11):e112625. https://doi.org/10.1371/journal.pone.0112625 Vaughan CL, Andrews JG, Hay JG (1982) Selection of body segment parameters by optimization methods. J Biomech Eng 104(1):38–44. https://doi.org/10.1115/1.3138301 Verriest JP (2012) Automatic anthropometric personalization of a digital human model from a set of subject’s photographs. Work 41(Suppl 1):4061–4068. https://doi.org/10.3233/WOR-20120071-4061 Winter DA (2009) Biomechanics and motor control of human movement. Wiley, New York Wu G, Siegler S, Allard P, Kirtley C, Leardini A, Rosenbaum D, Whittle M, D’Lima DD, Cristofolini L, Witte H, Schmid O, Stokes I (2002) ISB recommendation on definitions of joint coordinate system of various joints for the reporting of human joint motion – part I: ankle, hip, and spine. J Biomech 35(4):543–548. https://doi.org/10.1016/S0021-9290(01)00222-6 Wu G, van der Helm FCT, Veeger HEJ, Makhsous M, Van Roy P, Anglin C, Nagels J, Karduna AR, McQuade K, Wang X, Werner FW, Buchholz B (2005) ISB recommendation on definitions of joint coordinate systems of various joints for the reporting of human joint motion – part II: shoulder, elbow, wrist and hand. J Biomech 38(5):981–992. https://doi.org/10.1016/j.jbiomech. 2004.05.042 Yang F, Pai Y-C (2014) Can sacral marker approximate center of mass during gait and slip-fall recovery among community-dwelling older adults? J Biomech 47(16):3807–3812. https://doi. org/10.1016/j.jbiomech.2014.10.027 Yeadon MR, Morlock M (1989) The appropriate use of regression equations for the estimation of segmental inertia parameters. J Biomech 22(6):683–689. https://doi.org/10.1016/0021-9290 (89)90018-3 Young JW, Chandler RF, Snow CC, Robinette KM, Zehner GF, Lofberg MS (1983) Anthropometric and mass distribution characteristics of the adults female. FAA Civil Aeromedical Institute, Oklaoma City Zatsiorsky VM, Seluyanov VN, Chugunova LG (1990) Methods of determining mass-inertial characteristics of human body segments. In: Chernyi GG, Regirer SA (eds) Contemporary problems of biomechanics. CRC Press, Massachusetts, pp 272–291

Part II Discriminative Methods in Dynamic Pose Estimation

3D Dynamic Pose Estimation from Marker-Based Optical Data W. Scott Selbie and Marcus J. Brown

Abstract

The desire to capture images of human movement has existed since prehistoric times (see chapter “Observing and Revealing the Hidden Structure of the Human Form in Motion Throughout the Centuries”). However, it is only since the late nineteenth century and the development of cameras able to capture multiple sequential images that the recording and quantitative analysis of movement has become possible. With modern cameras and high computational power now available, it is commonplace for researchers and clinicians to make detailed measurements, from which an estimation of the position and orientation (pose) of a human body during motion can be computed. This chapter focuses on the estimation of dynamic 3D pose based on optical motion capture systems that record the 3D location of markers attached to the body (see Fig. 1). In this chapter, we describe the estimation of the pose of a multibody model comprising segments that are connected by joints that constrain the direction and range of motion between those segments. There are three common deterministic solutions to the problem of pose estimation; direct, single body, and multibody. This chapter focuses on the two optimization methods, single body and multibody, that provide a deterministic and a discriminative solution to the problem of pose estimation. Unlike the direct pose estimation, these two approaches mitigate, to some extent, uncertainty in the data.

W.S. Selbie (*) HAS-Motion, Inc., Kingston, ON, Canada C-Motion Inc., Germantown, MD, USA e-mail: [email protected] M.J. Brown HAS-Motion, Inc., Kingston, ON, Canada e-mail: [email protected] # Springer International Publishing AG, part of Springer Nature 2018 B. Müller, S.I. Wolf (eds.), Handbook of Human Motion, https://doi.org/10.1007/978-3-319-14418-4_152

81

82

W.S. Selbie and M.J. Brown

Keywords

Skeletal modeling • Pose estimation • Motion-capture • Inverse kinematics • Soft tissue artifact • Optimization

Contents Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . State of the Art . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Six Degree of Freedom (6DOF) Pose Estimation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Pose Estimation Using a Technical Reference Frame (TF) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Inverse Kinematics (IK) Pose Estimation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Weighting . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Observability of the Inverse Kinematics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . IK Optimization Algorithms . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Direction Search Methods (Newton’s Method) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Global Search Methods: Simulated Annealing . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6DOF Versus IK . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Future Directions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Cross-References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

82 84 84 86 88 90 90 94 94 96 97 98 99 99

Introduction For this chapter, the assumption underlying pose estimation is that the human body model is constructed from a set of rigid (nondeformable) segments (or bodies) (see ▶ “Three-Dimensional Reconstruction of the Human Skeleton in Motion”). While this is not literally true, it allows a straightforward model that, for many biomechanical analyses, provides an adequate representation of the underlying skeletal structure for describing motor coordination and functional performance. Each segment is defined by a local anatomical reference frame (Cartesian coordinate system). These subjectspecific anatomical reference frames (AF) are often defined by the location of anatomically palpable landmarks, by matching statistical shape models to surface geometry, or by system identification methods such as functional joints estimated from recorded movements. Regardless of the technique used to establish the reference frame, the common goal is to establish an anatomically relevant reference frame that can be determined reliably and reproducibly. The origin of the reference frame can be located anywhere, but for convenience in this chapter the origin is placed at the proximal end of a segment coincident with the distal end of an adjacent segment (a joint connecting to the parent segment) (Fig. 1). Each segment is restricted to having one parent segment, and the segment’s interaction with its parent segment is described by the specification of joint constraints acting at and around the origin of a segment relative to the parent segment. These joint constraints define the number of degrees of freedom and possibly a prescribed relative path of the segments comprising the joint. The number of degrees of freedom can be any integer value between zero and six. A joint constrained to zero degrees of freedom with no path constraint allows no relative motion between segments, while a joint with six degrees of freedom allows the segments to move independently of each other.

3D Dynamic Pose Estimation from Marker-Based Optical Data

83

Fig. 1 A rigid body (part of a multibody model) is defined by an anatomical reference frame (AF) drawn as red vectors. The left leg is shown with N tracking markers mi and vectors ai describing the location of mi in AF for the right thigh and shank. Note that this configuration of tracking markers is but one of many different configurations in common use. The right leg displays a cluster of markers (sometimes secured to a rigid shell) and the left leg displays skin-based tracking markers attached at palpable anatomical landmarks

To estimate the pose of the multibody model, the 3D location of reflective markers attached to the segments is recorded by one or more optical sensors. It is beyond the scope of this chapter to describe the algorithms for identifying these 3D locations from the optical sensors, but regardless of the optical technology, the resultant 3D locations are used consistently between approaches. The tracking of each segment (pose estimation during a dynamic trial) is accomplished by establishing the location of the markers in the segment’s anatomical reference frame to which they are attached, recording the location of these markers in each frame of a motion trial, and by satisfying the specified joint constraints. A fundamental assumption of the algorithms presented in this chapter is that segments are rigid and the markers attached to those segments are secured rigidly and do not move relative to the segment to which they are attached. The number of markers required, and the number of segments to which markers are attached, depends on the structure of the multibody model and the pose estimation algorithm being used. The most important concept within this methodology is observability. Observability is dealt with in more detail later in the chapter; however, in short, a system is observable if the data are sufficient to describe, uniquely, the pose of the model. If the markers were truly attached rigidly to the underlying skeleton, i.e., a marker’s coordinates in the AF were invariant during movement, and the segments of the multibody model were truly rigid, and the markers were never occluded, this would be a straightforward chapter as all the pose estimation methods described in the scientific literature and textbooks would yield reliable pose estimations, and we could choose the mathematically simplest approach.

84

W.S. Selbie and M.J. Brown

Any marker that is attached to the skin, however, can move relative to the underlying skeleton (Cappozzo et al. 1996). This relative motion occurs as flesh between the marker and skeleton deforms during movements, and is commonly known in the biomechanics community as soft tissue artifact (STA). It is, as yet, challenging to mitigate STA through mathematical approaches because, while STA is systematic, it varies on a case-by-case basis between individuals, between locations on the body, and between movements. Pose estimation algorithms that mitigate these “uncertainties” resulting from STA can improve the effectiveness of pose estimation dramatically. The two pose estimation algorithms discussed in this chapter are common in the biomechanics community and are deterministic and discriminative. In other words, they rely solely on the structure of the multibody model and instantaneous data to estimate pose. This is in contrast with probabilistic pose estimation, in which prior information (e.g., models of STA or predictions based on the statistics of past performance) are incorporated into the pose estimation algorithm (see chapter ▶ “3D Dynamic Probabilistic Pose Estimation From Data Collected Using Cameras and Reflective Markers”).

State of the Art Six Degree of Freedom (6DOF) Pose Estimation This section describes an algorithm for six degree of freedom (6DOF) pose estimation, sometimes referred to as a segment optimization algorithm (Lu and O’Connor 1999) or single-body optimization. To estimate the pose of a segment at each frame of data, the 6DOF algorithm requires that a set of not less than three noncollinear markers be attached to each segment. To clarify the need for three markers, we will describe the information available from 1, 2, or more markers on a segment. If a segment was to have a single marker attached to it, this marker would permit the estimation of translations of the segment along the three principal axes of the global reference frame (e.g., 3DOF). If a second marker was added, it would be possible to estimate rotations about two principal axes of the segment; however, rotations about an axis between the two markers would be undetectable (e.g., 5DOF). When a third marker is added, offset from the line between the first two, rotations about all three segmental axes become observable (e.g., 6DOF). Additional markers on a segment cannot increase the number of degrees of freedom but, as will be see below, can be useful in a least-squares sense. This method is referred to as a 6DOF method because each segment (or joint) is considered to have six independent variables that describe its pose; three variables describe the location of the segment’s origin within the global reference frame (its position) and three variables describe the rotation about each of the principal axes of the segment (its orientation). In principle, each segment can be tracked independently of any other

3D Dynamic Pose Estimation from Marker-Based Optical Data

85

segment. This independence infers that there is no explicit linkage defined, i.e., there are no preconceived assumptions about the properties of any joint connecting segments. This means that the endpoints of a segment, and those of its the proximal and distal adjacent neighbors, are free to move relative to each other, based directly and solely on the recorded motion capture (MoCap) data (Cappozzo et al. 1995). This independent estimation of the pose of the segments requires that markers used to track one segment are not used to track any other segment. It is quite common, however, for one marker to be used as a tracking marker on two adjacent segments. For example, a lateral knee marker may be used as a tracking marker on the thigh and the shank. In this situation, the thigh and shank segments are still 6DOF because six variables describe the motion of a segment, but in this case the segments are not actually independent of each other. The 6DOF algorithm we describe here estimates the pose of a segment using a least-squares procedure (Kepple and Stanhope 2000). Consider a point mi attached to a segment, whose location is represented by vector ai in the AF. The location of the same marker mi is represented by vector vi in the GF (vi = the data recorded). The relationship between ai and vi is given by: vi ¼ RAG ai þ OAG

(1)

where: RAG is a rotation matrix from AF to GF OAG is the translation from AF to GF. The rotation matrix RAG and translation vector OAG may be computed at any instant, given that at least three noncollinear vectors ai are assumed stationary in the AF, and vi are recorded in GF, by minimizing the sum of squares error expression: f ðRAG , OAG Þ ¼

N X

ðvi  RAG ai  OAG Þ2

(2)

i¼1

where N is equal to the number of tracking targets on the segment. There are an infinite number of solutions of RAG and OAG that will produce minima for Eq. 2. Not all of these solutions result in RAG being a rotation matrix, so we specify the orthonormal constraint RTAG RAG ¼ I as a boundary condition on the solution (Spoor and Veldpaus 1980): gðRAG Þ ¼ RTAG RAG  I ¼ 0

(3)

The method sets the gradient of Eq. 2 equal to the gradient of Eq. 3 times a set of Lagrangian multipliers: ∇f ðRAG , OAG Þ ¼ λ∇gðRAG Þ

86

W.S. Selbie and M.J. Brown

This results in a system of algebraic equations: ∇f ðRAG , OAG Þ  λ∇gðRAG Þ ¼ 0

(4)

for which there exists an exact solution as long as N  3 The 6DOF algorithm requires a minimum of three noncollinear tracking markers, but more can be accommodated because the 6DOF algorithm permits a solution for an over-specified system with an unlimited number of tracking markers on a segment. This over-specification means that, provided noise (or some features of STA) in the data is uncorrelated, the least-squares algorithm will act to minimize the effects of the noise. If one or more tracking targets are missing in any frame(s), the over-specification still allows a calculated segmental pose, provided at least three noncollinear targets are present. The observability for a 6DOF method is straightforward because it is simply N  3, provided the locations of the markers are fixed in the AF, and are not collinear. In principle, tracking markers can be placed anywhere on a rigid segment. In practice, marker placement on an anatomical segment is a compromise between distributing markers over the entire surface of a segment and placing markers in areas that exhibit minimal STA (Cappozzo et al. 1997). As concluded in a review article by Cereatti et al. (2006), there have been attempts to modify the 6DOF algorithm in order to mitigate the effects of STA (Cappozzo et al. 1997; Andriacchi et al. 1998), but none of these approaches have proved satisfactory.

Pose Estimation Using a Technical Reference Frame (TF) While this chapter is focused on estimating pose from marker data, it is convenient at this time to discuss briefly pose estimation from two other 6DOF sensors: electromagnetic sensors and Moiré-phase tracking. It is beyond the scope of this chapter to describe the theory behind the sensor technology, but in summary, electromagnetic systems record the 6DOF pose of a sensor relative to an emitted electromagnetic dipole field. The Moiré-phase tracking (MPT) 3D motion capture system (Weinhandl et al. 2010) is a single-camera 3D motion tracking technology that tracks the 6DOF pose of a Moiré target (a lightweight, multilayer passive optical target; Weinhandl et al. 2010). The important idea to note is that these sensors describe their pose relative to an internal reference frame, not an anatomical frame. To put these 6DOF sensors in the context of marker-based MoCap (the focus of the chapter), we consider a slightly different approach to the 6DOF algorithm. Consider the same markers mi from Fig. 1, but instead of creating vectors ai in the anatomical reference frame of the segment, we create vectors bi in a technical reference frame (TF) defined by the markers (Fig. 2). In this description, the segment origin is located at one of the markers (m2), the principal axis is defined by vector from (m2 to m1), and the reference frame is established from the principal axis and m3. This adds another “layer” to the pose estimation as it requires an additional step to include the transformation between this TF and the associated AF (RTA, OTA).

3D Dynamic Pose Estimation from Marker-Based Optical Data

87

Fig. 2 A rigid body (part of a multibody model) is defined by an anatomical reference frame (AF) drawn as red vectors. The left leg is shown with N tracking markers mi and vectors bi describing the location of mi in a technical reference frame TF for the right thigh and shank. The left leg displays a Moiré-phase tracking sensor (top) and electromagnetic sensor (middle)

Using the same markers (mi) from one frame of data, and assuming that the transformation from TF to AF is invariant, we can identify (RTA, OTA) from vector calculus using the same methods used to define AF in the first place. Consider a point mi attached to a segment, whose location is represented by vector bi in the TF. Eq. (1) is written as: vi ¼ RTG bi þ OTG

(5)

where: RTG is a rotation matrix from TF to GF OTG is the translation from TF to GF RTG and OTG are computed as in Eq. 4. The resulting pose estimation (RTG, OTG) in a local reference frame, defined by the markers independently of the anatomy, is similar to the pose estimates of the other 6DOF sensors. In marker-based MoCap, it is possible to define the relationship

88

W.S. Selbie and M.J. Brown

between the markers and the anatomy because markers can be placed in locations that have anatomical meaning. With the electromagnetic and MPT technologies, the anatomical locations can be identified with a pointer, or system identification methods can be used to identify the AF. As in the previous section, our goal is once again to identify (RAG,OAG), but in this case, the least-squares solution computes the transformation from the TF to the GF (RTG,OTG). The additional step is to include the additional transform from TF to AF (RTA,OTA). RAG ¼ RTTA RTG OAG ¼ OTG  OAT

(6)

There is a considerable benefit to the 6DOF approach to pose estimation, as it is straightforward to implement with results that are easy to understand. The 6DOF solution has no local minima, and requires no guidance from users. Notably, 6DOF estimates a pose that is an accurate representation of the data, which is useful for identifying local problems. An example of such a local problem would be the swapping of the names of two markers between trials, or even within a trial (something not uncommon when working with many passive reflected markerbased MoCap systems). Such mislabeling of markers will cause obvious discontinuities in the pose estimations of a 6DOF segment, which can be easily identified and corrected. The deterministic assumption that neither STA nor noisy marker data occur can result in pose estimations where the adjacent endpoints of segments are dislocated from each other or “merge” together. While these pose solutions reflect the true marker data, and thus highlight the presence of noise and/or STA, they can present estimations of pose that are anatomically impossible. To highlight the serious challenge of STA, if the entire set of markers translates in unison (e.g., through inertial forces or impact), the estimated pose of the segment can be quite wrong. There is, however, no information in the relative configuration of the tracking markers to indicate that anything has gone awry, so this artifact cannot be mitigated. The next section describing inverse kinematics discusses a deterministic approach to remove such an obvious artifact as joint disarticulation from the 6DOF model.

Inverse Kinematics (IK) Pose Estimation Inverse kinematics (IK) is the search for, and identification of, an optimal pose of a multibody model with explicit joint constraints, such that the overall differences between the measured and model-estimated marker coordinates are minimized, in a least-squares sense, at a system level. Lu and O’Connor (1999) termed this process global optimization, but in this chapter, we will refer to this as multibody optimization or IK. IK, as described here, is a least-squares solution that may be considered an extension to the 6DOF pose estimation because if a joint is ascribed six degrees of

3D Dynamic Pose Estimation from Marker-Based Optical Data

89

Fig. 3 A multibody model showing the pelvis as a root segment (e.g., 6DOF with respect to the global reference frame) and joint constraints at the hip, knee, and ankle of the left leg. In this figure, the hip has 3DOF, the knee 5DOF, and the ankle 3DOF, but many other multibody configurations can be found in the literature

freedom within the IK, the IK and 6DOF solutions are equivalent. Selecting appropriate joint constraints is idiosyncratically based on the number of markers being tracked, the context of the motion being analyzed, and many other factors; some of these factors will be discussed later in this section. As with 6DOF, the algorithms involved in IK pose estimation will be described in the context of marker-based optical 3D MoCap (Fig. 3). The solution to the IK is the pose of a multibody model that best matches the MoCap data, in terms of a least-squares criterion. In the Lu and O’Connor (1999) approach, the IK solution is found for each frame of data, independent of any previous or subsequent frames of data. Mathematically, van den Bogert and Su (2008) described this approach, based on the overall configuration of the multibody model, using a set of generalized coordinates q.

90

W.S. Selbie and M.J. Brown

Generalized coordinates are the minimum set of independent variables that describe the pose. In this case, R and O of Eq. 1 now consist of multiple transformations and become a function of the generalized coordinate vector q: vi ¼ RðqÞai þ OðqÞ

(7)

The expression that is minimized becomes: f ðR, OÞ ¼ f ðqÞ ¼

N X

ðvi  RðqÞai  OðqÞÞ2

(8)

i¼1

where N is the total number of targets on all the segments in the IK chain.

Weighting Within an IK model, it is possible to rely more on data that are known, a priori, to contain less noise or be less affected by STA. This can be achieved via a weighting. f ðR, OÞ ¼ f ðqÞ ¼

N X

αi ðvi  RðqÞai  OðqÞÞ2

(9)

i¼1

The selection of the weights, αi, can be made pragmatically and heuristically, or rules may be used that allow the computation of an optimal set of weights. Without a priori information, it is usually best to set αi to 1, but on occasion, when estimating pose, the user may want to ensure that certain segments follow the tracking targets with a higher degree of accuracy than other segments. For example, the user may want the distance between the foot and the floor (or recorded ground reaction force) to remain similar to the values that would be obtained using a 6DOF method because 6DOF is likely the best local estimate of the pose of the foot. Likewise, data from some markers may not be considered representative of the pose because they are noisy, so the weight of these data can be reduced. In some cases, the marker may be known to have substantial STA relative to one of the degrees of freedom (generalized coordinates) and the influence of the marker on this generalized coordinate can be removed.

Observability of the Inverse Kinematics As mentioned previously, the pose of a multibody model is observable if the data are sufficient to describe the pose uniquely. In the case of the 6DOF pose estimation, three or more rigidly attached, noncollinear targets are required to track each segment. When one target is placed on a rigid segment, three independent pieces of information can be obtained, the X, Y, and Z coordinates of the target. When a

3D Dynamic Pose Estimation from Marker-Based Optical Data

91

second target is placed on the segment, two further pieces of information are obtained. The number of new pieces of information for the second target is two, not three like the first target, because if we know the X and Y locations of the second target, then the Z coordinate is known because the distance between the first and second targets is fixed. Thus, two targets only supply five of the six unknowns. When a third target is added, one additional piece of information is supplied; note the third target only adds one new piece of information because the distance from the third target to the first target and the distance from the third target to the second target are fixed. Still with three noncollinear targets, we have sufficient information to fully solve the pose of a 6DOF segment. With IK, not only is there the assumption of rigid segments, but there are also constraints added at the joints. A consequence of the joint constraints is that fewer than three markers may be sufficient to fully determine the pose of a segment. For example, a segment that has only one degree of freedom (e.g., one connected to a parent segment by a hinge joint) only requires one marker to fully determine the joint angle. It is not possible to just count markers, however, because if this one marker is coincident with the hinge joint, it does not provide any information and the pose is nonobservable. Therefore, the question of whether the markers provide sufficient information to determine the model’s pose is far more complex when joint constraints exist. A straightforward approach to the problem would be to specify the number of targets required to track a segment, based solely on the type of joint connecting that segment to its parent. For example, Yeadon (1984) required two markers to track a segment connected to the parent via a ball joint (three degrees of freedom) or a universal joint (two degrees of freedom) and required only one marker when the segment was connected via a one degree of freedom hinge joint. Although this approach will guarantee that the system will likely be observable, if these requirements are met, it can be overly conservative and will occasionally consider the model to be unobservable, when in fact there is sufficient information available. For example, Schulz and Kimmel (2010) demonstrated that it is possible to track the pose of the thigh segment without actually placing any markers on the thigh. Yeadon’s method would declare this model to be unobservable. This is important because for many activities, the STA of markers on the thigh is detrimental to an accurate estimate of the pose and if Schulz’s assumption that the hip has three degrees of freedom and the knee has one degree of freedom is an accurate reflection of the movement, his approach could be useful for studying many activities. To demonstrate how it is possible to calculate a general solution to the observability problem, consider the simple example of a single segment constrained to its parent (in this example, the ground) by a ball joint. This system can be fully described by three degrees of freedom: the Euler rotations, θx , θy , and θz. For this case, the general IK objective function Eq. 8 becomes: P ð qÞ ¼

m X i¼1

fðR0 ðqÞAi Þg

2

(10)

92

W.S. Selbie and M.J. Brown

Assume there is only one target (m = 1) fixed to the segment, the local coordinates, in the AF, of that target (Ax, Ay, Az) and the global coordinates of the targets, in the GF, are (Px, Py, Pz). Applying Eq. (10) for this simple case of one segment connected to the ground via a ball joint with a single tracking target, the objective function f(q) is: Pðq Þ             AX cos ðθz Þ cos θy þ Ay cos ðθz Þ sin θy sin ðθx Þ þ sin ðθz Þ cos ðθx Þ þ Az  cos ðθz Þ sin θy cos ðθx Þ þ sin ðθz Þ sin ðθx Þ  Px               ¼  Ax sin ðθz Þ cos θy þ Ay  sin ðθz Þ sin θy sin ðθx Þ þ cosðθz Þ cos ðθx Þ þ Az sin ð θ Þ sin θ ð Þ þ cos ð θ Þ sin ð θ Þ  P cos θ y x z x y  z   AX sin θy þ Ay cos θy sin ðθx Þ þ Az cos θy cos ðθx Þ  Pz

If a change in rotation (some combination of a change in θx , θy , θz) exists for which the target does not move, then the system is not observable. In order to establish whether this is the case, it is necessary to discover if a situation exists where the cost function does not change with respect to changes in the joint angle. This exactly describes the Jacobian (or matrix of partial derivatives) of the cost function:    d ð Px Þ d ð Px Þ d ð Px Þ       d ðθx Þ d θy dðθz Þ         d P d Py d Py  y    Jacobian ¼    d ðθx Þ d θy d ðθ z Þ     d ð Pz Þ d ð Pz Þ d ð Pz Þ       d ðθx Þ d θy d ðθ z Þ  Calculating the Jacobian of the cost function described in Eq. 10:     d ðP x Þ ¼ Ay cos ðθz Þ sin θy cos ðθx Þ  sin ðθz Þ sin ðθx Þ d ðθ x Þ     þ Az cos ðθz Þ sin θy sin ðθx Þ þ sin ðθz Þ cos ðθx Þ       d ð Px Þ   ¼ Ax cos ðθz Þ sin θy þ Ay sin ðθz Þ cos θy sin ðθx Þ  sin ðθz Þ sin ðθx Þ d θy     þ Az  cos ðθz Þ cos θy cos ðθx Þ       d ð Px Þ ¼ Ax sin ðθz Þ cos θy þ Ay  sin ðθz Þ sin θy sin θx þ cos ðθz Þ cos ðθx Þ d ðθ z Þ     þ Az sin ðθz Þ sin θy sin ðθx Þ þ cos ðθz Þ sin ðθx Þ       d Py ¼ Ay  sin ðθz Þ sin θy cos ðθx Þ  cos ðθz Þ sin ðθx Þ d ðθ x Þ     þ Az  sin ðθz Þ sin θy sin ðθx Þ þ cos ðθz Þ cos ðθx Þ         d Py   ¼ Ax sin ðθz Þ sin θy þ Ay  sin ðθz Þ cos θy sin ðθx Þ  cos ðθz Þ sin ðθx Þ d θy     þ Az sin ðθz Þ cos θy cos ðθx Þ

3D Dynamic Pose Estimation from Marker-Based Optical Data

93

        d Py ¼ Ax cos ðθz Þ cos θy þ Ay  cos ðθz Þ sin θy sin ðθx Þ  sin ðθz Þ cos θx  d ðθz Þ     þ Az cos ðθz Þ sin θy sin ðθx Þ  sin ðθz Þ sin ðθx Þ     d ð Pz Þ ¼ Ay cos θy cos ðθx Þ  Az cos θy sin ðθx Þ d ðθ x Þ       d ð Pz Þ   ¼ Ax cos θy þ Ay sin θy sin ðθx Þ  Az sin θy cos ðθx Þ d θy d ð Pz Þ ¼0 d ðθ z Þ To simplify this equation, consider the state where θx = 0, θy = 0, The Jacobian now becomes:    0 Az Ay   0 Ax  Jacobian of cost function ¼  Az  Ay Ax 0 

θz = 0

The determinant of the Jacobian is:   Det ¼ 0ð0  Ax Ax Þ  AZ Ax Ay  0 þ Ay ð Az Ax  0Þ ¼ 0  Ax Ay Az þ Ax Ay Az Since the determinant of the Jacobian is zero, it is not invertible and its rank is not full; thus, one target is not sufficient to estimate the pose of a segment connected to ground via a ball joint. Assume now that two targets are attached to the segment: A1 and A2. In this case, the cost function (Eq. 10) for one segment that is connected to ground by a ball joint is: Pðp Þ             Ax1 cos ðθz Þ cos θy þ Ay cos ðθz Þ sin θy sin ðθx Þ þ sin ðθz Þ cos ðθx Þ þ Az1  cos ðθz Þ sin θy cos ðθx Þ þ sin ðθz Þ sin ðθx Þ  Px1  1              Ax1 sin ðθz Þ cos θy þ Ay  sin ðθz Þ sin θy sin ðθx Þ þ cos ðθz Þ cos ðθx Þ þ Az1 sin ðθz Þ sin θy cos ðθx Þ þ cos ðθz Þ sin ðθx Þ  Py  1           þ A sin θ cos θ A sin θ cos θ ð Þ þ A cos θ ð Þ  P x y y y x z y x z 1 1 1             ¼   Ax2 cos ðθz Þ cos θy þ Ay2 cos ðθz Þ sin θy sin  ðθx Þ þ sin ðθz Þ cos ðθx Þ þ Az2 cos ðθz Þ sin θy  cos ðθx Þ þ sin ðθz Þ sin ðθx Þ  Px2   Ax sin ðθz Þ cos θy þ Ay  sin ðθz Þ sin θy sin ðθx Þ þ cos ðθz Þ cos ðθx Þ þ Az sin ðθz Þ sin θy cos ðθx Þ þ cos ðθz Þ sin ðθx Þ  Py  2  2 2      2    Ax2 sin θy þ Ay2 cos θy sin ðθx Þ þ Az2 cos θy cos ðθx Þ  Pz2

Again, taking the simplest case and setting the orientation to θx ¼ 0, θy ¼ 0, θz ¼ 0 The Jacobian of the cost function now reduces to:    0 A1z A1y    A1 z 0 A1x    A1y A1x 0  Jacobian of cost function =  A2z A2y   0  A2 z 0 A2x    A2y A2x 0 

94

W.S. Selbie and M.J. Brown

If targets A1 and A2 have coordinates (0, 0, A1z) and (0, 0, A2z) which they are collinear along the Z axis, we would expect the system to be unobservable as the targets will not register rotation about the Z axis. For this case:    0 A1z 0    0 0  A1z   0 0  0 Jacobian of cost function =    0 A2z 0    0 0  A2z   0 0 0 Column 3 equals zero, not full column rank, and thus the system is not observable as expected. Now If the two targets do not form a line that points to the joint center, for example: A1 ¼ ð0:1, 0:1, 0:1Þ and A2 ¼ ð0:1, 0:1, 0:1Þ The Jacobian now is:   0   0:1   0:1 Jacobian of cost function ¼   0   0:1  0:1

 0:1 0:1   0 0:1   0:1 0   0:1 0:1   0 0:1   0:1 0

This matrix has a rank = 3, which is full column rank and thus marker information (A1 and A2) is independent and the model is fully observable. Therefore, the general solution for observability in inverse kinematics reduces to determining whether the Jacobian for cost function of Eq. 10 has full rank. If it does, we have sufficient information to determine the pose of the model. Conversely, if the rank of the Jacobian of the IK cost function is not full rank, there is not enough information to determine a unique pose for the model.

IK Optimization Algorithms In the general case, there is no analytic solution for the IK problem. We, therefore, summarize examples from two classes of implementation of a numerical solution to this optimization problem: direction search methods and global search methods.

Direction Search Methods (Newton’s Method) To understand Newton’s method, consider a function f(q) that starts at an initial vector q0, moves through a series of vectors qk, and converges to a solution at qmin.

3D Dynamic Pose Estimation from Marker-Based Optical Data

95

Newton’s method is a three step process: 1. Compute the search direction 2. Determine the length of the next step 3. Use the results of steps 1 and 2 to obtain a new point qk. These steps are repeated until a minimum is found To find the search direction (Step 1) using Newton’s method, consider a vector, q, on f(q) located near the current value qk .The vector q  qk can be approximated by a second-order Taylor series expansion: f ðq  qk Þ ¼ f ðqk Þ þ Δf ðqk Þðq  qk Þ þ

Bðqk Þðq  qk Þ2 þ ... 2

(11)

where Δf(qk) is the gradient of f at the current value qk and B is the Hessian, or matrix, of second partial derivatives at qk. Taking the derivative of the function in Eq. 11 with respect to q and ignoring the derivative of the third term (e.g., the Hessian), we obtain: Δf ðq  qk Þ ¼ Δf ðqk Þ þ Bðqk Þðq  qk Þ

(12)

The derivative has a minimum at: 0 ¼ Δf ðqk Þ þ Bðqk Þðq  qk Þ

(13)

and thus the search direction, (q  qk), can be obtained from: ðq  qk Þ ¼ Δf ðqk ÞB1 ðqk Þ

(14)

After solving for the search direction, (q  qk), the next point in the search, qk + 1, is found by moving in the direction of (q  qk). Ideally, the step size is determined by the magnitude of the eigenvalues of the movement to ensure that we obtain a sufficient decrease in the cost function, without taking excessively small steps. In practice, steps sizes that have worked for previous data sets are assumed to be sufficient. Once qk + 1 is obtained, it is checked against a termination criterion (is (q  qk) small). If the termination criterion is satisfied, then the minimum for the global IK problem is found. If the criteria is not met, the process is repeated, beginning at step 1 with qk + 1 acting as the new current value qk. Ideally, Δf(qk) and B (Hessian) are derived symbolically but this is not always straightforward. Furthermore, even if the symbolic version of the Hessian is derived, computing the inverse of the Hessian, B1, requires a series of linear equations to be solved, which can be computationally costly. An alternative to this method, called the quasi-Newton method, Δf(qk) and B are approximated numerically by the change in the gradient between steps. Several methods of approximation have been proposed that all follow three primary assumptions:

96

W.S. Selbie and M.J. Brown

1. The Hessian must be symmetrical 2. The model gradient must be equal to the function gradient at the current step and at the previous step 3. The Hessian cannot change drastically between successive steps The consequence of these assumptions is that convergence may be compromised. Unlike the 6DOF least-squares solution, there are many possible solutions to the IK optimization as the solution space typically has many local minima. If the initial estimated position, q0, is “close” to the global minimum, the solution will likely converge to the correct solution. The initial estimated position, or “seed,” is therefore critical to the success of the algorithm. For the first frame of data, it is possible to use a 6DOF solution as the seed. For subsequent frames, the seed for the optimization algorithm at any given frame is the state of the model at the previous frame. This could be problematic if the solution at the previous frame was an inappropriate local minimum, resulting in subsequent pose estimates diverging from the real solution due to being held in this local minimum. For example, the data collection volumes of most optical MoCap systems are smaller than the laboratory that they are in, and subjects often begin their movements outside the volume (for example, to perhaps ensure that they are at a constant speed while walking or running through the data collection volume). The first frame with complete data can often be relatively unreliable because it is captured near to the edge of the calibrated volume, and therefore the likelihood of the optimization solution becoming trapped in a local minimum increases. In order to avoid this, one potential improvement to the algorithm is to compute the solution both forward and backward, in the hope that one of the passes will provide a more optimal solution path.

Global Search Methods: Simulated Annealing Simulated annealing (Higginson et al. 2005; Ingber 2012) is a Monte Carlo method in which the solution space is explored probabilistically by randomly searching near the best known solution. Simulated annealing is not prone to finding a local minima and therefore, given “enough” computing time (unfortunately, “enough” cannot be calculated but needs to be learned from experience), finds the global minimum. It is modeled after annealing in metallurgy, in which the thermodynamic free energy of a metal decreases as its temperature cools. In simulated annealing, as the virtual temperature cools, the algorithm searches in a smaller and smaller region around the best known solution (Fig. 4). Simulated annealing functions using two nonobvious principles: 1. Some new values that do not actually reduce the minimum value are allowed so that more of the solution space can be explored. (The allowed values are determined by the Metropolis criteria.) 2. After making many estimates, and observing that the cost function declines slowly, one lowers the temperature and thus limits the size of allowed values

3D Dynamic Pose Estimation from Marker-Based Optical Data

97

Fig. 4 Flowchart of the simulated annealing algorithm. The size of the perturbation is based on current temperature and the Metropolis criteria: Randð0, 1Þ < e

f i f best T

that are larger than the current minimum. After lowering the temperature several times, only more optimal values are accepted, and the optimization approaches the global minimum. One of the biggest challenges to simulated annealing is that the algorithm is computationally expensive, and perhaps more problematically, it is not possible to determine if the current solution is actually a global minimum without continuing the optimization indefinitely. In other words, there is no threshold or criterion for identifying that the search is complete. The user must decide how many iterations to perform in the optimization and accept that the minimum found in that time period may not be the global minimum. Despite the computational cost (time), simulated annealing is a more robust algorithm than direction search algorithms. Despite the robustness, however, most IK users opt for direction search algorithms because of time constraints.

6DOF Versus IK In many circumstances, the IK solution is likely to be more anatomically congruent and therefore preferable to the 6DOF solution, but the user must attend to the

98

W.S. Selbie and M.J. Brown

determination of the appropriateness of the selected joint constraints. For example, an experiment that was focused on understanding the kinematics of an injured knee, where translations and rotations occur as a result of the injury (e.g., anterior cruciate ligament damage), would likely not benefit from an IK solution where the constraints, and consequent prescribed motion of the knee joint, “hide” the pathology. Finally, it is well known that residual errors, i.e., differences between model predictions and marker measurements, computed by IK algorithms are reflections of noise in the marker data, soft tissue artifact, and inaccurate marker placement. A limitation of the IK algorithm, however, is that it has no straightforward mechanism to compensate for systematic noise, even though it can be used to identify its presence.

Future Directions In this chapter, we have described the current state of deterministic pose estimation algorithms. The future evolution of deterministic algorithms is quite limited. Begon et al. (2016), for example, has introduced an approach that removes STA without modeling STA but rather by ignoring information in markers that are considered unreliable. For many segments of the human body, STA has a particularly disastrous effect on the axial rotation of the segment. In other words, the markers rotate about the long axis of the segment (upper arm, forearm, thigh to name a few). Begon’s solution was to ignore any information in the marker that would reflect axial rotation by projecting tracking markers onto the long axis of the segment. These projected markers influence five of the degrees of freedom of a segment only. The long axis rotation is then estimated based on the pose of adjacent-constrained segment. The example given by Begon is movement of the upper arm, in which the axial rotation of the upper arm is estimated by constraining the elbow joint to have only two rotational degrees of freedom, and therefore the axial rotation of the upper arm is based on the pose of the forearm. There is some potential for improvements to deterministic pose estimation algorithms based on similarly clever rejection of data in isolated/idiosyncratic cases. It is our believe that the future of marker-based pose estimation lies not in deterministic algorithms but in algorithms based on Bayesian Inference (Todorov 2007) (chapter ▶ “3D Dynamic Probabilistic Pose Estimation from Data Collected Using Cameras and Reflective Markers”) and algorithms based on optimal control theory (Miller and Hamill 2015) (▶ “Optimal Control Modeling of Human Movement”). Bayesian Inference allows a principled way to mitigate the effects of STA by modeling artifact and removing it. Optimal control theory is capable of generating motion independently of any recorded data based on generated simulated motion of the behavior based on some optimization criteria (e.g., minimum energy). The technique can be influenced by recorded data to ensure that the pose estimation is arbitrarily close to the recorded motion. Optimal control theory has the additional

3D Dynamic Pose Estimation from Marker-Based Optical Data

99

benefit of being able to generate solutions for unobservable, and even sparse, marker sets. Lastly, it is important to consider algorithms for which the soft tissue artifact is considered important data reflective of an individual subject instead of an artifact to be removed. Michael Black’s laboratory at the Max Planck Institute for Intelligent systems has been developing pose estimation algorithms based on statistical shape models (Loper et al. 2015). Instead of defining pose based on the position and orientation of an underlying skeleton, this research has focused on modeling the surface geometry of the subject and estimating the pose of the surface. Based on high-density surface scans of subjects performing movement, the statistical shape model is a parameterized surface that can be subsequently fit to sparse surface data (e.g., markers). These models are remarkably good at representing the surface of the body during motion. From a biomechanics perspective, a fundamental question is whether we can infer the multibody skeletal pose from this parameterized surface data.

Cross-References ▶ 3D Dynamic Probabilistic Pose Estimation from Data Collected Using Cameras and Reflective Markers ▶ Observing and Revealing the Hidden Structure of the Human Form in Motion Throughout the Centuries ▶ Optimal Control Modeling of Human Movement ▶ Three-Dimensional Human Kinematic Estimation Using Magneto-Inertial Measurement Units ▶ Three-Dimensional Reconstruction of the Human Skeleton in Motion

References Andriacchi TP, Alexander EJ, Toney MK, Dyrby C, Sum J (1998) A point cluster method for in vivo motion analysis: applied to a study of knee kinematics. J Biomech Eng 120:743–749 Begon M, Bélaise C, Naaim A, Lundberg A, Chèze L (2016) Multibody kinematics optimization with marker projection improves the accuracy of the humerus rotational kinematics. J Biomech (16):31111–31113 Cappozzo A, Catani F, Croce UD, Leardini A (1995) Position and orientation in space of bones during movement: anatomical definition and determination. Clin Biomech 10(4):171–178 Cappozzo A, Catani F, Leardini A, Benedetti MG, Della Croce U (1996) Position and orientation in space of bones during movement: experimental artefacts. Clin Biomech 11(2):90–100 Cappozzo A, Cappello A, Della Croce U, Pensalfini F (1997) Surface-marker cluster design criteria for 3-D bone movement reconstruction. IEEE Trans Biomed Eng 44(12):1165–1174 Cereatti A, Della Croce U, Cappozzo A (2006) Reconstruction of skeletal movement using skin markers: comparative assessment of bone pose estimators. J Neuro Eng Rehabil 3(1):7 Higginson JS, Neptune RR, Anderson FC (2005) Simulated parallel annealing within a neighborhood for optimization of biomechanical systems. J Biomech 38:1938–1942

100

W.S. Selbie and M.J. Brown

Ingber L (2012) In: Oliveira H, Petraglia A, Ingber L, Machado M, Petraglia M (eds) Adaptive simulated annealing, in stochastic global optimization and its applications with fuzzy adaptive simulated annealing. Springer, New York, pp 33–61 Kepple T, Stanhope S (2000) Moved software. In: Winters, Crago (eds) Biomechanics and neural control of posture and movement. Springer, New York Loper M, Mahmood N, Romero J, Pons-Mol G, Black MJ (2015) SMPL: a skinned multi-person linear model. ACM Trans Graph 34(6):248:1–248:16. ACM Lu TW, O’Connor JJ (1999) Bone position estimation from skin marker co-ordinates using global optimization with joint constraints. J Biomech 32:129–134 Miller R, Hamill J (2015) Optimal footfall patterns for cost minimization in running. J Biomech 48:2858–2864 Schulz BW, Kimmel WL (2010) Can hip and knee kinematics be improved by eliminating thigh markers?Clinical. Biomechanics 25(2010):687–692 Spoor C, Veldpaus F (1980) Rigid body motion calculated from spatial coordinates of markers. J Biomech 13(4):391–393 Todorov E (2007) Probabilistic inference of multijoint movements, skeletal parameters and marker attachment from diverse motion capture data. IEEE Trans on Biomed Eng 54:1927–1939 Van Den Bogert AJ, Su A (2008) A weighted least squares method for inverse dynamic analysis. Comput Methods Biomech Biomed Eng 11(1):3–9 Weinhandl JT, Armstrong BSR, Kusik TP, Barrows RT, O’Connor KM (2010) Validation of a single camera three-dimensional motion tracking system. J Biomech 43(7):1437–1440 Yeadon MR (1984) The mechanics of twisting somersaults. Doctoral thesis. University of Calgary

Measurement of 3D Dynamic Joint Motion Using Biplane Videoradiography Hans Gray, Shanyuanye Guan, Peter Loan, and Marcus Pandy

Abstract

Accurate measurement of in vivo joint kinematics is important for understanding normal and pathological human motion and for evaluating the outcome of surgical procedures. Biplane videoradiography is currently the most accurate method available for measuring in vivo joint kinematics noninvasively. The method uses two X-ray images obtained from different perspectives to deduce precise three-dimensional spatial information of the bones that meet at a joint. The abilities to collect high-quality X-ray images at high frame rates and to process these images in a time efficient manner are key factors determining the feasibility of using modern biplane videoradiography systems to measure human joint motion in vivo. The latest developments in this field include improvements in image quality, software for more efficient and accurate data processing, and the advent of mobile biplane videoradiography systems. Mobile systems enable data capture for a wider range of joints and activities by increasing the effective image capture volume, thereby addressing a major limitation of stationary systems. This chapter summarizes the most recent advances in human motion measurement using biplane videoradiography (also commonly referred to as biplane X-ray fluoroscopy). We begin with some basic considerations related to hardware setup, data capture, and data processing and then describe methods commonly used to evaluate system accuracy. The chapter concludes with a discussion of the relative merits of mobile versus stationary systems as well as some thoughts on potential future applications of biplane videoradiography in human joint motion measurement. H. Gray (*) • S. Guan • M. Pandy Department of Mechanical Engineering, The University of Melbourne, Parkville, VIC, Australia e-mail: [email protected]; [email protected]; [email protected] P. Loan C-Motion, Inc., Germantown, MD, USA e-mail: [email protected] # Springer International Publishing AG, part of Springer Nature 2018 B. Müller, S.I. Wolf (eds.), Handbook of Human Motion, https://doi.org/10.1007/978-3-319-14418-4_154

101

102

H. Gray et al.

Keywords

3D joint kinematics • Pose estimation • Six-degree-of-freedom joint motion • Biplane fluoroscopy • Mobile biplane fluoroscopy

Contents Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . State of the Art . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Hardware Configuration . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . X-Ray Unit Configuration . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Mobility of the X-Ray Imaging Systems . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Data Collection . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . X-Ray Generator Settings . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Frame Rate . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Exposure Time . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Motion Blur . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Data Processing . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Image Preparation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Calibration . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Pose Estimation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Validation of Measurement Accuracy . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Contrasting Mobile and Stationary Biplane Videoradiography Systems . . . . . . . . . . . . . . . . . . . . . . Conclusion and Future Directions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

102 103 104 104 105 105 106 106 107 107 107 107 108 108 111 112 113 113

Introduction Accurate measurement of in vivo joint kinematics is important for understanding normal and pathological human motion and for evaluating the outcome of surgical procedures (Dennis et al. 2005). Biplane videoradiography is currently the most accurate method available for measuring in vivo joint kinematics noninvasively. Two X-ray video imaging units are used to acquire time-synchronized sequences of two-dimensional (2D) X-ray images of a target joint from two different orientations (Fig. 1). Threedimensional (3D) geometric models of the individual bones or implant components comprising the joint are then registered to the 2D images in a 3D computer space through a process called “pose estimation.” Pose estimation defines the position and orientation of each of the bones or implants in a common reference frame and is used to calculate joint kinematics. Joint kinematics describes the relative positions and orientations of the bones using anatomically meaningful parameters (Grood and Suntay 1983). The aim of this chapter is to summarize recent advances in human motion measurement using biplane videoradiography (a technique that is also commonly referred to as biplane fluoroscopy). The next section titled “State of the Art” briefly summarizes the history and current state of biplane videoradiography. In the subsequent three sections, we follow a process flow-based structure focusing on issues related to the configuration of biplane systems, data collection, and data processing (see Fig. 2). Section “Validation of Measurement Accuracy” addresses matters concerning the accuracy with which biplane videoradiography systems are able to measure joint kinematics. Section “Contrasting Mobile and Stationary Biplane Videoradiography Systems” compares

Measurement of 3D Dynamic Joint Motion Using Biplane Videoradiography

103

Fig. 1 Design of a typical biplane videoradiography system. The X-ray tubes generate X-ray beams which pass through the joint of interest and enter the image intensifiers. Each image intensifier projects onto a phosphor screen an image that is subsequently captured by a camera mounted on the intensifier. The two cameras collect the images in a time-synchronized manner

Fig. 2 Schematic diagram illustrating the key steps involved with using biplane videoradiography to measure joint kinematics

and contrasts mobile and stationary biplane systems, while Section “Conclusion and Future Directions” concludes with some thoughts on the future of biplane videoradiography with respect to its potential applications in human motion measurement.

State of the Art The use of X-ray images to investigate the relative positions of bones, fractured segments of a bone, or foreign bodies dates back to the late nineteenth century following the discovery of X-ray imaging. The use of two X-ray images taken from two perspectives for precise measurement of the relative position of human tissue and foreign bodies emerged soon thereafter (Davidson 1898). Biplane X-ray images have

104

H. Gray et al.

been used to estimate 3D joint kinematics at least as early as the 1970s. These early studies were performed under static conditions and used either discrete bony landmarks (Matteri et al. 1976) or tantalum beads inserted in the bone (Selvik 1974; van Dijk et al. 1979) to calculate the pose (position and orientation) of the bone. The technique of using tantalum beads embedded in bones to precisely measure skeletal kinematics is called roentgen stereophotogrammetric analysis (RSA) and was introduced by Selvik (1974). Several early studies used this technique to measure joint kinematics under static conditions and slow dynamic conditions (Kärrholm et al. 1988; Uvehammer et al. 2000). More recent studies have used the technique to study more rapid dynamic movements (Anderst et al. 2009; Tashman et al. 2007). Although radiopaque markerbased pose estimation is still widely used (Anderst et al. 2008; Brainerd et al. 2010), its invasive nature makes it unsuitable for clinical studies involving large numbers of patients. In order to overcome limitations imposed by the need for beads, markerless techniques were developed to estimate pose from X-ray images for joint replacements (Banks and Hodge 1996; List et al. 2012) and for intact bone (Giphart et al. 2012; You et al. 2001; Zhu and Li 2011). Over the last two decades, refinements in sequential X-ray image acquisition systems have enabled higher quality images (with higher resolution and better contrast) acquired at higher frame rates. Improvements in pose estimation software and increased computing power at reduced costs have further expanded the use of biplane videoradiography systems in measuring 3D joint kinematics in vivo.

Hardware Configuration X-Ray Unit Configuration A videoradiography unit consists of an X-ray source and an image collector usually referred to as an image intensifier. The image capture volume of a single X-ray unit forms a cone with the X-ray source at the apex and the image intensifier forming the base. Because biplane videoradiography requires simultaneous imaging of a joint using two X-ray units, it is necessary for the joint to be within the intersection of the image capture volumes of both X-ray units during the activity of interest. There are several factors to consider when deciding on the configuration of the X-ray units, including anatomical direction used in imaging the joint, occlusion, and the size and shape of the capture volume. Anatomical Direction of Imaging and Occlusion: The anatomical direction of imaging has a bearing on the image quality. For example, anterior-posterior (AP) views may give clearer images of the hip joint than medial-lateral views due to reduced depth of bone and soft tissue in the AP direction. For the tibiofemoral joint, equally clear images may be obtained from the AP as well as the medial-lateral (ML) direction, whereas for the patellofemoral joint, the ML direction is preferred as it reduces occlusion by the femur. However, ML images of the knee will be occluded by the contralateral leg during walking (Fig. 3). Capture volume: The size, shape, and orientation of the image capture volume are critical as the joint must remain within the capture volume, which is relatively small.

Measurement of 3D Dynamic Joint Motion Using Biplane Videoradiography

105

Fig. 3 Samples of a series of biplane (Plane 1 and Plane 2) X-ray images of the knee of a total knee replacement patient collected during overground walking. Images taken at around 20% and 80% of the gait cycle were occluded by the contralateral leg for Plane 1. Similarly, images taken at around 30% and 70% were also occluded by the contralateral leg for Plane 2 but are not shown above

This feature of biplane X-ray systems limits the types of joints and activities that can be imaged. For example, measurement of knee kinematics during walking has been limited to a portion of the gait cycle such as the stance phase, and most studies of lower-limb joint kinematics during locomotion have been confined to treadmill gait (Anderst et al. 2009).

Mobility of the X-Ray Imaging Systems Mobile biplane X-ray systems offer concurrent tracking and imaging of dynamic joint motion during human activity, which overcomes the limitations imposed by the small capture volume inherent to stationary systems. A mobile biplane X-ray (MoBiX) imaging system developed at the University of Melbourne translates each of the two X-ray units in a vertical plane so that the target joint remains within the image capture volume during movement (Fig. 4). The ability to track the motion of a target joint increases the size of the effective image capture volume of the system and enables the kinematics of various joints such as the knee, hip, and shoulder to be studied for a wide range of activities, including overground and treadmill walking, lunging, squatting, jumping, and stair ambulation. Section “Contrasting Mobile and Stationary Biplane Videoradiography Systems” below presents a more comprehensive discussion of the advantages and disadvantages of mobile and stationary biplane videoradiography systems.

Data Collection Key factors that must be considered during data collection include X-ray generator settings, specifically, tube current and voltage, and camera settings such as frame rate and exposure time. These settings are important not only because they collectively influence the quality of the images captured but also because they affect the radiation dosage received by the subject.

106

H. Gray et al.

Fig. 4 A mobile biplane X-ray (MoBiX) imaging system developed at the University of Melbourne comprising of two X-ray units mounted on a custom robotic gantry mechanism. Each X-ray unit comprising an X-ray tube and an image intensifier is mounted on the robotic arms that translate vertically along the mobile columns. The mobile columns translate along the horizontal guides positioned on either side of the walkway. The X-ray units translate both horizontally and vertically, thus increasing the effective image capture volume

X-Ray Generator Settings Appropriate values of X-ray tube current and voltage must be selected to produce images of sufficient brightness and contrast. In addition, the X-ray beam may be operated in either continuous or pulse mode. These parameters determine the subject’s exposure to ionizing radiation and are therefore carefully scrutinized by ethics committees to ensure the benefits outweigh the risks. A wide range of X-ray settings have been used for studies reported in the literature. In studies involving the knee joint, for example, Giphart et al. (2012) used a continuous voltage and current of 60 kV and 60 mA, respectively, while Anderst et al. (2009) used a voltage and current of 90 kV and 100 mA, respectively.

Frame Rate The upper limit in the frame rate of commercially available videoradiography systems is typically 30 Hz. However, in order to capture detailed 3D joint kinematics, many biplane videoradiography systems have been fitted with high-speed cameras capable of imaging at much higher frame rates, for example, the systems described by Guan et al. (2016) and Ivester et al. (2015) are capable of maximum frame rates of 1000 Hz. In general, faster activities require higher frame rates to

Measurement of 3D Dynamic Joint Motion Using Biplane Videoradiography

107

capture more detailed kinematic measurements. Frame rates used in recent studies involving biplane videoradiography systems have typically ranged from 30 Hz to 500 Hz (e.g., Li et al. 2009; Myers et al. 2012; Tashman et al. 2004).

Exposure Time Exposure time or shutter speed (measured in s) is the duration the camera sensor is exposed to light for a single image frame. Shorter exposure times produce sharper images as they reduce motion blur. However, longer exposure times allow more light to reach the sensor and therefore make the images brighter and have potential to improve image contrast. Exposure times in recent studies involving biplane videoradiography systems have ranged from 0.5 ms to 8 ms (e.g., Li et al. 2009; Myers et al. 2012).

Motion Blur Motion blur is caused by relative motion between the imaging system and the object and is proportional to the distance the object moves relative to the imaging system in the imaging plane during the exposure time. Therefore, motion blur is proportional to both exposure time and the relative velocity between the object and the imaging system. Two approaches have been used to reduce motion blur. Mobile biplane videoradiography systems reduce motion blur by reducing the relative velocity between the object and the imaging system. Other systems reduce motion blur by reducing the exposure time. Reducing exposure time reduces the light falling on the image sensor causing the images to be dark. This problem can be overcome by employing a higher X-ray current to maintain sufficient image brightness and contrast leading to an increase in the X-ray dosage received by the subject. However, some systems are able to pulse the X-ray beam synchronously with camera exposure time ensuring that the X-ray beam is only emitted by the X-ray generator when the camera shutter is open. This enables sufficiently bright images with little blur to be obtained while only exposing the subject to radiation when the joint is being imaged (Ivester et al. 2015; Tashman 2016).

Data Processing Image Preparation Raw images obtained from X-ray equipment are processed prior to pose estimation by porting the images through a pipeline comprised of flat field correction, distortion correction, and other image enhancement processors (Fig. 5). Flat field correction is performed to compensate for the differences in sensitivity between pixels in the

108

H. Gray et al.

Fig. 5 An example of an image processing pipeline comprising flat field correction, distortion correction, and feature enhancement

camera sensor. X-ray image distortion occurs within the image intensifier and the lens of the camera and may affect the measurement accuracy of the system. An image of an object embedded with an array of equispaced radiopaque beads is used to model the distortion as polynomial functions. These functions are then used to correct the distortion of the X-ray images (Garling et al. 2005). Further image processing is often performed to reduce noise and blur and to enhance features such as edges on the X-ray images.

Calibration Calibration of a biplane videoradiography system involves determining the geometric configuration of the X-ray sources and the image intensifiers. This procedure is accomplished by using the biplane system to image a calibration object containing several radiopaque beads at precisely known locations. The positions of the beads on the x-ray images are then used to calculate the geometric configurations of the imaging units (Kaptein et al. 2011).

Pose Estimation Pose estimation is the process of calculating the three translational and three rotational parameters needed to fully describe the position and orientation of each bone (or implant) in a common reference frame. This procedure is usually accomplished by solving an optimization problem that minimizes a scalar cost function. The unknown variables in this optimization problem are the six independent variables that define the pose of a geometric model of the bone or implant. The cost is formulated as a function of the six independent variables and reflects the accuracy with which the computed projections of the model can be superimposed onto the biplane X-ray images recorded in each image plane. An initial guess for the six variables is provided, and the cost function is minimized to calculate the pose. Various possibilities exist for the formulation of the cost function, for example, an edge-based method where root mean squared (RMS) distances between the detected edges on the

Measurement of 3D Dynamic Joint Motion Using Biplane Videoradiography

109

Fig. 6 DRR generation using ray tracing. For each view, a ray is traced from the X-ray source (green and red spheres) to every pixel in the simulated X-ray image. The value assigned to the pixel is the weighted sum of the voxel intensities of the bone models through which the ray passes. Each voxel is weighted by the distance travelled through it by the ray

X-ray images and the projected edges of the geometric model have been used by some researchers (Bingham and Li 2006; Guan et al. 2016), while others have used the correlation between X-ray images and digitally reconstructed radiographs (Bey et al. 2006; Ohnishi et al. 2010). Digitally reconstructed radiographs (DRRs) are generated by simulating the X-ray imaging process. A ray is traced from the X-ray source to each pixel on the imaging plane. The sum of intensities of the voxels in the bone model that a ray passes through is assigned to the respective pixel in the DRR (Fig. 6). To create the geometric model of an implant, a CAD model can usually be obtained from the manufacturer. For bones, segmented image data and 3D surface models are typically reconstructed from CT images. There are many commercial and open-source software packages available for segmenting individual bones from CT images and generating smooth polyhedral surface models (e.g., Mimics (Materialise N.V., Leuven, Belgium), 3D Slicer (Fedorov et al. 2016)). The segmented CT voxel data are used to create the DRRs. Pose Optimization: There are two general methods of optimizing the poses of bones or implants in biplane X-ray images. Descent-direction optimizers, such as the Levenberg-Marquardt algorithm, use the gradient of the cost function at the current pose to iteratively calculate a subsequent pose with a smaller value of cost function. The challenge with these methods is that they find the local minimum relative to the initial pose, which is not necessarily the optimal solution. Global optimizers, such as simulated annealing, use deterministic or stochastic methods to search the entire solution space for the global minimum of the cost function. The challenge with these algorithms is that they are computationally expensive and the value of the cost function at the global minimum is not known in advance. Thus it is not known when the algorithm should terminate, so choosing the number of iterations involves a

110

H. Gray et al.

balance between finding the optimal pose and finishing in a reasonable amount of time. Both of these challenges can be mitigated by providing an initial pose that is close to the optimal solution. Manually positioning the bone using a graphical user interface for each time frame can be a time-consuming and tedious process. A more efficient technique is to use the optimal pose from one or more previously solved frames to calculate an initial pose for the next frame. This procedure usually requires manual pose estimation to be performed only for the first frame or two. Another solution to this problem is to use traditional motion capture (surface markers, markerless, or inertial measurement units) to measure the approximate bone poses during the activity and to use these as the initial poses during optimization (Bone tracking software, C-Motion, MD, USA). 4D Tracking: An alternative to frame-by-frame pose estimation as described above is to optimize the pose for several time frames simultaneously, a method which we refer to here as 4D tracking (Tracking software, C-Motion, MD, USA). This process involves fitting splines to the six degrees of freedom of the initial bone poses calculated from the traditional motion capture data, with nodes at specified intervals (e.g., every fifth time frame), and treating the nodes as the independent variables during optimization. In each iteration, the optimization algorithm generates a set of node values and calculates the corresponding value of the cost function. Specifically, the algorithm generates DRRs for every X-ray image in the motion sequence for both views. The pose of the bone for each DRR is calculated by evaluating the node splines at the time of the X-ray image corresponding to that DRR. Each DRR is then compared to its X-ray image, and the correlation metrics for all DRR/X-ray pairs are summed to generate the value of the cost function for that set of node values. Natural cubic splines are used to interpolate the nodes, with a userdefined low-pass cutoff frequency for smoothing. 4D tracking has two primary advantages over a sequential process in which each time frame is solved independently. First, 4D tracking exploits temporal coherence in the data, which guarantees a smooth transition between successive poses and enforces reasonable motion physics. The smoothness of the splines and the spacing of the nodes can be controlled by the user and are usually matched to the activity being measured. Typical low-pass cutoff frequencies are 10 Hz for walking and 20 Hz for running. The second advantage of 4D tracking relates to the timing of the X-ray devices. Most current biplane videoradiography systems are capable of synchronous recording of the two X-ray views. With synchronous recording, the X-ray emitters are activated at the same time, and the two images are recorded simultaneously. This synchronicity eases the burden of pose estimation because it enables independent optimization for each time frame reducing the number of independent variables associated with 4D tracking. However, synchronous activation of the X-ray devices introduces additional noise into the system. When one X-ray beam intersects soft tissue or bone, some of its X-rays scatter and hit the image intensifier of the other view. When imaging parts of the body with small bones and a large amount of soft tissue, such as the lumbar spine, X-ray scatter can be a significant source of noise. With 4D tracking, the node splines can be evaluated at any time between the first and last frames (Fig. 7), and the optimization process can proceed whether the X-ray images are synchronous or asynchronous.

Measurement of 3D Dynamic Joint Motion Using Biplane Videoradiography

111

60.0

50.0

angle (deg.)

40.0

30.0

20.0

10.0

0.0 0.28

0.29

0.30

0.31 time (sec.)

0.32

0.33

Fig. 7 Example illustrating spline fitting one of the angles expressing the poses of an object for 4D tracking with asynchronous X-ray data. The green (X-ray Plane 1) and red (X-ray Plane 2) lines represent the frame times for the X-ray views. The circles represent the control points (nodes) of a spline, which is smoothed using a user-specified cutoff frequency. The spline is interpolated at each time frame (green and red diamonds) to evaluate the bone poses while tracking

Validation of Measurement Accuracy The errors involved in measuring 3D joint motion are dependent on the system as well as the target joint, activity, and the speed at which the task is performed. In order to evaluate measurement errors, the kinematic measurements obtained from the system need to be compared against a “gold standard” method. The most widely accepted procedure for obtaining accurate kinematic measurements involves embedding X-ray opaque beads in the bones and using their positions in the X-ray images to quantify joint kinematics. The ideal method for determining measurement accuracy of a biplane videoradiography system is described below (Fig. 8): 1. Insert at least three radiopaque beads in each of the bones of the target joint of a living person. The beads should be positioned nonlinearly and near enough to the target joint so that they are clearly visible in the X-ray images. 2. Collect biplane videoradiography images of the joint while the subject performs the activity of interest. 3. Calculate the joint kinematics using the locations of the beads on the X-ray images as described by You et al. (2001). 4. Calculate joint kinematics from biplane X-ray images using a markerless pose estimation method. 5. Compare the kinematics obtained from the markerless pose estimation method in step 4 with those obtained from the more accurate marker-based method in step 3.

112

H. Gray et al.

Fig. 8 Procedure involved in quantifying the measurement accuracy of a biplane videoradiography system (See text for details)

The aforementioned procedure for estimating measurement errors is invasive and thus rarely used. Several alternative variations to this method have been reported that involve inserting radiopaque beads into human cadaver joints, animal cadaver joints, live animal joints, and inanimate objects. Studies using human or animal cadaver joints are either conducted under static conditions or at significantly lower speeds than the activity intended to be studied (Ohnishi et al. 2010). Studies using live animals provide more realistic joint kinematics and speeds although differences in the 3D geometric shapes and sizes between the animal bones and human bones may make the results less convincing. Dynamic joint motion simulators (DJMS) overcome the above limitations as they enable validation studies to be performed on human cadaver joints under dynamic conditions which closely replicate the conditions present in vivo. For example, Guan et al. (2016) used a DJMS with an intact human cadaver knee to simulate joint motion during normal walking over ground.

Contrasting Mobile and Stationary Biplane Videoradiography Systems Perhaps most significantly, mobile biplane videoradiography systems offer a much larger effective image capture volume compared to stationary systems, enabling the study of a wider range of joints and activities. Another advantage of simultaneously tracking and imaging a joint is the potential reduction in radiation dosage to the subject. Radiation dosage, motion blur, and camera exposure time are interrelated as described earlier in this chapter. Because the MoBiX tracks the joint while imaging it, the relative motion between the joint and the imaging system is reduced, enabling

Measurement of 3D Dynamic Joint Motion Using Biplane Videoradiography

113

longer exposure times while keeping motion blur to an acceptable level. This feature enables X-ray currents to be reduced while still maintaining sufficiently bright images. However, mobile biplane systems are not without their limitations. They are more difficult and more costly to design and build than their stationary counterpart. Furthermore, the moving parts in a mobile system are subject to wear and tear and require regular maintenance. Misalignment of the robotic actuator guideways during installation and vibrations during operation both have the potential to change the relative positions of the X-ray tubes and image intensifiers away from their configuration at the time of system calibration. This problem can be minimized through precision manufacture and installation, using components with high rigidity and minimizing vibrations through improved joint tracking algorithms (Guan et al. 2016).

Conclusion and Future Directions Biplane videoradiography is currently the most accurate method for noninvasive measurement of dynamic in vivo joint kinematics during human movement. This chapter summarizes the basic concepts and considerations in configuring a biplane fluoroscopy system and accurately capturing and processing the X-ray image outputs. It also discusses some aspects related to validation of system accuracy and compares and contrasts stationary and mobile systems. The authors believe that the near future will see significant developments in hardware and software in a more integrated and streamlined manner making such systems and data processing more user-friendly and thus more accessible to clinicians and healthcare specialists. The accuracy of such systems may also continue to improve with the development of improved imaging systems with higher image quality in contrast and resolution. The development of more advanced CT scanners and videoradiography units capable of using lower radiation dosage may make these systems safer and thus suitable for studying a wider range of subject populations, including young children. It is hoped that these devices may be used ultimately as diagnostic tools for designing patientspecific treatment regimens for patients with various musculoskeletal conditions.

References Anderst W, Zauel R, Bishop J, Demps E, Tashman S (2009) Validation of three-dimensional modelbased tibio-femoral tracking during running. Med Eng Phys 31:10–16. https://doi.org/10.1016/j. medengphy.2008.03.003 Anderst WJ, Vaidya R, Tashman S (2008) A technique to measure three-dimensional in vivo rotation of fused and adjacent lumbar vertebrae. Spine J 8:991–997. https://doi.org/10.1016/j. spinee.2007.07.390 Banks SA, Hodge WA (1996) Accurate measurement of three dimensional knee replacement kinematics using single-plane fluoroscopy. IEEE Trans Biomed Eng 46:638–649 Bey MJ, Zauel R, Brock SK, Tashman S (2006) Validation of a new model-based tracking technique for measuring three-dimensional, in vivo glenohumeral joint kinematics. J Biomech Eng 128:604–609. https://doi.org/10.1115/1.2206199

114

H. Gray et al.

Bingham J, Li G (2006) An optimized image matching method for determining in-vivo TKA kinematics with a dual-orthogonal fluoroscopic imaging system. J Biomech Eng 128:588–595. https://doi.org/10.1115/1.2205865 Brainerd EL, Baier DB, Gatesy SM, Hedrick TL, Metzger KA, Gilbert SL, Crisco JJ (2010) X-ray reconstruction of moving morphology (XROMM): precision, accuracy and applications in comparative biomechanics research. J Exp Zool A Ecol Genet Physiol 313:262–279. https://doi.org/ 10.1002/jez.589 Davidson JM (1898) Roentgen rays and localisation: an apparatus for exact measurement and localisation by means of roentgen rays. Br Med J 1:10–13 Dennis DA, Mahfouz MR, Komistek RD, Hoff W (2005) In vivo determination of normal and anterior cruciate ligament-deficient knee kinematics. J Biomech 38:241–253. https://doi.org/ 10.1016/j.jbiomech.2004.02.042 Fedorov A, Beichel R, Kalpathy-Cramer J, Finet J, Fillion-Robin J-C, Pujol S, Bauer C, Jennings D, Fennessy F, Sonka M, Buatti J, Aylward S, Miller JV, Pieper S, Kikinis R (2016) 3D slicer as an image computing platform for the quantitative imaging network. Magn Reson Imaging 30:1323–1341. https://doi.org/10.1016/j.mri.2012.05.001 Garling EH, Kaptein BL, Geleijns K, Nelissen RGHH, Valstar ER (2005) Marker configuration model-based roentgen fluoroscopic analysis. J Biomech 38:893–901. https://doi.org/10.1016/j. jbiomech.2004.04.026 Giphart JE, Zirker CA, Myers CA, Pennington WW, LaPrade RF (2012) Accuracy of a contourbased biplane fluoroscopy technique for tracking knee joint kinematics of different speeds. J Biomech 45:2935–2938. https://doi.org/10.1016/j.jbiomech.2012.08.045 Grood ES, Suntay WJ (1983) A joint coordinate system for the clinical description of threedimensional motions: application to the knee. J Biomech Eng 105:136–144 Guan S, Gray HA, Keynejad F, Pandy MG (2016) Mobile biplane X-ray imaging system for measuring 3D dynamic joint motion during overground gait. Med. Imaging IEEE Trans 35:326–336. https://doi.org/10.1109/TMI.2015.2473168 Ivester JC, Cyr AJ, Harris MD, Kulis MJ, Rullkoetter PJ, Shelburne KB (2015) A reconfigurable high-speed stereo-radiography system for sub-millimeter measurement of in vivo joint kinematics. J Med Device 9:41009. https://doi.org/10.1115/1.4030778 Kaptein BL, Shelburne KB, Torry MR, Giphart JE (2011) A comparison of calibration methods for stereo fluoroscopic imaging systems. J Biomech 44:2511–2515. https://doi.org/10.1016/j. jbiomech.2011.07.001 Kärrholm J, Selvik G, Elmqvist L-G, Hansson LI (1988) Active knee motion after cruciate ligament rupture. Acta Orthop Scand 59:158–164. https://doi.org/10.1080/17453678809169699 Li G, Kozanek M, Hosseini A, Liu F, Van de Velde SK, Rubash HE (2009) New fluoroscopic imaging technique for investigation of 6DOF knee kinematics during treadmill gait. J Orthop Surg Res 4:6. https://doi.org/10.1186/1749-799X-4-6 List R, Foresti M, Gerber H, Goldhahn J, Rippstein P, Stussi E (2012) Three-dimensional kinematics of an unconstrained ankle arthroplasty: a preliminary in vivo videofluoroscopic feasibility study. Foot Ankle Int 33:883–892. https://doi.org/10.3113/FAI.2012.0883 Matteri RE, Pope MH, Frymoyer JW (1976) A biplane radiographic method of determining vertebral rotation in postmortem specimens. Clin Orthop Relat Res 116 Myers C a, Torry MR, Shelburne KB, Giphart JE, LaPrade RF, Woo SL-Y, Steadman JR (2012) In vivo tibiofemoral kinematics during four functional tasks of increasing demand using biplane fluoroscopy. Am J Sports Med 40:170–178. https://doi.org/10.1177/03635465 11423746 Ohnishi T, Suzuki M, Nawata A, Naomoto S, Iwasaki T, Haneishi H (2010) Three-dimensional motion study of femur, tibia, and patella at the knee joint from bi-plane fluoroscopy and CT images. Radiol Phys Technol 3:151–158. https://doi.org/10.1007/s12194-010-0090-1 Selvik G (1974) Roentgen stereophotogrammetry: a method for the study of the kinematics of the skeletal system. University of Lund, Lund. https://doi.org/10.3109/17453678909154184

Measurement of 3D Dynamic Joint Motion Using Biplane Videoradiography

115

Tashman S (2016) Comments on “validation of a non-invasive fluoroscopic imaging technique for the measurement of dynamic knee joint motion”. J Biomech 41:3290–3291. https://doi.org/ 10.1016/j.jbiomech.2008.07.038 Tashman S, Collon D, Anderson K, Kolowich P, Anderst W (2004) Abnormal rotational knee motion during running after anterior cruciate ligament reconstruction. Am J Sports Med 32:975–983. https://doi.org/10.1177/0363546503261709 Tashman S, Kolowich P, Collon D, Anderson K, Anderst W (2007) Dynamic function of the ACL-reconstructed knee during running. Clin Orthop Relat Res 454:66–73. https://doi.org/ 10.1097/BLO.0b013e31802bab3e Uvehammer J, Karrholm J, Brandsson S, Herberts P et al (2000) In vivo kinematics of total knee arthroplasty: flat compared with concave tibial joint surface. J Orthop Res 18:856–864 van Dijk R, Huiskes R, Selvik G (1979) Roentgen stereophotogrammetric methods for the evaluation of the three dimensional kinematic behaviour and cruciate ligament length patterns of the human knee joint. J Biomech 12:727–731. https://doi.org/10.1016/0021-9290(79)90021-6 You BM, Siy P, Anderst W, Tashman S (2001) In vivo measurement of 3D skeletal kinematics from sequences of biplane radiographs: application to knee kinematics. IEEE Trans Med Imaging 20:514–525. https://doi.org/10.1109/42.929617 Zhu Z, Li G (2011) An automatic 2D–3D image matching method for reproducing spatial knee joint positions using single or dual fluoroscopic images. Comput Methods Biomech Biomed Engin:1–12. https://doi.org/10.1080/10255842.2011.597387

3D Musculoskeletal Kinematics Using Dynamic MRI Frances T. Sheehan and Richard M. Smith

Abstract

Until the early 1990s, the tools available to measure musculoskeletal motion were typically highly invasive. Thus, knowledge of musculoskeletal system dynamics was primarily derived through cadaver and modeling experiments. The rapid development of dynamic magnetic resonance (MR) imaging techniques changed this and opened vast new opportunities for the study of 3D musculoskeletal dynamics during volitional activities. Today, dynamic MR methodologies remain the only techniques that can noninvasively track in vivo 3D musculoskeletal movement. One difficulty in applying these dynamic MR techniques to the study of musculoskeletal motion is the complex interplay of parameters that affect the spatial/temporal resolution, accuracy, and precision. The purpose of this chapter is to first provide an explanation of the fundamental principles behind two of these dynamic imaging techniques, cine and cine phase-contrast MR. Tagged cine MR is another technique that has been primarily used to track muscle motion and strain but will not be addressed. In doing so, this will create a platform for future experimental designs using dynamic MR. This will be followed by a review of the accuracies, the advantages, and disadvantages of the these dynamic MR methods. Finally, several previously published studies will be highlighted to provide an explanation of how these techniques can be applied and what main challenges must be considered for future experiments using dynamic MR.

F.T. Sheehan (*) • R.M. Smith Rehabilitation Medicine Department, Functional and Applied Biomechanics Section, National Institutes of Health, Bethesda, MD, USA e-mail: [email protected]; [email protected]; [email protected] # This is a U.S. Government work and not under copyright protection in the US; foreign copyright protection may apply 2018 B. Müller, S.I. Wolf (eds.), Handbook of Human Motion, https://doi.org/10.1007/978-3-319-14418-4_155

117

118

F.T. Sheehan and R.M. Smith

Keywords

MR • Magnetic resonance imaging • Cine • Cine phase contrast • Fastcard • CPC • Fast-PC • Musculoskeletal • Kinematics • Strain • Moment arms • Tendon

Contents Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . State of the Art . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Static MR Imaging . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Dynamic MR Imaging . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Cine MR . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Fastcard (Fast Cine) MR . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Fast Cine: Phase-Contrast (Fast-PC) MR . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Advanced Applications of Dynamic MR Imaging . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Combining Fastcard Imaging with 3D Modeling . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Applications of CPC and Fast-PC . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Future Directions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Cross-References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

118 119 119 121 121 122 123 126 126 128 128 129 130

Introduction Until the early 1990s, the tools available to measure musculoskeletal motion were typically highly invasive (Regev et al. 2011; Edsfeldt et al. 2015; Lafortune et al. 1994; van Kampen and Huiskes 1990; Manal et al. 2000). Thus, knowledge of musculoskeletal system dynamics was primarily derived through cadaver and modeling experiments. The rapid development of dynamic magnetic resonance (MR) imaging techniques changed this and opened vast new opportunities for the study of 3D musculoskeletal dynamics during volitional activities (Burnett et al. 1987; Drace and Pelc 1994; Sheehan et al. 1998; Sheehan and Drace 2000; Asakawa et al. 2002; Pappas et al. 2002; Fujiwara et al. 2004; Finni et al. 2006; Cheng et al. 2008). Today, dynamic MR methodologies remain the only techniques that can noninvasively track in vivo 3D musculoskeletal movement. One difficulty in applying these dynamic MR techniques to the study of musculoskeletal motion is the complex interplay of parameters that affect the spatial/ temporal resolution, accuracy, and precision. The purpose of this chapter is to first provide an explanation of the fundamental principles behind two of these dynamic imaging techniques, cine (Glover and Pelc 1988) and cine phase-contrast (Feinberg et al. 1984; Wedeen et al. 1985; Pelc et al. 1991b, 1994) MR. Tagged cine MR is another technique that has been primarily used to track muscle motion and strain but will not be addressed (Moerman et al. 2012). In doing so, this will create a platform for future experimental designs using dynamic MR. This will be followed by a review of the accuracies, the advantages, and disadvantages of the these dynamic MR methods. Finally, several previously published studies will be highlighted to provide an explanation of how these techniques can be applied and what main challenges must be considered for future experiments using dynamic MR.

3D Musculoskeletal Kinematics Using Dynamic MRI

119

State of the Art The earliest use of MR technology was isolated to spectroscopy (NRM or nuclear magnetic resonance). Once researchers realized how to manipulate the MR signal to produce in vivo images of the human tissue, the technology was rapidly transferred to the clinical domain in the early 1980s, providing a wealth of new information. A primary drawback of this new technology was the acquisition time for a single image prohibited its use in imaging moving tissue, particularly the heart. Cine MR was developed to acquire anatomic images of moving tissue during a cyclic movement sequence by synchronizing the data collection to the cardiac cycle. This was rapidly followed by the development of tagged MR and cine phase-contrast (CPC) MR, which allowed the analytical tracking of tissue. Tagged MR has remained a tool primarily used for cardiac imaging, but both cine and CPC MR have been widely applied to the noninvasive study of 3D musculoskeletal kinematics. The original applications of cine and CPC MR to the study of musculoskeletal motions required long acquisition times with a high number of repeated cycles. Currently, the accuracy of using CPC to track musculoskeletal motion has dropped to less than 0.3 mm. In addition, the imaging times have dramatically dropped, enabling acquisitions of less than a minute.

Static MR Imaging MR imaging is distinct from other imaging techniques in two key ways. First, the measured signal is generated from the tissue being imaged. MR imaging is based on the principle that atoms with an odd number of neutrons and protons will precess about an external magnetic field at a frequency directly proportional to the field strength. Clinical MR imaging typically focusses on hydrogen atoms, often referred to as “spins.” With the introduction of a perpendicular magnetic field at the same frequency of precession, the spins will “tip” into the transverse plane, creating a transverse magnetization. The MR signal is the transverse magnetization of all spins within a voxel (the smallest volumetric unit of the image). As the signal for all voxels is collected together in a single summed signal, the location of a voxel’s signal is encoded in the phase and frequency of its signal using magnetic gradients (Riederer 1993). This encoding process results in another unique feature of MR imaging; the data collected during a scan is not spatial but spatial frequency data (k-space data, Fig. 1). More importantly, numerous data acquisitions are required to provide a complete spatial frequency map (k-space map) that can be converted into an image using inverse Fourier transforms. Various methods for collecting data in k-space have been developed (Foo et al. 1995; Markl et al. 2003; Pike et al. 1994; Thompson and McVeigh 2004; Thunberg et al. 2003), but the simplest, collecting a single line in k-space (phase encode) per acquisition (Fig. 2), will be used as the example for this chapter.

120

F.T. Sheehan and R.M. Smith

Fig. 1 K-space map to MR image. K-space map (left) of a 2D GRE sagittal image of the knee (right). The majority of the data in the k-space map resides at the lower frequencies, which represent the portions of the image where the contrast is not changing rapidly from one pixel to the next. For example, the magnitude at kx = ky = 0 represent the parts of the 2D spatial image where the contrast is not changing from pixel to pixel. One such region of low spatial frequency is shown on the left by the number 1. The number 2 denotes a region of high spatial frequency in the anteriorposterior direction

Fig. 2 K-space map representing a linear phase encode collection: in this example, the frequency direction is kx (anteriorposterior) and the phase direction is ky (superiorinferior). For visual clarity, the acquisition is represented with very low frequency and phase resolution (41
41). For this example, 41 individual phase encodes (each represented by a unique color) are required to produce one 2D spatial image

3D Musculoskeletal Kinematics Using Dynamic MRI

121

The speed at which a single phase encode can be captured is referred to as the repetition rate (TR). Thus, the time needed to acquire an entire image is: Imaging time ¼ #phases  TR

(1)

#phases = number of required phase encodes to create a single image If data averaging is used, to reduce the effect of random noise, then the imaging times increase linearly with the number of averages (NA) acquired: Imaging time ¼ NA  #phases  TR

(2)

As noted above (Eqs. 1 and 2), the imaging time is dependent on the number of phase encodes acquired. For a fixed field of view, #phases directly determines the spatial resolution of the final image. To put numbers to this, let us assume a 256
256 (frequency resolution
#phases) acquisition. This would result in a square image 256 pixels wide by 256 high. Assuming a TR = 5 ms and no data averaging, the imaging time (Eq. 2) would be 1.28 s for a single 2D image. If we set the spatial field of view at 200
200 mm, the spatial resolution becomes (200 mm/ 256) = 0.78 mm/pixel.

Dynamic MR Imaging Cine MR When the first scanners became commercially available in the early 1980s, it was not feasible to image the beating heart, because its period of motion was significantly greater than the image acquisition time. As noted in the example above for a 256
256 acquisition with a TR of 5 ms, the imaging time (1.28 s) is longer than the typical cardiac period of 1 s. A TR of 5 ms has only been feasible recently, when cine MR was first being developed TRs above 20 ms were typical (Keegan et al. 1994; Sheehan et al. 1998). Cine MR (Waterton et al. 1985; Burnett et al. 1987; Glover and Pelc 1988; Cadera et al. 1992) was developed to overcome this limitation, with a particular focus on cardiac imaging. The underlying assumption of cine MR is that the tissue being imaged is moving in a repeatable, cyclic motion path. By gating the data collection to the cardiac cycle, the required phase encodes in k-space (256 in our example) can be acquired over numerous cardiac cycles. Retrospective gating with cine MR compensates for variations in the period of motion by repeatedly acquiring the same k-space data line every TR during a single cycle. A trigger is used to detect when a new cycle begins, signaling a change in the k-space data line being acquired (Fig. 3). After #phases cycles, the data are retrospectively interpolated so that images are created at specified time intervals. Although any number of image time frames can be created through the interpolation process, the true temporal resolution is TR, whereas, the spatial resolution is dependent on #phases:

122

F.T. Sheehan and R.M. Smith

Fig. 3 Cine MR data capture: when the scanner receives a trigger that a new cycle has begun (black line on red signal), the ky data line (phase encode) being acquired is incremented. In the above example, all cycles are of even length and a true 16 frames is captured. The temporal resolution is equal to the TR (time between data captures). Compensating for variations in the periods of each cycle is accomplished with retrospective gating and interpolation to align the data to the correct temporal location in the motion cycle

cine temporal resolution ¼ TR

(3)

spatial resolution ¼ FOV=#phases

(4)

Thus, with a TR of 5 ms, 200 true frames of data are acquired with a period of 1 s. Other types of gating, including prospective, are available for cine MR. One issue with prospective gating is that the entire cycle is typically not captured. The imaging time for cine MR depends on the #phases and motion period (T) of the moving tissue being imaged: cine imaging time ¼ T  #phases

(5)

Following the example above, and assuming T = 1 s, the cine imaging time = 1 s*256 = 4.3 min (Eq. 5), requiring 256 cycles of motion. Although the cine techniques were motivated by a desire to image cardiac motion, its applicability to musculoskeletal motion was quickly realized. In 1987, a study using cine MR to visualize passive temporomandibular joint mechanics (Burnett et al. 1987) was published. This was quickly followed by a qualitative study of active ankle joint motion (Melchert et al. 1992) and a quantitative study of active knee joint motion (Brossmann et al. 1993). More recently, cine MR has been used to evaluate tongue movement (Stone et al. 2001), fetal motion (Verbruggen et al. 2016), and scaphoid-lunate motion (Langner et al. 2015).

Fastcard (Fast Cine) MR Fastcard (Foo et al. 1995), or fast cine MR, was developed to enable a trade-off between temporal resolution and the number of cycles required to capture a full image set. Instead of repeatedly capturing a single phase encode line per motion

3D Musculoskeletal Kinematics Using Dynamic MRI

123

cycle, fastcard collects multiple phase encodes during a single motion cycle. This reduces the overall imaging time, with a degradation in temporal resolution: fastcard imaging time ¼ T  #phases=#views

(6)

fastcard temporal resolution ¼ #views  TR

(7)

#views = the number of unique phase encode lines acquired during a single motion cycle For most joint motion studies using cine MR, the period of motion is longer than that of a beating heart. Using the above example, but assuming a typical knee joint extension/flexion type movement (Brossmann et al. 1993), with a motion period of 2 s, the 256 motion cycles would require 8.5 min to capture a cine MR image set. Yet, 400 temporal frames (a temporal resolution of 5 ms) would be produced to represent an arc of motion of approximately 40 . Using fastcard and assuming 20 views are collected during each motion cycle, the fastcard imaging time is just 26 s, requiring only 13 motion cycles. Here, only 20 true data frames are collected, with a temporal resolution equal to 100 ms. This trade-off between temporal resolution and number of required motion cycles afforded by fastcard is fundamental to the design of experiments that can be repeatedly performed by both healthy volunteers and individuals with musculoskeletal impairments/pathologies. Fastcard also enables a trade-off between motion period and the number of required motion cycles. Using the same example, if the period is increased by a factor of 2 (T = 4), the number of views acquired during each motion cycle could be doubled (40 views). Although the temporal resolution would increase by a factor of 2, the speed of motion would reduce by the same amount. As such, no increase in temporal blurring would be expected. This would reduce the number of required motion cycles to just 7, but the overall acquisition time would still remain at ~26 s. It is important to note that this example is assuming an acquisition matrix of 256
256. Using the spatial parameters from a previous study (Carlson et al. 2016) of patellofemoral kinematics (FOV = 180 mm and using a 75% phase acquisition), the acquisition matrix can be dropped to 180
135 pixels and still maintain a resolution of 1
1 mm. This would require just four motion cycles with an imaging time of 16 s. Such a reduced number of required cycles could potentially allow for studies with increased joint loading and is crucial to acquiring data during volitional joint motion in individuals with musculoskeletal pathologies/impairments (Fig. 4).

Fast Cine: Phase-Contrast (Fast-PC) MR CPC MR (Pelc et al. 1991b) combines the ability of phase-contrast imaging to produce three-dimensional quantitative velocity data with cine’s ability to produce a series of images throughout a gated motion cycle. Fast-PC MR (Foo et al. 1995) is identical to CPC MR but inherits fastcard’s ability to capture multiple views, or phase encode lines, during a single motion/cardiac cycle. Phase-contrast imaging

124

F.T. Sheehan and R.M. Smith

Fig. 4 Fastcard data acquisition: for this figure, five phase encode lines (#views = 5) are captured per motion cycle. Thus, phase encode lines ky = 16 through ky = 20 are captured during the first motion cycle. Temporal resolution is increased to 5*TR (the time between acquiring the same k-space line is 5*TR). Yet, on the positive side, the imaging time and the number of cycles required are also reduced by a factor of 5

Fig. 5 Acquiring fast-PC data. To capture the full fast-PC dataset, four separate data acquisitions are needed. This provides measures of velocity in the frequency (x), phase (y), and slice (z) directions. The temporal resolution is reduced by a factor of 4 (temporal resolution = 4 * TR) and 5 true frames of data are available. If the #views was increased from 1 to 2, then the temporal resolution would increase to 8*TR, with a reduction by a factor of 2 in the number of motion cycles and the imaging time

(Pelc 1995; Pelc et al. 1991a) manipulates the concept that the MR signal from spins moving in the direction of a magnetic gradient will accumulate phase proportional to the first moment of that gradient. In MR imaging, magnetic gradient fields are used in three perpendicular directions to isolate the slice being imaged (slice selection gradient) and to encode signal’s location into the phase and frequency of the signal (phase and frequency gradients) (Riederer 1993). Thus, for PC imaging, each of these gradients is sequentially modified to enable data an acquisition that is sensitive to velocity in the three perpendicular directions, frequency, phase, and slice (Fig. 5). In addition, a fourth acquisition is used that is insensitive to motion, as a reference. Thus, for each phase encode line in k-space, four acquisitions are needed to capture a

3D Musculoskeletal Kinematics Using Dynamic MRI

125

Fig. 6 Fast-PC acquisition of the knee (Sheehan et al. 2012): from top to bottom, the rows represent the anatomic images and the right-left (RL), the anterior-posterior (AP), and the superior-inferior (SI) velocity images. In the velocity images, pure white represents +30 cm/s and black represents 30 cm/s. The following parameters were used: TR = 5 ms, motion rate = 30 cycles/min; phase direction = AP; #views = 4; number of averages = 2; percent phase fov = 72%; 256
256 pixels; 200 mm fov, acquired resolution = 0.94
0.94
8 mm; reconstructed resolution = 0.78
0.78
8 mm; maximum velocity encoding = 30 cm/s. The temporal resolution = 80 ms = 25 frames of data (Eq. 8). Imaging time = 1 min 30 s, 45 motion cycles (Eq. 9). Note, the scan time and number of required cycles could easily be reduced by 50% if average was not used, but all subjects could tolerate the scan time and the data averaging reduced random noise in the images. The black edges seen on the right and left side of the images are due to the 75% phase fov, which reduced the imaging time by 25%, without reducing the spatial resolution

complete (velocity in three perpendicular direction) CPC data line. This increases the overall temporal resolution by a factor of 4, with no reduction in imaging time or number of cycles required: FastPC temporal resolution ¼ 4  TR  #views

(8)

FastPC imaging time ¼ T  #phases=#views

(9)

The complete acquisition produces a temporal series of images representing the anatomy within the imaging plane, plus images representing the velocity in the frequency, phase, and slice (x, y, and z) directions (Figs. 5 and 6). Thus, following from above, using a FOV = 180 mm, a 75% phase acquisition, a T = 2 s, a TR = 5 ms,

126

F.T. Sheehan and R.M. Smith

and 20 views, the temporal resolution would be 400 ms, allowing for only 5 true frames of data. By reducing the #views to 5, the temporal resolution returns to 100 ms with 20 true frames of data. Yet, this comes at a cost in imaging time, which would increase to 54 s. A final key parameter when using fast-PC MR to track musculoskeletal motion is the maximum velocity encoding, venc. As the velocity is encoded in the phase of the signal, a phase shift of 180 is designated as the venc. Any velocity greater than 180 will create velocity aliasing. For example, a velocity producing a phase shift of 270 could not be distinguished from a negative velocity that produced a 90 phase shift. Thus, keeping venc low will improve the velocity resolution, but potentially could create aliasing. Further, any reduction in venc results in increases in TR, which will negatively affect the temporal resolution. For both cine and CPC imaging, any tissue that is moving in the imaging plane that is not synchronized to the data capture will result in destructive noise (ghosts). The largest sources of such noise when using dynamic MR to track motion are inconsistent movement and blood flow within the image. The noise from the latter can be greatly reduced using spatial pre-saturation of the blood (Im et al. 2015; Wood and Wiang 1993). This pre-saturation destroys the MR signal in the blood prior to it entering the imaging plane, thus eliminating (or greatly reducing) the noise it generates.

Advanced Applications of Dynamic MR Imaging Combining Fastcard Imaging with 3D Modeling One disadvantage of cine MR is that quantifying musculoskeletal movement from 2D spatial images becomes less accurate when there is out-of-plane motion, as the points of reference cannot be directly tracked throughout the entire movement, which could lead to larger errors (Shibanuma et al. 2004, 2005). Yet, with 2D fastcard imaging times approaching just 13 s and newer low-resolution cine imaging becoming available (Kaiser et al. 2013), it is quite possible to capture a multiplane fastcard (MPC) image set with limited scanning time. Scan times from 1.5 min (45 cycles) (Borotikar et al. 2012) to 5 min (150 cycles) (Kaiser et al. 2016) have been reported. From each time frame, a low-resolution model of the bone being tracked can be created and fit to a static 3D model of the bone, providing a method to consistently track the 3D motion of the bone (Fig. 7). This model fitting methodology also provides a method for quantifying dynamic cartilage contact parameters (Borotikar et al. 2012; Kaiser et al. 2016). Model fitting using dynamic MR data has been applied in two ways. The first is based on fitting the sparse dynamic model from a single time frame to a 3D static model of the same bone (Borotikar et al. 2012). From a single time frame in the MPC image set, a sparse model of the bone is created. Next, a high-resolution model of the same bone is created from static images. The optimal rotation and translation that minimize errors between the surfaces of the two models of the same bone are found.

3D Musculoskeletal Kinematics Using Dynamic MRI

127

Fig. 7 3D kinematics derived from combining fastcard imaging and 3D modeling. This example set is based on a previous study (Borotikar et al. 2012), where fastcard imaging was used to produce a 4D image set representing 24 evenly spaced temporal increments throughout the motion (first column). Each 3D image (representing the anatomy at a single time frame) is comprised of seven images. A sparse dynamic model of the bone (2nd column) is extracted from each of the 24, 3D image sets. The femur is shown segmented, but this can be done for the patella and tibia as well. From a high-resolution 3D image static image set, a high-resolution model of the bone is created (third column). Using registration, typically iterative closest point algorithm, the rotation and translation of the rough dynamic model that places the dynamic and static model into the best alignment are calculated. This is independently accomplished at each time frame. These rotation and translation are expressed as a matrix (STT1 = transformation required to bring the dynamic model from time frame 1 into alignment with the static model). When this is completed for all time frames, the 3D kinematic motion of the bone is known (T1TTi). Applying these transforms to rigid models of the bone that include the cartilage enables an analysis of cartilage contact (fourth column)

Thus, the alignment of the 3D static model is now known for one dynamic time frame. The 3D displacement and rotation of the bone, derived from the fast-PC data, are then applied to the static model, so that its position is known for all time frames. This was done for the patellofemoral joint with an accuracy of less than 0.9 mm (average absolute error) reported (Borotikar et al. 2012). In addition, it is possible to create the sparse model for every time frame of interest (Kaiser et al. 2016; Borotikar et al. 2012). By fitting each sparse dynamic bone model to the 3D static bone model, the 3D kinematics of the bone can be backed out from the transformation matrices describing the sparse dynamic to high-resolution static model. This has been done for the tibiofemoral joint, with an accuracy of than 0.60 mm (RMS error)

128

F.T. Sheehan and R.M. Smith

(Kaiser et al. 2016), and for the patellofemoral joint with an accuracy less than 1.3 mm (average absolute error) (Borotikar et al. 2012). Ultimately, the final accuracy is affected by the signal-to-noise ratio, the spatial resolution of the images, the size and shape of the bone being tracked, and the type of motion being analyzed.

Applications of CPC and Fast-PC The patellofemoral and tibiofemoral joint dynamics were the first to be studied using CPC MR. With the addition of fast-PC, and continuously decreasing TRs, the time for acquiring a full CPC dataset of the knee has dropped from over 7 min (Sheehan et al. 1998) to just 1.5 min (Behnam et al. 2011). If data averaging had not been used, both of these times would be halved. This reduction in time opens up the use of CPC, as the number of required motion cycles dropped from just over 200 to 45 (motion rate = 30 cycles/min). During this time, improvements in scanner strength and coil design greatly enhanced the signal-to-noise ratio, which allowed an improved accuracy. Based on a nearly identical phantom experiment, the average absolute error for tracking both in- and out-of-plane motion dropped from a maximum 1.48 mm (Sheehan et al. 1998) to 0.33 mm (Behnam et al. 2011). A recent study by Jensen and colleagues (Jensen et al. 2015) did an excellent job exploring the various potential sources of error in tracking muscle strain with CPC, finding that the bias (average error) in the velocity measures was below 1.3 mm/s. In terms of tracking muscle, researchers have focused primarily on two types of studies. One is to track a single point on the muscle in order to calculate moment arms (Finni et al. 2006; Im et al. 2015; Sheehan 2012; Westphal et al. 2013; Wilson and Sheehan 2009), tendon paths (Wilson and Sheehan 2010), and muscle excursions (Wen et al. 2008). The other is to evaluate muscle deformation, particularly as a marker of pathology (Finni et al. 2006; Kinugasa et al. 2008; Pappas et al. 2002; Silder et al. 2010; Sinha et al. 2012; Zhou and Novotny 2007). A recent group has shown the feasibility of evaluating three-dimensional strain (Jensen et al. 2016). Unfortunately, the scan time for the 3D CPC images required for 3D strain measures is unrealistically long to be applied to the evaluation of in vivo muscle strain during a volitional activity. Yet, as scanners further improve, the available TRs continue to drop, and new 3D fast MR data acquisition algorithms are developed; the scan time will likely reduce enough to allow in vivo experiments.

Future Directions CPC and fast-PC MR remain the only methodologies available to noninvasively measure in vivo 3D musculoskeletal kinematics (Bey et al. 2008; Shih et al. 2003; Fregly et al. 2005; Manal et al. 2000; Moro-oka et al. 2007; Yamashita et al. 2007). The reported accuracy of tracking rigid motion (0.33 mm) is better than other noninvasive techniques that measure in vivo skeletal kinematics (e.g., fluoroscopy

3D Musculoskeletal Kinematics Using Dynamic MRI

129

and motion capture) and muscle dynamics (e.g., ultrasound). It has an advantage over x-ray-based techniques (e.g., fluoroscopy and 4D ultrasound) in that it does not expose subjects to ionizing radiation. Quantifying kinematics by combining model fitting with fastcard data is less accurate (Kaiser et al. 2016; Borotikar et al. 2012) than fast-PC imaging but does provide the opportunity to quantify cartilage contact. The accuracy of tracking musculoskeletal motion from a fast-PC acquisition is dependent on the strength of the magnet, the consistency of movement, the quality of the scanner, the TR, the number of data averages, the fov, the venc, and the signal-tonoise ratio within the imaging plane. The integration algorithms used to track the motion can help compensate for some of the errors due to noise and systematic errors from the MR scanner (Jensen et al. 2015; Pelc et al. 1995; Zhu et al. 1996). Thus, relying on past validation studies does provide a general framework for the accuracy of tracking musculoskeletal motion, but a validation based on the specific parameters being used for a particular study should be done in order to insure that the combination of parameters used leads to accurate tracking. As with all measurement techniques, dynamic MR imaging has its limitations. First, it relies on a costly imaging modality that is not readily available to all clinicians and researchers. For the most accurate data, the majority of studies use high field strength, closed-bore, MR units, which limits the types of movements that can be studied. Open-bore units are available and some cine studies have used this technology, but currently the images produced are of an inferior quality. Thus, as open-bore technology improves, the types of functional movements that can be evaluated will expand. Lastly, the dynamic MR techniques rely on repetitive movements. As acquisition methods and scanners continue to improve, this limitation is quickly being removed, and real-time dynamic MR imaging is becoming a reality. Another limitation is that the integration routines to track musculoskeletal motion are not widely available for all researchers. Yet, this will likely rapidly change in the next few years. Future advancements in various aspects of MRI technology will help expand the application of dynamic MR in the quantification of 3D musculoskeletal dynamics. Improvements in scanner and coil design will support a wider range of tasks that can be studied. Eventually, this will likely include the full range of motion for joints such as the shoulder, as well as the ability to study the kinematics of multiple joints during dynamic tasks. The development and sharing of integration algorithms and packages for CPC will support its expanded use in both the research and clinical setting. As the accuracy of real-time MR imaging reaches the level of the current CPC techniques, the types of motions and pathologies that can be studied will rapidly expand.

Cross-References ▶ Cross-Platform Comparison of Imaging Technologies for Measuring Musculoskeletal Motion

130

F.T. Sheehan and R.M. Smith

Acknowledgments This work was funded by the Intramural Research Program of the National Institutes of Health Clinical Center, Bethesda, MD, USA. This research was also made possible through the NIH Medical Research Scholars Program, a public-private partnership (http://fnih.org). We thank Judith Welsh for her help and support in the work.

References Asakawa DS, Blemker SS, Gold GE, Delp SL (2002) In vivo motion of the rectus femoris muscle after tendon transfer surgery. J Biomech 35(8):1029–1037 Behnam AJ, Herzka DA, Sheehan FT (2011) Assessing the accuracy and precision of musculoskeletal motion tracking using cine-PC MRI on a 3.0T platform. J Biomech 44(1):193–197. https://doi.org/10.1016/j.jbiomech.2010.08.029 Bey MJ, Kline SK, Tashman S, Zauel R (2008) Accuracy of biplane x-ray imaging combined with model-based tracking for measuring in-vivo patellofemoral joint motion. J Orthop Surg Res 3:38. https://doi.org/10.1186/1749-799x-3-38 Borotikar BS, Sipprell WH 3rd, Wible EE, Sheehan FT (2012) A methodology to accurately quantify patellofemoral cartilage contact kinematics by combining 3D image shape registration and cine-PC MRI velocity data. J Biomech 45(6):1117–1122 Brossmann J, Muhle C, Schroder C, Melchert UH, Bull CC, Spielmann RP, Heller M (1993) Patellar tracking patterns during active and passive knee extension: evaluation with motiontriggered cine MR imaging. Radiology 187(1):205–212. https://doi.org/10.1148/radiology.187. 1.8451415 Burnett KR, Davis CL, Read J (1987) Dynamic display of the temporomandibular joint meniscus by using “fast-scan” MR imaging. AJR Am J Roentgenol 149(5):959–962. https://doi.org/10.2214/ ajr.149.5.959 Cadera W, Viirre E, Karlik S (1992) Cine magnetic resonance imaging of ocular motility. J Pediatr Ophthalmol Strabismus 29(2):120–122 Carlson VR, Boden BP, Sheehan FT (2016) Patellofemoral kinematics and tibial tuberositytrochlear groove distances in female adolescents with patellofemoral pain. Am J Sports Med. https://doi.org/10.1177/0363546516679139 Cheng S, Butler JE, Gandevia SC, Bilston LE (2008) Movement of the tongue during normal breathing in awake healthy humans. J Physiol 586(17):4283–4294. https://doi.org/10.1113/ jphysiol.2008.156430 Drace JE, Pelc NJ (1994) Skeletal muscle contraction: analysis with use of velocity distributions from phase-contrast MR imaging. Radiology 193(2):423–429. https://doi.org/10.1148/ radiology.193.2.7972757 Edsfeldt S, Rempel D, Kursa K, Diao E, Lattanza L (2015) In vivo flexor tendon forces generated during different rehabilitation exercises. J Hand Surg Eur Vol 40(7):705–710. https://doi.org/ 10.1177/1753193415591491 Feinberg DA, Crooks L, Hoenninger J 3rd, Arakawa M, Watts J (1984) Pulsatile blood velocity in human arteries displayed by magnetic resonance imaging. Radiology 153(1):177–180. https:// doi.org/10.1148/radiology.153.1.6473779 Finni T, Hodgson JA, Lai AM, Edgerton VR, Sinha S (2006) Muscle synergism during isometric plantarflexion in achilles tendon rupture patients and in normal subjects revealed by velocityencoded cine phase-contrast MRI. Clin Biomech (Bristol, Avon) 21(1):67–74. https://doi.org/ 10.1016/j.clinbiomech.2005.08.007 Foo TK, Bernstein MA, Aisen AM, Hernandez RJ, Collick BD, Bernstein T (1995) Improved ejection fraction and flow velocity estimates with use of view sharing and uniform repetition time excitation with fast cardiac techniques. Radiology 195(2):471–478. https://doi.org/ 10.1148/radiology.195.2.7724769

3D Musculoskeletal Kinematics Using Dynamic MRI

131

Fregly BJ, Rahman HA, Banks SA (2005) Theoretical accuracy of model-based shape matching for measuring natural knee kinematics with single-plane fluoroscopy. J Biomech Eng 127(4): 692–699 Fujiwara T, Togashi K, Yamaoka T, Nakai A, Kido A, Nishio S, Yamamoto T, Kitagaki H, Fujii S (2004) Kinematics of the uterus: cine mode MR imaging. Radiogr Rev Publ Radiol Soc N Am Inc 24(1):e19. https://doi.org/10.1148/rg.e19 Glover GH, Pelc NJ (1988) A rapid-gated cine MRI technique. In: Magnetic resonance annual. Raven Press, New York, NY, pp 299–333 Im HS, Goltzer O, Sheehan FT (2015) The effective quadriceps and patellar tendon moment arms relative to the tibiofemoral finite helical axis. J Biomech 48(14):3737–3742. https://doi.org/ 10.1016/j.jbiomech.2015.04.003 Jensen ER, Morrow DA, Felmlee JP, Odegard GM, Kaufman KR (2015) Error analysis of cine phase contrast MRI velocity measurements used for strain calculation. J Biomech 48(1):95–103. https://doi.org/10.1016/j.jbiomech.2014.10.035 Jensen ER, Morrow DA, Felmlee JP, Murthy NS, Kaufman KR (2016) Characterization of three dimensional volumetric strain distribution during passive tension of the human tibialis anterior using cine phase contrast MRI. J Biomech 49(14):3430–3436. https://doi.org/10.1016/j. jbiomech.2016.09.002 Kaiser J, Bradford R, Johnson K, Wieben O, Thelen DG (2013) Measurement of tibiofemoral kinematics using highly accelerated 3D radial sampling. Magn Reson Med 69(5):1310–1316. https://doi.org/10.1002/mrm.24362 Kaiser J, Monawer A, Chaudhary R, Johnson KM, Wieben O, Kijowski R, Thelen DG (2016) Accuracy of model-based tracking of knee kinematics and cartilage contact measured by dynamic volumetric MRI. Med Eng Phys 38(10):1131–1135. https://doi.org/10.1016/j. medengphy.2016.06.016 Keegan J, Firmin D, Gatehouse P, Longmore D (1994) The application of breath hold phase velocity mapping techniques to the measurement of coronary artery blood flow velocity: phantom data and initial in vivo results. Magn Reson Med 31(5):526–536 Kinugasa R, Shin D, Yamauchi J, Mishra C, Hodgson JA, Edgerton VR, Sinha S (2008) Phasecontrast MRI reveals mechanical behavior of superficial and deep aponeuroses in human medial gastrocnemius during isometric contraction. J Appl Physiol (Bethesda, MD: 1985) 105(4): 1312–1320. https://doi.org/10.1152/japplphysiol.90440.2008 Lafortune MA, Cavanagh PR, Sommer HJ 3rd, Kalenak A (1994) Foot inversion-eversion and knee kinematics during walking. J Orthop Res 12(3):412–420. https://doi.org/10.1002/jor. 1100120314 Langner I, Fischer S, Eisenschenk A, Langner S (2015) Cine MRI: a new approach to the diagnosis of scapholunate dissociation. Skelet Radiol 44(8):1103–1110. https://doi.org/ 10.1007/s00256-015-2126-4 Manal K, McClay I, Stanhope S, Richards J, Galinat B (2000) Comparison of surface mounted markers and attachment methods in estimating tibial rotations during walking: an in vivo study. Gait Posture 11(1):38–45 Markl M, Alley MT, Pelc NJ (2003) Balanced phase-contrast steady-state free precession (PC-SSFP): a novel technique for velocity encoding by gradient inversion. Magn Reson Med 49(5):945–952. https://doi.org/10.1002/mrm.10451 Melchert UH, Schroder C, Brossmann J, Muhle C (1992) Motion-triggered cine MR imaging of active joint movement. Magn Reson Imaging 10(3):457–460 Moerman KM, Sprengers AM, Simms CK, Lamerichs RM, Stoker J, Nederveen AJ (2012) Validation of continuously tagged MRI for the measurement of dynamic 3D skeletal muscle tissue deformation. Med Phys 39(4):1793–1810. https://doi.org/10.1118/1.3685579 Moro-oka TA, Hamai S, Miura H, Shimoto T, Higaki H, Fregly BJ, Iwamoto Y, Banks SA (2007) Can magnetic resonance imaging-derived bone models be used for accurate motion measurement with single-plane three-dimensional shape registration? J Orthop Res 25(7):867–872. https://doi.org/10.1002/jor.20355

132

F.T. Sheehan and R.M. Smith

Pappas GP, Asakawa DS, Delp SL, Zajac FE, Drace JE (2002) Nonuniform shortening in the biceps brachii during elbow flexion. J Appl Physiol (Bethesda, MD: 1985) 92(6):2381–2389. https:// doi.org/10.1152/japplphysiol.00843.2001 Pelc NJ (1995) Flow quantification and analysis methods. Magn Reson Imaging Clin N Am 3(3):413–424 Pelc NJ, Bernstein MA, Shimakawa A, Glover GH (1991a) Encoding strategies for three-direction phase-contrast MR imaging of flow. J Magn Reson Imaging: JMRI 1(4):405–413 Pelc NJ, Herfkens RJ, Shimakawa A, Enzmann DR (1991b) Phase contrast cine magnetic resonance imaging. Magn Reson Q 7(4):229–254 Pelc NJ, Sommer FG, Li KC, Brosnan TJ, Herfkens RJ, Enzmann DR (1994) Quantitative magnetic resonance flow imaging. Magn Reson Q 10(3):125–147 Pelc NJ, Drangova M, Pelc LR, Zhu Y, Noll DC, Bowman BS, Herfkens RJ (1995) Tracking of cyclic motion with phase-contrast cine MR velocity data. J Magn Reson Imaging: JMRI 5(3):339–345 Pike GB, Meyer CH, Brosnan TJ, Pelc NJ (1994) Magnetic resonance velocity imaging using a fast spiral phase contrast sequence. Magn Reson Med 32(4):476–483 Regev GJ, Kim CW, Tomiya A, Lee YP, Ghofrani H, Garfin SR, Lieber RL, Ward SR (2011) Psoas muscle architectural design, in vivo sarcomere length range, and passive tensile properties support its role as a lumbar spine stabilizer. Spine 36(26):E1666–E1674. https://doi.org/ 10.1097/BRS.0b013e31821847b3 Riederer SJ (1993) Spatial encoding and image reconstruction. In: Bronskill MJ, Sprawls P (eds) The physics of MRI: 1992 AAP< summer school proceedings. American Instiutes of Physics, Woodbury, pp 135–165 Sheehan FT (2012) The 3D in vivo Achilles’ tendon moment arm, quantified during active muscle control and compared across sexes. J Biomech 45(2):225–230. https://doi.org/10.1016/j. jbiomech.2011.11.001 Sheehan FT, Drace JE (2000) Human patellar tendon strain. A noninvasive, in vivo study. Clin Orthop Relat Res 370:201–207 Sheehan FT, Zajac FE, Drace JE (1998) Using cine phase contrast magnetic resonance imaging to non-invasively study in vivo knee dynamics. J Biomech 31(1):21–26 Sheehan FT, Borotikar BS, Behnam AJ, Alter KE (2012) Alterations in in vivo knee joint kinematics following a femoral nerve branch block of the vastus medialis: implications for patellofemoral pain syndrome. Clin Biomech (Bristol, Avon) 27(6):525–531. https://doi.org/ 10.1016/j.clinbiomech.2011.12.012 Shibanuma N, Sheehan FT, Lipsky PE, Stanhope SJ (2004) Sensitivity of femoral orientation estimates to condylar surface and MR image plane location. J Magn Reson Imaging: JMRI 20(2):300–305. https://doi.org/10.1002/jmri.20106 Shibanuma N, Sheehan FT, Stanhope SJ (2005) Limb positioning is critical for defining patellofemoral alignment and femoral shape. Clin Orthop Relat Res 434:198–206 Shih YF, Bull AM, McGregor AH, Humphries K, Amis AA (2003) A technique for the measurement of patellar tracking during weight-bearing activities using ultrasound. Proc Inst Mech Eng H J Eng Med 217(6):449–457 Silder A, Reeder SB, Thelen DG (2010) The influence of prior hamstring injury on lengthening muscle tissue mechanics. J Biomech 43(12):2254–2260. https://doi.org/10.1016/j.jbiomech.2010.02.038 Sinha S, Shin DD, Hodgson JA, Kinugasa R, Edgerton VR (2012) Computer-controlled, MR-compatible foot-pedal device to study dynamics of the muscle tendon complex under isometric, concentric, and eccentric contractions. J Magn Reson Imaging: JMRI 36(2): 498–504. https://doi.org/10.1002/jmri.23617 Stone M, Davis EP, Douglas AS, NessAiver M, Gullapalli R, Levine WS, Lundberg A (2001) Modeling the motion of the internal tongue from tagged cine-MRI images. J Acoust Soc Am 109(6):2974–2982 Thompson RB, McVeigh ER (2004) Flow-gated phase-contrast MRI using radial acquisitions. Magn Reson Med 52(3):598–604. https://doi.org/10.1002/mrm.20187

3D Musculoskeletal Kinematics Using Dynamic MRI

133

Thunberg P, Karlsson M, Wigstrom L (2003) Accuracy and reproducibility in phase contrast imaging using SENSE. Magn Reson Med 50(5):1061–1068. https://doi.org/10.1002/ mrm.10634 van Kampen A, Huiskes R (1990) The three-dimensional tracking pattern of the human patella. J Orthop Res 8(3):372–382. https://doi.org/10.1002/jor.1100080309 Verbruggen SW, Loo JH, Hayat TT, Hajnal JV, Rutherford MA, Phillips AT, Nowlan NC (2016) Modeling the biomechanics of fetal movements. Biomech Model Mechanobiol 15(4): 995–1004. https://doi.org/10.1007/s10237-015-0738-1 Waterton JC, Jenkins JP, Zhu XP, Love HG, Isherwood I, Rowlands DJ (1985) Magnetic resonance (MR) cine imaging of the human heart. Br J Radiol 58(692):711–716. https://doi.org/10.1259/ 0007-1285-58-692-711 Wedeen VJ, Rosen BR, Chesler D, Brady TJ (1985) MR velocity imaging by phase display. J Comput Assist Tomogr 9(3):530–536 Wen H, Dou Z, Finni T, Havu M, Kang Z, Cheng S, Sipila S, Sinha S, Usenius JP, Cheng S (2008) Thigh muscle function in stroke patients revealed by velocity-encoded cine phase-contrast magnetic resonance imaging. Muscle Nerve 37(6):736–744. https://doi.org/10.1002/mus.20986 Westphal CJ, Schmitz A, Reeder SB, Thelen DG (2013) Load-dependent variations in knee kinematics measured with dynamic MRI. J Biomech 46(12):2045–2052. https://doi.org/ 10.1016/j.jbiomech.2013.05.027 Wilson NA, Sheehan FT (2009) Dynamic in vivo 3-dimensional moment arms of the individual quadriceps components. J Biomech 42(12):1891–1897. https://doi.org/10.1016/j.jbiomech. 2009.05.011 Wilson NA, Sheehan FT (2010) Dynamic in vivo quadriceps lines-of-action. J Biomech 43(11): 2106–2113. https://doi.org/10.1016/j.jbiomech.2010.04.002 Wood M, Wiang Q (1993) Motion artifacts and remedies. In: Bronskill MJ, Sprawls P (eds) The physics of MRI: 1992 AAP< summer school proceedings. American Instiutes of Physics, Woodbury, pp 383–411 Yamashita S, Isoda H, Hirano M, Takeda H, Inagawa S, Takehara Y, Alley MT, Markl M, Pelc NJ, Sakahara H (2007) Visualization of hemodynamics in intracranial arteries using time-resolved three-dimensional phase-contrast MRI. J Magn Reson Imaging: JMRI 25(3):473–478. https:// doi.org/10.1002/jmri.20828 Zhou H, Novotny JE (2007) Cine phase contrast MRI to measure continuum Lagrangian finite strain fields in contracting skeletal muscle. J Magn Reson Imaging: JMRI 25(1):175–184. https://doi.org/10.1002/jmri.20783 Zhu Y, Drangova M, Pelc NJ (1996) Fourier tracking of myocardial motion using cine-PC data. Magn Reson Med 35(4):471–480

Cross-Platform Comparison of Imaging Technologies for Measuring Musculoskeletal Motion Richard M. Smith and Frances T. Sheehan

Abstract

Human movement is integral to daily life, it defines our species (the ability to walk upright and manipulate objects using an opposable thumb), and it is central to our ability to interact with our environment. As such, the study of human motion is dually important in our ability to optimize human functional ability. It provides a platform for understanding how pathology or injury affects human motion, so that we can both prevent and treat such pathologies. The earliest studies of human motion were mainly observational to qualify types of movements, while the current discipline and subdisciplines of human movement studies aim to quantify musculoskeletal kinematics, at times with submillimeter accuracy. The aim of this chapter is to discuss invasive and noninvasive methodologies for studying human motion with a focus on the reported accuracies, advantages, and limitations for each technique. Accuracies are presented throughout this chapter if they were reported as maximum average absolute or root mean squared errors for accuracy data for translational (in millimeters) and rotational data (in degrees) in order to simplify the reporting of cumulative accuracies from relevant articles. Thus, this review will highlight the current state of each methodology, as a platform for future investigators to build on these technologies. Keywords

Validation • Accuracy • Magnetic resonance imaging • MRI • Cine MRI • Cine phase contrast • CPC motion capture • Fluoroscopy • Single-plane videoradiography • Biplane videoradiography • Ultrasound • Muscle • Skeletal • R.M. Smith • F.T. Sheehan (*) Rehabilitation Medicine Department, Functional and Applied Biomechanics Section, National Institutes of Health, Bethesda, MD, USA e-mail: [email protected]; [email protected] # This is a U.S. Government work and not under copyright protection in the US; foreign copyright protection may apply 2018 B. Müller, S.I. Wolf (eds.), Handbook of Human Motion, https://doi.org/10.1007/978-3-319-14418-4_194

135

136

R.M. Smith and F.T. Sheehan

Musculoskeletal • Computed Tomography • CT • Motion capture • Optoelectronic tracking system • OTS • Pose estimation

Contents Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . State of the Art . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Invasive Methods in Motion Analysis and Radiographic-Based Studies . . . . . . . . . . . . . . . . . . . . . . Motion Analysis with X-Ray Fluoroscopy and Computed Tomography . . . . . . . . . . . . . . . . . . . . . . X-ray Stereophotogrammetry . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Fluoroscopy . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Dynamic Computed Tomography . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Noninvasive Imaging . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Motion Capture . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Dynamic Ultrasound . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Magnetic Resonance Imaging (MRI) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Future Directions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

136 137 137 139 139 139 142 143 143 145 146 149 151

Introduction Human movement is integral to daily life. It defines our species (the ability to walk upright and manipulate objects using an opposable thumb) and is central to our ability to interact with our environment. As such, the study of musculoskeletal kinematics is crucial as it (1) provides qualitative and quantitative information about the musculoskeletal system, (2) aids our understanding of how pathology or injury affects human motion, (3) increases our ability to prevent and treat such pathologies, and (4) is a basis from which to optimize human functional performance. The aim of this chapter is to discuss invasive and noninvasive methodologies for estimating the static and dynamic joint pose (position and orientation) and muscle dynamics. There is a focus on the reported accuracies, advantages, limitations, as well as future directions for each technique. This review does not intend to summarize every published work in the field. However, great care was taken to perform a robust search of both the PubMed and EMBASE databases for any article on human musculoskeletal kinematics. Both authors systematically sorted the resulting references by title, abstract, and full text to include only those dealing directly with accuracy of musculoskeletal tracking techniques. For the purpose of this review, invasive techniques will refer to methods that introduce a break in skin and surface barriers to insert research instruments, such as bone-anchored pins or screws and radiopaque beads. Accuracies will be presented throughout the chapter as maximum average absolute or root mean squared (RMS) errors for translational (in millimeters) and rotational data (in degrees) to simplify discussion of the cumulative accuracies from relevant articles. This decision was made, as average error represents the bias of a measurement technique. Bias only defines accuracy when numerous measures (at least 30–50) of the same quantity can be acquired and then averaged to remove

Cross-Platform Comparison of Imaging Technologies for Measuring. . .

137

the inaccuracy from random error (Materials 2010). Additionally, this chapter seeks to distinguish the technologies which provide static two-dimensional analyses from others which provide truly three-dimensional assessments of musculoskeletal motion. Thus, this review will highlight the current state of each methodology as a platform for future investigators to build on these technologies.

State of the Art Historical studies of human motion date back through the millennia and may predate the ancient Greek philosophers, Hippocrates, Plato, and Aristotle (Abernethy 2013). However, the modern study of human movement evolved as an academic discipline in the mid-twentieth century with photographic and electrogoniometric techniques utilizing both in vivo and cadaver studies (Abernethy 2013). Early noninvasive gait studies were followed by more invasive methods (e.g., bone-anchored pins) as researchers sought greater accuracy in recording the underlying skeletal motion. In the following decades, fluoroscopy and computed tomography (CT) were adapted with and without invasive methods to study skeletal kinematics with the drawback of ionizing radiation exposure. Yet, advances in X-ray-based modalities have led to decreases in radiation exposure. Currently in the research community, there has been a refocus to noninvasive, nonionizing methods to study musculoskeletal dynamics. These include magnetic resonance imaging (MRI), ultrasonography, and motion capture systems. Both MRI and ultrasound provide the added benefit of being capable of tracking muscle motion. It is these non-irradiating, noninvasive techniques which have paved the way for greater research applicability especially in voluntary research studies involving vulnerable populations (e.g., children and pregnant women). From a global perspective, the earliest studies were based on observations to qualify the types of human movements, while the current discipline and subdisciplines of human movement studies aim to quantify musculoskeletal kinematics, at times with submillimeter accuracy. It is this current aim of human motion studies that makes it a truly evolving field. For details on the historical aspects of human motion studies, see the chapter on ▶ “Observing and Revealing the Hidden Structure of the Human Form in Motion Throughout the Centuries.”

Invasive Methods in Motion Analysis and Radiographic-Based Studies Highly invasive techniques with percutaneous skeletal markers and bone-anchored screws, intracortical pins, and radiopaque beads were adapted from earlier motion capture studies as a means of investigating skeletal kinematics without the effects of soft tissue artifacts. In the study of human kinematics, invasive procedures are far reaching given their adaptability to many, if not all, imaging modalities. Invasive

138

R.M. Smith and F.T. Sheehan

methodologies have been used to study the hip and pelvis (Holden et al. 1997; Neptune and Hull 1995) and both upper and lower appendages, especially the knee joint, ankle complex, and glenohumeral joints (Dal Maso et al. 2014; Lafortune 1984; Manal et al. 2000; Reinschmidt et al. 1997). Although intracortical pins are limited in their applicability to in vivo human studies, due to their invasiveness; they have become one of the gold standards for validating other measurement techniques. Initial work using photogrammetry coupled with intracortical pins by Levens et al. (1948) and later by Lafortune (1984) provided insights into multibody pose estimation during the gait cycle. Early validation studies focused on quantifying the accuracy of pins compared to surface trackers. Over time, intracortical pins became their own gold standard. Employing both photographic and radiographic methods, Lafortune et al. (1992, 1994) ascertained an average maximal detectable difference of 0.4 mm radiographically and 0.5 mm for their photogrammetric system calibration. Later photogrammetric work with intracortical pins showed that pins were superior to external markers when studying tibiofemoral rotations but not superior to surface markers for rotations of the tibiocalcaneal joint (Reinschmidt et al. 1997). More recent studies of the upper extremity with intracortical pins and computed tomography (CT) have demonstrated RMS error of 0.15 mm and 0.2 for glenohumeral kinematics using a computer-based filtering technique (Dal Maso et al. 2014, 2016). Besides their invasiveness, the other drawback to using intracortical pins is the potential for the pins to loosen or dislodge. In an effort to mitigate errors related to soft tissue artifact from skin-surface markers and to avoid the invasiveness of bone-mounted markers, Holden et al. (1997) developed and validated a percutaneous skeletal tracker, a clamp-like device attached to the periosteum (rather than cortex) of the underlying bone via halo pins (2 mm diameter). In a preliminary study, using unfiltered data, an absolute error of 1.7 mm was reported (Holden et al. 1997). Percutaneous trackers offer an advantage over intracortical pins because their construct provides a compressive force to minimize soft tissue movement. Percutaneous skeletal trackers are often considered “minimally invasive” when compared to larger and more depth-intrusive intracortical pins. These trackers allow for accurate analyses, on par with intracortical pin studies of (Dal Maso et al. 2014). Invasive methodologies coupled to ionizing radiation imaging technologies (Brainerd et al. 2010; Tashman and Anderst 2003; Veress et al. 1979; You et al. 2001) and nonionizing MRI studies (Pelc et al. 1995) have also been used to study motion. Radiopaque bone markers, such as lead, steel, or tantalum beads, have been used in vitro and in vivo studies with fluoroscopic modalities in a technique widely known as radiostereometric analysis. However, in vivo accuracy can be limited by an inability to visualize all markers, due to field of view (FOV) limitations or other anatomic structures blocking the view of the implants (e.g., the contralateral limb). Overall, invasive studies have been used as the gold standard for validation of other measurement techniques, due to their inherent accuracy. However, there are several disadvantages to their use, such as risk of infection, bleeding, and the use of anesthesia affecting the “natural” motion of the test subject. These limit their use for large-scale population studies, studies of pathology, and longitudinal assessments. In

Cross-Platform Comparison of Imaging Technologies for Measuring. . .

139

summary, invasive procedures have lost prestige in human research due to their low favorability in recruiting subjects for non-medically necessary research, especially among minors. Notwithstanding, invasive studies continue to direct veterinary studies, as well as cadaveric studies.

Motion Analysis with X-Ray Fluoroscopy and Computed Tomography X-ray Stereophotogrammetry The first sets of experiments with X-ray stereophotogrammetry provided two-dimensional information of relative bone pose and movement (Lippert et al. 1975; Tranberg et al. 2011; Veress et al. 1979). These early systems were limited by poor resolution, motion blurring, and artifacts due to magnification and perspective. Over time, digitization procedures, using bone-specific coordinate transformations, were adopted to correct for such distortions. A major disadvantage with these two-dimensional techniques was the constraints on what movements could be studied; for example, studies of the lower limbs were limited to small incremental angular movements between poses.

Fluoroscopy Fluoroscopy evolved from the need for a noninvasive method to track joint function, particularly of the spine and knee. Fluoroscopy exposes subjects to continuous ionizing X-rays which can then be recorded to provide looping video images. Hence, singleplane and biplane videoradiography are adaptations of this X-ray technique. In contrast to earlier plain film X-ray, registration of dynamic fluoroscopic images to 3D bone models enables dynamic noninvasive tracking of skeletal kinematics. In general, these techniques began with a library-based imaging matching methodology presented by Banks and Hodge (1996) for single-plane fluoroscopy. With increased access to computational power, the library-based technique gave way to the current 2D to 3D matching methodologies for both single-plane and biplane videoradiography. The advantages of fluoroscopy are its ability to noninvasively assess joint function under dynamic activities and, in the case of biplane videoradiography, its submillimeter accuracies. For single-plane fluoroscopy registered to CT-based bone models, the maximum in-plane errors across studies range from 1.6 to 2.0 mm and 1.0 to 1.6 (Table 1). The out-of-plane errors tend to be higher, and for this reason, single-plane fluoroscopy has been primarily recommended for studies evaluating planar motion (Fregly et al. 2005). Biplane videoradiography can provide submillimeter accuracies for tracking 3D motion (Table 2). One of the most realistic validation studies was based on comparing the in vivo tracking of implanted tantalum beads to quantify tibiofemoral motion during running in three human subjects (Anderst et al. 2009). The maximum reported RMS error for

140

R.M. Smith and F.T. Sheehan

Table 1 Single-plane videoradiography accuracies. Studies were included if a root mean square (RMS) or average absolute (AA) error was available Study

Year

Object imaged

Motion

San Juan

2010

Harvested cadaver

1D translation

Lina Cerciello Tang

2013 2011 2004

Fernandezb

2008

Humerus and scapula Cadaver spine Dried vertebrae Cadaver knee (PF joint) Dried tibial and femoral bones

Maximum RMS or ave. abs error Translational Rotational IP OP IP OP 0.43 mm

Static Static Static

0.3 mm 2.0 mm 1.6 mm

2.5 mm

Static

2.0 mm

1.7 mm

0.5 2.1

2.9

3.8 mm

1.0

1.9

Abbreviations: IP in-plane, OP out -of -plane, 1D one-dimensional, PF patellofemoral a Estimated from Fig. 10 in (Lin et al. 2013) b Based on a single pose

tracking tibiofemoral motion was 1.5 mm and 1.8 . For individual bones, these RMS errors dropped to a maximum of 0.9 mm. For all fluoroscopic techniques, the capture FOV is small, limiting the ranges and types of motion that can be studied. In one study (Wang et al. 2015), heel strike and toe off needed to be captured separately, as the length of the foot could not be captured within the available FOV. A recent study by Guan et al. (2016) presents a system that partially overcomes this limitation by mounting a fluoroscopy system onto a computer-controlled sliding track that follows a subject during a walk, or potentially a run, along a linear track. These authors reported excellent accuracies (0.8 mm and 0.8 , max RMS error), yet the simulated walking speed was approximately 50% slower than typical (Bohannon and Williams Andrews 2011). Although rarely discussed, another limitation of fluoroscopy studies is the data processing times. One study reported analysis times up to 12.5 h (Ohnishi et al. 2010). Longer processing times make the applicability of biplane videoradiography in clinical, non-research, settings less likely. Moving away from CT models to MR-based models, which typically require lengthy manual segmentation of the bones of interest, will serve to increase computation time, whereas ever-increasing computer processor speeds will serve to reduce computation time. For researchers looking to apply this modality, care must be given in terms of radiation exposure and the true accuracy of the system being used. The capture rate is system dependent, with the majority of commercially available systems having capture rates of just 30 frames/s. This limits studies to motion at unnaturally slow speeds (Tashman 2008). Several custom-built videoradiography systems with capture rates of 200 frames/s and 250 frames/s (Anderst et al. 2009; Guan et al. 2016)

Cross-Platform Comparison of Imaging Technologies for Measuring. . .

141

Table 2 Biplane videoradiography accuracies. Studies were included if a root mean square (RMS) or average absolute (AA) error was available. Note, the Anderst et al. (2009) and the Ohnishi study are listed three and two times, respectively, to highlight three and two unique validations done within these studies

Study You

Year 2001

deBruina

2008

Beyb

2006

Beyb

2008

Wangc

2008

Object imaged In vivo canine tibia Sawbone scapula Cadaver – humerus/ scapula Cadaver – PF Ovine spine

Anderst

2009

In vivo TF

Anderst

2009

In vivo tibia

Anderst Ohnishi Ohnishi Wang

2009 2010 2010 2015

In vivo TF Ovine knee Ovine knee Cadaver rear foot

Thorhauerd Guan

2015 2015

Cadaver TF Cadaver TF

Frame rate

Shutter speed

Motion Walking (1.5 m/s) Static

(Frames/s) 250

(μs) 0.5

Dynamic (unknown)

50

Dynamic (unknown) Translation (0.017 m/s) Running (2.5 m/s) Running (2.5 m/s) Static Static Static Simulated walking (1.0 m/s) Static Simulated walking (0.7 m/s)

60

Maximum RMS or AA error Trans (mm) Rot 0.8 3.9 0.3

0.4 a

2.0

0.4

0.9

2.0

0.4

0.9

30

0.2

250

0.5

1.5

250

0.5

0.9

250 30 30 100

0.5

1.0

0.26 0.5 0.4 0.7

250 200

0.5 5.0

1.5 0.8

1.8

0.9 0.7 0.6

0.8

Abbreviations: TF tibiofemoral a RMS error calculated from Table 2 in text b Motion rate not provided c Translation was a 1 degree of freedom movement and error increased to 0.4 mm when an MRI model was used d Accuracy listed for locating point of contact

have enabled biplane videoradiography to capture movement at realistic speeds. When these fast frames rates are coupled with fast shutter speeds, most issues associated with blurring are removed (Tashman 2008). Yet, based on numerous studies, quantifying dynamic joint kinematics is less accurate than quantifying a static pose, even with customized systems (Anderst et al. 2009; Ohnishi et al. 2010). The system parameters, the speed of motion, the shape/size of the bone being tracked, and the 3D model used can all affect tracking accuracy. For example,

142

R.M. Smith and F.T. Sheehan

tracking single bones is more accurate than tracking two bones relative to each other (Anderst et al. 2009; Ohnishi et al. 2010). Further, using an MRI model (de Bruin et al. 2008; Thorhauer and Tashman 2015; Wang et al. 2008) or a lower-resolution CT model (Fox et al. 2011) can increase errors. One key issue raised by Anderst et al. (2009) is that capturing dynamic bone kinematics versus static pose is less accurate. Thus, researchers must carefully discern if previous validation studies are truly applicable to their current experimental design or if a validation study for their individual study is needed. For example, two studies (Bey et al. 2006, 2008) reported excellent accuracies for tracking patellofemoral and humeral motion, yet the motion rate evaluated was not provided. Thus, it is unclear how to apply their results to future studies. Both single-plane and biplane videoradiography have advanced the study of skeletal kinematics. Single-plane fluoroscopy has its strengths for studies involving in-plane motion (Fregly et al. 2005; Cerciello et al. 2011; Fernandez et al. 2008), as such it is an indispensable clinical tool during intraoperative planning and postsurgical follow-up. Biplane videoradiography is a continually evolving technique which can provide submillimeter accuracies. Additionally, biplane shows great adaptability in its application to the study of functional tasks, such as walking and running (Anderst et al. 2009; Guan et al. 2016). Innovations in its use, such as creating a mobile videoradiography platform (Guan et al. 2016), allow for a wide variety of joints and joint range of motions to be captured with greater accuracy than motion capture systems, discussed later in this chapter. For details on biplane videoradiography, see the chapter on ▶ “Measurement of 3D Dynamic Joint Motion Using Biplane Videoradiography.”

Dynamic Computed Tomography Computed tomography (CT) is another common noninvasive clinical tool which uses ionizing radiation to reconstruct three-dimensional volumetric anatomical data from planar datasets. This technology has also progressed from static CT to both contrast-enhanced CT and four-dimensional (dynamic) CT. Similar to X-ray stereophotogrammetry, early studies in conventional CT (Dupuy et al. 1997; Fischer et al. 2001; Lee et al. 2014; Rogers et al. 2005; Schutzer et al. 1986; Zuhlke et al. 2009) provided information about static alignment for various joints. In one study, researchers utilized CT and electrogoniometry to study joint pose in 6 degrees of freedom with results validated up 1.2 mm and 0.5 ; RMS error (Jan et al. 2002). CT studies in rigid body pose estimation and registration block methodologies (Fischer et al. 2001) paved the way for future dynamic CT studies. For example, Buffi et al. (2013) utilized static datasets of relative CT bone poses to create manual computerbased bone models of the carpometacarpal joints of the hand. This in vivo study warrants commendation as it provided data on in vivo hand range of motion fitted to a biomechanical model of hand kinematics. These authors reported absolute error between the CT and digitized models up to 6.2 for supination pronation, up to 1.2 for flexion-extension, and up to 1.2 in abduction-adduction (Buffi et al. 2013).

Cross-Platform Comparison of Imaging Technologies for Measuring. . .

143

Currently, there are seemingly few validation studies for dynamic CT on skeletal kinematics. Of the studies that have focused on validation, numerous authors do not report accuracies in RMS or absolute errors, but rather they report bias or precision measurements (Gondim Teixeira et al. 2017; Kerkhof et al. 2016; Goto et al. 2014; Zuhlke et al. 2009). One validation study reported RMS localization errors between the dynamic and static CT data in the range of 0.023–0.139 mm (Zhao et al. 2015). Dynamic CT does have great clinical applicability and has already been used to study multiple joints (Dupuy et al. 1997; Kalia et al. 2009; Williams et al. 2016). Dynamic CT, like other X-ray technologies, is somewhat limited in its use due to ionizing radiation exposure. In addition, due to public perception and fears about prolonged radiation exposure, CT studies are further limited for use in underrepresented research populations, such as children and pregnant women or in anatomical regions close to the thyroids or genitals. Technical limitations of CT include capture FOV, data acquisition speed, and motion artifacts. Although the bore of CT scanners has increased in recent years to accommodate more complex positioning and subjects with increased body habitus (80–90 cm in most commercial scanners), the capture FOV for many systems is around 50–60 cm, which does not allow the full range of motion for many joints to be studied. Motion artifacts and data acquisition time will continue to improve with spiral (or helical), cone-beam, and multi-detector CT technologies. However, faster CT scan times are proportional to greater radiation exposure (Biswas et al. 2009).

Noninvasive Imaging Motion Capture Stereophotogrammetry is the science of utilizing dynamic video images, static photography, or radiography to quantify human motions. Photogrammetric studies of gait and motion date back to the late 1800s and are most attributed to the work of E. Muybridge, with his work on animal and human motion studies using freeze frame photographic analysis (Cappozzo and Paul 1998). This work in freeze frame photography was later advanced through a collaboration with E.J. Marey. Early studies (Levens et al. 1948; Eberhart and Inman 1951) in human motion capture involved a system of high-speed cameras and skin-surface markers. These skinmounted targets often consisted of retroreflective (passive) or light-emitting (active) markers clustered on the skin directly above a bone or other anatomical landmarks. In one experimental design, participants were photographically recorded walking along glass walkways to estimate human pose during gait, with an estimated accuracy of approximately 13 mm (Eberhart and Inman 1951). Progressively, researchers identified that the movement and deformation of the underlying tissue (e.g., adipose, muscle) caused movement of the skin-mounted markers, relative to the underlying bones. This affected the accuracy of tracking bone, with errors up to 40 mm, depending on the location of trackers and the activity studied (Karlsson and

144

R.M. Smith and F.T. Sheehan

Lundberg 1994; Cappozzo et al. 1996). As such, the previously discussed invasive methodologies (pins, beads, percutaneous trackers) were utilized as an attempt to directly measure and track anatomical points of interests. Initial studies of the knee with 2D goniometric systems reported average absolute errors up to 2 based on the relative placement of the device in relation to the central axis of the joint (Kettelkamp et al. 1970). Although goniometry showed consistent cyclogram patterns for certain linear activities such as walking, running, and knee flexion-extension tasks (Townsend et al. 1977); it was limited to angular motions, and the test-retest reliability depends on the correct placement of the device (Chao 1980). Modern motion capture combines multiple cameras (infrared or high-speed video-capture), calibration hardware and software, and multiprocessor computer systems to measure whole-body motion. These systems can also be tuned to evaluated small regions (e.g., facial motion). Optoelectronic tracking system (OTS) is one form of motion capture technology that utilizes this collection of hardware and software. Most OTS setups require a large open space, wherein the cameras can be oriented according to manufacturer’s specifications. This open space requirement results in motion capture being the one measurement system that can look at the largest variety of functional tasks (e.g., walking, running, pitching, golf swing, etc.). Submillimeter accuracies in locating the skin-mounted markers are achievable, but the actual accuracies for tracking skeletal kinematics using OTS are limited by lab setup (camera-subject operating range and software calibration) and anthropometric subject variables. The biggest issue in terms of the accuracy is the mismatch between the underlying bone motion and the motion tracked by skin markers. Although numerous algorithms (Wan and Nelson 2001; Todorov 2007) have been developed to compensate for this mismatch, errors attributed to skin marker artifact have a reported mean absolute error range of 0.5–7.7 mm (Chiari et al. 2005). Overall, probabilistic approaches to pose estimation have shown promise at reducing these errors (Todorov 2007). One study (Wilson et al. 2009) compensated for skin marker artifacts in motion capture by creating a thermoplast cup for the patella. By doing so, this study demonstrated the feasibility of tracking patellofemoral motion during a specific exercise (slow squat) using OTS. The overall accuracies were reported at 1.2 / 1.1 mm (average absolute errors). The term “motion capture” envelops more than just OTS systems. Electromagnetic tracking systems evaluate 3D positions relative to a magnetic field but are limited in their accuracy by artifacts created by various metals in the workspace (Milne et al. 1996; Meskers et al. 1999). Force platforms record the interaction forces between the body and the ground. These are critical data for inverse dynamics analyses of human motion, which can provide estimates of muscle forces and neuromuscular control strategies. Lastly, current advances in wearable technologies are fostering the development of portable magneto-inertial units (MIMU) that feature accelerometers, gyroscope, and magnetometers for measuring joint kinematics. Wearable sensors have been tested against known clinical tests such as “timed up and go” tasks (Salarian et al. 2010) and commercial OTS systems (Simoes 2011).

Cross-Platform Comparison of Imaging Technologies for Measuring. . .

145

MIMUs are currently limited by the accuracy of how the device estimates its pose relative to assigned reference frames (Cereatti et al. 2015). However, this technology shows great promise for future studies of multi-segment rigid body kinematics and the parameters of pathological gait in both clinical and research settings (Cereatti et al. 2015; Crabolu et al. 2016). Overall, motion capture analysis excels in its noninvasive nature and the ease at which it can quantify multi-joint movements. However, the results of such estimations have a relatively lower accuracy in tracking skeletal motion, compared to biplane videoradiography and dynamic MRI. An advantage, as well as a disadvantage, of OTS systems is the large quantity of output variables it can provide. Although this wealth of data can provide crucial insights into human movement, it can be difficult to extract the key variables that are relevant to a specific question. OTS and electromagnetic trackers must be calibrated on a case-by-case basis to ensure accurate analyses. Looking forward, small wearable activity monitors are popular consumer electronics for tracking general activity levels. In addition, there is an emerging market for powerful inertial trackers, which are currently being validated to track real-time gait measurements. Further details on motion capture systems including reflective markers or markerless systems can be found in the chapter(s) on: ▶ “Estimation of the Body Segment Inertial Parameters for the Rigid Body Biomechanical Models Used in Motion Analysis,” ▶ “3D Dynamic Pose Estimation from Marker-Based Optical Data,” ▶ “3D Dynamic Probabilistic Pose Estimation from Data Collected Using Cameras and Reflective Markers,” ▶ “3D Dynamic Pose Estimation from Markerless Optical Data,” ▶ “Physics-Based Models for Human Gait Analysis,” ▶ “Clinical Gait Assessment by Video Observation and 2D Techniques.” For further details on inertial trackers and wearable sensors, please refer to book chapter(s) on: ▶ “Gait Parameters Estimated Using Inertial Measurement Units,” ▶ “Three-Dimensional Human Kinematic Estimation Using Magneto-Inertial Measurement Units,” ▶ “Induced Acceleration and Power Analyses of Human Motion.”

Dynamic Ultrasound Diagnostic ultrasonography offers noninvasive, nonionizing imaging of the musculoskeletal system. It utilizes high frequency sound waves to produce a grayscale display based on tissue density and flow properties. Rigid or semirigid anatomic structures such as bones, blood vessels, and soft tissue reflect the transmission of acoustic waves causing propagated vibrations, which are in turn interpreted by a mechanical sensor to generate real-time dynamic images. There are several imaging protocols for diagnostic ultrasound such as Doppler (D), brightness (B), and motion (M) modes, among others. The mode and frequency of the selected ultrasound protocol depend on the depth and size of the region or structure imaged. Major advantages of ultrasound are its wide availability and its portability, which has fostered extensive use of ultrasound in both research and clinical settings. In addition, it is adaptable to studying a broad range of musculoskeletal structures

146

R.M. Smith and F.T. Sheehan

and movements, and it offers a wide field of study which allows the probe to be custom fitted to various body segments. Musculoskeletal sonography is capable of measuring structures at both the macroscopic and gross tissue level. Ultrasound technologies can track single muscle fascicle motion and tendon velocities (Tat et al. 2015). Ultrasound has also been coupled with surface or fine wire electromyography to study muscle fiber activations (Byrne et al. 2005; Pulkovski et al. 2008). For gross structure examinations, dynamic ultrasonography has wide applications in studying hyoid bone motion, pelvic floor musculature contraction (Braekken et al. 2008), ankle motion (Telfer et al. 2014), and patellar tracking (Shih et al. 2003). Static ultrasound has been coupled with motion capture and computer-assisted algorithms to study hip kinematics (Jia et al. 2016). Jia et al. (2016) produced faster processing times with their automated registration pipeline and reported maximum absolute errors up to 0.2 mm and 4 , for segmentation and registration, respectively. A major disadvantage of ultrasound is its operator bias, as data quality is dependent on the operator’s level of training and experience. As such, many ultrasound studies aim to standardize their results with measures of class or rater reliability. Shih et al. (2003) performed an in vivo dynamic study of medial-lateral patellar displacement during sitting and squatting using a custom knee brace fitted with a lateral mounted ultrasound transducer. These authors validated their technique against a static interventional MRI and even provided follow-up studies that showed increased accuracies, which were reported as bias (Shih et al. 2003, 2004). Current limitations of ultrasonography include image resolution, which is both mode and orientation dependent, its inability to capture data if an air interface is present, and its relative difficulty in studying deep musculoskeletal structures or ligaments and joints. As a whole, ultrasound offers great potential for noninvasive musculoskeletal studies involving tendons, ligaments, and muscle. For further discussion on ultrasound use in musculoskeletal kinematics, see the chapter on ▶ “Ultrasound Technology for Examining the Mechanics of the Muscle, Tendon, and Ligament.”

Magnetic Resonance Imaging (MRI) MRI uses the magnetic properties of hydrogen atoms in tissue to create 3D anatomical images of bone, soft tissue, and cartilage, as well as other structures. MRI is a noninvasive and nonionizing technique that provides three-dimensional, quantitative data with high-resolution and accuracy. High-resolution static MRI sequences (e.g., gradient echo, fat suppression, proton density weighted) can be used to capture anatomic images of joints in various poses. Static MRI does not allow for full dynamic analysis throughout the motion cycle. However, the advantage of this process is that multiple parallel scans can be segmented to provide 3D musculoskeletal models that have become the foundation for dynamic musculoskeletal models at the single and multiple joint levels. Static MRI has provided the ability to measure such quantities as musculotendon and ligament paths, muscle fascicle direction and length,

Cross-Platform Comparison of Imaging Technologies for Measuring. . .

147

tendon/ligament moment arms, relative muscle strength (in terms of muscle volume), bone shape, cartilage morphology, and numerous other properties. In addition, static MRI has been used to evaluate joint pose during active muscle activation, joint pose during weight bearing (Boden et al. 2009), and muscle volume as it relates to human movement (Eng et al. 2007; Im et al. 2014). In terms of evaluating musculoskeletal kinematics, there are three primary MRI techniques: cine MRI (Brossmann et al. 1993; Borotikar et al. 2012; Kaiser et al. 2016), MRI tagging (Pipe et al. 1991; Niitsu et al. 1992), and cine phase-contrast (CPC) MRI (Finni et al. 2006; Jensen et al. 2015; Pappas et al. 2002; Sheehan et al. 1998). CPC MRI is the only modality that can noninvasively track 3D in vivo skeletal and muscle movement (Table 3). It has the further advantage that it does Table 3 MRI Accuracies. Studies were included if a root mean square (RMS) or average absolute (AA) error was available. The table is broken into three sections. The top three studies focus primarily on muscle/soft tissue. The next three focus primarily on quantify skeletal kinematics by registering a low- to a high-resolution mode. The last three are focused primarily on using cine phase contrast (CPC) to measure skeletal motion

Study Drace

Year 1994

Pelc

1994

Moerman

2012

Object imaged Muscle (ex vivo bovine) In vivo canine heart Gelatin phantom

Fellowsa

2005

Cadaver PF joint

Borotikarb

2012

Kaiserb

2016

Sheehan

1998

Barrancec

2005

Behnam

2011

In vivo human PF joint Femoral and tibial bone segments Human cadaver bone and gelatin phantom Ex vivo bovine tibiofemoral Phantom

Motion Dynamic

Modality CPC

Dynamic

CPC

Linear depression Static

MRI tagging Low to high res Cine to model Cine to model

Dynamic Dynamic

System 1.5 T GE 1.5 T GE 3.0 T Philips 1.5 T GE 3.0 T Philips 3.0 T GE

Dynamic

CPC

1.5 T GE

Dynamic

CPC

Dynamic

CPC

1.5 T GE 3.0 T Philips

Maximum RMS or AA error Trans (mm) Rot 1.0 1.32 1.02 0.9

1.8

0.9

1.7

0.60

1.1

0.9

3.67 0.3

1.0

Abbreviations: bcine to model implies using a low spatial resolution dynamic capture and defining kinematics by registering the low-resolution images to the high-resolution images. Note two different types of cine imaging were used in these studies a Low to high res: A low-resolution static model was fitted to a high-resolution model b Cine to model: A multiplane cine model was fitted to a high-resolution model c Only a single degree of freedom was validated

148

R.M. Smith and F.T. Sheehan

not expose subjects to ionizing radiation. Yet, dynamic MRI is limited by the confined space within a closed-bore environment, which reduces the joint motions that can be studied, and it typically requires subjects to perform repetitive motion during data capture. Cine MRI was the first of the dynamic MRI techniques to be applied to the study of quantitative joint kinematics, with a focus on the temporomandibular (Burnett et al. 1987; Maniere-Ezvan et al. 1999) and patellofemoral (Brossmann et al. 1993; Ward et al. 2002) joints. This technique produces a series of anatomic images depicting cyclic motion by synchronizing the data collection to a repeated motion cycle. Recently, multiplane cine (MPC) MRI, which is cine MRI that captures multiple anatomic planes of data at each time point, has been used to track joint kinematics for the patellofemoral and tibiofemoral joints and cartilage contact for the patellofemoral joint, with submillimeter accuracy (Table 3). This was accomplished by creating spatially sparse 3D models (at each time point) of the bones being evaluated from MPC data (Borotikar et al. 2012; Kaiser et al. 2016). Independently fitting each of these temporal models to a static high-spatial-resolution 3D model of the same bone provided a transformation matrix that defined the required translation and rotation for transitioning the bone from the dynamic time frame to the static state. By acquiring the transformation matrices for all dynamic time frames to the static state, the kinematics throughout the motion were defined with submillimeter accuracy. MRI tagging enables muscle strain to be quantified by automatically tracking a signal void grid, which is established at the start of data capture. In short, this process destroys the MR signal of the tissue within the imaging plane in an even grid pattern (Pipe et al. 1991; Niitsu et al. 1992). This grid is inherent to the tissue; thus as the muscle deforms, the grid deforms with it. Visually, the image appears as a standard anatomical image with a black grid overlaid on it. Thus, tracking the grid points allows for tracking of muscle deformation with accuracies on the order of a millimeter (Table 3). Although many of the validation studies (Pipe et al. 1991; Niitsu et al. 1992; Sprengers et al. 2013) demonstrated the feasibility of using this technique to track skeletal muscle deformation, tagging has been primarily used in the arena of quantifying cardiac dynamic properties. Cine phase contrast (CPC) combines the properties of both PC and cine MRI sequences to allow cyclical anatomical and velocity data in three orthogonal directions (vx, vy, and vz) to be acquired. In total, these data provide images of the moving anatomy throughout the movement, along with the 3D velocity of each pixel within the imaging plane. Integrating the velocity data (Pelc et al. 1995) allows for tissues within the imaging plane to be tracked three-dimensionally over time with submillimeter accuracies. In doing so, dynamic musculotendon paths, moment arms, tendon strain, 3D joint kinematics, and muscle strains can be tracked throughout the motion. Unique to MRI, this modality captures dynamic data and static poses using similar, but distinctive, techniques. Since integration of CPC is used to determine the future position of an anatomical point, tracking skeletal kinematics using CPC (Behnam et al. 2011) has better accuracy than tracking joint pose using static MR (Fellows et al. 2005). In its earliest application, CPC showed high

Cross-Platform Comparison of Imaging Technologies for Measuring. . .

149

accuracy in tracking skeletal muscle up to 1 mm, ex vivo (Drace and Pelc 1994), and up to 1.32 mm, in vivo. With improvements in MR hardware and software design, along with increases in magnetic strength (1.5 to 3.0 Tesla), increased signal-to-noise (SNR) ratios, improved temporal resolution, and reductions in the number of required motion rates were achieved. Behnam et al. (2011) demonstrated an accuracy of 0.3 mm for tracking the 3D motion of a phantom using the more advanced 3 T MRI scanner. Although CPC can track motion in 3D, it is limited in its ability to quantify 3D volumetric data due to imaging time constraints. Thus, CPC has also been combined with the MPC-to-modeling methodology to provide an accurate method for tracking joint contact with an average absolute error of 0.9 mm/1.7 (Borotikar et al. 2012). In comparison, Fellows et al. (2005) registered a low-resolution static MRI model to a high-resolution static model to evaluate loaded knee pose at various flexion angles. They reported an accuracy of 0.9 mm/1.8 (Fellows et al. 2005). Overall, dynamic MRI sequences offer a safe, nonionizing method to quantify in vivo 3D musculoskeletal kinematics. While long acquisition times have been cited in past studies, MRI sequences, such as real-time MRI (Asakawa et al. 2003) and fast-PC MRI sequences, continue to advance the field by drastically reducing the scanning time and the required number of motion cycles. Current closed-bore MRI systems limit the field of study and the types of dynamic motions (such as weightbearing exercises) that can be evaluated. Open-bore systems exist but are not widely used due to lower resolution and reduced signal-to-noise SNR. Current progress in MRI technology shows promise for continued improvements in temporal resolution, imaging time, and SNR. For greater detail on dynamic MRI theory and applications, refer to the chapter on ▶ “3D Musculoskeletal Kinematics Using Dynamic MRI.”

Future Directions In total, these various tools for evaluating in vivo musculoskeletal kinematics have greatly enhanced our understanding of the function and dysfunction, in the case of injury and pathology, of the neuromuscular and musculoskeletal systems. Modern research is grounded in the principle of minimizing harm to subjects. Thus, there should be a fine balance between the use of invasive or otherwise harmful techniques and the intended research goal of advancing scientific knowledge. This is not to say that invasive techniques do not have a place. For example, tracking a total knee replacement (TKR) using single-plane fluoroscopy has been proven to be quite accurate (Banks and Hodge 1996), but it is inherently an invasive technique, based on the need for an implanted TKR. Yet, as the decision for surgery is a clinical one, using fluoroscopy post-surgery is a noninvasive, inexpensive, and expedient method for tracking the implant functionality over time. Next, the most accurate technique is not always the best technique for every question. For example, a sparsely accurate pedometer can be a perfect tool for evaluating a subject’s general level of activity, whereas to understand changes in muscle control, motion capture may be needed. In

150

R.M. Smith and F.T. Sheehan

the current state of human kinematic studies, researchers have an array of tools to select from depending on the type of movement they wish to quantify; some techniques only provide static information, while others offer full 3D kinematic data with varying accuracies. To date, CPC MRI is the most accurate means to noninvasively quantify in vivo, 3D musculoskeletal kinematics (Table 3). MRI is the only current technique that can track 3D muscle and skeletal kinematics and does not expose subjects to ionizing radiation. Dynamic MRI is limited by the type of motions that can be studied, due to the typically closed-bore environment. Yet, as scanner technology continues to advance, these limitations will likely be eliminated. Biplane videoradiography has greatly advanced the field of skeletal dynamics. Like MR, it offers a noninvasive methodology to evaluate skeletal kinematics with submillimeter accuracies, but it cannot track muscle movement and does expose subjects to ionizing radiation. The accuracy of these systems depends on the type of movement examined (slow walking vs. running), the shape of the bones under evaluation, and the inherent properties of the capture system. Additionally, the capture FOV and frame rate can limit the types of motion being studied. However, an array of custom-built biplane systems and innovative solutions (e.g., biplane systems mounted on motorized tracks) have emerged to mitigate some of the technical challenges. Lastly, dynamic (four-dimensional) CT is an emerging research methodology. While there is literature describing its qualitative properties, more validation studies are needed to assess the scope of its applicability. One ex vivo study (Zhao et al. 2015) showed promising results for achieving submillimeter accuracy. This review of human motion analysis highlights that each imaging modality has its strengths and limits. Although the measurement technologies reviewed within this chapter have shown and continue to show improved accuracy over time, it is the current trend of combining modalities that will be the true driver of novel information in the coming years. This combination of tools theoretically allows for the best aspects of each technique to be exploited, while downplaying the deficits of each technique. In vivo validation studies that closely mimic the true experimental conditions are crucial for the continued advancement of measuring musculoskeletal motion, regardless of the technique used. It is also important for researchers to report accuracies in a standardized form, such as root mean square or mean absolute errors, and not to simply report bias for previously listed reasons. Finally, through an exploration of these various techniques, it is obvious that the accuracy and precision of each modality should be considered and matched with each study’s intended goals. Acknowledgments We thank Judith Welsh for her help and support toward this project. This work was funded by the Intramural Research Program of the National Institutes of Health Clinical Center, Bethesda, MD, USA. This research was also made possible through the NIH Medical Research Scholars Program, a public-private partnership (http://fnih.org).

Cross-Platform Comparison of Imaging Technologies for Measuring. . .

151

References Abernethy B (2013) Historical origins of the academic study of human movement. In: Abernethy B, Kippers V, Stephanie HJ, Pandy MG, McManus AM, Mackinnon L (eds) Biophysical foundations of human movement. Human Kinetics, Champaign, pp 13–23 Anderst W, Zauel R, Bishop J, Demps E, Tashman S (2009) Validation of three-dimensional modelbased tibio-femoral tracking during running. Med Eng Phys 31(1):10–16. https://doi.org/ 10.1016/j.medengphy.2008.03.003 Asakawa DS, Nayak KS, Blemker SS, Delp SL, Pauly JM, Nishimura DG, Gold GE (2003) Realtime imaging of skeletal muscle velocity. J Magn Reson Imaging: JMRI 18(6):734–739. https:// doi.org/10.1002/jmri.10422 Banks SA, Hodge WA (1996) Accurate measurement of three-dimensional knee replacement kinematics using single-plane fluoroscopy. IEEE Trans Biomed Eng 43(6):638–649. https:// doi.org/10.1109/10.495283 Barrance PJ, Williams GN, Novotny JE, Buchanan TS (2005) J Biomech Eng 127(5):829–837 Behnam AJ, Herzka DA, Sheehan FT (2011) Assessing the accuracy and precision of musculoskeletal motion tracking using cine-PC MRI on a 3.0T platform. J Biomech 44(1):193–197. https://doi.org/10.1016/j.jbiomech.2010.08.029 Bey MJ, Zauel R, Brock SK, Tashman S (2006) Validation of a new model-based tracking technique for measuring three-dimensional, in vivo glenohumeral joint kinematics. J Biomech Eng 128(4):604–609. https://doi.org/10.1115/1.2206199 Bey MJ, Kline SK, Tashman S, Zauel R (2008) Accuracy of biplane x-ray imaging combined with model-based tracking for measuring in-vivo patellofemoral joint motion. J Orthop Surg Res 3:38. https://doi.org/10.1186/1749-799x-3-38 Biswas D, Bible JE, Bohan M, Simpson AK, Whang PG, Grauer JN (2009) Radiation exposure from musculoskeletal computerized tomographic scans. J Bone Joint Surg Am 91(8):1882–1889 Boden BP, Breit I, Sheehan FT (2009) Tibiofemoral alignment: contributing factors to noncontact anterior cruciate ligament injury. J Bone Joint Surg Am 91(10):2381–2389. https://doi.org/ 10.2106/JBJS.H.01721 Bohannon RW, Williams Andrews A (2011) Normal walking speed: a descriptive meta-analysis. Physiotherapy 97(3):182–189. https://doi.org/10.1016/j.physio.2010.12.004 Borotikar BS, Sipprell WH 3rd, Wible EE, Sheehan FT (2012) A methodology to accurately quantify patellofemoral cartilage contact kinematics by combining 3D image shape registration and cine-PC MRI velocity data. J Biomech 45(6):1117–1122 Braekken IH, Majida M, Ellstrom-Engh M, Dietz HP, Umek W, Bo K (2008) Test-retest and intraobserver repeatability of two-, three- and four-dimensional perineal ultrasound of pelvic floor muscle anatomy and function. Int Urogynecol J Pelvic Floor Dysfunct 19(2):227–235. https:// doi.org/10.1007/s00192-007-0408-7 Brainerd EL, Baier DB, Gatesy SM, Hedrick TL, Metzger KA, Gilbert SL, Crisco JJ (2010) X-ray reconstruction of moving morphology (XROMM): precision, accuracy and applications in comparative biomechanics research. J Exp Zool A Ecol Genet Physiol 313(5):262–279. https://doi.org/10.1002/jez.589 Brossmann J, Muhle C, Schroder C, Melchert UH, Bull CC, Spielmann RP, Heller M (1993) Patellar tracking patterns during active and passive knee extension: evaluation with motiontriggered cine MR imaging. Radiology 187(1):205–212. https://doi.org/10.1148/radiology.187. 1.8451415 de Bruin PW, Kaptein BL, Stoel BC, Reiber JH, Rozing PM, Valstar ER (2008) Image-based RSA: roentgen stereophotogrammetric analysis based on 2D-3D image registration. J Biomech 41(1):155–164. https://doi.org/10.1016/j.jbiomech.2007.07.002 Buffi JH, Crisco JJ, Murray WM (2013) A method for defining carpometacarpal joint kinematics from three-dimensional rotations of the metacarpal bones captured in vivo using computed tomography. J Biomech 46(12):2104–2108. https://doi.org/10.1016/j.jbiomech.2013.05.019

152

R.M. Smith and F.T. Sheehan

Burnett KR, Davis CL, Read J (1987) Dynamic display of the temporomandibular joint meniscus by using “fast-scan” MR imaging. AJR Am J Roentgenol 149(5):959–962. https://doi.org/10.2214/ ajr.149.5.959 Byrne CA, Lyons GM, Donnelly AE, O’Keeffe DT, Hermens H, Nene A (2005) Rectus femoris surface myoelectric signal cross-talk during static contractions. J Electromyogr Kinesiol 15(6):564–575. https://doi.org/10.1016/j.jelekin.2005.03.002 Cappozzo A, Paul J (1998) Instrumental observation of human movement: historical development. In: Allard P, Cappozzo A, Lundberg A, Vaughan C (eds) Three-dimensional analysis of human locomotion. New York, NY: Wiley, pp 1–25 Cappozzo A, Catani F, Leardini A, Benedetti MG, Della Croce U (1996) Position and orientation in space of bones during movement: experimental artefacts. Clin Biomech 11(2):90–100. https:// doi.org/10.1016/0268-0033(95)00046-1 Cerciello T, Romano M, Bifulco P, Cesarelli M, Allen R (2011) Advanced template matching method for estimation of intervertebral kinematics of lumbar spine. Med Eng Phys 33(10): 1293–1302. https://doi.org/10.1016/j.medengphy.2011.06.009 Cereatti A, Trojaniello D, Croce UD (2015) Accurately measuring human movement using magneto-inertial sensors: techniques and challenges. In: 2015 I.E. international symposium on Inertial Sensors and Systems (ISISS) proceedings, 23–26 March 2015, Hapuna Beach, HI, USA, pp 1–4. https://doi.org/10.1109/ISISS.2015.7102390 Chao EYS (1980) Justification of triaxial goniometer for the measurement of joint rotation. J Biomech 13(12):989–1006. https://doi.org/10.1016/0021-9290(80)90044-5 Chiari L, Della Croce U, Leardini A, Cappozzo A (2005) Human movement analysis using stereophotogrammetry. Part 2: instrumental errors. Gait Posture 21(2):197–211. https://doi. org/10.1016/j.gaitpost.2004.04.004 Crabolu M, Pani D, Raffo L, Cereatti A (2016) Estimation of the center of rotation using wearable magneto-inertial sensors. J Biomech 49(16):3928–3933. https://doi.org/10.1016/j.jbiomech. 2016.11.046 Dal Maso F, Raison M, Lundberg A, Arndt A, Begon M (2014) Coupling between 3D displacements and rotations at the glenohumeral joint during dynamic tasks in healthy participants. Clin Biomech (Bristol, Avon) 29(9):1048–1055. https://doi.org/10.1016/j.clinbiomech.2014.08.006 Dal Maso F, Blache Y, Raison M, Lundberg A, Begon M (2016) Glenohumeral joint kinematics measured by intracortical pins, reflective markers, and computed tomography: a novel technique to assess acromiohumeral distance. J Electromyogr Kinesiol 29:4–11. https://doi.org/10.1016/j. jelekin.2015.07.008 Drace JE, Pelc NJ (1994) Tracking the motion of skeletal muscle with velocity-encoded MR imaging. J Mag Reson Imaging: JMRI 4(6):773–778 Dupuy D, Hangen D, Zachazewski J, Boland A, Palmer W (1997) Kinematic CT of the patellofemoral joint. AJR Am J Roentgenol 169(1):211–215 Eberhart HD, Inman VT (1951) An evaluation of experimental procedures used in a fundamental study of human locomotion. Ann N Y Acad Sci 51(7):1213–1228 Eng CM, Abrams GD, Smallwood LR, Lieber RL, Ward SR (2007) Muscle geometry affects accuracy of forearm volume determination by magnetic resonance imaging (MRI). J Biomech 40(14):3261–3266. https://doi.org/10.1016/j.jbiomech.2007.04.005 Fellows R, Hill N, Gill H, MacIntyre N, Harrison M, Ellis R, Wilson D (2005) Magnetic resonance imaging for in vivo assessment of three-dimensional patellar tracking. J Biomech 38(8): 1643–1652 Fernandez JW, Akbarshahi M, Kim HJ, Pandy MG (2008) Integrating modelling, motion capture and x-ray fluoroscopy to investigate patellofemoral function during dynamic activity. Comput Methods in Biomech Biomed Eng 11(1):41–53. https://doi.org/10.1080/10255840802296814 Finni T, Hodgson JA, Lai AM, Edgerton VR, Sinha S (2006) Muscle synergism during isometric plantarflexion in achilles tendon rupture patients and in normal subjects revealed by velocityencoded cine phase-contrast MRI. Clin Biomech (Bristol, Avon) 21(1):67–74. https://doi.org/ 10.1016/j.clinbiomech.2005.08.007

Cross-Platform Comparison of Imaging Technologies for Measuring. . .

153

Fischer KJ, Manson TT, Pfaeffle HJ, Tomaino MM, Woo SL (2001) A method for measuring joint kinematics designed for accurate registration of kinematic data to models constructed from CT data. J Biomech 34(3):377–383 Fox AM, Kedgley AE, Lalone EA, Johnson JA, Athwal GS, Jenkyn TR (2011) The effect of decreasing computed tomography dosage on radiostereometric analysis (RSA) accuracy at the glenohumeral joint. J Biomech 44(16):2847–2850. https://doi.org/10.1016/j.jbiomech.2011. 08.009 Fregly BJ, Rahman HA, Banks SA (2005) Theoretical accuracy of model-based shape matching for measuring natural knee kinematics with single-plane fluoroscopy. J Biomech Eng 127(4):692–699 Gondim Teixeira PA, Formery AS, Hossu G, Winninger D, Batch T, Gervaise A, Blum A (2017) Evidence-based recommendations for musculoskeletal kinematic 4D-CT studies using wide area-detector scanners: a phantom study with cadaveric correlation. Eur Radiol 27(2):437–446. https://doi.org/10.1007/s00330-016-4362-y Goto A, Leng S, Sugamoto K, Cooney WP, Kakar S, Zhao K (2014) In vivo pilot study evaluating the thumb carpometacarpal joint during circumduction. Clin Orthop Relat Res 472(4): 1106–1113 Guan S, Gray HA, Keynejad F, Pandy MG (2016) Mobile biplane X-ray imaging system for measuring 3D dynamic joint motion during overground gait. IEEE Trans Med Imaging 35(1):326–336. https://doi.org/10.1109/tmi.2015.2473168 Holden JP, Orsini JA, Siegel KL, Kepple TM, Gerber LH, Stanhope SJ (1997) Surface movement errors in shank kinematics and knee kinetics during gait. Gait Posture 5(3):217–227. https://doi. org/10.1016/S0966-6362(96)01088-0 Im HS, Alter KE, Brochard S, Pons C, Sheehan FT (2014) In vivo pediatric shoulder muscle volumes and their relationship to 3D strength. J Biomech 47(11):2730–2737. https://doi.org/ 10.1016/j.jbiomech.2014.04.049 Jan SVS, Salvia P, Hilal I, Sholukha V, Rooze M, Clapworthy G (2002) Registration of 6-DOFs electrogoniometry and CT medical imaging for 3D joint modeling. J Biomech 35(11): 1475–1484 Jensen ER, Morrow DA, Felmlee JP, Odegard GM, Kaufman KR (2015) Error analysis of cine phase contrast MRI velocity measurements used for strain calculation. J Biomech 48(1):95–103. https://doi.org/10.1016/j.jbiomech.2014.10.035 Jia R, Mellon S, Monk P, Murray D, Noble JA (2016) A computer-aided tracking and motion analysis with ultrasound (CAT & MAUS) system for the description of hip joint kinematics. Int J Comput Assist Radiol Surg 11(11):1965–1977. https://doi.org/10.1007/s11548-016-1443-y Kaiser J, Monawer A, Chaudhary R, Johnson KM, Wieben O, Kijowski R, Thelen DG (2016) Accuracy of model-based tracking of knee kinematics and cartilage contact measured by dynamic volumetric MRI. Med Eng Phys 38(10):1131–1135. https://doi.org/10.1016/j. medengphy.2016.06.016 Kalia V, Obray RW, Filice R, Fayad LM, Murphy K, Carrino JA (2009) Functional joint imaging using 256-MDCT: technical feasibility. Am J Roentgenol 192(6):W295–W299 Karlsson D, Lundberg A (1994) Accuracy estimation of kinematic data derived from bone anchored external markers. In: Proceedings of the 3rd international symposium on 3D analysis of human movement, Stockholm, pp 27–30 Kerkhof FD, Brugman E, D'Agostino P, Dourthe B, van Lenthe GH, Stockmans F, Jonkers I, Vereecke EE (2016) Quantifying thumb opposition kinematics using dynamic computed tomography. J Biomech 49(9):1994–1999. https://doi.org/10.1016/j.jbiomech.2016.05.008 Kettelkamp DB, Johnson RJ, Smidt GL, Chao EY, Walker M (1970) An electrogoniometric study of knee motion in normal gait. J Bone Joint Surg Am 52(4):775–790 Lafortune MA (1984) The use of intra-cortical pins to measure the motion of the knee joint during walking. Doctoral dissertation, Pennsylvania State University, University Park Lafortune MA, Cavanagh PR, Sommer HJ 3rd, Kalenak A (1992) Three-dimensional kinematics of the human knee during walking. J Biomech 25(4):347–357

154

R.M. Smith and F.T. Sheehan

Lafortune MA, Cavanagh PR, Sommer HJ 3rd, Kalenak A (1994) Foot inversion-eversion and knee kinematics during walking. J Orthop Res 12(3):412–420. https://doi.org/10.1002/jor. 1100120314 Lee S, Kim YS, Park CS, Kim KG, Lee YH, Gong HS, Lee HJ, Baek GH (2014) CT-based threedimensional kinematic comparison of dart-throwing motion between wrists with malunited distal radius and contralateral normal wrists. Clin Radiol 69(5):462–467. https://doi.org/ 10.1016/j.crad.2013.09.023 Levens AS, Inman VT, Blosser JA (1948) Transverse rotation of the segments of the lower extremity in locomotion. J Bone Joint Surg Am 30a(4):859–872 Lin CC, Lu TW, Shih TF, Tsai TY, Wang TM, Hsu SJ (2013) Intervertebral anticollision constraints improve out-of-plane translation accuracy of a single-plane fluoroscopy-to-CT registration method for measuring spinal motion. Med Phys 40(3):031912. https://doi.org/10.1118/ 1.4792309 Lippert F, Veress S, Takamoto T, Spolek G (1975) Experimental studies on patellar motion using X-ray photogrammetry. In: Proceedings of symposium on close-range photogrammetric systems, Champaign, IL, USA, pp 186–208 Manal K, McClay I, Stanhope S, Richards J, Galinat B (2000) Comparison of surface mounted markers and attachment methods in estimating tibial rotations during walking: an in vivo study. Gait Posture 11(1):38–45 Maniere-Ezvan A, Havet T, Franconi JM, Quemar JC, de Certaines JD (1999) Cinematic study of temporomandibular joint motion using ultra-fast magnetic resonance imaging. Cranio J Craniomandibular Prac 17(4):262–267 Materials ASfTa (2010) Standard practice for use of the terms precision and bias in ASTM test methods, vol E177-08. ASTM International, West Conshohocken Meskers CG, Fraterman H, van der Helm FC, Vermeulen HM, Rozing PM (1999) Calibration of the “flock of birds” electromagnetic tracking device and its application in shoulder motion studies. J Biomech 32(6):629–633 Milne A, Chess D, Johnson J, King G (1996) Accuracy of an electromagnetic tracking device: a study of the optimal operating range and metal interference. J Biomech 29(6):791–793 Moerman KM, Sprengers AM, Simms CK, Lamerichs RM, Stoker J, Nederveen AJ (2012) Validation of continuously tagged MRI for the measurement of dynamic 3D skeletal muscle tissue deformation. Med Phys 39(4):1793–1810. https://doi.org/10.1118/1.3685579 Neptune RR, Hull ML (1995) Accuracy assessment of methods for determining hip movement in seated cycling. J Biomech 28(4):423–437 Niitsu M, Campeau NG, Holsinger-Bampton AE, Riederer SJ, Ehman RL (1992) Tracking motion with tagged rapid gradient-echo magnetization-prepared MR imaging. J Magn Reson Imaging 2(2):155–163 Ohnishi T, Suzuki M, Nawata A, Naomoto S, Iwasaki T, Haneishi H (2010) Three-dimensional motion study of femur, tibia, and patella at the knee joint from bi-plane fluoroscopy and CT images. Radiol Phys Technol 3(2):151–158. https://doi.org/10.1007/s12194-010-0090-1 Pappas GP, Asakawa DS, Delp SL, Zajac FE, Drace JE (2002) Nonuniform shortening in the biceps brachii during elbow flexion. J Appl Physiol (Bethesda, Md: 1985) 92(6):2381–2389. https:// doi.org/10.1152/japplphysiol.00843.2001 Pelc LR, Sayre J, Yun K, Castro LJ, Herfkens RJ, Miller DC, Pelc NJ (1994) Evaluation of myocardial motion tracking with cine-phase contrast magnetic resonance imaging. Investig Radiol 29(12):1038–1042 Pelc NJ, Drangova M, Pelc LR, Zhu Y, Noll DC, Bowman BS, Herfkens RJ (1995) Tracking of cyclic motion with phase-contrast cine MR velocity data. J Magn Reson Imaging 5(3):339–345 Pipe JG, Boes JL, Chenevert TL (1991) Method for measuring three-dimensional motion with tagged MR imaging. Radiology 181(2):591–595. https://doi.org/10.1148/radiology.181.2. 1924810 Pulkovski N, Schenk P, Maffiuletti NA, Mannion AF (2008) Tissue Doppler imaging for detecting onset of muscle activity. Muscle Nerve 37(5):638–649. https://doi.org/10.1002/mus.20996

Cross-Platform Comparison of Imaging Technologies for Measuring. . .

155

Reinschmidt C, van den Bogert AJ, Lundberg A, Nigg BM, Murphy N, Stacoff A, Stano A (1997) Tibiofemoral and tibiocalcaneal motion during walking: external vs. skeletal markers. Gait Posture 6(2):98–109. https://doi.org/10.1016/S0966-6362(97)01110-7 Rogers B, Wiese S, Blankenbaker D, Meyerand E, Haughton V (2005) Accuracy of an automated method to measure rotations of vertebrae from computerized tomography data. Spine 30(6): 694–696 Salarian A, Horak FB, Zampieri C, Carlson-Kuhta P, Nutt JG, Aminian K (2010) iTUG, a sensitive and reliable measure of mobility. IEEE Trans Neural Syst Rehabil Eng 18(3):303–310. https:// doi.org/10.1109/tnsre.2010.2047606 San Juan JG, Karduna AR (2010) Measuring humeral head translation using fluoroscopy: a validation study. J Biomech 43(4):771–774. https://doi.org/10.1016/j.jbiomech.2009.10.034 Schutzer SF, Ramsby GR, Fulkerson JP (1986) The evaluation of patellofemoral pain using computerized tomography: a preliminary study. Clin Orthop Relat Res 204:288–293 Sheehan FT, Zajac FE, Drace JE (1998) Using cine phase contrast magnetic resonance imaging to non-invasively study in vivo knee dynamics. J Biomech 31(1):21–26 Shih YF, Bull AM, McGregor AH, Humphries K, Amis AA (2003) A technique for the measurement of patellar tracking during weight-bearing activities using ultrasound. Proc Inst Mech Eng H J Eng Med 217(6):449–457 Shih YF, Bull AM, McGregor AH, Amis AA (2004) Active patellar tracking measurement: a novel device using ultrasound. Am J Sports Med 32(5):1209–1217. https://doi.org/10.1177/ 0363546503262693 Simoes MA (2011) Feasibility of wearable sensors to determine Gait parameters. Masters thesis, University of South Florida Sprengers AM, Caan MW, Moerman KM, Nederveen AJ, Lamerichs RM, Stoker J (2013) A scale space based algorithm for automated segmentation of single shot tagged MRI of shearing deformation. Magma (New York, NY) 26(2):229–238. https://doi.org/10.1007/ s10334-012-0332-9 Tang TS, MacIntyre NJ, Gill HS, Fellows RA, Hill NA, Wilson DR, Ellis RE (2004) Accurate assessment of patellar tracking using fiducial and intensity-based fluoroscopic techniques. Med Image Anal 8(3):343–351. https://doi.org/10.1016/j.media.2004.06.011 Tashman S (2008) Comments on “validation of a non-invasive fluoroscopic imaging technique for the measurement of dynamic knee joint motion”. J Biomech 41(15):3290–3291. https://doi.org/ 10.1016/j.jbiomech.2008.07.038. author reply 3292–3293 Tashman S, Anderst W (2003) In-vivo measurement of dynamic joint motion using high speed biplane radiography and CT: application to canine ACL deficiency. J Biomech Eng 125(2): 238–245 Tat J, Kociolek AM, Keir PJ (2015) Validation of color Doppler sonography for evaluating relative displacement between the flexor tendon and subsynovial connective tissue. J Ultrasound Med 34(4):679–687. https://doi.org/10.7863/ultra.34.4.679 Telfer S, Woodburn J, Turner DE (2014) An ultrasound based non-invasive method for the measurement of intrinsic foot kinematics during gait. J Biomech 47(5):1225–1228. https://doi. org/10.1016/j.jbiomech.2013.12.014 Thorhauer E, Tashman S (2015) Validation of a method for combining biplanar radiography and magnetic resonance imaging to estimate knee cartilage contact. Med Eng Phys 37(10):937–947. https://doi.org/10.1016/j.medengphy.2015.07.002 Todorov E (2007) Probabilistic inference of multijoint movements, skeletal parameters and marker attachments from diverse motion capture data. IEEE Trans Biomed Eng 54(11):1927–1939. https://doi.org/10.1109/tbme.2007.903521 Townsend MA, Izak M, Jackson RW (1977) Total motion knee goniometry. J Biomech 10(3):183–193 Tranberg R, Saari T, Zügner R, Kärrholm J (2011) Simultaneous measurements of knee motion using an optical tracking system and radiostereometric analysis (RSA). Acta Orthop 82(2): 171–176. https://doi.org/10.3109/17453674.2011.570675

156

R.M. Smith and F.T. Sheehan

Veress SA, Lippert FG, Hou MC, Takamoto T (1979) Patellar tracking patterns measurement by analytical x-ray photogrammetry. J Biomech 12(9):639–650 Wan EA, Nelson AT (2001) Chapter 5- dual extended Kalman filter methods. In: Haykin S (ed) Kalman filtering and neural networks. New York, NY: Wiley, pp 123–173 Wang S, Passias P, Li G, Li G, Wood K (2008) Measurement of vertebral kinematics using noninvasive image matching method-validation and application. Spine 33(11):E355–E361. https://doi.org/10.1097/BRS.0b013e3181715295 Wang B, Roach KE, Kapron AL, Fiorentino NM, Saltzman CL, Singer M, Anderson AE (2015) Accuracy and feasibility of high-speed dual fluoroscopy and model-based tracking to measure in vivo ankle arthrokinematics. Gait Posture 41(4):888–893. https://doi.org/10.1016/j. gaitpost.2015.03.008 Ward SR, Shellock FG, Terk MR, Salsich GB, Powers CM (2002) Assessment of patellofemoral relationships using kinematic MRI: comparison between qualitative and quantitative methods. J Magn Reson Imaging 16(1):69–74. https://doi.org/10.1002/jmri.10124 Williams AA, Elias JJ, Tanaka MJ, Thawait GK, Demehri S, Carrino JA, Cosgarea AJ (2016) The relationship between tibial tuberosity–trochlear groove distance and abnormal patellar tracking in patients with unilateral patellar instability. Arthrosc J Arthrosc Relat Sur 32(1):55–61 Wilson NA, Press JM, Koh JL, Hendrix RW, Zhang LQ (2009) In vivo noninvasive evaluation of abnormal patellar tracking during squatting in patients with patellofemoral pain. J Bone Joint Surg Am 91(3):558–566. https://doi.org/10.2106/jbjs.g.00572 You BM, Siy P, Anderst W, Tashman S (2001) In vivo measurement of 3-D skeletal kinematics from sequences of biplane radiographs: application to knee kinematics. IEEE Trans Med Imaging 20(6):514–525. https://doi.org/10.1109/42.929617 Zhao K, Breighner R, Holmes D 3rd, Leng S, McCollough C, An KN (2015) A technique for quantifying wrist motion using four-dimensional computed tomography: approach and validation. J Biomech Eng 137(7). https://doi.org/10.1115/1.4030405 Zuhlke T, Fine J, Haughton VM, Anderson PA (2009) Accuracy of dynamic computed tomography to calculate rotation occurring at lumbar spinal motion segments. Spine 34(6):E215–E218. https://doi.org/10.1097/BRS.0b013e318199700d

Ultrasound Technology for Examining the Mechanics of the Muscle, Tendon, and Ligament Glen Lichtwark

Abstract

Ultrasound imaging provides a means to look inside the body and examine how tissues respond to mechanical stress or muscle contraction. As such, it can provide a valuable tool for understanding how muscle, tendon, and ligament mechanics influence the way we move, or vice versa, in health and disease, or to understand how and why these tissues might get injured due to chronic or acute loading. This chapter explores the basic concepts of ultrasound and how it can be used to examine muscle, tendon, and ligament structure and mechanical function. It introduces different techniques, like conventional B-mode imaging, threedimensional ultrasound, and various forms of elastography that can be used to quantify geometrical and mechanical properties of the muscle, tendon, and ligament. Furthermore, methods to quantify muscle and tendon mechanical function during dynamic human movement are explored, and recommendations provided on which techniques are most suitable for different biomechanical investigations. Finally, some predictions about how new ultrasound imaging technologies might continue to advance our understanding of human motion are proposed and explored. Keywords

Biomechanical imaging • Stress • Strain • 3D ultrasound • Tissue tracking • Elastography

G. Lichtwark (*) Centre for Sensorimotor Performance, School of Human Movement and Nutrition Sciences, The University of Queensland, St Lucia, QLD, Australia e-mail: [email protected] # Springer International Publishing AG, part of Springer Nature 2018 B. Müller, S.I. Wolf (eds.), Handbook of Human Motion, https://doi.org/10.1007/978-3-319-14418-4_156

157

158

G. Lichtwark

Contents Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . State of the Art . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Static Measurement of Muscle, Tendon or Ligament Architecture . . . . . . . . . . . . . . . . . . . . . . . . . . . . The Basics of Conventional B-Mode Ultrasound . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Measuring Relevant Biomechanical Parameters . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Extended Field of View Ultrasound . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Three-Dimensional Ultrasound . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . In Vivo Mechanical Properties of the Muscle, Tendon, and Ligament . . . . . . . . . . . . . . . . . . . . . . . . Dynamic Tissue Response to Forces and/or Movement . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Elastography and Tissue Strain Measurement . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Shear-Wave Imaging . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . In Vivo Determination of the Muscle and Tendon Length Changes . . . . . . . . . . . . . . . . . . . . . . . . . . . Dynamic Imaging of Muscle . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Dynamic Imaging of the Tendon . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Dynamic Imaging of Ligaments . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Other Applications for Ultrasound in Human Motion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Future Directions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

158 159 159 159 162 163 163 165 165 166 166 168 168 170 172 172 172 173

Introduction A key area of basic, clinical, and applied biomechanics is the quantification of soft tissue mechanical properties and deformation of such tissues in response to force or motion (see chapter ▶ “Simulation of Soft Tissue Loading from Observed Movement Dynamics”). Of particular interest are the mechanical properties of the muscle, tendon, and ligament, as these tissues play a critical role in enabling efficient, powerful, or highly precise movement. In addition, these tissues are also often injured in response to either acute or chronic loads. Ultrasound imaging provides a noninvasive and in vivo method for examining the biomechanical properties and function of the muscle, tendon, and ligament in humans. The dynamic function of the muscle, tendons, and ligament in human motion has traditionally been inferred based on the mechanical properties of tissue that is harvested from human or animal specimens and the estimated forces that these tissues experience during movement. For instance, in a classic study Alexander and Bennet-Clark (1977) estimated the elastic energy that is stored and returned from the Achilles tendon during running by estimating the forces applied to the tendon (through inverse dynamics and estimates of tendon moment arm) and stressstrain properties of the tendon. This approach reveals that much of negative work in early stance is actually stored in the tendon (rather than dissipated by the muscle) and then returned in late stance (rather than being generated by the muscle). Therefore understanding the mechanical interaction of the muscle, tendons, and ligaments is essential to understanding the energetics of human movement. In general, muscles, ligaments, and tendons are commonly characterized by similar compositions across different people (e.g., fiber-type composition in muscle, collagen type and content in the tendon and ligament). However, different

Ultrasound Technology for Examining the Mechanics of the Muscle, Tendon, and. . .

159

individuals do display variations in the precise material compositions and structural makeup of the muscle, tendon, and ligament, which can ultimately influence the mechanical function of these tissues. Variation in mechanical properties of soft tissues in the human body are likely to depend on extrinsic factors, like loading commonly experienced by tissues during activities of daily living or athletic training, intrinsic factors like gender, age, body size/structure, and various clinical pathologies or genetic factors. Techniques that can help determine which of these extrinsic and intrinsic factors are important for performance or injury risk is a key area of research in human motion. Ultrasound imaging provides a relatively affordable and low-risk method to look inside the body and examine the mechanical properties of soft tissues like the muscle, tendon, and ligament.

State of the Art Ultrasound imaging can be used for three broad purposes in investigating the mechanics of the muscle, tendon, and ligaments: (1) measurement of tissue architecture to infer mechanical properties, (2) direct (in vivo) determination of tissue mechanical properties, and (3) direct measurement of tissue function during force production or movement. Advances in ultrasound technology, including improvements in image collection and processing, extended field of view measurements, 3D ultrasound imaging, elastography, shear-wave imaging, and advanced methods to track tissue deformation will be discussed below with respect to achieving the above broad purposes.

Static Measurement of Muscle, Tendon or Ligament Architecture A primary determinant of the basic mechanical properties of the muscle, tendon, and ligament is the architecture or geometry of the tissue. For instance, the maximum force generating capacity of muscle is strongly related to its physiological crosssectional area (PCSA, muscle volume divided fiber length (Powell et al. 1984)) and the peak stress of the tendon and ligaments is directly related to its cross-sectional area (Pollock and Shadwick 1994). Ultrasound imaging allows for low-risk, in vivo measurements of the geometry of such tissue, which can be useful for understanding potential performance of such tissues.

The Basics of Conventional B-Mode Ultrasound Ultrasound imaging has been used as a medical diagnosis tool for characterizing tissue dimensions since the 1940s and was quickly adopted for use in musculoskeletal assessment (Kane et al. 2004). Ikai and Fukunaga (1968) used ultrasound to report the relationship between human biceps brachii muscle cross-sectional area

160

G. Lichtwark

Transverse plane (top view)

Sagittal plane (side view)

Image plane

Transducer

Transducer

Fig. 1 Conceptual diagram of scanning plane viewed from the sagittal plane (left) and transverse plane (right), where the transducer is held horizontally to produce a transverse plane image of the underlying tissues (grayscale image overlaid on transverse view). Note that depth is not sufficient to view through the leg and the width is constrained by the size of the transducer

and strength in 1968. However, the systematic use of ultrasound to study the geometry of muscles, tendon, or ligaments for the purpose of understanding human motion and performance really only became prominent in the 1990s (Kallinen and Suominen 1994; Kawakami et al. 1993). Brightness mode, or B-mode, imaging is the most common ultrasound imaging modality, as it can generate an image which is equivalent to a cross section through the tissues in which the ultrasound transducer is imaging (Fig. 1a, b). Tissues that have higher echogenicity (the ability to reflect a sound wave), such as tendinous connective tissue, show up as white on B-mode scans, while tissues with low echogenicity show up as black or gray. As such, contrasts between tissues of different echogenicity are easily identifiable. Muscle fibers typically have low echogenicity, due to their high water content, while the connective tissue that binds the muscle (fascia/aponeurosis) and binds the fibers as fascicles (perimyosim) has high echogenicity (Fig. 2a, b). As such, muscle fascicles can be seen as striated patterns within muscles when the fascicles are viewed within the plane of the image (Fig. 2b). Tendons and ligaments typically have high collagen content, which is relatively highly echogenic; therefore this tissue, and particularly the borders, typically shows up as white pixels in the image (Fig. 2c, d). Ligaments are often less echogenic than tendons, partially because of the less regular filament pattern and the difficulty in aligning to these structures through an image plane. The echogenicity is also dependent on the frequency of the sound waves emitted by the transducer, with superficial structures like the tendon and ligaments optimally imaged with frequencies between 10–12 MHz and deeper muscle at lower frequencies (e.g., 6–8 Hz). The reflection of sound waves depends not only on the composition of the tissue but also on the orientation of the tissue relative to the direction of travel of the sound waves. The “angle of incidence” is the angle at which the sound waves encounter the

Ultrasound Technology for Examining the Mechanics of the Muscle, Tendon, and. . .

161

c Tendon – transverse plane

a Muscle – transverse plane Skin

Stand-off pad

Skin/tissue

Subcutaneous fat Muscle tissue

Muscle fascia

Achilles tendon

Muscle tissue

d Tendon – sagittal plane

b Muscle – sagittal plane Tendon Muscle

Achilles tendon Muscle – tendon junction

Muscle tissue

Muscle fascia aponeurosis

Fig. 2 Ultrasound images of the muscle (a–b) and tendon (c–d) in both transverse and sagittal plane relative to the leg. Bright white regions indicate tissues with high echogenicity (e.g., muscle fascia in (a), muscle fascicles in (b)). To ensure contact with the skin with a flat transducer, a standoff pad can be used between the transducer and the tissue, which deforms around the region and allows sound waves to be transmitted, as used in (c)

surface of the tissue of interest (Ihnatsenka and Boezaart 2010) and is optimum when the direction of the structure is approximately perpendicular to the direction of travel of the sound waves. As the angle of incidence increases, so that the sound waves become more and more parallel to the surface of the tissue, the amount of reflection decreases and therefore reduces the definition of the structure within the tissue, as more of the sound waves are scattered and less reflected (Fig. 3a). Therefore to enhance the clarity of structures within an image, the transducer should be positioned such that it is as perpendicular to the structures of interest as possible, or additionally in many modern ultrasound machines the angle at which the sound waves are transmitted can be changed relative to the transducer to enhance the image quality (Fig. 3b). Understanding how the angle of incidence influences image quality is important, because in some instances the surfaces of tissues may have a circular shape (e.g., imaging a tendon cross section, Fig. 3a) or the tissues may move (e.g., during muscle contraction) and hence it is unavoidable that the image quality may not be optimum across the image or over time. Therefore the correct placement of the transducer to maximize the image quality for the purposes of the measurement is essential for measuring muscle, tendon, or ligament architecture or function. Because ligaments often lie between bones, it is often only possible to image superficial ligaments

162

G. Lichtwark

a

b

Transverse view of Achilles tendon (cross section)

Clear echo

Dispersed echo

Conventional

Steering Angle = -10 °

Steering Angle = 10 °

Fig. 3 (a) Conceptual diagram of effect of the angle of incidence or angle of reflecting tissue relative to the sound wave transmission direction. When the tissues are less orthogonal to the sound waves, the reflected sound is dispersed and is not detected by the ultrasound transducers receiving crystals. This can decrease the clarity of borders of tissues that effectively increase the angle of incidence, like tendons in cross section. (b) Pennate muscle fascicles act at an angle to the direction of the sound waves, which reduces the signal strength. Changing the angle of the sound waves through changing the steering angle or by adjusting the direction of the transducer can improve or degrade the clarity of the fascicles, depending on the direction of the fascicles

(e.g., collateral ligaments of the knee) or small part of ligaments between bones. Therefore a critical limiting factor in using ultrasound is its dependence on a good imaging site of the structures relative to the skin, which limits the structures that can be accurately imaged.

Measuring Relevant Biomechanical Parameters The main architectural parameters of interest that can be measured statically in the muscle, tendons, and ligaments are measures of thickness, cross-sectional area, and length (e.g., fascicle or ligament length) (see chapter ▶ “Cross-Platform Comparison of Imaging Technologies for Measuring Musculoskeletal Motion”). The borders of a muscle, tendon, and ligament contain connective tissue that is reasonably echogenic and therefore relatively easily identifiable. Therefore it is often possible to visualize clear cross sections through these structures in different planes (Fig. 2). However, the

Ultrasound Technology for Examining the Mechanics of the Muscle, Tendon, and. . .

163

size of the cross section that is visible is limited by the length of the imaging transducer, which defines the maximum width of the image. Therefore for larger muscles or longer tendons (e.g., Achilles) or ligaments (e.g., plantar fascia), it is hard to capture the entire cross section in either the sagittal or transverse planes. While the field of view of ultrasound can limit the potential structures that can be measured, new technologies have provided solutions to overcome this limitation.

Extended Field of View Ultrasound One method to overcome the limitation of the field of view of the transducer is to use ‘extended field of view’ or ‘panoramic’ imaging methods that require the user to move the transducer along a straight line to image consecutive regions of the muscle, which can be stitched back together using image processing techniques (Cooperberg et al. 2001). Such imaging methods have been shown to have good validity and reliability for measuring lengths of muscle fibers and cross-sectional area in various large human muscles and in different planes of imaging (Noorkoiv et al. 2010a, b); however the valid reconstruction of the plane is highly dependent on the ability to move the transducer in a single plane on the surface of the structure.

Three-Dimensional Ultrasound Three-dimensional (3D) imaging of soft tissues makes it possible to make geometrical measurements in 3D space without the limitations of measurements made in a single plane. Freehand 3D ultrasound is a technique that uses conventional two-dimensional imaging (e.g., B-mode), however utilizes multiple image slices along a structure to reconstruct the area of interest, much in the same way that magnetic resonance imaging works. To overcome the issue of the image planes not being perpendicular from one image to the next, it is necessary to track the orientation and position of the transducer and apply a known calibration of the position of the image relative to the transducer, so that images can be correctly projected into the 3D space and accurate voxel information generated (Treece et al. 2003). This method is particularly useful as it allows small spaces between image slices and hence has good spatial resolution, although this is subject to the resolution of the instruments used to track the position and orientation of the transducer (e.g., magnetic or optical systems). Freehand 3D ultrasound also makes it possible to make volumetric measures (e.g., muscle volume; Fig. 4) as well as examine the geometry of large tissue structures that may not be imaged within the field of view of the transducer in conventional B-mode ultrasound [e.g., muscle aponeurosis (Raiteri et al. 2016), tendon length, and cross-sectional area (Obst et al. 2014)]. Three-dimensional ultrasound imaging of smaller volumes can also be performed in real time using specialized transducers. The first method involves capturing a small volume by mechanically sweeping a linear array through an angular motion and

164

G. Lichtwark

b 3D image location

a Transvers scan image

Sagittal reconstruction plane

Transverse image scan

Tibialis anterior muscle border Reconstructed muscle volume

c

Sagittal reconstruction Current image plane

Tibialis anterior muscle border

Fig. 4 Freehand 3D ultrasound uses conventional B-mode images that are collected sequentially along the length of the tissue of interest while the position and orientation of the image is recorded so that the images can be stacked together to generate a 3D volume. (a) Transverse scan of the tibialis anterior muscle in mid-region of muscle. (b) Position of transverse scan relative to muscle and a 3D reconstruction of the muscle volume created from segmenting the muscle borders in sequential images along the muscle. (c) Sagittal plane reconstruction of the muscle through the mid-region

reconstructing the volume in a similar manner to freehand 3D ultrasound. This method is typically used in obstetrics and gynecology, but has found limited use at the present for biomechanics, possibly because of the limited capture volume. However, for imaging small ligamentous structures or sites of muscle injury, this method may be useful because it can collect and reconstruct data in near real time. The second method developed was real-time 3D ultrasound, which uses a matrix of ultrasound elements (instead of an array) to reconstruct volumes at discrete time points; hence this method is often known as four-dimensional (4D) ultrasound (three spatial dimensions and a time dimension). Because of the large amount of data that must be transmitted at high frequencies, the size of the matrix is currently limited and this technology is primarily used to examine cardiac function (e.g., valve mechanics); however there has been some investigation on the function of the pelvic floor muscles (Braekken et al. 2009). This technique should be considered further in both static and dynamic measurement of small muscle, tendon, and ligaments where it is well suited.

Ultrasound Technology for Examining the Mechanics of the Muscle, Tendon, and. . .

165

In Vivo Mechanical Properties of the Muscle, Tendon, and Ligament While static measures of soft tissue geometry are valuable for biomechanical assessment of musculoskeletal capacity, the true value of ultrasound comes from being able to determine subject-specific material properties of tissues (see chapters ▶ “Induced Acceleration and Power Analyses of Human Motion,” ▶ “Optimal Control Modeling of Human Movement,” and ▶ “Physics-Based Models for Human Gait Analysis”). Such imaging allows for the characterization of how soft tissues adapt in the event of different loading or exercise, various clinical conditions, and also across the life span. This information is important for understanding capacity to perform movements, requirements for rehabilitative interventions, or prevention of injury.

Dynamic Tissue Response to Forces and/or Movement B-mode ultrasound is the most accessible tool to assess deformation or changes in geometry of human muscle, tendon, and ligaments. Conventional ultrasound machines have good temporal (time) resolution, with frame rates that vary from 10–100 frames per second, depending on imaging parameters, scan depth, and computer processing power. This allows users to record changes in geometry during tasks where tissues are deformed (e.g., muscle contraction, passive stretching). One of the first demonstrations of this capacity was measurement of muscle fascicle length changes during passive length changes (Herbert and Gandevia 1995; Narici et al. 1996) and during isometric contractions (Fukunaga et al. 1997; Fukashiro et al. 1995). These studies clearly showed that human fascicle length changes are not necessarily concomitant with the whole muscle-tendon unit length. This is because the muscle fibers connect to the skeleton via elastic tendons which stretch when force (either passive or active) is applied by the muscle. Utilizing this knowledge, the first in vivo estimates of tendon stiffness in muscles of the lower limb (e.g., gastrocnemius, tibialis anterior) were reported through measuring the shortening of muscle or movement of the muscle-tendon junction during isometric contractions, which was assumed to be the equivalent of the stretch of the elastic tendons (Fukashiro et al. 1995; Maganaris and Paul 1999). Providing that adequate estimates of forces applied to tissues can be determined, typically through measurement of external forces/torques and estimates or direct measurement of muscle moment arms (Maganaris 2005), then these measurements can be used to provide estimates of stress versus strain relationships of both the muscle and tendon. There are numerous limitations to using fascicle length changes during isometric contractions to infer strain or material properties of external tendons. One of the main limitations is that it is difficult to prevent rotation of joints during isometric contractions, which induce fascicle shortening independent of tendon strain (Maganaris 2005; Karamanidis et al. 2005). Another difficulty is the ability to accurately estimate muscle forces using external force measurement techniques

166

G. Lichtwark

(Lichtwark et al. 2013) or ability to accurately synchronize the force and length change data (Finni et al. 2013). Detailed methods to correct for joint rotation errors have been developed (Karamanidis et al. 2005), along with methods to track the movement of the ultrasound transducer to limit errors in displacement measurements and more accurately synchronize signals; however it remains uncertain whether strain measured at one end of a tissue is representative of strain occurring throughout the tissue. In some tendons, like the patella tendon, it is possible to image the entire tendon within the field of the tendon, and hence this has served as a good model to understand tendon adaptation to exercise (Hansen et al. 2006; Onambele et al. 2007; Pearson et al. 2007). However, recent advancements in ultrasound technology have concentrated on methods to examine local strains or material properties of the muscle, tendon, and ligament.

Elastography and Tissue Strain Measurement Ultrasound elastography was first developed as a method to distinguish strain of soft tissues in response to compression. Tissues that strain more for a given compressive force are considered more compliant or elastic. Ultrasound is an ideal medium for examining this mechanical response because the radio frequency (RF) data received from the return sound wave is in the same line of action as the line of compression of the tissue and hence will also compress in response to the tissue deformation. Changes in the compression across a region are indicative of different tissue stiffness values and can be quantified visually using a map overlaid across the image. Methods to compress the tissue are varied and include quasi-static or hand-driven compression, as well as mechanical compression or vibrations or an acoustic radiation force (force generated using the transducer itself) (for reviews of these methods see Varghese 2009; Treece et al. 2011; Nightingale 2011). While all of these methods can be used to quantify the strain of tissues in response to the force, it is difficult to quantify the material properties of the tissue (e.g., Young’s modulus), and these methods only quantify the stiffness of the tissue in the direction of the sound-wave beams. For tissues such as muscles and ligaments, where the interesting material properties are often orthogonal to the sound-wave beams (or parallel with the skin where the line of action of these tissues is usually most prominent), various other exciting methods have been developed.

Shear-Wave Imaging Shear-wave imaging is a form of elastography that utilizes a mechanical perturbation to the tissue, but instead of determining the compression of the tissue, the speed of the resulting propagation of the mechanical shear wave along the tissue is examined through speckle tracking (effective movement of the tissues) across the image. The technique most commonly used to examine the muscle, ligament, and tendon is known as “super-sonic shear imaging” or SSI (Hug et al. 2015). This technique uses

Ultrasound Technology for Examining the Mechanics of the Muscle, Tendon, and. . .

167

Fig. 5 Example of supersonic shear imaging (SSI) technique. The bottom image is a conventional B-mode image of the muscle, and the top image has the shear modulus map of a particular region of interest. In this case the muscle is contracting slightly, which increases the shear modulus of the muscle tissue

an acoustic radiation force and ultrafast imaging to quantify the wave propagation speed in real time. Making some assumptions about the tissue density enables a region-specific quantification of shear modulus of the tissue (for full review see Hug et al. (2015), example image in Fig. 5). Quantification of the shear modulus of tissues has been shown to have many biomechanical applications that are useful for understanding human motion. For instance, there is considerable evidence showing that the average shear modulus across an area of muscle is highly related to muscle isometric force across a range of forces that can be assessed (Ates et al. 2015; Bouillard et al. 2012). Therefore, this technique may be a valuable method to understand which muscles contribute to forces that generate human motion (Hug et al. 2015). However, at present the low acquisition rate and low saturation level limit the potential to look at dynamic muscle contractions (Hug et al. 2015). SSI has also been used extensively to examine differences in material properties in the tendon (Helfenstein-Didier et al. 2016;

168

G. Lichtwark

Hug et al. 2013). For instance, the shear modulus of young and old Achilles tendon seems to be different (Slane et al. 2016); however, due to the low saturation level, it is only possible to look at the tendon at very low (often passive) forces. Currently the technology has limitations in being able to understand dynamic muscle, tendon, or ligament function during human motion; however, as this technology improves, it will likely become an invaluable tool for understanding the material properties of the muscle, ligament, and tendon in both healthy and clinical populations.

In Vivo Determination of the Muscle and Tendon Length Changes Ultrasound has a relatively high temporal resolution for characterizing strains of the tissue, and the imaging location is not constrained because the transducer is freely moveable. As such, it is possible to use ultrasound imaging to examine the muscle and tendon length changes during muscular contractions and movement and therefore assess muscle mechanical function of individual muscles.

Dynamic Imaging of Muscle The early measures of muscle fascicle length changes in response to changes in joint angle or isometric contractions clearly demonstrated that the muscle fascicles apparently changed length in a manner that wasn’t consistent with the whole muscletendon unit (Fukashiro et al. 1995; Fukunaga et al. 1997; Kawakami et al. 1998; Narici et al. 1996, Herbert and Gandevia 1995). Since this time, it has become apparent that the length changes of muscle fascicles may be very different to the length changes of the muscle-tendon unit, particularly in the lower limb muscles like the gastrocnemius or soleus. For instance, during the early to mid-stance phase of human walking, it has been demonstrated that muscle fascicles operate relatively isometrically while the muscle-tendon unit is lengthened (Fukunaga et al. 2001; Lichtwark and Wilson 2006). As such, it can be concluded that the tendon tissue must be stretching to store elastic energy that can later be used to help power propulsion. Because muscle fascicles are relatively simple to image through the skin, it is possible to measure dynamic length changes during human motion; however there are numerous technical considerations that must be considered. Firstly, for such imaging to be successful, one must ensure that the ultrasound transducer remains in a similar plane to that in which the fascicles lengthen and shorten. Even in static measures of muscle length, it can be difficult to do this in a way that accurately measures muscle fascicle length. Firstly, this requires that most (if not all) of the length of the muscle fascicle is imaged within the field of view. This limits the size of the muscles that are capable of being imaged. Many of the human lower limb muscles have fiber lengths that are less than the width of the transducer and hence are suitable for dynamic imaging. Secondly, the image plane must be aligned with the line of the fascicles. Recent comparisons between ultrasound imaging in different

Ultrasound Technology for Examining the Mechanics of the Muscle, Tendon, and. . .

169

planes and magnetic resonance imaging (more specifically diffusion tensor imaging) in the gastrocnemius muscle concluded that errors in fascicle length measurement of up to 20% could be found with misalignment of the transducer (Bolsterlee et al. 2016a, b). These studies have also found that the best location was approximately perpendicular to the skin, parallel to the tibia, and in the mid-region of the muscle. Whether this alignment remains consistent through a dynamic task is difficult to determine. The general rule of thumb in imaging muscle fascicle lengths during dynamic tasks is that the lines that constitute the connective tissue around the fascicles should remain continuous and clear, as should the fascia to which the muscle connects, throughout the movement (see Fig. 2b). During dynamic contraction the muscle fascicles shorten and their pennation angle increases. As such the image quality decreases due to the change in the angle of incidence (Fig. 3b). This can make it difficult to interpret whether the fascicles are still in the plane of the image. A secure attachment to the site of imaging, such that there is little movement or rotation relative to the skin, is a key requirement. Various flat-shaped ultrasound transducers that can be strapped to the leg seem to be best suited to such tasks (e.g., T-shaped or veterinary rectal transducers). However, as secure attachment requires pressure, this can lead to artificial changes in muscle geometry which need to be considered as they may impact results (Wakeling et al. 2013). Temporal resolution is an important factor when assessing ultrasound data. As is the case with any analysis of human motion, the capture rate must be sufficient to detect the event of interest (typically greater than twice the Nyquist frequency). The rate at which ultrasound machines can collect a complete frame of B-mode ultrasound data depends on many factors including the machine and transducers (processing power, size, and resolution of the transducer) and the imaging settings (e.g., depth of image, image processing techniques). The range of acquisition speeds varies from 5 to 10, 000 frames per second. For slow movements (e.g., passive rotations of joints) a slow frame rate is sufficient, but for activities that require high temporal resolution (e.g., high speed running, impact during landing), a high frame rate is essential. A novel application of ultrahigh speed ultrasound has been the ability to accurately measure the electromechanical delay – the delay between an electrical impulse transmitted along a muscle to induce contraction and the time at which force is transmitted. By examining the timing of length changes of the muscle fibers, which represents force transmission due to stretch of the connective tissue, Nordez and colleagues (Nordez et al. 2009) were able to determine that the electromechanical delay is as small as 6 milliseconds in the ankle plantar flexor muscles, although the time taken to transmit the force along a long tendon like the Achilles may contribute to further delays in force transmission. Automatic tracking of muscle fascicle length changes during dynamic human motion is one major advance that has reduced the time required to assess dynamic length changes. There have been numerous different approaches. The most common approach has been to attempt to track homologous structures from one image frame to the next, either through cross-correlation or optic flow techniques (Korstanje et al.

170

G. Lichtwark

2010; Lee et al. 2008; Loram et al. 2006). These techniques are useful especially for slower movements or those where it is possible to visualize the same structures across consecutive frames. Another method that is commonly used is an optic flow algorithm with an affine fit to estimate deformation across a region of the muscle (Cronin et al. 2011; Farris and Lichtwark 2016; Gillett et al. 2013). Optic flow algorithms create a vector field estimate of the displacement of multiple regions across an image, from one frame to the next. An affine transformation (horizontal and vertical displacement, rotation, dilation, shear in horizontal and vertical direction) can then be fit to the vector field so as to create a smoothed distortion map from one image to the next based on movement across the entire image or area of interest. The distortion map can then be applied to any points within or outside the image, for example, the end points of fascicles, fascia, or other regions of interest. This has been shown to be a useful method for tracking dynamic tasks like walking or isometric contraction. However, the major limitation to these frame-by-frame approaches are that small errors in frame-by-frame estimations of movement can accumulate over time and cause the lengths to drift. High frame rates (such that the movement between frames is small, as is the potential error) can alleviate tracking errors, while other methods have been proposed to correct drift (Farris and Lichtwark 2016); however there are cases where it may be more suitable to treat individual frames as separate problems and to identify structural measurements directly from individual images (e.g., aponeurosis locations, average pennation angle). For instance, there have been attempts to quantify pennation angle automatically in individual frames (Rana et al. 2009) or machine learning algorithms that can detect the length of regions of fascicles (Darby et al. 2012). However such approaches are still subject to noise and can be highly time consuming (e.g., having to train a data set for subsequent analysis) and hence the algorithm of choice is very much dependent on the question being asked and accuracy required.

Dynamic Imaging of the Tendon There has been less focus on examining dynamic function of the tendon when compared to the muscle. As was the case described earlier, much of the literature examining dynamic function of the tendon (e.g., strain responses during movement) have been deduced based on measures of muscle fascicle length changes. Estimates of muscle-tendon length changes are determined using kinematics and various models based on cadaveric data (Hawkins and Hull 1990; Grieve et al. 1978) or geometrical models (Delp et al. 2007), and the difference between length changes of the muscle-tendon unit and the fascicles (often corrected for by the pennation angle) is attributed to the strain of the tendinous tissues (Fukunaga et al. 2001). While this approach gives a global understanding of the function of tendinous tissue, its application to understanding differences between or within different populations is limited because it is hard to assess where potential differences lie within a tendon (regionally within external tendon or within muscle tendinous tissues like aponeurosis).

Ultrasound Technology for Examining the Mechanics of the Muscle, Tendon, and. . .

171

One method to determine the length changes of tendons is to examine the movement of the end points. As previously mentioned, in tendons like the patella tendon, that are reasonably short, it may be possible to image the end points of the tendon to assess overall strain (Hansen et al. 2006), although this author is unaware of any publications where this has been done outside of isometric contractions. An alternative for longer tendons, like the Achilles tendon, is to image the muscle-tendon junction (one end point of the tendon, Fig. 2b) and track the position of this junction within the image while also tracking the position and orientation of the image using conventional motion capture techniques (e.g., motion capture markers mounted on the transducer) (Lichtwark and Wilson 2005). This provides a dynamic measure of where the proximal end of the tendon is located, and this can be combined with information about where the tendon inserts onto the skeleton (e.g., marker placed on a bony landmark) and a length between the two points determined. While this is a useful technique for examining tendon strains during dynamic tasks like walking, running, or hopping, it is still limited in that it ignores factors like curvature and/or spiral twists (Obst et al. 2014) in tendons and does not give an indication of regions of high strain along the tendon. One of the most exciting new developments in tendon research has been the development of new elastography methods that use the natural deformation of the tendon during dynamic tasks (i.e., when forces are applied or removed from the tendon) to estimate strain in local regions. Using speckle tracking algorithms that require the raw radio frequency data and which have been customized and validated for estimating tendinous tissue strain (Chernak Slane and Thelen 2014), it has now become possible to examine dynamic tendon strain in the Achilles tendon during dynamic tasks like eccentric contracts (Slane and Thelen 2014) or walking (Franz et al. 2015) (Fig. 6). The potential for this technology to look at region-specific strains (including regions attaching to the gastrocnemius vs. soleus muscles) in various populations (Franz and Thelen 2015) and in different tendons is likely to provide the next major advance in understanding the relationships between movement and tissue strain. However, as was the case with the dynamic muscle imaging, there needs to be numerous considerations

Fig. 6 Example of dynamic speckle tracking (elastography) of the Achilles tendon during dynamic contractions (eccentric). The transducer is placed over along the length of the tendon (imaging in the sagittal plane), and the displacement of individual nodes is tracked using the radio frequency content from the raw ultrasound signal. Note the difference in regional movement from superficial to deep parts of the Achilles tendon (Figure from Slane and Thelen 2014. Permission from Elsevier (License Number 3976471248270))

172

G. Lichtwark

about the potential accuracy of measures depending on factors like transducer movement and misalignment. Dynamic elastography techniques where tissue strain is induced by movement or muscle contraction also has potential for assessing local muscle and ligament strains; however there is currently little or no research in this area.

Dynamic Imaging of Ligaments There is very little literature using ultrasound to examine the dynamic strain of human ligaments. Ultrasound is routinely used as a diagnostic tool to assess ligamentous damage; however there is a dearth of literature on the mechanical function of ligaments in dynamic activities. This may partially be because many ligaments are difficult to image because they reside in locations with high concentrations of bone, which can cause large distortions and unwanted reflections within the image. It may also be because it is difficult to maintain the position of an ultrasound transducer when attached to a joint. Finally, ligaments often twist and turn around bones and hence it may be difficult to align a transducer well with the ligaments. However, the potential to use methods already established in tendon research, including tracking the end points of the bones (Hansen et al. 2006) or using elastography, provides a large avenue of research for understanding ligament function in human motion.

Other Applications for Ultrasound in Human Motion While ultrasound has traditionally been used as an imaging modality to assess soft tissue, it is also increasingly being used to assess bone geometry and bone motion. For instance, 3D ultrasound techniques are being used to assess the location of bony landmarks or bone surfaces (Jia et al. 2016; Passmore and Sangeux 2016) or to assess the position of the joint center of rotation (Peters et al. 2010). Alternatively, measuring the position and orientation of an ultrasound transducer while imaging bony landmarks within the plane of the image can also be used to determine bony translation and/or rotation beneath the surface of the skin (Telfer et al. 2014). This method could potentially be used as a way to correct for soft tissue movement artifact associated with conventional 3D motion analysis techniques (see chapter ▶ “3D Dynamic Probabilistic Pose Estimation from Data Collected Using Cameras and Reflective Markers”). The current limitation to these techniques is the small field of view and the fact that only the surface can be imaged due to the fact that ultrasound does not penetrate the bone.

Future Directions Ultrasound imaging has clearly provided a significant advance in the ability to assess the material properties or mechanical function of human muscle, tendon, and ligaments. However, there are numerous limitations to current approaches that either

Ultrasound Technology for Examining the Mechanics of the Muscle, Tendon, and. . .

173

limits the capacity to make accurate biomechanical measurements or to fully understand how muscle, tendon, or ligament mechanical properties influence or adapt to human motion. The future will see continued improvement in image resolution (both space and time), potential to track strain or displacement of tissues and methods to perturb tissues for assessment of mechanical properties. Developments in elastography techniques are likely to have the greatest potential in biomechanics. For instance, increases in the rate at which SSI can be collected and the range over which it can measure the shear modulus will allow tissues to be examined while contracting in more realistic conditions and therefore potentially allow researchers to better assess when and how much individual muscles contribute to various tasks. Examining dynamic and localized strains in tendinous and ligamentous tissues using speckle tracking techniques also has great potential at examining injury mechanics in these tissues. One of the biggest limitations of ultrasound is that it is highly localized and only represents a single plane of the tissue of interest. Advancements in 3D technology will be a major area of advance in ultrasound imaging. For instance, 3D elastography has already been developed (Lindop et al. 2006) and may be used to precisely determine areas of tissue “weakness” or structural difference. 4D ultrasound will continue to be developed with larger acquisition volumes that will enable researchers to examine precise muscle, tendon, or ligament strains across multiple planes with good time resolution. A final possibility is that transducers will also advance so that they may also deform along with the tissues of interest to enable more accurate 3D representations with less influence of pressure on the transducer. Despite the relentless advancement of such technology, it is always necessary to validate the ability of these technologies in actually quantifying the measurement of interest. Such validation is often difficult to do in humans in vivo, and hence this is likely to be time-limiting factor in advancing this area of biomechanics and human motion.

References Alexander RM, Bennet-Clark HC (1977) Storage of elastic strain energy in muscle and other tissues. Nature 265:114–117 Ates F, Hug F, Bouillard K, Jubeau M, Frappart T, Couade M, Bercoff J, Nordez A (2015) Muscle shear elastic modulus is linearly related to muscle torque over the entire range of isometric contraction intensity. J Electromyogr Kinesiol 25:703–708 Bolsterlee B, Gandevia SC, Herbert RD (2016a) Effect of transducer orientation on errors in ultrasound image-based measurements of human medial gastrocnemius muscle fascicle length and Pennation. PLoS One 11:e0157273 Bolsterlee B, Gandevia SC, Herbert RD (2016b) Ultrasound imaging of the human medial gastrocnemius muscle: how to orient the transducer so that muscle fascicles lie in the image plane. J Biomech 49:1002–1008 Bouillard K, Hug F, Guevel A, Nordez A (2012) Shear elastic modulus can be used to estimate an index of individual muscle force during a submaximal isometric fatiguing contraction. J Appl Physiol (1985) 113:1353–1361

174

G. Lichtwark

Braekken IH, Majida M, Engh ME, Bo K (2009) Test-retest reliability of pelvic floor muscle contraction measured by 4D ultrasound. Neurourol Urodyn 28:68–73 Chernak Slane L, Thelen DG (2014) The use of 2D ultrasound elastography for measuring tendon motion and strain. J Biomech 47:750–754 Cooperberg PL, Barberie JJ, Wong T, Fix C (2001) Extended field-of-view ultrasound. Semin Ultrasound CT MR 22:65–77 Cronin NJ, Carty CP, Barrett RS, Lichtwark G (2011) Automatic tracking of medial gastrocnemius fascicle length during human locomotion. J Appl Physiol (1985) 111:1491–1496 Darby J, Hodson-Tole EF, Costen N, Loram ID (2012) Automated regional analysis of B-mode ultrasound images of skeletal muscle movement. J Appl Physiol (1985) 112:313–327 Delp SL, Anderson FC, Arnold AS, Loan P, Habib A, John CT, Guendelman E, Thelen DG (2007) OpenSim: open-source software to create and analyze dynamic simulations of movement. IEEE Trans Biomed Eng 54:1940–1950 Farris DJ, Lichtwark GA (2016) UltraTrack: software for semi-automated tracking of muscle fascicles in sequences of B-mode ultrasound images. Comput Methods Prog Biomed 128:111–118 Finni T, Peltonen J, Stenroth L, Cronin NJ (2013) Viewpoint: on the hysteresis in the human Achilles tendon. J Appl Physiol (1985) 114:515–517 Franz JR, Slane LC, Rasske K, Thelen DG (2015) Non-uniform in vivo deformations of the human Achilles tendon during walking. Gait Posture 41:192–197 Franz JR, Thelen DG (2015) Depth-dependent variations in Achilles tendon deformations with age are associated with reduced plantarflexor performance during walking. J Appl Physiol (1985) 119:242–249 Fukashiro S, Itoh M, Ichinose Y, Kawakami Y, Fukunaga T (1995) Ultrasonography gives directly but noninvasively elastic characteristic of human tendon in vivo. Eur J Appl Physiol Occup Physiol 71:555–557 Fukunaga T, Ichinose Y, Ito M, Kawakami Y, Fukashiro S (1997) Determination of fascicle length and pennation in a contracting human muscle in vivo. J Appl Physiol (1985) 82:354–358 Fukunaga T, Kubo K, Kawakami Y, Fukashiro S, Kanehisa H, Maganaris CN (2001) In vivo behaviour of human muscle tendon during walking. Proc Biol Sci 268:229–233 Gillett JG, Barrett RS, Lichtwark GA (2013) Reliability and accuracy of an automated tracking algorithm to measure controlled passive and active muscle fascicle length changes from ultrasound. Comput Methods Biomech Biomed Engin 16:678–687 Grieve DW, Pheasant S, Cavanagh PR (1978) Prediction of gastrocnemius length from knee and ankle joint posture. Biomechanics vi-a. University Park Press, Baltimore Hansen P, Bojsen-Moller J, Aagaard P, Kjaer M, Magnusson SP (2006) Mechanical properties of the human patellar tendon, in vivo. Clin Biomech (Bristol, Avon) 21:54–58 Hawkins D, Hull ML (1990) A method for determining lower extremity muscle-tendon lengths during flexion/extension movements. J Biomech 23:487–494 Helfenstein-Didier C, Andrade RJ, Brum J, Hug F, Tanter M, Nordez A, Gennisson JL (2016) In vivo quantification of the shear modulus of the human Achilles tendon during passive loading using shear wave dispersion analysis. Phys Med Biol 61:2485–2496 Herbert RD, Gandevia SC (1995) Changes in pennation with joint angle and muscle torque: in vivo measurements in human brachialis muscle. J Physiol 484(Pt 2):523–532 Hug F, Lacourpaille L, Maisetti O, Nordez A (2013) Slack length of gastrocnemius medialis and Achilles tendon occurs at different ankle angles. J Biomech 46:2534–2538 Hug F, Tucker K, Gennisson JL, Tanter M, Nordez A (2015) Elastography for muscle biomechanics: toward the estimation of individual muscle force. Exerc Sport Sci Rev 43:125–133 Ihnatsenka B, Boezaart AP (2010) Ultrasound: basic understanding and learning the language. Int J Shoulder Surg 4:55–62 Ikai M, Fukunaga T (1968) Calculation of muscle strength per unit cross-sectional area of human muscle by means of ultrasonic measurement. Int Z Angew Physiol 26:26–32

Ultrasound Technology for Examining the Mechanics of the Muscle, Tendon, and. . .

175

Jia R, Mellon S, Monk P, Murray D, Noble JA (2016) A computer-aided tracking and motion analysis with ultrasound (CAT & MAUS) system for the description of hip joint kinematics. Int J Comput Assist Radiol Surg 11:1965–1977 Kallinen M, Suominen H (1994) Ultrasonographic measurements of the Achilles tendon in elderly athletes and sedentary men. Acta Radiol 35:560–563 Kane D, Grassi W, Sturrock R, Balint PV (2004) A brief history of musculoskeletal ultrasound: ‘from bats and ships to babies and hips’. Rheumatology (Oxford) 43:931–933 Karamanidis K, Stafilidis S, Demonte G, Morey-Klapsing G, Bruggemann GP, Arampatzis A (2005) Inevitable joint angular rotation affects muscle architecture during isometric contraction. J Electromyogr Kinesiol 15:608–616 Kawakami Y, Abe T, Fukunaga T (1993) Muscle-fiber pennation angles are greater in hypertrophied than in normal muscles. J Appl Physiol (1985a) 74:2740–2744 Kawakami Y, Ichinose Y, Fukunaga T (1998) Architectural and functional features of human triceps surae muscles during contraction. J Appl Physiol (1985b) 85:398–404 Korstanje JW, Selles RW, Stam HJ, Hovius SE, Bosch JG (2010) Development and validation of ultrasound speckle tracking to quantify tendon displacement. J Biomech 43:1373–1379 Lee SS, Lewis GS, Piazza SJ (2008) An algorithm for automated analysis of ultrasound images to measure tendon excursion in vivo. J Appl Biomech 24:75–82 Lichtwark GA, Cresswell AG, Ker RF, Reeves ND, Maganaris CN, Magnusson SP, Svensson RB, Coupe C, Hershenhan A, Eliasson P, Nordez A, Foure A, Cornu C, Arampatzis A, MoreyKlapsing G, Mademli L, Karamanidis K, Vagula MC, Nelatury SR (2013) Commentaries on viewpoint: on the hysteresis in the human Achilles tendon. J Appl Physiol (1985) 114:518–520 Lichtwark GA, Wilson AM (2005) In vivo mechanical properties of the human Achilles tendon during one-legged hopping. J Exp Biol 208:4715–4725 Lichtwark GA, Wilson AM (2006) Interactions between the human gastrocnemius muscle and the Achilles tendon during incline, level and decline locomotion. J Exp Biol 209:4379–4388 Lindop JE, Treece GM, Gee AH, Prager RW (2006) 3D elastography using freehand ultrasound. Ultrasound Med Biol 32:529–545 Loram ID, Maganaris CN, Lakie M (2006) Use of ultrasound to make noninvasive in vivo measurement of continuous changes in human muscle contractile length. J Appl Physiol (1985) 100:1311–1323 Maganaris CN (2005) Validity of procedures involved in ultrasound-based measurement of human plantarflexor tendon elongation on contraction. J Biomech 38:9–13 Maganaris CN, Paul JP (1999) In vivo human tendon mechanical properties. J Physiol 521(Pt 1):307–313 Narici MV, Binzoni T, Hiltbrand E, Fasel J, Terrier F, Cerretelli P (1996) In vivo human gastrocnemius architecture with changing joint angle at rest and during graded isometric contraction. J Physiol 496(Pt 1):287–297 Nightingale K (2011) Acoustic radiation force impulse (ARFI) imaging: a review. Curr Med Imaging Rev 7:328–339 Noorkoiv M, Nosaka K, Blazevich AJ (2010a) Assessment of quadriceps muscle cross-sectional area by ultrasound extended-field-of-view imaging. Eur J Appl Physiol 109:631–639 Noorkoiv M, Stavnsbo A, Aagaard P, Blazevich AJ (2010b) In vivo assessment of muscle fascicle length by extended field-of-view ultrasonography. J Appl Physiol (1985) 109:1974–1979 Nordez A, Gallot T, Catheline S, Guevel A, Cornu C, Hug F (2009) Electromechanical delay revisited using very high frame rate ultrasound. J Appl Physiol (1985) 106:1970–1975 Obst SJ, Newsham-West R, Barrett RS (2014) In vivo measurement of human achilles tendon morphology using freehand 3-D ultrasound. Ultrasound Med Biol 40:62–70 Onambele GN, Burgess K, Pearson SJ (2007) Gender-specific in vivo measurement of the structural and mechanical properties of the human patellar tendon. J Orthop Res 25:1635–1642 Passmore E, Sangeux M (2016) Defining the medial-lateral axis of an anatomical femur coordinate system using freehand 3D ultrasound imaging. Gait Posture 45:211–216

176

G. Lichtwark

Pearson SJ, Burgess K, Onambele GN (2007) Creep and the in vivo assessment of human patellar tendon mechanical properties. Clin Biomech (Bristol, Avon) 22:712–717 Peters A, Baker R, Sangeux M (2010) Validation of 3-D freehand ultrasound for the determination of the hip joint centre. Gait Posture 31:530–532 Pollock CM, Shadwick RE (1994) Allometry of muscle, tendon, and elastic energy storage capacity in mammals. Am J Phys 266:R1022–R1031 Powell PL, Roy RR, Kanim P, Bello MA, Edgerton VR (1984) Predictability of skeletal muscle tension from architectural determinations in guinea pig hindlimbs. J Appl Physiol Respir Environ Exerc Physiol 57:1715–1721 Raiteri BJ, Cresswell AG, Lichtwark GA (2016) Three-dimensional geometrical changes of the human tibialis anterior muscle and its central aponeurosis measured with three-dimensional ultrasound during isometric contractions. PeerJ 4:e2260 Rana M, Hamarneh G, Wakeling JM (2009) Automated tracking of muscle fascicle orientation in B-mode ultrasound images. J Biomech 42:2068–2073 Slane LC, Thelen DG (2014) Non-uniform displacements within the Achilles tendon observed during passive and eccentric loading. J Biomech 47:2831–2835 Slane LC, Martin J, Dewall R, Thelen D, Lee K (2016) Quantitative ultrasound mapping of regional variations in shear wave speeds of the aging Achilles tendon. Eur Radiol 27(2):474–482 Telfer S, Woodburn J, Turner DE (2014) An ultrasound based non-invasive method for the measurement of intrinsic foot kinematics during gait. J Biomech 47:1225–1228 Treece GM, Gee AH, Prager RW, Cash CJ, Berman LH (2003) High-definition freehand 3-D ultrasound. Ultrasound Med Biol 29:529–546 Treece G, Lindop J, Chen L, Housden J, Prager R, Gee A (2011) Real-time quasi-static ultrasound elastography. Interface Focus 1:540–552 Varghese T (2009) Quasi-static ultrasound Elastography. Ultrasound Clin 4:323–338 Wakeling JM, Jackman M, Namburete AI (2013) The effect of external compression on the mechanics of muscle contraction. J Appl Biomech 29:360–364

Part III Generative Methods in Dynamic Pose Estimation

3D Dynamic Probabilistic Pose Estimation from Data Collected Using Cameras and Reflective Markers Thomas M. Kepple and Alan R. De Asha

Abstract

Position and orientation (Pose) estimations of the human body during motion that are derived from data collected using any marker-based camera system have inherent errors related to a combination of measurement noise, soft tissue artifact (STA), and inaccuracies due to incorrect marker placement. Individually, and in combination, these errors reduce the overall accuracy of marker-based Pose estimation. Optimization and multibody dynamics methods have been formulated to reduce these errors. However it has been argued that uncertainty in data, such as that caused by sensor noise, soft tissue deformation, marker movement, or inaccurate marker placement, cannot be directly accounted for using traditional deterministic approaches. We postulate that uncertainty can be more appropriately addressed by casting the Pose estimation problem within the general framework of probabilistic inference. In this chapter, we will introduce Bayes theorem, the basis for probabilistic inference, and give a general example of how a Bayesian approach can take advantage of prior knowledge to improve estimation. We will then formulate Bayes theorem in the context of mitigating uncertain marker motion. Finally, we will apply this approach on some sample data to demonstrate how this method can, in practice, produce substantially better measurement of knee joint motion then the previously established deterministic methods. Keywords

Bayesian inference • Pose estimation • Motion capture • Markers • Multibody models • Probabilistic • X-ray

T.M. Kepple (*) • A.R. De Asha C-Motion Inc., Germantown, MD, USA e-mail: [email protected]; [email protected] # Springer International Publishing AG, part of Springer Nature 2018 B. Müller, S.I. Wolf (eds.), Handbook of Human Motion, https://doi.org/10.1007/978-3-319-14418-4_158

179

180

T.M. Kepple and A.R. De Asha

Contents Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . State of the Art . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Probabilistic Pose Estimation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Bayes Theorem . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Probabilistic Pose Estimation from Marker-Based Data: Theory . . . . . . . . . . . . . . . . . . . . . . . . . . The Importance of the Covariance Matrices . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Probabilistic Pose Estimation from Marker-Based Data: An Example . . . . . . . . . . . . . . . . . . . . . Casting the 6-DOF and IK Methods in a Statistical/Generative Framework . . . . . . . . . . . . . . . . . . Future Directions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

180 181 182 182 184 186 187 192 193 194

Introduction For those of us who have always used discriminative or deterministic models for solving the Pose estimation problem, the generative approach about to be described requires a conceptual leap. The solution to the position and orientation (Pose) of a model, for a given set of data, is, perhaps counterintuitively, not to solve for the Pose directly. Instead it is to solve for the possible data sets that are consistent with the measured Pose. In this chapter, we consider the Pose estimations of multi-segment, rigid body (multibody) models based on recordings of 3D optical marker-based motion data. The principal assumption of these Pose estimation algorithms (▶ “3D Dynamic Pose Estimation from Marker-Based Optical Data”) is that the markers move rigidly along with the body segments to which they are attached. It is known, however, that marker measurements have errors related to noise and marker movement relative to the underlying skeleton (soft tissue artifact, STA) (▶ “Estimation of the Body Segment Inertial Parameters for the Rigid Body Biomechanical Models Used in Motion Analysis”) and can have inaccuracies due to incorrect marker placement (Leardini et al. 2005; Taylor et al. 2005; Peters et al. 2010). All of these errors reduce the accuracy of the Pose estimation. The consequence of STA alone is that the accuracy with which bone motion can be measured by noninvasive clinical motion analysis is typically insufficient for tissue-/joint-level biomechanical analysis (Cappello et al. 2005). It should be noted that these nonrigid marker motions can be represented mathematically (Dumas et al. 2014; Grimpampi et al. 2014), but it has proven difficult to incorporate this information into a deterministic Pose estimation algorithm. It has been argued (Todorov 2007) that uncertainties related to STA cannot be mitigated directly from the currently used discriminative Pose estimation methods and that Pose estimation from motion capture (MoCap) data is best estimated by assuming uncertainty in the data. Todorov suggested using well-established probabilistic algorithms, based on Bayesian inference. These algorithms provide a principled way for making optimal inferences from uncertain data in combination with previous experience. Probabilistic Pose estimation may be understood conceptually by realizing that an improved estimate of the Pose of the model for any given set of

3D Dynamic Probabilistic Pose Estimation from Data Collected Using Cameras. . .

181

data is not to solve for the Pose directly (a deterministic or discriminative model) but rather to solve for the Pose using a generative model that minimizes a combination of the estimated errors of a discriminative Pose and a predicted Pose. Probabilistic Pose estimation provides a principled way to include models of STA in the Pose estimation algorithms and to produce better Pose estimates of the skeleton.

State of the Art Current methodologies for estimating the Pose of rigid bodies from marker data, such as those described in the previous chapter, are still based on deterministic methodologies. The simplest deterministic approaches to 3D movement analysis use direct methods (Kadaba et al. 1989; Davis et al. 1991) that are based on vector algebra to compute the Pose of a model, made up of rigid anatomical segments. Spoor and Veldpaus (1980) cast Pose estimation as an optimization problem in order to try and mitigate STA and other measurement errors. Their optimization approach was based on the assumption that the configuration of targets on a segment remains constant with respect to each other and with respect to the underlying bone. Andriacchi et al. (1998) enhanced the segmental optimization (6 DOF) approach by distributing a large number of markers on a given segment and reducing the contribution of unreliable markers to the solution in the hope of further reducing the effects of STA. Lu and O’Connor (1999) extended the segment optimization approach to a multibody optimization, or inverse kinematic (IK), solution discussed in the previous chapter, by adding joint constraints to the model in order to further reduce errors. Other methods, based on dynamics, including residual elimination (Remey and Thelen 2009), residual reduction (Hamner et al. 2008), and optimal control (Kaplan and Heegaard 2001; van den Bogert et al. 2011; Miller et al. 2016; Koelewijin et al. 2016) have also been used to improve estimations of Pose. In particular, optimal control is an emerging method for treating multibody Pose estimation (Kaplan and Heegaard 2001; van den Bogert et al. 2011; Miller and Hamill 2015) (chapter ▶ “Optimal Control Modeling of Human Movement”; ▶ “Physics-Based Models for Human Gait Analysis”). The objective is to determine the controls, u(t), which produce a movement that is, in some sense, “optimal.” Optimal control casts Pose estimation as a trade-off between an initial estimate of the model’s Pose and multibody dynamics. Multibody model dynamics can be represented symbolically by the function f of the model’s state variables, s, and control variables, u (i.e., the equations of motion): f ðs, s0 , uÞ ¼ 0

(1)

where the controls u specify the joint moments and the state vector s that contains the model’s generalized coordinates (the Pose) and generalized speeds. The tracking error in the equations of motion between the model (s) and motion capture (q) generalized coordinates is given by:

182

T.M. Kepple and A.R. De Asha

dðs, uÞ ¼ q 

ðT 0

f ðs, s0 , uÞ

(2)

d(s, u) represents the difference between the Pose Ð T estimate generated from the motion capture data q and the simulated Pose, 0 f ðs, s0 , uÞ , estimated from the multibody model at time T. These optimal control-based methods generally require either additional data from one or more force platforms (GRF) or an anatomically congruent full-body model. For example, the equations of motion typically contain a discrete-element viscoelastic/Coulomb friction model of foot-ground contact (Miller and Hamill 2015). An additional error term can then be represented as the difference between the model-generated ground reaction force (GRF) EGRF and the measured GRF (GRF(t)): gðs, uÞ ¼ EGRF ðs, τðuÞÞ  GRFðtÞ

(3)

g(s, u) is added as a constraint on the objective function to be minimized in the optimization. The result is that the optimal control solutions do not supply a general solution for tracking rigid bodies using surface markers. Because these dynamics-based methods generally require either additional data from one or more force platforms or an anatomically congruent full-body model, they do not supply a general solution for tracking rigid bodies using surface markers. Probabilistic inference, unlike the deterministic approaches of optimal control or 6 DOF and IK (described in the previous chapter), is able to deal with uncertainty and thereby makes optimal use of available data. The “noise” is captured by a generative model, which defines the conditional probability of the data given the Pose. The most advanced use of probabilistic Pose estimation is a new more generalized approach to Pose estimation based on real-time state estimation (Lowrey et al. 2017). This approach, however, is currently used in robotics not human motion measurement.

Probabilistic Pose Estimation Bayes Theorem Estimating the Pose of a model probabilistically begins with Bayes theorem: Pðqj vÞ ¼

pðvj qÞpðqÞ pð v Þ

(4)

To explain the basis of Bayes theorem, we shall step through a simple example. A female soccer player comes into the office of a sports medicine doctor complaining of knee pain. The doctor conducts an anterior draw test (an imperfect

3D Dynamic Probabilistic Pose Estimation from Data Collected Using Cameras. . .

183

diagnostic test) on the patient and finds more laxity in the painful knee than in the unaffected knee. What is the probability that the patient has an ACL injury? Using a Bayesian approach to the problem, we start by applying the Bayesian likelihood: p(v|q), that is, we know that the probability of the data, q (a positive anterior draw test), given the hypothesis, v (an ACL injury), is: pðvj qÞ ¼ 0:55 Using our prior knowledge of ACL injury, p(q), that female soccer players are more likely to suffer ACL injuries than participants in other sports, we use this past experience to determine the probability that our soccer-playing patient has an ACL injury: pðqÞ ¼ 0:65 Finally, we need to find the probability of the data, the normalization term, p(v). The normalization term is based on the probability that the patient is not ACL injured ( p(~q) = 1.0  0.65 = 0.35). We also need the probability that a patient has an ACL injury but is not a female soccer player. (For our example, we assume this value: p(v| q) = 0.5). To compute the normalization term, we use: pðvÞ ¼ pðvj qÞ  pðqÞ þ pð qÞ  pðvj  qÞ

(5)

which for our example is: pðvÞ ¼ ð0:55Þ  ð0:65Þ þ ð0:35Þ  0:5 ¼ 0:5325 So Bayes theorem becomes: Pðqj vÞ ¼

0:55  0:65 ¼ 0:67 0:5325

Note that in this case we use our prior knowledge, that a female soccer player presenting with knee pain is more likely to have an ACL injury than another athlete, to improve the predictability of the anterior draw test from 0.55 to 0.67. It is important to note that in order for this approach to work properly, this prior must be veridical, not simply based on personal bias; that is, it should come from some valid source of prior knowledge (published epidemiological research would be a valid prior source in our example). In this way, the prior will improve accuracy while minimizing any reduction in the value of available data to the outcome. Armed with this understanding of how Bayes theorem works, we will now apply these same principles to use prior knowledge about how markers move relative to underlying bone (i.e., STA). We shall do this in order to obtain better estimates of a model’s Pose using data from a motion capture system, by mitigating the effects of STA.

184

T.M. Kepple and A.R. De Asha

Probabilistic Pose Estimation from Marker-Based Data: Theory Throughout this chapter, we discuss the application of Bayes theorem to improve Pose estimation for traditional marker-based MoCap data. To improve Pose estimates from MoCap data (v), the terms in Bayes theorem (Eq. 4) become: P(q| v) = the posterior, which is the probability of the Pose, q, given the marker data, v P(v| q) = the likelihood, which is the probability of the marker data, v, given the Pose, q P(q) = the prior which is the initial probability of the Pose, q, given some prior knowledge of the model (i.e., an expected state) p(v) = the probability of the data or normalization term. In our probabilistic inference approach, the normalization term can be ignored because it is considered to be a constant within an optimization problem (Todorov 2007). To frame our Pose estimation problem via Bayes theorem, we start by defining the likelihood from a discriminative IK Pose estimation. From IK (Lu and O’Connor 1999), each marker has a static vector ai that represents the location of the marker in the anatomical reference frame of a segment (AF) and a dynamic vector vi that represents the location of the marker in the global reference frame (GF) (for markers i = 1,2,. . ..N). The transformation for any marker i from AF to GF is given by a set of rotation matrices R and translation vectors O, which are a function of the model’s total generalized coordinates q: vi ¼ RðqÞai þ OðqÞ

(6)

Due to measurement error and STA, the location of markers relative to the underlying bone will change during movement, thus there will always be an error vector (residual) r ðq, vi Þ for each marker: r ðq, vi Þ ¼ vi  RðqÞai þ OðqÞ

(7)

We will assume that all error r ðq, vi Þ comes from marker uncertainty and that the error is normal and independent. For convenience, we define the matrix v which contains all components from all of the marker vectors. For consistency with the probabilistic Pose estimation, the IK residuals are represented as a generative model, with the IK residuals taking the form of the conditional probability distribution: Pðvj qÞ  N d ðr ðq, vÞ, V Þ

(8)

where Nd(r(q, v), V ) is a multivariate normal distribution centered around the mean of residual error r(q, v) with a covariance matrix V generated for all components of each marker. Expressing the likelihood as a Gaussian distribution in canonical form (Stroupe et al. 2001)

3D Dynamic Probabilistic Pose Estimation from Data Collected Using Cameras. . .

Pðvj qÞ ¼

1 1 pffiffiffiffiffiffi eð2Þ 2Π jV j



rðq, vÞ



T





V 1 rðq, vÞ

185

(9)

To find the best Pose estimate, we define our problem as solving q for the maximum likelihood P(v| q). For convenience, we take the negative log of the likelihood: logðPðvj qÞÞ ¼ C þ

  1 ðr ðq, vÞÞT V 1 ðr ðq, vÞÞ 2

(10)

Maximizing P(v| q) is now equivalent to finding q to minimize  log (P(v| q), for the set of marker coordinates, v (Todorov 2007), which we express as: f ðqÞ ¼ r ðq, vÞT V 1 r ðq, vÞ

(11)

In other words, the solution to the IK Pose estimation is to find q that minimizes Eq. 11 over all marker data, v. The fundamental advantage of the probabilistic Pose estimation is that we can introduce prior knowledge of biomechanics in a principled way. We can represent this knowledge of the state of the system (^q) as the expected Pose estimate obtained from joint constraint(s) and/or a simple STA model. The difference between the current state of the system and the expected state is: Prior error ¼ ðq  ^ qÞ

(12)

Consistent with our expectation of the uncertainty in the Pose estimate, we assume that the differences between our current and predicted state estimates, ðq  ^ q Þ are normal and independent with a covariance of U: Pðq  ^q Þ  N d ðr ðq  ^ q Þ, U Þ

(13)

Expressing the prior as a Gaussian distribution in canonical form (Stroupe et al. 2001) Pðq  ^q Þ ¼

1 1 pffiffiffiffiffiffi eð2Þ 2Π jV j





rðq^ qÞ

T



U1 rðq^q Þ

(14)

Referring back to Bayes theorem (Eq. 4), if we declare that both our likelihood and prior are Gaussian distributions and express them in canonical form, P(v| q) takes the form: Pðvj qÞ ¼

   T 1 1 T 1 1 1 1 pffiffiffiffiffiffi eð2Þ rðq, vÞ V rðv, xÞ  pffiffiffiffiffiffiffi eð2Þðq^q Þ U ðq^q Þ 2Π jV j 2Π jU j

(15)

186

T.M. Kepple and A.R. De Asha

where ^ q were the predicted positional state variables, V1 is the inverted covariance matrix obtained from the variation of the IK residuals (Likelihood) over multiple trials of data (i.e., not the identity matrix), and U1 is the inverted covariance matrix of the difference between measured states q and expected states ^q . As before, we solve for the maximum likelihood P(v| q) by minimizing the negative log of:   1 ðr ðq, vÞÞT V 1 ðr ðq, vÞÞ 2   1 þ ðq  ^q ÞT U 1 ðq  ^q Þ 2

logðPðvj qÞÞ ¼ C þ

(16)

which we express as: f ðqÞ ¼ r ðq, vÞT V 1 r ðq, vÞ þ ðq  ^ q ÞT U 1 ðq  ^q Þ

(17)

So the solution to the Bayesian Pose estimation is to find q that minimizes Eq. 17 over all marker data, v.

The Importance of the Covariance Matrices In Eq. 17, the covariance matrices, V1 and U1, serve two vital roles. Firstly, consider that r(q, v) is a measure of the marker residual error (Eq. 7), and thus this difference between the measured and expected marker locations will always be a distance. Conversely, ðq  ^q Þ is a measure of the difference between the measured and expected values of the Pose. Note the model Pose, q, will contain both position and orientation degrees of freedom, and therefore the values in ðq  ^q Þ (Eq. 12) can appear as both distances and angular measures. Thus, since the two terms in Eq. 17 contain matrices whose elements may have different units, these elements need to be normalized to allow the two terms in the cost function to be added. Since V is a measure of the covariance of r(q, v), the elements of variance matrix V will have the same units as r(q, v). Likewise, all the elements of U are obtained from the variance in ðq  ^ q Þ and thus have the same units as the elements of ðq  ^q Þ. From this, we can see that inverting these matrices, V1 and U1, will normalize all the elements in Eq. 17 so that they can be used in a meaningful cost function. The second role for the covariance matrices comes from the fact that, at each time step, V1 will weigh the solution to Eq. 17 inversely to the variation found by the likelihood r(q, v), while U1 will weigh the solution to Eq. 17 inversely to the variation expected by the prior (q  ^q ). Thus if the data for either the likelihood or the prior at any given instant during the movement become more variable, we tend to trust those data less. This means they will contribute less to the overall solution. Conversely, when the variation in the data decreases, the values in the inverted covariance matrices will increase, and these data will get extra weight in the solution.

3D Dynamic Probabilistic Pose Estimation from Data Collected Using Cameras. . .

187

Since our assumption is that both our prior and likelihood errors are normally distributed (Eqs. 7 and 12), weighing the data in this manner is consistent with Bayesian inference. Practically speaking, in order to obtain the covariance matrices V1 and U1, it is best to capture a large number of independent trials from a group of subjects that matches the population of interest. The data from these trials can then be time normalized to a fixed number of frames between the start and the end of the movement. The covariance of the fit of the MoCap data to the expected target locations, as obtained from Eq. 7, gives us the matrix V. Likewise, at each time step during the motion, we have both a measured and an expected (prior) value for model Pose, q, with the difference between the two being given by Eq. 12. We obtain the covariance matrix, U, by applying Eq. 12 at each time step over our large sample size of subjects. Finally by inverting both Vand U, we get V1 and U1 which weight the solution to Eq. 17 (under the assumption that our errors are normally distributed).

Probabilistic Pose Estimation from Marker-Based Data: An Example Now that we have considered and discussed this probabilistic theory, let’s investigate some data to illustrate how it can be used in practice. The test data for this chapter were collected in collaboration with Dr. Scott Tashman at the biodynamics laboratory of the University of Pittsburgh. Three subjects walked on a treadmill while data were captured synchronously using a traditional marker-based motion capture system (a synchronized 12 camera Vicon MX system) with a conventional lower extremity gait marker set and a 3D dynamic stereo X-ray (DSX) system (chapter ▶ “Measurement of 3D Dynamic Joint Motion Using Biplane Videoradiography”). The DSX system had been previously validated, both in static and dynamic joint rotation conditions (Anderst et al. 2009). For the static testing, the DSX translational accuracy was found to be on the order of 0.2 mm and rotational accuracy on the order of 0.2 degrees. During the dynamic testing, the accuracy of the DSX system was found to be better than 1.0 degree for all rotations (flexion/extension, ab/adduction, and axial rotation), and translational accuracy was found to be better than 0.7 mm in all three planes. Thus the DSX system serves as our “gold standard” throughout this example. A calibration cube (11.5 cm per side), containing both tantalum beads and optical motion capture markers, was used to calibrate the DSX and also to define the transformation between the DSX and motion capture space. Shank and thigh positions and orientations were measured by both the DSX system and markerbased MoCap system. We compared the DSX Pose values against three different marker-based Pose estimation algorithms: 6-DOF, IK, and a probabilistic method. The errors associated with the different marker-based methods were computed from the RMS translation errors, determined by the Euclidean distance between the segment origins obtained from the MoCap data and from the “gold standard” DSX system.

188

T.M. Kepple and A.R. De Asha

As noted above, our probabilistic inference approach relies on having some prior knowledge about the measured motion. The “biomechanically inspired” prior used in this probabilistic Pose example was based on the twin assumptions that the knee behaves as a three rotational degree of freedom joint (no translations allowed) and that a model of STA could be developed from the DSX data in order to improve the knee joint angle estimates. This STA model was based on the location of the lateral knee marker relative to the knee joint center. It is important to note that although this method used a purely rotational knee model for the prior, the overall Bayesian cost function does not constrain the final knee motion to be rotation-only; it merely guides the solution to that suggestion. This is because our Bayesian likelihood, which is also part of the cost function, allows five degrees of freedom at the knee (three rotational degrees of freedom plus anterior/posterior and superior/inferior translation). The results of the example study revealed that the 6-DOF approach produced a poor Pose estimation compared to the DSX data (Fig. 1a, Table 1). This was, at least in part, because the three markers used to track the shank segment (on the lateral femoral epicondyle, the lateral aspect of the tibia, and the lateral ankle) were almost collinear (as is common for the conventional gait marker set). This near collinearity caused the 6-DOF solution to be hypersensitive to small measurement errors and STA. We next tested an IK model that had constrained the knee joint to permit no mediolateral translations. Thus the knee was, identically to the Bayesian likelihood, a 5-DOF joint (three rotations, plus anterior/posterior and inferior/superior translations), which permitted knee translations in the sagittal plane (Fig. 1b, Table 1). This model produced a poorer Pose estimation than 6-DOF for the given marker set when compared to the DSX results.

Fig. 1 A comparison of the DSX data (gold thigh and shank) and motion capture data (white thigh and shank) for an indicative sample frame. (a) When segment optimization (6 DOF) Pose estimation was used to track the thigh and shank. (b) When inverse kinematics was used to track the thigh and shank with a five degree of freedom knee constraint was used

3D Dynamic Probabilistic Pose Estimation from Data Collected Using Cameras. . .

189

Table 1 Test results (RMS error) of the tibia and femoral origin locations (motion capture vs. DSX) over all frames measured by the X-ray system Pose method 6 DOF Inverse kinematics (5 DOF) Inverse kinematics (3 DOF) Bayesian with motion and soft tissue prior

Left thigh error (in mm) 15.6 12.9 9.7 10.6

Left shank error (in mm) 32.0 41.9 13.2 4.4

Fig. 2 A comparison of the dynamic stereo X-ray data (gold thigh and shank) and motion capture data (white thigh and shank) for an indicative sample frame tracking the thigh and shank. (a) A three rotational degree of freedom inverse kinematic knee model was used to track the segments from the motion capture data. (b) Bayesian inference, with a prior that included a three degree of freedom knee and a soft tissue artifact model obtained from DSX data, was used to tack the segments from the motion capture data

The third and final test of the deterministic algorithms compared the DSX results to those when using an IK model with a 3-DOF knee that was constrained to permit no translation but allow rotations in all three planes. This produced the best non-probabilistic inference match between the motion capture data and the dynamic stereo X-ray data (Fig. 2a). However, a 3-DOF IK will never, by definition, allow any knee translation to be measured. Thus it is limited as a long-term solution to our Pose estimation problem, especially in a clinical setting, when joint translation may be a critical piece of information. For our probabilistic approach, we implemented a prior based on two assumptions: a 3 (rotational)-DOF knee joint and an STA model. The STA model, which is discussed in the next section, was based on a simple linear relationship between the axial rotation of the hip joint and the soft tissue motion of the lateral knee marker. It is important to reemphasize that the assumption of a 3-DOF knee prevented joint translation in the prior (second term on the right-hand side in Eq. 17); however, perhaps counterintuitively, the final overall solution (minimization of Eq. 17) still allowed translation at the knee joint because our Bayesian likelihood (the first term on the right side of Eq. 17) was based on a 5-DOF knee. The end result was a probabilistic Pose estimation that provided superior shank translational agreement

190

T.M. Kepple and A.R. De Asha A/P Knee Translation: X-Ray (black) and Bayesian (red)

A/P Translation (meters)

0.0000

−0.0050

−0.0100 0.0

50.0 % DSX Interval

100.0

Fig. 3 Comparison of the anterior/posterior knee translation as measured by the DSX system (black) and from the MoCap data after the Bayesian model was applied (red)

than the discriminative models and equivalent agreement for thigh translations (Fig. 2b and Table 1). To further demonstrate the potential of using a STA model without limiting joint translation, we can examine the data for the anterior/posterior knee translation (Fig. 3) as measured by the DSX system (black) compared to the anterior/posterior knee translation obtained from a Bayesian model (red). Note the translational displacement between the two models match up very well over the first 75% of the data interval. To develop our prior used in this model, we collected data on two subjects during walking at each’s self-selected walking speed on a treadmill. We then tested the algorithm on a third subject during the same task. We modeled the soft tissue motion of the lateral knee marker because the IK solution proved to be most sensitive to “accuracy” of this marker. Upon examination of the DSX and MoCap data, we identified a relationship between the soft tissue error of the lateral knee marker and the internal/external rotation of the hip (Fig. 4). When looking at the data from our initial two subjects, we found that the linear relationship between internal/external rotation angle versus the anterior/posterior soft tissue motion had an average r2 = 0.78 with an average slope of 1.0 mm/degree of rotation. Unfortunately, due to the limited amount of DSX data available, we did not find a strong enough relationship to predict vertical soft tissue motion and vertical knee translation. From Fig. 3, it is clear that although our prior included the expectation that the knee is a three rotational-DOF joint, the overall solution to the Bayesian approach allowed translation due to the inclusion of knee translation in the likelihood. We can

3D Dynamic Probabilistic Pose Estimation from Data Collected Using Cameras. . .

191

LHip Internal/External Rotation

degrees

7.4

2.1

−3.2 0.0

50.0

100.0

Ap / Soft Tissue (meters)

LKNE Targer A/P soft-tissue motion 0.00

−0.00

−0.01 0.0

50.0

100.0

% DSX Interval

Fig. 4 Plot of the hip internal/external rotation angle (top) versus the anterior/posterior soft tissue motion for the lateral knee marker (bottom) from one of the subjects

also observe from Fig. 3 that, for the first 75% of the X-ray-viewing interval, the Bayesian approach did a reasonably good job of predicting the magnitude and direction of knee translation. During the last 25% of the interval, the Bayesian approach did not match the DSX-measured knee translation well. This was because there was a weak soft tissue relationship described by our simple regression equation during the latter part of the viewing interval (Fig. 4). Ideally, during this latter part of the DSX interval, the prior should have been weighted less. Examining the DSX data for the two subjects used to develop the soft tissue model, we found that the soft tissue relationship was variable during the latter part of the DSX interval. Thus it could be postulated that, with enough DSX data to generate a motion prior that included more robust covariance matrices, our results would be improved during this period. These data represent only a proof of concept; a much larger sample size for generating the covariance matrices would improve the results further. Figures 3 and 4 illustrate the potential power of the probabilistic algorithm. From this sample work, we were able to demonstrate that a probabilistic Pose estimation approach greatly reduced translation estimation errors compared to traditional deterministic approaches (Table 1) while still allowing a good precision for our knee translation estimates throughout the first three quarters of the DSX interval (Fig. 3). We realize that developing better soft tissue models and generating improved

192

T.M. Kepple and A.R. De Asha

covariance matrices to properly weight the Bayesian terms at the appropriate times during a movement could further improve probabilistic Pose estimates.

Casting the 6-DOF and IK Methods in a Statistical/Generative Framework It should be noted that both the 6-DOF and IK deterministic methods for estimating model Pose may also be cast within a statistical framework. Although they both implicitly define a discriminative model and attempt to maximize the probability of the data under their respective model, we can restate these methods within a statistical/generative framework. This will hopefully allow the reader to better able to appreciate the relationship between the deterministic Pose estimations of the previous chapter and the probabilistic Pose estimation described in this chapter. To understand how both the 6-DOF and IK methods are a subset of our Bayesian optimization, let’s take another look at Eq. 17 above: f ðqÞ ¼ r ðq, vÞT V 1 r ðq, vÞ þ ðq  ^ q ÞT U 1 ðq  ^q Þ

(18)

This is the cost function for our probabilistic Pose estimation optimization problem. In this equation, r(q, v) is the residual of the fit between the measured and expected target locations, and ðq  ^q Þ is the difference between the current generalized coordinate, q, and the expected generalized coordinate ^q, based on prior knowledge. V1 is the inverted covariance matrix for the X, Y, Z components of the residual, and U1 is the inverted covariance matrix for the difference between the measured and expected generalized coordinates. Now, in the 6-DOF method, there are no prior expectations of the values of the generalized coordinates. In the inverse kinematics method, the only prior knowledge we apply to the joint motion is that the motion becomes fully constrained. The net effect is that, for both the 6-DOF and IK methods, U1 becomes a zero matrix and the cost function reduces to: f ðqÞ ¼ r ðq, vÞT V 1 r ðq, vÞ

(19)

Another underlying assumption of both the basic 6-DOF and IK methods is that the residuals at any given frame and for any given target are weighted equally, which is theoretically equivalent to setting the covariance matrix V1 to the identity matrix. This would further reduce the cost function to: f ðqÞ ¼ r ðq, vÞT r ðq, vÞ

(20)

Examination of this cost function reveals that it represents nothing more than a minimization of the sum of squares error between the expected and measured target

3D Dynamic Probabilistic Pose Estimation from Data Collected Using Cameras. . .

193

locations. This is the equivalent of the cost function for an unweighted 6 DOF where the model Pose, q, is contained in the rotation matrix, T, and translation vector, O: r ðq, vÞ ¼

XN i¼1

ðvi  Rai  OÞ2

(21)

and an unweighted inverse kinematics problem: r ðq, vÞ ¼

XN i¼1

ðvi  RðqÞai  OðqÞÞ2

(22)

This derivation demonstrates that the deterministic solution is simply a subset of the generative solutions to the estimation of Pose.

Future Directions In this chapter, we have highlighted our initial success at using a model of soft tissue artifact as a prior. Although the simple STA model used in our test example worked reasonably well for our simple test example, establishing a general prior for STA that will be applicable to different marker sets, patient populations, and movement patterns is unlikely. In other words, establishing priors for STA and other expectations of the Pose will be quite challenging. Thus, in our view, it is imperative that probabilistic Pose estimation evolve in a manner that will facilitate the development of custom Bayesian priors. In addition to dealing with STA, custom priors will allow probabilistic Pose estimation to aid new motion capture technologies, such as markerless motion capture (▶ “3D Dynamic Pose Estimation from Markerless Optical Data”), to move into clinical and research settings. Note that once a new prior is created, a new set of covariance matrices is required for use with that prior. When insufficient data are available to generate the covariance matrices from a large sample, the approach can still be used by either using a smaller sample size and only the diagonals of the covariance matrices or simply by weighting the matrices so that the likelihood and prior terms are equal. It is expected that diagonal covariance matrices or ad hoc weighting will likely diminish the accuracy of the probabilistic method but will still provide improved Pose estimations better than discriminative methods. One statistical technique that might possibly allow for adapting priors, in order to facilitate their being generalized across populations (e.g., a patient group), is waveform principal component analysis (PCA) (Deluzio and Astephen 2007). PCA can be used to describe a normalized waveform (e.g., one of the generalized coordinates describing the Pose). Given a mean signal of a generalized coordinate for a motion (Pose being described by this collection of signals), and a small number of principal components, a signal may be constructed that could approximate signals from all subjects in the control data set (subject-specific signals differ by a set of coefficients

194

T.M. Kepple and A.R. De Asha

or PCA scores). Consider now a rich data set that includes multi-subject metadata such as height, weight, age, sex, and disease or condition, as well as the condition severity. In addition to representing subject-specific signals, PCA can be used to discriminate or differentiate groups; therefore the shape of the prior could, potentially, be predicted using these metadata and the principal components alone. The power in this method is that it is predictive and could be calculated prior to collection. Accordingly, it could be used as a Bayesian prior for estimating Pose. This approach would also have the advantage that it would always have a prior, even in the absence of data, and thus it would provide a solution (e.g., the prior), rather than no solution, in situations when the model is not observable. In a deterministic solution, as described in the previous chapter, an unobservable system state results in the entire frame of data being unreliable and therefore “empty.” In the future, it should be possible to expand Bayesian priors to include multibody dynamics via optimal control. Treating multibody Pose estimation as an optimal control problem was described earlier in this chapter and is well established. The objective is usually to determine controls that produce a movement that is, in some sense, “optimal.” Optimal control, via direct collocation (DC) methods (van den Bogert et al. 2011; Miller and Hamill 2015) (▶ “Optimal Control Modeling of Human Movement”), casts Pose estimation as a trade-off between an initial estimate of the Pose from the MoCap data, multibody dynamics, and from matching the recorded ground reaction force. In the future, we can envisage use of the output of the DC solution as a Bayesian prior that would, in essence, drive the Pose estimation from the marker based MoCap data toward dynamic consistency in a controlled way. This would represent a “full circle,” as the use of multibody dynamics-based priors was the starting point for this probabilistic approach (Todorov 2007), and has already been extended by Todorov’s lab for robotics (Lowrey et al. 2017) and may be the most likely path of future progression for human motion.

References Anderst W, Zauel R, Bishop J, Demps E, Tashman S (2009) Validation of three-dimensional model based tibio-femoral tracking during running. Med Eng Phys 31(1):10–16 Andriacchi TP, Alexander EJ, Toney MK, Dyrby C, Sum J (1998) A point cluster method for in vivo motion analysis: applied to a study of knee kinematics. J Biomech Eng 120:743–749 Cappello A, Stagni R, Fantozzi S, Leardini A (2005) Soft tissue artifact compensation in knee kinematics by double anatomical landmark calibration: performance of a novel method during selected motor tasks. IEEE Trans Biomed Eng 52:992–998 Davis R, Ounpuu S, Tyburski D, Gage J (1991) A gait analysis data collection and reduction technique. Hum Mov Sci 10:575–587 Deluzio KJ, Astephen JL (2007) Biomechanical features of gait waveform data associated with knee osteoarthritis. An application of principal component analysis. Gait Posture 25:86–93. PMID: 16567093 Dumas R, Camomilla V, Bonci T, Cheze L, Cappozzo A (2014) Generalized mathematical representation of the soft tissue artefact. J Biomech 47:476–481

3D Dynamic Probabilistic Pose Estimation from Data Collected Using Cameras. . .

195

Grimpampi E, Camomilla V, Cereatti A, de Leva P, Cappozzo A (2014) Metrics for describing softtissue artefact and its effect on Pose, size, and shape of marker clusters. IEEE Trans Biomed Eng 61(2):362–367 Hamner S, John C, Anderson FC, Higginson J, Delp S (2008) Reducing residual forces and moments in a three-dimensional simulation of running. Proceedings of the North American Congress on Biomechanics IV, August 2008 Kadaba M, Ramakrishnan H, Wootten M, Gainey J, Gorton G, Cochran G (1989) Repeatability of kinematic, kinetics and electromyographic data in normal adult gait. J Orthop Res 7:849–860 Kaplan ML, Heegaard JH (2001) Predictive algorithms for neuromuscular control of human locomotion. J Biomech 34:1077–1083. PMID: 11448699 Koelewijin A, Richter H, van den Bogert A (2016) Trajectory optimization in stochastic multibody systems using direct collocation. In Proceedings of the 4th joint international conference on multibody system dynamics Leardini A, Chiari L, DellaCroce U, Cappozzo A (2005) Human movement analysis using stereophotogrammetry. Part3. Soft tissue artifact assessment and compensation. Gait Posture 21:212–225 Lowrey K, Dao J, Todorov E (2017) Real-time state estimation with whole-body multi-contact dynamics: a modified UKF approach. In: Humanoid Robots (Humanoids), 2016, IEEE-RAS, 16th International Conference Lu TW, O’Connor JJ (1999) Bone position estimation from skin marker co-ordinates using global optimization with joint constraints. J Biomech 32:129–134 Miller R, Hamill J (2015) Optimal footfall patterns for cost minimization in running. J Biomech 48:2858–2864 Miller R, Kepple T, Selbie WS (2016) Direct collocation as a filter for inverse dynamics. In Proceedings of the 40th annual meeting of the American Society of Biomechanics Peters A, Galna B, Sangeux M, Morris M, Baker R (2010) Quantification of soft tissue artifact in lower limb human motion analysis: a systematic review. Gait Posture 31:1–8 Remey C, Thelen D (2009) Optimal estimation of dynamically consistent kinematics and kinetics for dynamic simulation of gait. ASME J Biomech Eng 13(3):31005 Spoor C, Veldpaus F (1980) Rigid body motion calculated from spatial coordinates of markers. J Biomech 13(4):391–393 Stroupe AW, Martin MC, Tucker B (2001) Distributed sensor fusion for object position estimation by multi-robot systems. In IEEE international conference on robotics & automation (ICRA-01) Taylor WR, Ehrig RM, Duda GN, Schell H, Seebeck P, Heller MO (2005) On the influence of soft tissue coverage in the determination of bone kinematics using skin markers. J Orthop Res 23:726–734 Todorov E (2007) Probabilistic inference of multijoint movements, skeletal parameters and marker attachment from diverse motion capture data. IEEE Trans Biomed Eng 54:1927–1939 van den Bogert A, Blana D, Heinrich D (2011) Implicit methods for efficient musculoskeletal simulation and optimal control. Procedia IUTAM 2(2011):297–316

3D Dynamic Pose Estimation from Markerless Optical Data Steven Cadavid and W. Scott Selbie

Abstract

This chapter provides an overview of three-dimensional (3D) dynamic Pose (position and orientation) estimation of human movement without the use of markers or sensors, more commonly known as Markerless Motion Capture (Markerless Mocap). As with Marker-based Motion Capture (Marker-based Mocap), the methods presented estimate the Pose of an underlying multibody subject-specific model comprising rigid segments with anatomically defined local reference frames and joint constraints. In addition, the model has an overlying surface representing the skin, or clothing, depending on the context. The focus of this chapter is on Markerless Mocap algorithms best suited to biomechanical analyses of human movement. In other words, those techniques appropriate for estimating 3D Pose directly, and accurately, from recorded data. Of all the approaches to Markerless Mocap, 3D-to-3D Pose estimation is most similar to Marker-based Mocap techniques because it requires arrays of multiple, time synchronous, video cameras encircling the capture volume. In addition to the underlying multibody skeletal model that marker-based and markerless techniques have in common, during Markerless Mocap, the subject is identified by a surface model overlying the skeleton. In each frame of motion data, a pixelated surface, comprised of a dense collection of points lying on the surface, is extracted from the scene and registered to the model. Neither marker-based nor 3D-to-3D Markerless Mocap is typically accurate enough to record the Pose of the bones at a resolution for studying joint dynamics. S. Cadavid (*) KinaTrax Inc., Palm Beach, FL, USA e-mail: [email protected] W.S. Selbie HAS-Motion Inc, Kingston, ON, Canada C-Motion Inc., Germantown, MD, USA e-mail: [email protected] # Springer International Publishing AG, part of Springer Nature 2018 B. Müller, S.I. Wolf (eds.), Handbook of Human Motion, https://doi.org/10.1007/978-3-319-14418-4_160

197

198

S. Cadavid and W.S. Selbie

An alternative markerless approach to joint level biomechanics has emerged. Biplanar videogradiography (or Dynamic Stereo X-ray) uses a 3D-to-2D approach to Markerless Mocap, whereby only two views of the subject are acquired because of space limitations and to minimize radiation exposure. A brief introduction to 3D-to-2D registration will be presented because this is covered in more detail in another chapter. Keywords

Markerless Mocap • Marker-based Mocap • Multibody 3D Pose estimation • Articulated registration • Space carving • Stereo reconstruction • Biplanar videoradiography • 3D-to-3D registration • 3D-to-2D registration • Visual hull

Contents Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . State of the Art . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Creating a Visual Hull Representation of the Surface of the Subject . . . . . . . . . . . . . . . . . . . . . . Modeling the Surface from Dense Voxelization . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Identifying Subject-Specific Reference Frames . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Pose Estimation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Tracking the Pose . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3D-2D Pose Estimation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Future Directions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Cross-References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

198 199 201 205 206 208 210 212 214 216 216

Introduction There are numerous challenges when recording human movement using markers or sensors attached to a subject’s skin or clothing. Data collection protocols for threedimensional Marker-based Mocap are time-consuming and expensive and require a constrained laboratory/studio environment. Marker-based protocols require the individual collecting data to have a high degree of technical expertise, as segmental Pose (position and orientation) estimation can be extremely sensitive to marker placement precision (Della Croce et al. 2005). In common with all Marker/sensor-based Mocap systems, the encumbrance of the markers or sensors and the instructions to participants, to behave “naturally” on demand, can cause many subjects to act “unnaturally.” In this chapter, the exemplar motion capture being illustrated is the recording of a Major League Baseball pitcher throwing during live game action. In this game scenario, the Mocap protocol must not affect the game or the pitcher, and all players must be unaware of the recording. Lastly, and in some respects most importantly, markers and sensors move relative to the underlying skeleton to which they are attached (Cappozzo et al. 1996). This soft tissue artifact is idiosyncratic and challenging to mitigate mathematically. Markerless Mocap requires no markers or sensors and thus reduces the time and expertise required to collect data. Markerless Mocap allows subjects to move

3D Dynamic Pose Estimation from Markerless Optical Data

199

naturally through a motion capture volume that may be indoors or outdoors. Because there are no errors associated with marker/sensor placement, Pose estimation is dependent, predominantly, on the mathematical algorithms used for processing the data. These can always be refined as mathematical techniques improve, so no data are lost to operator expertise. Markerless Mocap therefore has the potential to increase the availability and reduce the cost, of performing accurate 3D movement analysis. The ability to collect reliable motion capture data efficiently and independently of the operator facilitates the collection of large amounts of data across a broad spectrum of environments. Markerless Tracking is the process for estimating the Pose of segments that comprise a model of a subject, in each frame of a recorded sequence. This chapter focuses on optimization-based approaches to Pose estimation including methods that register a 3D articulated multibody model to the surface of the subject (3D-to-3D registration; Corazza et al. 2010; Cheung et al. 2003), and methods that generate multiple virtual two dimensional (2D) images (digitally reconstructed video image) and register these 2D images to recorded video images (3D-to-2D registration; Stoll et al. 2011; Balan et al. 2007, Bey et al. 2006, 2008; Brainerd et al. 2010; Giphart et al. 2012; Haque et al. 2013; Tashman et al. 2017). Omitted from this chapter are methods of sensor fusion, including Markerless Mocap and IMU (Marcard et al. 2016), and methods predominantly focused on animation and vision, such as 2D-to3D Markerless Mocap from one video camera (Bogo et al. 2015), or estimations from sparse markers and statistical shape models (Loper et al. 2014).

State of the Art As in other chapters of this Handbook, the assumption underlying markerless Pose estimation is that the body (Fig. 1a) is comprised of a skeletal model (Fig. 1b) constructed from a set of rigid (nondeformable) segments (or bodies) and, unique to a markerless approach, a surface model (Fig. 1d) representing the skin and/or clothing. Each skeletal segment is defined by a local anatomical reference frame (right handed Cartesian coordinate system). The origin of the reference frame is placed at the proximal end of a segment, coincident with the distal end of an adjacent segment. This creates a joint connecting the child (usually the distal segment) to the parent (usually the proximal segment, Fig. 1b). Each segment is restricted to having only one parent segment, and the segment’s interaction with its parent segment is described by the specification of joint constraints acting at, and around, the origin of a segment, relative to the parent segment. These joint constraints define the number of degrees of freedom allowed at the joint. For Markerless Mocap, the number of degrees of freedom can be any integer value between zero and six. In practice, however, only the root segment is prescribed 6 DOF relative to the laboratory, and all other segments have between 1 and 3 DOF relative to the parent, i.e., segments are allowed to rotate relative to each other, but translation tends to be constrained. For marker-based Mocap, the local reference frames defining the segments of the multibody model are typically generated from the location of markers placed on

200

S. Cadavid and W.S. Selbie

Fig. 1 (a) The subject is defined by an underlying skeleton and surface. (b) For the baseball example in this chapter the skeleton is represented by a set of geometric primitives linked as a multibody model. (c) An example of segment anatomical reference frames typical of Marker-based Mocap. (d) The surface of the body is defined by a polygonal mesh rigged to the skeleton. An exemplar frame showing the skeleton and surface model (Visual Hull) of a pitcher

several anatomically defined locations (palpable bony landmarks) in a reference trial, or by the relative motion of sets of homologous landmarks markers in a functional reference trial (Fig. 1c). With respect to the local segmental reference frames, the Pose of a template is defined by a set of tracking markers attached to the segments. In each frame of motion data, this template is registered to the 3D location of homologous markers from the recorded scene. From this registration, the Pose of the multibody model is computed. Similarly, a 3D-to-3D markerless Mocap solution identifies the locations of a set of surface points with respect to local segment reference frames. The articulated skeleton is an underlying structure that defines the Pose of the subject, but because it is hidden within the body, it does not provide data needed for the markerless Pose estimation. An independent approximation to the subject’s

3D Dynamic Pose Estimation from Markerless Optical Data

201

skin/clothing is established as a polygonal surface model that is rigged to the underlying skeleton (Fig. 1c). In order to represent a specific subject, this skeletal/surface model must be scaled and transformed to match the size and shape of the subject. Instead of discrete identified markers, however, markerless Mocap uses a dense set of unidentified 3D points representing an entire pixelated surface of the subject (Template Point Cloud). All 3D-to-3D methods are based on identifying a cloud of 3D points that define the surface of the subject (a Visual Hull) and map this surface to an underlying representation of the skeleton. In each frame of motion data, the Template Point Cloud is segmented from the scene and registered to the Visual Hull. Instead of the one-to-one mapping used in a marker-based approach, however, the two point clouds are registered as a collection (3D-to-3D registration). From this registration, the Pose of the multibody model is computed. Of the many techniques described in the literature to recover the 3D surface shape of an object from multiple calibrated viewpoints; In this chapter, Space Carving (Kutulakos and Seitz 2000) will be presented in detail, and Stereo Reconstruction (Seitz et al. 2006) and depth sensors (Weiss et al. 2011) in summary.

Creating a Visual Hull Representation of the Surface of the Subject Of the many approaches to markerless Mocap, 3D-to-3D Pose estimation is most similar to marker-based Mocap techniques because it requires arrays of multiple time synchronous video cameras encircling the capture volume. As with marker-based Mocap, the number and configuration of cameras is dependent on the type of motion to be captured and the size of the capture volume. The camera lenses and calibrated volume are adjusted such that every camera’s field of view contains the movement being analyzed with as wide a variety of unobstructed views of the movement as possible. In the Major League Baseball parks, arrays of 7 to 16 cameras are mounted to the exterior structure of the stadiums (Fig. 2). Space carving is a method for identifying the Visual Hull of an object from projections of a silhouette of the subject (in the example of this chapter, the baseball pitcher) onto multiple cameras. This 3D-to-3D Pose estimation is often referred to as a “silhouette based” approach. The prerequisite, therefore, is to extract a silhouette of the subject from the background. In order to identify pixels in a video that correspond solely to the subject (in this example, the pitcher), a statistical model of the background (all nonpitcher pixels) must be created and removed from each video image. Of the many ways to model the background (Piccardi 2004), the simplest approach is to use a single image of the capture volume prior to the subject entering the scene. The limitation of this approach is that the background may vary during the pitch resulting in background pixels that substantively change in appearance being detected as foreground pixels (i.e., as part of the subject). Background models that can cope with changes in illumination conditions and appearance are typically statistical methods, such as the Gaussian Mixture Model (Zivkovic 2004), which compute a multimodal distribution at each pixel location across multiple exemplar background images. For a series of background images, each pixel is assigned a likelihood (of being a background pixel) based on its intensity and the respective

202

S. Cadavid and W.S. Selbie

Fig. 2 A schematic drawing of a baseball stadium with 7 cameras aimed at the pitcher’s mound. The figure displays the image of a pitcher during the delivery of the ball from each of the 7 cameras. At the bottom of the image is a controlling computer synchronizing the cameras and receiving the video data

distribution of intensities of every other pixel in the image. Incorporating multiple background images, with varying appearances (e.g. at different times during the game), to build pixel intensity distributions enables the background model to be more robust to changes in background appearance during the data collection. The silhouette of the subject is isolated by subtracting the background and thresholding the resultant difference image to yield a binary image for each video image frame. Holes and rough edges in the silhouette, where a background pixel was inadvertently identified, are corrected through a series of binary morphological operations that fill in holes and smooth the resulting silhouette (Fig. 3). For Space Carving, each voxel (volume pixel) is projcted onto the image plane of a camera; all voxels that do not intersect with the silhouette are assigned a value of zero (e.g., carved out of the voxel array). This process is repeated for each camera view. The resulting nonzero voxels represent a carved representation (Convex Hull) of the subject for that frame of data. Space Carving is robust to false positives in the foreground detection (i.e., background pixels incorrectly labeled as foreground pixels) because the resultant Visual Hull is the intersection of the back-projections and foreground. Note that the resulting shape cannot recover the smoothness and concave surfaces of the subject because many of these features lie within the silhouette. If, however, these “enclosed” surfaces have voxels with a unique color from the rest of the silhouette, the silhouette can be assigned multiple colors and a color consistency check may be used to carve out these regions independently (Kutulakos and Seitz 2000).

3D Dynamic Pose Estimation from Markerless Optical Data

203

Fig. 3 A silhouette of a baseball pitcher at a frame in the middle of the throwing motion extracted from a video image

Fig. 4 Two views of a pitcher at different time points in the pitch carved from multiple cameras. Note the problem created by shadows, and the concave regions (for example, the ankle) that have been filled by the space carving technique

A challenge facing the Space Carving technique, when collecting data in outdoor settings (as in our baseball example), is inconsistent lighting that casts shadows of the subject (Fig. 4). The simplest and most effective approach to remove shadows is to place a camera close to the ground with its optical axis parallel to the floor. This low-placed camera could “carve away” the cast shadow because it is not visible from this viewpoint, and hence discarded during the intersection of the silhouette backprojections (KaewTraKulPong and Bowden 2002). In our baseball example, however, cameras cannot be placed on the field, so the cast shadow must be identified by the Pose estimation as noise and ignored. The voxelization obtained by Space Carving is referred to as a dense Visual Hull voxelization because all surface and internal voxels are included in the Visual Hull.

204

S. Cadavid and W.S. Selbie

This dense voxelization then undergoes shelling and water-tight meshing, using the Marching Cubes algorithm, to obtain a Visual Hull. The surface of the hull, a thinshelled voxelization, is produced by applying binary morphological cleaning to the dense voxelization. This thin-shelled Visual Hull voxelization is an approximation to the subject’s skin (Fig. 4). An alternative approach to generating the Visual Hull is Stereo Reconstruction. In this approach, a set of common fiducial (reference) landmarks are identified across adjacent cameras. Methods such as the Harris Corner Detector (Harris and Stephens 1988) and the FAST Corner Detector (Rosten and Drummond 2006) can be used to identify these fiducial landmarks. This is done within a small window of the entire image where a strong gradient, within that window, is present. For each fiducial window, a feature vector is generated to represent the texture within this local window surrounding a fiducial point. The feature vector is generated using rotation and scale invariant feature extraction methods such as SIFT (Lowe 1999), SURF (Bay et al. 2006), and ORB (Rublee et al. 2011). Candidate keypoint correspondences are then established by pairing key points in two images that are most similar in the feature space. Outlier correspondences are filtered by aligning the key points in one image to their correspondences in the second image. This is achieved by applying an affine transformation that minimizes the root mean square distance error between the point sets. Corresponding key points that exceed a predefined distance threshold are then discarded. The optimal affine transformation used to align the correspondences can be computed using Procrustes Analysis (Gower 1975). This process yields a sparse set of keypoint correspondences between the two images, which alone is insufficient to generate a dense point cloud reconstruction of a subject. To address this, the images undergo rectification to transform them to a common image plane. This results in all epipolar lines being parallel to the horizontal axis and corresponding points having identical vertical coordinates. The problem of establishing dense point correspondences is then greatly simplified, because the correspondence search is reduced to a 1D search in the same row of the counterpart image. Simple and efficient pixel intensity comparisons can then be applied to identify a correspondence for each pixel in the two images. A disparity map (the apparent pixel difference or motion between a pair of stereo images) is then generated from the pixel location distances between correspondences. A disparity map can be transformed into a dense 3D point cloud by triangulation. The Stereo Reconstruction method can be extended to more than two views to form a complete 3D model by reconstructing a point cloud from every adjacent pair of cameras. Feature correspondences can also be established across more than two views to obtain a more accurate measure of depth using Bundle Adjustment (Triggs et al. 1999). Stereo Reconstruction can produce accurate 3D reconstructions recovering convex and concave surfaces; however, the primary disadvantage is that it relies on the veracity of point correspondences to produce an accurate result. Objects with little texture, such as skin and some articles of clothing, yield poor correspondences since the resultant feature vectors are insufficiently discriminative. Therefore, most researchers opt for Space Carving because it does not require the establishment of point correspondences.

3D Dynamic Pose Estimation from Markerless Optical Data

205

Lastly, and presented only briefly, active depth sensing technologies such as Structured Light Sensors and Time of Flight (ToF) sensors (e.g., Microsoft Kinect) can be used to generate a Visual Hull. Depth sensors project a constant pattern of infrared (IR) light (dots) into the capture volume. As the light strikes, and is reflected from a surface, the pattern is distorted. This distortion is recorded by depth cameras located at an offset from the IR transmitter. From the difference between the recorded and expected dot positions, the depth of each pixel of an RGB (red, green, blue) color camera is computed. Perhaps the best known implementation of this approach is the Microsoft Kinect. The Kinect algorithm (Shotton et al. 2013; Keskin et al. 2013) extracts an efficient feature vector from each pixel in a depth image and classifies the pixel as belonging to a specific body part. To create a complete Visual Hull from depth sensors, multiple sensors are typically used in which the depth map identifies pixels on the surface of the subject. These pixels are converted into 3D points on the surface of the object (i.e., the Visual Hull). To cover the entire surface of the object, the 3D points captured by multiple depth cameras from different viewpoints can be integrated. Points on background objects (i.e., objects except for the subject) are excluded typically by the difference in depth from the subject. This approach tends to be useful in a small volume because the depth sensors have only a short range. In the case of the baseball example in this chapter, in which there is no possibility of getting close to the pitcher at any time, this approach is not appropriate.

Modeling the Surface from Dense Voxelization The next step in this markerless approach is to model this dense point cloud of data as a mathematical surface. In this chapter, two types of subject-specific body surface models are presented: generic surface templates and reconstructed Visual Hull models. Generic surface template models consist of a generic skin surface that, by default, represents a parameterized body shape. Typically the model also contains an underlying generic skeleton that is rigged to the skin surface. A simple parameterized segment surface can be represented by a small series of spheres along the axis between joints, as used successfully to track non human primates (Nakamura et al. 2016) This model must be transformed to register with the size and shape of the subject. This model is perhaps best represented by a Statistical Shape Model (Anguelov et al. 2005; Pons-Moll et al. 2015) that is derived from a collection of similar subjects with the surface synthesized from a weighted set of shape modes of variation. The shape modes of variation are computed by applying principal component analysis to a dataset of varying body morphologies. The skin surface consists of a mesh, where each skin vertex on the mesh, p, is assigned a set of weights, wi. These weights control the amount of deformation that a vertex can undergo as a result of n neighboring segments being rotated during registration:

206

S. Cadavid and W.S. Selbie

pi ¼

n X i¼1

w i Ri p

n X

wi ¼ 1

(1)

i¼1

where Ri corresponds to the rotation matrix applied to the ith neighboring segment and pi signifies the transformed skin vertex position. In a process that is commonly referred to as skinning, the generic model is made subject-specific by applying a nonrigid registration between the generic model and a 3D representation of the subject (i.e., the Visual Hull described in the previous section) that has been captured from the acquisition sensors. This registration process is challenging because, at this stage, the subject’s segment lengths, body morphology, and the texture information for the subject are all unknown. When registered well, the generic template models tend to produce a clean smooth skin surface and underlying skeleton because the model is usually handmade by a graphic artist or animator. However, the registration optimization procedure is often ill-constrained and the resultant skin surface may not fully conform to the actual morphology of the subject. Reconstructed Visual Hull models can be obtained by a 3D scan of the body (Corazza et al. 2010) or can be constructed directly from the Visual Hull extracted from the subject performing the movement of interested (e.g., representing many of the Poses inherent to the motion). Space Carving is typically used to reconstruct the texture-mapped point cloud of the subject’s skin. A skin mesh is then constructed from the point cloud using the Marching Cubes algorithm (Lorensen and Cline 1987). Because Visual Hull models are directly constructed from the actual morphology of the subject, the skin surface may contain artifacts due to the limitations of the Space Carving procedure.

Identifying Subject-Specific Reference Frames Given the surface approximation of the subject, the next step is to identify the Pose of a skeleton within this surface. To construct the underlying skeleton, joint centers (i.e., the origin of each segment anatomical reference frame) must be identified within the Visual Hull. The lengths of the segments in a multibody model can often be measured or identified from selected postures. In the baseball example, the data are collected during a live Major League baseball game, and it is essential that the pitchers are not interfered with (even before the game) and that the game itself is not interfered with in any way. Therefore, recording a predefined A-Pose or T-Pose is not practicable; instead, the joint centers are digitized manually from video images of a recorded pitch. The carved Visual Hull deforms during the pitch, so several key frames are digitized. At each keyframe, the joints of the subject are annotated (digitized) manually from two or more camera views (Fig. 5). A subject-specific multibody model with joint-specific constraints suitable for pitching is derived from a collection of the 3D location of joint centers (Fig. 5). Identifying subjects’ joint centers during a baseball pitch requires multiple

3D Dynamic Pose Estimation from Markerless Optical Data

207

Fig. 5 Joint centers as digitized manually, and displayed as colored circles

keyframes to account for deformation of the Visual Hull at different times during the pitch; thus, joint center locations vary between frames. A mean joint center position is estimated using a Sparse Bundle Adjustment algorithm (Lourakis and Argyros 2009). Each segment of a template skeleton, at each digitized video frame, is scaled to the distance between its proximal and distal joint centers. This collection of joint centers is then registered to each other using Procrustes Analysis. Using an iterative scheme, an optimal Pose of the constrained model, which minimizes the distance between the descendants of a given joint (distal segments) over all video images, is estimated. The lengths of the segments are computed as the mean lengths, calculated across the keyframes, and are used in the skeletal model for all subsequent processing. The resulting underlying skeleton, therefore, consists of rigid segments connected by joints. Given constant segment lengths and joint constraints, it is possible that for some Poses of the model during a movement, the computed average joint centers may not be anatomically congruent, relative to the surface. For example, a large asymmetry in the side-to-side anthropometrics or anatomically inconsistent joint placement (e.g., right and left shoulder joints are “flipped”). If an error is detected, the 2D annotations are adjusted manually and the average joint centers are recomputed. This process is repeated iteratively until an anatomically credible 3D joint center is obtained over multiple pitches. Given the multibody skeleton and Visual Hull, the individual voxels comprising the Hull must be categorized with respect to the skeleton. Depth Map Classification-based methods (Shotton et al. 2013) can be used to assign each pixel in the image to a specific segment of the skeleton. The classification features are the set of distances in depth between groups of two pixels at predetermined offsets from the pixel being classified. Pixels corresponding to each body part are clustered and fit to a skeletal model on a frame-by-frame basis. The classifier is usually trained on a large set of human movement Mocap data, in which pixels associated with each segment are identified manually. Given this training set, machine learning algorithms are typically used to estimate the Pose of the multibody model from the depth map and the association between a pixel and a segment. In this way, a skeletal model can

208

S. Cadavid and W.S. Selbie

be extracted from the depth map. In other words, each voxel must be assigned to one or two of the underlying nearest skeleton segments. Skin weights are computed for a given skin vertex by determining the closest segment to the vertex using point-to-line segment distance. The skin vertex is assigned to the nearest segment, and a secondary weight is computed for the nearest segment that is connected to the assigned segment. The weights are computed by conducting a k nearest neighbor search at the location of the skin vertex. The weight associated with a segment is equal to the fraction of nearest neighbors that are assigned to the segment. The value of k determines the extent of smoothing during the skin deformation. These weights dictate the deformation that a skin vertex undergoes as the model moves. Appropriate skin weights ensure smooth deformation of the skin around the joint centers during articulation. Visual Hull models constructed across multiple, varying Poses of a subject can improve the quality of both the skin surface reconstruction and the accuracy of the underlying skeletal Pose estimate. Given the segment lengths and locations of the joint centers, the alignment of the anatomical reference frames (AFs) must be computed. Unlike marker-based systems in which anatomical landmarks can be identified, the Visual Hull is not sufficiently unique to be considered in a reference alignment. The surface model and Visual Hull are globally aligned by applying a similarity transformation to the surface model, such that the centroids and first two principal components of both models coincide. The first two principal components, corresponding to the two largest eigenvalues, are computed by applying Singular Value Decomposition to the surface hulls. The models are scaled such that the distance between the centroids and their respective farthest point cloud neighbors is the same. A nonlinear registration process is used to optimally align each body region of the model to the surface hull. To obtain an optimal alignment, each body region of the model undergoes a nonrigid transformation comprising a rotation and nonuniform scaling about the axes of its anatomical reference frame (AF) while maintaining symmetry in the side-to-side anthropometrics. The resulting alignment defines the AFs for the model.

Pose Estimation At each frame of a movement trial, the Visual Hull of the subject is segmented from the background using the same methods described earlier in this chapter. Pose estimation is the process for registering the multibody skeleton and surface model to the Visual Hull. 3D-to-3D registration methods commonly use a variant of the Iterative Closest Point algorithm (Besl and McKay 1992), with scaling capability, to register the point clouds of the 3D articulated model and 3D frame reconstruction on a segment-by-segment basis. It should be noted that additional, nonrigid transformations can be incorporated into the objective function beyond affine scaling, to better conform to the subject body morphology. This, however, is at the cost of having additional variables in the optimization problem.

3D Dynamic Pose Estimation from Markerless Optical Data

209

At all frames of the movement, the Pose of the subject specific skeletal model is estimated. Using the same background subtraction and space carving algorithms as in the model identification, the voxels (points) comprising the subject are segmented, and from this dense voxelization, the surface voxels (point cloud) are identified (Note: for the movement trial the Visual Hull is not computed as only the point cloud is required for the Pose estimation). At each video frame, the Pose is estimated by registering the Visual Hull (rigged to the skeleton) to the surface point cloud. This is done using the articulated iterative closest point method. Notably, this does not require the correspondence between model vertices and the point cloud generated at a given frame, as global optimization methods are used to identify the optimal fit of the Pose in each frame. Seed Pose estimation involves identifying the subject in the scene and registering a multibody model to the subject in all video images from one frame. If possible, the frame of data selected for the Seed Pose is typically based on a frame of data in which the registration is considered straightforward on the basis of the subject being in a “known” or “predictable” Pose. Approaches to seed Pose detection differ depending on whether cooperation from the subject can be expected. In cooperative scenarios, a multibody model of the subject, including segment scaling, can be created by measuring the subject or by computing the location of joint centers functionally. For example, the subject could be instructed to assume a known position, such as the common T-Pose (subject standing upright with feet shoulderwidth apart and arms straight out to the sides, horizontal to the floor, palms facing forwards). The surface model is then configured in the known Pose and that is used as the initial seed Pose estimate. Virtual markers can also be manually placed at anatomical locations similar to Marker-based Mocap to help guide the nonrigid registration. For instance, a virtual marker at the elbow may be useful if the subject is in a T-Pose where the upper arm and forearm may be difficult to disambiguate due to lack of flexion at the elbow. The subject could be requested to wear specific clothing that is colored consistent with colors assigned to the Surface model. In this registration procedure, only the rotational degrees of freedom are retained as variables in the optimization provided fixed segment lengths are specified by the skeletal model. In uncooperative scenarios, where the subject is unaware of the motion capture, or where the subject cannot cooperate (e.g., during a competition), a known Pose may be detected within the trial. There are a variety of ways to detect a known Pose; however, one straight forward way in 3D-to-3D tracking methods is to construct the Visual Hull of the known Pose from a set of exemplar frames of the subject in that Pose. The iterative closest point algorithm can be applied between the Visual Hull Pose template and the 3D reconstruction obtained at each frame of the motion trial. Frame instances that yield residual alignment errors below a specified threshold are accepted as containing the known Pose. If a known Pose cannot be detected, the method of last resort is to manually digitize the joint centers in all of the video images of one frame (see Fig. 5).

210

S. Cadavid and W.S. Selbie

Tracking the Pose Once a Pose estimate has been computed for the initial frame, the tracking procedure is applied for each subsequent frame. The Pose estimate computed for a frame instance, t, can be used as the initial seed to compute the estimates at frame instances t  1 (backwards tracking) and t + 1 ( forward tracking). Backwards and forward tracking can be performed independently and concurrently to potentially reduce processing time. The concept of template matching to detect known Poses in the motion trial can be extended to also detect multiple key Poses in the motion trial. For instance, in baseball pitching, there are key Poses such as the setup, cocking phase, ball release, and follow through that must be present in each pitching sequence because they are important in the biomechanical analysis. These events can be detected and used as sample Poses to build a Pose interpolant. For instance, a spline interpolant can be built for each Euler angle component of each joint to estimate the trajectory of the joints. An interpolated Pose may result in joint angles that violate the joint constraints. Therefore, each joint in the interpolated Pose must be adjusted to ensure it conforms with the defined constraints. One example procedure to building the Pose interpolants is as follows. Consider the case where a key Pose is detected at frames 0 and n, n > 2. A Pose interpolant, γ 0, is built using the estimates at frames 0 and n to compute initial seed Pose estimates at frames 1 and n  1. At the following iteration, Pose interpolant, γ 1, is built using the Pose estimates at frames 0, 1, n  1, and n to compute initial seed Pose estimates at frames 2 and n  2. This process is repeated until all frames between 0 and n have a computed Pose estimate. In cases where a given frame to be tracked is not located between two key Posedetected frames, extrapolation can be used to refine the initial seed of the frame. To alleviate the computational bottleneck of Space Carving in 3D-to-3D tracking methods, the initial seed Pose estimate can be used to reduce the space carved volume to a subvolume within the capture volume enclosing the initial seed Pose estimate. The reduction of the space carved volume on a frame-by-frame basis will also reduce the effect of any nonsubject artifacts present in the Space Carving reconstruction. The articulated registration procedure employed in tracking can utilize local optimization techniques, provided a good initial seed Pose estimate is available. The optimization consists of only rotational degrees of freedom being allowed at each joint in addition to translational and rotational degrees of freedom at the root joint, in cases where fixed segment lengths across the Pose estimates of a motion trial are imposed. Constraints that are applied to the tracking optimization include joint constraints, displacement constraints that restrict the amount of rotation possible between adjacent frames, shape constraints, texture constraints, kinetic constraints, and surface collision constraints. In 3D-to-3D tracking methods, shape constraints can be imposed by computing surface normals at each skin vertex on the articulated model, as well as on the points of the 3D frame reconstruction. The registration procedure can account for the angular difference between the surface normals of registered points to ensure the registered shapes are similar. Articulated registration techniques typically apply segment-wise registration that follows the hierarchical

3D Dynamic Pose Estimation from Markerless Optical Data

211

order of the kinematic chain. For instance, the root segment (typically the pelvis) is firstly registered to the frame reconstruction by translating and rotating the articulated model to align it to the frame reconstruction. Shape constraints are also imposed by discarding all points in the frame reconstruction that are within proximity of the articulated model’s root segment points for the remainder of the articulated registration process. This ensures that points assigned to nonroot segments in the articulated model are not incorrectly registered to points in the root segment region of the frame reconstruction. Secondly, the child segments of the root segment are registered to the frame reconstruction using only rotation. Once again, the frame reconstruction points in proximity to the registered child segment are discarded for the remainder of the articulated registration process. If the articulated model and frame reconstruction are identical, all points belonging to the frame reconstruction should be discarded by the end of the articulated registration process. In 3D-to-3D tracking methods, texture constraints can be applied by representing each surface point in 4D (grayscale texture) or 6D (color texture) instead of just in the three dimensions corresponding to the x, y, and z coordinates. Note that the texture values must be normalized since the spatial coordinates are in different units. The time series of Pose estimates (joint rotations) is referred to as a Pose map. At times, the Pose may drift from the true solution because of a local minimum in the computed solution. One possible cause of this is that the surface of the body, at any given Pose, does not match the Visual Hull because the surface is deformable. The residual error from the articulated registration process can be used to generate a confidence score on the tracking. The initial Pose detection phase can be retriggered whenever the confidence score falls below a predefined threshold. To address this, the surface hull at one of the keyframes is used to register that frame instead of the average hull. These updated “local” hulls can be used in place of the generic hull at or near the relevant keyframes, thereby facilitating continuously enhanced Pose estimations (e.g., “resetting” the tracking process), which then continues. The nonrigid registration process can be automated by solving an optimization problem to minimize the residual error between the transformed generic model and the 3D subject representation. The articulated model can also be parameterized to facilitate the application of joint constraints in the segmental local coordinate system, or AF by Euler angles. Although prone to Gimbal lock, Tait-Bryan angles, consisting of independent rotations about the three principal axes, are also commonly used to parameterize the joint rotations because their representation is intuitive, and also because joint constraints can be directly applied. For example, joint constraints can be applied to the elbow to allow only two rotational degrees of freedom for flexion/extension and forearm pronation/supination by restricting one Tait-Bryan angle to zero rotation. Model parameterization can also include scaling parameters that facilitate adjustment of segmental geometry and thus segmental inertial properties (Fig. 6). The ability to record and analyze 3D markerless Mocap data in a live Major League Baseball game setting is an extraordinary advance in performance analytics for baseball (Fig. 6). In baseball, the important features that have been explored are the kinematics of the shoulder and elbow joints. At this point in time the level of

212

S. Cadavid and W.S. Selbie

Fig. 6 The computed Pose map for three different times in a pitch. The images display the underlying skeleton and the Visual Hull

accuracy is comparable to marker-based Mocap, partly because Marker-based Mocap does a relative poor job of tracking the arm (and shoulder joint center) compared to other segments. In other words, both methods of MoCap, Markerless and Marker-based, can struggle at times to deal with such a highly complex and ballistic movement. The benefit of a markerless approach, in this situation, is the ability to record actual in-game data, which would be impossible using markers. These markerless techniques, however, are proving to be sufficiently accurate for whole body motion and temporal analyses. This means that markerless MoCap could well be applicable in a variety of situations. Given this, the ability to record and analyze 3D markerless Mocap data for clinical assessment would be an exciting evolution in clinical movement analysis.

3D-2D Pose Estimation When only a few cameras are used to record the motion, the 3D-to-3D techniques described earlier cannot generate a sufficiently accurate surface Visual Hull from which the Stereo Reconstruction or Space Carving methods can generate accurate 3D Pose estimation. Biplanar Videoradiography (also known as Dynamic Stereo X-ray; DSX) was selected as the exemplar 3D-to-2D application for this chapter because there is evidence that abnormal mechanical joint function contributes significantly to the development and progression of many types of joint disease. Notably, joint translations of only a few millimeters are critical to estimating tissue stress or joint impingement during loaded functional movements. DSX is the only currently available technology that can achieve submillimeter bone Pose estimation accuracy during a wide variety of functional movements. The details of the DSX application are described elsewhere in this handbook, but because of its importance, it is highlighted in this chapter as well. DSX is a form of 3D-to-2D markerless MoCap that tracks and reconstructs the underlying skeletal structures of a body. The following section will give a brief

3D Dynamic Pose Estimation from Markerless Optical Data

213

Fig. 7 3D representation of a biplanar X-ray configuration when the two X-rays are synchronized. The distal femur, reconstructed from the CT data, is shown in the middle. The inline X-ray image (in line with the X-axis of the lab reference frame) is shown in the red frame; the red line is the perpendicular from the center of the X-ray image plane to the X-ray source. The offset X-ray image (offset from the X-axis of the lab reference frame) is shown in the green frame; the green line is the perpendicular from the center of the X-ray image plane to the X-ray source. For illustration, the inline X-ray image is shown after processing (smoothing and edge detection) and the offset image is shown unprocessed

overview of the techniques involved in DSX. Given a 3D representation of a bone extracted from a high-resolution CT scan of the subject, a local reference frame assigned to the bone, and a time series of X-ray images containing the bone, a Pose map is the solution of the DSX across all frames. The DSX algorithm (Bey et al. 2006, 2008; Ohnishi et al. 2010) solves for the 3D Pose by registering two noncoplanar X-ray images of a bone to two digitally reconstructed radiographs (DRRs) (Fig. 7). Given the position and orientation of an X-ray source, an X-ray image plane, and volumetric CT bone, a DRR is the projection of the CT bone onto a virtual X-ray image, using a simplified X-ray generation model (Siddon 1985; Zhao and Reader 2003). In other words, rays from the X-ray source are cast through the bone to generate a simulated X-ray with the same size and resolution as the actual X-ray. The objective function for the DSX optimization is based on a measure of the overlap between an X-ray image and a DRR. The similarity (or overlap) of the DRR of a given 3D bone Pose and an X-ray image at time stamp j is quantified as a normalized correlation (rj) for all pixels (x, y): P 

  I Xrayj ðx, yÞ  I Xrayj I XDRRj ðx, yÞ  I XDRRj r j ¼ qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi 2 qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi  2ffi I Xrayj ðx, yÞ  I Xrayj I XDRRj ðx, yÞ  I XDRRj xy

214

S. Cadavid and W.S. Selbie

such that IXDRR(x, y) represents the intensity of the DRR at pixel x,y; IXDRR is the mean of the DRR and IXRay(x, y) represents the intensity of the actual X-ray at pixel x,y, and IXRay ðx, yÞ is the mean of the X-ray. Registration is considered maximum similarity or minimum normalized correlation. High speed X-ray images are particularly challenging for this approach because the bone and tissue do not display uniform (or consistent) density across the images. This inconsistency means that the objective function for the registration (e.g., optimization) may identify Poses that do not appear visually to be the best solution. To compensate for the existence of local minima, this optimization problem is solved using global algorithms such as simulated annealing (Higginson et al. 2005; Ingber 2012), a Monte Carlo method in which the solution space is explored probabilistically by randomly searching near the best-known solution. It is modeled after annealing in metallurgy, in which the thermodynamic free energy of a metal decreases as its temperature decreases. In simulated annealing, as the virtual temperature cools, the algorithm searches in a smaller and smaller region around the best-known solution. Simulated annealing is an ideal optimization technique as it is not prone to finding local minima and thus a user can be confident that given a reasonable initial seed, the Pose it returns will be the optimal Pose Map across all frames. DSX has been demonstrated to have the capability to capture dynamics movements with high sampling rates (150-200 Hz) and submillimeter spatial accuracy. DSX has been used to characterize a variety of joint disorders, including changes in joint contact kinematics with knee injuries (ACL,PCL, meniscus) (Tashman et al. 2004; Gill et al. 2009; Van de Velde et al. 2009; Hoshino et al. 2013; Goyal et al. 2012; Marsh et al. 2014), dynamic aspects of patello-femoral disorders (Fernandez et al. 2008; Bey et al. 2008), femoro-acetabular impingement of the hip (Martin et al. 2011; Kapron et al. 2014), shoulder function after rotator cuff injury (Bey et al. 2011) and arthroplasty (Massimini et al. 2010), changes in intervertebral kinematics with lumbar disc degeneration (Anderst et al. 2008; Li et al. 2011), and deformation of the joint capsule and intervertebral discs with cervical spine disc fusion (Anderst et al. 2013, 2014).

Future Directions Two future directions of markerless Mocap for biomechanical applications and one future direction for DSX are highlighted; first, advances in defining subject-specific multibody segment/surface models that do not require manual identification of anatomical references frames and rigged surfaces; second, advances in physicsbased simulations that influence the markerless 3D Pose estimation and that permit the use of only one video camera; and third, 4D and hierarchical tracking of DSX data.

3D Dynamic Pose Estimation from Markerless Optical Data

215

First, impressive results achieved through Deep Learning in image classification tasks in uncontrolled environments have spawned interest in applying such techniques to human Pose estimation. Traditional classification algorithms employ hand-designed feature extraction schemes, which are naturally limited because they require substantial engineering efforts to develop, and are limited to what humans perceive as being “good” features. The advantage of Deep Learning is that Deep Learning methods employ an unsupervised approach to learning features directly from the training data. Convolutional Neural Networks (CNN), for instance, employ kernels that are convolved with an image to extract a feature representation. The kernel weights are trained at multiple resolutions to capture features at different scales. Toshev and Szegedy (2014) proposed a method consisting of a cascaded set of CNNs to estimate Pose from a single 2D view in uncontrolled environments. A vast dataset of normalized images and corresponding Poses, spanning a variety of activities, is used to train each predictor. The initial predictor estimates an initial Pose for a subject from a single viewpoint, which aims to obtain the same output as the initial Pose detection method described in the previous section. Additional Pose predictors are then trained for each body segment in order to refine the localization of the segments. The authors report state-of-the-art or better performance on four academic benchmarks including the Frames Labeled In Cinema dataset, the Leeds Sports Dataset, the Buffy dataset, and the Image Parse dataset. Future directions in Markerless Mocap include the application of Deep Learning to point clouds to estimate a 3D Pose (Wu et al. 2015), incorporating multiple calibrated views of a Pose to train a Deep Learning classifier, and validating these methods against the HumanEva II dataset (Sigal et al. 2010) and Faust dataset (Bogo et al. 2014), which consists of marker based ground-truth data corresponding to motion trial video data. Second, for the markerless Mocap solution, at each video frame, the estimated Pose is a kinematic solution based on registering the vertices of the Visual Hull to the multibody model. The solution, however, is not guaranteed to be dynamically consistent (i.e., to satisfy the laws of mechanics). Physics-based models can be used to drive the Pose estimation from the markerless Mocap system toward dynamic consistency in a controlled way through a Bayesian approach where the markerless Pose is fused statistically with a physical simulation. Motivated by the abundance of conventional video footage, recent approaches based on Deep Learning are aiming to elaborate the physics-based simulations to estimate 3D Pose from a single optical camera (Toshev and Szegedy 2014; Wandt et al. 2016; Wang et al. 2014). Third, for the DSX application, recent research on 4D and hierarchical tracking is promising (Tashman et al. 2017). In a 4D solution, the Pose is not solved at each frame, but rather solved across all frames of data simultaneously by treating the Pose map as a spline across time rather than a collection of discrete Poses. One of the advantages of this approach is that the X-ray emitters can be run slightly out of phase, which dramatically reduces the X-ray scatter in each image.

216

S. Cadavid and W.S. Selbie

Cross-References ▶ 3D Dynamic Pose Estimation from Marker-Based Optical Data ▶ 3D Dynamic Probabilistic Pose Estimation from Data Collected Using Cameras and Reflective Markers ▶ Estimation of the Body Segment Inertial Parameters for the Rigid Body Biomechanical Models Used in Motion Analysis ▶ Measurement of 3D Dynamic Joint Motion Using Biplane Videoradiography ▶ Physics-Based Models for Human Gait Analysis ▶ Three-Dimensional Reconstruction of the Human Skeleton in Motion

References Anderst WJ, Vaidya R, Tashman S (2008) A technique to measure three-dimensional in vivo rotation of fused and adjacent lumbar vertebrae. Spine J 8:991–997 Anderst WJ, Donaldson WF, Lee JY, Kang JD (2014) In vivo cervical facet joint capsule deformation during flexion-extension. Spine J 39(8):514–520 Anderst W, Donaldson W, Lee J, Kang J (2013) Cervical disc deformation during flexionextension in asymptomatic controls and single-level arthrodesis patients. J Orthop Res 31 (12):1881–1889 Anguelov D, Srinivasan P, Koller D, Thrun S, Rodgers J, Davis J, (2005) SCAPE: shape completion and animation of people. In: ACM transactions on graphics (TOG), vol 24, no 3. ACM, pp 408–416 Balan AO, Sigal L, Black MJ, Davis JE, Haussecker HW (2007) Detailed human shape and pose from images. In: 2007 IEEE conference on computer vision and pattern recognition. IEEE, pp 1–8 Bay H, Tuytelaars T, Van Gool L (2006) Surf: speeded up robust features. In: European conference on computer vision. Springer, Berlin/Heidelberg, pp 404–417 Besl PJ, McKay ND (1992) Method for registration of 3-D shapes. In: Robotics-DL tentative. International Society for Optics and Photonics, pp 586–606 Bey MJ, Zauel R, Brock SK, Tashman S (2006) Validation of a new model-based tracking technique for measuring three-dimensional, in vivo glenohumeral joint kinematics. J Biomech Eng 128:604–609 Bey MJ, Kline SK, Tashman S, Zauel R (2008) Accuracy of biplane X-ray imaging combined with model-based tracking for measuring in-vivo patellofemoral joint motion. J Orthop Surg Res 3:38 Bey MJ, Peltz CD, Ciarelli K, Kline SK, Divine GW, van Holsbeeck M, Muh S, Kolowich P, Lock T, Moutzouros V (2011) In vivo shoulder function after surgical repair of a torn rotator cuff: Glenohumeral joint mechanics, shoulder strength, clinical outcomes, and their interaction. Am J Sports Med 10:2117–2129 Bogo F, Romero J, Loper M, Black MJ (2014) FAUST: Dataset and evaluation for 3D mesh registration In: Proceedings IEEE Conference on computer vision and pattern recognition (CVPR), pp 3794–3801 Bogo F, Black MJ, Loper M, Romero J (2015) Detailed full-body reconstructions of moving people from monocular RGB-D sequences. In: International conference on computer vision (ICCV), pp 2300–2308 Brainerd EL, Baier DB, Gatesy SM, Hedrick TL, Metzger KA, Crisco JJ (2010) X-ray reconstruction of moving morphology (XROMM): precision, accuracy and applications in comparative biomechanics research. J Exp Zool 313A:262–279

3D Dynamic Pose Estimation from Markerless Optical Data

217

Cappozzo A, Catani F, Leardini A, Benedetti MG, Della Croce U (1996) Position and orientation in space of bones during movement: experimental artefacts. Clin Biomech 11(2):90–100 Cheung KMG, Baker S, Kanade T (2003) Shape-from-silhouette of articulated objects and its use for human body kinematics estimation and motion capture. In: Computer vision and pattern recognition, 2003. Proceedings. 2003 I.E. computer society conference on, vol 1, IEEE, pp 1–77 Corazza S, Mündermann L, Gambaretto E, Ferrigno G, Andriacchi TP (2010) Markerless motion capture through visual hull, articulated icp and subject specific model generation. Int J Comput Vis 87(1–2):156–169 Della Croce U, Leardini A, Chiari L, Cappozzo A (2005) Human movement analysis using stereophotogrammetry. Part 4: assessment of anatomical landmark misplacement and its effects on joint kinematics. Gait Posture 21(2):226–237 Fernandez JW, Akbarshahi M, Kim HJ, Pandy MG (2008) Integrating modelling, motion capture and x-ray fluoroscopy to investigate patellofemoral function during dynamic activity. Comput Methods Biomech Biomed Engin 11(1):41–53 Gill TJ, Van de Velde SK, Wing DW, Oh LS, Hosseini A, Li G (2009) Tibiofemoral and Patellofemoral kinematics following reconstruction of an isolated posterior cruciate ligament injury: in vivo analysis during lunge. Am J Sports Med 37(12):2388–2385 Giphart JE, Zirker C, Myers C, Pennington WW, LaPrade RF (2012) Accuracy of a contour-based biplane fluoroscopy technique for tracking knee joint kinematics of different speeds. J Biomech 45:2935–2938 Gower JC (1975) Generalized procrustes analysis. Psychometrika 40(1):33–51 Goyal K, Tashman S, Wang JH, Li K, Zhang X, Harner C (2012) In vivo analysis of the isolated posterior cruciate ligament-deficient knee during functional activities. Am J Sports Med 40(4):777–785 Haque MA, Anderst W, Tashman S, Mari GE (2013) Hierarchical model-based tracking of cervical vertebrae from dynamic biplane radiographs. Med Eng Phys 35(7):994–1004 Harris C, Stephens M (1988) A combined corner and edge detector. In: Alvey vision conference, vol 15, p 50 Higginson JS, Neptune RR, Anderson FC (2005) Simulated parallel annealing within a neighborhood for optimization of biomechanical systems. J Biomech 38:1938–1942 Hoshino Y, Fu FH, Irrgang JJ, Tashman S (2013) Can joint contact dynamics be restored by anterior cruciate ligament reconstruction? Clin Orthop Relat Res 471(9):2924–2931 Ingber L (2012) Adaptive simulated annealing. In: Oliveira H, Petraglia A, Ingber L, Machado M, Petraglia M (eds) Stochastic global optimization and its applications with fuzzy adaptive simulated annealing. Springer, New York, pp 33–61 KaewTraKulPong P, Bowden R (2002) An improved adaptive background mixture model for realtime tracking with shadow detection. In: Video-based surveillance systems. Springer, New York, pp 135–144 Kapron AL, Aoki SK, Peters CL, Maas SA, Bey MJ, Zauel R, Andersen A (2014) Accuracy and feasibility of dual fluoroscopy and model-based tracking to quantify in vivo hip kinematics during clinical exams. J Appl Biomech 30(3):461–470 Keskin C, Kıraç F, Kara YE, Akarun L (2013) Real time hand pose estimation using depth sensors. In: Consumer depth cameras for computer vision. Springer, London, pp 119–137 Kutulakos KN, Seitz SM (2000) A theory of shape by space carving. Int J Comput Vis 38(3):199–218 Li W, Wang S, Xia Q, Passias P, Kozanek M, Wood K (2011) Lumbar facet joint motion in patients with degenerative disc disease at affected and adjacent levels: an in vivo biomechanical study. Spine 36(10):629–637 Loper M, Mahmood N, Black MJ (2014) MoSh: motion and shape capture from sparse markers. ACM Trans Graph 33(6):220:1–220:13 Lorensen WE, Cline HE (1987) Marching cubes: a high resolution 3D surface construction algorithm. In: ACM siggraph computer graphics, vol 21, no 4. ACM, pp 163–169

218

S. Cadavid and W.S. Selbie

Lourakis MI, Argyros AA (2009) SBA: a software package for generic sparse bundle adjustment. ACM Trans Mathemat Software 36(1):2 Lowe DG (1999) Object recognition from local scale-invariant features. In: Computer vision, 1999. The proceedings of the seventh IEEE international conference on, vol 2, IEEE, pp 1150–1157 Marsh C, Martin DE, Harner C, Tashman S (2014) Effect of posterior horn medial meniscus root tear on in vivo knee kinematics. Orthop J Sports Med 2(7):1–7 Martin DE, Greco NJ, Klatt BA, Wright VJ, Anderst WJ, Tashman S (2011) Model-based tracking of the hip: implications for novel analyses of hip pathology. J Arthroplast 26(1):88–97 Massimini DF, Li G, Warner JP (2010) Glenohumeral contact kinematics in patients after total shoulder arthroplasty. J Bone Joint Surg Am 92(4):916–926 Nakamura T, Matsumoto J, Nishimaru H, Bretas RV, Takamura Y, Hori E, Ono T, Nishijo H (2016) A markerless 3D computerized motion capture system incorporating a skeleton model for monkeys. PLoS One 11(11):e0166154 Ohnishi T, Suzuki M, Nawata A, Naomoto S, Iwasaki T, Haneishi H (2010) Three-dimensional motion study of femur, tibia, and patella at the knee joint from bi-plane fluoroscopy and CT images. Radiol Phys Technol 3:151–158 Piccardi M (2004) Background subtraction techniques: a review. In: Systems, man and cybernetics, 2004 I.E. international conference on, vol 4, IEEE, pp 3099–3104 Marcard T, Pons-Moll G, Rosenhahn B (2016) Human pose estimation efrom video and IMUs. Trans Patt Anal Mach Intellig 38:1533–1547 Pons-Moll G, Romero J, Mahmood N, Black M (2015) Dyna: a model of dynamic human shape in motion. ACM Trans Graph 34(4):120:1–120:14 Rosten E, Drummond T (2006, May) Machine learning for high-speed corner detection. In: European conference on computer vision. Springer, Berlin/Heidelberg, pp 430–443 Rublee E, Rabaud V, Konolige K, Bradski G (2011) ORB: An efficient alternative to SIFT or SURF. In: 2011 international conference on computer vision, IEEE, pp 2564–2571 Seitz, S.M., Curless, B., Diebel, J., Scharstein, D. and Szeliski, R., 2006, June. A comparison and evaluation of multi-view stereo reconstruction algorithms. In: 2006 I.E. computer society conference on computer vision and pattern recognition (CVPR’06), vol 1, IEEE, pp 519–528 Shotton J, Sharp T, Kipman A, Fitzgibbon A, Finocchio M, Blake A, Cook M, Moore R (2013) Real-time human pose recognition in parts from single depth images. Commun ACM 56(1):116–124 Siddon RL (1985) Fast calculation of the exact radiological path for a three-dimensional CT array. Med Phy 12:252–255 Sigal L, Balan AO, Black MJ (2010) Humaneva: synchronized video and motion capture dataset and baseline algorithm for evaluation of articulated human motion. Int J Comput Vis 87(1–2):4–27 Stoll C, Hasler N, Gall J, Seidel HP, Theobalt C (2011) Fast articulated motion tracking using a sums of gaussians body model. In: 2011 international conference on computer vision, IEEE, pp 951–958 Tashman S, Collon D, Anderson K, Kolowich P, Anderst W (2004) Abnormal rotational knee motion during running after anterior cruciate ligament reconstruction. Am J Sports Med 32(4):975–983 Tashman S, Princehorn J, Penatto S, Andherst W (2017) Intelligent algorithms for tracking threedimensional skeletal movement from radiographic image sequences. US patent # 9538940 B2 Toshev A, Szegedy C (2014) Deeppose: human pose estimation via deep neural networks. In: Proceedings of the IEEE conference on computer vision and pattern recognition, pp 1653–1660 Triggs B, McLauchlan PF, Hartley RI, Fitzgibbon AW (1999) Bundle adjustment – a modern synthesis. In: International workshop on vision algorithms. Springer, Berlin/Heidelberg, pp 298–372 Van de Velde SK, Gill TJ, Li G (2009) Evaluation of kinematics of anterior cruciate ligamentdeficient knees with use of advanced imaging techniques, three-dimensional modeling techniques, and robotics. J Bone Joint Surg Am 91(Suppl 1):108–114

3D Dynamic Pose Estimation from Markerless Optical Data

219

Wandt B, Ackermann H, Rosenhahn B (2016) 3d reconstruction of human motion from monocular image sequences. Trans Pattern Analy Mach Intellig 38(8):1505–1516 Wang C, Wang Y, Lin Z, Yuille A, Gao W (2014). Robust estimation of 3d human poses from a single image. In: Conference on computer vision and pattern recognition (CVPR) Weiss A, Hirshberg D, Blanc MJ (2011) Home 3D body scans from noisy image and range data. In: ICCV ’11 proceedings of the 2011 international conference on computer vision, pp 1951–1958 Wu Z, Song S, Khosla A, Yu F, Zhang L, Tang X, Xiao J (2015) 3d shapenets: A deep representation for volumetric shapes. In: Proceedings of the IEEE conference on computer vision and pattern recognition, pp 1912–1920 Zhao H, Reader AJ (2003). Fast ray-tracing technique to calculate line integral paths in voxel arrays. In: Proceedings of the IEEE nuclear science symposium and medical imaging conference, pp 2808–2812 Zivkovic Z (2004) Improved adaptive Gaussian mixture model for background subtraction. In: Pattern recognition, 2004. ICPR 2004. Proceedings of the 17th international conference on, vol 2, IEEE, pp 28–31

Three-Dimensional Human Kinematic Estimation Using Magneto-Inertial Measurement Units Andrea Cereatti, Ugo Della Croce, and Angelo M. Sabatini

Abstract

This chapter deals with the estimation of human kinematics using magneto and inertial sensing technology. A magneto-inertial measurement unit typically embeds a triaxial gyroscope, a triaxial accelerometer, and a triaxial magnetic sensor in the same assembly. By combining the information provided by each sensor within a sensor fusion framework, it is possible to determine the unit orientation with respect to a common global coordinate system. Recent advances in the construction of microelectromechanical system devices have made possible the manufacturing of small and light devices. These advances have widened the range of possible applications to include areas such as human movement. This chapter aims at providing the reader with a picture of the state of the art in the measurement and estimation methods for the description of human joint kinematics using magneto-inertial sensing technology. In the first section, fundamental concepts of rigid body kinematics are introduced with special reference to magneto-inertial measurements. Then a short description of the operational A. Cereatti (*) Department POLCOMING, University of Sassari, Sassari, Italy Interuniversity Centre of Bioengineering of the Human Neuromusculoskeletal System, University of Sassari, Sassari, Italy Department of Electronics and Telecommunications, Politecnico di Torino, Turin, Italy e-mail: [email protected] U. Della Croce Department POLCOMING, University of Sassari, Sassari, Italy Interuniversity Centre of Bioengineering of the Human Neuromusculoskeletal System, University of Sassari, Sassari, Italy e-mail: [email protected] A.M. Sabatini The BioRobotics Institute, Scuola Superiore Sant’Anna, Pisa, Italy e-mail: [email protected] # Springer International Publishing AG, part of Springer Nature 2018 B. Müller, S.I. Wolf (eds.), Handbook of Human Motion, https://doi.org/10.1007/978-3-319-14418-4_162

221

222

A. Cereatti et al.

characteristics of accelerometers, gyroscopes, and magnetometers is provided. The third section reports theory and methods for the estimation of the orientation and position of magneto-inertial measurement units along with the implementation of a Kalman filter for 3D orientation estimate as an example. In the last section, a critical review of the most common methodologies for the joint kinematic estimation is reported. Keywords

Joint mechanics • Acceleration • Angular velocity • Orientation • Position • Multisegmental model • Multibody • Anatomical coordinate system • Joint kinematics • Wearable sensors • Kalman filter • Pose Abbreviations

ALI ARW ACS BCS CoR CS DoFs EKF FUN KF GCS IMU MCS MEMS (M)IMU MUL NEMS VRW h, i  [q]

Anatomical landmark identification Angle Random Walk Anatomical coordinate system Body-fixed coordinate system Center of rotation Coordinate system Degree of freedom Extended Kalman filter Functional Kalman filter Global coordinate system Inertial measurement unit MIMU coordinate system Microelectromechanical systems (Magneto)-inertial measurement unit Manual Unit Alignment Nano-electromechanical systems Velocity Random Walk Dot product between vectors Quaternion multiplication Skew-symmetric operator

Contents Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . State of the Art . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Rigid Body Kinematics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Magneto-Inertial Measurement Technology . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Orientation and Position Estimates Using MIMU . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Three-Dimensional Human Joint Kinematics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Joint Positional Kinematic Estimation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Cross-References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

223 224 224 228 230 233 240 241 242 242

Three-Dimensional Human Kinematic Estimation Using Magneto-Inertial. . .

223

Introduction Human movement kinematics requires the description of the displacements, velocities, and accelerations, with respect to a global coordinate system (GCS), of every bony segment modeling the portion of skeletal system under analysis. In general, the body segments are assumed to be perfectly rigid and therefore to constitute a single rigid body with the underlying bone. Adjacent body segments are constrained through ideal joint models. Since the majority of the human joints allow only small relative linear displacements, whose amplitude is comparable to the errors associated to their estimate, only joint angular displacements are generally considered. This chapter deals with the estimation of the human kinematics using magneto and inertial sensing technology. The term inertial navigation refers to a set of techniques that exploit measurements of linear accelerations and angular velocities, with the aim of estimating the position and orientation (pose) of an object in the three-dimensional (3D) space relative to a known starting point, orientation, and velocity (Titterton and Weston 2004). An inertial measurement unit (IMU) is commonly used to measure linear accelerations and angular velocities. Typically, an IMU embeds a triaxial gyroscope and a triaxial accelerometer; another type of sensor that is commonly integrated in an IMU is a triaxial magnetic sensor, which measures the strength and direction of the local magnetic field, allowing the north direction to be found (Barbour and Schmidt 2001). Triaxial means that the sensor sensitivity axes are three, and they are mutually orthogonal to span the whole space and define the IMU-based coordinate system. The term MIMU (magneto-inertial measurement unit) is popularly used to denote a device that integrates accelerometer, gyroscope, and magnetic sensor in the same assembly (Bergamini et al. 2014). Inertial navigation was first developed for applications including navigation of aircraft, tactical and strategic missiles, submarines, and ships. Recent advances in the construction of microelectromechanical system (MEMS) devices have made possible the manufacturing of small and light IMUs. These advances have widened the range of possible applications to include areas such as human motion. Several key factors are behind the success of these sensing technologies. First, (M)IMUs are self-contained, in the sense that they do not rely on any external infrastructure to be operational. Second, since inertial and magnetic sensors are heavily used in the consumer electronics market, their price keeps dropping, while their performance improves. Lastly, the move from wearable measurement systems to pervasive systems made possible by the MEMS/NEMS (nano-electromechanical systems: NEMS) technology opens up new perspectives for motor performance assessment and monitoring. This chapter aims at providing the reader with a picture of the state of the art in the measurement and estimation methods for the description of the human joint kinematics using (M)IMUs. In the first section of the chapter, fundamental concepts of rigid body kinematics are introduced with special reference to magneto-inertial measurements. Then a short description of the operational characteristics of accelerometers, gyroscopes, and magnetometers is provided. The third section reports

224

A. Cereatti et al.

theory and methods for the estimation of the orientation and position of (M)IMUs along with the implementation of a Kalman filter for 3D orientation estimate as an example. In the final section, a critical review of the most common methodologies for joint kinematics estimation is reported.

State of the Art Rigid Body Kinematics Let B be a rigid body and P a point of B (Fig. 1). Two coordinate systems are introduced: the global (earth-fixed) coordinate system (GCS), specified by the origin OG and the right-handed orthonormal basis G = {Gx Gy Gz}, and the body-fixed coordinate system (BCS), specified by the origin OB and the right-handed orthonormal basis B = {Bx By Bz}. The motion of the body B is described by the translation of the origin OB and the rotation of


! BCS with respect to GCS. The translation is given by the vector G b ¼ OG OB of the

! origin OB relative to GCS. The vector G p ¼ OG P gives the coordinates of P relative

! to GCS; finally, the vector G r ¼ OB P is the vector from the origin OB to P (as seen from GCS): G

p ¼ G b þ G r,

(1)

With explicit notation, all vectors in Eq. 1 are measured in GCS. The vector Bω represents the angular velocity, resolved in the BCS, and it describes the rotational speed of B and its axis of rotation (Fig. 1). Since the body is rigid, the magnitude of G r is constant during the body motion. The vector Gr can be expressed in terms of the time-independent vector Br from OB to P as seen in BCS:

Fig. 1 Rigid body motion

Three-Dimensional Human Kinematic Estimation Using Magneto-Inertial. . . B

r ¼ BG CG r,

225

(2)

where C ¼ BG C is called the rotation matrix from GCS to BCS (Shuster 1993). An alternative representation is provided by the quaternion q ¼ BG q. The quaternion q is defined by: q ¼ ½q q4 T :

(3)

It is composed of the scalar component q4 and the vector component given by q = [qx qy qz]T. For the quaternion to be a valid parameterization of rotation, the following normalization constraint must be enforced: jqj2 þ q24 ¼ q2x þ q2y þ q2z þ q24 ¼ 1:

(4)

The quaternion multiplication between two generic quaternions q and h is defined as:  qh¼

 q4 h þ h4 q  q  h : q 4 h 4  qT h

(5)

A generic vector Gp can be transformed from one coordinate system (e.g., GCS) to another (e.g., BCS), by pre- and post-multiplying its quaternion (with scalar part equal to zero) by the rotation quaternion BG q and its inverse (i.e., the same quaternion with the vector part changed by sign), respectively: B

p ¼ BG q 

G  p  BG q1 0

(6)

According to the Euler’s theorem, the most general motion of a rigid body with one point fixed is a rotation by an angle θ (rotation angle) about some axis n (rotation axis). This yields another representation of the rigid body orientation in terms of a rotation vector: θ ¼ θn:

(7)

The rotation vector is related to the quaternion as follows: 

q ¼ qx

qy

qz

T

  θ ¼ sin n, 2

  θ q4 ¼ cos 2

(8)

The rotation matrix C can be expressed as a function of the quaternion q (Shuster 1993)

C ¼ 2q24  1 I33  ½q  2q4 þ qqT  2,

(9)

226

A. Cereatti et al.

where In  n is the n  n identity matrix (n = 3) and the skew-symmetric operator: 2

0 ½q ¼ 4 qz qy

qZ 0 qx

3 qy qx 5, 0

(10)

and denotes the matrix notation for the cross product. In other words, the skewsymmetric matrix can be used to represent the cross product between the vector q and a generic vector t in terms of the matrix-vector multiplication [q]t. When the BCS is moving with respect to the GCS, the rotation matrix C ¼ G BC can be shown to be the solution to the matrix differential equation: C_ ¼ ½ω C,

(11)

where ω = Bω. A triaxial gyroscope with the sensitivity axes aligned along the directions of BCS is customarily used in (M)IMUs to provide the (noisy) measurements of the angular velocity vector. Equation 11 can be reformulated in an equivalent form involving the quaternion q¼G B q and its first-order time derivative:  1 1 ½ω q_ ¼ ΩðωÞq ¼ 2 2 ωT

 ω q: 0

(12)

The solution to this system of first-order linear differential equations from known initial conditions provides therefore the orientation of the rigid body relative to the GCS. Furthermore, time differentiation of (11) yields:  € ¼ ½ω _  þ ½ω2 C: C

(13)

Using Eqs. 1 and 2, the velocity and the acceleration of the point P (relative to GCS) can then be written: G_ G _B p ¼ G b_ þ G B C r ¼ b þ ½ω r  G€ € B r ¼ Gb € þ GC € þ ½ω _  þ ½ω2 G r, p ¼ Gb B G_

A triaxial accelerometer located in P senses, resolved in BCS, the so-called €  B g , where Bg is the constant-orientation gravity vecspecific force B a ¼ B p tor (Ligorio and Sabatini 2016); for example, when the basis vector Gz is aligned parallel to the gravity vector, Gg = [0 0 g]T, with g = 9.81 m/s2. Let u = Br and θi , i = x , y , z denote the location and sensing directions of the accelerometer with respect to BCS, which are all time independent. Using (14), the components of the

Three-Dimensional Human Kinematic Estimation Using Magneto-Inertial. . .

227

specific force (accelerometer output) a = [ax ay a z]Tcan also be expressed with all quantities resolved in the BCS as: ai ðu, θi Þ ¼

D

 E _  þ ½ω2 u, θi , i ¼ 1, 2, 3 b  B g þ ½ω

B€

(15)

where h, i denotes the dot product between vectors. The first term within the dot product expresses the specific force as the additive combination of linear acceleration, gravity vector, and angular acceleration; the angular acceleration includes the _ tangential acceleration ½ωu and the centripetal acceleration [ω][ω]u. Suppose that the location of the accelerometer is where the origin of the BCS lies, i.e., u = 0. The angular acceleration is therefore null; without loss of generality, we can also assume that the

sensitive axes  of the triaxial accelerometer are oriented along the directions of Bx By Bz , namely, θ1 = [1 0 0]T , θ2 = [0 1 0]T , θ3 = [0 0 1]T. As for tracking the origin of the BCS, it is necessary, first, to rotate the (noisy) measured specific force from the BCS to the GCS using the rotational matrix computed from the integration of Eq. 11, or equivalently Eq. 12; second, the gravity vector contribution must be canceled by adding the known expression of Gg to the rotated measured specific force: G

G B BC a

€¼ p

þ G g:

(16)

When available in the GCS, the linear acceleration can be integrated once to obtain velocity and again to obtain displacement (strap-down approach to inertial navigation).

Determination of the Center of Rotation of a Rigid Body Let us consider the rigid body B constrained to a rigid frame through a spherical joint; the body can then only experience a pure rotational motion around the center of rotation (CoR). According to Eq. 14, assuming OBOG and coinciding with the CoR, the acceleration of a point P can be expressed as: G

 € ¼ ½ω _  þ ½ω2 G r, p

(17)

After some algebraic manipulation, Eq. 2 can be rearranged as: €, KG r ¼ G p

(18)

where 2

ω2y  ω2z



6

K¼6 4 ω_ z þ ωx ωy

ωx ωz  ω_ y



ωx ωy  ω_ z 2

ωx  ω2z

ω_ x þ ωy ωz



ω_ y þ ωx ωz

3



7 7 ωy ωz  ω_ x 5 ω2x  ω2y

(19)

228

A. Cereatti et al.

Equation 18 is linear in the unknown vector Gr, which represents the position € can be obtained from the accelerometric vector of CoR in the GCS. The vector G p measurements according to (16), and ω can be obtained from the gyroscope readings. The position vector of CoR can be then expressed in the BCS using either the rotation matrix or the rotation quaternion from GCS to BCS. In the presence of noise, a more reliable estimate of Gr can be obtained by computing (18), for each of the N sampled instants of time, recorded during a pure rotational motion of the rigid body, to obtain an over determined linear system which can be solved using a least-square technique.

Determination of the Axis of Rotation of a Rigid Body Let us consider the rigid body B constrained to a rigid frame through a revolute joint, which is rotating with an angular velocity ω around the single axis of rotation n. The most straightforward solution to compute the direction of n is from its angular velocity: n¼

ω , kωk

(20)

Alternatively, n can be also obtained from Eq. 8: n¼

q   , θ ¼ 2cos1 ðq4 Þ, θ sin 2

(21)

In the presence of noise, a more reliable estimate of r can be estimated either by selecting only angular velocity above a given threshold or by averaging the quaternion over the N observations (Prentice 1986).

Magneto-Inertial Measurement Technology Measurement System Description (M)IMUs for applications in the areas of human motion fall in the category of so-called strap-down systems. Since inertial sensors are rigidly mounted on the device, output quantities are measured in the BCS rather than the GCS (Titterton and Weston 2004). To track the (M)IMU orientation, the signals from the gyroscopes are time integrated. To track position, the signals from the accelerometer must be resolved into earth-fixed coordinate system (GCS) using the computed orientation and then integrated from known initial conditions. This procedure is shown in Fig. 2. In strap-down IMUs, the signals produced by the inertial sensors are resolved mathematically, prior to the calculation of navigational information. This reduces the mechanical complexity of the inertial navigation system, as it is implemented in the classical applications of inertial navigation technology, i.e., stable platform technique, thus decreasing the cost and size of the system and consequently increasing its

Three-Dimensional Human Kinematic Estimation Using Magneto-Inertial. . .

229

Fig. 2 Strap-down approach to inertial navigation

reliability. Nowadays, the processing speed and low cost of modern computers and microcomputers allow for the implementation of wearable strap-down IMUs for applications in human motion.

Measurement Characteristics of State-of-the-Art MEMS It is commonplace to distinguish different categories, or grades, of inertial sensors, which group them according to their expected performance, namely, in decreasing order of performance, “marine and navigation,” “tactical,” “industrial,” and “automotive and consumer” grades (Yazdi et al. 1998). Table 1 depicts the expected performance in terms of parts per million (ppm) of scale-factor stability (i.e., how well the sensor reproduces the sensed angular velocity or acceleration) and  /h or m/s2/h of inherent bias stability (i.e., the error independent of angular velocity or acceleration). While these performance factors are not the only ones that influence sensor selection, they are useful for comparison purposes. Quite invariably, the MEMS technologies that, for reasons of cost, complexity, size, and weight, are compatible with the requirements of human motion studies fall within the “automotive and consumer” grade. It is noted that in the absence of rotation (acceleration), the gyroscope (accelerometer) output is the sum of white noise and a slowly varying function (bias). The parameters Angle Random Walk (ARW) and Velocity Random Walk (VRW) reported in Table 1 are customarily used to quantify the white noise strength (in alternative to the RMS per square root of measurement bandwidth). These noise specifications describe the average deviation occurring when signals from the gyroscope and the accelerometer are integrated and when their actual estimation is based on Allan variance computation (El-Sheimy et al. 2008). Although the bias drift could be defined in different ways, the values reported in Table 1 are intended as the peak-to-peak value of the bias (Yazdi et al. 1998). The accuracy specifications of inertial sensors reported in Table 1 cannot be directly translated into pose error estimates. Qualitatively, longer time horizons and higher accuracy are achievable when higher-grade inertial sensors are used in a given application. However, other problems exist, which can be particularly critical because of the low performance requirements of MEMS/NEMS technologies. First, the difficulty of correctly interpreting the acceleration signals, when the component due to the gravity field (vertical reference) coexists with the component related to the

230

A. Cereatti et al.

Table 1 IMU accuracy specifications

Gyroscopes Angle random walk [ /h/√Hz] Bias drift [ /h] Scale factor stability [ppm] Accelerometers Velocity random walk [g/√Hz] Bias drift [m/s2/h] Scale factor stability [ppm]

Automotive and consumer

Tactical

Marine and navigation

>1 >100 >2000

0.1–0.5 5–50 500–1500

5000

50–75 500–1,000 1000–3000

100 steps recorded in a 24-h period. They also noted that only 75% of their sample was able to capture 5 days of monitoring. These results may be influenced by the attachment of the SW with a strap versus a knit cuff, as has been employed in previous work (Bjornson et al. 2007, 2014).The definition of the metric of “day” should be clearly defined and consistent in the processing of raw SW data with the proprietary software, to allow accurate interpretation pre-/post-intervention and across studies. The construct validity of SW watch monitoring in day-to-day life was examined by comparison to gait-lab-based walking speed, a summary score of gait deviation (Gait Deviation Index, GDI) and distance walked in 6 min (Wilson et al. 2015). Examining a cohort of 55 youth with CP, primarily at GMFCS level I, the authors documented a moderate relationship between GDI and SW average strides/day (r = 0.58) and that as the strides/day increase so does GDI. SW stride activity was also significantly correlated to lab-based walking speed and walking distance. This work suggests that interventions that improve gait kinematics (orthopedic surgery, orthotics) or decrease overall gait deviations (increase in GDI score toward normal) may enhance community walking activity as well. Replication of this work across lower functioning ambulatory children and pre-/post-intervention is needed to further understand this potential relationship. StepWatch capture of habitual walking activity has been employed as an outcome within a randomized intervention testing the effect of a 6-month physical activity stimulation program in the Netherlands (van Wely et al. 2014). No significant difference was documented in walking level or intensity between intervention groups at 6 and 12 months post. Habitual walking activity with and without current AFO prescription was examined with daily walking activity and intensity with the SW in 2016 (Bjornson et al. 2016). This work documented no group level condition effect (AFO ON vs. OFF) for community walking levels for clinically prescribed AFOs in 11 children with diplegia, across GMFCS levels I–III and various barefoot walking patterns (equinus, jump gait, and crouch gait). Two participants who did exhibit improved walking activity levels were prescribed similar AFO prescriptions. These studies suggest that monitoring with the SW is feasible and has potential to provide “real-world” information to inform clinical care and research outcomes.

Physical Activity in Children with CP (Actigraph) A 2012 study utilizing the ActiGraph GT1M examined 23 ambulatory and non-ambulatory adolescents with CP (13.5, SD 2.6 year) at GMFCS levels I–IV over a 7-day period. Participants wore the device from 540.5 to 859.2 min per day and engaged in 89.5  47.1 min of LPA, 17.8  16.9 min of MPA, 12.0  14.4 min of VPA, and 30.7  30.3 min of MVPA. Youth classified at GMFCS level IV presented with lower levels of LPA, MPA, and MVPA compared with level I (P < 0.05). Similarly, youth at level III demonstrated lower levels of

Walking and Physical Activity Monitoring in Children with Cerebral Palsy

1023

MPA and MVPA compared with level I. No differences were seen between levels I and II for any intensity. MVPA and GMFCS levels were negatively correlated (minutes/day: τ = 0.65, P < 0.001) (Gorter et al. 2012). A 2015 study of 102 children (11  2 years) with spastic hemiplegia classified at GMFCS levels I and II found that only 25% of participated in 60 min of MVPA on at least one of 4 days. The mean totals for the group were 438 counts/min and 7,541 steps per day and spent 8:36 h (72% of recorded time) in sedentary time, 2:38 h in light activities (22% of recorded time), and 0:44 h (6% of recorded time) in MVPA. There were no significant differences in physical activity between children classified at GMFCS levels I and II. Children were significantly more physically active than adolescents. Boys were significantly more physically active than girls. More steps were recorded on weekdays than on weekend days (Mitchell et al. 2015).

Physical Activity in Children with CP Uptimer The Uptimer is a device validated to measure time spent upright and was used in a study of 300 children and youth with CP. The authors found that compared to 5.6 h per day of “uptime” in able-bodied peers, youth with CP described to have hemiplegia (n = 115) spent 5.1 h/day in upright time, youth with diplegia (n = 113) spent 2.5 h/day in upright time, and youth with quadriplegia (n = 72) spent 0.5 h/day in upright time. Analysis of variance revealed significant differences between all groups (P < 0.001), including the nondisabled comparison group (Pirpiris and Graham 2004). Walking Accelerometry Combined with Global Positioning System (GPS) in Children with CP A synchronization of SW walking activity and Global Positioning System (GPS) data has been piloted in 12 ambulatory children with CP (R21 HD 077186) who underwent 20 sessions of short burst interval treadmill training (SBLTT) (Bjornson, Moreau, Hurvitz, Kerfeld 2016 unpublished data). SW accelerometry walking activity data were time matched with GPS records to document walking in the home and community over a 7-day sample through distance walked, average strides per day, and percent of time each day spent in low (1–30 strides/min), moderate (31–60 stride/min), and high (>60 stride/min) stride rates. Combined StepWatch/GPS pilot data demonstrated an increase in the percentage of overall strides per day ambulated in the community setting from 44.2% to 49.8%, suggesting an increase in community walking participation at 6 weeks post-SBLTT training (Fig. 2). Figure 3 illustrates a StepWatch/GPS a synchronization map for a child with CP at GMFCS level I over one measurement day of walking activity in the Seattle, Washington, area. The StepWatch data were classified into intensity levels based on stride rate. Each GPS location was matched with the walking intensity level at the same time point, allowing measurement of where walking across these levels occurred in a spatial context. For this child, the moderate stride activity occurred at school and a soccer field. This preliminary data provides feasibility of this novel combined SW/GPS methodology and potential sensitivity for the amount and location of community strides and intensity levels.

1024

K.F. Bjornson and N. Lennon

Fig. 2 Percentage of total walking (avg. strides/day) in the community (blue) vs in the home (red) after 20 sessions of interval treadmill training (n = 12)

Fig. 3 Synchronized StepWatch/GPS map in Seattle for a child with CP (GMFCS I). X = no walking (nw); Ο = low stride rates; ☐ = moderate stride rates; Δ = home

Summary Our knowledge of the real-world physical activity and walking activity habits of children and youth with CP has expanded over the past decade. The StepWatch, the Actigraph, and Uptimer have provided data to show that ambulatory youth with CP, those at GMFCS levels I, II, and III, are more sedentary, spend less time upright, spend less time being physically active, and take fewer steps per day than TDY. Studies using these devices have shown that time upright, time active, and amount of walking are strongly associated with motor ability, as reported by GMFCS classification levels. Walking and activity habits have similarities to TDY in that school-age

Walking and Physical Activity Monitoring in Children with Cerebral Palsy

1025

children with CP take more steps and have higher PA levels than adolescents with CP. The PA and walking activity habits of school-age children have also revealed that youth with CP tend to take more steps on school days compared to non-school days. This recent addition of knowledge about the activity habits of children and youth with CP can be utilized on a broad scale to tailor PA opportunities and programs to promote greater activity and participation and can be used on an individual level in physical therapy and home programming. We did not identify for this review studies utilizing the BMSW or the VitaMove monitors to report on PA and WA for children and youth with CP, although we acknowledge that this is not an exhaustive review.

Monitoring of Walking and Physical Activity: Clinical Implications for Children CP Interventions for Walking Activity The literature suggests that the use of walking and physical activity monitoring in daily life is clinically feasible and has significant potential to be employed as a clinical outcome to inform the care and management of children with cerebral palsy. Community-based monitoring with devices reviewed in this chapter can be employed within numerous rehabilitation strategies in this population including but not limited to medications for movement disorders, injection therapy to lower and upper extremity, upper extremity constraint-induced therapy, orthotic management, use of assistive mobility devices, (walker versus crutches) gait training (i.e., overground, treadmill), orthopedic surgery, and neurosurgical interventions (intrathecal baclofen, selective dorsal rhizotomy) as well as lifestyle intervention to enhance physical activity. We will review preliminary and published data describing the application of monitoring devices to describe community-based walking and physical activity in children with CP. All children with cerebral palsy (CP) exhibit some type of movement disorder (i.e., spasticity, dystonia etc.). Often oral medications are employed to decrease the influence of these movement disorders on activities of daily life such as sitting, walking, and upper extremity tasks of daily life. A monitoring device (Actigraph on the wrist) can be employed to document the influence of oral medications on overall excessive movement in a child with upper extremity dystonia during feeding, for example. In this clinical scenario, we would hope to see less extraneous limb movement if the medication is having the desired effect. Similarly, pre-/post-upper extremity injections to optimize upper extremity use for feeding could employ the Actigraph during feeding of one arm. Devices (i.e., three-dimensional accelerometers) on both wrists could be used to document the relative change in bilateral upper extremity use post-constraint-induced therapy treatment. Walking activity in daily life captured by wearable devices can offer valuable information to guide interventions which are employed to optimize walking within the context of the child’s environment. A 2016 clinical pilot study of 11 children with CP and bilateral impairment employed 2 weeks monitoring with the StepWatch to

1026

K.F. Bjornson and N. Lennon

examine the influence of current orthotic prescriptions on community walking (Bjornson et al. 2016). StepWatch data captured strides/day, time walking, as well as walking intensity. Depending on the walking (SW) outcome examined, only two to four of the 11 participants demonstrated improvements with their current orthotic prescription as compared to walking without their orthotics. Recent work has applied StepWatch monitoring pre- and post-short-burst interval locomotor treadmill training (SBLTT) in ambulatory school-aged children with CP (unpublished Bjornson et al. 2016 R21 results). R21 pilot data of pre-/post-SBLTT (n = 12, Figs. 4 and 5) documented enhanced community walking levels (average strides/day +948 strides/day, p < 0.001) at 6 weeks post-SBLTT training. Percent time walking at higher stride/min intensities of medium/high stride (>39 strides/ min) rates increased (+3.8%, p = 0.04) including absolute number of strides/day at medium/high stride rates (+627, p < 0.001). Studies of orthopedic surgery outcomes in lab settings for youth with cerebral palsy (CP) who undergo surgical correction to improve gait reveal positive changes in body structure/function measures such as kinematic and kinetic patterns (Wren et al. 2013) and improvement in activity capacity measures such as gait speed (Gannotti et al. 2007) after surgery. The validation of physical activity (PA) monitors for youth with CP now allows for investigation of performance level outcomes such as change in habitual level of walking activity post-surgery. Lennon and colleagues used the StepWatch to examine the recovery of walking activity in youth who underwent orthopedic surgery to correct gait (Lennon et al. 2015). Preoperative clinical gait analysis at the Nemours duPont Hospital for Children includes collection of habitual walking activity pre- and postoperatively. The surgeons, physiatrists, and physical therapists at the hospital evaluate walking activity data as part of rehabilitation planning during the child’s recovery from

Fig. 4 Average strides/day at baseline, post-SBLTT, and 6 weeks post-SBLTT (n = 12)

Walking and Physical Activity Monitoring in Children with Cerebral Palsy

1027

Fig. 5 Percent of total strides/day in low, moderate and high stride rates for 368 typically developing youth (TDY), (Bjornson et al. 2014) 209 youth with CP and 12 youth with CP, Pre/post short-burst LTT. Green- high= >60 stride/min, Red-moderate=30-60 stride/min and Blue- low 0.05) (Lennon et al. 2015). Figure 6 shows recovery trends of both groups relative to expected strides for GMFCS level (Bjornson et al. 2007).

1028

K.F. Bjornson and N. Lennon

Interventions for Physical Activity The Carol and Paul Hatfield Cerebral Palsy Sports and Rehabilitation Center at The St. Louis Children’s Hospital provide children with cerebral palsy and related childhood disabilities the opportunity to experience sport activities in fun, social environments. Known as “Camp Independence,” it is a summer program of intensive physical activities in a sports camp format. Participants are assisted by physical therapists, nurses, aides, and volunteers to play many different sporting activities. Campers spend 7 h a day for up to 4 weeks participating in sport activities such as tennis, swimming, yoga, martial arts, basketball, soccer, cycling, baseball, and dance. Each sport is adapted as necessary to meet the needs of the child. During the summer of 2014, Miros and colleagues recruited 34 campers to wear a FitBit Flex activity tracking device for 2 weeks before camp, while they were attending camp, and a minimum of 2 weeks after camp (St. Louis Childrens Hospital 2015). All activity data were collected using the FitBit app and the fitbit.com website (FitBit 2016). Campers also completed a number of standardized physical outcome measures at the beginning and end of camp to evaluate the functional benefits of camp. Participants were between the ages of 7 and 18 and ranged in physical ability from independently being able to run to requiring a power wheelchair for community mobility. Although there were no significant changes in standardized physical outcome measures, participants were more active on camp days versus non-camp days. The participants averaged 8,693 steps on days they attended camp and 5,730 steps on days they did not attend camp. Activity levels decreased after camp, which families attribute to the lack of community physical activity opportunities accommodating children with CP (St. Louis Childrens Hospital 2015). Active video games (AVGs) are another promising strategy to provide access to fun, active recreation for youth with CP. O’Neil and colleagues examined PA intensity levels in 57 youth with CP (mean = 12 years) at GMFCS levels I (28), II (16), and III (13) during 5 min of play on the X-Box 360 Kinect™ Adventures: (River Rush and Space Pops) games (O’Neil et al. 2916). Data was collected utilizing Polar heart rate monitors, the Cosmed K4b2 indirect calorimeter device, the OMNI rate of perceived exertion (RPE) scale and the Actigraph SW, and BMSW activity monitors. They found that youth played AVGs at light-to-moderate PA intensity. Median MET values were River Rush = 3.0, IQR = 2.5–3.8, and Space Pops = 3.3, IQR = 2.6–3.8. Median OMNI RPE values were River Rush = 3.8, IQR = 1–4, and Space Pops = 4.0 (IQR = 2–6). Findings suggest that commercial AVGs may provide opportunities to promote PA in ambulatory youth with CP but are limited in their ability to promote higher levels of PA intensity and longer bouts of PA (O’Neil et al. 2916, 2016). Researchers and entrepreneurs are now working together to develop custom gaming platforms to meet the needs of youth with physical disabilities. A game that allows physical therapists to change the speed and intensity of the game through a back-end portal allows the therapist to structure game conditions to reach the heart rate and level of physical activity intensity individuals most benefit from. At the same time, the game system can collect data on a player’s progress over a session,

Walking and Physical Activity Monitoring in Children with Cerebral Palsy

1029

several sessions, or several weeks. The flexibility of gaming parameters allows the therapist to adjust the game to address therapeutic goals. Having the players wear heart rate monitors and activity monitors during game play allows the therapist to see if the child is reaching the aerobic and activity levels that are health promoting. To date, O’Neil’s pilot study has enrolled 12 children and youth to play three newly developed 20-min games (O’Neil et al. 2016, unpublished work). Youth have enjoyed playing, which is an important factor. Having a fun way to keep kids active has the potential to mitigate. In 2014, van Wely and colleagues conducted a randomized trial examining the effect of 6-month physical activity stimulation program (van Wely et al. 2010, 2014). The investigators employed the StepWatch to capture habitual walking activity. No significant difference was documented in walking level or intensity between intervention groups at 6 and 12 months post. A randomized control trial of an internetbased physical activity intervention was completed by Maher and colleagues in 2010 (Maher et al. 2010). Employing a waist-mounted pedometer (NL1000), no significant differences between intervention and comparison group were documented for number of steps taken per week or self-reported physical activity at 10 and 20 weeks post.

Considerations and Limitations of Walking and Physical Activity Monitoring When selecting a monitor or device to capture WA or PA, there are several factors to consider. First, it is important to select a device that is validated and most directly related to the behavior you are aiming to capture and/or influence with intervention. For example, for walking activity post-lower extremity surgical intervention, you may choose the StepWatch, while physical activity after a home-based AVG program maybe most appropriately measured with the Actigraph. Second, the costs of these devices vary from $30 to $500 per device, and this may or may not include the processing software to access meaningful information to interpret. Third, the ease of processing and getting to outputs that can be readily interpreted should be considered as well. The addition of bluetooth versions (wireless/cloud-based) of devices will now make downloading relatively easier (i.e., StepWatch and Actigraph). Lastly, in a clinical setting, billing for the assessment of walking and/or physical activity can be included within a clinical gait or mobility evaluation and/or a post-intervention outcome assessment physical therapy charge by a physical therapist. The emerging data on WA and PA in children and youth with CP reveals high variability among many of the study samples. Thus, it is important that this variability is taken into account in developing study designs/sample sizes to insure that work is adequately powered to capture clinically meaningful as well as statistical significance. For WA activity in children with CP, a potential “minimal clinical important difference” (MCID) benchmark could be the mean difference between walking activity between GMFCS levels. For example, enroll a sample size adequate

1030

K.F. Bjornson and N. Lennon

to capture the mean difference between WA of children with CP at GMFCS level I versus II (based on current published WA levels) (Bjornson et al. 2014). PA monitors have potential weaknesses that should be considered with their implementation as well. The accuracy and precision of instruments worn on the waist or upper arm can be limited for certain types of upright behaviors that have a low ambulatory component and that may involve upper body work. Similarly WA devices worn on the wrist may not adequately capture the abnormal walking patterns of children with CP and/or after interventions (i.e., orthotics, orthopedic surgery). Recent efforts to enhance the ability of physical activity monitors to capture these behaviors include more densely sampled data and more sophisticated prediction equations. Information about the location or purpose of individual activities is limited, unless information from other sources is integrated with information from the monitor. Technological solutions synchronizing GPS with accelerometers previously employed in populations without physical disability are being examined in children with CP (Bjornson et al. 2016, unpublished data).

Conclusions/Summary In summary, accelerometers have been validated and reliably employed to document walking and physical activity of children with cerebral palsy across functional levels. Relative to walking activity, published information to date offers reference information for walking activity levels, walking patterns, and intensity of walking by GMFCS levels as compared to children without motor limitations with the StepWatch. Physical activity has been documented for ambulatory and non-ambulatory children with CP with several devices (i.e., Actigraph, BodyMedia, etc.) and validated to energy cost in ambulatory children. Cut points for interpretation of PA intensity with the Actigraph in children with CP of varying ages are emerging in the literature. Based on the body of knowledge reviewed in this chapter, accelerometry can be effectively employed to capture walking and physical activity in children and youth with CP. This knowledge should inform clinical and/or research questions relative to performance by functional levels, documenting natural history and directing rehabilitation strategies as ecologically based outcomes of WA and PA within the context of daily life. Future development of WA and PA monitoring in children with CP should focus on the capturing upper extremity functional activities, expand the use of bluetooth technology and/or cloud-based capture of data and synchronization accelerometry with GPS and patient/parent reported qualitative outcomes.

Cross-References ▶ Assessing Clubfoot and Cerebral Palsy by Pedobarography ▶ Assessing the Impact of Aerobic Fitness on Gait ▶ Clinical Gait Assessment by Video Observation and 2D Techniques

Walking and Physical Activity Monitoring in Children with Cerebral Palsy

1031

▶ Diagnostic Gait Analysis Use in the Treatment Protocol for Cerebral Palsy ▶ EMG Activity in Gait: The Influence of Motor Disorders ▶ Foot and Ankle Motion in Cerebral Palsy ▶ Functional Effects of Foot Orthoses ▶ Gait Scores: Interpretations and Limitations ▶ Interpreting Spatiotemporal Parameters, Symmetry, and Variability in Clinical Gait Analysis ▶ Natural History of Cerebral Palsy and Outcome Assessment ▶ Oxygen Consumption in Cerebral Palsy ▶ Spasticity Effect in Cerebral Palsy Gait ▶ Strength Related Stance Phase Problems in Cerebral Palsy ▶ Swing Phase Problems in Cerebral Palsy

References Alotaibi M, Long T, Kennedy E, Bavishi S (2014) The efficacy of GMFM-88 and GMFM-66 to detect changes in gross motor function in children with cerebral palsy (CP): a literature review. Disabil Rehabil 36:617–627 Andreacci JL, Dixon C, Dube JJ, McConnell TR (2007) Validation of SenseWear Pro2 armband to assess energy expenditure during treadmill exercise in children 7–10 years of age. J Exerc Physiol Online 10:35–42 Bagley AM, Gorton GE, Bjornson K, Bevans K, Stout JL, Narayanan U, Tucker CA (2011) Factorand item-level analyses of the 38-item activities scale for kids–performance. Dev Med Child Neurol 53:161–166 Barreira TV, Katzmarzyk PT, Johnson WD, Tudor-Locke C (2012) Cadence patterns and peak cadence in US children and adolescents: NHANES, 2005–2006. Med Sci Sports Exerc 44:1721–1727 Bassett DR, Dinesh J (2010) Use of pedometers and accelerometers in clinical populations: validity and reliability issues. Phys Ther Rev 15(3):135–142 Beckung E, Hagberg G (2002) Neuroimpairments, activity limitations, and participation restrictions in children with cerebral palsy. Dev Med Child Neurol 44:309–316 Beckung E, Carlsson G, Carlsdotter S, Uvebrant P (2007) The natural history of gross motor development in children with cerebral palsy aged 1 to 15 years. Dev Med Child Neurol 49:751–756 van den Berg-Emons HJ, Saris WH, de Barbanson DC, Westerterp KR, Huson A, van Baak MA (1995) Daily physical activity of school children with spastic diplegia and of healthy control subjects. J Pediatr 127:578–584 Bjornson KF, Belza B, Kartin D, Logsdon R, McLaughlin JF (2007) Ambulatory physical activity performance in youth with cerebral palsy and youth who are developing typically. Phys Ther 87:248–257 Bjornson K, Song K, Lisle J, Robinson S, Killien E, Barrett T, Zhou C (2010) Measurement of walking activity throughout childhood: influence of leg length. Pediatr Exerc Sci 22:581–595 Bjornson KF, Song K, Zhou C, Coleman K, Myaing M, Robinson SL (2011) Walking stride rate patterns in children and youth. Pediatr Phys Ther 23:354–363 Bjornson KF, Zhou C, Stevenson RD, Christakis D (2014a) Relation of stride activity and participation in mobility-based life habits among children with cerebral palsy. Arch Phys Med Rehabil 95:360–368 Bjornson K, Zhou C, Stevenson RD, Christakis D, Song K (2014b) Walking activity patterns in youth with cerebral palsy and youth developing typically. Disabil Rehabil 36:1279–1284

1032

K.F. Bjornson and N. Lennon

Bjornson K, Zhou C, Fatone S, Orendurff M, Stevenson R, Rashid S (2016) The effect of ankle-foot orthoses on community-based walking in cerebral palsy: a clinical pilot study. Pediatr Phys Ther 28:179–186 Bohannon R (1997) Comfortable and maximum walking speed of adults aged 20–79 years: reference values and determinants. Age Ageing 26:15–19 Bürgi R, Tomatis L, Murer K, de Bruin ED (2015) Localization of physical activity in primary school children using accelerometry and global positioning system. PLoS One 10:e0142223 Bussmann JBJ, Ebner-Priemer U, Fahrenberg J (2009) Ambulatory activity monitoring: progress in measurement of activity, posture, and specific motion patterns in daily life. Eur Psychol 14:142–152 Cain K, Salis JF, Conway TL, van Dyck D, Calhoon L (2013) Using accelerometers in youth physical activity studies: a review of methods. J Phys Act Health 10:437–450 Capio CM, Sit CHP, Abernethy B, Masters RSW (2012) Fundamental movement skills and physical activity among children with and without cerebral palsy. Res Dev Disabil 33:1235–1241 Carlson JA, Schipperijn J, Kerr J, Saelens BE, Natarajan L, Frank LD, Glanz K, Conway TL, Chapman JE, Cain KL, Sallis JF (2016) Locations of physical activity as assessed by GPS in young adolescents. Pediatrics 137:e20152430 Caspersen CJ, Powell KE, Christenson GM (1985) Physical activity, exercise, and physical fitness: definitions and distinctions for health-related research. Public Health Report 100:126–131 Cesari M, Kritchevsky SB, Penninx BWHJ, Nicklas BJ, Simonsick EM, Newman AB, Tylavsky FA, Brach JS, Satterfield S, Bauer DC, Visser M, Rubin SM, Harris TB, Pahor M (2005) Prognostic value of usual gait speed in well-functioning older people – results from the health, aging and body composition study. J Am Geriatr Soc 53:1675–1680 Chillón P, Panter J, Corder K, Jones AP, van Sluijs EMF (2015) A longitudinal study of the distance that young people walk to school. Health Place 31:133–137 Chrysagis N, Skordilis EK, Koutsouki D (2014) Validity and clinical utility of functional assessments in children with cerebral palsy. Arch Phys Med Rehabil 95:369–374 Clanchy K, Tweedy S, Boyd R, Trost S (2011) Validity of accelerometry in ambulatory children and adolescents with cerebral palsy. European J Appl Physiol 111(12):2951–9. https://doi.org/ 10.1007/s00421-011-1915-2 Claridge EA, McPhee PG, Timmons BW, Martin Ginis KA, Macdonald MJ, Gorter JW (2015) Quantification of physical activity and sedentary time in adults with cerebral palsy. Med Sci Sports Exerc 47:1719–1726 Corder K, Brage S, Ramachandran A, Snehalatha C, Wareham N, Ekelund U (2007) Comparison of two Actigraph models for assessing free-living physical activity in Indian adolescents. J Sports Sci 25:1607–1611 Del Pilar Duque Orozco M, Abousamra O, Church C, Lennon N, Henley J, Rogers KJ, Sees JP, Connor J, Miller F (2016) Reliability and validity of Edinburgh visual gait score as an evaluation tool for children with cerebral palsy. Gait Posture 49:14–18 Demant Klinker C, Schipperijn J, Toftager M, Kerr J, Troelsen J (2015) When cities move children: development of a new methodology to assess context-specific physical activity behaviour among children and adolescents using accelerometers and GPS. Health Place 31:90–99 Diggory P, Gorman M, Schwarz J, Helme R (1994) An automatic device to measure time spent upright. Clin Rehabil 8:353–357 Evenson KR, Catellier DJ, Gill K, Ondrak KS, McMurray RG (2008) Calibration of two objective measures of physical activity for children. J Sports Sci 26:1557–1565 FITBIT (2016) Fitbit Products [Online]. Available: https://www.fitbit.com/ Foster RC, Lanningham-Foster LM, Manohar C, McCrady SK, Nysse LJ, Kaufman KR, Padgett DJ, Levine JA (2005) Precision and accuracy of ankle-worn accelerometer-based pedometer in step counting and energy expenditure. Prev Med 41:778–783 Fritz S, Lusardi M (2009) White paper: “walking speed: the sixth vital sign”. J Geriatr Phys Ther 32:46–49

Walking and Physical Activity Monitoring in Children with Cerebral Palsy

1033

Gannotti ME, Gorton GEI, Nahorniak MT, Masso PD, Landry B, Lyman J, Sawicki R, Hagedorn K, Ross E, Warner J (2007) Postoperative gait velocity and mean knee flexion in stance of ambulatory children with spastic diplegia four years or more after multilevel surgery. J Pediatr Orthop 27:451–456 Gor-García-Fogeda MD, Cano de la Cuerda R, Carratalá Tejada M, Alguacil-Diego IM, MolinaRueda F (2016) Observational gait assessments in people with neurological disorders: a systematic review. Arch Phys Med Rehabil 97:131–140 Gorter JW, Noorduyn SG, Obeid J, Timmons BW (2012) Accelerometry: a feasible method to quantify physical activity in ambulatory and nonambulatory adolescents with cerebral palsy. Int J Pediatr 2012:6 Gorton GE III, Stout JL, Bagley AM, Bevans K, Novacheck TF, Tucker CA (2011) Gillette functional assessment questionnaire 22-item skill set: factor and Rasch analyses. Dev Med Child Neurol 53:250–255 Graham HK, Harvey A, Rodda J, Nattrass GR, Pirpiris M (2004) The functional mobility scale (FMS). J Pediatr Orthop 24:514–520 Granger CV, Hamilton BB (1993) The uniform data system for medical rehabilitation report of first admissions for 1991. Am J Phys Med Rehabil 72:33–38 Haak P, Lenski M, Hidecker MJC, Li M, Paneth N (2009) Cerebral palsy and aging. Dev Med Child Neurol 51:16–23 Health M (2015) StepWatch [Online]. Available: https://modushealth.com/. Accessed 17 Nov 2015 Healthcare O (2016) Omron Pedometers [Online]. Available: https://omronhealthcare.com/fitness/ activity-trackers-and-pedometers/ Imms C, Reilly S, Carlin J, Dodd K (2008) Diversity of participation in children with cerebral palsy. Dev Med Child Neurol 50:363–369 Iannotti RJ, Wang J (2013) Patterns of physical activity, sedentary behavior, and diet in U.S. adolescents. J Adolescent Health 53(2):280–6. https://www.ncbi.nlm.nih.gov/pubmed/ 23642973 Ishikawa S, Kang M, Bjornson KF, Song K (2013) Reliably measuring ambulatory activity levels of children and adolescents with cerebral palsy. Arch Phys Med Rehabil 94:132–137 Kamp FA, Lennon N, Holmes L, Dallmeijer AJ, Henley J, Miller F (2014) Energy cost of walking in children with spastic cerebral palsy: relationship with age, body composition and mobility capacity. Gait Posture 40:209–214 Kang M, Bjornson KF, Barreira TV, Ragan BG, Song K (2014) The minimum number of days required to establish reliable physical activity estimates in children aged 2–15 years. Physiol Meas 35:2229 Karabulut M, Crouter S, Bassett DR (2005) Comparison of two waist-mounted and two anklemounted electronic pedometers. Eur J Appl Physiol 96:334–335 Keawutan P, Bell K, Davies PSW, Boyd RN (2014) Systematic review of the relationship between habitual physical activity and motor capacity in children with cerebral palsy. Res Dev Disabil 35:1301–1309 Keawutan P, Bell KL, Oftedal S, Davies PSW, Boyd RN (2016) Validation of accelerometer cut-points in children with cerebral palsy aged 4 to 5 years. Pediatr Phys Ther 28:427–434 Kerr C, Parkes J, Stevenson M, Cosgrove AP, McDowell BC (2008) Energy efficiency in gait, activity, participation and health status in children wtih cerebral palsy. Dev Med Child Neurol 50:204–210 Ko J, Kim M (2013) Reliability and responsiveness of the gross motor function measure-88 in children with cerebral palsy. Phys Ther 93:393–400 Koehler K, Abel T, Wallmann-Sperlich B, Dreuscher A, Anneken V (2015) Energy expenditure in adolescents with cerebral palsy: comparison of the SenseWear armband and indirect calorimetry. J Phys Act Health 12:540–545 Kozey SL, Staudenmayer JW, Troiano RP, Freedson PS (2010) Comparison of the ActiGraph 7164 and the ActiGraph GT1M during self-paced locomotion. Med Sci Sports Exerc 42:971–976

1034

K.F. Bjornson and N. Lennon

Lee J-M, Kim Y, Welk GJ (2014) Validity of consumer-based physical activity monitors. Med Sci Sports Exerc 46:1840–1848 Lee J-M, Kim Y, Bai Y, Gaesser GA, Welk GJ (2016) Validation of the SenseWear mini armband in children during semi-structure activity settings. J Sci Med Sport 19:41–45 Lennon N, Hulbert R, Church C, Miller F (2015) Surgical burden and recovery of walking performance in youth with cerebral palsy. Dev Med Child Neurol 57:96–97 Maher CA, Williams MT, Olds TIM, Lane AE (2010) An internet-based physical activity intervention for adolescents with cerebral palsy: a randomized controlled trial. Dev Med Child Neurol 52:448–455 McCrorie PRW, Fenton C, Ellaway A (2014) Combining GPS, GIS, and accelerometry to explore the physical activity and environment relationship in children and young people – a review. Int J Behav Nutr Phys Act 11:93 McDonald CM, Widman L, Abresch T, Walsh S, Walsh D (2005) Utility of a step activity monitor for the measurement of daily ambulatory activity in children. Arch Phys Med Rehabil 86:793–801 Melanson EL, Knoll JR, Bell ML, Donahoo WT, Hill JO, Nysse LJ, Lanningham-Foster L, Peters JC, Levine JA (2004) Commercially available pedometers: considerations for accurate step counting. Prev Med 39:361–368 Mitchell LE, Ziviani J, Boyd RN (2015) Habitual physical activity of independently ambulant children and adolescents with cerebral palsy: are they doing enough? Phys Ther 95:202 Mitre N, Lanningham-Foster L, Foster R, Levine JA (2009) Pedometer accuracy in children: can we recommend them for our obese population? Pediatrics 123:e127–e131 Morris C, Kurinczuk JJ, Fitzpatrick R, Rosenbaum PL (2006) Do the abilities of children with cerebral palsy explain their activities and participation? Dev Med Child Neurol 48:954–961. Dev Med Child Neurol. 2007 Feb;49(2):122 Nicholson K, Lennon N, Hulbert R, Church C, Miller F (2017) Pre-operative walking activity in youth with cerebral palsy. Res Dev Disabil 60:77–82 Nooijen CF, de Groot JF, Stam HJ, van den Berg-Emons RJ, Bussmann HB (2015) Validation of an activity monitor for children who are partly or completely wheelchair-dependent. J Neuroeng Rehabil 12:11 O’Neil ME, Fragala-Pinkham MA, Forman JL, Trost SG (2014) Measuring reliability and validity of the ActiGraph GT3X accelerometer for children with cerebral palsy: a feasibility study. J Pediatr Rehabil Med 7:233–240 O’Neil ME, Fragala-Pinkham M, Lennon N, George A, Forman J, Trost SG (2016) Reliability and validity of objective measures of physical activity in youth with cerebral palsy who are ambulatory. Phys Ther 96:37–45 O’Neil ME, Fragala-Pinkham M, Lennon N, Trost SG (2916) Active video games to promote physical activity in youth with cerebral palsy. In: IV STEP 2016. Columbus: Pediatric and Neurology Sections, American Physical Therapy Association Oeffinger D, Bagley A, Rogers S, Gorton G, Kryscio R, Abel M, Damiano D, Barnes D, Tylkowski C (2008) Outcome tools used for ambulatory children with cerebral palsy: responsiveness and minimum clinically important differences. Dev Med Child Neurol 50:918–925 Oftedal S, Bell KL, Davies PSW, Ware RS, Boyd RN (2014) Validation of accelerometer cut points in toddlers with and without cerebral palsy. Med Sci Sports Exerc 46:1808–1815 Oliver M, Badland H, Mavoa S, Duncan MJ, Duncan S (2010) Combining GPS, GIS, and accelerometry: methodological issues in the assessment of location and intensity of travel behaviors. J Phys Act Health 7:102–108 Palisano R, Rosenbaum P, Bartlett D, Livingston MH (2007) GMFCS- R & E gross motor function classification system expanded and revised, 2007. CanChild Centre for Childhood Disability, McMasters University, Hamilton Pirpiris M, Graham HK (2004) Uptime in children with cerebral palsy. J Pediatr Orthop 24:521–528 Play G (2016) Virtual walk treadmill or GPS [Online]. Available: https://play.google.com/store/ apps/details?id=com.virtualwalk.android&hl=en

Walking and Physical Activity Monitoring in Children with Cerebral Palsy

1035

Portney LW, Watkins MP (2009) Foundations of clinical research: applications to practice. Pearson/ Prentice Hall, Upper Saddle River Postma K, van den Berg-Emons HJG, Bussmann JBJ, Sluis TAR, Bergen MP, Stam HJ (2005) Validity of the detection of wheelchair propulsion as measured with an activity monitor in patients with spinal cord injury. Spinal Cord 43:550–557 Puyau MR, Adolph AL, Vohra FA, Butte NF (2002) Validation and calibration of physical activity monitors in children. Obes Res 10:150–157 Puyau MR, Adolph AL, Vohra FA, Zakeri I, Butte NF (2004) Prediction of activity energy expenditure using accelerometers in children. Med Sci Sports Exerc 36:1625–1631 Rothney MP, Apker GA, Song Y, Chen KY (2008) Comparing the performance of three generations of ActiGraph accelerometers. J Appl Physiol 105:1091–1097 Ryan JM, Walsh M, Gormley J (2014) A comparison of three accelerometry-based devices for estimating energy expenditure in adults and children with cerebral palsy. J Neuroeng Rehabil 11:116 Services, U. S. D. O. H. A. H. (2012) Physical activity guidelines for Americans: fact sheet for professionals [Online]. Available: http://www.health.gov/paguidelines/factsheetprof.aspx. Accessed 06 Feb 2013 Sirard JR, Trost SG, Pfeiffer KA, Dowda M, Pate RR (2005) Calibration and evaluation of an objective measure of physical activity in preschool children. J Phys Act Health 3:345–357 Slaman J, Dallmeijer A, Stam H, Russchen H, Roebroeck M, van den Berg-Emons R (2013) The six-minute walk test cannot predict peak cardiopulmonary fitness in ambulatory adolescents and young adults with cerebral palsy. Arch Phys Med Rehabil 94:2227–2233 Song KM, Bjornson KF, Capello T, Coleman K (2006) Use of the StepWatch activity monitor for characterization of normal activity levels in children. J Pediatr Orthop 26:245–249 St Louis Childrens Hospital (2015) Fitbit fun: does Camp Independence increase activity of chil dren with cerebral palsy? Pediatr Perspect [Online], Fall Available: http://www.stlouischildrens. org/sites/default/files/health_professionals/images/PedPerspective_Fall2015web.pdf. Accessed 11 Dec 2016 Sullivan E, Barnes D, Linton JL, Calmes J, Damiano DL, Oeffinger D, Abel M, Bagley A, Gorton G, Nicholson D, Rogers S, Tylkowski C (2007) Relationship among functional outcome measures used for assessing children with ambulatory CP. Dev Med Child Neurol 49:338–344 Thomas SS, Buckon CE, Piatt JH, Aiona MD, Sussman MD (2004) A 2-year follow-up of outcomes following orthopedic surgery or selective dorsal rhizotomy in children with spastic diplegia. J Pediatr Orthop B 13:358–366 Thompson P, Beath T, Bell J, Jacobson G, Phair RSNM, Wright FV (2008) Test-retest reliability of the 10-metre fast walk test and 6 minute walk test in ambulatory school-aged children with cerebral palsy. Dev Med Child Neurol 50:370–376 Trost SG (2007) State of the art reviews: measurement of physical activity in children and adolescents. Am J Lifestyle Med 1:299–314 Trost SG, McIver KL, Pate RR (2005) Conducting accelerometer-based activity assessments in field-based research. Med Sci Sports Exerc 37:S531–S543 Trost SG, Way R, Okely AD (2006) Predictive validity of three Actigraph energy expenditure equations for children. Med Sci Sports Exerc 38:380–387 Trost SG, Fragala-Pinkham M, Lennon N, O’Neil ME (2016) Decision trees for detection of activity intensity in youth with cerebral palsy. Med Sci Sports Exerc 48:958–966 Tudor-Locke CE, Myers AM (2001) Methodological considerations for researchers and practitioners using pedometers to measure physical (ambulatory) activity. Res Q Exerc Sports 72:1–12 Tudor-Locke C, Williams JE, Reis JP, Pluto D (2002) Utility of pedometers for assessing physical activity: convergent validity. [Review] [49 refs] Sports Med 32:795–808 Tudor-Locke C, Craig C, Beets M, Belton S, Cardon G, Duncan S, Hatano Y, Lubans D, Olds T, Raustorp A, Rowe D, Spence J, Tanaka S, Blair S (2011a) How many steps/day are enough? For children and adolescents. Int J Behav Nutr Phys Act 8:78

1036

K.F. Bjornson and N. Lennon

Tudor-Locke C, Craig CL, Brown WJ, Clemes SA, de Cocker K, Giles-Corti B, Hatano Y, Inoue S, Matsudo SM, Mutrie N, Oppert J-M, Rowe DA, Schmidt MD, Schofield GM, Spence JC, Teixeira PJ, Tully MA, Blair SN (2011b) How many steps/day are enough? For adults. Int J Behav Nutr Phys Act 8:79–79 Tudor-Locke C, Brashear M, Katzmarzyk PT, Johnson WD (2012) Peak stepping cadence in freeliving adults: 2005–2006 NHANES. J Phys Act Health 9:1125–1129 Tyron WW (1991) Activity measurement in psychology and medicine. Plenum Press, New York U.S. Dept of Health and Human Services (2008) Physical activity guidelines for Americans summary [Online]. Available: http://www.health.gov/PAGuidelines/. Accessed 30 Nov 2016 Voorman JM, Dallmeijer AJ, Schuengel C, Knol DL, Lankhorst G, Becher JG (2006) Activities and participation of 9- to 13-year-old children with cerebral palsy. Clin Rehabil 20:937–948 Warms CA, Whitney JD, Belza B (2008) Measurement and description of physical activity in adult manual wheelchair users†. Disabil Health J 1:236–244 van Wely L, Becher J, Reinders-Messelink H, Lindeman E, Verschuren O, Verheijden J, Dallmeijer A (2010) LEARN 2 MOVE 7-12 years: a randomized controlled trial on the effects of a physical activity stimulation program in children with cerebral palsy. BMC Pediatr 10:77 van Wely L, Becher JG, Balemans ACJ, Dallmeijer AJ (2012) Ambulatory activity of children with cerebral palsy: which characteristics are important? Dev Med Child Neurol 54:436–442 van Wely L, Balemans ACJ, Becher JG, Dallmeijer AJ (2014) Physical activity stimulation program for children with cerebral palsy did not improve physical activity: a randomised trial. J Physiother 60:40–49 Wen LM, van der Ploeg HP, Kite J, Cashmore A, Rissel C (2010) A validation study of assessing physical activity and sedentary behavior in children aged 3 to 5 years. Pediatr Exerc Sci 2:408–420 Wieters KM, Kim J, Lee C (2012) Assessment of wearable global positioning system units for physical activity research. J Phys Act Health 9:913–923 Wilson NC, Signal N, Naude Y, Taylor D, Stott NS (2015) Gait deviation index correlates with daily step activity in children with cerebral palsy. Arch Phys Med Rehabil 96:1924–1927 Wilson NC, Mudge S, Stott NS (2016) Variability of total step activity in children with cerebral palsy: influence of definition of a day on participant retention within the study. BMC Res Notes 9:411 Wininger M, Bjornson K (2016) Filtering for productive activity changes outcomes in step-based monitoring among children. Physiol Meas 37(12):2231 Wood E, Rosenbaum P (2000) The gross motor function classification system for cerebral palsy: a study of reliability and stability over time. Dev Med Child Neurol 42:292–296 World Health Organization (2002) International classification of functioning, disability and health (ICF). World Health Organization, Geneva Wren TAL, Lening C, Rethlefsen SA, Kay RM (2013) Impact of gait analysis on correction of excessive hip internal rotation in ambulatory children with cerebral palsy: a randomized controlled trial. Dev Med Child Neurol 55:919–925 YAMAX-DIGIWALKER.COM (2016) Yamax-Digiwalker [Online]. Warminster. Available: https://www.yamax-digiwalker.com/

Spasticity Effect in Cerebral Palsy Gait Marlene Cristina Neves Rosa and André Gonçalo Gomes Roque

Abstract

A high number of children with cerebral palsy (CP) have spastic gait and consequently abnormalities in joint patterns. Several factors have been contributing to the lack of consensus on the spasticity effect in cerebral palsy gait and would be summarized and discussed in this chapter, e.g., spastic gait patterns are in constant evolution during the process of growth; there are still considerable limitations in the methodologies used to assess spasticity during gait; a wide range of rehabilitation strategies have been explored to control spasticity during gait. Spastic gait patterns are divided in hemiplegic (5 types) and diplegic (4 types) with the most prevalent joint abnormalities described in the sagittal plane. Ashworth, Tardieu, and DAROM scales and Pendulum Tests are widely used to assess spasticity but do not reliably explain the spasticity effects during gait. Orthotics, adequate exercise or handling techniques, botulinum injections, or surgical procedures have been used to manage spasticity effects. Keywords

Cerebral palsy • Gait • Spasticity

Contents Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1038 State of the Art . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1039 Spastic Gait Disorders . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1039

M.C.N. Rosa (*) Piaget Institute, Viseu, Portugal e-mail: [email protected] A.G.G. Roque (*) Physiotherapy, University of Averio, Aveiro, Portugal e-mail: [email protected] # Springer International Publishing AG, part of Springer Nature 2018 B. Müller, S.I. Wolf (eds.), Handbook of Human Motion, https://doi.org/10.1007/978-3-319-14418-4_55

1037

1038

M.C.N. Rosa and A.G.G. Roque

Methodologies to Assess Spasticity During Gait in Cerebral Palsy . . . . . . . . . . . . . . . . . . . . . . . . . . . Spasticity Treatments with Impact in CP Gait . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Future Directions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

1043 1046 1050 1050

Introduction Spasticity is one of the sequelae of neuromuscular disabilities, affecting 260–340 per 100,000 children with cerebral palsy (Dana and Cub 2013). Spasticity is the result of pyramidal tract lesions, but there are several theories to explain the neurophysiologic mechanisms related with this phenomenon (Dana and Cub 2013). The most commonly used definition of spasticity is that of Lance (1980) that explains it as “a velocity-dependent increase in tonic stretch reflexes (muscle tone) with exaggerated tendon jerks. . .” (Hobart et al. 1994). It depends on overexcitability of spinal alpha motor neurons, as a consequence of the interruption of descending modulatory carried by the corticospinal, vestibulospinal, and reticulospinal tracts (Filloux 1996). In specific, spasticity can be explained by the following pathophysiologic mechanisms: reduced reciprocal inhibition of antagonist motor neuron pools by Ia afferents, decreased presynaptic inhibition of Ia afferents, and decreased nonreciprocal inhibition by Ib terminals (Hobart et al. 1994). Despite this lack of consensus explaining the neurophysiologic mechanisms underlying spasticity, it is one of the most common disabling consequences in cerebral palsy, causing contractures, joint subluxations, and, indirectly, fatigue, loss of dexterity and coordination, and balance disorders (Dana and Cub 2013). As cerebral palsy (CP) is a life-lasting condition, spasticity also affects the growth of the individual, hampering both muscle and skeletal development and resulting in symmetric or asymmetric biomechanical deficiencies depending on the PC type (Bar-On et al. 2015). Also, one has to consider that CP originates in the infancy or childhood, and the consequences of spasticity have to be separated from that resulting from adult lesions. In the second case, motor control has been normally developed, as opposed by the first case in which spasticity-related gait alterations will direct the motor learning lifelong. These consequences have serious impact in patients´ functionality, specifically in gait. Gait disorders due to spasticity can assume different patterns, depending on the muscles affected, which may cause different effects in joint kinematics (Miller 2004). The first section (“Spastic Gait Disorders”) of this chapter will summarize the types of spastic gait disorders and its characteristics, according to the muscles affected. In addition, efficient methodologies to assess spasticity are crucial to understand its effect in gait pattern (Scholtes et al. 2006). Most of the instruments described in the literature are not able to assess spasticity under rhythmic motor tasks such as gait (Scholtes et al. 2006). Alternative methodologies (e.g., electromyography) have critical aspects that would be discussed in this chapter as they are not aligned with the original definition of spasticity. Section two (“Methodologies to Assess Spasticity During Gait in Cerebral Palsy”) will summarize weaknesses and strengths of

Spasticity Effect in Cerebral Palsy Gait

1039

different scales and instruments that have been used to assess spasticity during gait or the related phenomenon. Section three (“Spasticity Treatments with Impact in CP Gait”) of this chapter will give a simple perspective of the widely accepted approaches to manage spasticity and its effects on CP gait, not focusing on specific results or merits of each.

State of the Art The knowledge of CP spastic gait patterns is well established and despite some children may present with mixed patterns, often they walk in a typical bilateral equinus. Also spasticity contribution to the different gait patterns is defined in the literature; however, the specific association between spasticity and different components and phases of gait is only recently subjected of throughout investigation and will not therefore be exploited in this chapter. This fact is related to the newly available research methods that allows to study spasticity during gait but also to the capacity to establish associations between clinical and research measures of CP gait and spasticity. The methods and therapeutics that address spasticity and CP gait are numerous; some of them are well established for a long time with different degrees of success (orthopedic surgery, stretching, botulinum toxin A, orthotics), while some are more recent yet promising (selective dorsal rhizotomy, intrathecal baclofen, virtual reality, and transcranial magnetic stimulation). The choice between different approaches is a truly clinical decision, which should be based on CP gait analysis, the available evidence, and the individual presented to evaluation. The only conclusion that our readers should obtain toward the treatment theme is that a throughout evaluation of the characteristics of CP gait is fundamental to the success of each of the selected interventions. As such, it is not our objective in this chapter to demonstrate the results of the different approaches but to indicate the existing managing possibilities and, when they were studied, the expected gait outcomes of those interventions and possible implications for the future of the adolescent and adult with CP.

Spastic Gait Disorders Cerebral palsy (CP) is traditionally classified by the clinical type (motor dysfunctional) and topography (anatomical region of the lesion). The motor dysfunctional patterns are classified as (i) spastic, which is the most common pattern, and (ii) ataxic, hypotonic, dyskinetic, and mixed (Morais Filho et al. 2014). Approximately two thirds of all patients with CP suffer from spasticity (Awaad and Rizk 2012). The most affected muscles are the antigravity muscles. Commonly, and as a result of this spasticity, legs are in extended and adducted position and arms are in flexed, internal rotation and pronated position (Awaad and Rizk 2012). In gait, several deviations have been reported as a consequence of spasticity in hemiplegic and diplegic patients, most of them based on joint deviations in the sagittal plane (O’Byrne et al. 1998). Ultimately, these deviations reduce walking

1040

M.C.N. Rosa and A.G.G. Roque

speed and increase energy expenditure, which cause functional inability (Piccinini et al. 2007; Novacheck et al. 2000). Spastic motor deviations are consistent from stride to stride and day to day (Rodda and Graham 2001). Therefore, detectable changes are generally seen as a result of an intervention or as a change related with the aging process and therefore should be monitored during the rehabilitation process (Rodda and Graham 2001). There are at least four different gait patterns in hemiplegic patients with CP all characterized by more distal involvement, which means that true equinus is the basis of the most common gait hemiplegic patterns. In hemiplegia type 1, the most prominent characteristic is drop foot during swing phase caused by inability to control dorsiflexors. Two subtypes characterize hemiplegia type 2. The subtype 2.A includes equinus, neutral knee, and extended hip, while subtype 2.B is characterized by equinus, recurvatum knee, and extended hip. This is the most common type found in clinical practice and it is observed in the stance phase (spasticity in gastrocnemius). Drop foot is also noted during the swing phase because tibialis anterior is impaired. Associated with the predominance of ankle plantarflexion, the knee may adopt the position of recurvatum or extension. The main characteristics of the hemiplegic gait type 3 is the impaired ankle flexion in swing due to gastrocnemius spasticity or contracture and the “stiff knee gait” because of co-contraction between hamstrings and quadriceps. Finally, the hemiplegia type 4 is similar to the pattern of spastic diplegia (but unilateral, therefore asymmetric) with increasing proximal involvement. The characteristics in the sagittal plane are flexed knee, flexed hip, and anterior pelvic tilt. Two other planes present abnormalities in lower limb posture. In the coronal plane, there is hip adduction and in the transverse plane, there is internal rotation. Table 1 summarizes gait hemiplegic patterns in people with cerebral palsy (Rodda and Graham 2001). In contrast with hemiplegic gait, diplegic gait is characterized by more proximal involvement, which means that apparent equinus and crouch gait are very common (Rodda et al. 2004). The five different diplegic gait patterns are schematically described in Table 2. In the up-down direction across the table, there is a decreasing equinus, an increasing proximal involvement, and a change in the direction of the ground reaction force, i.e., from the front of the knee to behind it. The key muscles in these patterns are hip and knee flexors and ankle plantarflexors. These deformities in sagittal gait plane are probably responsible for others found in transverse plane such as torsional deformities of the femur and tibia. The true equinus hemiplegic gait pattern (type 1) is one of the most incident patterns and it is characterized by an equinus and a fully extended knee in late stance. About the jump gait (type II), it is characterized by hip and knee flexion during all stance phase, and then, in late stance, the main characteristic is the equinus. These deformities give the appearance that the individual is jumping up and down, which explains the identification of this gait pattern. In apparent equinus (type III), there are knee and hip flexion contractures that explain why the heel is not in contact with the ground, despite it is not a true equinus foot. In the crouch gait (type IV), three simultaneous conditions are confirmed: hip and knee flexed and ankle kinematics in calcaneus range. Finally, each lower limb presents a different kinematics pattern in the asymmetrical gait

Spasticity Effect in Cerebral Palsy Gait

1041

Table 1 Characteristics of gait patterns in spastic cerebral palsy hemiplegia (Rodda and Graham 2001) Spastic hemiplegia Type I Drop foot

Joint deviation Drop foot

Type II 2.a True equinus 2.b. True equinus/ recurvatum knee

Equinus + neutral knee + extended hip Equinus + recurvatum knee + extended hip Ankle dorsiflexion impaired Flexed stiff knee

Type III True equines/ jump knee

Type IV Equinus/ jump knee Pelvis rotation/hip flexion/ adducted/ internal rotation

Equinus + flexed stiff knee + flexed hip + anterior pelvic tilt Hip adduction Hip internal rotation

Characteristics Plane of Problematic movement gait phase Sagittal SW

Origin of the problem Loss of selective control in dorsiflexors No calf contractures Spasticity/ contracture of gastrocnemius/ soleus muscles

Sagittal

ST

Sagittal

SW

Spasticity/ contracture of gastrocnemius/ soleus muscles Hamstring/ quadricep co-contraction

Sagittal Coronal Transverse

ST SW

Spasticity/ contracture of gastrocnemius/ soleus muscles Hamstring/ quadricep co-contraction Spasticity/ contracture of hip adductors and hip flexors

Management Ankle foot orthosis

Botulinum toxin type A (spasticity) Orthotic support (mild contracture) Tendo Achilles + calf lengthening (fixed contractures) Botulinum toxin type A (spasticity) Tendo Achilles + hamstring lengthening (contractures) Orthotic support (plantar flexion + knee extension couple) Botulinum toxin type A (spasticity) Tendo Achilles + hamstrings + hip adductors + iliopsoas lengthening (contractures) Orthotic support (plantar flexion + knee extension couple)

(type V), i.e., one lower limb is usually classified as belonging to group III and the other is considered as belonging to group II (Rodda et al. 2004). Joint problems may be primary or secondary, with the related problem in the joint or elsewhere. Secondary joint adjustments can be spontaneously reversed with the solving of the primary affection. However, if left untreated, it can become a primary problem that progresses into adulthood. Usually asymmetrical gait has one lower limb toe walking due to spasticity; however, if the child is strong enough, the tendency is for gait symmetry. Thus, the contralateral lower limb will also toe

1042

M.C.N. Rosa and A.G.G. Roque

Table 2 Characteristics of gait patterns in spastic cerebral palsy diplegia (Rodda et al. 2004) Spastic diplegia Type I True equinus Type II Jump gait

Type III Apparent equinus Type IV Crouch gait

Type V Asymmetric gait

Joint deviation Anterior/normal pelvic tilt + recurvatum/normal knee position + drop foot Anterior/normal pelvic tilt + knee and hip from excessively flexed to extended Anterior/normal pelvic tilt + knee and hip flexed Posterior/normal pelvic tilt + knee and hip flexed + foot in excessive dorsiflexion One lower limb – pattern III The other – pattern II

Characteristics Plane of Problematic movement gait phase Sagittal Stance phase

Origin of the problem Contracture of the hamstrings Contracture of calf muscles Flexion contractures of knee and hip

Sagittal

Stance phase

Sagittal

Stance phase

Hamstrings and psoas contraction

Sagittal

Stance phase

Hamstrings and psoas contraction

Sagittal

Stance phase

Lower limb (pattern II) – Flexion contractures of knee and hip Lower limb (pattern III) – Hamstrings and psoas contraction

walk and develop a fixed contracture that will need surgical addressing, becoming itself a primary joint problem (Miller 2004). During adolescence there is a rapid decrease in the strength ratio (Miller 2004). The association of this factor with a spasticity growth impairs normal muscle development, leaving the young adults even weaker. In adults, muscle contractures due to spasticity develop much more rapidly than in childhood and results in severe amplitude limitations. The following fixed contractures are not associated with joint structural alterations, but they reduce the variability of movements available to motor control, promoting its decrease to a more gross pattern (Miller 2004). Nevertheless, joint deformities occur in adults with CP independently of the ambulatory status, diagnosis, or severity of involvement (Kembhavi et al. 2011). The combined impairments contribute to a sense of frustration in the adolescents, who lose the motivation to walk (Miller 2004). Adolescents with CP are usually heavier, and as such they are more prone to fall and be hurt during the adolescent clumsy stage (a normal feature of gait development). These falls may cognitively repress the gait behavior in this population, worsening the functional prognosis. However, it is difficult to assess if deterioration of mobility is due to the pathophysiology associated with CP, or the physical effects associated with aging, or an interaction of both (Kembhavi et al. 2011).

Spasticity Effect in Cerebral Palsy Gait

1043

Height and weight growth contributes to the development of the crouch gait. The muscles and joints become unable to support the body in the child typical toe-walking pattern, and knee flexion, hip flexion, and dorsiflexion increases, which can cause a midfoot and hindfoot collapse and a consequential severe planovalgus (Miller 2004). Young adults may also develop a back-kneeing pattern, particularly those using walking aids, with severe weakness of the gastrocnemius and often those who had tendon Achilles transections (Miller 2004).

Methodologies to Assess Spasticity During Gait in Cerebral Palsy Clinical evaluation of spasticity and its contribution to gait alterations in PC was traditionally a subjective analysis. Manual testing of spasticity uses non standardized velocities, which are frequently under the threshold for eliciting the stretch reflex, and is dependent on a relaxed state of the individual. As such it cannot differentiate between the contributions of neural and nonneural factors to spasticity. This is of relevance in the selection of treatment options, with medication used primarily in the first situation and casts or orthotics in the second. Also selective dorsal rhizotomy produces results only in a neural predominant spasticity (Bar-On et al. 2015). Different subjective scales have been used for clinical tone assessment, such as (i) the Ashworth Scale (AS)/Modified Ashworth scales (MAS) (Table X), (ii) the Tardieu Scale (TS) and the Modified Tardieu Scale (MTS) (Table Y), (iii) the Dynamic Evaluation Range of Motion (DAROM), and (iv) the Pendulum Test. However, as stated before, the reliability of these instruments have been criticized because the results obtained are dependent on some differences across inter-raters or day-testing conditions, e.g., the range of motion (ROM), the velocity of movement, and the position of the tested muscle (Boyd and Graham 1999). In specific, the Tardieu Scale and the Dynamic Evaluation of Range of Movement (DAROM) have reported more specifications that improved reliability in assessing spasticity, as, for example, the ROM is defined as slow or fast passive stretching. Moreover, the DAROM identifies a “range of motion deficit,” which means a value from the minimal muscle stretch position (Pandyan et al. 1999). When using the Ashworth Scales, an assessor tests the resistance to passive movement about a joint. Then, the resistance perceived while moving a joint through its full range of movement – except in grade “4” (Table 3) – is graded in a 5-Likert scale. The Ashworth Scale and the Modified Ashworth Scale can be used as measures of resistance to passive movement, but not as an ordinal level measure of spasticity (Pandyan et al. 1999; Damiano et al. 2002). The Tardieu Scale (Table 4) rates the spasticity as the difference between the reactions to stretch at two extreme velocities (Gracies et al. 2005): – The slowest: below the threshold of any significant stretch reflex (output = passive range of motion). – The fastest: maximizes the involvement of the stretch reflex, and if any spasticity is present, then the rater feels the sensations of catch and release/clonus/fatigable.

1044

M.C.N. Rosa and A.G.G. Roque

Table 3 The Ashworth and modified Ashworth scales Score 0 1

Ashworth scale No increase in tone Slight increase in tone giving a catch when the limb was moved in flexion or extension

1+

Slight increase in tone giving a catch when the limb was moved in flexion or extension

2

More marked increase in tone but limb easily flexed

3

Considerable increase in tone, passive movement difficult Limb rigid in flexion or extension

4

Modified Ashworth scale No increase in muscle tone Slight increase in muscle tone, manifested by a catch and release or by minimal resistance at the end of the range of motion when the affected part(s) is moved in flexion or extension Slight increase in muscle tone, manifested by a catch, followed by minimal resistance throughout the remainder (less than half) of the range of movement (ROM) More marked increase in muscle tone through most of the ROM, but affected part(s) easily moved Considerable increase in muscle tone, passive movement difficult Affected part(s) rigid in flexion or extension

Table 4 The Tardieu Scale principles and grading system Principles 1. Grading always performed: Muscle at rest before the stretch maneuver Reproducible velocity of stretch: once the fast velocity is selected for a muscle, it remains always the same At the same time of the day 2. Velocity of stretch Slow: V1, as slow as possible/slower than the rate of natural drop of the limb segment under gravity Fast: either V2 and V3; V2 is equivalent to the speed of the limb segment falling under gravity; V3 is equivalent to faster than the rate of natural drop of the limb segment under gravity

Grading X = spasticity angle (threshold) (Angle of arrest at slow speed (V1) – angle of catch at fast speed (V3)) Y = spasticity grade 0 = no resistance throughout passive movement 1 = slight resistance throughout passive movement 2 = clear catch at precise angle, interrupting passive movement, followed by release 3 = fatigable clonus (10s when maintaining pressure occurring at a precise angle) Catch without release Catch with minimal release Angle 0 For grades 0 and 1, spasticity angle X = 0 by definition

Comparing the characteristics of Ashworth and Tardieu Scales, it can be concluded that spasticity measured by the AS is confounded by muscle contracture, while the TS is able to differentiate spasticity from contracture and therefore is a more valid tool (Emily Patrick and Ada 2006). The Dynamic Evaluation of Range of Motion (DAROM) and the Pendulum Test complete the most widely used instruments for assessing spasticity in children with CP. The DAROM relies on the same velocity principle of TS, as it considers at least two different velocities of passive muscle stretching. The DAROM identifies a

Spasticity Effect in Cerebral Palsy Gait

1045

Table 5 Dynamic Evaluation of Range of Motion (DAROM) and pendulum tests outcomes Measurement Range of motion deficit (DROM) for DAROM tests: T1, T2, T3, T4

Parameter DROM I DROM II ASO

Pendulum Test

Ex.

t n

Description Range of motion deficit following a slow velocity stretch (V1) – expressed in degrees Range of motion deficit after a fast velocity stretch (V3) – expressed in degrees Value calculated as the difference between the DROM II and DROM I – expressed in degrees First swing excursion – difference between the starting angle and the first angle of reversal of the swinging limb; expressed in degrees Duration of the pendulum swings (sec.) Counting the maxima of the sinusoidal waves produced by the swinging limb after the heel was released

“range of motion deficit” (DROM), defined as a value from the minimal muscle stretch position. Using this instrument, two joint angles are measured: DROM I, defined as the range of motion deficit following a slow velocity stretch, and DROM II, defined as the angle of catch after a fast velocity stretch. The difference between DROM II and DROM I indicates the examined muscle group’s level of contracture and is called the angle of spasticity (ASO) (Bax et al. 2005). Otherwise, the Pendulum Test was first described by Wartenberg and it consists of a biomechanical method that evaluates muscle tone (discriminating between various degrees of spasticity) using gravity to induce the muscle stretch reflex during passive swinging of the lower leg. The following table (Table 5) shows a summary of the outcome measures that can be collected using the DAROM and Pendulum Tests (Domagalska et al. 2013). When considering how to treat spasticity, in addition to documenting changes in the resistance to passive movement (Pandyan et al. 1999), it may also be relevant to document changes in function. To further understand the role of spasticity in functional status of patients with CP, the literature has been focused in exploring possible relationships between spasticity and other gait parameters. Considering this problematic, there are still relevant controversies that may affect the quality of spasticity management in gait. For example, while Damiano and Abel (Abel et al. 2003) reported a significant correlation between the knee extensor Ashworth score and gait impairments, Gage (2004) and Boyd and Graham (1999) confirmed that rectus femoris spasticity is a possible cause of stiff knee gait; however, Wren et al. (2007) questioned the correlations between gait deviations and knee flexors spasticity and contractures. Moreover, spasticity in the knee (both in quadriceps and hamstrings) and hip muscles (adductors) (Ross and Engsberg 2007) has also been related with gait velocity and stride length decreasing (Damiano et al. 2006). This finding corroborates the clinical importance of spasticity in general functioning, as gait velocity is a major determinant of patients’ integration in community, i.e., someone that walks faster can easily and safety cross a street and therefore would not avoid community environments (Middleton et al. 2015). Spasticity and co-contraction have been critically assessed as being the same phenomenon (Diane et al. 2000). Despite the co-contraction is most of the times

1046

M.C.N. Rosa and A.G.G. Roque

related with spasticity and therefore simultaneously present in CNS disorders, these are not exactly the same phenomenon. In fact, co-contraction becomes excessive, and the agonist force decreases because spasticity transforms, at long term, the peripheral muscle components (configuration and properties of muscle fibers) and, consequently, generates abnormalities in dynamic muscle activation patterns. The application of EMG has been tested as a novel approach to quantitative study of muscle activity during dynamic motor activities. For example, during gait, the CP children stretching of calf muscles (agonist) in the post-contact stride period is known to be frequently accompanied by synchronous bursts in triceps surae (antagonist), generating abnormal muscle co-contraction patterns. This abnormal muscle patterns tend to be congruent with mechanical changes in ankle joint (e.g., absence of dorsiflexion). Even though EMG is able to detect these abnormal muscle patterns, it cannot reliably separate two possible causes for this phenomenon: the augmented myotatic reflex effect or the peripheral component contributions (muscle rigidity) (Crenna 1998). Instrumented gait analysis (IGA) has the unique ability to measure and provide information critical to the distinction between complex gait patterns. By assessing the tridimensional position of the different segments, including pelvis, at any moment of the gait cycle, it facilitates pattern recognition and problem addressing (Miller 2004). The integration of video analysis with dynamic electromyography (D-EMG) allows the identification of the contribution of muscles in each phase of the gait cycle, enhancing understanding of the pathological gait patterns (Miller 2004). Also, it makes it possible to assess the tridimensional gait results of the different proposed interventions (Miller 2004). By also incorporating clinical measures, IGA is the gold standard in gait evaluation in the CP population, with none other video analysis protocol reaching its consistency (Rathinam et al. 2014). Therefore, the simultaneous analysis of IGA and D-EMG during gait does not provide a direct information on the spasticity effect in locomotion patterns in children with cerebral palsy but can reliably assess consequences of spasticity in segment postures and spatiotemporal parameters. As a conclusion, the original definition of spasticity (Lance and McLeod 1981) “does not include impaired movement and an abnormal posture” and only considered observations under static conditions (muscles relaxed; muscles tonically activated), which explains the poor report of correlations between the pathological responsiveness to stretch measured at rest and motor deficits assessed during natural actions, such as gait (Crenna 1998). Therefore, the lack of adequate instruments to assess spasticity during gait has been limiting the consensus about the effect of spasticity in gait pattern.

Spasticity Treatments with Impact in CP Gait For some years now, a spinal cord mechanism that controls the excitability of the stretch reflex has been described, which does not depend on supraspinal inhibition. The so-called postactivation depression is an intrinsic property of the Ia afferent fibers that is associated with a decreased release of neurotransmitters and has been

Spasticity Effect in Cerebral Palsy Gait

1047

shown to be reduced with limb immobilization in a shortened position. As such, muscle atrophy and reduced activity seen in CP also results in spasticity. More importantly this mechanism seems to be affected by active movement, passive mobilization, and prolonged muscle stretching, all of which can prevent muscle hypertonia (Trompetto et al. 2014). This could be one mechanism subjacent to improvements seen with both splits/orthotics and physiotherapy movement modalities. Bracing and orthotics are classic treatment options for lower limb muscle and tendon lengthening in CP. Orthotics applied in the ankle, holding the foot in a plantigrade position and correcting planovalgus foot deformities, may provide the stable fundamental base that allow CP children to develop balance and train taskspecific activities in stance but also give a movement focus in the knee and hip. This favors a more symmetric and “normal” gait pattern. However, by restraining active plantar flexion, ankle-foot orthotics (AFO) reduce the lift-off moment of gait, producing secondary adaptations, such as the hip extensors being the main responsible for the push-off phase and forward movement, which is frequently combined with pelvis rotation. A flexible leaf-spring AFO may allow some degree of plantar flexion but only maintains the foot in dorsiflexion if there is absent spasticity of the gastrocnemius and soleus. Baclofen is a gamma-aminobutyric acid (GABA) agonist and thus mimics an inhibitory neurotransmitter, reducing spasticity. It has been used both orally and, more recently, intrathecally. The latter reduces some side effects such as excessive sedation and also increases the bioavailability of the drug in its action sites: the cerebrospinal fluid and brain. Its evidence is insufficient in regard to decreasing spasticity, improving gross motor actions, and transferring achievements to activities and participation (Martin et al. 2010). Despite the lack of evidence, baclofen is a treatment option for generalized spasticity, such as the one presented in tetraparetic CP. Other alternatives to the management of generalized spasticity are pharmacological, including diazepam and tizanidine with moderate evidence and dantrolene with insufficient evidence. However, the pharmacological options have not demonstrated changes in gross motor function or quality of life (Martin et al. 2010). Botulinum toxin A (BTA) is a neurotoxin that inhibits acetylcholine release into the neuromuscular junction, reducing the muscle response to efferent neural stimuli. As such, the tonic and phasic stretch reflexes and consequently spasticity are diminished in the selected injected muscles or muscular groups. However, the duration of its effect is estimated in 3–6 months and the action potential of the muscle and active movements are reduced, which means that BTA injections by themselves are not a treatment for spasticity-derived gait alterations but can provide the means to facilitate a more biomechanical correct movement and posture during gait (Klemens Fheodoroff et al. 2016). This facilitation is achieved by providing the correct afferent exteroceptive and proprioceptive stimulus in an intensive physiotherapy program or the use of orthotics. Its direct impact is therefore in the structural and functional dimensions of the International Classification of Functioning, Disability and Health (ICF), with only a latter translation to activities and participation,

1048

M.C.N. Rosa and A.G.G. Roque

due to the time required to motor relearning of the treated limb to occur (Love et al. 2010). Precocious use of BTA in young children with CP may reduce the evolution to a crouch gait and the need for surgical corrective procedures but also increase the probability of the child to achieve the motor milestones crucial to a normal neurodevelopment (Klemens Fheodoroff et al. 2016). Furthermore it is considered a safe procedure in CP children with localized spasticity (Delgado et al. 2010). An international consensus has stated in 2010 that BTA is effective in the management of spastic equines to improve gait (level A), however, only probably effective to improve goal attainment and function in the management of spastic equinus (level B). Also, adductor injections may help attain some goals (level B) and may delay hip displacement (level A) but does not improve gross motor function and does not affect long-term outcome (level A). Multiple lower limb injections have conflicting information as respect to gait, goal attainment, and function (level U) (Love et al. 2010). Another treatment option for spasticity-related gait disorders in PC is selective dorsal rhizotomy (SDR). By selectively impairing afferent stimuli to the medullae, this procedure inhibits the efferent response of the stretch reflex, resulting in decreased tone and therefore reducing spasticity. Since the clinical effects are somewhat similar to BTA or baclofen, it is expected to act merely in the structural and functional dimensions of ICF, with the considerations made to BTA regarding transfer to activities and participation, being also valid. However, the benefits observed in the sagittal plane at all lower limb joints with SDR are associated with an augmented misalignment in the pelvis. These postural alterations include crouching and horizontalization of the sacrum (Roberts et al. 2015), which will progress to the adult age due to the irreversibility of the technique. Furthermore, the procedure may reveal underlying muscle weakness that worsens the performance of previous ambulatory children, especially if the antigravity postural capacity is affected. Thus, a multidimensional evaluation and biomechanical analysis of gait should be performed to guide patients’ election to this procedure (Grunt et al. 2014). Indeed, spasticity of the extensor apparatus of the lower limb may be a useful characteristic of CP gait, as it maintains the possibility of placing the foot (even if in equinus) on the ground and thus relieve the body mass loading in the contralateral limb. This is a fundamental characteristic of a normal gait. When spasticity is removed, for instance, by the abovementioned dorsal rhizotomy, this extension maintenance is impaired, resulting in incapacity of body mass transfer to the affected limb. The correct consideration of the contribution of spasticity to CP gait is crucial to the success or failure of a spasticity management procedure, when it relates to activity, participation, and transfer to the child environment (Grunt et al. 2014). Despite the absent role in the management of tone abnormalities of physiotherapy, it is a fundamental adjunctive therapy to increase muscle strength and aerobic conditioning but also in improving lower extremity coordination and speed (Martin et al. 2010). Excessive exercise and strength training was said to aggravate spasticity in CP and therefore a cause to gait deterioration; however, studies have challenged this traditional belief (DL 2006). As such intensive task-oriented practice is nowadays a central practice in CP rehabilitation, which in the case of gait precludes

Spasticity Effect in Cerebral Palsy Gait

1049

intensive treadmill training with or without assistive weight supports, despite a lack of demonstrated evidence (Martin et al. 2010). Other physiotherapy modalities that address spasticity and gait have long been used in clinical practice. Neurodevelopmental therapy (NDT) and functional training (FT) evaluate the individual as a whole, directing its practice according to the identified incapacities, limitations, or handicaps (Martin et al. 2010). By focusing in the particular characteristics of the different presentations of gait in CP instead of intervening in a cluster of typical patterns, these modalities have struggled to produce conclusive evidence in investigational studies designed as randomized control trials (RCT) but have nevertheless a role in neurorehabilitation of the consequences of spasticity in CP gait, either individually or as multimodal approaches. Virtual reality (VR) is an emerging computer-assisted multisensorial feedback approach to neurorehabilitation that promotes neuroplasticity induced by repetition in an enjoyable environment but depends on the capacity of the individual to integrate the “scene.” By conferring the possibility of manipulation of the environment, VR facilitates motor learning which in the case of the present chapter is directed to improvements in the spatiotemporal features of CP gait (E. Monge Pereiraa et al. 2014). Transcranial magnetic stimulation (TMS) has also recently been therapeutically used, and it is characterized by noninvasive stimulation of the motor cortex, promoting supraspinal inhibition of spasticity and therefore reducing CP spasticity-induced gait changes (Gunduz et al. 2014). When not efficiently managed, spasticity can progress to the complications previously mentioned, and orthopedic surgery has a pivotal role in addressing those consequences. As deformity evolves from dynamic to fixed, the focal tone management modalities already referred become contraindicated, as bone deformity or soft tissue contractures are not tone dependent and therefore susceptible. A wide range of orthopedic interventions are available in CP management, some of which address soft tissues such as tendon lengthening and ligament or capsular releases, while others intervene in the skeletal system with osteotomies and arthrodeses as the mainstay of the possibilities. After a biomechanical evaluation, there is also the possibility of combining surgical procedures to different anatomical locations, in order to optimize gait, being referred to as single-event multilevel surgery (SEMS) (Thomason et al. 2013). We previously described the pattern of flexed knee gait, characterized by the incapacity to perform full hip and knee extension in late stance. The surgical lengthening of the biarticular gastrocnemius, hamstrings, and iliopsoas is one component of surgical procedures targeting improved stance and gait in spastic diplegia. Indeed, a surgical procedure that lengthens the gastrocnemius may overcome the hip secondary compensation abovementioned, as it allows some active plantar flexion and also reduces the contracture in midstance, resulting in global gait improvements. Also hamstring lengthening is effective in correcting knee extension; however, since the hamstrings contribute to hip extension in stance, the procedure affects to some degree pelvic control. Other surgical options have been developed, such as the supracondylar extension osteotomy to correct the flexed knee deformity combined with patellar tendon advancement, which is described to improve knee extensor lag,

1050

M.C.N. Rosa and A.G.G. Roque

knee extension in gait, and overall gait. Due to increased stability in the knee, these procedures are also associated with reduced spasticity in the knee extensors (Sossai et al. 2010). Surgical selection should thus be decided based in gait analysis of the CP individual potentiating the after gains. From the treatments referred above, orthopedic surgery (including tendon transfer, muscle lengthening, or others), orthotics, and spasticity control (including rhizotomy, BTA, phenol blocks, or others) have succeeded in improving gait velocity, decreasing cadence, and increasing stride length (Paul et al. 2007). Other treatments such as serial casting, muscle strengthening exercises, and biofeedback have not accumulated the necessary evidence to affirm their effects in improving CP spatiotemporal components of gait. A multifactorial approach to gait alterations should thus be considered in detrimental of more isolated approaches.

Future Directions There is a scarce of consolidated information on the evolution of the gait pattern throughout adolescent and adulthood, specially from CP individuals who are subjected to orthopedic and invasive tone management procedures (Wilson et al. 2014). This is an issue that must be address in future studies, as the best medical, developmental, and rehabilitative science for optimizing mobility, motor functioning, and fitness for adults with CP are unknown (Frisch and Msall 2013). Also, it is not clear for now what the activity limitations that impair adults with CP from participating in their community are and also which environmental factors and in what measure they affect CP adults’ handicaps, despite some works investigating this theme (Frisch and Msall 2013). Given the difficulty to distinguish between spasticity and intrinsic hypertonia with the use of current evaluation measures, question remains if the actual treatments for lower limb hypertonia really address spasticity or are actually just influencing the biomechanical factors responsible for the intrinsic hypertonia as a result of immobilization. This is an important issue when deciding for intervention use, orthopedic or tone management. In other words where and when are the limits to tone management potential? When does rigidity become a primary issue, contraindicating spasticity control? The use of biomechanical measures with EMG recordings and perhaps accelerometers should be included in future clinical trials of CP gait evaluation and spasticity management, in order to differentiate between the focus of the treatment and provide a more comprehensive knowledge of the individual alterations (Chang et al. 2010), and on where in the gait cycle the velocity threshold to elicit the tonic or phasic stretch reflex occurs.

References Abel MF, Damiano DL, Blanco JS, Conaway M, Miller F, Dabney K, Sutherland D, Chambers H, Dias L, Sarwark J, Killian J, Doyle S, Root L, LaPlaza J, Widmann R, Snyder B (2003) Relationships among musculoskeletal impairments and functional health status in ambulatory cerebral palsy. J Pediatr Orthop 23(4):535–541

Spasticity Effect in Cerebral Palsy Gait

1051

Awaad Y, Rizk T (2012) Spasticity in children. J Taibah Univ Med Sci 7(2):53–60 Bar-On L, Molenaers G, Aertbelien E, Van Campenhout A, Feys H, Nuttin B, Desloovere K (2015) Spasticity and its contribution to hypertonia in cerebral palsy. BioMed Res Int 2015:1–10 Bax M, Goldstein M, Rosenbaum P, Leviton A, Paneth N, Dan B, Jacobsson B, Damiano D (2005) Proposed definition and classification of cerebral palsy. Dev Med Child Neurol 47:571–576 Boyd R, Graham HK (1999) Objective measurement of clinical findings in the use of botulinum toxin type A for the management of children with CP. Eur J Neurol 6:23–35 Chang FM, Rhodes JT, Flynn KM, Carollo JJ (2010) The role of gait analysis in treating gait abnormalities in cerebral palsy. Orthop Clin N Am 41(4):489–506 Crenna P (1998) Spasticity and ‘Spastic’ gait in children with cerebral palsy. Neurosci Biobehav Rev 22(4):571–578 Damiano DL (2006) Activity, activity, activity: rethinking our physical therapy approach to cerebral palsy. Phys Ther 86(11):1534–1540 Damiano DL, Quinlivan JM, Owen BF, Payne P, Nelson KC, Abel MF (2002) What does the Ashworth scale really measure and are instrumented measures more valid and precise? Dev Med Child Neurol 44(2):112–118 Damiano DL, Laws E, Carmines DV, Abel MF (2006) Relationship of spasticity to knee angular velocity and motion during gait in cerebral palsy. Gait Posture 23(1):1–8 Dana R, Cub DA (2013) Cerebral palsy gait, clinical importance. MAEDICA – J Clin Med 8(4):388–393 de Morais Filho MC, Kawamura CM, Lopes JAF, Neves DL, Cardoso M d O, Caiafa JB (2014) Most frequent gait patterns in diplegic spastic cerebral palsy. Acta Ortop Bras 22:197–201 Delgado MR, Hirtz D, Aisen M, Ashwal S, Fehlings DL, McLaughlin J, Morrison LA, Shrader MW, Tilton A, Vargus-Adams J (2010) Practice parameter: pharmacologic treatment of spasticity in children and adolescents with cerebral palsy (an evidence-based review): report of the quality standards subcommittee of the American academy of neurology and the practice committee of the child neurology society. Neurology 74(4):336–343 Diane L, Martellotta L, Sullivan J, Granata P, et al. (2000) Muscle force production and functional performance in spastic cerebral palsy: Relationship of cocontraction. Archives of Physical Medicine and Rehabilitation 81(7):895–900 Domagalska M, Szopa A, Syczewska M, Pietraszek S, Kidoń Z, Onik G (2013) The relationship between clinical measurements and gait analysis data in children with cerebral palsy. Gait Posture 38(4):1038–1043 Fheodoroff K, Jacinto J, Geurts A, Molteni F, Franco JH, Santiago T, Rosale R, Gracies J-M (2016) How can we improve current practice in spastic paresis? Eur Neurol Rev 11(2) (Epub ahead of print) Filloux FM (1996) Neuropathophysiology of movement disorders in cerebral palsy. J Child Neurol 11(Suppl 1):S5–S12 Frisch D, Msall ME (2013) Health, functioning, and participation of adolescents and adults with cerebral palsy: a review of outcomes research. Dev Disabil Res Rev 18(1):84–94 Gage JR (2004) The treatment of gait problems in cerebral palsy. Mac Keith, London Gracies J-M, Burke K, Clegg NJ, Browne R, Rushing C, Fehlings D, Matthews D, Tilton A, Delgado MR (2005) Reliability of the Tardieu Scale for assessing spasticity in children with cerebral palsy. Arch Phys Med Rehabil 91(3):421–428 Grunt S, Fieggen AG, Vermeulen RJ, Becher JG, Langerak NG (2014) Selection criteria for selective dorsal rhizotomy in children with spastic cerebral palsy: a systematic review of the literature. Dev Med Child Neurol 56(4):302–312 Gunduz A, Kumru H, Pascual-Leone A (2014) Outcomes in spasticity after repetitive transcranial magnetic and transcranial direct current stimulations. Neural Regen Res 9(7):712–718 Hobart JC, Riazi A, Thompson AJ (1994) Spasticity: a review. Neurology 44(11):S12–S20 Kembhavi G, Johanna D, Kelsey P, Devon P (2011) Adults with a diagnosis of cerebral palsy: a mapping review of long-term - Review Development medicine and child neurology 53(7):610–614 Lance JW (1980) Symposium synopsis. In: Feldman RG, Young RR, Koella WP (eds) Spasticity: disordered motor control. Chicago, Yearbook Medical Publishers

1052

M.C.N. Rosa and A.G.G. Roque

Lance JW, McLeod JG (1981) A physiological approach to clinical neurology. Butterworths, London Love SC, Novak I, Kentish M, Desloovere K, Heinen F, Molenaers G, O’Flaherty S, Graham HK (2010) Botulinum toxin assessment, intervention and after-care for lower limb spasticity in children with cerebral palsy: international consensus statement. Eur J Neurol 17:9–37 Martin L, Baker R, Harvey A (2010) A systematic review of common physiotherapy interventions in school-aged children with cerebral palsy. Phys Occup Ther Pediatr 30(4):294–312 Middleton A, Fritz SL, Lusardi M (2015) Walking speed: the functional vital sign. J Aging Phys Act 23(2):314–322 Miller F (2004) Cerebral palsy. Springer, New York Monge Pereiraa E, Rueda FM, Diego IMA, de la Cuerda RC, de Mauroc A, Miangolarra JC (2014) Empleo de sistemas de realidad virtual como método de propiocepción en parálisis cerebral: guía de práctica clínica. Neurologia 29(9):550–559 Novacheck TF, Stout JL, Tervo R (2000) Reliability and validity of the Gillette Functional Assessment Questionnaire as an outcome measure in children with walking disabilities. J Pediatr Orthop 20(1):75–81 O’Byrne JM, Jenkinson A, O’Brien TM (1998) Quantitative analysis and classification of gait patterns in cerebral palsy using a three-dimensional motion analyzer. J Child Neurol 13(3):101–108 Pandyan AD, Johnson GR, Price CIM, Curless RH, Barnes MP, Rodgers H (1999) A review of the properties and limitations of the Ashworth and modified Ashworth Scales as measures of spasticity. Clin Rehabil 13(5):373–383 Patrick E, Ada L (2006) The Tardieu Scale differentiates contracture from spasticity whereas the Ashworth Scale is confounded by it. Clin Rehabil 20(2):173–182 Paul SM, Siegel KL, Malley J, Jaeger RJ (2007) Evaluating interventions to improve gait in cerebral palsy: a meta-analysis of spatiotemporal measures. Dev Med Child Neurol 49(7):542–549 Piccinini L, Cimolin V, Galli M, Berti M, Crivellini M, Turconi AC (2007) Quantification of energy expenditure during gait in children affected by cerebral palsy. Eura Medicophys 43(1):7–12 Rathinam C, Bateman A, Peirson J, Skinner J (2014) Observational gait assessment tools in paediatrics – a systematic review. Gait Posture 40(2):279–285 Roberts A, Stewart C, Freeman R (2015) Gait analysis to guide a selective dorsal rhizotomy program. Gait Posture 42(1):16–22 Rodda J, Graham HK (2001) Classification of gait patterns in spastic hemiplegia and spastic diplegia: a basis for a management algorithm. Eur J Neurol 8:98–108 Rodda JM, Graham HK, Carson L, Galea MP, Wolfe R (2004) Sagittal gait patterns in spastic diplegia. J Bone Joint Surg Br 86(2):251–258 Ross SA, Engsberg JR (2007) Relationships between spasticity, strength, gait, and the GMFM-66 in persons with spastic diplegia cerebral palsy. Arch Phys Med Rehab 88(9): 1114–1120 Scholtes VAB, Becher JG, Beelen A, Lankhorst GJ (2006) Clinical assessment of spasticity in children with cerebral palsy: a critical review of available instruments. Dev Med Child Neurol 48(1):64–73 Sossai R, Vavken P, Brunner R, Camathias C, Graham HK, Rutz E (2010) Patellar tendon shortening for flexed knee gait in spastic diplegia. Gait Posture 41(2):658–665 Thomason PJ, Selber P, Graham HK (2013) Single event multilevel surgery in children with spastic cerebral palsy: 5 year prospective cohort study. Gait and Posture 37(1):23–28 Trompetto C, Marinelli L, Mori L, Pelosin E, Currà A, Molfetta L, Abbruzzese G (2014) Pathophysiology of spasticity: implications for neurorehabilitation. BioMed Res Int 2014:1–8 Wilson NC, Chong J, Mackey AH, Stott NS (2014) Reported outcomes of lower limb orthopaedic surgery in children and adolescents with cerebral palsy: a mapping review. Dev Med Child Neurol 56(9):808–814 Wren TA, Do KP, Hara R, Dorey FJ, Kay RM, Otsuka NY (2007) Gillette gait Index as a gait analysis summary measure: comparison with qualitative visual assessments of overall gait. J Pediatr Orthop 27(7):765–768

Natural History of Cerebral Palsy and Outcome Assessment Erich Rutz and Pam Thomason

Abstract

Cerebral palsy (CP) is defined as “a group of permanent disorders of the development of movement and posture, causing activity limitation, that are attributed to nonprogressive disturbances that occurred in the developing fetal or infant brain. The motor disorders of CP are often accompanied by disturbances of sensation, perception, cognition, communication, and behavior, by epilepsy; and by secondary musculoskeletal problems.” Though the neurological disturbance is nonprogressive, the natural history is for deterioration in gait and motor function due to musculoskeletal pathology. The natural history and outcomes of musculoskeletal problems are discussed. The measurement tools and comprehensive assessment used to evaluate gross motor function and gait impairments in children with CP are presented. Keywords

Cerebral palsy • Musculoskeletal deformity • Assessment • Gait • Hip displacement • Outcome

Contents State of the Art . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Natural History, Spasticity, Muscle Force, and Growth . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . The Hip and Spine . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Orthopedic Surgery to Improve Gait and Function . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Classification . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

1054 1054 1056 1058 1059

E. Rutz (*) Pediatric Orthopaedic Department, University Children’s Hospital Basel, Basel, Switzerland e-mail: [email protected] P. Thomason Hugh Williamson Gait Analysis Laboratory, Royal Children’s Hospital, Melbourne, VIC, Australia e-mail: [email protected] # Springer International Publishing AG, part of Springer Nature 2018 B. Müller, S.I. Wolf (eds.), Handbook of Human Motion, https://doi.org/10.1007/978-3-319-14418-4_49

1053

1054

E. Rutz and P. Thomason

Outcome Assessment . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Body Structure and Function . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Activity . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Functional Mobility Scale . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Functional Assessment Questionnaire . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Cross-References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

1061 1063 1066 1066 1066 1068 1068 1069

State of the Art Natural History, Spasticity, Muscle Force, and Growth Natural history of gait in children with cerebral palsy (CP) is one of deterioration. Several studies report on these findings (Norlin and Odenrick 1986; Johnson et al. 1997; Bell et al. 2002; Rose et al. 2010; Morgan and McGinley 2014). The most important factors that affect the natural history of deterioration, in our opinion, are the following: increased muscle tone during growth, the development of deformities, and a decrease of muscular force. Primary features in children with CP are spasticity, muscle weakness, and lack of selective motor control. All these factors are in conflict with growth in children with CP. Lower limbs deformities are worsened by the increase of body weight (Davids et al. 2015). A longitudinal study (Johnson et al. 1997) over a mean of 32 months on 18 subjects with spastic diplegia, ranging in age from 4 to 14 years was conducted. Instrumented gait analyses (IGA) were performed to compare the temporal and kinematic data across the two time intervals. The comparison revealed a deterioration of gait stability, evidenced by increases in double support and decreases in single support time with time and growth. In this study, kinematic analysis revealed a loss of excursion about the knee, ankle, and pelvis. Additionally, passive range-ofmotion analysis revealed a decrease in the popliteal angle over time. They concluded that, in contrast to the gait of children with intact motor function, ambulatory ability tends to worsen over time in spastic CP (Johnson et al. 1997). Bell et al. (2002) showed a deterioration of gait temporal and stride parameters and kinematics over time in 28 children with CP. There was no surgical intervention in this cohort during the study period of 4.4 years. Of the 28, 19 children were classified with the diagnosis of diplegia, seven with hemiplegia, and two with quadriplegia. In general, CP is considered a static deformity. But in this study (Bell et al. 2002), changes in function commonly result with growth and age. At the time of the first 3DGA, recommendations for 14 were conservative (e.g., bracing, botulinum toxin A). After 4.4 years, at the end of the study, surgery was recommended in all but two children. Hagglund and Wagner, (2008) reported the development of spasticity with age in children with CP. In 1994, a register and a health-care program for children with CP in southern Sweden were initiated. In the program, the child’s muscle tone according to the Modified Ashworth Scale is measured twice a year until 6 years of

Natural History of Cerebral Palsy and Outcome Assessment

1055

age and then once a year. They used this data to analyze the development of spasticity with age in a total population of children with CP. In their study, all measurements of muscle tone in the gastrocnemius-soleus muscle in all children with CP from 0 to 15 years during the period 1995–2006 were analyzed. They found that in the total sample, the degree of muscle tone increased up to 4 years of age. After 4 years of age, the muscle tone decreased each year up to 12 years of age. A similar development was seen when excluding the children operated with selective dorsal rhizotomy, intrathecal baclofen pump, or tendo achilles lengthening. At 4 years of age, 47% of the children had spasticity in their gastro-soleus muscle graded as Ashworth 2–4. After 12 years of age, 23% of the children had that level of spasticity. Hagglund and Wagner, (2008) concluded that, in children with CP, the muscle tone as measured with the Modified Ashworth Scale increases up to 4 years of age and then decreases up to 12 years of age. Morgan and Mc Ginley conducted a systematic review on gait function and decline in adults with CP. Sixteen studies met the inclusion criteria, and the reported mean ages of the study groups varied from 22 to 42.6 years of age. Results suggest that mobility decline occurs in 25% or more of adults with CP. Those at higher risk of gait decline are those with worse initial gait ability, bilateral rather than unilateral motor impairment, older age, and higher levels of fatigue and pain. They conclude that many ambulant adults with CP experience mobility decline earlier than their nondisabled peers (Morgan and McGinley 2014). The relationship of strength, weight, age, and function in ambulatory children with CP was investigated by Davids et al. (2015). The study design was prospective, case series of 255 subjects, aged 8–19 years, with diplegic type of CP. Linear regression was used to predict the rate of change in lower extremity muscle strength, body weight, and strength normalized to weight (STR-N) with age. The cohort was analyzed as a whole and in groups based on functional impairment as reflected by Gross Motor Function Classification System (GMFCS) level. In their results, they found that strength increased significantly over time for the entire cohort at a rate of 20.83 N/y (P = 0.01). Weight increased significantly over time for the entire cohort at a rate of 3.5 kg/y (P < 0.0001). Lower extremity STR-N decreased significantly over time for the entire cohort at a rate of 0.84 N/kg/y (P < 0.0001). The rate of decline in STR-N (N/kg/y) was comparable among age groups of the children in the study group. Interestingly they stated that there were no significant differences in the rate of decline of STR-N (N/kg/y) among GMFCS levels. There was a 90% chance of independent ambulation (GMFCS levels I and II) when STR-N was 21 N/kg (49% predicted relative to typically developing children). The results of this study (Davids et al. 2015) support the long-standing clinically based observation that STR-N decreases with age in children with CP. This decrease occurs throughout the growing years and across GMFCS levels I to III. Independent ambulation becomes less likely as STR-N decreases. The natural history of gross motor development in children with CP aged 1–15 years was reported by Beckung et al. (2007). These curves may be useful for monitoring and predicting motor development, for planning treatment, and for evaluating outcome after interventions.

1056

E. Rutz and P. Thomason

Fig. 1 Indicates the typical posture of a child with spastic diplegia

In all of these studies, the natural history of ambulation in children with CP consists of deterioration over time. (Norlin and Odenrick 1986; Bell et al. 2002; Rose et al. 2010; Rutz et al. 2012; Morgan and McGinley 2014; Davids et al. 2015; Graham et al. 2016). Figures 1 and 2 present illustrative postures of the lower extermities (1 in bilateral spastic CP, 2 in unilateral spastic CP).

The Hip and Spine Soo et al. (2006) reported that hip displacement is common in children with CP, with an overall incidence of 35% found in their study. The risk of hip displacement is directly related to gross motor function as graded with the GMFCS (Palisano et al. 1997). There are four studies from hip surveillance programs, and all report very similar results (Soo et al. 2006; Connelly et al. 2009; Terjesen 2012; Hagglund et al. 2014). Hips in ambulatory children with CP are usually stable or easily made so, but hips in children with GMFCS levels IV and V have unstable hips (Rutz 2012; Rutz et al. 2015). The untreated hip dislocation is not an easy problem. Pritchett (1983) reported the incidence of pain and other complications associated with the untreated spastic dislocated or subluxed hip in 80 institutionalized patients. The average age was 19 years, and the average length of follow-up study was 12 years. Eighty-five percent of the patients had scoliosis, and 56% had significant pelvic obliquity. A dislocated hip predisposed to lower extremity fractures but did not present problems of pain, decubitus ulcers, or difficulties with perineal hygiene. The loss of sitting balance, scoliosis, and pelvic obliquity are correlated with the severity of neurologic involvement rather than with the mechanics of a dislocated hip. Several studies reported pain and other problems as a significant finding (Samilson et al. 1972; Moreau et al. 1979). Knapp et al. (Knapp and Cortes 2002) reported seven dislocated hips (18%, out of 29 hips) were definitely painful and four hips (11%) produced only mild or intermittent pain. Twenty-seven hips (71%) were not painful. Pruszczynski and Miller (2015) performed a review and reported that when hip displacement in children with CP is identified early, treatment is more successful.

Natural History of Cerebral Palsy and Outcome Assessment

1057

Fig. 2 Presents a child with spastic hemiplegia on the right side

The review included ten studies with sample sizes greater than 20 children. They found that the GMFCS level has a strong impact on subluxation risk and that the risk continues to the end of growth. Wawrzuta et al. 2016 studied “hip health” in a population-based cohort of adolescents and young adults with CP to investigate associations between hip morphology, pain, and gross motor function. Ninety-eight young adults were reviewed at a mean age of 18 years, 10 months. Hip morphology was classified using the Melbourne Cerebral Palsy Hip Classification Scale (MCPHCS) (Murnaghan et al. 2010). Hip pain was reported in 72% of participants. Poor hip morphology at skeletal maturity was associated with high levels of pain. Limited hip surveillance and access to surgery, rather than GMFCS, were associated with poor hip morphology. The majority of young adults with access to hip surveillance and preventive and reconstructive surgery had satisfactory hip morphology at skeletal maturity and less pain. The role of pelvic obliquity is not entirely clear, but its management deserves closer scrutiny in children and adolescents with CP (Heidt et al. 2015). Natural history of scoliosis in spastic CP was described very well in the classical paper of Saito et al. (1998). They concluded that the risk factors for progression of scoliosis in spastic CP are having a spinal curve of 40 before age 15 years, having

1058

E. Rutz and P. Thomason

total body involvement, being bedridden, and having a thoracolumbar curve. Patients with these risk factors might benefit from early surgical intervention to prevent progression to severe scoliosis (Saito et al. 1998). Gu et al. (2011) performed a retrospective study to analyze the development and progression of scoliosis in children and adolescents with nonambulatory spastic quadriplegic CP. A total of 110 children and adolescents 40 by the age of 12 years were associated with greater increases in Cobb angle with age (Gu et al. 2011).

Orthopedic Surgery to Improve Gait and Function The current concept for the management of musculoskeletal deformities in children with CP is single-event multilevel surgery (SEMLS). In a systematic review of SEMLS evidence was found for large improvements in gait dysfunction, moderate improvements in health-related quality of life, and only small changes in gross motor function (McGinley et al. 2012). In the first randomized clinical trial of SEMLS, a 50% improvement in gait function and a 4.9% improvement in gross motor function (GMFM-66) were reported (Thomason et al. 2011). The 5-year results of this clinical trial show that these improvements were largely maintained at 5 years after SEMLS (Thomason et al. 2013). In the SEMLS approach, the gait pattern is identified and evaluated by IGA as part of the diagnostic matrix (Davids et al. 2003). SEMLS in bilateral spastic CP can be considered an exercise in correcting anatomical deformities based on the clinical findings. It is necessary to consider all components of the matrix so that surgical planning is optimized for the individual child. Careful preoperative discussions about setting realistic goals help to ensure that the patient’s, parents’, and surgeon’s goals are consistent and achievable (Thomason and Graham 2013). A comprehensive plan is then developed for the correction of all muscle tendon contractures, torsional malalignments, and joint instabilities in one operative session. Rehabilitation requires at least 1 year and improvements continue into the second year, postoperatively. A carefully tailored and carefully monitored rehabilitation program can ensure that the child will reach a higher level of function (Thomason et al. 2013; Thomason and Graham 2013). Follow-up until skeletal maturity is recommended, to detect for new or recurrent deformities. The orthotic prescription must be carefully monitored throughout the first year after surgery. Functional recovery and orthotic prescription can be monitored by a gait laboratory visit every 3 months for the first year after surgery and yearly thereafter. Our approach to SEMLS rehabilitation has been described in more detail elsewhere (Thomason and Graham 2013).

Natural History of Cerebral Palsy and Outcome Assessment

1059

Multilevel orthopedic surgery for older children and adolescents with severe crouch gait is effective for relieving stress on the knee extensor mechanism, reducing knee pain, and improving function and independence (Rodda et al. 2006). Surgical treatment for equinus gait in children with spastic diplegia was successful, at a mean of 7 years, in the majority of cases when combined with multilevel surgery, orthoses, and rehabilitation. No patient developed crouch gait, and the rate of revision surgery for recurrent equinus was 12.5% (Firth et al. 2013).

Classification Classification of functional levels in children with CP permits description of the level of impairment, encourages clear communication, and is important to establish gross motor prognosis, monitoring progress and in choice of appropriate interventions (Rosenbaum et al. 2002, 2007a). Development of the Gross Motor Functional Classification System (GMFCS) has given clinicians and parents a common language to communicate about CP (Palisano et al. 1997, 2008). The GMFCS is a classification system, and its use is essential when discussing gross motor function in children with CP. It should be used alongside the classification of upper limb function (Manual Ability Classification System) and communication abilities (Communication Function Classification System) to provide the essential context for considering the individual child’s prognosis, goal setting, management, and outcome evaluation (Eliasson et al. 2006; Hidecker et al. 2011). The descriptors and accompanying illustrations are found in Fig. 3a, b. Although GMFCS levels are considered to be stable over time, it is important to note that the descriptors for youth aged 13–18 years allow for deterioration in gross motor function, a decrease in independent walking abilities, and an increase in the need for both assistive devices and wheeled mobility. This is particularly true at GMFCS levels II, III, and IV (McCormick et al. 2007). Certain musculoskeletal features and deformities are closely related to GMFCS level. The shape of the proximal femur shows a strong correlation with GMFCS level. (Robin et al. 2008). The incidence and severity of hip displacement are directly predicted by GMFCS level (Soo et al. 2006). The relationship between GMFCS and hip displacement has implications for screening and management protocols. Changes in GMFCS levels should be carefully documented. The GMFCS may not be stable in the very young child. However, the most common reason for a change in GMFCS level is an error in interpretation in the previous or current examination. Given that the GMFCS is a categorical grading system, true changes in GMFCS level sometimes occur which may be in both directions, namely, improvement or deterioration. After major intervention such as selective dorsal rhizotomy (SDR) or single-event multilevel surgery (SEMLS), a few children move up a level. This is uncommon and should not be expected in more than 5–10% of children (Rutz et al. 2012). Deterioration in GMFCS level is more common. For example, lengthening of the Achilles’ tendons in children in GMFCS level II can result in progressive crouch gait and the need for assistive devices. For these children, their gait and function

Fig. 3 (a) GMFCS levels descriptors for ages 6–12 years (a) and 13–18 years (b)

1060 E. Rutz and P. Thomason

Natural History of Cerebral Palsy and Outcome Assessment

1061

deteriorate, and GMFCS level changes from II to III (Rodda et al. 2006). The relative stability of the GMFCS means that this is not an outcome measure and was never meant to be one. The GMFCS provides an excellent guide to long-term prognosis and influences our decision making and management plan. The development of gross motor function in children with CP can be described by a series of curves (Rosenbaum et al. 2002). Understanding the position of a child’s development in relation to their gross motor curve provides a rational basis for understanding management strategies, goal setting, and long-term gross motor function (Hanna et al. 2008). The popularity of many forms of intervention in early childhood in children with CP is the mistaken attribution of improvements in gross motor function to the intervention, when natural history has such a large effect. Association is not causation. In most children, gross motor function reaches a plateau between 3 and 6 years, with some regression in later childhood. One of the causes for this regression in gross motor function is progressive musculoskeletal pathology (Graham 2004). After age 6 years, gait parameters deteriorate as contractures and bony deformities increase. It should be noted that the gross motor curves (Rosenbaum et al. 2002) include GMFM data up to the age of 15 years. There is much less information regarding changes in gross motor function in the 15–20 age group and beyond. Changes in gross motor function and gait during the plateau/early decline phase can be more realistically attributed to intervention.

Outcome Assessment Previously we have discussed the natural history of the musculoskeletal pathology in CP. Musculoskeletal impairments affect many aspects of the child’s physical functioning, limiting their levels of physical activity and participation (Graham et al. 2016; Rosenbaum et al. 2007b). Children with CP often undergo interventions designed to modify the natural history of musculoskeletal pathologies and improve their gross motor and gait function, including botulinum toxin A injections, physiotherapy, and orthopedic surgery (Graham et al. 2016). It is extremely important to be able to accurately assess the outcomes of these interventions. Outcome measures must consider a child’s level of function across multiple domains (Oeffinger et al. 2009). The World Health Organization’s International Classification of Functioning, Disability and Health (ICF) provides a useful framework for the measurement of outcomes following interventions in children with CP. The ICF considers health conditions with regard to three domains: body structure and function and activities and participation. These domains are influenced by environmental and personal factors (WHO 2001). Various tools exist to measure outcomes relevant to children with CP in the ICF domains, and new measurement tools are being developed. These are shown in Fig. 4. When considering measurement of outcomes in children with CP, it is important to consider all of the components of the ICF. It is also important to choose the correct measurement tool to use at any given time or after an intervention. Choice of measurement tool should be based on the psychometric properties of the tool, the

1062

E. Rutz and P. Thomason

Fig. 4 Measurement tools used for gross motor function in children with CP per ICF domains

aspect of the ICF being measured, as well as the age and GMFCS level of the child. A detailed discussion of measurement tools is beyond the scope of this chapter. Some of the tools that we use commonly are discussed below.

The Diagnostic Matrix Assessment tools can be considered in the context of a diagnostic matrix (Davids et al. 2003). Davids and colleagues described a diagnostic matrix consisting of

Natural History of Cerebral Palsy and Outcome Assessment

1063

clinical history, physical examination, diagnostic imaging, instrumented gait analysis (IGA), and examination under anesthesia as five components of a diagnostic matrix useful in clinical decision making, in relation to gait correction surgery in children with CP. The diagnostic matrix consists of measurements across multiple domains of the ICF and includes many aspects of gross motor function. Incorporating these measurements builds up a comprehensive picture of the individuals’ function which assists with diagnosis and identification of impairments and assist treatment planning especially in relation to major interventions such as gait improvement surgery. Major interventions, such as selective dorsal rhizotomy and multilevel orthopedic surgery, are designed to improve gait and functioning in children with CP. Ideally these major interventions are most appropriately conducted following the most comprehensive and objective assessment possible. Over time we have added additional measurement tools to the diagnostic matrix used in the gait laboratory. Figure. 5 shows the diagnostic matrix and assessment tools included in the matrix.

Body Structure and Function Instrumented Gait Analysis The role of instrumented gait analysis (IGA) is crucial to evaluating gait dysfunction in children with CP, especially in relation to planning and assessing the outcome of major interventions such as selective dorsal rhizotomy (SDR) and single-event multilevel surgery (SEMLS), and is central to the diagnostic matrix. Three-dimensional kinematics, kinetics, and dynamic electromyography provides a comprehensive description of joint movements, moments, and powers and muscle timing which is essential to the management of gait disorders in children with CP (Baker 2013). The objectivity and relative freedom from bias of IGA are factors of major importance in establishing objective outcomes. The use of IGA and a composite measure of gait such as the Gillette Gait Index (Schutte et al. 2000), the Gait Deviation Index (McMulkin and MacWilliams 2015), or the gait profile score (GPS) (Baker et al. 2009) are useful to describe outcomes, of prospective cohort studies as well as randomized trials of multilevel surgery in children with CP. The movement analysis profile (MAP) and overall gait profile score (GPS) have been developed to summarize kinematic data. The root mean square (RMS) difference between nine clinically relevant kinematic variables for a particular child and the average values of that variable from typically developing children are calculated. This represents a clinical meaningful measure of gait variables as it measures difference in degrees and is of value both clinically and in the research setting to evaluate change following surgery (Baker et al. 2009). These measures have revolutionized gait outcome assessment. We are now able to document change in gait, and we can compare kinematic data collected from different centers, allowing the possibility of large cohort collaborative studies.

1064

E. Rutz and P. Thomason

Fig. 5 Assessment Matrix used for assessment of motor function in CP

Video Gait Analysis Video gait analysis (VGA) is a central part of the diagnostic matrix. A visual record of a child’s gait and functioning on digital video can be of much greater value than observational gait analysis and a written report. Digital video is objective and can be shared by multiple observers over a long period. It can be replayed in slow motion and can be reviewed repeatedly. VGA can be used when IGA is either not appropriate or not available. We found VGA particularly useful for the objective documentation of younger children commencing BoNT-A therapy. At this stage, they are too small, too young, and uncooperative for IGA. It is also useful for the selection and monitoring of the use of ankle-foot orthoses (AFOs) and for the monitoring of children after major intervention such as selective dorsal rhizotomy (SDR) or multilevel surgery. In an effort to quantify and objectify the outcome of observational gait analysis, several gait scores have been developed. These include the Physician Rating Scale, the Observational Gait Scale, and the Edinburgh Visual Gait Score (Koman et al. 1993; Mackey et al. 2003; Read et al. 2003; Wren et al. 2005). We do not consider VGA an adequate substitute for IGA when decisions regarding major intervention such as selective dorsal rhizotomy or multilevel surgery have

Natural History of Cerebral Palsy and Outcome Assessment

1065

to be made. Nor is VGA adequate for outcome measurement in clinical trials of gait correction surgery.

Clinical History and Standardized Physical Examination The clinical history is obtained by a careful review of all current and previous medical records complemented by an up to date interview of the child in the context of his family or care providers. Associated medical comorbidities and the response to previous interventions are also crucial to the planning of interventions. Goals and expectations can be established using specific tools such as goal attainment scales, Canadian Occupational Performance Measure and the newly developed Gait Outcomes Assessment List. A standardized routine physical examination is an essential part of the diagnostic matrix. The clinical protocol used in The Hugh Williamson Gait Analysis Laboratory has been published elsewhere (Thomason et al. 2013) and will not be described in detail here. Sagittal Gait Patterns We believe that classification of sagittal gait patterns initially from VGA and then from the information from IGA (sagittal gait kinematics) to be very important. The classification of sagittal gait patterns in spastic hemiplegia (Winters et al. 1987) is a valid and reliable tool which helps in framing logical management strategies. The sagittal gait classification described by Rodda and Graham can be useful in planning intervention in spastic diplegia (Rodda et al. 2004). Both gait classifications suggest common patterns of musculoskeletal deformity and may assist in identifying deformities that require intervention. Standardized Radiology Medical imaging is important in the diagnostic matrix as high proportions of children with CP have skeletal deformities including torsional deformities of long bones and instability of the hip and foot specifically the subtalar and midtarsal joints. Radiology of the hips, including plain radiographs, supplemented by CT measurements of femoral torsion and tibial torsion and more recently, biplanar radiography using EOS (Escott et al. 2013), can be very useful as additional information in the planning of multilevel surgery. Weight-bearing radiographs can be analyzed by the measurement of a series of key radiological indices which can help identify segmental malalignments in the hindfoot, midfoot, and forefoot in a systematic manner (Davids et al. 2005). This contributes greatly to the analysis of segmental foot deformity and the planning of intervention. Instrumented gait analysis most typically interprets the foot as a rigid segment and does not provide detailed information on segmental malalignments within the foot. Until better foot models are in routine use, standardized weight-bearing radiographs remain the cornerstone of analysis of deformities within the foot. This information can be augmented by dynamic pedobarography which is in use in a number of gait laboratories.

1066

E. Rutz and P. Thomason

Longitudinal Assessments with Radiology: Hip Surveillance Centers in Europe and Australia have developed formal “hip surveillance programs” (Dobson et al. 2002; Hagglund et al. 2005; Kentish et al. 2011). Children with a confirmed diagnosis of CP are offered regular clinical and radiographic examination of their hips and access to both preventive and reconstructive surgery. In both Victoria, Australia, and Southern Sweden, the prevalence of late dislocation has decreased, and the need for salvage surgery has been reduced (Dobson et al. 2002; Hagglund et al. 2005). The most useful radiographic index for measuring hip displacement in children with CP is the migration percentage of Reimers 1980. This measures the percentage of the femoral head that lies outside the acetabulum. Migration percentage can be reliably measured from anterior-posterior hip radiographs, taken in supine with good positioning and a standardized technique (Parrott et al. 2002). It is the key index for making decisions about surgical management and to monitor hip displacement both before and after operative intervention. The incidence of hip displacement (migration percentage >30%) is 35% of children with CP in population-based studies and is directly related to the child’s GMFCS level (Soo et al. 2006; Hagglund et al. 2007; Connelly et al. 2009). Early hip displacement is silent, and formal screening by radiographs of the hips is advised. The frequency of such radiographs should be directly related to the risk of hip displacement, which is in turn related to the child’s GMFCS level. Evidence-based recommendations on hip surveillance in children with CP have been published (Wynter et al. 2011).

Activity Functional Mobility Scale The FMS is a six level ordinal scale that rates the mobility of children with CP over three distances according to their need for assistive devices (Graham et al. 2004). The three distances of 5, 50, and 500 Meters represent mobility in the home, school, and wider community settings, respectively. The scale is clinician administered through parent or child report and should reflect performance rather than capability, i.e., what the child actually does do rather than what they can do. For each of the three distances, a rating of 1–6 is assigned. The FMS was designed as an outcome measure and is sensitive to change (Thomason et al. 2011). Following optimum biomechanical realignment and correction of spastic contractures, children can often progress to lesser levels of support than used preoperatively (Thomason et al. 2011). These important changes can be monitored and reported using the FMS (Graham et al. 2004; Harvey et al. 2007).

Functional Assessment Questionnaire The Functional Assessment Questionnaire (FAQ) is a 10-level, parent-reported walking scale, which describes a range of walking abilities across the entire

Natural History of Cerebral Palsy and Outcome Assessment

1067

spectrum of CP, from nonambulatory to independent ambulation at a high level (Novacheck et al. 2000). In addition to the 10-level walking scale, there is an additional list of 22 items describing a variety of higher-level functional activities requiring varying degrees of walking ability, balance, strength, and coordination. The FAQ is a valid and reliable scale and has been shown to be sensitive to change (Gorton et al. 2011). It is a simple scale, which can be quickly completed by parents or caregivers and provides an excellent longitudinal view of the child’s gross motor and walking abilities. It is a good measure of parental perspective however and covers a wide variety of activities of daily living. The FMS and FAQ are complementary scales and are used to assess outcomes in children with CP after intervention.

The Gross Motor Function Measure The Gross Motor Function Measure (GMFM) (Russell and Rosenbaum 1989) is the gold standard for the measurement of gross motor function in children with CP and has been shown to be valid, reliable, and responsive to change (Russell and Rosenbaum 1989; Nordmark et al. 1997, 2000; Bjornson et al. 1998a, b; Russell et al. 2000; Russell and Leung 2003; Russell and Gorter 2005). There are two versions of the GMFM. The original version consisted of 88 items, which were grouped into five dimensions of gross motor function: lying and rolling, sitting, crawling and kneeling, standing and walking, and running and jumping (Russell and Rosenbaum 1989). Following Rasch analysis, a revised version, which consists of 66 items, was developed, is interval scaled, and features item maps. It is quicker to administer. However, it is limited for children who are very young or severely involved (Russell and Leung 2003). The use of GMFM requires an experienced physiotherapist and will take between 45 and 60 min to administer. It is an essential tool in clinical outcome studies to assess change in gross motor function. It is also a useful clinical measure to assess function in CP and guide the management and treatment planning. The Pediatric Evaluation of Disability Inventory (PEDI) (Haley 1997) may be used to assess motor function. The PEDI may be a more appropriate tool for use with children in GMFCS levels IV and V. A detailed examination of the child’s level of activity and participation, using measures such as the Canadian Occupational Performance Measure, Children’s Assessment of Participation and Enjoyment, or the Activity Scale for Kids (Law et al. 1990; Young et al. 2000; King et al. 2004), may be useful. Self-Reported Questionnaires and Health-Related Quality of Life In recent years, the assessment of quality of life has become a major goal in health management including children with CP. A number of generic and specific instruments have been developed which address aspects of health, functioning, and quality of life. The Child Health Questionnaire (CHQ) (Waters et al. 2000) is a widely used tool and is not disease specific. It has the advantage that the scores of children with CP can be compared to children with typical health or with other disease conditions. The Pediatric Orthopaedic Data Collection Instrument (PODCI) (Daltroy et al. 1998)

1068

E. Rutz and P. Thomason

has a more musculoskeletal focus and contains several domains directly relevant to children with CP and gait problems. Although some information exists on the use of both of these instruments in children with CP, the responsiveness to change and the value of using these as outcome instruments are not yet fully established. It is also the case that neither can be considered to be a true quality of life measure. Other questionnaires include the PedsQL and the CPQoL-Child. The CPQoLChild is a specific quality of life measure developed for children with CP; however, its responsiveness to physical interventions such as gait improvement surgery is not yet known (Waters et al. 2006). In order to judge the effectiveness of any intervention in children with CP, it is important to understand the priorities and expectations of the child and parent (Oeffinger et al. 2009; Novak et al. 2012). The Gait Outcomes Assessment List (GOAL) is a new outcome measure to evaluate gait priorities and functional mobility for ambulant children with CP (Narayanan et al. 2015). There are two versions of the questionnaire, parent and child. There are 48 items in both versions grouped into seven domains. Recent research has establishes the validity of the GOAL in measuring the gross motor and gait function of ambulant children with CP. Evidence was found for the discriminative validity of the GOAL, and correlations were demonstrated with standard measures of gross motor function and gait. The GOAL provides meaningful information about a child’s function across multiple dimensions, accounts for the environmental and personal factors that may contribute to function, and measures the priorities and expectations of children and their parents. The GOAL will allow clinicians to better understand the motor abilities, priorities, and expectations of ambulant children with CP and enable better decision making about appropriate interventions. The GOAL will be an invaluable addition to the measurement tools available for gross motor function in CP.

Summary In conclusion, natural history of gait in children with CP is one of deterioration. There is level II evidence that single-event multilevel surgery (SEMLS) improves the gait of children with spastic diplegic CP 12 months after surgery (Thomason et al. 2011). SEMLS (Rutz et al. 2013) results in clinically and statistically significant improvements in gait and function, in children with bilateral spastic CP, which were maintained at 5 years after surgery (Rutz et al. 2013; Thomason et al. 2013), and GMFCS stability was confirmed in the majority of children (Rutz et al. 2012).

Cross-References ▶ Assessing Clubfoot and Cerebral Palsy by Pedobarography ▶ Clinical Gait Assessment by Video Observation and 2D Techniques ▶ Diagnostic Gait Analysis Use in the Treatment Protocol for Cerebral Palsy

Natural History of Cerebral Palsy and Outcome Assessment

1069

▶ EMG Activity in Gait: The Influence of Motor Disorders ▶ Foot and Ankle Motion in Cerebral Palsy ▶ Functional Effects of Foot Orthoses ▶ Gait Scores: Interpretations and Limitations ▶ Interpreting Ground Reaction Forces in Gait ▶ Interpreting Joint Moments and Powers in Gait ▶ Kinematic Foot Models for Instrumented Gait Analysis ▶ Optimal Control Modeling of Human Movement ▶ Skeletal Muscle Structure in Spastic Cerebral Palsy ▶ Strength Related Stance Phase Problems in Cerebral Palsy ▶ Surface Electromyography to Study Muscle Coordination ▶ Swing Phase Problems in Cerebral Palsy ▶ The Conventional Gait Model - Success and Limitations ▶ Trunk and Spine Models for Instrumented Gait Analysis ▶ Variations of Marker Sets and Models for Standard Gait Analysis

References Baker R (2013) Measuring walking: a handbook of gait analysis. Mac Keith Press, London Baker R, McGinley JL, Schwartz MH et al (2009) The gait profile score and movement analysis profile. Gait Posture 30:265–269 Beckung E, Carlsson G, Carlsdotter S, Uvebrant P (2007) The natural history of gross motor development in children with cerebral palsy aged 1 to 15 years. Dev Med Child Neurol 49(10):751–756 Bell KJ, Ounpuu S, DeLuca PA, Romness MJ (2002) Natural progression of gait in children with cerebral palsy. J Pediatr Orthop 22(5):677–682 Bjornson KF, Graubert CS, Buford VL, McLaughlin J (1998a) Validity of the gross motor function measure. Pediatr Phys Ther 10:43–47 Bjornson KF, Graubert CS, McLaughlin J, Kerfeld CI, Clark EM (1998b) Test retest reliability of the gross motor function measure in children with cerebral palsy. Phys Occup Ther Pediatr 18:51–61 Connelly A, Flett P, Graham HK, Oates J (2009) Hip surveillance in Tasmanian children with cerebral palsy. J Paediatr Child Health 45(7–8):437–443 Daltroy LH, Liang MH, Fossel AH, Goldberg M (1998) The POSNA pediatric musculoskeletal functional health questionnaire: report on reliability, validity and sensitivity to change. The pediatric outcomes instrument development group. J Pediatr Orthop 18:561–571 Davids JR, Ounpuu S, DeLuca PA, Davis RB (2003) Optimization of walking ability of children with cerebral palsy. J Bone Joint Surg Am 85:2224–2234 Davids JR, Gilson T, Pugh LI (2005) Quantitative segmental analysis of weight-bearing radiographs of the foot an ankle for children. J Pediatr Orthop 25(6):769–776 Davids JR, Oeffinger DJ, Bagley AM, Sison-Williamson M, Gorton G (2015) Relationship of strength, weight, age, and function in ambulatory children with cerebral palsy. J Pediatr Orthop 35(5):523–529 Dobson F, Boyd RN, Parrott J, Nattrass GR, Graham HK (2002) Hip surveillance in children with cerebral palsy: impact on the surgical management of spastic hip disease. J Bone Joint Surg Br 84:720–726 Eliasson A-C, Krumlinde-Sundholm L, Rosblad B et al (2006) The manual ability classification system (MACS) for children with cerebral palsy: scale development and evidence of validity and reliability. Dev Med Child Neurol 48:549–554

1070

E. Rutz and P. Thomason

Escott BG, Ravi B, Weathermon AC, Acharva J, Gordon CL, Babyn PS, Kelley SP, Narayanan UG (2013) EOS low-dose radiography: a reliable and accurate upright assessment of lower-limb lengths. J Bone Joint Surg Am 95(23):e1831–e1837 Firth GB, Passmore E, Sangeux M, Thomason P, Rodda J, Donath S, Selber P, Graham HK (2013) Multilevel surgery for equinus gait in children with spastic diplegic cerebral palsy: medium-term follow-up with gait analysis. J Bone Joint Surg Am 95(10):931–938 Gorton GE 3rd, Stout JL, Bagley AM, Bevans K, Novacheck TF, Tucker CA (2011) Gillette functional assessment questionnaire 22.-item skill set: factor and rasch analyses. Dev Med Child Neurol 53:250–255 Graham HK (2004) Mechanisms of deformity. In: Scrutton D, Damiano D, Mayston M (eds) Management of the motor disorders of children with cerebral palsy, Clinics in developmental medicine, vol 161, 2nd edn. Mac Keith Press, London, pp 105–129 Graham HK, Harvey A, Rodda J, Nattrass GR, Pirpiris M (2004) The functional mobility scale (FMS). J Pediatr Orthop 24(5):514–520 Graham HK, Rosenbaum P, Paneth N, Dan B, Lin JP, Damiano DL, Becher JG, Gaebler-Spira D, Colver A, Reddihough DS, Crompton KE, Lieber RL (2016) Cerebral palsy. Nat Rev Dis Primers 2:15082 Gu Y, Shelton JE, Ketchum JM, Cifu DX, Palmer D, Sparkman A, Jermer-Gu MK, Mendigorin M (2011) Natural history of scoliosis in nonambulatory spastic tetraplegic cerebral palsy. PM R 3(1): 27–32 Hagglund G, Wagner P (2008) Development of spasticity with age in a total population of children with cerebral palsy. BMC Musculoskelet Disord 9:150 Hagglund G, Andersson S, Duppe H, Lauge-Pedersen H, Nordmark E, Westbom L (2005) Prevention of dislocation of the hip in children with cerebral palsy. The first ten years of a population-based prevention programme. J Bone Joint Surg Br 87:95–101 Hagglund G, Lauge-Pedersen H, Wagner P (2007) Characteristics of children with hip displacement in cerebral palsy. BMC Musculoskelet Disord 8:101–107 Hagglund G, Alriksson-Schmidt A, Lauge-Pedersen H, Rodby-Bousquet E, Wagner P, Westbom L (2014) Prevention of dislocation of the hip in children with cerebral palsy: 20-year results of a population-based prevention programme. Bone Joint J 96-B(11):1546–1552 Haley SM (1997) The pediatric evaluation of disability inventory. J Rehab Outcome Meas 1:61–69 Hanna SE, Bartlett DJ, Rivard LM, Russell DJ (2008) Reference curves for the gross motor function measure: percentiles for clinical description and tracking over time among children with cerebral palsy. Phys Ther 88:596–607 Harvey A, Graham HK, Morris ME, Baker RJ, Wolfe R (2007) The functional mobility scale: ability to detect changed following single event multilevel surgery. Dev Med Child Neurol 49(8):603–607 Heidt C, Hollander K, Wawrzuta J, Molesworth C, Willoughby K, Thomason P, Khot A, Graham HK (2015) The radiological assessment of pelvic obliquity in cerebral palsy and the impact on hip development. Bone Joint J 97-B(10):1435–1440 Hidecker MJ, Paneth N, Rosenbaum PL, Kent RD, Little J, Eulenberg JB, Chester K Jr, Johnson B, Michalsen L, Evatt M, Taylor K (2011) Developing and validating the communication function classification system (CFCS) for individuals with cerebral palsy. Dev Med Child Neurol 53:704–710 Johnson DC, Damiano DL, Abel MF (1997) The evolution of gait in childhood and adolescent cerebral palsy. J Pediatr Orthop 17(3):392–396 Kentish M, Wynter M, Snape N, Boyd R (2011) Five-year outcome of state-wide hip surveillance of children and adolescents with cerebral palsy. J Pediatr Rehabil Med 4:205–217 King G, Law M, King S et al (2004) Children’s assessment of participation and enjoyment (CAPE) and preferences for activities of children (PAC). Harcourt Assessment, San Antonio Knapp DR Jr, Cortes H (2002) Untreated hip dislocation in cerebral palsy. J Pediatr Orthop 22(5): 668–671 Koman LA, Mooney JF, Smith B (1993) Management of cerebral palsy with botulinum-A toxin: preliminary investigation. J Pediatr Orthop 4:489–495

Natural History of Cerebral Palsy and Outcome Assessment

1071

Law M, Baptiste S, McColl M, Opzoomer A, Polatajko H, Pollock N (1990) The Canadian occupational performance measure: an outcome measure for occupational therapy. Can J Occup Ther 57:82–87 Mackey AH, Lobb GL, Walt SE, Stott NS (2003) Reliability and validity of the observational gait scale in children with spastic diplegia. Dev Med Child Neurol 45(1):4–11 McCormick A, Brien M, Plourde J, Wood E, Rosenbaum P, McLean J (2007) Stability of the gross motor function classification system in adults with cerebral palsy. Dev Med Child Neurol 49:265–269 McGinley J, Dobson F, Ganeshalingam R, Shore BJ, Rutz E, Graham HK (2012) Single-event multilevel surgery for children with cerebral palsy: a systematic review. Dev Med Child Neurol 54:117–128 McMulkin ML, MacWilliams BA (2015) Application of the Gillette gait index, gait deviation index and gait profile score to multiple clinical pediatric populations. Gait Posture 41:608–612 Moreau M, Drummond DS, Rogala E, Ashworth A, Porter T (1979) Natural history of the dislocated hip in spastic cerebral palsy. Dev Med Child Neurol 21(6):749–753 Morgan P, McGinley J (2014) Gait function and decline in adults with cerebral palsy: a systematic review. Disabil Rehabil 36(1):1–9 Murnaghan ML, Simpson P, Robin JG, Shore BJ, Selber P, Graham HK (2010) The cerebral palsy hip classification is reliable: an inter- and intra-observer reliability study. J Bone Joint Surg Am 92:436–441 Narayanan U, Moline R, Encisa C, Yeung C, Weir S (2015) Validation of the GOAL questionnaire: an outcome measure for ambulatory children with cerebral palsy. Dev Med Child Neurol. 57(29) https://doi.org/10.1111/dmcn.45_12887 Nordmark E, Hagglund G, Jarnlo GB (1997) Reliability of the gross motor function measure in cerebral palsy. Scand J Rehabil Med 29:25–28 Nordmark E, Jarnlo GB, Hagglund G (2000) Comparison for the gross motor function measure and paediatric evaluation of disability inventory in assessing motor function in children undergoing selective dorsal rhizotomy. Dev Med Child Neurol 42:245–252 Norlin R, Odenrick P (1986) Development of gait in spastic children with cerebral palsy. J Pediatr Orthop 6(6):674–680 Novacheck TF, Stout JL, Tervo R (2000) Reliability and validity of the Gillette functional assessment questionnaire as an outcome measure in children with walking disabilities. J Pediatr Orthop 20:75–81 Novak I, Hines M, Goldsmith S, Barclay R (2012) Clinical prognostic messages from a systematic review on cerebral palsy. Pediatrics 130(5):e1285–312. Oeffinger DJ, Rogers SP, Bagley A, Gorton G, Tylkowski CM (2009) Clinical applications of outcome tools in ambulatory children with cerebral palsy. Phys Med Rehabil Clin N Am 20:549–565 Palisano R, Rosenbaum P, Walter S, Russell D, Wood E, Galuppi B (1997) Development and reliability of a system to classify gross motor function in children with cerebral palsy. Dev Med Child Neurol 39(4):214–223 Palisano RJ, Rosenbaum P, Bartlett D, Livingston MH (2008) Content validity of the expanded and revised gross motor function classification system. Dev Med Child Neurol 50:744–750 Parrott J, Boyd RN, Dobson F et al (2002) Hip displacement in spastic cerebral palsy: repeatability of radiologic measurement. J Pediatr Orthop 22:660–667 Pritchett JW (1983) The untreated unstable hip in severe cerebral palsy. Clin Orthop Relat Res 173:169–172 Pruszczynski B, Sees J, Miller F (2015) Risk factors for hip displacement in children with cerebral palsy: systematic review. J Pediatr Orthop 36(8):829–833 Read HS, Hazlewood ME, Hillman SJ, Prescott RJ, Robb JE (2003) Edinburgh visual gait score for use in cerebral palsy. J Pediatr Orthop 23:296–301 Reimers J (1980) The stability of the hip in children: a radiological study of the results of muscle surgery in cerebral palsy. Acta Orthop Scand Suppl 184:1–100

1072

E. Rutz and P. Thomason

Robin J, Graham HK, Selber P, Dobson F, Smith K, Baker R (2008) Proximal femoral geometry in cerebral palsy. A population-based cross-sectional study. J Bone Joint Surg Br 90:1372–1379 Rodda JM, Graham HK et al (2004) Sagittal gait patterns in spastic diplegia. J Bone Joint Surg Br 86(2):251–258 Rodda JM, Graham HK, Nattrass GR, Galea MP, Baker R, Wolfe R (2006) Correction of severe crouch gait in patients with spastic diplegia with use of multilevel orthopaedic surgery. J Bone Joint Surg Am 88(12):2653–2664 Rose GE, Lightbody KA, Ferguson RG, Walsh JC, Robb JE (2010) Natural history of flexed knee gait in diplegic cerebral palsy evaluated by gait analysis in children who have not had surgery. Gait Posture 31(3):351–354 Rosenbaum PL, Walter SD, Hanna SE, Palisano RJ, Russell DJ, Raina P, Wood E, Bartlett DJ, Galuppi BE (2002) Prognosis for gross motor function in cerebral palsy: creation of motor development curves. JAMA 288:1357–1363 Rosenbaum P, Paneth N, Leviton A et al (2007a) A report: the definition and classification of cerebral palsy April 2006. Dev Med Child Neurol Suppl 109:8–14 Rosenbaum PL, Livingston MH, Palisano RJ, Galuppi BE, Russell DJ (2007b) Quality of life and health-related quality of life of adolescents with cerebral palsy. Dev Med Child Neurol 49:516–521 Russell DJ, Gorter JW (2005) Assessing functional differences in gross motor skills in children with cerebral palsy who use an ambulatory aid or orthoses: can the GMFM-88 help? Dev Med Child Neurol 47:462–467 Russell DJ, Leung KM (2003) Accessibility and perceived clinical utility of the GMFM-66: evaluating therapists’ judgements of a computer-based scoring program. Phys Occup Ther Pediatr 23(2):45–58 Russell DJ, Rosenbaum PL (1989) The gross motor function measure: a means to evaluate the effects of physical therapy. Dev Med Child Neurol 31(3):341–352 Russell DJ, Avery LM, Rosenbaum PL, Raina PS, Walter SD, Palisano RJ (2000) Improved scaling of the gross motor function measure for children with cerebral palsy: evidence of reliability and validity. Phys Ther 80:873–885 Rutz E (2012) Are hips stable in children with cerebral palsy? Dev Med Child Neurol 54(10):878 Rutz E, Tirosh O, Thomason P, Barg A, Graham HK (2012) Stability of the gross motor function classification system after single-event multilevel surgery in children with cerebral palsy. Dev Med Child Neurol 54(12):1109–1113 Rutz E, Donath S, Tirosh O, Graham HK, Baker R (2013) Explaining the variability improvements in gait quality as a result of single event multi-level surgery in cerebral palsy. Gait Posture 38(3): 455–460 Rutz E, Vavken P, Camathias C, Haase C, Juenemann S, Brunner R (2015) Long-term results and outcome predictors in one-stage hip reconstruction in children with cerebral palsy. J Bone Joint Surg Am 97(6):500–506 Saito N, Ebara S, Ohotsuka K, Kumeta H, Takaoka K (1998) Natural history of scoliosis in spastic cerebral palsy. Lancet 351(9117):1687–1692 Samilson RL, Tsou P, Aamoth G, Green WM (1972) Dislocation and subluxation of the hip in cerebral palsy. Pathogenesis, natural history and management. J Bone Joint Surg Am 54(4): 863–873 Schutte LM, Narayanan U, Stout JL, Selber P, Gage JR, Schwartz MH (2000) An index for quantifying deviations from normal gait. Gait Posture 11:25–31 Soo B, Howard JJ, Boyd RN, Reid SM, Lanigan A, Wolfe R, Reddihough D, Graham HK (2006) Hip displacement in cerebral palsy. J Bone Joint Surg Am 88(1):121–129 Terjesen T (2012) The natural history of hip development in cerebral palsy. Dev Med Child Neurol 54(10):951–957 Thomason P, Graham HK (2013) Rehabilitation of cerebral palsy. In: Iansek R, Morris M (eds) Rehabilitation in movement disorders. Cambridge University Press, Cambridge, UK

Natural History of Cerebral Palsy and Outcome Assessment

1073

Thomason P, Baker R, Dodd K, Taylor N, Selber P, Wolfe R, Graham HK (2011) Single-event multilevel surgery in children with spastic diplegia: a pilot randomized controlled trial. J Bone Joint Surg Am 93(5):451–460 Thomason P, Selber P, Graham HK (2013) Single event multilevel surgery in children with bilateral spastic cerebral palsy: a 5 year prospective cohort study. Gait Posture 37(1):23–28 Waters E, Salmon L, Wake M, Hesketh K, Wright M (2000) The child health questionnaire in Australia: reliability, validity and population means. Aust N Z J Public Health 24:207–210 Waters E, Davis E, Boyd R, Reddihough D, Mackinnon A, Graham HK, Lo SK, Wolfe R, Stevenson R, Bjornson K, Blair E, Ravens-Sieberer U (2006) Cerebral palsy quality of life questionnaire for children (CP QOL-child) manual. Deakin University, Melbourne Wawrzuta J, Willoughby KL, Molesworth C, Ang SG, Shore BJ, Thomason P, Graham HK (2016) Hip health at skeletal maturity: a population-based study of young adults with cerebral palsy. Dev Med Child Neurol 58(12):1273–1280 Winters TF, Gage JR, Hicks R (1987) Gait patterns in spastic hemiplegia in children and young adults. J Bone Joint Surg Am 69:437–441 World Health Organization (2001) International classification of functioning, disability and health. World Health Organization, Geneva, pp 121–160; cited 2016. http://www.who.int/classifica tions/icf/en/ Wren TAL, Rethlefsen SA, Healy BS, Do KP, Dennis SW, Kay RM (2005) Reliability and validity of visual assessments of gait using a modified physician rating scale for crouch and foot contact. J Pediatr Orthop 25:646–650 Wynter M, Gibson N, Kentish M, Love S, Thomason P, Graham HK (2011) The consensus statement on hip surveillance for children with cerebral palsy; Australian standards of care. J Pediatr Rehab Med 4:183–195 Young N, Williams JI, Yoshida KK, Wright JG (2000) Measurement properties of the activities scale for kids. J Clin Epidemiol 53:125–137

Skeletal Muscle Structure in Spastic Cerebral Palsy Adam Shortland

Abstract

The structure of skeletal muscle in cerebral palsy (CP) is altered at the molecular level, at the cellular level, and at the level of the tissue. These abnormalities in structure have implications for active and passive muscle performance and for the functional capacity of the individual, particularly in the long term. Appreciating the deficits of muscle structure may well encourage clinicians to focus on muscle growth when managing this group and lead researchers to novel therapeutics targeted at normalizing muscle structure. Keywords

Skeletal muscle • Growth • Functional reserve • Cerebral palsy

Contents Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . State of the Art . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Conclusions and Future Directions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Implications for Treatment . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Cross-References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

1075 1077 1086 1086 1087 1087

Introduction Cerebral palsy (CP) is a lifelong condition arising from a brain lesion affecting motor development in early life. The brain injury is considered unchanging but subsequent central nervous system (CNS) and musculoskeletal development is affected. A. Shortland (*) One Small Step Gait Laboratory, Evelina Children’s Hospital, Guy’s and St Thomas’ NHS Foundation Trust, London, UK e-mail: [email protected] # Springer International Publishing AG, part of Springer Nature 2018 B. Müller, S.I. Wolf (eds.), Handbook of Human Motion, https://doi.org/10.1007/978-3-319-14418-4_51

1075

1076

A. Shortland

Muscles are the only tissues in the human body which generate forces and movements. Adequate muscle size and quality in the lower limbs are required to carry out the routine activities of daily life such as walking, getting up from a chair, and climbing the stairs. Of particular concern here is the alteration of the development of muscle which may limit the ability of the person with CP to carry out the functions of daily living that require muscular strength and the coordination of lower limb movement. In this chapter, we will discuss the prerequisites of good long-term muscle function and how these are compromised in the individual with spastic CP. We will consider the failure of muscle development in children with spastic CP as potential significant factor in the decline of these individuals as they pass into adulthood. Muscle is a complex tissue. Each muscle is an organized composite of individual muscle cells (or fibers) bound together within an extracellular matrix formed from different forms of collagen and proteoglycans. Within each muscle fiber are structures which enable its activation by the repeated release and sequestration of calcium into the cytoplasm (the sarcolemma or cell membrane, the t-tubule system, and the sarcoplasmic reticulum). Muscle cells have mitochondria which are responsible for much of the production of adenosine triphosphate (ATP), the currency of energy in the cell. However, the feature of muscle fibers that differentiate them from other cells are the long chains of force-producing units known as sarcomeres. This repeating subdivision of muscle fibers is the only unit in skeletal muscle able to actively develop significant movement and force. Sarcomeres contain parallel interacting macromolecules of myosin and actin. Molecular extensions of the myosin macromolecule, consisting of a long neck and a globular head, form bonds (crossbridges) with regularly spaced sites on the actin long-chain molecule. Relative movement of myosin and actin (and contraction of the sarcomere) occurs on the interaction of available ATP with the myosin neck causing a transformation of chemical into mechanical energy and change in conformation of the myosin neck. Sequestration of calcium and the presence of adenosine diphosphate (ADP) and phosphorus into the sarcoplasm allows the myosin and actin to associate and the myosin head and neck to return to its resting state. Under unloaded conditions, a single cycle of excitation, activation, and contraction should lead to a shortening of the sarcomere of 430 angstroms (the distance between neighboring active sites on the actin molecules in the longitudinal direction of the muscle fiber). A detailed description of the mechanism of contraction of the sarcomere is not warranted in this chapter but the development of active force by an individual sarcomere is proportionate to the number of crossbridges that can be formed which is dependent on the overlap between the actin and myosin long-chain molecules and the relative speed of movement of these molecules. This dependency of force on length and velocity is characteristic of the sarcomere and, by consequence, of the whole muscle. The synchronized contraction of many sarcomeres contributes to the shortening of individual muscle fibers and of the muscle as a whole. Via tendons linking muscles and bones, muscular contractions lead to joint rotation and joint torques.

Skeletal Muscle Structure in Spastic Cerebral Palsy

1077

State of the Art Skeletal muscle has many functions including the regulation of blood sugars and body temperature. It also has a role as a store of energy being able to be broken down in times of famine. But, the key distinguishing feature of muscle is the development of force and movement. To generate the appropriate levels of force over the lifespan, muscles must have the following qualities (or prerequisites). • Functional flexibility: Ability to produce appropriate forces and contractions speeds for a variety of physical tasks. Collectively, muscles must support the body while also being able to rapidly change the configuration of the skeleton. • Structural integrity: Muscles must be able to resist the large forces to which they are exposed to without frequent failure. • Adaptability secondary to changing demands due to exercise and disuse. • Capacity to self-repair or grow rapidly during childhood. In the following sections, we will discuss these ideal muscle properties and see how they may be altered in individuals with spastic CP. The first prerequisite of muscle function is that muscles must be capable of generating forces and velocities for a variety of physical tasks. For successful movement, the nervous system must present the skeletal musculature with a series of activation sequences and sense the resultant joint rotations and joint forces for modifying or correcting the movement. Further, the activated musculature should be capable of developing the forces and velocities necessary to execute the “planned” movements. Both intracellular properties and the gross morphological features of muscle contribute to the development of muscular force and speed. Muscles rely on energy sources for contraction. The speed at which energy sources can be broken down to transform ADP (adenosine diphospate) into ATP and for that ATP to be delivered to the site of the crossbridges and utilized for the transformation of chemical to mechanical energy determines the velocity of contraction and to some extent the specific tension within a muscle (the active force a muscle can produce per unit area). There is certainly a trade-off between muscle speed and endurance. In slow muscle (muscle containing predominantly slow phenotypes of myosin and large numbers of mitochondria), where the conversion of ADP to ATP for contraction largely takes place within the mitochondria and the delivery of ATP to crossbridges occurs at a lower rate, muscular forces develop slowly but may be maintained for long periods. In contrast, in faster muscle, a significant proportion of the reformation of ATP available for muscular contraction takes place in the cytoplasm. Here, large quantities of ATP are made available but only for short periods (maybe 10–30 s of contractions). Furthermore, there are variants (isoforms) of the myosin molecule in faster muscle which allows a more rapid crossbridge cycling and higher velocity contractions. Largely, each skeletal muscle in the lower limb consists of a mixture of slow and fast fiber type (with the honorable exception of the soleus muscle which has a predominance of slow muscle

1078

A. Shortland

A

C

FORCE B

B D

A

C 1.5 D

2.5

4.0 LENGTH (μm)

Fig. 1 The active force-length relationship of the sarcomere. The active force developed depends on the number of crossbridges that can be formed between actin (red thin strand) and the myosin (the thicker blue strand)

fibers). However, they are highly adaptive tissues in which fiber-type transformation can occur in response to changes in levels of activation and tension. For a more detailed description of the crossbridge cycle, please refer to a muscle physiology text. Although much of the attention in muscle studies is focused on the metabolic and molecular properties, gross morphology and architecture has a much stronger influence on muscle speed and strength than fiber type because they determine the serial and parallel arrangement of sarcomeres within a muscle (Lieber and Fridén 2000). To understand the influence of sarcomere arrangement on muscle dynamics, one must have an appreciation of the force generating capacity of the individual sarcomere. Fig. 1 depicts the active force-length relationship of an individual sarcomere. At longer lengths (~4.0 μ), there is little overlap of the actin and myosin molecules, fewer crossbridges are formed, and little force is produced. Similarly, at shorter lengths, the opportunity from crossbridge formation is limited and little active force is developed. The range over which a sarcomere may produce active forces is around 2.4 μ. At maximal overlap of the actin and myosin macromolecules, the greatest number of crossbridges are formed and the force developed is maximal. In normal muscle, we would expect the resting length of the sarcomere (the length adopted in the muscle’s passive state with no external forces applied) to be close to the position of optimal force (at about 2.4 μ). This would allow muscles to produce their maximum forces quickly from rest. The potential significance of an altered sarcomere resting length will become clear when we discuss skeletal muscle in spastic CP. An individual sarcomere develops little force and acts over a very limited range. However, when linked together, they have the capacity to produce the magnitude of forces and speeds that we require to maintain our posture against gravity and to move our joints through significant ranges of motion. Imagine two small (2 dimensional) muscles, each created from just 12 sarcomeres (Fig. 2).

Skeletal Muscle Structure in Spastic Cerebral Palsy

1079

Muscle A

FORCE

Muscle A Muscle B

Muscle B LENGTH

Fig. 2 The effect of sarcomere arrangement on muscle force development

Each sarcomere can produce the same peak forces, has the same active range, and has the same maximal velocity of contraction. In muscle A, the sarcomeres are arranged with 4 in parallel and 3 in series while in muscle B there are 3 sarcomeres in parallel and 4 in series. Even though these two hypothetical muscles have the same number of sarcomeres, they behave very differently. Muscle A can produce 4/3 times the peak force of muscle B, while muscle B has 4/3 the active range and 4/3 the speed of muscle A. In brief, the number of sarcomeres acting in parallel within a muscle determines its peak active force while the number of sarcomeres acting in series determines its active range and speed. If we scale up these tiny muscles to something more approaching the sizes of those of the human body, we can appreciate that the peak active force developed by a muscle is proportionate to the area of the muscle offered in the direction of its external tendon and the range and velocity of the muscle is proportionate to its fiber type. Of course, these relationships assume that the area of the muscle is largely composed of sarcomeric (or myofibrilar) tissue and that the force length relationship of the sarcomere is homogenous throughout the muscle. There are few muscles in the human body that have the linear appearance of muscles A or B. Most muscles have a pennate structure in which muscle fibers make an angle with the tendons with which they are joined. This arrangement is thought to allow the efficient packing of sarcomeres to optimize force production and active range. Fukanaga and his colleagues (Fukunaga et al. 1992) used a simple mathematical expression (Eq. 1) in units of area which is proportionate to the number of sarcomeres acting in parallel in a pennate muscle in the direction of the tendon which

1080

A. Shortland

they termed the physiological cross-sectional area (PCSA). They were able to show that for individual muscle PCSA is a strong linear predictor of peak muscular force. Physiological Cross-Sectional Area PCSA ¼

V  cos θ lf

(1)

where V is the muscle volume, θ is the angle of pennation and lf is the length of the muscle fibres. Some interesting studies of the morphology and architecture of the lower limbs in typically developing adult subjects have been conducted. Notably, and most recently, Ward and coworkers (2009) analyzed gross muscle structure from a large number of cadavers. It is clear that function follows form in the human musculature with antigravity muscles (such as the soleus) having large PCSAs and relatively short muscle fibers while the flexors of the limb (such as the ankle dorsiflexors) have longer fiber lengths but smaller PCSAs. A limited number of studies have been conducted of the metabolic phenotype and cross-sectional shape and size of muscle fibers in spastic CP. Ito et al. found fibers to be predominantly slow in the lower limb muscles of ambulant individuals with increased variation in fiber cross-sectional area (Rose et al. 1994; Araki et al. 1996). The largest number of samples reported in a single study were taken by Castle and colleagues (1979). These were taken from multiple muscles in the lower limb and showed a large heterogeneity in muscle fiber type, increased variation in fiber diameter. It should be stated that these studies were poorly controlled (it is difficult to find samples from representative muscles in control subjects) or that the studies were of limited sample sizes, and sometimes, did not fully take into account the clinical presentation of the individuals from which the samples were taken, and that the samples were taken from upper and lower limbs. More recently, in samples of tissue taken from the semitendinosis of children with CP and those from children with ACL injury about to have surgeries suggested a predominance of fibers with a slower phenotype and a reduced cross-sectional area (Zogby et al. 2016). Gantelius and colleagues (2012) found muscles in the forearms of children with unilateral CP to have a greater expression of myosin heavy chain IIx, consistent with a faster fiber phenotype These studies collectively show alterations in muscles at the level of the fiber in CP that are not consistent with simple model of chronic over-activity (where a uniform slow phenotype may emerge) or a model of disuse (where we may expect faster smaller fibers to be present) (Foran et al. 2005). In contrast to the variable results reported for muscle fiber type, a consistent feature of spastic CP is reduction in lower limb muscle size. There is a lack of data for muscles of the upper limb, but the majority of muscles in the lower limb are affected with reductions in muscle volume (normalized to muscle mass) of between 20% and 50% reported (Noble et al. 2014a; Fry et al. 2007; Shortland 2009; Barber et al. 2011; Handsfield et al. 2016). These large reductions in muscle volume are likely to have a large effect on the functional capacity of an individual with

Skeletal Muscle Structure in Spastic Cerebral Palsy

1081

spastic CP. Musculoskeletal simulations suggest that muscle volume deficits of 40%-60% would severely compromise ambulatory function in children with unimpaired gait moderate crouch (Steele et al. 2012). A global reduction in muscle volume would be an interesting enough finding, however, in those studies, where a near complete set of lower limb muscle volumes have been measured (Noble et al. 2014a; Handsfield et al. 2016; Lampe et al. 2006) the distal musculature appear more affected while certain muscles such as the quadriceps appear relatively spared. This is an interesting observation from two perspectives. Firstly, an uneven distribution of muscle deficits contradicts the notion of a disuse model of muscle atrophy while greater distal involvement implies a neurological influence on muscle development (Gough and Shortland 2012). Herskind and colleagues recently demonstrated that gastrocnemius muscle volumes and cross-sectional areas were reduced in children below the age of 18 months implying that muscles of infants with CP are on a different developmental track with reduced rates of growth compared to their typically developing peers (Herskind et al. 2016). Noble et al. (in review) show similar results in a cross-sectional study of adolescents and young adults, reporting a disparity in rate of muscle growth in relation to the accrual of bodyweight. Secondly, the distal musculature is very important for standing and walking in typically developing subjects. Larger deficits in the distal musculature of individuals in CP may particularly compromise these activities. Greater muscle deficits in children with higher gross motor functional classification system (GMFCS) scores (Noble et al. in review) gives further support to the idea that reduction in muscle volume are related to ambulatory mobility. However, studies with greater numbers and a wider range of GMFCS levels are required to fully map out the associations between muscle volume and functional capacity. Further, longitudinal studies are required to elaborate any causal relationships between reduced muscle growth rates and trajectories of functional improvement and decline. There continues to be a dearth of studies of muscle volume in the upper limb. The active range of muscles depend on the fiber length of muscles and on the number of sarcomeres per unit fiber length. It is difficult to measure the length of fibers directly but B-mode ultrasound imaging allows the length of bundles of fibers (or fascicles) to be estimated (Fig. 3). It had been a long-held belief that the cause of muscular deformity (normally, manifesting itself as a limitation in passive joint range) is short muscle fibers. However, the results of imaging studies are rather equivocal (Barber et al. 2011; Malaiya et al. 2007; Khan et al. 2008; Mathewson et al. 2015) with some workers claiming no measureable difference of the length of fascicles between children with spastic CP and control subjects while others have measured small reductions in fascicle length in children with CP. There are few studies reporting measurements from muscles other than the plantarflexors, probably, because of the technical difficulties associated with imaging longer fascicle lengths with standard 2D ultrasound probes. A notable exception is from the work of Moreau et al. (2009) in which the researchers measured fascicle lengths and fascicle angles in the quadriceps of children with CP. Shortness of the muscle fibers, in itself, now seems unlikely to be a significant contributor to measured passive muscle stiffness. Recently, Smith et al. (2011)

1082

A. Shortland

Fig. 3 Composite longitudinal ultrasound image of the tibialis anterior in a typically developing adult subject. As indicated by the red lines, fascicle length and fascicle angle may be measured directly from the ultrasound images

demonstrated that the increased stiffness of hamstring muscles in children with CP having surgery to lengthen these muscles had increased quantities of extracellular matrix (ECM) and that the increased Young’s modulus associated with this extra material could account for increased stiffness of fascicular bundles of tissue. Perhaps, the more extraordinary finding from this paper was that the authors found resting sarcomere lengths that were longer than in control subjects (children having ACL repairs). Certainly, sarcomeres being close to the limits of their extension (at about 4 μ) would contribute to the passive resistance to stretch but the implications of this finding may be more wide-reaching, affecting the active properties of the muscle. If the resting length of the sarcomere is increased, then the overlap between the sarcomeric macromolecules actin and myosin would be reduced with a consequent reduction in the number of crossbridges formed and force produced. In order to reach optimal sarcomere lengths to produce maximal forces, each sarcomere would need to contract by more than a micron (Fig. 4). If the sarcomere contracted by the distance between each active site on the actin molecule per activation cycle, then it would take the sarcomere about 20 activation cycles to go from its resting state to its optimal state. Under those conditions, the muscle would generate its maximum force from rest at more than half a second after it was initially activated. Downing and colleagues (2009) demonstrated that the time to maximal activation was more than 0.5 s in a group of children with CP, consistently twice as slow as their able-bodied peers. The long resting length of sarcomeres may also have implications for the joint position at which optimal joint torques are developed. Matthewson et al. (2015) estimated the number of sarcomeres in series in the soleus muscle. She found resting sarcomere lengths of around 4 μ in children about to undergo surgery in contrast to lengths of 2.17 μ in typically developing children. With similar fascicle lengths in both groups as measured by ultrasound, one can estimate the number of sarcomeres in series (about 10,000). If each sarcomere has to contract by 1.5 μm to operate at its optimal force, then the soleus would need to contract by 1.5 cm from its resting length. If we assume a moment arm of the Achilles tendon at the ankle of 4 cm, then the ankle joint would need to plantarflex by 21 degrees. Could the disparity between the active and passive force length properties of the sarcomere in children with spastic CP explain phenomena such as

Skeletal Muscle Structure in Spastic Cerebral Palsy Fig. 4 A schematic of the resting length of sarcomeres in typically developing subjects and subjects with spastic cerebral palsy. The greater resting length in subjects with CP suggests a reduced overlap between the sarcomeric proteins of actin and myosin implying a greater time period for the muscle to reach its optimal (peak) force from a resting position

1083

3.6m

CP

TD

2.6mm dynamic equinus or crouch gait? At the moment, the extraordinary finding of “overstretched” sarcomeres in the lower limb is limited to a small number of muscles and has been made by a single, albeit, expert group. Certainly, more studies are required if we are to confirm a relationship between overstretched sarcomeres and dynamic shortness in these children. One interesting coda to the story of the overstretched sarcomere is what happens to muscle fascicle length after surgery. Shortland et al. (2004) in the medial gastrocnemius demonstrated reduction in fascicle length after surgery. Such a change could be attributed to sarcomere loss, or it could be that sarcomere length becomes normalized after surgery by a mechanism that relieves passive tension in the muscle. At the moment, we do not know the answer, but the observation poses the question about how normal sarcomere function may be restored. While a number of biopsy studies have found increased ECM in muscles of the upper and lower limbs of children with spastic CP, fewer studies have investigated changes in muscle composition at the level of the whole muscle (Johnson et al. 2009; Noble et al. 2014b). These studies show increased level of subcutaneous, intermuscular and intramuscular fat and raised levels of connective tissue in individuals with bilateral CP. The effect of a raised fraction of nonmyofibrillar tissue would tend to decrease the active specific tension within a muscle and by implication the force developing capacity of that muscle. Our second prerequisite is that muscle should have mechanical integrity. Muscle is a complex composite material. Muscle fibers consist of multiple parallel myofibrils (long chains of sarcomeres). Muscles themselves consist of fibers positioned (largely) in parallel with each other. If these structures produced forces independently or were exposed individually to externally imposed tensile forces, then the specific tensions generated would be highly destructive. Instead, forces within muscles are shared across the muscular components by molecular linkages within and between muscle fibers.

1084

A. Shortland

Fig. 5 (a) The myosin filament is linked to the z-disc of the sarcomere by the large macromolecule titin, providing integrity for sarcomeres in series. (b) Sarcomeres are bound together in parallel at the z-disc by molecules such as α-actinin. (c) Muscle fibers are connected together through a complex of molecules including desmin and dystrophin which join the z-discs with the extracellular matrix

At the level of the sarcomere, there are long chain proteins that prevent dissociation of actin and myosin. Of specific interest is titin. This enormous molecule connects myosin to the z-disc (the stiff network of actin-like molecules that define the ends of each sarcomere. Titin has an elastic subcomponent which allows the molecule to stretch (presumably to allow different degrees of overlap between actin and myosin). This molecule becomes stiff at long sarcomere lengths and therefore is responsible for maintaining serial integrity between sarcomeres along the length of the myofibril (Fig. 5a). But how is mechanical integrity maintained in parallel? The thin filament, actin is bound to the z-disc by the dimer α-actinin. This binding helps to form a 3D stiff structure among neighboring sarcomeres and thus parallel integrity within the muscle fiber is maintained (see Fig. 5b). Mechanical connections between neighboring fibers are maintained though a complex of proteins collectively known as costameres (Fig. 5c) which attach to the z-discs of peripheral sarcomeres within the fiber to sites in the ECM. The combination of these mechanical proteins enables lateral force transmission between neighboring sarcomeres and neighboring muscle fibers allowing the forces generated by the activation of individual muscle fibers to be distributed across the muscle, avoiding local raised stress concentrations. Further, these molecules probably help to regulate sarcomere length across the muscle. There are no reports of significant alterations in these “mechanical” proteins in spastic CP. Indeed, Smith et al. found that titin was not changed and did not contribute to the increased mechanical stiffness of fascicular bundles (bundles of 20 or more fibers) taken from the hamstrings of children with spastic CP just prior to surgery to lengthen these muscles. However, the authors did note that the increased stiffness was associated with increased levels of collagen with the ECM of the bundle. This finding suggests that internal forces are generated between the ECM and the muscle fibers that causes overstretching of sarcomeres. This could be related to differential distances between costameric bindings to the ECM and costameric attachments at the z-discs of the sarcomeres. Our third perquisite for muscle function is adaptability. Muscle is, in fact, a highly adaptable biological material. Cell density is high within muscle and protein turnover is great. This makes muscle a plastic material, responsive to changes in demand.

Skeletal Muscle Structure in Spastic Cerebral Palsy

1085

The rates of sarcomeric protein generation and breakdown are regulated by systemic factors such as myostatin, tumor necrosis factor, and insulin-like growth factor and by additional transcriptional factors that reside within the cell. Until recently, it was not known precisely how mechanical or electrical stimuli were translated into muscle growth. There is an excellent review of muscle adaptability by Braun and Gautel (2011) and the reader is referred here for a deeper understanding. Briefly, transcriptional factors associated with the sarcomere are mobilized by stretching of the sarcomere and these are transported to the nucleoplasm where they become linked with a particular gene associated with the production of an associated protein. Similarly, on activation of a muscle fiber, calcium ions are released which provokes the calcineurin pathway and stimulates production (upregulates) trophic pathways. So, when a muscle is stretched or activated there are molecular signals which stimulate trophic pathways and building of muscle. Equally, there are pathways that depress (downregulate) protein production and cause atrophy. Therefore, muscle hypertrophy and atrophy are directly related to muscle use and disuse. In spastic CP, many of the pathways promoting the building of muscle are depressed and many of the pathways that causes atrophy are activated (Smith et al. 2012). It is likely that the downregulation of these trophic pathways is directly related to the reductions in the rates of muscle growth in spastic CP that are observed macroscopically. Our fourth prerequisite for muscle function concerns the capacity for muscles to repair themselves and to grow rapidly to match growth of the body. Muscles produce large forces when active and necessarily cause large stresses within their tissues. This can lead to cell damage and potential cell death. Muscles have additional cells that during early development were not differentiated to myoblasts (the precursor cells of muscle fibers). These “satellite” cells reside close to muscle fibers (between the cell membrane and the basal lamina). When a muscle cell is damaged, satellite cells proliferate with some differentiating to form myoblasts which fuse with the damage cell and donate their nuclei, while others are dedicated to maintaining the satellite cell population (for review, see (Dayanidhi and Lieber 2014)). The addition of nuclei to an already multinucleated cell allows the cell to recover and grow. In the typically developing mature subject, about 5% of muscle nuclei reside in satellite cells. The capacity of satellite cells to activate or maintain a quiescent state depends upon the signaling milieu within which they exist – the satellite cell niche – thus rates of self-renewal and of functional differentiation may be affected by the properties of tissue in their immediate vicinity including those of the ECM. Conversely, satellite cells modify the ECM and local fibroblast activity. In typically developing infants, the nuclei of satellite cells are thought to make up about 30% of the total muscle nuclei. This high density of stem cells supports an extraordinary rate of muscle growth in early life as these cells donate their nuclei to the developing muscle fibers. Lieber and colleagues found large deficits in the number of satellite cells in the hamstrings of children with spastic cerebral palsy just prior to surgery (Smith et al. 2013). The low number of muscle stem cells may explain the reduced rates of longitudinal and cross-sectional growth in spastic CP31.The reasons for the reduced

1086

A. Shortland

population of SCs are unclear but it may be related to the extensive ECM that is present in these children and also of the ECM’s altered mechanical properties.

Conclusions and Future Directions In summary, the force developing capacity and the speed of muscles in the lower limbs of children with spastic CP are compromised by multiple aberrant structural features including reduced muscle volume (and physiological cross section), reduced serial sarcomere number, altered sarcomere length, and increased nonmyobrillar content. These abnormalities in microscopic and macroscopic properties may be related to a depressed population of satellite cells, those cells that potentiate growth and repair. While researchers and clinicians are getting closer to a fuller understanding of nature of muscular contracture in spastic CP, the origins of these problems are unknown. It is likely, that the developmental trajectory of muscle in affected children is set very early in life when the innervation of skeletal muscle is being organized and refined. Certainly, we understand that that it takes some time for the ratio of myofibrillar tissue to ECM to increase during postnatal growth and that this is related to the sustained activation of the musculature (Eken et al. 2008). In children with spastic CP, there is an injury to the developing brain that compromises the development of the descending neural tracts, particularly of the cortico-spinal tract (Clowry 2007). Among other things, these tracts are responsible for the lowering of the excitation thresholds of alpha motor neurons and the increase in efferent output from the spinal cord to the developing musculature (Lemon 2008). If efferent output from the spinal cord to the musculature in young children with spastic CP is diminished, then it is probable that the development of the myofibrillar fraction is compromised and the ECM is not refined during the first 1 or 2 years of childhood. The maldevelopment of the ECM may well then affect the regulation of sarcomere length and the proliferation and activation of satellite cells. Subsequent muscle growth during childhood would be reduced and the relationship between the active and passive properties of muscles deranged. As yet there is very little in the literature documenting early muscle development in those children at risk of CP18. Certainly, there are significant ethical and logistical problems with sampling muscle tissue from infants. However, there is an opportunity to use simple imaging methods such as echogenicity from B-mode ultrasound to estimate the nonmyofibrillar fraction in muscles during development (Pillen et al. 2003).

Implications for Treatment The altered trajectory of muscle development in spastic CP has important implications for mobility across the lifespan. Shortland (2009) proposed a conceptual framework for understanding the decline of mobility in the affected young adult. In brief, he suggested that due to the reduced rate of accretion of muscle mass, the acquisition of important functional milestones (such as standing walking, getting up

Skeletal Muscle Structure in Spastic Cerebral Palsy

1087

from a chair) would be delayed. Furthermore, he thought that the reduced rates of muscle growth would lead to a reduction in the muscular reserve of an individual (i.e., the muscular forces that the individual could develop above the threshold required to perform the activity). This would result in increased levels of fatigue for the individual performing that activity and reduce the number of years that the individual would be able to perform that activity as they become exposed to the deleterious effects of aging on muscle properties. Clinical management, therefore, should be directed at developing and maintaining a muscular reserve and avoiding interventions that result in long-term atrophy. Studies of resistance training (McNee et al. 2009) and of muscle stimulation suggest that muscle mass can be developed in short-term programs. However, long-term muscle hypertrophy in this group may be limited by the reduced population of satellite cells available to support growth and repair. Increasing the potential for growth and self-repair by cellular and molecular interventions is the holy grail of muscle management in spastic CP, but at the moment, we lack a precise understanding of the mechanism of the development of muscular deformity that might be challenged by early molecular intervention.

Cross-References ▶ 3D Musculoskeletal Kinematics Using Dynamic MRI ▶ Cross-Platform Comparison of Imaging Technologies for Measuring Musculoskeletal Motion ▶ EMG Activity in Gait: The Influence of Motor Disorders ▶ Optimal Control Modeling of Human Movement ▶ Spasticity Effect in Cerebral Palsy Gait ▶ Ultrasound Technology for Examining the Mechanics of the Muscle, Tendon, and Ligament

References Barber L, Hastings-Ison T, Baker R, Barrett R, Lichtwark G (2011) Medial gastrocnemius muscle volume and fascicle length in children aged 2 to 5 years with cerebral palsy. Dev Med Child Neurol 53(6):543–548. https://doi.org/10.1111/j.1469-8749.2011.03913.x Braun T, Gautel M (2011) Transcriptional mechanisms regulating skeletal muscle differentiation, growth and homeostasis. Nat Rev Mol Cell Biol 12(6):349–361. https://doi.org/10.1038/ nrm3118 Castle ME, Reyman TA, Schneider M (1979) Pathology of spastic muscle in cerebral palsy. Clin Orthop Relat Res 142:223–232. http://www.ncbi.nlm.nih.gov/pubmed/159152. Accessed 1 Apr 2012 Clowry GJ (2007) The dependence of spinal cord development on corticospinal input and its significance in understanding and treating spastic cerebral palsy. Neurosci Biobehav Rev 31(8):1114–1124. https://doi.org/10.1016/j.neubiorev.2007.04.007 Dayanidhi S, Lieber RL (2014) Skeletal muscle satellite cells: mediators of muscle growth during development and implications for developmental disorders. Muscle Nerve 50(5):723–732. https://doi.org/10.1002/mus.24441

1088

A. Shortland

Downing AL, Ganley KJ, Fay DR, Abbas JJ (2009) Temporal characteristics of lower extremity moment generation in children with cerebral palsy. Muscle Nerve 39(6):800–809. https://doi. org/10.1002/mus.21231 Eken T, Elder GCB, Lømo T (2008) Development of tonic firing behavior in rat soleus muscle. J Neurophysiol 99(4):1899–1905. https://doi.org/10.1152/jn.00834.2007 Foran JRH, Steinman S, Barash I, Chambers HG, Lieber RL (2005) Structural and mechanical alterations in spastic skeletal muscle. Dev Med Child Neurol 47(10):713–717. https://doi.org/ 10.1017/S0012162205001465 Fry NR, Gough M, McNee AE, Shortland AP (2007) Changes in the volume and length of the medial gastrocnemius after surgical recession in children with spastic diplegic cerebral palsy. J Pediatr Orthop 27(7):769–774. https://doi.org/10.1097/BPO.0b013e3181558943 Fukunaga T, Roy RR, Shellock FG et al (1992) Physiological cross-sectional area of human leg muscles based on magnetic resonance imaging. J Orthop Res 10(6):928–934. https://doi.org/ 10.1002/jor.1100100623 Gantelius S, Hedström Y, Pontén E (2012) Higher expression of myosin heavy chain IIx in wrist flexors in cerebral palsy. Clin Orthop Relat Res. https://doi.org/10.1007/s11999-011-2035-3 Gough M, Shortland AP (2012) Could muscle deformity in children with spastic cerebral palsy be related to an impairment of muscle growth and altered adaptation? Dev Med Child Neurol 54(6):495–499. https://doi.org/10.1111/j.1469-8749.2012.04229.x Handsfield GG, Meyer CH, Abel MF, Blemker SS (2016) Heterogeneity of muscle sizes in the lower limbs of children with cerebral palsy. Muscle Nerve 53(6):933–945. https://doi.org/ 10.1002/mus.24972 Herskind A, Ritterband-Rosenbaum A, Willerslev-Olsen M et al (2016) Muscle growth is reduced in 15-month-old children with cerebral palsy. Dev Med Child Neurol 58(5):485–491 Ito J, Araki A, Tanaka H, Tasaki T, Cho K, Yamazaki R (1996) Muscle histopathology in spastic cerebral palsy. Brain and Development. https://doi.org/10.1016/0387-7604(96)00006-X Johnson DL, Miller F, Subramanian P, Modlesky CM (2009) Adipose tissue infiltration of skeletal muscle in children with cerebral palsy. J Pediatr 154(5):715–720. https://doi.org/10.1016/j. jpeds.2008.10.046 Mohagheghi AA, Khan T, Meadows TH, Giannikas K, Baltzopoulos V, Maganaris CN (2008) In vivo gastrocnemius muscle fascicle length in children with and without diplegic cerebral palsy. Dev Med Child Neurol 50(1):44–50. https://doi.org/10.1111/j.1469-8749.2007.02008.x Lampe R, Grassl S, Mitternacht J, Gerdesmeyer L, Gradinger R (2006) MRT-measurements of muscle volumes of the lower extremities of youths with spastic hemiplegia caused by cerebral palsy. Brain and Development 28(8):500–506. https://doi.org/10.1016/j.braindev.2006.02.009 Lemon RN (2008) Descending pathways in motor control. Annu Rev Neurosci 31:195–218. https:// doi.org/10.1146/annurev.neuro.31.060407.125547 Lieber RL, Fridén J (2000) Functional and clinical significance of skeletal muscle architecture. Muscle Nerve 23:1647–1666. https://doi.org/10.1002/1097-4598(200011)23:113.0.CO;2-M. [pii] Malaiya R, McNee AE, Fry NR, Eve LC, Gough M, Shortland AP (2007) The morphology of the medial gastrocnemius in typically developing children and children with spastic hemiplegic cerebral palsy. J Electromyogr Kinesiol 17(6):657–663. https://doi.org/10.1016/j.jelekin.2007. 02.009 Mathewson MA, Ward SR, Chambers HG, Lieber RL (2015) High resolution muscle measurements provide insights into equinus contractures in patients with cerebral palsy. J Orthop Res. https:// doi.org/10.1002/jor.22728 McNee AE, Gough M, Morrissey MC, Shortland AP (2009) Increases in muscle volume after plantarflexor strength training in children with spastic cerebral palsy. Dev Med Child Neurol 51(6):429–435. https://doi.org/10.1111/j.1469-8749.2008.03230.x Moreau NG, Teefey SA, Damiano DL (2009) In vivo muscle architecture and size of the rectus femoris and vastus lateralis in children and adolescents with cerebral palsy. Dev Med Child Neurol 51(10):800–806

Skeletal Muscle Structure in Spastic Cerebral Palsy

1089

Noble JJ, Chruscikowski E, Fry NR, Lewis AP, Gough M, Shortland AP. Reduced lower limb muscle growth in relation to body mass in a cross-sectional study of ambulant individuals with bilateral cerebral palsy aged 10 to 23. Article in review Noble JJ, Fry NR, Lewis AP, Keevil SF, Gough M, Shortland AP (2014a) Lower limb muscle volumes in bilateral spastic cerebral palsy. Brain and Development 36:294–300. https://doi.org/ 10.1016/j.braindev.2013.05.008 Noble JJ, Charles-Edwards GD, Keevil SF, Lewis AP, Gough M, Shortland AP (2014b) Intramuscular fat in ambulant young adults with bilateral spastic cerebral palsy. BMC Musculoskelet Disord 15:236. https://doi.org/10.1186/1471-2474-15-236 Pillen S, Scholten RR, Zwarts MJ (2003) Verrips a. Quantitative skeletal muscle ultrasonography in children with suspected neuromuscular disease. Muscle Nerve 27(6):699–705. https://doi.org/ 10.1002/mus.10385 Rose J, Haskell WL, Gamble JG, Hamilton RL, Brown DA, Rinsky L (1994) Muscle pathology and clinical measures of disability in children with cerebral palsy. J Orthop Res 12(6):758–768. https://doi.org/10.1002/jor.1100120603 Shortland A (2009) Muscle deficits in cerebral palsy and early loss of mobility: can we learn something from our elders? Dev Med Child Neurol 51(Suppl 4):59–63. https://doi.org/10.1111/ j.1469-8749.2009.03434.x Shortland AP, Fry NR, Eve LC, Gough M (2004) Changes to the muscle architecture of the medial gastrocnemius after surgical intervention in spastic diplegia. Dev Med Child Neurol 46:667–673 Smith LR, Lee KS, Ward SR, Chambers HG, Lieber RL (2011) Hamstring contractures in children with spastic cerebral palsy result from a stiffer extracellular matrix and increased in vivo sarcomere length. J Physiol 589(Pt 10):2625–2639. https://doi.org/10.1113/jphysiol.2010. 203364 Smith LR, Chambers HG, Subramaniam S, Lieber RL (2012) Transcriptional abnormalities of hamstring muscle contractures in children with cerebral palsy. PLoS One 7(8):e40686. https:// doi.org/10.1371/journal.pone.0040686 Smith LR, Chambers HG, Lieber RL (2013) Reduced satellite cell population may lead to contractures in children with cerebral palsy. Dev Med Child Neurol 55:264–270. https://doi. org/10.1111/dmcn.12027 Steele KM, van der Krogt MM, Schwartz MH, Delp SL (2012) How much muscle strength is required to walk in a crouch gait? J Biomech 45(15):2564–2569. https://doi.org/10.1016/j. jbiomech.2012.07.028 Ward SR, Eng CM, Smallwood LH, Lieber RL (2009) Are current measurements of lower extremity muscle architecture accurate? Clin Orthop Relat Res 467:1074–1082. https://doi. org/10.1007/s11999-008-0594-8 Zogby AM, Dayanidhi S, Chambers HG, Schenk S, Lieber RL (2016) Skeletal muscle fiber-type specific succinate dehydrogenase activity in cerebral palsy. Muscle Nerve. https://doi.org/ 10.1002/mus.25379

Part XIV Movement Deviations in Cerebral Palsy

Swing Phase Problems in Cerebral Palsy Ana Presedo

Abstract

The normal human gait cycle is commonly divided in two phases: stance and swing. The objective of stance phase is to provide support, stability, and propulsion and contribute to the advancement of the limb in swing. Ground clearance and appropriate pre-positioning of the foot are prerequisites of normal gait during swing. Normal characteristics of stance and swing phases ensure adequate step length and energy conservation during walk. Swing phase problems in cerebral palsy tend to be related to spasticity, abnormal muscular control, poor opposite limb balance, and lack of power. From a clinical point of view, knee stiffness and ankle insufficient dorsiflexion constitute the two major abnormalities in the sagittal plane, whereas in the frontal and transverse planes, excessive hip adduction and foot deviations can cause problems with limb advancement. Rectus femoris (RF) abnormal activity has been proposed as the main cause of knee stiffness in patients with cerebral palsy. Spasticity and/or contractures of the hip adductors and tibialis posterior muscles can cause abnormal hip and ankle kinematics and can lead to foot deformities if left untreated. Indications for treatment should be based on patient’s age and functional status. Different therapeutic approaches are described and commented in this chapter. A review of the literature based on the major articles in this field is provided. Keywords

Swing phase • Stiff-knee gait • Rectus femoris release • Split anterior tibial tendon transfer

A. Presedo (*) Pediatric Orthopaedics Department, Robert Debré University Hospital, Paris, France e-mail: [email protected]; [email protected] # Springer International Publishing AG, part of Springer Nature 2018 B. Müller, S.I. Wolf (eds.), Handbook of Human Motion, https://doi.org/10.1007/978-3-319-14418-4_53

1093

1094

A. Presedo

Contents Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . State of the Art . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Swing Phase Abnormal Kinematic Patterns and Causes of Pathological Function . . . . . . . . . . . Hip and Pelvis Pathology . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Knee Pathology . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Ankle and Foot Pathology . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Management of Swing Phase Problems in Cerebral Palsy . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Excessive Hip Adduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Stiff-Knee Gait . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Insufficient Ankle Dorsiflexion and Foot Varus Deviation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

1094 1095 1096 1096 1098 1101 1103 1103 1103 1105 1106 1107

Introduction Swing phase represents about 40% of the gait cycle. The purposes of swing are to advance the limb, to provide foot clearance, to allow variation in cadence, and to participate to the energy conservation mechanisms. The swing phase is commonly subdivided in three periods: initial swing, mid-swing, and terminal swing. The objective of the first period is to accelerate the limb, and this function is dependent on muscular work occurring during terminal stance and pre-swing, the second period represents a transition period, and the third period’s functional goal is to decelerate the limb to prepare it for initial contact. At normal self-selected walking speed, the limb swings with little muscle intervention; only the tibialis anterior is active throughout swing (Piazza and Delp 1996). In terminal swing, hamstring contraction generates simultaneous hip extension and knee flexion. The action of hip extension contributes to level the pelvis and to maintain trunk alignment, whereas the knee flexion moment decelerates the shank and controls knee extension at initial contact. The main role of the rectus femoris (RF) muscle in normal gait is to act as a hip flexor to propel the limb forward into swing (Perry 1992). Using fire-wire electrodes, two bursts of activity have been described (Annaswamy et al. 1999). The first burst occurs during the loading response phase of gait where the RF acts along with the vastii, and the second burst occurs during pre- and initial swing. During fast walking, the RF contraction helps to accelerate the shank in initial swing, and the hamstrings act to decelerate it in terminal swing (Nene et al. 1999). Swing phase important kinematic events occur mainly in the sagittal plane. During initial swing, the peak of knee flexion (PKF), which accounts for the maximum degree of knee flexion during the gait cycle, compensates for ankle plantar flexion and helps with foot clearance (Anderson et al. 2004). Since the lower limb acts as a compound pendulum, hip and knee flexion occur with very little muscular intervention during normal swing phase (Piazza and Delp 1996). In mid-swing, this mechanism generates knee extension, whereas the hip keeps going into flexion. During terminal swing, hip flexion reaches its maximum and the knee is almost extended before initial contact. The only active muscle throughout swing phase is

Swing Phase Problems in Cerebral Palsy

1095

the tibialis anterior that brings the ankle into neutral position acting against gravity. The actions of hamstrings and tibialis anterior stabilize the knee and ankle positions for landing. In the transverse and coronal planes, pelvic internal rotation and hip adduction in terminal swing provide adequate step length and proper foot orientation at heel contact. Hip abduction is observed during initial swing and helps with limb advancement. Abnormal kinematics during swing are mainly represented by insufficient or delayed peak of knee flexion (stiff-knee gait) and excessive ankle plantar flexion. Transverse plane abnormal kinematics are frequently seen in patients with hemiplegia and tend to be related to excessive pelvic rotation and foot deviations.

State of the Art Swing phase problems in cerebral palsy (CP) are mainly represented by limited knee flexion (stiff knee) and insufficient ankle dorsiflexion. An insufficient or delayed PKF and similarly excessive ankle plantar flexion during swing compromise foot clearance and adequate foot position at initial contact. Although terminal stance kinematics and kinetics influence swing phase events, spasticity of the RF and poor distal motor control constitute the main causes of these gait deviations. Stiff-knee management is addressed to diminish RF spasticity or to treat RF contractures. In children with CP abnormal RF activity is almost always present. Rectus femoris distal tendon transfer to knee flexor muscles has been advocated to avoid the knee extensor moment generated by RF abnormal activity during swing and to create a knee flexion moment. Studies based on cine phase-contrast magnetic resonance imaging (MRI) showed that the RF is not converted to a knee flexor after its distal tendon is transferred to the posterior side of the knee, but its capacity for knee extension is diminished by the surgery. Also, the three-dimensional reconstruction (MRI) of RF muscle geometry after the transfer demonstrated that the muscle didn’t follow a straight course from its origin to its new insertion, but a sharp deviation was frequently seen (Asakawa et al. 2004). In addition, scar tissue between RF and the underlying muscles was visible and could be the cause of relapses after RF transfer. Long-term follow-up studies showed variable results regarding improvement of the magnitude of PKF, but improvements in timing of knee flexion and overall knee range of motion remained stable over time (Dreher et al. 2012; Thawrani et al. 2012). Distal RF tendon resection has been proposed as an alternative to the transfer. A short-term follow-up study showed that RF distal tendon resection was associated with improved knee range of motion and timing of PKF in swing, and the absolute values of PKF became normal for those patients who showed abnormal preoperative values (Presedo et al. 2012). Ankle excessive plantar flexion and dynamic varus deviations are commonly observed in patients with spastic hemiplegia (Gage 1991). Botulinum toxin and orthoses are currently used to diminish spastic plantar flexion and varus deviations and to improve foot clearance in patients with ankle dorsiflexor weakness. Surgical muscular-tendon lengthenings may be indicated to treat fixed contractures. Davids

1096

A. Presedo

et al. (2011) demonstrated an improvement in ankle dorsiflexion following ankle plantar flexor surgery in selected children with CP. They explained this by the fact that correction of a fixed constraining equinus contracture of the ankle plantar flexors may unmask preexisting ankle dorsiflexion function. Split anterior tibial tendon transfer, combined with gastrocnemius release and/or tibialis posterior lengthening, has been advocated to treat spastic equinovarus deformity. Hoffer et al. (1985) demonstrated that this procedure was reliable and effective to correct dynamic varus deviations in patients with CP, after a follow-up of 10 years. Other softtissue procedures like tibialis posterior transfer to the forefoot tend to be abandoned in patient with spasticity because of danger of overcorrection and resultant calcaneus-valgus deformity. Bony procedures (calcaneal osteotomies and triple arthrodesis) may be indicated to correct fixed deformities.

Swing Phase Abnormal Kinematic Patterns and Causes of Pathological Function Computerized gait analysis has become the gold standard for the evaluation of complex gait problems in patients with cerebral palsy (Narayanan 2007; Gough and Shortland 2008). Based on the kinematic study of lower limb motion, common abnormal patters for the individual joints have been described. The recognition of these patterns has conducted to the elaboration of gait classifications (Rodda and Graham 2001). Without any intervention, these patterns of gait are extremely predictable over time. Although therapy decisions are particular to each patient, the knowledge of these abnormal patterns should allow for a better comprehension of pathology and conduct to a more appropriate treatment.

Hip and Pelvis Pathology The hip and pelvis play a major role during gait, ensuring power generation and balance of the upper body segment. Hip extensor muscles provide 50% of power during normal walking (the other 50% coming from ankle plantar flexors), and in patients with CP, since motor control is somehow preserved around the hip, compensatory mechanisms occur mainly at this level (Winter 1991; Õunpuu 1994). During swing phase, pelvic internal rotation and hip adduction in terminal swing provide adequate step length and proper foot orientation at heel contact. Hip abduction is observed during initial swing and helps with limb advancement. Excessive hip flexion throughout the cycle is commonly seen in patients with severe involvement (GMFCS III–IV). The typical kinematic pattern called “doublebump pelvic pattern” (Fig. 1) is characterized by an increased anterior pelvic tilt in stance phase and a second one in swing phase (Gage 1991). This abnormal pattern is related to a poor dissociation between pelvic and hip motion. Anterior pelvic tilt peak occurs always at single-limb stance under the influence of spastic hip flexors and/or weak hip extensors. “Single-bump pelvic pattern” is seen in patients with

Swing Phase Problems in Cerebral Palsy

1097 Pelvic Tilt

30.0

Angle (degrees)

Ant

25.0

20.0

15.0

10.0

5.0

Post

0.0 −5.0

0

10

20

30

40

50

60

70

80

90

100

Angle (degrees)

Fig. 1 The green line represents normal pelvis kinematics in the sagittal plane. The blue and red lines represent abnormal “double-bump pattern” kinematics for the right and left lower limb, respectively

hemiplegia (Fig. 2). Anterior pelvic tilt peak occurs in terminal stance, only on the hemiplegic side, and pelvis returns to normal alignment in swing phase as the stretch of the hip flexors ends when the hip moves into flexion. Anterior pelvic tilt throughout the gait cycle is commonly seen in young children with spastic diplegia who walk with increased cadence to compensate for the lack of stability related to spasticity of ankle plantar flexors. In this case, the shank does not advance normally over the foot, and so the trunk and pelvis bend forward in order to move the center of mass along the line of progression. Hip and pelvic coronal and transverse plan abnormal patterns present often combined. In diplegic and hemiplegic patients with distal involvement, these patterns may represent coping responses. In hemiplegic patients, abnormal rotation of pelvis and hip are very common and may be related to asymmetric neurological involvement or to torsional bony deviations. Internal hip rotation on the hemiplegic side is compensated by external rotation of the ipsilateral hemipelvis (Fig. 3). Increased pelvic rotation can be a compensation for reduced sagittal plane motion, in order to increase step length. During swing, hip abduction or circumduction represents a typical compensation for reduced sagittal plane motion. As a consequence of increased pelvic rotation, coronal hip motion will also increase. Increased

1098

A. Presedo Pelvic Tilt

Ant

30.0

Angle (degrees)

20.0

10.0

Post

0.0

0

10

20

30

40

50

60

70

80

90

100

Normalised (percent)

Fig. 2 In a patient with right hemiplegia, a “single-bump” pattern of pelvis kinematics is typically observed

coronal pelvic motion may be seen to compensate for a reduced hip abductor moment in stance (hip abductor insufficiency). In this case, the patient will lean the trunk laterally to shift the body center of mass and bring the GRF vector close to the hip center of rotation. Kinematic curves will show ipsilateral pelvic depression in stance and elevation in swing with decreased hip adduction in mid-stance and increased hip adduction in terminal stance and initial swing. In quadriplegic and diplegic patients with more severe involvement, hip dysfunction is commonly related to excessive femoral anteversion and adductor muscle contractures.

Knee Pathology The peak of knee flexion is generated by the action of ankle plantar flexors in terminal stance and by the hip flexors in pre-swing and initial swing, so the lower limb acts as a compound pendulum (Anderson et al. 2004). During initial swing, additional knee flexion is performed by the short head of biceps femoris to facilitate foot clearance (Perry 1992). The magnitude of the motion and the timing are critical to achieve the necessary knee flexion to clear the foot. The isometric contractions of RF and hamstrings during swing phase act to regulate knee flexion/extension in order to adjust to changes in walking speed (Nene et al. 1999). Excessive knee flexion pattern is characterized by greater than normal knee flexion throughout the stance phase. Increased knee flexion is usually maintained

Swing Phase Problems in Cerebral Palsy

Add

a

1099 Ab/Adduction Hanche

30

20

Angle (degrees)

10

0

Abd

–10

–20

–30 0

b

10

20

30

20.0

60 40 50 Normalized (percent)

70

80

90

100

70

80

90

100

70

80

90

100

Obliquité bassin

up

15.0

Angle (degrees)

10.0 5.0 0.0 –5.0

Down

–10.0 –15.0 –20.0 0

Int

c

10

20

30

40 50 60 Normalized (percent) Rotation bassin

30

20

Angle (degrees)

10

0

–10

Ext

–20

–30 0

10

20

30

40

50

60

Normalized (percent)

Fig. 3 (a) Patient with right hemiplegia. Hip frontal plane kinematics show an asymmetric pattern, with the right side being abducted throughout the cycle. This pattern of hip motion is related to asymmetric pelvic frontal (b) and transverse (c) alignment

1100

A. Presedo

during swing phase, and global knee range of motion tends to be very limited (“crouch gait”). In cerebral palsy, this pattern is usually associated to hamstring spasticity or contracture but can be also secondary to plantar flexor or hip extensor weakness. Insufficient knee flexion in swing phase has been defined as diminished and/or delayed peak of knee flexion and is referred to as “stiff-knee gait” (Sutherland and Davids 1993). Stiff-knee gait was first described as one of the gait classification patterns; however, this pattern of gait is not considered as a separate entity by Rodda and Graham classification (Rodda and Graham 2001) since knee stiffness can be part of different types of gait. Stiff-knee gait is frequent in patients with jump and crouch gait patterns and also in hemiplegic patients with knee involvement. This kinematic pattern has been initially described as a deficit of knee flexion during swing (Fig. 4). However, since patients who walk with a permanent knee flexion throughout the

Fig. 4 (a) This patient shows knee stiffness and difficulty to clear the foot during swing phase. (b) Kinematics curve shows insufficient peak knee flexion during initial swing with decreased knee range of motion. (c) There is a premature and prolonged activity of the rectus femoris muscle during terminal stance and swing phases

Swing Phase Problems in Cerebral Palsy

1101

cycle may have normal amount and timing of PKF but very limited knee range of motion that may be difficult to clear the foot, decreased knee range of motion is also considered one of the criteria to define stiff-knee gait (Moreau and Tinsley 2005). In children with CP, stiff-knee gait is almost always associated to abnormal activity of the rectus femoris during swing. Abnormal knee function is almost always present in children with spastic diplegia. In patients with spasticity, selective distal motor control is impaired, and so they are stance stability and propulsion. Young patients with mild involvement can compensate for this with an increase of cadence, so they preserve speed and swing knee flexion. However, propulsion tends to deteriorate over time and so it does swing knee function. Because of the lack of distal motor control, biarticular muscles (rectus femoris and hamstrings) tend to favor proximal function and act as hip flexors/extensors rather than control knee motion. It is common in diplegic and hemiplegic patients to show prolonged activity of RF and hamstring co-contraction during swing phase. Thus, both muscles are active through mid-swing when they should be silent. The RF acts as a primary hip flexor and secondarily extends the knee, reducing the peak knee flexion, and the hamstrings act primarily as hip extensors and not being able to counterbalance the extension caused by the rectus (Gage et al. 1987; Perry 1987; Sutherland et al. 1990). Given the pathology of these muscles, the ability to allow variation of cadence can be largely compromised in patients with CP. In addition to that, poor propulsion and deficit of passive hip flexion that results are frequent in patients with slow walks and also contribute to knee stiffness.

Ankle and Foot Pathology In patients with CP, there is almost always a dominance of the triceps over the ankle dorsiflexors (Gage 1991). During swing phase, the shape of the ankle kinematic curve can be normal or it can lack dorsiflexion (Fig. 5a). In patients with spastic diplegia, foot position relative to the shank is usually normal, and forefoot initial contact is commonly related to abnormal knee flexion rather to ankle plantar flexion. The gastrocnemius is usually spastic and can develop some degree of contraction, but the soleus tends to maintain normal length, so fixed plantar flexion due to triceps contracture is relatively uncommon in these children. During swing phase, although the dorsiflexors may still be overwhelmed by the plantar flexors, some degree of dorsiflexion is in general possible, and thus, ankle kinematics tends to occur within normal limits. In children with hemiplegia, contracture is common in both the soleus and gastrocnemius. As a result of the overactivity of the triceps and the dominance of the tibialis anterior and posterior over the peroneals and toe extensors, the foot is typically postured in equinovarus. This deformity tends to become rigid over time. Thus, patients with hemiplegia may show a permanent plantar flexion kinematic pattern (Fig. 5b). During swing phase, problems with foot clearance and inadequate foot position in terminal swing can result from abnormal ankle kinematics.

1102

A. Presedo

Fig. 5 (a) Ankle kinematics in a spastic diplegic patient is characterized by plantar flexion at initial contact, followed by limited dorsiflexion in stance and insufficient dorsiflexion in swing phase. (b) Ankle kinematics in a hemiplegic patient with gastrocsoleus contracture shows permanent plantar flexion

Swing Phase Problems in Cerebral Palsy

1103

Management of Swing Phase Problems in Cerebral Palsy Treatment strategies should be based on the age of the patient and the severity of neurologic involvement. Primary problems (spasticity) should receive specific treatment with the aim of prevent secondary problems (muscular contractures and bone deformities). If orthopedic surgery procedures are needed, these should be planned within a multidisciplinary team, taking into consideration the whole child and not just his motor-skeletal parts.

Excessive Hip Adduction In diplegic and hemiplegic patients with distal involvement, abnormal hip and pelvis kinematic patterns may represent coping responses. In hemiplegic patients, abnormal rotation of pelvis and hip are very common and may be related to asymmetric neurological involvement and/or to torsional bony deviations. Increased hip adduction in terminal stance and initial swing is commonly seen in diplegic patients with more severe involvement and tends to be related to excessive femoral anteversion and adductor muscle contractures. Excessive hip adduction in swing can compromise foot clearance and cause tripping and falling. Therapeutic goals are oriented to treat adductor muscle contractures and bony torsions.

Stiff-Knee Gait Stiff-knee gait management in children with cerebral palsy will focus on treatment of RF spasticity and/or contractures. Rectus femoris surgical procedures are often performed as part of multilevel surgery; therefore, improvement on stance phase parameters can largely contribute to a better knee flexion during swing. Adequate knee flexion velocity at toe-off has been proved to be the most important kinematic factor to achieve normal swing phase knee flexion (Goldberg et al. 2004). Hip flexion moment, generated by the iliopsoas muscle, and ankle plantar flexion, generated by the gastrocnemius, were identified as the parameters that contribute most to increasing knee velocity during double support. Since prolongation of phasic, late stance, RF activity into swing phase constitutes an electromyographic (EMG) pattern commonly associated with stiff-knee gait, RF proximal or distal release procedures have been proposed to improve knee flexion in patients with abnormal RF activity in swing phase (Sutherland et al. 1990). According to these authors, RF proximal release did not influence pelvic alignment or hip range of motion. They concluded that RF release would reduce the extensor properties of the muscle and thus facilitate passive knee flexion in swing. Since patients with cerebral palsy tend to walk slower than normal, hip flexion moment in pre-swing is often diminished. For this reason, Perry suggested to abandon proximal RF release and transfer the RF posterior to the axis of knee to enhance active knee flexion (Perry 1987). Indications for transfer included (a) diminished range of knee

1104

A. Presedo

flexion during swing phase, (b) excessive RF EMG activity during swing, and (c) a positive Duncan-Ely test. Õunpuu et al. (1993a) compared the outcomes of RF release versus transfer. Although the differences were not significant in terms of amount of variation in PKF and time to PKF in swing, this study showed that there was a tendency toward more normal values after the transfer. This tendency could be explained by the notion that the RF, transferred posterior to the axis of knee flexion, would generate a knee flexion moment in swing. The choice of the transfer site did not influence PKF, knee ROM, or transverse plan kinematics values (Õunpuu et al. 1993b). Hemo et al. (2007) compared the outcomes of two different techniques: RF distal release and RF transposition to the iliotibial band. They found similar improvement in knee ROM, PKF, and time to PKF after 1-year follow-up. Factors related to good results after RF transfer have been indicated: (a) preoperative RF prolonged activity limited to swing phase (Miller et al. 1997), (b) preoperative positive Duncan-Ely test (Kay et al. 2004), and (c) postoperative increase in knee flexion velocity at toe-off (Goldberg et al. 2006). More recent studies attempted to explain the action of the RF muscle following distal tendon transfer. Riewald and Delp (1997) investigated whether the RF converts to a knee flexor after being transferred to the semitendinosus muscle or to the iliotibial band. Rectus femoris EMG activity showed that the muscle generated an extensor moment in all of their subjects. Based on cine phase-contrast MRI, Asakawa et al. (2002, 2004) examined RF motion in vivo and muscle geometry after tendon transfer surgery. In the tendon transfer group, the RF moved in the direction of the knee extensors, and fiber excursions were reduced compared to vastus intermedius. These authors concluded that the RF was not converted to a knee flexor after its distal tendon was transferred to the posterior side of the knee, but its capacity for knee extension was diminished by the surgery. They suggested that scar tissue could form after RF transfer, making the RF adhere to the underlying muscles. They also examined three-dimensional MRI of patients who had a RF transfer and observed abnormal, low-signal intensity images that could represent scar tissue between the transferred muscle and the underlying vastii in each of the patients. Tridimensional models showed that the transferred muscles followed angular, deviated paths to their new insertions, suggesting that RF tendons were probably constrained by adhesions to the underlying muscles. Following these results, we believed there was insufficient evidence supporting transferring the RF, rather than performing a distal release. We also thought a complete tendon resection would reduce adherences and prevent relapses (Fig. 6). We therefore determined to assess the outcome of children with spastic diplegia following RF distal tendon resection as a part of multilevel surgery, in order to (a) evaluate the improvement of knee ROM during gait, (b) measure changes in maximum knee flexion and time to PKF during swing phase, and (c) compare our results with those reported in the literature after RF distal release or transfer, since we are not aware of any previous reports on RF distal tendon resection. Our results, after 2-year follow-up, showed that RF distal tendon resection was associated with improved knee ROM and timing of peak knee flexion in swing, and the absolute values of peak knee flexion became normal for those patients who showed abnormal preoperative values (Presedo et al. 2012).

Swing Phase Problems in Cerebral Palsy

1105

Fig. 6 Rectus femoris distal tendon resection. The tendon is transected at the muscular junction and removed completely in order to avoid subsequent adherences to the underlying muscles

Long-term follow-up studies showed variable results in terms of PKF magnitude and timing but agreement on knee range of motion improvement that was maintained at final follow-up. Moreau and Tinsley (2005) compared the outcomes of a group of patients who had RF transfer as part of multilevel surgical procedures to another group of patients who did not undergo any RF procedure. They found the RF transfer helped to maintain knee ROM and PKF values over time, after a minimum follow-up of 3 years. Dreher et al. (2012) found a significant improvement in timing of PKF, knee range of motion, and knee flexion velocity after 9 years follow-up. Thawrani et al. (2012) reported an improvement in PKF magnitude and timing after 7-year follow-up.

Insufficient Ankle Dorsiflexion and Foot Varus Deviation Sagittal and coronal plan deviations of the ankle and foot are common in hemiplegic patients. These patients may show a permanent plantar flexion pattern, combined with hind foot varus deviation. A gait classification based on kinematic abnormal patterns in the sagittal plan is commonly used for hemiplegic patients (Winters et al. 1987). Type 1 hemiplegia is characterized by a “drop foot” during swing phase. This is due to a lack of control of ankle dorsiflexors. There is not plantar flexor contracture although they tend to be spastic. Because of the absence of muscle contracture, ankle dorsiflexion is normal in mid-stance, but the ankle is in plantar flexion at the initial contact. The treatment of this type of pattern is a leaf spring or an articulated ankle foot orthosis (AFO). Spasticity treatment can be indicated in order to improve ankle dorsiflexion during swing in patients with relative good motor control and active dorsiflexion in physical exam. Type 2 hemiplegia is characterized by a permanent ankle plantar flexion, during stance and swing phases. There is spasticity and/or contracture of plantar flexors. The management of this type of gait is based on spasticity treatment by botulinum

1106

A. Presedo

toxin injections and/or muscular lengthenings, if necessary. If there is a mild contracture, spasticity management combined with walking casts can be very effective to improve ankle range of motion. We tend to use knee immobilizers during nighttime in order to maintain the gastrocnemius in a stretched position. In case of more severe contracture, gastrocnemius lengthening or tendo-Achilles lengthening will be indicated. Most of the children will require the use of orthoses. Leaf spring orthoses will be adequate in absence of knee recurvatum; however, if the knee is fully extended or in recurvatum, articulated AFO with limitation of plantar flexion may be more convenient. Most of the diplegic and some hemiplegic patients with proximal involvement (type 4 hemiplegia) show combined coronal and transverse plane problems. In the coronal plane, excessive hip adduction secondary to adductor spasticity or contracture and hip subluxation represent the main issues. In the transverse plane, pelvic and hip abnormal rotations, femoral anteversion, tibial torsion, and foot deviations are commonly observed. Whereas management of sagittal plane problems is mainly address to spasticity and muscle contracture, bony surgery is often required to treat coronal and transverse plane abnormalities. Patients with type 4 hemiplegia present proximal involvement and a pattern of sagittal kinematics that can be similar to that seen on diplegic patients. As a result of the triceps overactivity and the dominance of the tibialis anterior and posterior over the peroneals, the foot typically in equinovarus tends to become rigid over time. During swing phase, due to spasticity or contracture of tibialis posterior and the predominance of tibialis anterior and extensor hallucis longus, the foot is deviated in varus and supination. In patients with severe spasticity and/or contracture of hip flexors and adductors, hip subluxation is not uncommon. Management of this type of hemiplegia may include spasticity treatment, but more often, muscular lengthening is required. Split anterior tibial tendon transfer, combined with gastrocsoleus and tibialis posterior lengthening, is recommended to correct dynamic equinovarus foot deviation (Hoffer et al. 1985; Davids et al. 2011). Bony surgery to address foot deformities and femoral torsion can be also part of the surgical program. After surgery, the most common type of orthoses used is a solid AFO that will help with knee extension and will also avoid ankle plantar flexion during swing. With time, most of the patients tend to keep a leaf spring AFO to help with clearance.

Summary Swing phase abnormal kinematics can compromise foot clearance and adequate limb pre-positioning for initial contact. In children with cerebral palsy, these problems are related to spasticity, muscular weakness, and poor motor control. Stiff-knee gait represents the most common swing phase kinematic problem in these patients. Although knee flexion during swing is related to terminal stance and pre-swing kinematics, rectus femoris prolonged activity during swing is almost always present, and stiff-knee treatment is oriented to diminish the extensor effect of this muscle. Long-term studies have demonstrated that after rectus femoris surgery (distal tendon

Swing Phase Problems in Cerebral Palsy

1107

resection or transfer), swing phase knee kinematics and knee range of motion tend to improve, and this improvement lasts over time. Difficulties with foot clearance caused by ankle dorsiflexor weakness, tendo-Achilles contracture, or hindfoot varus deviation are frequent in hemiplegic patients. Treatment focuses on spasticity management, musculotendinous releases, tibialis anterior transfer, or bony procedures to correct fixed deformities.

References Anderson FC, Goldberg SR, Pandy MG, Delp SL (2004) Contributions of muscle forces and toe-off kinematics to peak knee flexion during the swing phase of normal gait: an induced position analysis. J Biomech 37:731–737 Annaswamy TM, Giddings CJ, Della Croce U, Kerrigan DC (1999) Rectus femoris: its role in normal gait. Arch Phys Med Rehabil 80:930–934 Asakawa DS, Blemker SS, Gold GE, Delp SL (2002) In vivo motion of the rectus femoris muscle after tendon transfer surgery. J Biomech 35:1029–1037 Asakawa DS, Blemker SS, Rab GT, Bagley A, Delp SL (2004) Three dimensional muscle-tendon geometry after rectus femoris tendon transfer. J Bone Joint Surg 86:348–354 Davids JR, Rogozinski BM, Hardin JW, Davis RB (2011) Ankle dorsiflexion function after plantar flexor surgery in children with cerebral palsy. J Bone Joint Surg Am 93:e 138, 1–7 Dreher T, Wolf SI, Maier M, Hagmann S, Vegvari D, Gantz S, Heitzmann D, Wenz W, Braatz F (2012) Long-term results after distal rectus femoris transfer as a part of multilevel surgery for the correction of stiff-knee gait in spastic diplegic cerebral palsy. J Bone Joint Surg Am 94:e 142, 1–10 Gage JR (1991) Gait analysis in cerebral palsy. Mac Keith Press, London, pp 61–95 Gage JR, Perry J, Hicks RR, Koop S, Werntz JR (1987) Rectus femoris transfer to improve knee function of children with cerebral palsy. Dev Med Child Neurol 29:159–166 Goldberg SR, Anderson FC, Pandy MG, Delp SL (2004) Muscles that influence knee flexion velocity in double support: implications for stiff-knee gait. J Biomech 37:1189–1196 Goldberg SR, Õunpuu S, Arnold AS, Gage JR, Delp SL (2006) Kinematic and kinetic factors that correlate with improved knee flexion following treatment for stiff-knee gait. J Biomech 39:689–98. Gough M, Shortland AP (2008) Can clinical gait analysis guide the management of ambulant children with bilateral spastic cerebral palsy? J Pediatr Orthop 28:879–883 Hemo Y, Aiona MD, Pierce RA, Dorociack R, Sussman M (2007) Comparison of rectus femoris transposition with traditional transfer for the treatment of stiff knee gait in patients with cerebral palsy. J Child Orthop 1:37–41 Hoffer MM, Barakat G, Koffman M (1985) 10-year follow-up of split tibial tendon transfer in cerebral palsied patients with spastic equinus deformity. J Pediatr Orthop 5:432–434 Kay RM, Rethlefsen SA, Kelly JP, Wren TAL (2004) Predictive value of the Duncan-Ely test in distal rectus femoris transfer. J Pediatr Orthop 24:59–62 Miller F, Dias R, Lipton GE, Albarracin JP, Dabney KW, Castagno P (1997) The effect of rectus EMG patterns on the outcome of rectus femoris transfers. J Pediatr Orthop 17(5):603–607 Moreau N, Tinsley S (2005) Progression of knee joint kinematics in children with cerebral palsy with and without rectus femoris transfers: a long-term follow up. Gait Posture 22:132–137 Narayanan UG (2007) The role of gait analysis on the orthopaedic management of ambulatory cerebral palsy. Curr Opin Pediatr 19:38–43 Nene A, Mayagoitia R, Veltink P (1999) Assessment of rectus femoris function during initial swing phase. Gait Posture 9:1–9 Õunpuu S (1994) The biomechanics of walking and running. Clin Sports Med 13:843–863

1108

A. Presedo

Õunpuu S, Muik E, Davis RB 3rd, Gage JR, DeLuca PA (1993a) Rectus femoris surgery in children with cerebral palsy. Part I. The effect of rectus femoris transfer location on knee motion. J Pediatr Orthop 13:325–330 Õunpuu S, Muik E, Davis RB 3rd, Gage JR, DeLuca PA (1993b) Rectus femoris surgery in children with cerebral palsy. Part II. A comparison between the effect of transfer and release of the distal rectus femoris on knee motion. J Pediatr Orthop 13:331–335 Perry J (1987) Distal rectus femoris transfer. Dev Med Child Neurol 29:53–58 Perry J (1992) Gait analysis: normal and pathological function. Slack, Thorofare, pp 1–19 Piazza SJ, Delp SL (1996) The influence of muscles on knee flexion during the swing phase of gait. J Biomech 29:723–733 Presedo A, Megrot F, Ilharreborde B, Mazda K, Penneçot GF (2012) Rectus femoris distal tendon resection improves knee motion in patients with spastic diplegia. Clin Orthop Relat Res 470(5):1312–1319 Riewald SA, Delp SL (1997) The action of the rectus femoris muscle following distal tendon transfer: does it generate knee flexion moment? Dev Med Child Neurol 39:99–105 Rodda J, Graham HK (2001) Classification of gait patterns in spastic hemiplegia and spastic diplegia: a basis for a management algorithm. Eur J Neurol 8(Suppl 5):98–108 Sutherland DH, Davids JR (1993) Common gait abnormalities of the knee in cerebral palsy. Clin Orthop Relat Res 288:139–147 Sutherland DH, Santi M, Abel MF (1990) Treatment of stiff-knee gait in cerebral palsy: a comparison by gait analysis of distal rectus femoris transfer versus proximal rectus release. J Pediatr Orthop 10:433–441 Thawrani D, Haumont T, Church C, Holmes L, Dabney KW, Miller F (2012) Rectus femoris transfer improves stiff-knee gait in children with spastic cerebral palsy. Clin Orthop Relat Res 470(5):1303 Winter DA (1991) The biomechanics and motor control of human gait: normal, elderly and pathological, 2nd edn. University of Waterloo Press, Waterloo, pp 35–52, 75–85 Winters TF, Gage JR, Hicks R (1987) Gait patterns in spastic hemiplegia in children and young adults. J Bone Joint Surg Am 69:437–441

Strength Related Stance Phase Problems in Cerebral Palsy Justin Connor and Mutlu Cobanoglu

Abstract

Normal human gait results from a combination of several complex coordinated activities. The variable loss of control that is associated with cerebral palsy (CP) lesions can cause multiple interruptions in the gait cycle. To understand and properly evaluate pathologic gait patterns and to provide proper medical remedies for gait impairments due to CP, the clinician must understand the normal gait cycle. Instrumented gait analysis is a tool that provides detailed information and quantitative measurements throughout the gait cycle to evaluate individual gait patterns that help surgeons plan appropriate interventions. There are several prevailing abnormal gait patterns associated with spastic CP. These patterns are categorized into those affecting stance vs swing phases of gait. The patterns at the knee most often responsible for impeding the gait cycle by causing stance phase instability are crouch gait and back-kneeing in sagittal plane kinematic. Patterns such as back-kneeing cause knee hyperextension which is especially bad for delaying push-off and propelling forward; crouch knee gait causes dropping down into hip and knee flexion. These two types of gait patterns require different types of intervention due to differing muscle responses. Identification of the correct gait pattern in cerebral palsy is required to create the correct management algorithm.

J. Connor Nemours A.I. duPont Hospital for Children, Wilmington, DE, USA e-mail: [email protected] M. Cobanoglu (*) Department of Orthopedics and Traumatology, Adnan Menderes University Faculty of Medicine, Aydın, Turkey e-mail: [email protected] # Springer International Publishing AG, part of Springer Nature 2018 B. Müller, S.I. Wolf (eds.), Handbook of Human Motion, https://doi.org/10.1007/978-3-319-14418-4_54

1109

1110

J. Connor and M. Cobanoglu

Keywords

Cerebral palsy • Stance phase • Crouch gait • Back-knee • Ankle-foot orthosis • Knee-ankle-foot orthosis • Ground reaction force Abbreviations

AFO CP GRAFO HAT KAFO

Ankle-foot orthosis Cerebral palsy Ground reaction ankle-foot orthosis Head, arm, trunk Knee-ankle-foot orthosis

Contents Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . State of the Art . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Back-Kneeing . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Treatment . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Crouch Knee . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Treatment . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Cross-References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

1110 1111 1112 1113 1113 1116 1118 1118 1118

Introduction Normal human gait is composed of a complex combination of coordinated motor activities which require precise control from the central nervous system. These motions control balance, motor activity, and cognitive decision-making. The gait cycle is divided into stance and swing phases. For gait to be productive and effective, there are five prerequisites as described by Perry: stability in stance, swing phase clearance, foot preposition in terminal swing, adequate step length, and energy conservation (Gage 1993). The role of stance phase (comprises 60% of the normal gait cycle) is to provide support for weight bearing and stability to provide power for push-off. Swing phase (comprises 40% of the normal gait cycle) advances the limb forward, thus allowing forward motion. These phases are supported by forces which are created at the ankle, knee, and hip joints by internal moments produced by attached muscles. The measurement of these forces at each joint through instrumented gait analysis allows for evaluation of kinetic moments. These kinetic measures create resultant vector forces by muscles on the bones of the body. Directly measuring muscle force is clinically impractical; however, net joint forces may be indirectly measured as the opposing ground reaction force. The residual of the ground reaction force at each joint has a direction and distance from the defined center of the joint that impacts gait. Concentric contraction of hip extensors, knee extensors, and ankle plantar flexors in stance phase maintain a ground reaction force in front of the knee joint during the stance phase, leading the erect posture during gait (Fig. 1) (Kadhim and Miller 2014).

Strength Related Stance Phase Problems in Cerebral Palsy

1111

Fig. 1 As shown in the figure, concentric contraction of hip extensors, knee extensors, and ankle plantar flexors in stance phase maintain a ground reaction force (red arrows) in front of the knee joint during the stance phase, leading the erect posture during gait

What is happening during stance phase? On the first contact to the ground, the knee is slightly flexed (approximately 5 ) to absorb the shock during initial heel strike. At this moment the quadriceps and hamstrings contract isometrically to stabilize the knee joint. Then the knee further flexes to 20 , shifting the center of gravity forward to allow advancement. After the short period of flexion, second half of stance starts. In this phase, the knee starts to extend with the activation of plantar flexion–knee extension couple: the soleus contracts and slows the progression of the tibia. This extension moment without activation of quadriceps muscle keeps the ground reaction force anterior to the knee. The hamstrings function in a closed chain to act as hip extensors. In the terminal phase of stance, limb acceleration occurs with the concentric contraction of the plantar flexors, providing power. This complex coordination of muscular activities is often difficult to maintain due to the neurologic insult associated with CP. Orthopedists must understand normal and abnormal gait cycle patterns and measurement techniques to be able to accurately evaluate pathologic gait and to improve function.

State of the Art Muscular spasticity and contractures leading to loss of functional abilities can be observed in ambulatory children with cerebral palsy resulting in pathologic gait patterns (Bell et al. 2002; Johnson et al. 1997). These primary motor deficits can lead to secondary biomechanical and growth problems such as muscle tightness, joint contractures, and secondary bony malformations. Rodda et al. described a classification based on the pelvis, hip, knee, and ankle position during stance to assist in understanding these abnormal gait patterns. According to this classification, there is a subset of patients in which the ankle is in equinus and knee and hip extend fully. Spasticity of the hamstrings and hip flexors increases gradually and leads to flexion of the knee and hip. This has been described as a jump gait. The classic crouch gait is manifested by ankle dorsiflexion with hip and knee flexion throughout the stance

1112

J. Connor and M. Cobanoglu

phase (Rodda et al. 2004). Increased body mass associated with the pubertal growth spurt may advance a child from a jump gait pattern to a crouch gait pattern (Rodda et al. 2006). The gait pattern also becomes asymmetric (Rodda et al. 2004). Crouch gait and back-kneeing (genu recurvatum) are the common types of gait patterns in ambulatory children with CP (Wren et al. 2005; Klotz et al. 2013). Both patterns affect stance phase stability by affecting pre-postponing the limb for proper push-off. The aim of this chapter is to investigate the properties and treatment results of these gait patterns.

Back-Kneeing The back-kneeing position is defined as hyperextension of the knee during the midstance phase and is frequently found in children with bilateral spastic cerebral palsy (Fig. 2). Eventually it may cause instability of the knee as the result of progressive stretching of the posterior knee capsule and ligamentous structures and over time cause secondary malformations of the tibial condyles as growth proceeds (Simon et al. 1978; Klotz et al. 2014). This condition can primarily occur as a result of three patterns. The first pattern is characterized by overactivity of the

Fig. 2 (a) The back-kneeing position is defined as hyperextension of the knee during the midstance phase of stance. (b) Ground reaction force in stance phase of the left lower limb is shown as red arrow which goes anterior to the knee axis. Average moments of lower limb joints in sagittal plan are shown. (c) Ten to forty percent of the gait cycle increased the internal hip extension moment seen because during the stance phase, the hip extensors try to compensate excessive hip flexion. This creates an increased internal extension moment. (d) Due to the severe back-kneeing during the stance phase, the knee flexor and capsule activate excessively to balance the high internal flexor moment caused by the hyperextension. (e) Because of persistent plantar flexion, the length of plantar flexors is shortened which limits the maximum internal plantar flexor moment needed after midstance (red line demonstrates right leg; blue line demonstrates left leg in c, d, and e)

Strength Related Stance Phase Problems in Cerebral Palsy

1113

gastrocsoleus muscles. The overactive gastrocsoleus causes the hyperextended knee through the coupling of ankle plantar flexion and knee extension. The second pattern is movement of the HAT (head, arm, trunk) center of gravity anterior to the knee in the presence of a weak gastrocnemius. The third pattern is movement of the HAT center of gravity posterior of the hip joint but anterior of the knee joint (Simon et al. 1978). Klotz et al. reported that equinus was a major underlying factor in primary back-knee gait (Klotz et al. 2014). Furthermore, this problem may occur secondarily after hamstring lengthening for the correction of flexed knee gait (Dreher et al. 2012).

Treatment The first consideration in the treatment of back-kneeing is to assess whether the gastrocnemius has adequate length to allow for dorsiflexion past neutral with knee extension. A hinged ankle-foot orthosis (AFO) with a 90 degree plantar flexion stop is the preferred orthosis for children whose back-kneeing results from overactivity of gastrocnemius spasticity. If dorsiflexion of the ankle with knee extension is possible, articulated AFO that allows 3–5 of dorsiflexion while limiting plantar flexion to minus 5 should be prescribed (Miller 2005). By setting the plantar flexion stop at 5 of dorsiflexion, these children will be forced into knee flexion in stance if they are independent ambulators. If the ground reaction force is moving either significantly in front or behind the knee in the presence of a plantar flexor weakness, a non-articulated solid ankle AFO should be used to control the ankle. If individuals have any shortness of the gastrocsoleus, the knee will hyperextend and go into back-kneeing. On the other hand, if individuals have plantar flexor weakness, back-kneeing can also occur due to increased ankle dorsiflexion and a HAT segment anterior to the knee. Instrumented gait analysis has shown an increased trunk lean in patients walking with assistive devices. In spite of appropriate orthotics, these assistive devices lead to an increased anterior trunk lean which may cause progressive back-kneeing and the development of pain (Simon et al. 1978; Klotz et al. 2014; Miller 2005). In the presence of this kind of progressive back-kneeing, the treatment method with AFO must be changed with knee-ankle-foot-orthosis (KAFO) with extension blocking hinges. But it is important to make sure that there is no contracture of the gastrosoleus. Ankle dorsiflexion has to be 5–10 in knee extension or the gastrocnemius should be lengthened (Dreher et al. 2012; Miller 2005).

Crouch Knee Crouch gait is common in ambulatory children with cerebral palsy (CP). The definition of a crouch gait is increased knee flexion in midstance with increased ankle dorsiflexion and usually increased hip flexion (Fig. 3). The crouch pattern may be seen in all levels of severity; however, it is primarily encountered in moderate and

1114

J. Connor and M. Cobanoglu

Fig. 3 (a) The definition of a crouch gait is increased knee flexion in midstance with increased ankle dorsiflexion and usually increased hip flexion. (b) Ground reaction force in stance phase of the lower limbs is shown as red arrow which goes posterior to the knee axis. Average moments of lower limb joints in sagittal plan are shown. (c) In this the gait cycle increased internal hip flexion moment seen at the end of stance phase in order to prepare the clearance during swing phase. (d) Due to the persistent knee flexion, knee extensors activate throughout the stance phase to balance the high knee internal extension moment caused by the knee flexion to prevent the possible collapse. (e) Because of persistent dorsiflexion, the plantar flexors activate to balance high ankle internal dorsiflexion moment; however, due to weakness of plantar flexors, the balance has to be maintained with strong quadriceps contraction (red line demonstrates right leg; blue line demonstrates left leg in c, d, and e)

severe diplegia. The ground reaction force is maintained close to the centers of the hip, knee, and ankle joints, reducing the demands on the antigravity support muscles. Failure of the total body extensor moment as a result of diminished ability of the hip, knee, or ankle plantar flexor moments may result in collapse of the extension posture into a flexion posture, described as crouch gait (Rodda et al. 2006). The rapid onset of weight and height growth may cause knee flexion contractures, hamstring contractures, deficient foot moment arm, and gastrocsoleus weakness which can contribute to the classic crouch gait. Knee flexion in stance phase increases and the foot starts to dorsiflex, and severe planovalgus foot deformities, characterized by heel equinus, talonavicular joint dislocation, and midfoot break during stance, develop and reduce lever arm (Kadhim and Miller 2014). The toe walking with knee flexion pattern (jump gait) is typically not seen in full adolescence or nearly adult-sized individuals. The muscles and joints are not strong enough to support the body weight for chronic ambulation with the typical early childhood toe-walking pattern. The other lever arm dysfunctions affecting stance stability are rotational deviations at the femurs, tibias, and feet. External tibial torsion reduces the extension capacity of the soleus and hip extensors, as it misdirects the forward propulsion vector, thus decreasing power, especially if the torsion is greater than 30 of normal (Hicks et al. 2007). Furthermore, the planovalgus foot is usually associated with

Strength Related Stance Phase Problems in Cerebral Palsy

1115

external tibial torsion. In the presence of gastrosoleus weakness in crouch gait, the ground reaction force creates a flexion moment on the knee and hip by passing through the knee posteriorly and the hip anteriorly. To prevent the knee from collapsing into flexion, quadriceps activity must increase during stance. The rectus femoris however not only acts as a hip extensor of the quadriceps but also contributes to hip flexion in a closed chain. Hip extensors are mechanically disadvantaged and the hamstrings cause increasing knee flexion. Over time, fixed hamstrings and knee flexion contracture, which are responsible for further exacerbation of crouch, may develop. The patellar tendon elongates gradually and extensor mechanism weakens. A secondary etiology for crouch may be a significant hip flexion contracture, which can limit knee extension in midstance. Although hamstring contracture is considered a primary cause for crouch, many patients have normal length hamstrings (Arnold et al. 2006; Delp et al. 1996). And also it has also been observed that excessive knee flexion typically accompanies excessive hip flexion due to shortened hip flexors throughout the gait cycle (Delp et al. 1996). The shortened hip flexors lead to anterior pelvic tilt facilitating hamstring tightness and causing knee flexion in stance phase (Kedem and Scher 2016). As a result of increasing crouch, the stress on the knee extensor mechanisms to support weight bearing may lead to the complaint of knee pain. Standing with >30 of knee flexion increases the forces acting on the quadriceps, patella, and proximal part of the tibia and requires the quadriceps muscle to work at >50% of its maximum moment-generating capacity in order to stabilize the knee joint (Perry et al. 1975). Tibial tubercle apophysitis or even chronic fractures through the distal pole of the patella may occur, especially during rapid growth. Progressive failure of the knee extensor mechanism is associated with gait deterioration, increased dependence on walkers or crutches, and the need for wheelchair use in the community (Rodda et al. 2006). As gait deteriorates the position of the feet may progress into increased. Orthotics lose their ability to support the collapsing feet, and ultimately the child may lose the ability and motivation to walk (Miller 2005). A child’s weight plays an important role in the evolution of gait patterns through maturity. It should be monitored on every clinic visit particularly throughout the pubertal growth spurt. During this period the child may exhibit new complaints of pain due to increasing stress on the knees or feet. Also, the physical examination should be monitored, being mindful of passive knee extension and popliteal angle measurements, to monitor progressive hamstring contractures or fixed knee flexion contractures that may impede gait. Objective measures to evaluate crouch gait through instrumented gait analysis include kinematic evaluation of the magnitude of the knee flexion in midstance, excessive dorsiflexion of the ankle, and the knee excessive internal knee extension moments in midstance. If the range of motion (ROM) of ankle is normal or below normal, and ROM does not increase with the knee flexion, the ankle weakness and foot moment arm are the most likely reasons of the crouch. If the knee extends to the limits of the fixed knee flexion contracture measured on physical examination, the knee joint contracture is most likely a cause. If the ankle has a high plantar

1116

J. Connor and M. Cobanoglu

flexion moment with a high knee flexion moment, a combination of contracture of the gastrocnemius and the hamstrings is most likely a cause. If the hip extension peak occurs early, and a significant hip flexion contracture is positive in the physical examination, hip flexion contracture may also contribute to knee flexion deformity in the midstance phase (Miller 2005).

Treatment Appropriate treatment for crouch gait should focus on early detection, by monitoring examinations every 6 months during middle childhood and intervention before the problem becomes severe. Mild crouch can be initially controlled conservatively, with spasticity management, physical therapy, and foot orthoses. But the role of physical therapy and strengthening exercises in the treatment of crouch is controversial. A meta-analysis of randomized trials concluded that strengthening interventions had no effect on strength, walking speed, or activity level in children with CP (Scianni et al. 2009), though there are reports that find that strengthening exercises may lead to some functional improvement in patients with crouch (Damiano et al. 1995, 2009). The aim of the interventions in these children is to reduce knee flexion and prevent gait deterioration ultimately maintaining gait efficiency and walking activity in daily life. Initially a solid AFO provides stability to the ankle and foot and provides a stable base of support for children to stand. These can usually be used at the preambulatory stage between the ages of 18 and 24 months. It is easy to done and works well for child less than 30 kg. As children gain better stability and start to walk using a walker, usually between the ages of 3 and 4 years, the ankle hinge can be added to allow dorsiflexion but limit plantar flexion. Most children who have good walking ability with diplegic and hemiplegic pattern involvement benefit from the transition to a hinged AFO at approximately 3 years of age. But hinged AFO is contraindicated if children have severe planovalgus or varus foot deformity and increased knee flexion in stance (Miller 2005). A rigid AFO can compensate for weakness of ankle plantar flexors and may normalize knee kinematics and kinetics effectively, but it has the disadvantage of inhibiting push-off power (Kerkum et al. 2015). Although AFO may affect ankle joint dorsiflexion during stance, ankle power and ankle plantar flexion moment, it does not affect proximal joint movement during gait (Rethlefsen et al. 1999). A ground reaction ankle-foot orthosis (GRAFO) can be used to enhance push-off power and is a commonly applied intervention in children with CP walking with crouch. A GRAFO features a solid pretibial shell to more effectively redirect the GRF vector and slow tibial progression as the center of pressure moves distally under the foot during stance (Kane et al. 2010). Extremities with knee and hip flexion contractures of 15 (Rogozinski et al. 2009). The GRAFO limits ankle dorsiflexion and reduces knee flexion in stance. This improved ankle and knee

Strength Related Stance Phase Problems in Cerebral Palsy

1117

position reduces the knee extensor moment in stance. The ankle can be brought to neutral dorsiflexion with the knee in full extension. If this is not possible, the orthosis cannot work and these children will require gastrocnemius and hamstring lengthening to accommodate the orthosis. The other point is that the foot-to-knee axis has to be in a relatively normal alignment which defines less than 20 of internal or external tibial torsion (Miller 2005). Transverse plane skeletal malalignments and foot segmental malalignments (planovalgus deformities) may lead to an increased external foot progression angle which moves the line of action of the ground reaction force lateral to the knee joint center. Recent studies utilizing dynamic computer models of the musculoskeletal system have demonstrated that increased external tibial torsion impedes the capacity of the soleus to extend the knee during stance, which supports the hypothesis of lever arm dysfunction as an important contributor to crouch gait (Hicks et al. 2007). As these children get heavier, this orthosis becomes more effective; however, it also has to be made stronger. As children approach 50–70 kg, the orthosis has to be constructed with a composite of carbon fiber or laminated copolymer to withstand the applied forces (Miller 2005). The surgery for crouch gait often includes many procedures at different joints. Single-event multilevel surgery is widely performed in cerebral palsy and focuses on improving gait function and pattern (Rodda et al. 2006). The preferred order is to start from proximal to distal. First the hip rotation should be corrected, with iliopsoas lengthening if needed. Distal hamstring lengthening if indicated should also be performed. Repeat hamstring lengthening may prevent or delay progressive crouch in patients with CP but does not result in long-term improvement in crouch gait (Rethlefsen et al. 2013). The hamstring lengthening followed by knee capsulotomy or femoral extension osteotomy (Novacheck et al. 2009) is performed if indicated. The surgery for foot deformity is done next; then an intraoperative assessment of the torsional alignment is used to make the final determination of the need for a tibial osteotomy. After the tibial osteotomy, the hip fully extends and the knee can be fully extended and lies in approximately 10 of external rotation. The foot-to-thigh alignment should be 20 external to neutral with neutral dorsiflexion (Kadhim and Miller 2014; Miller 2005). If the foot has a significant planovalgus or a midfoot break, it must be corrected. A stable and correctly aligned foot is mandatory in the correction of crouch because the ground reaction force has to be controlled through the foot as a functional moment arm (Kadhim and Miller 2014). Postoperative rehabilitation should start in the hospital with the goal of having children at least stand before discharge and plan for immediate home rehabilitation. Parents need to expect that the acute rehabilitation will take 3 months until these individuals are close to their preoperative function, and then it will take at least 1 year of rehabilitation to reach maximum function. If there is weakness or a tendency for the gastrocsoleus not to have good strength, a GRAFO has to be used postoperatively. This is the ideal time to use the articulated GRAFO, which will allow the gastrocsoleus to gain strength, and over 1–2 years, the orthotic can be weaned away and the correction will be maintained (Miller 2005).

1118

J. Connor and M. Cobanoglu

Conclusion There are a variety of factors that affect stance phase stability of the lower limb segment in gait. These factors may lead to progressive deterioration of gait and function and may ultimately lead to pain and disability. Through the use of routine monitoring of pubertal maturation, physical examination, and analysis of kinematics and moment kinetics through instrumented gait analysis, appropriate treatment plans can be acted upon to prevent progression of deformity and loss of function. Appropriate bracing adjuvants should also be utilized in addition to surgical interventions.

Cross-References ▶ Diagnostic Gait Analysis Use in the Treatment Protocol for Cerebral Palsy ▶ Natural History of Cerebral Palsy and Outcome Assessment ▶ Spasticity Effect in Cerebral Palsy Gait ▶ Swing Phase Problems in Cerebral Palsy

References Arnold AS, Liu MQ, Schwartz MH, Ounpuu S, Delp SL (2006) The role of estimating muscletendon lengths and velocities of the hamstrings in the evaluation and treatment of crouch gait. Gait Posture 23(3):273–81. Epub 2005 Jun 17 Bell KJ et al (2002) Natural progression of gait in children with cerebral palsy. J Pediatr Orthop 22(5):677–682 Damiano DL, Kelly LE, Vaughn CL (1995) Effects of quadriceps femoris muscle strengthening on crouch gait in children with spastic diplegia. Phys Ther 75(8):658–667. discussion 668–71 Damiano DL et al (2009) Can strength training predictably improve gait kinematics? A pilot study on the effects of hip and knee extensor strengthening on lower-extremity alignment in cerebral palsy. Phys Ther 90(2):269–279 Delp SL et al (1996) Hamstrings and psoas lengths during normal and crouch gait: implications for muscle-tendon surgery. J Orthop Res 14(1):144–151 Dreher T et al (2012) Development of knee function after hamstring lengthening as a part of multilevel surgery in children with spastic diplegia. J Bone Joint Surg 94(2). https://doi.org/ 10.2106/jbjs.j.00890 Gage JR (1993) Gait analysis. An essential tool in the treatment of cerebral palsy. Clin Orthop Relat Res 288:126–134 Hicks J et al (2007) The effect of excessive tibial torsion on the capacity of muscles to extend the hip and knee during single-limb stance. Gait Posture 26(4):546–552 Johnson DC, Damiano DL, Abel MF (1997) The evolution of gait in childhood and adolescent cerebral palsy. J Pediatr Orthop 17(3):392–396 Kadhim M, Miller F (2014) Crouch gait changes after planovalgus foot deformity correction in ambulatory children with cerebral palsy. Gait Posture 39(2):793–798 Kane K, Kyra K, John B (2010) Comparison of ground reaction and articulated ankle-foot orthoses in a child with lumbosacral myelomeningocele and tibial torsion. J Prosthetics Orthot 22(4): 222–229 Kedem P, Scher DM (2016) Evaluation and management of crouch gait. Curr Opin Pediatr 28(1): 55–59

Strength Related Stance Phase Problems in Cerebral Palsy

1119

Kerkum YL et al (2015) The effects of varying ankle foot orthosis stiffness on gait in children with spastic cerebral palsy who walk with excessive knee flexion. PLoS One 10(11):e0142878 Klotz MCM et al (2013) Reduction in primary genu recurvatum gait after aponeurotic calf muscle lengthening during multilevel surgery. Res Dev Disabil 34(11):3773–3780 Klotz MCM et al (2014) The association of equinus and primary genu recurvatum gait in cerebral palsy. Res Dev Disabil 35(6):1357–1363 Miller F (2005) Cerebral palsy. Springer Science & Business Media, New York Novacheck TF et al (2009) Distal femoral extension osteotomy and patellar tendon advancement to treat persistent crouch gait in cerebral palsy. Surgical technique. J Bone Joint Surg 91(Suppl 2): 271–286 Perry J, Antonelli D, Ford W (1975) Analysis of knee-joint forces during flexed-knee stance. J Bone Joint Surg 57(7):961–967 Rethlefsen S et al (1999) The effects of fixed and articulated ankle-foot orthoses on gait patterns in subjects with cerebral palsy. J Pediatr Orthop 19(4):470–474 Rethlefsen SA et al (2013) Repeat hamstring lengthening for crouch gait in children with cerebral palsy. J Pediatr Orthop 33(5):501–504 Rodda JM et al (2004) Sagittal gait patterns in spastic diplegia. J Bone Joint Surg 86(2):251–258 Rodda JM et al (2006) Correction of severe crouch gait in patients with spastic diplegia with use of multilevel orthopaedic surgery. J Bone Joint Surg 88(12):2653–2664 Rogozinski BM, Davids JR, Davis RB 3rd, Jameson GG, Blackhurst DW (2009) The efficacy of the floor-reaction ankle-foot orthosis in children with cerebral palsy. J Bone Joint Surg Am 91(10): 2440–2447 Scianni A et al (2009) Muscle strengthening is not effective in children and adolescents with cerebral palsy: a systematic review. Aust J Physiother 55(2):81–87 Simon SR et al (1978) Genu recurvatum in spastic cerebral palsy. Report on findings by gait analysis. J Bone Joint Surg Am 60(7):882–894 Wren TAL, Rethlefsen S, Kay RM (2005) Prevalence of specific gait abnormalities in children with cerebral palsy: influence of cerebral palsy subtype, age, and previous surgery. J Pediatr Orthop 25(1):79–83

Foot and Ankle Motion in Cerebral Palsy Jon R. Davids and Sean A. Tabaie

Abstract

There are three common ankle/foot segmental malalignment patterns seen in children with cerebral palsy (CP): equinus, equinoplanovalgus, and equinocavovarus. Each type of ankle/foot malalignment can be classified into three levels based upon the presence of dynamic muscle imbalance, fixed or myostatic deformity of the muscle-tendon unit, and fixed or skeletal deformities. Management of foot and ankle deformities in children with CP can consist of both nonsurgical (orthotics and pharmacotherapy) and surgical (muscle-tendon unit lengthening or transfer, skeletal osteotomies or arthrodeses) modalities. The goals of these surgeries are to rebalance muscle activity, restore range of motion, and realign the skeletal segments of the foot. Clinical decision-making for the management of the ankle/foot in children with CP involves a diagnostic matrix utilizing data from the clinical history, physical examination, radiographic imaging, and quantitative gait analysis. Keywords

Cerebral palsy • Foot • Ankle • Management

Contents Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Normal Gait . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Disruption of Foot, Ankle Function, and Its Effect on Gait in Cerebral Palsy . . . . . . . . . . . . . . . Clinical Decision-Making: Management of Foot and Ankle Problems in CP . . . . . . . . . . . . . . . . Clinical History . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Physical Examination . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Radiographic Evaluation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

1122 1122 1130 1133 1133 1134 1134

J.R. Davids (*) • S.A. Tabaie Northern California Shriner’s Hospital for Children, Sacramento, Sacramento, CA, USA e-mail: [email protected]; [email protected] # Springer International Publishing AG, part of Springer Nature 2018 B. Müller, S.I. Wolf (eds.), Handbook of Human Motion, https://doi.org/10.1007/978-3-319-14418-4_58

1121

1122

J.R. Davids and S.A. Tabaie

Quantitative Gait Analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Levels of Deformity . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Treatment Principles . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Surgical Treatment Goals . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Surgical Treatment Techniques . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

1135 1137 1138 1138 1140 1147 1148

Introduction Foot and ankle problems are common in children with cerebral palsy (CP). They affect both the ambulatory and nonambulatory children either by compromising the efficiency of gait or leading to issues with orthotic and shoe wear (Davids 2010). To approach the management of foot and ankle problems in a child with CP, it is important to understand normal gait and how it is disrupted in children with CP. This chapter will discuss normal gait at the foot and ankle and the disruption of normal gait in CP and then present the current paradigm for clinical decision-making for surgical management about the foot and ankle in children with CP.

Normal Gait The interaction between the ankle, foot, and the floor is the basis of normal gait. The function of the foot and ankle is based on a complex interplay between anatomy, physiology, and physics. Proper leads to efficient function of the knee and hip during gait (Davids et al. 2007, Inman 1969; Inman et al. 1981; Perry 1992). Standardized, consistent terminology should be used to describe the alignment of the separate segments of the ankle and foot (Ponseti et al. 1981). The understanding of foot and ankle function during normal gait is facilitated by considering the lower leg as consisting of four segments and the foot as three segments (Inman 1966; Inman et al. 1981; Perry 1992; Davids 2009a, 2010, 2016; Fig. 1): • • • •

Tibial or shank Hindfoot (talus and calcaneus) Midfoot (navicular, cuneiforms, and cuboid) Forefoot (metatarsals and phalanges) The foot should also be considered as consisting of two columns (Fig. 2):

• Medial column (talus, navicular, cuneiforms, great toe metatarsal, and phalanges) • Lateral column (calcaneus, cuboid, lesser toe metatarsals, and phalanges) Movement of the segments should be described as (Fig. 3):

Foot and Ankle Motion in Cerebral Palsy

1123

Fig. 1 Three segments of the foot. (a) Diagram of the AP view of the foot. The hindfoot segment is red, the midfoot segment is yellow, and the forefoot segment is white. (b) Diagram of the lateral view of the foot. The hindfoot segment is red, the midfoot segment is yellow, and the forefoot segment is white

Fig. 2 Two columns of the foot. (a) Diagram of the AP view of the foot. The medial column is red, and the lateral column is yellow. (b) Diagram of the lateral view (medial side) of the foot. The medial column is red. (c) Diagram of the lateral view (lateral side) of the foot. The lateral column is yellow

1124

J.R. Davids and S.A. Tabaie

Fig. 3 (a) Terminology for describing foot segmental alignment, utilizing the hindfoot as an example. When the plantar aspect of the hindfoot is deviated toward the body’s midline (red arrow), the segment is described as being inverted. When the plantar aspect is deviated away from the midline (yellow arrow), the segment is described as being everted. (b) When the distal aspect of the hindfoot is deviated toward the midline (red arrow), the segment is described as being adducted. When the distal aspect is deviated away from the midline (yellow arrow), the segment is described as being abducted

• Plantar aspect of segment – Inversion (toward the midline) – Eversion (away from the midline) • Distal portion of segment – Adduction (toward the midline) – Abduction (away from the midline) • Multi-planar motion – Supination (inversion and adduction) – Pronation (eversion and abduction) • Rotation of the segment about its longitudinal axis – External rotation (away from midline) – External rotation (away from midline) Function of the foot and ankle during the gait cycle is best appreciated by consideration of kinematics, kinetics, EMG (electromyography), and skeletal alignment (Fig. 4).

Foot and Ankle Motion in Cerebral Palsy

1125

Fig. 4 Illustration of kinematics, moment, and EMG throughout the gait cycle (IC initial contact, MSt midstance, TSt terminal stance, PSw preswing, ISw initial swing, MSw midswing, TSw terminal swing)

The interaction of the foot and ankle with the floor during the stance phase is best understood utilizing the concept of three rockers (Inman et al. 1981; Perry 1992; Davids 2009a, 2010). • First or heel rocker (Fig. 5) – Occurs from heel strike to foot flat during the loading response subphase of stance. – As the body progresses forward over the foot, the ground reaction force moves progressively distally through the foot, creating an external dorsiflexion moment. – The tibia advances forward over the foot, which is achieved by ankle dorsiflexion.

1126

J.R. Davids and S.A. Tabaie

Fig. 5 The first or heel rocker. The clinical photograph shows a heel strike at initial contact. The normal ankle kinematics, kinetics, and dynamic EMG during the gait cycle are plotted below the clinical photograph

35.0 Dorsiflexion Ankle Angle (degrees) 7.5 Plantar Flexion –20.0 2.0 Plantar Flexor Moment (Nm/kg)

0.8

Dorsiflexor –0.5 Gastrocnemius

Tibialis Anterior

0

25 50 75 % Gait Cycle

100

– This motion is controlled by eccentric activity of the ankle plantar flexor muscle group, which generates an internal plantar flexion moment. – This provides shock absorption during loading response. • Second or ankle rocker (Fig. 6) – Occurs as the tibia advances over the foot during the midstance subphase of stance. – With forward progression, the ground reaction force remains distal to the ankle joint, creating an increasing external dorsiflexion moment.

Foot and Ankle Motion in Cerebral Palsy

1127

Fig. 6 The second or ankle rocker. The clinical photograph shows the tibia advancing over the foot in midstance. The normal ankle kinematics, kinetics, and dynamic EMG during the gait cycle are plotted below the clinical photograph

35.0 Dorsiflexion Ankle Angle (degress) 7.5 Plantar Flexion –20.0 2.0 Plantar Flexor Moment (Nm/kg)

0.8

Dorsiflexor –0.5 Gastrocnemius

Tibialis Anterior

0

25 50 75 % Gait Cycle

100

– This motion is controlled by the concentric activity of the ankle plantar flexor muscle group, which continues to generate an internal plantar flexion moment. – This provides stability during midstance. • Third or forefoot rocker (Fig. 7) – Prior to the initial contact of the opposite foot, the heel of the reference foot rises off the ground, and dorsiflexion occurs through the metatarsophalangeal joints of the forefoot.

1128

J.R. Davids and S.A. Tabaie

Fig. 7 The third or forefoot rocker. The clinical photograph shows the heel rising and toes dorsiflexing in terminal stance. The normal ankle kinematics, kinetics, and dynamic EMG during the gait cycle are plotted below the clinical photograph

35.0 Dorsiflexion Ankle Angle (degress) 7.5 Plantar Flexion –20.0 2.0 Plantar Flexor Moment (Nm/kg)

0.8

Dorsiflexor –0.5 Gastrocnemius

Tibialis Anterior

0

25 50 75 % Gait Cycle

100

– The ankle starts to plantar flex as the reference limb is unloaded during preswing subphase of stance. – Largest moment generated by any single muscle group during the gait cycle is the internal plantar flexion moment generated by the plantar flexor group during third rocker in terminal stance. – This is essential for normal gait as it provides a rigid lever (i.e., stability) during terminal stance.

Foot and Ankle Motion in Cerebral Palsy

1129

Normal gait is the consequence of couple movements between the segments of the foot described earlier in this chapter. The alignments of the ankle and foot, and the activity of the ankle plantar flexor muscles, determine the location of the ground reaction force about the knee in mid and terminal stance which influences knee alignment. In general, ankle plantar flexion is associated with knee extension, and ankle dorsiflexion is associated with knee flexion. This biomechanical relationship is known as the plantar flexion/knee extension couple, and it is through this mechanism that control of the foot and ankle alignment can contribute to the control of knee alignment in stance phase (Inman et al. 1981; Inman 1966; Perry 1992; Davids 2009a; Fig. 8). In the swing phase of normal gait, the foot and ankle contribute to clearance and pre-positioning for the subsequent stance phase. • Initial swing – The combination of hip flexion (active), knee flexion (passive), and ankle dorsiflexion (active) serves to functionally shorten the limb to promote clearance. – The tibia or shank segment is externally rotating, and the ankle is plantarflexing.

Fig. 8 The plantar flexion/ knee extension couple. In the midstance of subphase of the gait cycle, increasing ankle plantar flexion (red arrow) shifts the ground reaction force (red arrow) anterior to the knee joint center, which causes the knee to extend. Increasing ankle dorsiflexion (black arrow) shifts the ground reaction force (black) behind the knee joint center, which causes the knee to flex

1130

J.R. Davids and S.A. Tabaie

• Midswing – The tibia or shank segment is rotating internally and the ankle is dorsiflexing. – Maximum shortening is achieved and clearance within 1 cm of the ground (proprioception) occurs. • Terminal swing – The coupled movements described in midswing continue, and the foot is maintained in a plantigrade position, perpendicular to the anatomical axis of the tibia or shank segment. This pre-positioning of the foot during terminal swing will result in a heel strike at the initial contact.

Disruption of Foot, Ankle Function, and Its Effect on Gait in Cerebral Palsy Foot and ankle deformities in children with CP are the consequence of a dynamic imbalance between the extrinsic muscles of the lower leg that control segmental foot and ankle alignment (Davids 2009b, 2010). The causes of imbalance are linked to: • Spasticity • Disrupted motor control • Impaired balance function Most often, the ankle plantar flexor muscles are overactive, and the ankle dorsiflexor muscles are ineffective. Additional imbalance patterns can be evident between the muscle groups that control foot and ankle supination and pronation. Ultimately, these muscle imbalances can lead to three common coupled foot and ankle segmental malalignment patterns in children with spastic-type CP. Other, more complex, or uncoupled segmental malalignment may occur but are much less common (Davids 2009b, 2010; Stevens 1988; Mosca 1995; Etnyre et al. 1993; Tylkowski et al. 2009). • Equinus deformity (Fig. 9) – Characterized by excessive plantar flexion of the hindfoot relative to the ankle. – In a purely equinus deformity, there is normal midfoot and forefoot alignment. • Equinoplanovalgus deformity (Fig. 10) – Characterized by equinus deformity of the hindfoot, along with pronation deformities of the midfoot and forefoot. – In this deformity pattern, the lateral column of the foot is shorter (anatomically and functionally) than the medial column. – Associated deformities of ankle and/or hallux valgus are commonly seen with equinoplanovalgus foot segmental malalignment. • Equinocavovarus deformity (Fig. 11) – Characterized by equinus deformity of the hindfoot, along with supination deformity of the midfoot and variable malalignment of the forefoot. – In this deformity pattern, the lateral column is functionally and/or anatomically longer than the medial column.

Foot and Ankle Motion in Cerebral Palsy

1131

Fig. 9 Plain radiographs of the foot in a child with equinus deformity. (a) AP view shows normal segmental alignment. (b) Lateral view shows hindfoot plantar flexion (diminished calcaneal pitch, normal is 25–30 ), with otherwise normal segmental alignment

Fig. 10 Plain radiographs of the foot in a child with equinoplanovalgus deformity. (a) AP view shows hindfoot pronation, talonavicular uncoverage, and forefoot abduction. (b) Lateral view shows hindfoot plantar flexion, midfoot pronation, and forefoot pronation

1132

J.R. Davids and S.A. Tabaie

Fig. 11 Plain radiographs of the foot in a child with equinovarus deformity. (a) AP view shows hindfoot supination and forefoot adduction. (b) Lateral view shows hindfoot varus, midfoot supination, and forefoot supination

– Associated deformities of the ankle may be seen with equinocavovarus segmental foot malalignment. The three common malalignment patterns described above are often correctable on manipulation in younger children with milder forms of CP. However, the foot and ankle deformities can become rigid and uncorrectable due to the following factors: • Increasing age and growth • Fixed shortening or myostatic deformities of the muscles • Development of permanent skeletal deformities in response to the malalignment pattern Foot and ankle segmental malalignment in children with CP can alter function during both the stance and swing phases of the gait cycle. The three segmental malalignment patterns are associated with a variety of gait disruptions and deviations (Davids 2010). • Common alterations to the normal gait cycle seen with all three segmental malalignment patterns: – Lack of heel strike at initial contact, disrupting the first rocker and shock absorption function in the loading response of the gait cycle – Compromised ankle plantar flexor muscles ability to generate an adequate internal plantar flexor moment during the second and third rockers

Foot and Ankle Motion in Cerebral Palsy

1133

– Inhibition of ankle dorsiflexion in swing phase, compromising clearance in midswing and proper positioning of the foot and ankle in terminal swing • Common alterations to the normal gait cycle seen with equinus and equinocavovarus malalignment patterns: – Disruption of the second rocker by blocking ankle dorsiflexion, thereby compromising stability in midstance. – Due to the hindfoot malalignment with these two deformity patterns, the lengths of the plantar flexor muscles are shortened, and their ability to generate tension is adversely affected. • Common alterations of the normal gait cycle with equinoplanovalgus segmental malalignment: – In this deformity pattern, the midfoot and forefoot segments remain in an unlocked alignment leading to excessive loading of the plantar, medial portion of the midfoot. – The malalignment of the midfoot and forefoot further compromises the momentgenerating capacity of the ankle plantar flexor muscles and effectively shortens the lever arm available to this muscle group during the third rocker. – External tibial torsion seen with equinoplanovalgus malalignment may contribute to an external foot progression angle, further affecting the lever arm available to the ankle plantar flexor muscles in terminal stance.

Clinical Decision-Making: Management of Foot and Ankle Problems in CP The management of foot and ankle deformities in CP can be very challenging. In order to provide the best care for this patient population, it is best to utilize a diagnostic matrix that incorporates the following (Davids 2009b, 2010; Davids et al. 2005, 2007; Inman et al. 1981; Perry 1992): • • • •

Clinical history Physical examination including observational gait analysis Plain radiographs Quantitative gait analysis

Clinical History Often patients present with specific complaints of gait abnormalities, such as pain with ambulation, tripping when ambulating, and/or issues with shoe wear. It is imperative to inquire and document the specifics of the symptoms, inciting and

1134

J.R. Davids and S.A. Tabaie

alleviating factors, duration of symptoms and their progression over time, and finally the treatment history to date.

Physical Examination The sequence of the physical exam in children with CP who have foot and ankle deformities should begin with a thorough inspection and progress to a hands-on evaluation of the foot segmental alignment in both weight-bearing and non-weightbearing conditions. Inspection of the plantar aspect of the feet should be done to evaluate for the presence of excessive or inadequate skin callous formation that could be due to abnormal loading patterns and/or problems with shoe/orthotic wear. Manual examination is performed to determine and document: • • • • •

Static standing alignment – assessed from the front, back, and both sides of the patient Intra- and intersegmental flexibility Rotational profile from the hips to the foot Range of motion – passive and active Muscle strength and selective motor control

The final key portion of the physical exam is the observational gait analysis. Observing ambulation needs to be done from multiple viewpoints in both the coronal and sagittal planes. The patient should be barefoot and wearing tight cycling style shorts in order for the examiner to adequately visualize the entire lower extremity including the thigh, knee, lower leg, ankle and foot. When observing a patient’s gait, the following events of the gait cycle related to dynamic foot function should be documented: • Foot position at initial contact – heel strike, flatfoot, or toe strike • Foot alignment in midstance – varus/valgus in the coronal plane and internal/ external in the transverse plane (foot progression angle) • Foot alignment at toe-off – varus/valgus in the coronal plane and dorsi-/plantarflexed in the sagittal plane • Foot clearance in swing phase

Radiographic Evaluation Radiographic images for evaluation of foot and ankle deformities must be routinely performed with the child fully weight bearing. Non-weight-bearing images are not of value in assessing foot segmental alignment and should not be utilized. Three standardized views should always be obtained: • AP foot • Lateral foot • AP ankle

Foot and Ankle Motion in Cerebral Palsy

1135

Table 1 Radiographic measurements of normative values for feet in children Radiographic measurements: quantitative and categorical definitions Normal (mean  Abnormal high value 3.75) (> mean +1 SD) Hindfoot Tibiotalar angle (degrees) 1.1  3.75 Eversion Calcaneal pitch (degrees) 17  6.0 Calcaneus Tibio-calcaneal angle 69  8.4 Equinus (degrees) Talocalcaneal angle (degrees) 49  6.9 Eversion Midfoot Naviculo-cuboid overlap (%) 47  13.8 Pronation Talonavicular coverage angle 20  9.8 Abduction (degrees) Lateral talo-first metatarsal 13  7.5 Pronation angle (degrees) Forefoot Anteroposterior talo-first 10  7.9 Abduction metatarsal angle (degrees) Metatarsal stacking angle 82  9 Supination (degrees) Columns Medial-lateral column ratio 0.9  0.3 Abduction

Abnormal low value (> mean +1 SD) Inversion Equinus Calcaneus Inversion Supination Adduction Supination

Adduction Pronation

Adduction

When analyzing radiographs and designating deformities, it is best to divide the foot into three segments and two columns as was described earlier in this chapter, then determining the relative alignment of each segment and the relative length of each column. A comprehensive technique of quantitative segmental analysis of the foot and ankle has been developed by Davids et al. (2005) and was derived from the foot model originally developed by Inman and colleagues (Inman et al. 1981). The technique of quantitative segmental analysis was developed from normative values of a cohort of 60 normal feet in children between ages of 5 and 17 years old, utilizing ten radiographic measurements to determine the alignment of the segments and the lengths of the columns of the foot and ankle (Davids et al. 2005; Table 1).

Quantitative Gait Analysis The calculation of foot and ankle kinematics and kinetics involves modeling assumptions and approximations concerning the relationship between the skin markers and the underlying skeletal anatomy. The standard ankle and foot model most commonly used in clinical gait analysis was developed in the 1980s, uses

1136

J.R. Davids and S.A. Tabaie

markers at the malleoli and forefoot, and considers the foot as a single segment (Perry 1992). It is assumed that the foot segment is rigid from the hindfoot to the forefoot. Ankle motion in the sagittal plane is calculated from the location of the foot axis relative to the tibial axis. Any movement between the three segments of the foot (i.e., hindfoot to midfoot, midfoot to forefoot) that occurs between the malleolar and forefoot markers is captured by this simple foot model and described as ankle motion. Significant measurement artifact occurs when the normal foot segmental alignment is disrupted (i.e., equinoplanovalgus foot malalignment in children with CP). This artifact creates apparent discrepancies within the diagnostic matrix between the data derived from the physical examination, observational gait analysis, and quantitative gait analysis. Failure to appreciate the causes for these apparent discrepancies may result in confusion for clinicians and compromise clinical decision-making. Technological improvements have allowed for the development of more sophisticated, multisegment foot models that more accurately approximate the complex anatomy and biomechanics of the foot (MacWilliams et al. 2003). However, these models are difficult to apply to children with CP, because of small foot size and deformity. Additionally, the dynamic EMG, which is part of the quantitative gait analysis, is very relevant in the evaluation foot deformities in children with CP (Sutherland 1993). The incorporation and use of surface/fine-wire EMG provides information on the timing of muscle activity during the gait cycle (Hoffer et al. 1985; Scott and Scarborough 2006). This data is imperative in determining: • The relative activity of tibialis anterior and posterior muscles in both stance and swing phases • Selection of a particular muscle-tendon unit for lengthening or transfer The final component of quantitative gait analysis is dynamic pedobarography, which measures the spatial and temporal distribution of force over the plantar aspect of the foot during the stance phase of the gait cycle (Jameson et al. 2008; Davids 2009a). Pedobarography provides the following information regarding the patient’s dynamic foot function: • Foot contact patterns • Foot pressure distribution and magnitude • Progression of the center of pressure through the foot As described above, foot function during gait in children with CP is disrupted by several patterns of skeletal segmental malalignment that can ultimately affect all three rockers in stance phase. This biomechanical disruption has been termed lever arm deficiency and is best characterized by the center of pressure progression (COPP) relative to the foot. Deviations in the location and duration of the COPP relative to the segments of the foot when compared to normative values of COPP in children can be used to describe common abnormal loading patterns (Davids 2009a; Fig. 12).

Foot and Ankle Motion in Cerebral Palsy

1137

Fig. 12 Examples of abnormal foot loading patterns (each figure is of a left, as if looking down from above, toes at the top, lateral to the left, medial to the right): (a) displacement of the center of pressure progression (COPP – solid red line) medially, describing a valgus loading pattern. (b) Displacement of the COPP laterally, resulting in a varus loading pattern. (c) Displacement of COPP distally, describing an equinus loading pattern). (d) Displacement of the COPP proximally (i.e., excessive duration of the COPP in the hindfoot segment, reflected by increased thickness of the red line) resulting in a calcaneus loading pattern

• Valgus loading pattern – Displacement of the COPP medially due to an everted, abducted, or pronated segmental malalignment of the foot segment • Varus loading pattern – Displacement of the COPP laterally, which is usually due to an inverted, adducted, or supinated segmental malalignment of the foot segment • Equinus loading pattern – Prolonged duration of the COPP in the forefoot segment • Calcaneus loading pattern – Prolonged duration of the COPP in the hindfoot segment

Levels of Deformity Foot and ankle deformities in children with CP are sequential and progressive with growth and development. The deformities can be classified into three levels (Davids 2010): • Level I – Dynamic soft tissue imbalance with no skeletal pathoanatomy

1138

J.R. Davids and S.A. Tabaie

• Level II – Fixed or myostatic soft tissue imbalance with flexible and correctable skeletal segmental malalignment • Level III – Fixed myostatic soft tissue imbalance with associated structural skeletal deformities Understanding the levels of deformity in children with CP can guide in both the nonoperative and surgical interventions for these patients.

Treatment Principles Management of foot and ankle deformities in children with CP can consist of both nonsurgical and surgical modalities. Treatment options include (Davids et al. 2004, 2007; Gage 1994, 1995, 2004): • Orthotics – May be used to protect the outcome of a surgical procedure during the healing and rehabilitation phases, to prevent the development or worsening of musculoskeletal deformities with growth, and to improve gait. – When the goal is to improve gait, the physician should clearly identify the gait deviations and functional deficits to be addressed by the orthosis. – Clinicians must be familiar with the biomechanical characteristics and clinical indicators of six common orthotic designs (Table 2): 1. Foot orthosis 2. Supramalleolar orthosis 3. Posterior leaf spring orthosis 4. Articulating ankle-foot orthosis 5. Floor-reaction ankle-foot orthosis 6. Solid ankle-foot orthosis • Pharmacologic/Neurosurgical – Level I deformities resulting from dynamic soft tissue imbalance with no skeletal deformities can be treated with interventions to manage muscle tone and spasticity (Preiss et al. 2003; Boyd et al. 2000; Bjornson et al. 2007). 1. Pharmacologic – botulinum toxin (or phenol) injection 2. Neurosurgical – selective dorsal rhizotomy or intrathecal baclofen – Early management with the abovementioned options is favored to avoid development of fixed deformities of the muscle-tendon unit.

Surgical Treatment Goals Surgical management to correct foot and ankle deformities in children with CP may be selected to improve gait or function, shoe wear, and cosmesis. These goals may be

Dorsiflexion to neutral

Dorsiflexion to neutral thigh-foot angle 15 extension

Normal

Normal

Normal

Extension  20

Normal

Normal

Normal

Extension  30

Articulating AFO

Solid AFO

FRAFO

Moderate, partially correctible

Moderate, partially correctible

Mild, correctible

Foot Mild, correctible Mild, correctible Mild, correctible

Absent heel strike

Absent heel strike Absent heel strike Absent heel strike

Normal

Increased plantar flexion, increased knee extension (mild) Increased plantar flexion, increased knee extension, or increased knee flexion (mild) Increased dorsiflexion, increased knee flexion, increased hip flexion

Normal

Normal

Gait deviation Initial contact Midstance Normal Normal

Increased dorsiflexion

Increased plantar flexion

Normal

Normal

Normal

Terminal stance Normal

Increased dorsiflexion

Increased plantar flexion Increased plantar flexion Increased plantar flexion

Normal

Swing phase Normal

UCBL University of California at Berkeley Laboratory, PLSO posterior leaf spring orthotic, AFO ankle-foot orthotic, FRAFO floor reaction ankle-foot orthotic

Dorsiflexion to 5

Dorsiflexion to 5

Normal

Normal

Normal

Supramalleolar orthosis PLSO

Ankle Normal

Knee Normal

Hip Normal

Orthosis UCBL

Indications for use of orthoses to improve gait in children with cerebral palsy Physical exam

Table 2 Indications for the use of common orthoses to improve gait in children with CP

Foot and Ankle Motion in Cerebral Palsy 1139

1140

J.R. Davids and S.A. Tabaie

achieved by surgical procedures that are designed to improve foot shape (Davids 2010; Sutherland 1993; Mosca 1998). • Improved foot shape following soft tissue and skeletal surgery can restore the stability function of the foot during: – The second rocker in midstance and the skeletal lever arm function of the foot during the third rocker in terminal stance • Cosmetic improvements following foot surgery are related to: – Improved visual assessment of static standing foot alignment – restoration of the medial longitudinal arch and toe alignment – Improved foot progression angle during stance phase Additionally, it is presumed that surgically improved foot shape can correct pain by improving foot loading and stability in stance phase. In regard to pain, it is important to make the distinction between the younger children with CP who may tolerate mild or moderate foot deformities versus the older teenage/adult population. Deformities that may be tolerated in a younger CP population are often poorly tolerated in teenage and adult life as the body weight increases, leading to a greater magnitude of the abnormal loading. The cumulative effect of the magnitude of abnormal loading results in premature degenerative changes of the joints of the foot and ankle. Presumptively, surgery to improve foot shape in childhood will improve the loading of the foot and decrease the possibility of early degenerative arthritis in adulthood.

Surgical Treatment Techniques As noted earlier in this chapter, there are three common coupled foot and ankle segmental malalignment patterns in children with spastic-type CP: equinus, equinoplanovalgus, and equinocavovarus. Additionally, hallux valgus is a common secondary malalignment that may be associated with any of the three principle malalignment patterns. The surgical options for these coupled segmental deformities are based on the level of deformity and will be discussed below. • Equinus – Pure plantar flexor malalignment is usually the consequence of overactivity (level I) or tightness (level II) of the ankle plantar flexor muscle group. Assessment of foot segmental alignment with plain radiographs is essential to establishing that there are no associated deformities at the level of the midor forefoot (Davids 2010). – Surgical management starts with level II deformities. – Careful assessment is required to determine the relative contributions of the gastrocnemius and soleus muscles to the fixed shortening of the plantar flexor group. 1. This is best achieved by assessing ankle dorsiflexion range of motion with the knee both flexed and extended.

Foot and Ankle Motion in Cerebral Palsy

1141

2. Isolated limitation with the knee flexed suggests involvement of the soleus muscle, limitation with the knee extended suggests involvement of the gastrocnemius muscle, and limitation regardless of knee position represents involvement of both muscles. – The goal of surgical lengthening of the plantar flexor group is to achieve five degrees of dorsiflexion when the knee is extended. 1. When only the gastrocnemius is involved, a selective fractional lengthening is best performed proximally at the level of the muscle belly. 2. When both muscles are involved and 15 or less of correction is required, selective fractional lengthening mid-calf at the level of the myotendinous junction is preferred. 3. When both muscles are involved and greater than 15 of correction is required, nonselective lengthening distally at the Achilles tendon level is necessary. – In most cases, a fractional lengthening at the level of the muscle belly or myotendinous junction is sufficient. However, with proper patient selection and careful surgical technique, all three techniques described above may be effective and excessive lengthening and weakness of the ankle plantar flexor muscle groups can be avoided (Etnyre et al. 1993; Tylkowski et al. 2009). • Equinoplanovalgus – This malalignment pattern, which is often referred to as flatfoot, is the consequence of overactivity (level I) or tightness (level II) of the ankle plantar flexor and evertor muscle groups. Physical examination is needed to determine whether the muscle-tendon unit deformities are dynamic or mysostatic. Fixed skeletal segmental malalignment (level III) is assessed with plain radiographs which are essential for preoperative planning (Davids 2010). – As with equinus malalignment, in this pattern orthopedic surgical intervention starts with level II deformity and is usually seen in children between 4 and 7 years of age. 1. The treatment choice is lengthening of the ankle plantar flexor muscle group, as described earlier in the equinus section. 2. Additionally, transfer of the peroneus brevis muscle to the peroneus longus muscle should be performed when fixed deformity of the former is present. 3. At the time of surgery, it is essential to determine that normal foot segmental alignment has been restored following soft tissue surgeries (Fig. 13). – Failure to restore normal skeletal alignment with soft tissue surgery alone reclassifies the deformity to level III (Davids 2009b; Mosca 1996, 1998; Yoo et al. 2005; Danko et al. 2004). 1. The primary procedure used to correct level III equinoplanovalgus malalignment is lateral column lengthening, which can be performed at the following locations in the foot: (a) Neck of the calcaneus (b) Calcaneocuboid (CC) joint (c) Body of the cuboid 2. Lateral column lengthening usually allows for correction of all three segments of the foot, most likely secondary to ligamentotaxis (Fig. 14). (a) The lengthening is approximately 1–2.5 cm. (b) Interposition grafting with tricortical iliac crest or allograft.

1142

J.R. Davids and S.A. Tabaie

Fig. 13 The technique for the intraoperative stress assessment of foot segmental alignment. The assistant stabilizes the knee while the surgeon loads the foot and ankle with the foot pusher. A fluoroscopic image is taken in the loaded position and is used to assess the segmental alignment

(c) Lengthening through the CC joint (arthrodesis) is often done in children over 12 years of age. Internal fixation should be used when the graft size is greater than 1.5 cm. in order to promote early mobilization and minimize late graft collapse during phases of healing. 3. Following lateral column lengthening for a level III equinoplanovalgus foot deformity, assessment of the medial column is performed. (a) If there is residual forefoot varus deformity, or if the medial column is hypermobile in the sagittal plane (i.e., unmasked deformities), then a plantar flexion osteotomy (or a combination of both) is made in the medial cuneiform or base of the great toe metatarsal (if the child is skeletally mature. (b) If there is residual forefoot abduction deformity (i.e., incomplete correction), then a talonavicular arthrodesis needs to be performed. (c) If lengthening of the lateral column fails to correct the hindfoot deformity, then arthrodesis of the subtalar, talonavicular, and calcaneocuboid joints are required to achieve optimal alignment of the foot. Fortunately, this is rarely required, even for feet with significant malalignments. • Equinocavovarus – This malalignment pattern that is often referred to as varus foot is usually the consequence of overactivity (level I) or tightness (level II) of the ankle plantar flexor and invertor muscle groups. As with the other malalignment patterns, physical examination is needed to determine whether the muscle-tendon unit deformities are dynamic or mysostatic. Fixed skeletal

Foot and Ankle Motion in Cerebral Palsy

1143

Fig. 14 Correction of moderate level III equinoplanovalgus deformity. (a, b) Lateral and AP radiographs in a child with CP showing equinoplanovalgus segmental malalignment. (c, d) Lateral and AP radiographs of the foot after gastrocsoleus fractional lengthening and lateral column lengthening through the neck of the calcaneus

segmental malalignment (level III) is assessed with plain radiographs and is essential for preoperative planning (Davids 2010). – In children over the 6 years of age with equinocavovarus foot malalignment, surgical management with muscle-tendon transfer can be an option. 1. Information from the physical exam, kinematics, kinetics, dynamic EMG, and pedobarography are used to determine the relative contributions of the tibialis anterior and tibialis posterior muscles to the dynamic varus deformity that occurs during stance and swing phases (Sutherland 1993; Scott and Scarborough 2006; Hoffer et al. 1985).

1144

J.R. Davids and S.A. Tabaie

(a) Split transfer of the tibialis anterior muscle should only be performed when there is ankle dorsiflexion appreciated in midstance. (b) Split transfer of the tibialis posterior should only occur when the dynamic EMG shows the timing of the activation of this muscle corresponds with the presence of varus malalignment during specific subphases of the gait cycle. (c) When kinematics and dynamic EMG assessments are not available, dynamic varus deformity is best treated by concomitant split transfer of the tibialis anterior muscle and fractional lengthening of the tibialis posterior muscle (Barnes and Herring 1991). – Level II deformity involves sequential correction of the hindfoot and midfoot soft tissue contractures (Fig. 15). 1. Correction of the deformity is performed by fractional lengthening of the following muscle groups: (a) Ankle plantar flexor muscle group

Fig. 15 Technique for the correction of level II equinocavovarus deformity. (a) Three incisions are used to lengthen the appropriate soft tissue structures. (b) The plantar fascia and intrinsic muscles of the foot are released through the plantar incision. (c) Fractional lengthening of the abductor hallucis muscle is performed through the distal medial incision. (d) Fractional lengthening of the gastrocsoleus muscle group is performed through the medial calf incision. (e) Fractional lengthening of the tibialis posterior muscle is performed through the same medial calf incision. The flexor digitorum and hallucis longus muscles can also be lengthened, when necessary, though this incision

Foot and Ankle Motion in Cerebral Palsy

1145

(b) Tibialis posterior muscle (c) Flexor hallucis and digitorum longus muscles – rarely necessary (d) Abductor hallucis (e) Plantar fascia and short intrinsic muscle of the foot 2. At the time of surgery, it is essential to determine the normal foot segmental alignment has been restored following the soft tissue surgeries. (a) Confirmation of normal alignment is best achieved with intraoperative stress radiographs of the foot as was described earlier in the chapter. (b) Failure to restore normal skeletal alignment should result in reclassification of the deformity to level III. – Correction of level III equinocavovarus segmental malalignment differs from correction of equinoplanovalgus malalignment, because there is not one single skeletal procedure that can achieve adequate correction of all three segments of the foot (Fig. 16).

Fig. 16 Correction of level III equinocavovarus deformity. (a, b) Lateral and AP radiographs of the foot in a child with CP showing equinocavovarus segmental malalignment. (c, d) Lateral and AP radiographs of the foot after gastrocsoleus fractional lengthening, posterior tibialis fractional lengthening, radical plantar fascia release, calcaneal slide osteotomy, and medial column osteotomy through the cuneiform

1146

J.R. Davids and S.A. Tabaie

1. Gross alignment and dynamic loading of the foot may be greatly improved by performing sequential osteotomies that create deformities to compensate for segmental malalignments. (a) Residual hindfoot varus malalignment may be corrected by calcaneal slide or laterally based closing wedge osteotomies (Koman et al. 1993). (b) Residual midfoot supination deformity may be corrected by lateral column shortening through the cuboid. (c) Residual forefoot pronation may be corrected by dorsiflexion osteotomy of the medial column. (d) Residual forefoot supination deformity may be corrected by plantar flexion osteotomy of the medial column. 2. Arthrodesis strategy is reserved for the most severe cases of level III equinocavovarus segmental malalignment. (a) Double arthrodesis of the calcaneocuboid and talonavicular joints may be required to correct extreme midfoot cavus deformity. (b) Triple arthrodesis of the subtalar, calcaneocuboid, and talonavicular joints may be required to achieve optimal foot alignment. • Hallux Valgus – This deformity in children with CP may be the consequence of intrinsic and/or extrinsic factors. The gait deviations secondary to these factors tend to medialize the forces across the great to metatarsophalangeal (MTP) joint, resulting in an external abduction or valgus moment (Davids 2010). – Radiographic assessment is essential for determining the elements and magnitude of deformity at the great to MTP joint and the presence of associated foot segmental malalignments. – Correction of the deformity is indicated to address the following issues: 1. Pain at the great toe MTP joint 2. Treat hygiene problems related to toe positioning 3. Facilitate shoe and orthotic wear – For hallux valgus deformities in this patient population, there is little evidence to support lengthening or release to correct or improve foot intrinsic muscle tightness (level II deformity). 1. Poor results for soft tissue balancing procedures are most likely a consequence of the failure to address significant extrinsic causes of hallux valgus deformity in children with CP. – The preferred treatment of level III hallux valgus deformity in children with CP is great toe MTP arthrodesis (Davids et al. 2001; Bishay et al. 2009; Fig. 17). 1. Due to the proximal location of the physis of the proximal phalanx, this procedure should not be performed in children who have more than 2 years of growth remaining. 2. The shape of the foot and the child’s gait pattern determines optimal alignment of the MTP arthrodesis.

Foot and Ankle Motion in Cerebral Palsy

1147

Fig. 17 Correction of hallux valgus deformity (in association with comprehensive foot segmental malalignment correction with lateral column lengthening through the neck of the calcaneus). (a) Standing AP radiograph of the foot in a child with CP. The great toe deformity consists of metatarsus varus and MTP joint valgus. (b) Intraoperative fluoroscopy AP showing great toe MTP arthrodesis. (c) Standing AP radiograph of the foot 1 year after great toe arthrodesis (note the indirect correction of the increased inter-metatarsal angle)

(a) Correction of the coronal plane should align the phalanges of the great toe with the lesser toes. Sagittal plane alignment of the arthrodesis should include between 15 and 20 of dorsiflexion (relative to the floor when the foot is weight bearing) to facilitate the forefoot rocker.

Summary • The normal interaction of the ankle/foot with the floor is best described by the three rockers (heel, ankle, and forefoot). • There are three common ankle/foot segmental malalignment patterns seen in children with CP: equinus, equinoplanovalgus, and equinocavovarus. • Each type of ankle/foot malalignment can be classified into three levels based upon the presence of dynamic muscle imbalance, fixed or myostatic deformity of the muscle-tendon unit, and fixed skeletal deformities. • Clinical decision-making for the management of the ankle/foot in children with CP involves a diagnostic matrix utilizing data from the clinical history, physical examination, radiographic imaging, and quantitative gait analysis.

1148

J.R. Davids and S.A. Tabaie

• Common orthopedic surgical interventions used to address ankle/foot problems in children with CP include muscle-tendon unit lengthening or transfer and skeletal osteotomies or arthrodeses. The goals of these surgeries are to rebalance muscle activity, restore range of motion, and realign the skeletal segments of the foot. Acknowledgments Sean Brown, M.A. for assistance in preparation of figure.

References Barnes MJ, Herring JA (1991) Combined split anterior tibial-tendon transfer and intramuscular lengthening of the posterior tibial tendon. Results in patients who have a varus deformity of the foot due to spastic cerebral palsy. J Bone Joint Surg Am 73(5):734–738 Bishay SN, El-Sherbini MH, Lotfy AA, Abdel-Rahman HM, Iskandar HN, El-Sayed MM (2009) Great toe metatarsophalangeal arthrodesis for hallux valgus deformity in ambulatory adolescents with spastic cerebral palsy. J Child Orthop 3(1):47–52 Bjornson K, Hays R, Graubert C, Price R, Won F, McLaughlin JF, Cohen M (2007) Botulinum toxin for spasticity in children with cerebral palsy: a comprehensive evaluation. Pediatrics 120(1):49–58 Boyd RN, Pliatsios V, Starr R, Wolfe R, Graham HK (2000) Biomechanical transformation of the gastroc-soleus muscle with botulinum toxin A in children with cerebral palsy. Dev Med Child Neurol 42(1):32–41 Danko AM, Allen B Jr, Pugh L, Stasikelis P (2004) Early graft failure in lateral column lengthening. J Pediatr Orthop 24(6):716–720 Davids JR (2009a) Normal function of the ankle and foot: biomechanics and quantitative analysis. In: McCarthy JJ, Drennan JC (eds) Drennan’s the child foot and ankle. Lippincott Williams and Wilkins, Philadelphia, pp 54–63 Davids JR (2009b) Orthopaedic treatment of foot deformities. In: Schwartz M, Gage J, Koop S, Novacheck T (eds) The identification and treatment of gait problems in cerebral palsy. MacKeith Press, London, pp 514–533 Davids JR (2010) The foot and ankle in cerebral palsy. Orthop Clin North Am 41(4):579–593 Davids JR (2016) Biomechanically based clinical decision making in pediatric foot and ankle surgery. In: Sabharwal S (ed) Pediatric lower limb deformities. Springer, Cham, pp 153–162 Davids JR, Mason TA, Danko A, Banks D, Blackhurst D (2001) Surgical management of hallux valgus deformity in children with cerebral palsy. J Pediatr Orthop 21(1):89–94 Davids JR, Ounpuu S, DeLuca PA, Davis RB 3rd (2004) Optimization of walking ability of children with cerebral palsy. Instr Course Lect 53:511–522 Davids JR, Gibson TW, Pugh LI (2005) Quantitative segmental analysis of weight-bearing radiographs of the foot and ankle for children: normal alignment. J Pediatr Orthop 25(6):769–776 Davids JR, Rowan F, Davis RB (2007) Indications for orthoses to improve gait in children with cerebral palsy. J Am Acad Orthop Surg 15(3):178–188 Davis RB, Jameson E, Davids JR, Christopher LM, Rogozinski BM, Anderson JP (2007) The design, development, and initial evaluation of a multisegment foot model for routine clinical gait analysis. In: Smith P, Harris GF, Marks R (eds) Foot and ankle motion analysis: clinical treatment and technology. CRC Press, Boca Raton, pp 425–444 Etnyre B, Chambers CS, Scarborough NH, Cain TE (1993) Preoperative and postoperative assessment of surgical intervention for equinus gait in children with cerebral palsy. J Pediatr Orthop 13(1):24–31 Gage JR (1994) The role of gait analysis in the treatment of cerebral palsy. J Pediatr Orthop 14(6):701–702 Gage JR (1995) The clinical use of kinetics for evaluation of pathologic gait in cerebral palsy. Instr Course Lect 44:507–515

Foot and Ankle Motion in Cerebral Palsy

1149

Gage JR (2004) Orthotics: a comprehensive clinical approach. The treatment of gait problems in cerebral palsy. MacKeith Press, London, pp 273–285 Hoffer MM, Barakat G, Koffman M (1985) 10-year follow-up of split anterior tibial tendon transfer in cerebral palsied patients with spastic equinovarus deformity. J Pediatr Orthop 5(4):432–434 Inman VT (1966) The human foot. Manit Med Rev 46(8):513–515 Inman VT (1969) The influence of the foot-ankle complex on the proximal skeletal structures. Artif Limbs 13(1):59–65 Inman VT, Ralston J, Todd F (1981) Human walking. Williams & Wilkins, Baltimore Jameson EG, Davids JR, Anderson JP, Davis RB 3rd, Blackhurst DW, Christopher LM (2008) Dynamic pedobarography for children: use of the center of pressure progression. J Pediatr Orthop 28(2):254–258 Koman LA, Mooney JF 3rd, Goodman A (1993) Management of valgus hindfoot deformity in pediatric cerebral palsy patients by medial displacement osteotomy. J Pediatr Orthop 13(2):180–183 MacWilliams BA, Cowley M, Nicholson DE (2003) Foot kinematics and kinetics during adolescent gait. Gait Posture 17(3):214–224 Mosca VS (1995) Calcaneal lengthening for valgus deformity of the hindfoot. Results in children who had severe, symptomatic flatfoot and skewfoot. J Bone Joint Surg Am 77(4):500–512 Mosca VS (1996) Flexible flatfoot and skewfoot. Instr Course Lect 45:347–354 Mosca VS (1998) The child’s foot: principles of management. J Pediatr Orthop 18(3):281–282 Perry J (1992) Gait analysis: normal and pathologic function. Slack, Thorofare Ponseti IV, El-Khoury GY, Ippolito E, Weinstein SL (1981) A radiographic study of skeletal deformities in treated clubfeet. Clin Orthop Relat Res (160):30–42 Preiss RA, Condie DN, Rowley DI, Graham HK (2003) The effects of botulinum toxin (BTX-A) on spasticity of the lower limb and on gait in cerebral palsy. J Bone Joint Surg Br 85(7):943–948 Scott AC, Scarborough N (2006) The use of dynamic EMG in predicting the outcome of split posterior tibial tendon transfers in spastic hemiplegia. J Pediatr Orthop 26(6):777–780 Stevens PM (1988) Effect of ankle valgus on radiographic appearance of the hindfoot. J Pediatr Orthop 8(2):184–186 Sutherland DH (1993) Varus foot in cerebral palsy: an overview. Instr Course Lect 42:539–543 Tylkowski CM, Horan M, Oeffinger DJ (2009) Outcomes of gastrocnemius-soleus complex lengthening for isolated equinus contracture in children with cerebral palsy. J Pediatr Orthop 29(7):771–778 Yoo WJ, Chung CY, Choi IH, Cho TJ, Kim DH (2005) Calcaneal lengthening for the planovalgus foot deformity in children with cerebral palsy. J Pediatr Orthop 25(6):781–785

The Arm Pendulum in Gait Jaques Riad

Abstract

Because of the physical length of the arms and the relatively large range of motion in the shoulder and elbow in gait, any deviation from normal is detected immediately and attracts attention. The deviation can consist of increased flexion in the elbow, decreased range of motion, movement out of phase with the lower extremities, and asymmetry between the movement of right and left arms, either in isolation or in many more or less noticeable combinations. Although human evolution means that we no longer walk on our arms, arm movement has impact on our stability, balance, and appearance while walking. In addition, we can carry things, make gestures, or do other things with the arms and hands while walking. Despite an evolution toward corticospinal control of arm and hand movements, quadrupedal limb coordination persists during locomotion. We do not think about how we coordinate our arms and legs when walking. It just happens. Individuals with deformity, limited range of motion, or movement disorders affecting the arm show a disturbance of the normal arm pendulum in gait. It can be difficult to understand the consequences of the primary pathology and the influence on the movement pattern as well as the possible development of compensation mechanisms. Studying the arm pendulum is important for diagnosis and treatment and to follow progression over time. In addition, our sensitivity to deviations from normal highlights the importance of arm movement for communication and appearance.

Keywords

Arm movement • Gait • Motion analysis • Upper extremity • Cerebral palsy • Symmetry J. Riad (*) Skaraborg Hospital Skövde, Skövde, Sweden e-mail: [email protected] # Springer International Publishing AG, part of Springer Nature 2018 B. Müller, S.I. Wolf (eds.), Handbook of Human Motion, https://doi.org/10.1007/978-3-319-14418-4_56

1151

1152

J. Riad

Contents Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . State of the Art . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Arm Pendulum and Gait . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Mechanical Effects of Arm Movement on Gait . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . How Information on Arm Pendulum Can Help in Medicine . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Psychological and Social Aspects of Deviation in Arm Pendulum . . . . . . . . . . . . . . . . . . . . . . . Future Directions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Cross-References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

1152 1153 1155 1155 1156 1160 1163 1163 1163

Introduction Smooth and symmetrical arm movements in gait are expected and part of the general movement pattern in human locomotion. The movement in gait sends signals about the walker’s personality, health, and functional status. Symmetry of movement in particular is important for appearance and well-being. Deviation from normal draws attention to abnormalities and implies poor and inadequate function. Asymmetry of movements in gait is immediately apparent to the beholder and can be or become a physiological and psychological problem for the walker. Arm movements in gait, as well as movements of the lower extremity, may be difficult to define and understand. Is the movement pattern primarily an expression of functioning at the lowest possible energy cost? Or are there secondary changes as a consequence of impaired control, injury, deformity, or any other cause? Even tertiary changes might develop owing to previous treatment or progressive deformity with continued growth in children. Obviously, it is essential to separate and identify these differences to make it possible to plan follow-up and treatment. It goes without saying that it is not ideal to treat the secondary or compensatory changes if the primary cause itself can be addressed. This is especially true when the treatment is irreversible, such as surgical interventions. Arm and hand function are essential for independence, something that is in focus in the modern industrial society and crucial for survival of the family in the developing world. In addition, upper extremity movement and function reflect personality and – along with facial expression – contribute strongly in human communication and appearance (Chouchourelou et al. 2006; Ikeda and Watanabe 2009; Meeren et al. 2005; Montemare 1987; Schneider et al. 2014). However, arm movements during gait also play a role in locomotion, contributing to stability of the trunk and keeping energy costs low. Therefore, arm pendulum in gait is of importance. Braune and Fischer included arm movement in their description “The gait of humans” from 1895, but not much attention has been directed specifically toward arm pendulum (Braune 1895). The purpose and mechanism of arm pendulum are still under discussion, although modern techniques have given us a better understanding and to some extent allowed us to appreciate their medical value.

The Arm Pendulum in Gait

1153

State of the Art Different models for analysis of the three-dimensional (3D) motion of the upper extremity have been developed, although overall the system is largely consistent with lower extremity gait analysis (Kadaba et al. 1990). A 3D analysis provides more detailed information than a two-dimensional assessment (video). Nevertheless, video capturing provides a lot of information on the dynamics and appearance of the movement of a moving subject. Video can be especially useful to provide information about transitions such as gait initiation and accelerating to walking or running at a higher speed. Arm posturing with increased elbow flexion and decreased range of motion in the sagittal plane (flexion/extension) is a typical observation in video captures when a child with hemiplegic cerebral palsy (CP) walks or starts running (Fig. 1). Hemiplegic CP is defined as increased muscle tone (spasticity) in the arm and the leg on either the right or the left side and is caused by a brain injury before the age of 2 years. However, the data from a video are difficult to quantify, and the exact degree of flexion and the range of motion can only be measured with a 3D analysis. In addition, the video assessment naturally does not include the rotational profile (i.e., the third dimension).

Fig. 1 Arm posture. In hemiplegic cerebral palsy, the typical one-sided spastic position is with elbow flexion, forearm pronation, and palmar flexion and with little range of motion when walking

1154

J. Riad

Previously, only the lower extremity – what Perry called the locomotor system – was segmented for calculations (two feet, two shanks, two thighs, and the pelvis) (Perry 1992). The head, arms, and trunk were treated as a single unit, the passenger segment. With higher computer capacity, most human motion analysis laboratories now segment also this passenger segment into the head, the trunk, two upper arms, two forearms, and two hands. For recording purposes, the segments are defined by several retroflective markers, which are attached to the skin at specific anatomical landmarks. Several cameras identify the markers when the person walks in the laboratory, and movements relative to other segments can be calculated in three dimensions. If the subject walks on the force plates situated in the floor, information on the forces acting on the body (ground reaction forces) can be obtained. From the movement and the force data, the moments over separate joints can be calculated. Hence, movements (kinematics) and forces (kinetics) are collected, and the information is presented in graphs with the gait cycles for the right and the left side and usually the mean. The gait cycle consists of stance phase (60% of the gait cycle) and swing phase (40%) (Fig. 2). From these data we can obtain exact degrees of movement at any point during the gait cycle. In addition, temporal-spatial data, information on gait speed and step length, stride length, support time, step width, etc., is captured. To obtain a measure of degree of deviation from normal in the upper extremity in gait, the Arm Posture Score (Riad et al. 2011) and the Arm Profile Score (Jaspers et al. 2011) were developed, analogous to the Gait Profile Score (Baker et al. 2009). The deviation is calculated as the root mean square during the entire gait cycle compared to the laboratory’s age- and gender-controlled references (Fig. 3). This score can be useful as a comprehensive assessment of the degree of deviation and to calculate symmetry between both upper and lower extremities. However, the degree of deviation does not tell us what sorts of changes are occurring, merely the degree of deviation. For example, it does not reveal whether the movement is increased or decreased or if the range of motion is increased or decreased relative to normal. It is especially in children with cerebral palsy that modern gait analysis has been evolved as a medical assessment tool. Bonnefoy-Mazure and co-workers, in

Fig. 2 Elbow motion (y-axis) during the gait cycle (x-axis) with the first 60% being the stance phase and the last 40% the swing phase. The gray band in the background is the normal/reference group here; the left (red) and right (blue) sides are plotted with all the gait trials obtained from each side. To the right the means from both sides are illustrated. It is clear that the left side (red) has increased flexion and decreased range of motion

The Arm Pendulum in Gait

1155

Control Patient RMSD

Non-involved side

Involved side Flexion 60.0

60.0

40.0

40.0

20.0

20.0

0.0

0.0

60.0

60.0

40.0

40.0

20.0

20.0

0.0

0.0

GPS Knee Extension Flexion

APS Elbow Extension

Fig. 3 The mean of the sagittal movements of knee and elbow on the involved (hemiplegic) side and the noninvolved side in a child with hemiplegic cerebral palsy. The area between the mean represents the root mean square and is the measurement of deviation

addition, reported comprehensively on upper limb patterns during gait in individuals with CP (Bonnefoy-Mazure et al. 2014). In this chapter we discuss only free arm movements unless stated otherwise. That is, we assume an independent gait without assistive devices other than possible bracing and orthosis of the foot and ankle.

Arm Pendulum and Gait In this chapter, arm pendulum in gait, we will discuss the possible purpose of arm pendulum and the mechanical effects on gait. We will discuss how arm pendulum can be useful in medicine to make diagnoses, how to follow development over time, and how arm pendulum assessment can help in treatment. The medical part will be focused on individuals with neurological impairments, with special emphasis on cerebral palsy, past stroke, and Parkinson’s disease. We will also discuss psychological and social aspects of deviations from normal arm movement in gait.

Mechanical Effects of Arm Movement on Gait Arm pendulum has been reported to serve the purposes of decreasing energy consumption, increasing gait stability, and improving balance (Kuhtz-Buschbeck and Jing 2012; Meyns et al. 2013; Ortega et al. 2008). Some state that arm pendulum is “a relic from quadrupedal walking” without purpose (Jackson et al. 1983; Murray et al. 1967). There is also a debate about the mechanism of arm pendulum.

1156

J. Riad

Muscle activity of the upper limb in gait has been assessed by electromyography (EMG) (Ballesteros et al. 1965; Kuhtz-Buschbeck and Jing 2012) under different conditions: free arms, arms held by the side, bound to the body, and in gait at different speeds. The purpose was to determine whether arm pendulum is active or passive. Muscle activity was noted in normal gait but also in some of the other conditions, and Kuhtz-Buschbeck and Jing concluded that normal arm swing most likely has both active and passive components. In a simulation study with the aim to quantify the components, a significant decrease of arm swing amplitude was found when there was no EMG activity (Goudriaan et al. 2014). The same authors could also show alteration in the pattern (interlimb coordination) when muscle activity was absent. It is not fully clear, but it appears that arm pendulum contributes toward lowering energy consumption, even though there is an energy cost for the swinging (Collins et al. 2009; Elftman 1939). The mechanism is the decrease of vertical ground reaction moment, and the net effect is a reduction of energy consumption by 8% compared to walking with arms constrained (Umberger 2008). Although there are some contradictory results regarding arm swing and lateralmedial stability (Collins et al. 2009), it seems that arm swing contributes to stability (Ortega and Farley 2015; Ortega et al. 2008). Obviously, the small child, the toddler, has the high guard position for stability and balance, and the elderly individual achieves increased trunk stability through increased arm swing (Nakakubo et al. 2014; Kubo and Ulrich 2006). Elderly also had longer recovery time after perturbation to return to steady state gait, with normalized arm swing (Nakakubo 2014). In running, step width is narrower than in walking, and arms swing to minimize energy cost and improve lateral-medial balance (Arellano and Kram 2011). In the neural control of arm pendulum in gait, it appears that a remnant from our quadrupedal days is still in charge of the coordination between the forelimbs and hind limbs, which is mediated by long neurons from the cervical and lumbar spine (Dietz 2011). An uncoupling occurs when the human wants to perform a voluntary arm and hand task, at which point direct control from the cortex takes over. Dietz points out that in patients with Parkinson’s disease, the so-called quadrupedal limb coordination is intact; therefore, verbal, visual, and other stimuli of arm swing can be beneficial in the treatment. In patients who have had a stroke, however, the afferent nerve signal does not lead to an appropriate response (Dietz 2011).

How Information on Arm Pendulum Can Help in Medicine Diagnosis Posturing with the arm – the “high guard” position – is common in toddlers for stabilization (Kubo and Ulrich 2006) and may persist in children with cerebral palsy (CP) (Meyns et al. 2012a). Typically in diplegic CP gait, there is high variability in arm movement and increased arm abduction, as a sign of balance disturbance owing to spasticity and decreased motor control in the lower extremity (Meyns et al. 2012a; Romkes et al. 2007).

The Arm Pendulum in Gait

1157

When attempting to differentiate between hemi- and diplegic cerebral palsy in a young child, the arm posturing often reveals the hemiplegic involvement even in mild cases of CP. In hemiplegic CP, the one-sided spastic position with elbow flexion, forearm pronation, and palmar flexion and with little range of motion (Riad et al. 2007; Romkes et al. 2007) is different from the bilateral elbow flexion pattern in diplegic CP (Fig. 1). It has been found that in children born at term with hemiplegic CP that the arm is more involved than the leg, whereas in children born prematurely, the arm and leg are more equally involved (Uvebrant 2000). Thus, the patient’s history of birth and the clinical sign of arm posturing when walking (and especially when running, which enhances the arm posturing) can contribute toward making a diagnosis. Spastic paraparesis and diplegic CP may be difficult to differentiate. According to Bonnefoy-Mazure and co-workers (Bonnefoy-Mazure et al. 2013), 3D analysis of the upper extremity can help in the diagnoses; they found differences by studying the trunk and arm movement patterns that compensate for gait deviation in the lower extremity. Apart from the fact that patients with spastic paraparesis were older (mean 16.7 years) than the diplegic CP group (mean 12.3 years) and walked slower (mean 1.13 m/s vs. mean 1.42 m/s in the diplegic CP group), no differences between the groups were found: their lower extremity movement in the sagittal plane during gait was indistinguishable in 3D analysis. However, the diplegic group showed increased flexion and more variable arm movements in the arms, similar to the high guard position described in toddlers. In those with spastic paraparesis, compensation for the lower extremity deviation occurred in the trunk and pelvis, which was interpreted as an attempt to stabilize the head. Hence, upper extremity 3D analysis can be helpful in this clinical situation. The gait of patients with Parkinson’s disease is characterized by low velocity, short, shuffling steps and the highly typical reduced arm swing, beside the freezing and difficulties initiating gait. Early diagnosis can be tricky, but is important, since early treatment may have a positive impact on the progression of the disease (Fahn et al. 2004). Reduced arm swing is an early sign of Parkinson’s disease (Lewek et al. 2010), and a study by Mirelman and co-workers reported that use of body-affixed sensors to examine arm swing might provide a new prodromal marker for the disease (Mirelman et al. 2016).

How One Limb Affects the Other Limbs Interlimb coordination is studied to determine if the arms and legs are moving in phase, i.e., if they are normally synchronized, in gait. Meyns and co-workers reported that several children with hemiplegic CP had a 2:1 arm-to-leg ratio on the involved side and a 1:1 ratio on the noninvolved side (1:1 ratio being normal). In children with diplegic CP, also with altered interlimb coordination, increased walking speed improved the synchronization, and the authors suggest that exercise increasing walking speed could be beneficial for the general gait pattern (Meyns et al. 2012b). However, this has neither been tested nor validated, and of course the question arises, what influences what?

1158

J. Riad

In early Parkinson’s disease, newly diagnosed patients show changes in interlimb coordination, and thus coordination analyses help in the diagnosis (Winogrodzka et al. 2005). In addition, it has been suggested that improving interlimb coordination could improve gait in Parkinson’s disease (Dietz 2011). Another interesting study, investigating the effect of elbow contracture on gait revealed significant decreases in gait speed with three different degrees of elbow flexion contractures simulated with a brace in healthy participants (Trehan et al. 2015). In other words, the arm pendulum affects gait also in subjects with no neurological impairment. The arm posture in hemiplegic CP may have an impact on gait and vice versa, and studying the possible influence of different limbs can be quite complex (Lundh et al. 2014). How does the gait respond to the increased elbow flexion and the decreased range of elbow motion? And how does the noninvolved lower extremity/leg respond or rather compensate for the hemiplegic leg? What does this do to the noninvolved arm? The interactions and deviations in relation to each of the two upper and two lower extremities in gait in hemiplegic CP were studied by Lundh (Lundh et al. 2014). By calculating the degree of deviation and symmetry between involved and noninvolved side in 47 adolescents and young adults with hemiplegic CP, four distinct different groups were identified: close to normal, deviations mainly in the leg, deviations mainly in the arm, and deviation in both the leg and the arm. This information can help in diagnoses, as previously described, but can also be useful in long-term follow-up. It is important to identify and understand changes of movement pattern caused by the development of deformity and/or new compensation mechanisms, so as to differentiate between what to address and attempt to treat and what to leave alone.

Natural Development and Treatment of Arm Posture in Children and Young Adults with Cerebral Palsy In children with hemiplegic CP, the arm posture of the involved side is mainly dependent on spasticity and decreased motor control. Children with diplegic CP, however, use their arms more for balance to compensate for lower limb spasticity and impaired motor control. The natural development of arm posturing, elbow flexion in gait, in patients with hemiplegic CP, is that it spontaneously resolves with time (Riad et al. 2007). In a study with 175 individuals (mean age 9.2 years, range 4 to 21 years), there was a significant decrease of elbow flexion with increased age in gait on the hemiplegic side, but not on the noninvolved side. Elbow extension normalized on both sides with increased age. The range of motion increased significantly on the noninvolved side with age, but not on the hemiplegic side (Fig. 4). It was also noted that the variability of elbow flexion in gait decreased with increased age. In the same study, the outcome of elbow flexor lengthening on the natural history of elbow flexion and motion during gait was investigated. Among the 175 children who had severe elbow flexion, some had surgery with elbow flexor lengthening and some did not. The outcome for the two groups was examined by comparison of two

The Arm Pendulum in Gait

1159

Fig. 4 Elbow flexion in gait in 175 individuals with hemiplegic cerebral palsy, with no surgical treatment. The dark area represents the hemiplegic side and reveals decreasing elbow flexion with increasing age, which is the natural development

gait analyses of the same patient over time. At a follow-up time of 3.3 years, the outcome in the surgical and nonsurgical groups was similar, suggesting that elbow flexor surgery had no effect on elbow motion during gait. The explanation for this might be the indication and in particular the individuals and their treatment goals. In addition, children with hemiplegic CP born at term has a higher degree of brain abnormalities, which can have implications when considering surgical correction of arm posture and hand and arm function, since these children would not be expected to cooperate in the rehabilitation program as compliantly as those with normal mental development. Taken together, these observations highlight how important it is to inform the child and concerned parents about the natural history of elbow flexion, and that one should be careful with early surgical intervention. The indications for and optimal timing of surgical intervention are not clear. Attention should be directed toward the child’s possible concerns, and treatment should be offered if the spontaneous correction is not satisfactory (Riad et al. 2007). Treatment with botulinum toxin has been reported to benefit young, mainly highfunctioning adolescents with hemiplegic CP; the patients experienced a very positive cosmetic effect, even though parents and caregivers could not detect any change in movement pattern (Corry et al. 1997). The explanation could be decreased tone, leading to a perception of better control. It is possible that having better control – especially emotional in situations or in transitions of movement (gait initiation, running, etc.) – was beneficial for these patients.

1160

J. Riad

In Treatment of Parkinson’s Disease Verbal instructions including “walking while deliberately swinging the arms” were found to normalize gait pattern regarding gait speed and step length, in addition to arm swing (Behrman et al. 1998). The authors stated that this cognitive strategy could be superior to use of visual cues such as lines drawn in the floor and also compared to hearing cues that are not accessible in daily life. Of course numerous studies have evaluated medical/drug treatment for normalization of arm and gait pattern, and many included symmetry assessments. Here, the 3D analysis can prove to be a useful tool for early diagnosis, evaluation of treatment, and following the progress of the disease. In Treatment of Adults Poststroke In rehabilitation after stroke, handheld training devices for arm function have been reported to have good effect. Virtual reality games have increased patients’ motivation and noticeably facilitated training. After stroke, the arm movement in gait is often abnormal/heavily affected, and this most likely affects exercise and training to a great extent. A study using movable handrails in a treadmill setup demonstrated changes in muscle activation in the lower limb, most likely due to changes in postural stability that occurred when the patient performed arm movements (Stephenson et al. 2010). Interestingly, during swing phase, muscle activity was significantly increased in the tibialis anterior muscle when walking with free/more normal arm movements. This increased muscle activity can help clear the foot in swing phase, which is otherwise a common problem after stroke (Stephenson et al. 2010). Botulinum toxin in elbow flexors resulted in increased gait speed (Esquenazi et al. 2008), and botulinum toxin in the hand, the forearm, and the elbow flexor brachioradialis improved the range of motion in the knee and ankle of “slow striding” patients (Hirsch et al. 2005).

Psychological and Social Aspects of Deviation in Arm Pendulum Our appearance – how we look and move – is of importance and contributes to our behavior, to how we are perceived, and, consequently, to how we feel. One method to study how humans perceive other humans’ movements is to use point-light biological motion sequences. This involves attaching lights to major joints and filming a body in motion. Although the resulting visual sequence shows only the movement patterns of the joints, without showing limbs or body, viewers are readily able to recognize the point lights as representing a human walking. It has been found that the visual detection of motion occurs in the superior temporal sulcus, an area of the brain that is closely connected with the emotional processing area, the amygdala (Johansson 1973). There is evidence that gait characteristics, including arm swing, differentiate emotions (Montemare 1987). There is a clear visual sensitivity to “angry walkers” (Chouchourelou et al. 2006; Ikeda and Watanabe 2009). In

The Arm Pendulum in Gait

1161

addition, arm movement in gait is important for the recognition of individuals, and “exaggerated” movements of the arms facilitated the task of recognition (Hill and Pollick 2000). In their paper “Show me how you walk and I tell you how you feel. . .,” Schneider and co-workers suggested that full understanding of emotion perception requires the investigation of dynamic representation and means of expression other than the face (Schneider et al. 2014). It has been reported that self-esteem and self-concept can be affected by physical impairment even in highly functioning patients with unilateral cerebral palsy (Russo et al. 2008). Arm posturing and gait deviations may be perceived as cosmetic and social impediments when the individual enters adolescence and becomes more selfconscious (Decety and Grezes 2006; Wake et al. 2003). In a group of high-functioning adolescents and young adults with hemiplegic CP, self-esteem was significantly lower than in a control group (Riad et al. 2013). In addition, movement deviations in the upper extremity in gait, mainly elbow flexion, correlate with both self-esteem and sense of coherence. The higher the deviation, the lower the self-esteem and sense of coherence (Fig. 5). Interestingly, no such correlations were seen with deviations in the lower extremity. The conclusion was that movement pattern in the upper extremity should not be assessed solely from a functional perspective. Possible concerns about appearance and influence on selfesteem should also be considered.

Fig. 5 Self-esteem assessed by questionnaire “I think I am” plotted with the Gait Profile Score (GPS) and the Arm Posture Score (APS) expressing degree of deviation from normal. Correlation coefficient ( 0.397) with APS. The higher the deviation, the lower the self-esteem

1162

J. Riad

Symmetry In ancient times, both the Egyptians and the Greeks noted that facial symmetry was important in human judgment of beauty; this has also proven to be true for body symmetry (Brown et al. 2008) (Figs. 6 and 7). In one study, dances performed by symmetrical men, as opposed to asymmetrical men, were rated as more attractive, suggesting that dynamic movement can signal underlying quality independently of static appearance (Brown et al. 2005). It is not well studied or clear what possible impact asymmetrical gait pattern, including arm swing, could have on the individual’s own perceptions and those of other people. Fig. 6 Facial symmetry

Fig. 7 Body symmetry

The Arm Pendulum in Gait

1163

Future Directions In the future, gait assessments should also include the upper extremity, head, and trunk, since this contributes to a more comprehensive understanding of human movement patterns in locomotion. Without doubt, inclusion of total body movement is important for a better understanding, not only for medical purposes but also for a wider understanding of human communication and behavior.

Cross-References ▶ 3D Dynamic Pose Estimation from Marker-Based Optical Data ▶ Clinical Gait Assessment by Video Observation and 2D Techniques ▶ Diagnostic Gait Analysis Use in the Treatment Protocol for Cerebral Palsy ▶ EMG Activity in Gait: The Influence of Motor Disorders ▶ Gait During Real-World Challenges: Gait Initiation, Gait Termination, Acceleration, Deceleration, Turning, Slopes, and Stairs ▶ Gait scores: Interpretations and Limitations ▶ Interpreting Spatiotemporal Parameters, Symmetry, and Variability in Clinical Gait Analysis ▶ Measures to Determine Dynamic Balance ▶ Motor Patterns Recognition in Parkinson’s Disease ▶ Gait Disorders in Persons After Stroke ▶ Slip and Fall Risk Assessment ▶ Spasticity Effect in Cerebral Palsy Gait ▶ Strength Related Stance Phase Problems in Cerebral Palsy ▶ EMG Activity in Gait: The Influence of Motor Disorders ▶ Swing Phase Problems in Cerebral Palsy ▶ Shoulder Joint Replacement and Upper Extremity Activities of Daily Living ▶ Upper Extremity Models for Clinical Movement Analysis

References Arellano CJ, Kram R (2011) The effects of step width and arm swing on energetic cost and lateral balance during running. J Biomech 44:1291–1295 Baker R, Mcginley JL, Schwartz MH, Beynon S, Rozumalski A, Graham HK, Tirosh O (2009) The gait profile score and movement analysis profile. Gait Posture 30:265–269 Ballesteros ML, Buchthal F, Rosenfalck P (1965) The pattern of muscular activity during the arm swing of natural walking. Acta Physiol Scand 63:296–310 Behrman AL, Teitelbaum P, Cauraugh JH (1998) Verbal instructional sets to normalise the temporal and spatial gait variables in Parkinson’s disease. J Neurol Neurosurg Psychiatry 65:580–582

1164

J. Riad

Bonnefoy-Mazure A, Turcot K, Kaelin A, De Coulon G, Armand S (2013) Full body gait analysis may improve diagnostic discrimination between hereditary spastic paraplegia and spastic diplegia: a preliminary study. Res Dev Disabil 34:495–504 Bonnefoy-Mazure A, Sagawa Y Jr, Lascombes P, De Coulon G, Armand S (2014) A descriptive analysis of the upper limb patterns during gait in individuals with cerebral palsy. Res Dev Disabil 35:2756–2765 Braune W (1895) Der gang des Mencschen I. Abh K Sachs Ges Wiss Math-Phys 21:153 Brown WM, Cronk L, Grochow K, Jacobson A, Liu CK, Popovic Z, Trivers R (2005) Dance reveals symmetry especially in young men. Nature 438:1148–1150 Brown WM, Price ME, Kang J, Pound N, Zhao Y, Yu H (2008) Fluctuating asymmetry and preferences for sex-typical bodily characteristics. Proc Natl Acad Sci U S A 105:12938–12943 Chouchourelou A, Matsuka T, Harber K, Shiffrar M (2006) The visual analysis of emotional actions. Soc Neurosci 1:63–74 Collins SH, Adamczyk PG, Kuo AD (2009) Dynamic arm swinging in human walking. Proc Biol Sci 276:3679–3688 Corry IS, Cosgrove AP, Walsh EG, Mcclean D, Graham HK (1997) Botulinum toxin A in the hemiplegic upper limb: a double-blind trial. Dev Med Child Neurol 39:185–193 Decety J, Grezes J (2006) The power of simulation: imagining one’s own and other’s behavior. Brain Res 1079:4–14 Dietz V (2011) Quadrupedal coordination of bipedal gait: implications for movement disorders. J Neurol 258:1406–1412 Elftman H (1939) The function of arms in walking. Hum Biol 11:529–535 Esquenazi A, Mayer N, Garreta R (2008) Influence of botulinum toxin type A treatment of elbow flexor spasticity on hemiparetic gait. Am J Phys Med Rehabil 87:305–310 quiz 311, 329 Fahn S, Oakes D, Shoulson I, Kieburtz K, Rudolph A, Lang A, Olanow CW, Tanner C, Marek K, Parkinson Study Group (2004) Levodopa and the progression of Parkinson’s disease. N Engl J Med 351:2498–2508 Goudriaan M, Jonkers I, Van Dieen JH, Bruijn SM (2014) Arm swing in human walking: what is their drive? Gait Posture 40:321–326 Hill H, Pollick FE (2000) Exaggerating temporal differences enhances recognition of individuals from point light displays. Psychol Sci 11:223–228 Hirsch MA, Westhoff B, Toole T, Haupenthal S, Krauspe R, Hefter H (2005) Association between botulinum toxin injection into the arm and changes in gait in adults after stroke. Mov Disord 20:1014–1020 Ikeda H, Watanabe K (2009) Anger and happiness are linked differently to the explicit detection of biological motion. Perception 38:1002–1011 Jackson KM, Joseph J, Wyard SJ (1983) The upper limbs during human walking. Part 2. Function. Electromyogr Clin Neurophysiol 23:435–446 Jaspers E, Feys H, Bruyninckx H, Klingels K, Molenaers G, Desloovere K (2011) The Arm Profile Score: a new summary index to assess upper limb movement pathology. Gait Posture 34:227–233 Johansson G (1973) Visual perception of biological motion and a model for its analysis. Percept Psychophys 14:201–211 Kadaba MP, Ramakrishnan HK, Wootten ME (1990) Measurement of lower extremity kinematics during level walking. J Orthop Res 8:383–392 Kubo M, Ulrich B (2006) A biomechanical analysis of the ‘high guard’ position of arms during walking in toddlers. Infant Behav Dev 29:509–517 Kuhtz-Buschbeck JP, Jing B (2012) Activity of upper limb muscles during human walking. J Electromyogr Kinesiol 22:199–206 Lewek MD, Poole R, Johnson J, Halawa O, Huang X (2010) Arm swing magnitude and asymmetry during gait in the early stages of Parkinson’s disease. Gait Posture 31:256–260 Lundh D, Coleman S, Riad J (2014) Movement deviation and asymmetry assessment with three dimensional gait analysis of both upper- and lower extremity results in four different clinical relevant subgroups in unilateral cerebral palsy. Clin Biomech (Bristol, Avon) 29:381–386

The Arm Pendulum in Gait

1165

Meeren HK, Van Heijnsbergen CC, De Gelder B (2005) Rapid perceptual integration of facial expression and emotional body language. Proc Natl Acad Sci U S A 102:16518–16523 Meyns P, Desloovere K, Van Gestel L, Massaad F, Smits-Engelsman B, Duysens J (2012a) Altered arm posture in children with cerebral palsy is related to instability during walking. Eur J Paediatr Neurol 16:528–535 Meyns P, Van Gestel L, Bruijn SM, Desloovere K, Swinnen SP, Duysens J (2012b) Is interlimb coordination during walking preserved in children with cerebral palsy? Res Dev Disabil 33:1418–1428 Meyns P, Bruijn SM, Duysens J (2013) The how and why of arm swing during human walking. Gait Posture 38:555–562 Mirelman A, Bernad-Elazari H, Thaler A, Giladi-Yacobi E, Gurevich T, Gana-Weisz M, SaundersPullman R, Raymond D, Doan N, Bressman SB, Marder KS, Alcalay RN, Rao AK, Berg D, Brockmann K, Aasly J, Waro BJ, Tolosa E, Vilas D, Pont-Sunyer C, Orr-Urtreger A, Hausdorff JM, Giladi N (2016) Arm swing as a potential new prodromal marker of Parkinson’s disease. Mov Disord 31(10):1527–1534 Montemare JM (1987) The identification of emotions from gait information. J Nonverbal Behavior 11:33–42 Murray MP, Sepic SB, Barnard EJ (1967) Patterns of sagittal rotation of the upper limbs in walking. Phys Ther 47:272–284 Nakakubo S, Doi T, Sawa R, Misu D, Tsutsumimoto K, Ono R (2014) Does arm swing emphasized deliberately increase the trunk stability during walking in the elderly adults? Gait Posture 40:516–520 Ortega JD, Farley CT (2015) Effects of aging on mechanical efficiency and muscle activation during level and uphill walking. J Electromyogr Kinesiol 25:193–198 Ortega JD, Fehlman LA, Farley CT (2008) Effects of aging and arm swing on the metabolic cost of stability in human walking. J Biomech 41:3303–3308 Perry J (1992) Gait analysis, normal and pathological function. Slack Incorporated, Thorofare Riad J, Coleman S, Miller F (2007) Arm posturing during walking in children with spastic hemiplegic cerebral palsy. J Pediatr Orthop 27:137–141 Riad J, Coleman S, Lundh D, Brostrom E (2011) Arm posture score and arm movement during walking: a comprehensive assessment in spastic hemiplegic cerebral palsy. Gait Posture 33:48–53 Riad J, Brostrom E, Langius-Eklof A (2013) Do movement deviations influence self-esteem and sense of coherence in mild unilateral cerebral palsy? J Pediatr Orthop 33:298–302 Romkes J, Peeters W, Oosterom AM, Molenaar S, Bakels I, Brunner R (2007) Evaluating upper body movements during gait in healthy children and children with diplegic cerebral palsy. J Pediatr Orthop B 16:175–180 Russo RN, Goodwin EJ, Miller MD, Haan EA, Connell TM, Crotty M (2008) Self-esteem, selfconcept, and quality of life in children with hemiplegic cerebral palsy. J Pediatr 153:473–477 Schneider S, Christensen A, Haussinger FB, Fallgatter AJ, Giese MA, Ehlis AC (2014) Show me how you walk and I tell you how you feel – a functional near-infrared spectroscopy study on emotion perception based on human gait. NeuroImage 85(Pt 1):380–390 Stephenson JL, De Serres SJ, Lamontagne A (2010) The effect of arm movements on the lower limb during gait after a stroke. Gait Posture 31:109–115 Trehan SK, Wolff AL, Gibbons M, Hillstrom HJ, Daluiski A (2015) The effect of simulated elbow contracture on temporal and distance gait parameters. Gait Posture 41:791–794 Umberger BR (2008) Effects of suppressing arm swing on kinematics, kinetics, and energetics of human walking. J Biomech 41:2575–2580 Uvebrant G (2000) Congenital hemiplegia. Mac Keith Press, London Wake M, Salmon L, Reddihough D (2003) Health status of Australian children with mild to severe cerebral palsy: cross-sectional survey using the Child Health Questionnaire. Dev Med Child Neurol 45:194–199 Winogrodzka A, Wagenaar RC, Booij J, Wolters EC (2005) Rigidity and bradykinesia reduce interlimb coordination in Parkinsonian gait. Arch Phys Med Rehabil 86:183–189

Upper Extremity Movement Pathology in Functional Tasks Lisa Mailleux, Cristina Simon-Martinez, Hilde Feys, and Ellen Jaspers

Abstract

Children with unilateral cerebral palsy (uCP) typically present with largely divergent upper extremity sensorimotor impairments that impact on their independence and quality of life. While clinical assessment tools have vastly contributed to our understanding of arm and hand problems in children with uCP, the availability of three-dimensional movement analysis has allowed a more detailed and objective assessment of upper extremity movement pathology in these children. This chapter firstly discusses the typical and atypical development of arm and hand function. The second part provides an overview of measurement protocols and outcome parameters and summarizes upper extremity movement pathology in children with uCP. Keywords

Upper extremity • Assessment • Three-dimensional analysis • Movement pathology • Unilateral • Cerebral palsy • Kinematics • Reaching • Grasping • Functional tasks

Contents Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . State of the Art . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Typical Development of Upper Extremity Function . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Upper Extremity Dysfunction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

1168 1168 1171 1172

Lisa Mailleux and Cristina Simon-Martinez contributed equally to this work. L. Mailleux • C. Simon-Martinez • H. Feys Research Group for Neuromotor Rehabilitation, KU Leuven, Leuven, Belgium e-mail: [email protected]; [email protected]; [email protected] E. Jaspers (*) Neural Control of Movement Lab, ETH Zurich, Zurich, Switzerland e-mail: [email protected] # Springer International Publishing AG, part of Springer Nature 2018 B. Müller, S.I. Wolf (eds.), Handbook of Human Motion, https://doi.org/10.1007/978-3-319-14418-4_57

1167

1168

L. Mailleux et al.

Measuring Upper Extremity Movement Pathology . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Movement Protocols . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Spatiotemporal Parameters . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Joint Kinematics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Summary Indices . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Future Directions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Cross-References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

1172 1173 1173 1175 1176 1179 1182 1182

Introduction In life, good arm and hand functionality is indispensable to carry out common daily tasks, such as eating, grooming, and personal hygiene. Seemingly easy as these activities may appear, various processes are involved to ensure successful task execution. Both simple everyday tasks and highly skilled activities require a finely tuned coordination between head, trunk, arm, and hand movements. Such finely tuned coordination constitutes a correct order of movement sequences with adequate timing, movement trajectory, and force generation. The redundant number of degrees of freedom of the human motor system at both the muscular and joint levels vastly challenges upper extremity coordination. To overcome this intrinsic redundancy, our nervous system imposes some coordinative constraints, which lead to adequate and unique movement solutions (Flash and Mussa-Ivaldi 1990; Aarts et al. 2007). These solutions produce stereotyped kinematic patterns during the execution of goal-directed arm movements, i.e., straight hand paths with a bell-shaped velocity profile (Morasso 1983). However, the acquisition of these finely tuned motor skills, going from simple to complex, can only take place through the process of motor development. As goal-directed movements are involved in many functional tasks, alterations in this motor development due to neurological deficits, such as cerebral palsy, will undoubtedly result in a challenging disability in everyday life.

State of the Art With a prevalence of 2–3 in 1000 live births, cerebral palsy (CP) is the leading cause of childhood physical disability (Himmelmann et al. 2010). CP describes a group of permanent disorders of the development of movement and posture, causing activity limitation, that are attributed to nonprogressive disturbances that occurred in the developing fetal or infant brain. Within the spastic CP subtypes, a further distinction is made between unilateral or bilateral CP, reflecting the topography of the movement disorder. The term hemiplegia refers to unilateral spastic involvement and occurs in about 38% of the children with CP (Himmelmann et al. 2010). Diplegia and quadriplegia refer to bilateral spastic CP, with predominant involvement of the lower extremities or of all four extremities, respectively (Himmelmann and Sundh 2015).

Upper Extremity Movement Pathology in Functional Tasks

1169

Taken together, up to 60% of children with CP have problems with their arms and/or hands, affecting mainly children with hemiplegia or quadriplegia. Upper extremity motor and sensory impairments compromise effective arm and hand use and thereby hamper the execution of many everyday activities. As a result, many children with upper extremity involvement become increasingly dependent in daily life. Targeted treatment planning is crucial to maximize a child’s arm and hand function, and this requires a thorough understanding of all upper extremity dysfunctions and underlying mechanisms. In standard clinical practice, the severity of upper extremity involvement is classified using the Manual Ability Classification System (MACS; (Eliasson et al. 2006)). The MACS provides a systematic method to classify typical use of both hands when handling objects in daily activities, in terms of five different levels. Children with unilateral CP (uCP) are usually classified in the first three levels, indicating that they are independent in most age-related activities (MACS I or II) or that they need some help to prepare and/or modify activities (MACS III). Children with bilateral CP with upper extremity involvement generally fall in the last two levels, i.e., they need constant support and can only perform simple actions in adapted situations. The vast difference in functional abilities between children with MACSscores I–III and those with MACS-score IV or V is also reflected in their respective therapy goals and concurrent treatment planning. For the remainder of this chapter, we will focus on children with spastic 69 unilateral CP (uCP). To further guide treatment planning and identify therapy goals, a more detailed evaluation of upper extremity dysfunctions is performed at the level of body function, activity, and participation according to the ICF framework (Fig. 1). At the level of body function, active or passive range of motion, muscle weakness, and spasticity, as well as sensory function, can be evaluated (Klingels et al. 2010a). At the level of activity, the available clinical measures allow to differentiate between the child’s capacity, i.e., the ability to execute a task on the highest probable level of functioning that the child may reach in a standardized environment, and his/her performance level, i.e., the spontaneous use of the impaired hand during activities or play. The Quality of Upper Extremity Skills Test (QUEST; (Thorley et al. 2012)) and the Melbourne Assessment 2 (Melbourne Assessment; (Randall et al. 2014)) measure upper extremity movement quality during various unimanual items, such as reaching, grasping, and object manipulation. The Shriners Hospital for Children Upper Extremity Evaluation (SHUEE; (Davids et al. 2006)) additionally assesses range of motion and muscle tone, as well as the spontaneous functional use, dynamic positioning, and grasp-release abilities during unimanual tasks. Movement speed can be assessed with the Jebsen-Taylor Hand Function Test (Taylor et al. 1973). The Assisting Hand Assessment (AHA; (Holmefur and KrumlindeSundholm 2016)) and Video Observations Aarts and Aarts (VOAA; (Aarts et al. 2007)) both measure impaired arm and hand use during bimanual activities. Although the reliability and validity of the abovementioned measurements have been well established (Klingels et al. 2010a, b; Sakzewski et al. 2007), their main disadvantage lies within the scoring system. Motor and sensory impairments are evaluated at the level of single joints using ordinal ratings. The scoring of upper

1170

L. Mailleux et al.

Fig. 1 The international classification of functioning, disability, and health (ICF) describes functioning of a patient from the perspective of functioning and disability while integrating contextual factors. Levels of functioning and disability are divided in (1) body functions and structure (impairments), (2) activity (limitations in task or action execution), and (3) participation (restrictions in the involvement in life situations). Contextual factors are divided into (1) “environmental” and (2) “personal” factors. “Environmental factors” have an impact on all components of functioning and disability, but “personal factors” are not classified in the ICF (http://www.who.int/ classifications/icf/en/)

extremity movement quality using the functional tests is mostly based upon the assessor’s visual observation and judgment, again using ordinal ratings at the level of individual joints. Questionnaires are subject to the caregiver’s perception of the child’s functioning in everyday life. Hence, combining a clinical assessment with objective and quantitative measurements provides a more integrated approach to evaluate the pathological and complex upper extremity movement patterns observed in children with uCP. Quantitative measures range from electrogoniometers at single joint level to advanced optoelectronic or inertial measurement systems and allow the simultaneous evaluation of multiple joints during various upper extremity tasks. While inertial measurement systems can be integrated into wearable systems and thus provide information regarding actual arm and hand use in daily life, these systems have not yet been developed for the child with uCP. Conversely, upper extremity three-dimensional movement analysis (3DMA) based on optoelectronic systems in a lab setting has gained increasing attention and is being widely used in various clinical motion analysis labs. This has increased our knowledge beyond the information typically gained from standard clinical assessments. In contrast to the clinical scales, 3DMA provides a detailed and objective description of selective anatomical motions and movement patterns at the different joint levels. The first part of this chapter will give a brief overview of the development of arm and hand function in typically developing children and in children with uCP. The second part will further elaborate on upper extremity movement pathology during functional tasks in children with uCP.

Upper Extremity Movement Pathology in Functional Tasks

1171

Typical Development of Upper Extremity Function Coordinated upper extremity task execution requires the synchronized movement of all joints of the arm and hand, which is challenged by the large amount of degrees of freedom of the upper extremity chain. The ability to perform a task in an accurate, consistent, and flexible manner is acquired through the process of motor learning, i.e., improving motor skills, which reflects the maturation of diverse cognitive, sensory, and motor systems. In typically developing children, skillful reaching and grasping develops throughout early childhood, following distinctive developmental adaptations toward finetuned motor performances (von Hofsten 1991; Ronnqvist and Domellof 2006). In general, reaching and grasping can be divided into three components: (1) the transport movement of the arm while moving the hand toward the object; (2) anticipatory hand opening, consisting of an increasing opening of the hand, followed by a closing of the hand to ensure a proper pre-shaping according to object size and shape; and (3) grasping and manipulating the object, requiring differentiated finger movement. It is well described that the first reaching movements in children are observed at the age of 3–4 months. These infantile reaching movements are characterized by considerably variable and irregular hand movement trajectories, with predominant trunk involvement when bringing the hand closer to the target (Konczak and Dichgans 1997; Berthier et al. 1999). Midline crossing and symmetrical bimanual hand use also initiate at the age of 3–4 months (van Hof et al. 2002; Thelen et al. 1993), followed by asymmetrical bimanual movements at around 7 months of age, which take another 5 months to become refined (Kimmerle et al. 2010). The ability to grasp develops somewhat later, i.e., around 6–9 months (Shumway-Cook and Woollacott 1995), with the anticipatory pre-shaping of the hand emerging around 8 months and greatly improving during the following months (von Hofsten and Ronnqvist 1988). Role-differentiated bimanual manipulation is developed at the age of 13 months, and children start using their preferred hand for manipulating the objects, whereas the non-preferred hand has a support or stabilizing role. This type of bimanual use is the most common when handling objects or when performing daily life activities (Kimmerle et al. 2010). Vast improvements of the hand trajectories during reaching and bimanual hand use are seen over the first 2 years of life, characterized by a proximal to distal maturation (Berthier and Keen 2006). By the age of 4 years, children show a mature pattern of temporal coordination of arm and trunk movements (Schneiberg et al. 2002). However, further maturation toward adult multi-joint coordination, with smooth and stable hand trajectories, continues until the age of 8–10 years (Schneiberg et al. 2002; Kuhtz-Buschbeck et al. 1998). Movement time seems rather constant from 4 to 12 years of age, whereas grip formation, force control of the precision grip, and bimanual coordination continue to mature until 12 years of age (Kuhtz-Buschbeck et al. 1998; Forssberg et al. 1991).

1172

L. Mailleux et al.

Upper Extremity Dysfunction For the child with spastic uCP, upper extremity motor impairments typically include muscle weakness, spasticity, and reduced muscle length. Most limitations in passive range of motion occur at the level of elbow extension and supination, whereas increased muscle tone and muscle weakness are generally more prominent in the forearm, wrist, and finger muscles (Klingels et al. 2012). Stereognosis and two-point discrimination are the most frequently impaired sensory deficits (Klingels et al. 2012). These upper extremity motor and sensory impairments compromise effective arm and hand use, which is reflected in the difficulties these children experience during reaching, grasping, manipulating, and releasing objects with their impaired upper extremity. Difficulties during such basic activities consequently hamper the performance of bimanual activities. These upper extremity impairments usually become clear before the age of 18 months. Children with uCP show a preference for the use of one hand only while keeping the other hand in a fist and thereby delaying the initial use of the impaired hand during bimanual play. Clinically, the absence of general movements and the presence of stiff movements at age 3 months in children at risk of developing a neurological disorder have been shown to predict upper extremity dysfunction in those children with CP (Hamer et al. 2011; Romeo et al. 2008). Other clinical scales focus on asymmetries in unimanual and bimanual upper extremity reaching at 4–5 months of age (Grasp and Reach Assessment of Brisbane (GRAB) (Perez et al. 2016)) or the spontaneous use of the assisting upper extremity during bimanual play between 3 and 18 months of age (Hand Assessment for Infants and Mini-Assisting Hand Assessment (Greaves et al. 2013; Krumlinde-Sundholm et al. 2015)). As children with uCP mature, the primary impairments (increased muscle tone, muscle weakness, sensory deficits) can also result in secondary problems such as contractures or bony deformities. These problems undoubtedly contribute to further unimanual activity limitations and bimanual coordination problems. Moreover, the interference of the more impaired hand during bimanual activities does not facilitate the use of both hands during these tasks which contributes to the “developmental nonuse,” as typically seen in children with uCP (Gordon et al. 2013).

Measuring Upper Extremity Movement Pathology 3DMA is a powerful tool for the quantitative assessment of multiple joints of the upper extremity in all degrees of freedom. However, the transfer of knowledge and experience gained from gait analysis to the upper extremity is not straightforward. The lack of cyclic movements, the variety of functions, and abundant degrees of freedom make the upper extremity analysis considerably more complex (Rau et al. 2000). To attain comparable and reliable results from the 3DMA, a general consensus on which tasks to analyze, as well as the biomechanical model, is of utmost importance. The construction of a relevant and clinically meaningful upper extremity movement protocol requires a thorough and patient-specific task selection. For

Upper Extremity Movement Pathology in Functional Tasks

1173

children with uCP, the movement protocol should thus contain several tasks that challenge the upper extremity motor performance in a variety of ways depending on the child’s functional potential. Secondly, to promote the standardization of upper extremity 3DMA, the International Society of Biomechanics (ISB) has published guidelines on the definition of joint coordinate systems and rotation sequences for the trunk, scapula, shoulder, elbow, wrist, and hand (Wu et al. 2005). The output of a 3DMA results in a large amount of data, including spatiotemporal parameters, joint kinematics, and summary indices that quantify the severity of movement pathology for each of the different tasks. In the following section, we will further address the movement protocols and the different parameters that have been used to describe upper extremity movement pathology in children with uCP.

Movement Protocols Depending on the research question or specific treatment goals, one must choose the most appropriate upper extremity tasks for the movement protocol and the movement pathology of interest. In children with uCP, tasks range from simple forward reaching to grasping and object manipulation and more gross motor tasks such as hand to mouth or hand to back pocket (Jaspers et al. 2009, 2011a; Butler et al. 2010a; Brochard et al. 2012; Klotz et al. 2014). Any child can perform reaching tasks, irrespective of their arm and hand function. However, grasping or gross motor tasks require a minimal ability to actively grasp an object in order to be able to successfully execute the task and maintain marker visibility. Hence, adequate task selection ultimately depends on the functional potential of the child and will be different for children with uCP with a good versus a poor arm and hand function. Although mostly unimanual tasks are studied, most daily life activities involve the use of both hands, characterized by asymmetrical hand movements that have to be coupled without interfering with each other. Tasks measuring upper extremity kinematics during symmetrical or asymmetrical movements include simultaneous reaching (Steenbergen et al. 1996; Utley and Sugden 1998) and opening a drawer with one hand and manipulating its content with the other hand (Hung et al. 2004; Rudisch et al. 2016). Importantly, irrespective of the task(s) studied, standardization is crucial and includes adjusting foot and back support, as well as reaching distance and/or height to the individual child’s anthropometrical measures (an example of standardization is provided in Fig. 2).

Spatiotemporal Parameters Spatiotemporal characteristics of upper extremity movements typically include parameters capturing movement duration, speed, and smoothness (i.e., hand trajectory path, number of movement units). Children with uCP are slower, move less smooth, and produce lower maximum movement speed compared to typically developing children (Jaspers et al. 2009; Klotz et al. 2014; Mackey et al. 2006;

1174

L. Mailleux et al.

Fig. 2 Example of a custommade chair with adjustable foot (1) and back support (2) ensuring 90 of hip and knee flexion. Object to be grasped is placed at shoulder height (3) and arm length distance (4)

Ronnqvist and Rosblad 2007; Jaspers et al. 2011b). These movement deviations further deteriorate with increasing severity of uCP (Butler et al. 2010a; Klotz et al. 2014; Ronnqvist and Rosblad 2007) and are seen during various upper extremity tasks, including reaching for a target, grasping, and even during gross motor tasks such as hand to mouth or hand to head (Butler et al. 2010a; Mackey et al. 2006; Ronnqvist and Rosblad 2007; Jaspers et al. 2011b). While children with uCP produce lower maximum velocities when reaching for a target, they also show an impaired ability to decelerate their movement at object contact, resulting in difficulties when the task becomes more demanding (e.g., a cup of water) (Ronnqvist and Rosblad 2007). Typical deficits during grasping include problems with anticipatory hand shaping (Ronnqvist and Rosblad 2007; Wolff et al. 2015) and slower hand positioning on the object to be grasped (Ronnqvist and Rosblad 2007; Steenbergen et al. 2000). Bimanual coordination is typically assessed based on total task completion time, goal synchronization of both hands, and movement overlap time. Studying symmetrical movements shows that children with uCP do have the ability to coordinate both hands by improving the temporal coupling. They compensate through prolonging the reaction and response time of the less impaired hand in order to match the decreased speed of the impaired hand. However, this is only true for simple, straightforward tasks, and increased accuracy demands compromise the temporal coupling (Steenbergen et al. 1996; Utley and Sugden 1998). Bimanual coordination during asymmetrical tasks is decreased in children with uCP compared to their typically developing peers (Hung et al. 2004; Rudisch et al. 2016). However, with increasing speed, bimanual coordination strikingly improves in children with uCP by decreasing the duration of goal synchronization of both hands and increasing movement overlap time (Hung et al. 2004).

Upper Extremity Movement Pathology in Functional Tasks

1175

Joint Kinematics One can choose among an enormous amount of parameters to describe joint kinematics, of which active range of motion and angles at point of task achievement are most often used. For example, during forward reaching, we can describe the total active range of motion of elbow flexion extension as the difference between maximal and minimal elbow flexion, as well the amount of elbow extension at the end of the reach task (point of task achievement, PTA) (see Fig. 3). These parameters are used to describe movement restrictions in distal joints, i.e., wrist and elbow, as well as in proximal joints, i.e., the shoulder, scapula, or trunk. When performing a reaching task, children with uCP use increased wrist flexion and less elbow extension and shoulder elevation, which is accompanied by increased shoulder external rotation and scapula lateral rotation compared to typically

Fig. 3 Example of the movement pattern of elbow flexion extension during reaching forward of an 11-year old boy with right-sided uCP (MACS-score III, mild spasticity in elbow flexors, and submaximal strength of the elbow extensors). Kinematic data points are time normalized from movement start to PTA (0–100%). The start position for the reach forward task is hand on ipsilateral knee; PTA is defined as the point where the target at arm length distance and shoulder height is touched. The gray line indicates the average movement pattern of 60 typically developing children (shaded bar represents 1 standard deviation); the blue line indicates the performance of the child with uCP. Abbreviations: uCP unilateral cerebral palsy, TDC typically developing children, MAX maximum elbow flexion angle, PTA elbow flexion angle at point of task achievement, MIN minimum elbow flexion angle, ROM range of motion calculated as the difference between the minimum and maximum elbow flexion angle

1176

L. Mailleux et al.

developing children (Jaspers et al. 2009, 2011b; Mackey et al. 2006). During reachto-grasp tasks, children with uCP additionally use increased trunk compensatory movements to grasp the object (Coluccini et al. 2007; Kreulen et al. 2007) (Jaspers et al. 2009, 2011b; Butler et al. 2010b). In case the object requires an active supination, such as a glass, elbow supination deficits also become apparent (Jaspers et al. 2011b; Kreulen et al. 2007; Butler et al. 2010b). During gross motor tasks, such as bringing the hand to the mouth, to the head, the back pocket, or across the midline to the other shoulder, children with uCP exhibit increased wrist flexion and reduced elbow supination, which is accompanied by increased scapular and trunk movements (Mackey et al. 2006; Jaspers et al. 2011b). Independent of the task, children with uCP have an altered scapular position in rest, with increased protraction, medial rotation, and anterior tilt (Brochard et al. 2012; Jaspers et al. 2011b). These differences in scapular orientation at rest influence muscle length and muscle activation of the scapulothoracic and glenohumeral muscles, which negatively impacts on scapulothoracic and glenohumeral control (McClure et al. 2001). An example of altered scapulothoracic starting position and movement pattern for a representative child with spastic uCP can be found in Fig. 4. In children with uCP, reported deficits strongly depend on the task requirements. True deficits are most often present in wrist and elbow joints, which generally lead to compensatory movements at the level of the trunk, scapula and shoulder joint to successfully perform the upper extremity task. Compensatory movements are typically seen in case of limited upper extremity mobility or when the effort to activate the “normal” movement pattern exceeds the effort to use compensatory movements (Michaelsen et al. 2001). For example, in uCP active elbow supination can be very strenuous, whereas the activation of other degrees of freedom such as shoulder external rotation or trunk lateral bending might be preferred in order to reach their goal (see Fig. 5). Joint kinematics during asymmetrical bimanual tasks do not provide any additional information with respect to restrictions in joint range of motion compared to unimanual upper extremity tasks (Klotz et al. 2014).

Summary Indices Summary indices go beyond the joint angles at specific events in the movement, by taking into account the amount of joint angle deviation at each point of the movement pattern. The Arm Profile Score (APS) (Jaspers et al. 2011c) and the Pediatric Upper Limb Motion Index (PULMI) (Butler and Rose 2012) are two examples of summary indices that quantify upper extremity movement pathology. Both are based on the Gait Profile Score, a mathematical method based on root-mean-square differences between kinematic data of an individual child with uCP and the mean of a reference database of typically developing children (Baker et al. 2009). The APS is a composite score based on the amount of movement pathology of 13 individual joint angles, i.e., Arm Variable Scores (AVS) of the trunk (flexion/extension, axial rotation, lateral bending), scapula (protraction/retraction,

Fig. 4 Scapulothoracic kinematics during the execution of reaching and grasping a vertically oriented target (cylinder). Scapular kinematics relative to the thorax results in three movements: protraction/retraction in the transverse plane, medial/lateral rotation in the frontal plane, and anterior/posterior tilting in the sagittal plane. The gray line indicates the average movement pattern of a typically developing group (shaded bar represents 1 standard deviation); the blue line indicates the performance of a representative child with spastic uCP (11-year old boy with right-sided uCP; MACS-score III, mild spasticity in elbow flexors, and submaximal strength of the elbow extensors). Children with uCP have an altered scapulothoracic starting position characterized by increased protraction, medial rotation, and anterior tilting, as well as an altered movement pattern in the first half of the movement cycle, which points toward a decreased scapular control. Scapular protraction normalizes toward the end of the movement, whereas increased medial rotation and posterior tilt persist at movement end. Abbreviations: uCP unilateral cerebral palsy, TDC typically developing children, MACS manual ability classification system

Upper Extremity Movement Pathology in Functional Tasks 1177

Fig. 5 Elbow, shoulder, and trunk kinematics during the execution of reaching and grasping a vertically oriented target (cylinder). The elbow kinematics shows a supination deficit at the point of task achievement, with concurrent increased external rotation of the shoulder and lateral bending of the trunk toward the moving arm. The compensations at the level of the shoulder and trunk allow a successful grasping of the cylinder. The gray line indicates the average movement pattern of a group of typically developing children (shaded bar represents 1 standard deviation); the blue line indicates the performance of a representative child with spastic uCP (11-year old boy with right-sided uCP; MACS-score III, mild spasticity in elbow flexors, and submaximal strength of the elbow extensors). Abbreviations: uCP unilateral cerebral palsy, TDC typically developing children, PTA point of task achievement

1178 L. Mailleux et al.

Upper Extremity Movement Pathology in Functional Tasks

1179

medial/lateral rotation, tilting), shoulder (elevation plane, elevation, internal/external rotation), elbow (flexion/extension, pronation/supination), and wrist (flexion/extension, ulnar/radial deviation). The AVS allows to investigate the severity of the movement deviations for each specific joint angle. An example of the AVS and APS during forward reaching is presented in Fig. 6. In contrast, the PULMI provides only one composite score for the total amount of movement pathology based on eight joint angles, including the trunk (flexion/extension, axial rotation), shoulder (internal/external rotation, elevation), elbow (flexion/extension, pronation/supination), and wrist (flexion/extension, ulnar/radial deviation). Irrespective of the task, children with uCP have a higher total amount of movement pathology compared to typically developing children. Movement pathology also increases with increasing severity of uCP (Jaspers et al. 2011c; Butler and Rose 2012). When breaking down the total amount of movement pathology into movement deviations at the single joint level (i.e., AVS scores), the main contributors are deviant wrist and elbow flexion during reach and grasp tasks, as well as increased elbow pronation in case active supination is required, such as during gross motor tasks (Jaspers et al. 2011c). Proximally, movement deviations at the trunk and especially scapula tilting are common across all tasks. The amount of movement pathology for shoulder elevation and scapula protraction typically increases during gross motor tasks (Jaspers et al. 2011c). While the APS, AVS, or PULMI provide clinically valuable information about the amount of movement pathology, these summary indices do not allow any interpretation regarding the direction of the movement deviations (e.g., shoulder external or internal rotation). Hence, these indices always need to be interpreted together with the individual kinematic waveforms of the child. Nevertheless, the use of summary scores substantially reduces the number of parameters obtained from the 3DMA, which in turn facilitates the clinical interpretation of the measured movement pathology.

Future Directions Movement patterns in children with uCP are characterized by increased wrist and elbow flexion during reaching and reach to grasp, along with an elbow supination deficit that becomes more pronounced when supination is a specific task requisite. To successfully accomplish the upper extremity task, proximal joint deviations also become apparent, including increased trunk movements or shoulder rotations. The scapula forms a crucial part of the shoulder complex, and the quantitative assessment of scapulothoracic movements adds to our insights in upper extremity movement pathology in children with uCP. 3DMA is currently the only method to capture these movements noninvasively during upper extremity task performance. Measuring single joint movement and movement pathology constitutes an essential asset of 3DMA and might help guiding and optimizing targeted treatment interventions at the level of the individual joint, i.e., training of scapulothoracic control, botulinum toxin type A injections, or surgery.

1180

a

L. Mailleux et al.

80

RMS difference (degrees)

70 60 50 40 30 20 10 0

Fl Ext Lat FI

Rot

Rot SCAPULA

TRUNK

b

Pro Med Lat Tilt Retr

Elev Plane

Elev

Rot

FI Ext

Pro Sup

ELBOW

SHOULDER

FI Ext UlnRad APS Dev WRIST

80

RMS difference (degrees)

70 60 50 40 30 20 10 0

Fl Ext Lat Fl TRUNK

Rot

Pro Med Lat Tilt Retr Rot SCAPULA

Elev Plane

Elev

SHOULDER

Rot

FI Ext

Pro Sup

ELBOW

FI Uln Rad APS Ext Dev WRIST

Fig. 6 Example of the AVS and APS during forward reaching for two representative children with uCP: (a) low functioning child (10-year old girl with right-sided uCP; MACS-score III, clear spasticity in elbow and finger flexors, and submaximal strength of the elbow and wrist muscles) and (b) high functioning child (10-year old girl with right-sided uCP; MACS-score I, very mild spasticity in elbow flexors and pronators, and good strength of the elbow and wrist muscles). Each column represents the AVS of one specific joint angle. The height corresponds to the root-meansquare difference across time between the individual child with uCP and the average movement cycle from 20 typically developing children (aged 5–15 years). The APS for the overall upper extremity movement pathology is presented in the rightmost column and equals the average rootmean-square difference across time for all 13 AVS. Gray columns indicate the AVS and APS of the typically developing group; colored columns represent the AVS and APS of the low (blue) and high (green) functioning child with uCP. Abbreviations: uCP unilateral cerebral palsy, AVS arm variable score, APS arm profile score, Fl Ext flexion/extension, La tFl lateral flexion, Rot rotation, Pro Retr protraction/retraction, Med Lat Rot medial/lateral rotation, Tilt tilting, Elev elevation, Pro Sup pronation/supination, Uln Rad Dev ulnar/radial deviation

Upper Extremity Movement Pathology in Functional Tasks

1181

Despite the vast potential of upper extremity 3DMA in treatment planning and the evaluation of treatment efficacy, this assessment tool has not yet been implemented in standard clinical practice. One of the primary reasons is the variety of tasks that can be measured, along with the lack of agreement regarding which task provides most valuable information, as there is still little knowledge on the sensitivity of the different tasks to detect specific movement deviations. Moreover, while the summary indices are a broad measure to identify movement pathology in a clinical setting, these do not contain accurate information that could aid in further optimization of upper extremity treatment planning. Statistical parametric mapping is a novel approach specifically developed to analyze kinematic waveforms that could help overcoming this problem (SPM1d (Pataky 2010)). SPM1d has been specifically designed for continuous field analysis, i.e., it takes into account the dependency between data points of an entire movement pattern. The main practical advantage of this analysis lies within its ability to identify specific parts of the movement pattern where children present with a pathological movement, information that is not provided by the summary indices or the discrete joint angles extracted at specific time points in the waveform. SPM1d analysis seems promising to identify those tasks that are more sensitive to change after treatment or those that are more discriminative between children with different upper extremity impairments. In the future, SPM1d might also contribute to the classification of upper extremity movement patterns in children with uCP. On the other hand, children with a unilateral impairment barely use their impaired hand to perform unimanual activities such as reaching or reach to grasp. While measuring upper extremity performance during asymmetrical tasks provides little additional information regarding joint pathology, these tasks might be of great interest to assess changes in bimanual coordination and motor control. Thus, the integration of an asymmetrical bimanual task into the movement protocol seems a valuable next step. For instance, 3DMA during a functional bimanual task such as opening a bottle could provide novel insights in the strategies of the preferred and impaired hand during object manipulation and stabilization. The implementation of force sensors or electromyography measures would further increase our understanding of how kinetic and kinematics are integrated in the movement patterns during upper extremity functional tasks. Future developments of the upper extremity 3DMA include the optimization of data analysis and parameter extraction (SPM1d), as well as the improvement of data acquisition via adequate task selection, the integration of bimanual asymmetrical tasks, and kinetic measures. These developments will vastly add to the clinical value of the upper extremity 3DMA, which will improve our understanding of changes in motor strategies due to uCP, after treatment, or with increasing age. Hence, 3DMA will prove to be a great tool to assist clinicians in their decision-making and in the individual child’s upper extremity treatment planning and follow-up.

1182

L. Mailleux et al.

Cross-References ▶ Gait Scores: Interpretations and Limitations ▶ Spasticity Effect in Cerebral Palsy Gait ▶ The Arm Pendulum in Gait ▶ Trunk and Spine Models for Instrumented Gait Analysis ▶ Upper Extremity Models for Clinical Movement Analysis

References Aarts PB, Jongerius PH, Aarts MA, Van Hartingsveldt MJ, Anderson PG, Beumer A (2007) A pilot study of the video observations Aarts and Aarts (VOAA): a new software program to measure motor behaviour in children with cerebral palsy. Occup Ther Int 14:113–122 Baker R, McGinley JL, Schwartz MH, Beynon S, Rozumalski A, Graham HK et al (2009) The gait profile score and movement analysis profile. Gait Posture 30:265–269 Berthier NE, Keen R (2006) Development of reaching in infancy. Exp Brain Res 169:507–518 Berthier NE, Clifton RK, McCall DD, Robin DJ (1999) Proximodistal structure of early reaching in human infants. Exp Brain Res 127:259–269 Brochard S, Lempereur M, Mao L, Remy-Neris O (2012) The role of the scapulo-thoracic and gleno-humeral joints in upper-limb motion in children with hemiplegic cerebral palsy. Clin Biomech (Bristol, Avon) 27:652–660 Butler EE, Rose J (2012) The pediatric upper limb motion index and a temporal-spatial logistic regression: quantitative analysis of upper limb movement disorders during the Reach & Grasp Cycle. J Biomech 45:945–951 Butler EE, Ladd AL, Lamont LE, Rose J (2010a) Temporal-spatial parameters of the upper limb during a Reach & Grasp Cycle for children. Gait Posture 32:301–306 Butler EE, Ladd AL, Louie SA, Lamont LE, Wong W, Rose J (2010b) Three-dimensional kinematics of the upper limb during a reach and grasp cycle for children. Gait Posture 32:72–77 Coluccini M, Maini ES, Martelloni C, Sgandurra G, Cioni G (2007) Kinematic characterization of functional reach to grasp in normal and in motor disabled children. Gait Posture 25:493–501 Davids JR, Peace LC, Wagner LV, Gidewall MA, Blackhurst DW, Roberson WM (2006) Validation of the Shriners Hospital for Children Upper Extremity Evaluation (SHUEE) for children with hemiplegic cerebral palsy. J Bone Joint Surg Am 88:326–333 Eliasson AC, Krumlinde-Sundholm L, Rosblad B, Beckung E, Arner M, Ohrvall AM et al (2006) The manual ability classification system (MACS) for children with cerebral palsy: scale development and evidence of validity and reliability. Dev Med Child Neurol 48:549–554 Flash T, Mussa-Ivaldi F (1990) Human arm stiffness characteristics during the maintenance of posture. Exp Brain Res 82:315–326 Forssberg H, Eliasson AC, Kinoshita H, Johansson RS, Westling G (1991) Development of human precision grip. I: basic coordination of force. Exp Brain Res 85:451–457 Gordon AM, Bleyenheuft Y, Steenbergen B (2013) Pathophysiology of impaired hand function in children with unilateral cerebral palsy. Dev Med Child Neurol 55(Suppl 4):32–37 Greaves S, Imms C, Dodd K, Krumlinde-Sundholm L (2013) Development of the mini-assisting hand assessment: evidence for content and internal scale validity. Dev Med Child Neurol 55:1030–1037 Hamer EG, Bos AF, Hadders-Algra M (2011) Assessment of specific characteristics of abnormal general movements: does it enhance the prediction of cerebral palsy? Dev Med Child Neurol 53:751–756

Upper Extremity Movement Pathology in Functional Tasks

1183

Himmelmann K, Sundh V (2015) Survival with cerebral palsy over five decades in western Sweden. Dev Med Child Neurol 57:762–767 Himmelmann K, Hagberg G, Uvebrant P (2010) The changing panorama of cerebral palsy in Sweden. X. Prevalence and origin in the birth-year period 1999–2002. Acta Paediatr 99:1337–1343 van Hof R, van der Kamp J, Savelsbergh GJ (2002) The relation of unimanual and bimanual reaching to crossing the midline. Child Dev 73:1353–1362 von Hofsten C (1991) Structuring of early reaching movements: a longitudinal study. J Mot Behav 23:280–292 von Hofsten C, Ronnqvist L (1988) Preparation for grasping an object: a developmental study. J Exp Psychol Hum Percept Perform 14:610–621 Holmefur MM, Krumlinde-Sundholm L (2016) Psychometric properties of a revised version of the assisting hand assessment (kids-AHA 5.0). Dev Med Child Neurol 58:618–624 Hung YC, Charles J, Gordon AM (2004) Bimanual coordination during a goal-directed task in children with hemiplegic cerebral palsy. Dev Med Child Neurol 46:746–753 Jaspers E, Desloovere K, Bruyninckx H, Molenaers G, Klingels K, Feys H (2009) Review of quantitative measurements of upper limb movements in hemiplegic cerebral palsy. Gait Posture 30:395–404 Jaspers E, Feys H, Bruyninckx H, Cutti A, Harlaar J, Molenaers G et al (2011a) The reliability of upper limb kinematics in children with hemiplegic cerebral palsy. Gait Posture 33:568–575 Jaspers E, Desloovere K, Bruyninckx H, Klingels K, Molenaers G, Aertbelien E et al (2011b) Three-dimensional upper limb movement characteristics in children with hemiplegic cerebral palsy and typically developing children. Res Dev Disabil 32:2283–2294 Jaspers E, Feys H, Bruyninckx H, Klingels K, Molenaers G, Desloovere K (2011c) The arm profile score: a new summary index to assess upper limb movement pathology. Gait Posture 34:227–233 Kimmerle M, Ferre CL, Kotwica KA, Michel GF (2010) Development of role-differentiated bimanual manipulation during the infant’s first year. Dev Psychobiol 52:168–180 Klingels K, De Cock P, Molenaers G, Desloovere K, Huenaerts C, Jaspers E et al (2010a) Upper limb motor and sensory impairments in children with hemiplegic cerebral palsy. Can they be measured reliably? Disabil Rehabil 32:409–416 Klingels K, Jaspers E, Van de Winckel A, De Cock P, Molenaers G, Feys H (2010b) A systematic review of arm activity measures for children with hemiplegic cerebral palsy. Clin Rehabil 24:887–900 Klingels K, Demeyere I, Jaspers E, De Cock P, Molenaers G, Boyd R et al (2012) Upper limb impairments and their impact on activity measures in children with unilateral cerebral palsy. Eur J Paediatr Neurol 16:475–484 Klotz MC, van Drongelen S, Rettig O, Wenger P, Gantz S, Dreher T et al (2014) Motion analysis of the upper extremity in children with unilateral cerebral palsy – an assessment of six daily tasks. Res Dev Disabil 35:2950–2957 Konczak J, Dichgans J (1997) The development toward stereotypic arm kinematics during reaching in the first 3 years of life. Exp Brain Res 117:346–354 Kreulen M, Smeulders MJ, Veeger HE, Hage JJ (2007) Movement patterns of the upper extremity and trunk associated with impaired forearm rotation in patients with hemiplegic cerebral palsy compared to healthy controls. Gait Posture 25:485–492 Krumlinde-Sundholm L, Ek L, Eliasson AC (2015) What assessments evaluate use of hands in infants? A literature review. Dev Med Child Neurol 57(Suppl 2):37–41 Kuhtz-Buschbeck JP, Stolze H, Johnk K, Boczek-Funcke A, Illert M (1998) Development of prehension movements in children: a kinematic study. Exp Brain Res 122:424–432 Mackey AH, Walt SE, Stott NS (2006) Deficits in upper-limb task performance in children with hemiplegic cerebral palsy as defined by 3-dimensional kinematics. Arch Phys Med Rehabil 87:207–215

1184

L. Mailleux et al.

McClure PW, Michener LA, Sennett BJ, Karduna AR (2001) Direct 3-dimensional measurement of scapular kinematics during dynamic movements in vivo. J Shoulder Elb Surg 10:269–277 Michaelsen SM, Luta A, Roby-Brami A, Levin MF (2001) Effect of trunk restraint on the recovery of reaching movements in hemiparetic patients. Stroke 32:1875–1883 Morasso P (1983) Three dimensional arm trajectories. Biol Cybern 48:187–194 Pataky TC (2010) Generalized n-dimensional biomechanical field analysis using statistical parametric mapping. J Biomech 43:1976–1982 Perez M, Ziviani J, Guzzetta A, Ware RS, Tealdi G, Burzi V et al (2016) Development, and construct validity and internal consistency of the grasp and reach assessment of Brisbane (GRAB) for infants with asymmetric brain injury. Infant Behav Dev 45:110–123 Randall M, Imms C, Carey LM, Pallant JF (2014) Rasch analysis of the melbourne assessment of unilateral upper limb function. Dev Med Child Neurol 56:665–672 Rau G, Disselhorst-Klug C, Schmidt R (2000) Movement biomechanics goes upwards: from the leg to the arm. J Biomech 33:1207–1216 Romeo DM, Guzzetta A, Scoto M, Cioni M, Patusi P, Mazzone D et al (2008) Early neurologic assessment in preterm-infants: integration of traditional neurologic examination and observation of general movements. Eur J Paediatr Neurol 12:183–189 Ronnqvist L, Domellof E (2006) Quantitative assessment of right and left reaching movements in infants: a longitudinal study from 6 to 36 months. Dev Psychobiol 48:444–459 Ronnqvist L, Rosblad B (2007) Kinematic analysis of unimanual reaching and grasping movements in children with hemiplegic cerebral palsy. Clin Biomech (Bristol, Avon) 22:165–175 Rudisch J, Butler J, Izadi H, Zielinski IM, Aarts P, Birtles D et al (2016) Kinematic parameters of hand movement during a disparate bimanual movement task in children with unilateral cerebral palsy. Hum Mov Sci 46:239–250 Sakzewski L, Boyd R, Ziviani J (2007) Clinimetric properties of participation measures for 5- to 13-year-old children with cerebral palsy: a systematic review. Dev Med Child Neurol 49:232–240 Schneiberg S, Sveistrup H, McFadyen B, McKinley P, Levin MF (2002) The development of coordination for reach-to-grasp movements in children. Exp Brain Res 146:142–154 Shumway-Cook AW, Woollacott MH (1995) Motor control: theory and practical applications. Williams and Wilkins, Baltimore Steenbergen B, Hulstijn W, de Vries A, Berger M (1996) Bimanual movement coordination in spastic hemiparesis. Exp Brain Res 110:91–98 Steenbergen B, van Thiel E, Hulstijn W, Meulenbroek RGJ (2000) The coordination of reaching and grasping in spastic hemiparesis. Hum Mov Sci 19:20 Taylor N, Sand PL, Jebsen RH (1973) Evaluation of hand function in children. Arch Phys Med Rehabil 54:129–135 Thelen E, Corbetta D, Kamm K, Spencer JP, Schneider K, Zernicke RF (1993) The transition to reaching: mapping intention and intrinsic dynamics. Child Dev 64:1058–1098 Thorley M, Lannin N, Cusick A, Novak I, Boyd R (2012) Reliability of the quality of upper extremity skills test for children with cerebral palsy aged 2 to 12 years. Phys Occup Ther Pediatr 32:4–21 Utley A, Sugden D (1998) Interlimb coupling in children with hemiplegic cerebral palsy during reaching and grasping at speed. Dev Med Child Neurol 40:396–404 Wolff AL, Raghavan P, Kaminski T, Hillstrom HJ, Gordon AM (2015) Differentiation of hand posture to object shape in children with unilateral spastic cerebral palsy. Res Dev Disabil 45–46:422–430 Wu G, van der Helm FC, Veeger HE, Makhsous M, Van Roy P, Anglin C et al (2005) ISB recommendation on definitions of joint coordinate systems of various joints for the reporting of human joint motion – part II: shoulder, elbow, wrist and hand. J Biomech 38:981–992

Part XV Other Neurologic Gait Disorders

Idiopathic Toe Walking Karen Davies, Lise Leveille, and Christine Alvarez

Abstract

Idiopathic toe walking in children is a diagnosis of exclusion. Toe walkers require investigation with a thorough clinical examination, observational gait assessment, and potential additional review to exclude an underlying diagnosis. Threedimensional instrumented gait analysis is helpful to characterize the observed gait pattern and identify children that may benefit from intervention. It is most useful in the child with reduced range of motion at the ankle who has a toe-toe gait pattern but appears to have a heel strike on observational gait analysis. These children often compensate with either hyperextension of the knee in mid-stance, external foot progression, and/or an early heel rise, frequently missed with observational gait analysis. In this clinical scenario, instrumented threedimensional gait analysis enables characterization of the toe walking pattern and identification of objective intervention outcomes. Keywords

Idiopathic toe walking • Gait analysis • Ankle kinematics • Ankle kinetics • Electromyography • Pedobarography

K. Davies (*) Shriners Gait Lab, Sunny Hill Health Centre for Children, Vancouver, BC, Canada e-mail: [email protected] L. Leveille • C. Alvarez Shriners Gait Lab, Sunny Hill Health Centre for Children, Vancouver, BC, Canada British Columbia Children’s Hospital, Vancouver, BC, Canada e-mail: [email protected]; [email protected] # Springer International Publishing AG, part of Springer Nature 2018 B. Müller, S.I. Wolf (eds.), Handbook of Human Motion, https://doi.org/10.1007/978-3-319-14418-4_60

1187

1188

K. Davies et al.

Contents Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . State of the Art . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Instrumented Gait Analysis in Idiopathic Toe Walking . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Typical Function of the Foot and Ankle . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Function of the Foot and Ankle in Idiopathic Toe Walking . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Gait Accommodations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Pedobarography . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Electromyography . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Temporal Spatial Characteristics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Severity Classification of Idiopathic Toe Walking . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Clinical Perspective . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Case Examples . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Case Study No. 1: Mild (Type 1) Idiopathic Toe Walking (Fig. 4) . . . . . . . . . . . . . . . . . . . . . . . . Case Study No. 2: Moderate (Type 2) Idiopathic Toe Walking (Fig. 5) . . . . . . . . . . . . . . . . . . Case Study No. 3: Severe (Type 3) Idiopathic Toe Walking (Fig. 6) . . . . . . . . . . . . . . . . . . . . . Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Cross-References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

1188 1189 1190 1190 1190 1191 1192 1194 1194 1194 1195 1197 1197 1197 1200 1200 1202 1202

Introduction Idiopathic toe walking (ITW) is a condition of childhood characterized by a bilateral toe-toe gait pattern of unknown cause. In typically developing gait, heel strike at initial contact is consistently present between 3 and 50 weeks following the emergence of independent ambulation (Sutherland et al. 1980; Burnett and Johnson 1971). Persistent toe walking is considered abnormal in typically developing children after 2–3 years of age (Engström and Tedroff 2012; Sobel et al. 1997). ITW is a diagnosis of exclusion when other musculoskeletal and neurological pathology have been ruled out. Potential differential diagnoses include cerebral palsy, autistic spectrum disorders, muscular dystrophy, peripheral neuropathy, neuromuscular disorders, spinal cord abnormalities, talipes equinovarus, dystonia, ankylosing spondylitis, leg length discrepancy, venous malformation or tumor in the gastrocnemius muscle, schizophrenia, and trauma (Hicks et al. 1988; Shulman et al. 1997; Le Cras et al. 2011; Engelbert et al. 2011). The reported prevalence of ITW in all children ranges between 4.9% and 24% with an estimate of 2.1% by the age of 5.5 years (Engström and Tedroff 2012; Sobel et al. 1997; Accardo et al. 1992). ITW is reported more frequently in males at an occurrence rate of 55–68% (Engström and Tedroff 2012; Fox et al. 2006; Eastwood et al. 2000; Stricker and Angulo 1998). Furthermore, a positive family history is reported in 10–40% of both males and females (Engström and Tedroff 2012; Sobel et al. 1997; Fox et al. 2006; Stricker and Angulo 1998; Pomarino et al. 2016). Family history studies suggest a genetic component may play a role in the etiology of toe walking. A relationship between ITW and language delays, learning disabilities, prematurity, persistent elements of the tonic labyrinthine reflex, and immature motor control has also been described (Accardo et al. 1992; Stricker and Angulo 1998; Fox

Idiopathic Toe Walking

1189

et al. 2006; Accardo and Barrow 2015; Clark et al. 2010). Further, challenges with motor skills and sensory processing as well as diminished thresholds for vibration perception have been linked with ITW (Williams et al. 2012, 2013), potentially indicating the presence of subtle neurological changes in children with ITW. Toe walking has been reported in autism and pervasive developmental disorders with an incidence as high as 19–20% (Ming et al. 2007; Barrow et al. 2011). While these more recent studies suggest subtle neurological abnormalities are associated with ITW, the original study describing ITW proposed that it was due to a congenital short tendo calcaneus (Hall et al. 1967). However, older and exclusive toe walkers show more restricted ankle dorsiflexion than younger or intermittent toe walkers, alternatively suggesting that prolonged toe walking leads to diminished range of motion (ROM) as opposed to limited ROM leading to ITW (Furrer and Deonna 1982; Sobel et al. 1997; Stricker and Angulo 1998). Increased time spent toe walking may also explain the high prevalence of Type 1 muscle fibers found in the gastrocnemius of children with ITW histologically (Eastwood et al. 1997). Despite these established associations with ITW, a distinct etiology remains unclear. Children with ITW may experience pain in their legs or feet, diminished balance, recurrent tripping or falling, difficulty with footwear, and reduced passive ankle dorsiflexion (DF) ROM (Sobel et al. 1997; Engelbert et al. 2011; Fox et al. 2006; Clark et al. 2010; Hirsch and Wagner 2004). Diminished ankle DF has been linked to injuries or pathology of the ankle, forefoot, midfoot, and/or hindfoot (Tabrizi et al. 2000; DiGiovanni et al. 2002; Hill 1995). The reported pathology and clinical manifestations associated with ITW are considered impairments in body structures and body functions from the International Classification of Functioning, Disability and Health for Children and Youth (ICF-CY) frame of reference (World Health Organization 2007), requiring appropriate assessment and recommendations for management. Parental and peer pressure to walk with a typical heel-toe gait pattern may lead to accommodations in gait that are imperceptible to observational gait or video analysis alone, thereby indicating the usefulness of an instrumented threedimensional (3D) gait analysis.

State of the Art The role of clinical gait analysis in children with idiopathic toe walking is to identify the condition and describe the necessary accommodations and impairments affecting their gait. It is based on objective instrumented measurement of individual joints or segments and the interpretation of the biomechanics and pathology of these measurements (Baker 2013; Pierz and Õunpuu 2013). Prior to clinical gait analysis, published evidence-based care guidelines are recommended for obtaining history; assessment of pain, speech and language development, and sensory processing function; physical examination; gross motor skill assessment; and observational gait analysis (Le Cras et al. 2011). Physical examination determines active and passive ROM, muscle strength, muscle tone, rotational alignment, and selective motor control. Imaging results are also used to rule out suspected underlying

1190

K. Davies et al.

neurological and orthopedic causes of toe walking. An understanding of typical gait is required to identify and interpret the atypical gait patterns observed in idiopathic toe walking.

Instrumented Gait Analysis in Idiopathic Toe Walking Typical Function of the Foot and Ankle In typical gait, weight bearing takes place on a stable foot, muscles absorb and generate power adequately, the knee and ankle joints progress in the appropriate plane, and there is sufficient ROM at the ankle. Ankle ROM approximates an arc from neutral to 10 of dorsiflexion (DF) and neutral to 20 of plantar flexion (PF) in mature heel-toe gait (Perry 1992). Typical passive ankle DF is 54 at birth and decreases to approximately 41 passive ankle DF by the time children are learning to walk and to 8–25 by young adulthood (Walker 1991; Grimston et al. 1993; Bovens et al. 1990). Ankle position contributes to stability in stance, foot clearance, pre-positioning the foot in swing, appropriate step length, and energy conservation – all considered determinants of normal gait (Perry 1985; Gage 2004). Normal gait is further divided into periods of initial contact, loading response, mid-stance, terminal stance, pre-swing, mid-swing, and terminal swing (Perry 1992). Perry (1992) also describes ankle and foot movement during stance in terms of three ankle rockers: the first (heel) rocker occurs from initial contact to flat foot weight bearing with controlled PF, the second (ankle) rocker takes place when the tibia moves forward over the foot in stance, and the third (forefoot) rocker happens when the foot advances into PF for push-off. The first rocker starts with initial contact, continuing to about 8% of the gait cycle. It overlaps somewhat with the second rocker, which occurs from approximately 5–45% of the gait cycle. The third rocker starts with heel lift at approximately 30% of the gait cycle, overlapping with the second rocker and continuing until the end of stance (Perry 1992). Typical ankle kinetics demonstrate an early ankle DF moment at loading response with a transition to a late PF moment corresponding with the third rocker and toe off.

Function of the Foot and Ankle in Idiopathic Toe Walking Clinically, absence of normal ankle motion during gait presents as forefoot or flat foot weight acceptance at initial contact with an early heel rise prior to mid-stance. Children with ITW demonstrate an absent first rocker and/or shortened loading response (Crenna et al. 2005). Following initial contact, there is progressive ankle DF instead of the controlled PF typically observed (Hicks et al. 1988). Diminished ankle DF ROM limits the forward progression of the tibia over the foot resulting in a reversal of the second rocker and early transition from progressive DF into progressive PF in mid-stance (Hicks et al. 1988). This can also be described as an early heel rise or an early third rocker, which is a premature progression into PF at or before

Idiopathic Toe Walking

1191 Ankle Moment

Flex-Ext

1.6

AM1

AM2

0.5

–0.5 0.0

50.0

100.0

% Gait Cycle

Fig. 1 Representative sagittal plane internal ankle moment. Graphical illustration demonstrating a predominant internal ankle plantar flexion moment in early stance (AM1) and smaller peak plantar flexion moment in late stance (AM2). The gray band represents sagittal ankle kinetics based on normative data from 34 healthy children aged 5 to 18 years within 1 SD of the mean

30% of the gait cycle (Alvarez et al. 2007). It is the atypical timing of peak ankle DF during stance that differentiates ITW from other gait pathology. Swing phase ankle kinematics are also helpful in distinguishing ITW from other foot pathology such as a contralateral drop foot. Idiopathic toe walkers typically present with an early excursion toward ankle DF from early to mid-swing followed by PF or diminished peak DF in mid- to late swing, leading to equinus pre-positioning by initial contact (Kelly et al. 1997; Westberry et al. 2008; McMulkin et al. 2016). In contrast, children with mild spastic diplegia typically demonstrate progressive excursion toward DF throughout swing phase (Kelly et al. 1997). Ankle kinetics in ITW demonstrate an early internal ankle moment resulting from a premature heel rise or early third rocker. Alvarez et al. (2007) label the early predominant internal ankle moment seen in ITW as the first ankle moment (AM1), to delineate the peak PF moment in early stance from the typical peak PF moment seen in late stance (AM2) (Fig. 1). The early first internal ankle moment (AM1) may be larger or smaller than the typical late peak ankle PF moment (AM2). This is alternatively described as a “double bump” ankle PF moment pattern (Stott et al. 2004; Hemo et al. 2006), indicating an early increased internal PF moment. It is important to note that children with ITW often have significant variability in their gait patterns, spontaneously alternating between bilateral toe-toe and heel-toe gait patterns with variable ankle motion (Hicks et al. 1988; Crenna et al. 2005; Westberry et al. 2008). In contrast, children with spastic diplegia tend to have more repeatable gait patterns with less deviation in ankle motion (Hicks et al. 1988). This subtle finding may be useful to differentiate these patient populations.

Gait Accommodations Gait deviations commonly seen with ITW include an increased anterior pelvic tilt, mild knee hyperextension in late stance, and external foot progression (out-toeing)

1192

K. Davies et al.

(Hicks et al. 1988; Westberry et al. 2008; Stott et al. 2004; McMulkin et al. 2006). These gait deviations are accommodations for the abnormal kinematics and kinetics seen at the ankle. It has been postulated that increased PF in stance due to toe walking may cause the center of body mass to shift forward over the base of support, increasing the anterior pelvic tilt (McMulkin et al. 2016). The mid-stance knee hyperextension commonly observed in toe walkers results from limited ankle DF in stance. Restricted motion at the ankle limits forward tibial progression during the second rocker resulting in knee hyperextension. The limited forward tibial progression with continued forward femoral progression over the tibia may further contribute to stance phase knee hyperextension in ITW (Hicks et al. 1988). Limited ankle DF with maximum knee extension leads to further gait accommodation with increased external foot progression to shorten the lever arm and avoid the forward tibial progression typically occurring during the second rocker (Hicks et al. 1988; McMulkin et al. 2016). It is unknown whether the increased external foot progression angle is due to increased hip external rotation, increased external torsion of the tibia, or a combination of both. It has been suggested that this increase in external foot progression is due to increased gastrocnemius tightness; however, static clinical ankle DF ROM is not correlated with peak ankle DF in stance (Stott et al. 2004; McMulkin et al. 2016). Five-year postsurgical intervention follow-up data (n = 8) also showed an increase in external foot progression angle; however, the overall mean was not significant (McMulkin et al. 2016). Surgical procedures in this follow-up study were either a gastrocnemius/soleus recession or tendo-Achilles lengthening (TAL). It is currently unclear whether increased external foot progression in ITW is a true gait accommodation for gastrocnemius tightness or a habitual pattern.

Pedobarography Pedobarography is used to demonstrate abnormal foot loading and center of pressure progression. Bowen et al. (1998) and Alvarez et al. (2008) have described loading patterns in the feet of typically developing children. Restricting the assessment to the dynamic phase of stance and dividing the foot into medial and lateral forefoot, midfoot, and hindfoot simplifies the evaluation of foot pressures to six segments. Consideration of pressure location, timing, and time spent in each of the six segments provides descriptive information of the loading experience of each foot. In children who exclusively walk with a toe-toe pattern, the pedobarography report is entirely forefoot loading with reasonably symmetrical medial and lateral pressure distribution (Fig. 2). Less severe toe walkers can be identified by minimal pressure on the heel segments and reduced time spent on the heel when compared to normal (Fig. 3). These toe walkers will quickly unload the hindfoot, equivalent to an early heel rise, spending the majority of stance phase on the forefoot. Pressure at initial contact in the most subtle toe walkers is seen simultaneously on the fore, midfoot, and hindfoot, or a reversal of loading where the loading is initiated in the forefoot but rapidly transmitted to the hindfoot and

Idiopathic Toe Walking

1193

Fig. 2 Plantar pressure profile of severe toe walker demonstrating forefoot loading

Fig. 3 Plantar pressure profile of less severe toe walker demonstrating minimal pressure on heel segments

then back to the forefoot. Lastly, accommodation to toe walking can be seen as increased external foot progression resulting in increased medial forefoot pressures on pedobarography.

1194

K. Davies et al.

Electromyography Dynamic electromyography (EMG) can be applied by surface or indwelling electrodes to assess muscle activity and contraction patterns throughout the gait cycle and is useful for detecting neuromuscular abnormalities. EMG data in children with ITW demonstrates abnormal co-contraction and out-of-phase muscle activity with premature firing of gastrocnemius in swing as well as low-amplitude tibialis anterior firing throughout stance and swing (Griffin et al. 1977). Brunt et al. (2004) showed that early swing phase gastrocnemius activity and early and restricted duration of tibialis anterior activity commonly result in flatfoot or forefoot weight bearing at initial contact. Children with ITW, equinus deformities, and cerebral palsy all demonstrated comparable EMG findings, differentiating them from a matched control group of children with typical gait walking on their toes voluntarily (Kalen et al. 1986). Therefore, EMG is not necessarily useful for differentiating between causes of toe walking. Additionally, EMG studies have shown large variability and may not be conclusive in providing objective diagnoses due to inconsistent findings (Papariello and Skinner 1985; Kalen et al. 1986).

Temporal Spatial Characteristics Few studies describe temporal spatial characteristics in ITW. McMulkin et al. (2006) found the velocity in 14 children with ITW (mean age 8.9 years, range 5.6–12.6 years) was comparable to a normative database of 44 healthy subjects; however, the toe walkers had a significant increase in cadence and diminished stride length. In contrast, Westberry et al. (2008) studied 51 children with ITW (mean age 9.3 years, range 6.0–18.0 years) and found diminished velocity secondary to a decrease in stride length and decreased cadence when compared to typically developing children.

Severity Classification of Idiopathic Toe Walking A classification framework has been developed based on a cohort of 133 children with a history of toe walking to quantitatively describe ITW, differentiate it from other pathology, and delineate severity gradation (Alvarez et al. 2007). In this classification framework, toe walkers are classified into three severity groups (mild, moderate, or severe) based on the absence or presence of a first rocker, an early third rocker, and an early predominant internal ankle moment. These criteria were developed from preliminary work addressing gait adaptations of the contralateral limb in children with hemiplegic cerebral palsy (Sawatzky et al. 1999). As described previously, Alvarez et al. (2007) refer to the early predominant ankle moment as the AM1, to delineate the peak PF moment in early stance from the peak PF moment typically seen in late stance (AM2) (Fig. 1). Mild (Type 1) toe walking is defined by presence of a first rocker, typical third rocker, and typical ankle PF

Idiopathic Toe Walking

1195

moment. Refer to Case Study No.1 in Case Examples. In moderate (Type 2) toe walkers, there may or may not be a first rocker and typical third rocker. Kinetics distinguish moderate toe walkers by the presence or absence of an early ankle moment (AM1) which is lower than the normally timed but diminished second moment (AM2). Refer to Case Study No. 2 in Case Examples. Severe (Type 3) toe walking demonstrates no first rocker, premature third rocker timing, and an early predominant ankle moment (AM1 greater than AM2). Refer to Case Study No. 3 in Case Examples. Alvarez et al. (2007) did not find clinically significant differences in knee position between the three levels of toe walking severity.

Clinical Perspective Idiopathic toe walking in children causes considerable consternation with parents, teachers, and coaches. Initially the concern is in regard to a potentially undiagnosed condition in the child. Given the multitude of possible underlying diagnoses that present in part with equinus gait, the toe walker cannot be set aside as idiopathic until proven otherwise. This becomes increasingly important as the child ages into adolescence. Children who have toe walked since initiation of walking, presenting with a normal birth history, normal lower extremity tone, no weakness, and no known significant past medical history, are commonly idiopathic toe walkers. However, a child who started walking with a heel-toe gait and later developed a toe-toe pattern is a far more concerning scenario and requires the appropriate assessments and consultations with orthopedic surgery and neurology. Once a diagnosis of idiopathic toe walking is made, concern from the family is often alleviated. Considering the severity of toe walking in the clinical setting is important to determine if further assessment by 3D gait analysis is necessary. Observational gait analysis and ROM assessment may delineate which children require further gait evaluation. With respect to visual gait observation – can the child produce a normal heel strike without knee hyperextension or excessive external rotation? Can the child walk exclusively on their heels elevating the midfoot and forefoot off the ground? What degree of motion does the child have at the ankle? Do they have passive ankle DF greater than 10 with knee extension? Le Cras et al. (2011) recommend referral to the appropriate specialist if a child is unable to achieve 10 of passive ankle DF ROM with knee extension in spite of previous therapeutic intervention. A child who cannot heel walk and has ROM at the ankle below 5 PF is at one end of the spectrum, and at the other end is the child who has the ability to walk on their heels and has ROM in excess of 10 DF with the knee extended. It is often the moderate toe walker in between these two ends of the spectrum who needs further delineation by instrumented gait analysis. The intervention for a child who is either a severe toe walker or a moderate toe walker with significant compensations at other joint levels is worth exploring. There are limited long-term studies evaluating interventions for ITW. Through systematic review of the ITW literature, precursory evidence from a limited number of randomized controlled trials and a large number of non-randomized retrospective

1196

K. Davies et al.

cohort studies suggest that serial casting and surgical treatment have short-term improvements in passive ankle DF, gait kinematics and kinetics, and toe walking (van Bemmel et al. 2014; van Kuijk et al. 2014). Despite the fact that passive ankle DF improved more in those treated surgically versus by casting, there were no significant differences between treatment type with respect to reports of persistent toe walking (van Bemmel et al. 2014; van Kuijk et al. 2014). A more recent cohort study, without a concurrent control group, evaluated gait outcomes in 8 subjects 5 years postoperatively and found significant improvements in pelvic tilt, mean peak ankle DF during stance and swing, internal ankle moments, and ankle power (McMulkin et al. 2016). Reports from adult foot and ankle orthopedic surgeons describing reconstructive or salvage orthopedic procedures including a gastrocnemius recession or open TAL are also limited (Cychosz et al. 2015). Although persistent concerns of parents and patients regarding clumsiness and bullying around severe toe walkers exist, it is difficult to determine if these complaints are sufficient to consider them as indications for intervention. Previous gait studies of children with idiopathic clubfeet have shown that a TAL accounts for 27% loss of gastrocnemius/soleus muscle strength and subsequent 23% decrease in PF power (Karol et al. 1997). Given the long-term sequelae of untreated equinus deformity is unknown in otherwise typically developing children, a TAL may be an excessive intervention, particularly in the growing child. However, children with severe equinus have limited potential to generate forward propulsive energy through dynamic PF at the ankle due to loss of ankle ROM. Power generation at the ankle is the product of the mass moment of inertia and angular acceleration of the anatomical segment. Therefore, normal power generation requires normal PF power (ankle moment) and normal ankle ROM (angular acceleration). In a child or adolescent with severe equinus, a TAL may be warranted to restore normal ankle ROM and improve power – yet one must be cautious not to overlengthen the tendo-Achilles. An overlengthened tendo-Achilles will result in loss of PF strength and a decreased internal ankle moment, leading to decreased power generation at the ankle. The premise of being a toe walker is still unknown, but it is an engrained gait behavior pattern. To attempt elimination of the pattern, treatment must take advantage of neural plasticity theory and give the child the opportunity to relearn how to walk. First, given a physical abnormality exists, in terms of a tight gastrocnemius/ soleus muscle complex, treatment needs to achieve a correction such that the necessary ROM is possible. Secondly, retraining the gait pattern by preventing recurrent equinus is pivotal. Maintenance of normal heel to toe loading and properly timed and loaded rockers is crucial. Achievement of correction can be done with serial casting, casting following Botulinum Toxin, casting following a gastrocnemius/soleus recession, casting following a percutaneous TAL, or casting following an open TAL. A randomized controlled trial by Engström et al. (2013), however, has shown that the addition of Botulinum Toxin type A prior to casting did not improve passive ankle DF ROM, ankle DF strength, or gait outcomes of cast-only treatment at 1 year. Further clinical care guidelines for conservative management of ITW are outlined by Le Cras et al. (2011). Care must be taken to expose the child to the least possible risks for complications and emotional trauma. Maintenance is achieved

Idiopathic Toe Walking

1197

with the use of articulated ankle foot orthoses that allow DF and block PF. Initially this is used to retrain the child’s gait pattern, and then the orthoses can be weaned. Reassessment using instrumented gait analysis is important, and in the experience of the authors rewarding, as often the abnormalities approximate normal studies 1 year posttreatment.

Case Examples Case Study No. 1: Mild (Type 1) Idiopathic Toe Walking (Fig. 4) This is a 6-year-old male with ITW. On physical examination he has mild gastroc nemius tightness with 10 and 5 ankle DF in knee extension on the right and left side, respectively. On video observation, he demonstrates a variable weight acceptance pattern with occasional forefoot weight acceptance and inconsistent early heel rise bilaterally. Mild external foot progression angle is also noted. Ankle kinematics are nearly normal (Fig. 4a). There is a decreased, but present, first rocker on the right, with normal second rocker and third rocker formation bilaterally. There is normal ankle DF in swing bilaterally. Internal ankle moments demonstrate a normal AM2 bilaterally without evidence of an AM1 (Fig. 4b). Sagittal ankle joint powers are normal on the left and mildly decreased on the right (Fig. 4c). This child is accommodating with a mild external foot progression angle that is still within normal range (Fig. 4d). The foot-floor graph demonstrates the angle of the foot with respect to the floor during stance phase. As defined by the authors, normal heel rise occurs after the contralateral leg passes, indicated by the vertical lines in the graphs (Fig. 4e). This child has a normal foot-floor angle with no evidence of early heel rise. At the knee, this child demonstrates near normal kinematics with peak knee extension seen at initial contact (Fig. 4f). However, there is intermittent knee hyperextension in late stance on the left, which is a sign of adjacent level gait accommodation for a tight gastrocnemius. This case is an excellent example of the variability in gait pattern that is commonly seen in children with ITW. In some trials this child demonstrates normal gait kinematics and kinetics, while in other trials there is evidence of a tendency for early heel rise and the associated gait accommodations discussed earlier, including increased external foot progression angle and knee hyperextension in mid- to late stance phase.

Case Study No. 2: Moderate (Type 2) Idiopathic Toe Walking (Fig. 5) This is a 6-year-old male with ITW. On physical examination he has moderate   gastrocnemius tightness with 0 and 5 ankle DF in knee extension on the right and left side, respectively. On video observation, he demonstrates a variable weight acceptance pattern with early heel rise bilaterally. Mild external foot progression angle is present, right greater than left. Instrumented 3D gait analysis confirmed these findings. Ankle kinematics show an absent first rocker, abbreviated second

1198

K. Davies et al.

a

Ankle Joint Angles (deg) (Right)

Ankle Joint Angles (deg) (Left) 13.9 Plant-Dors

Plant-Dors

14.5

–5.3

–25.1

–5.3

–24.5 0.0

50.0

0.0

100.0

% Gait Cycle

b

Dors-Plant

Dors-Plant

1.4

0.5

0.0

50.0

0.5

–0.5 0.0

–0.4 100.0

c

Sagittal Joint Powers (Watts/kg) (Right)

100.0

Sagittal Joint Powers (Watts/kg) (Left)

3.5

3.1 Ankle (Abs-Gen)

Ankle (Abs-Gen)

50.0 % Gait Cycle

% Gait Cycle

1.3

–0.9 0.0

50.0

1.1

–0.8 0.0

100.0

d

e Toes Up-Down

Foot Progression 2.6 –14.4 –31.5 0.0

50.0

100.0

100.0

Foot-Floor Angle (Rep) 70.8

23.0 –24.9 0.0

% Gait Cycle

f

50.0 % Gait Cycle

% Gait Cycle

Ext-Int

100.0

Ankle Joint Moment (Nm/kg) (Left)

Ankle Joint Moment (Nm/kg) (Right) 1.5

50.0

100.0

% Stance Phase

Knee Joint Angles (deg)

Knee Joint Angles (deg) 62.7

Ext-Flex

65.7

Ext-Flex

50.0 % Gait Cycle

29.9

25.6

–11.5

–6.0 0.0

50.0 % Gait Cycle

100.0

0.0

50.0

100.0

% Gait Cycle

Fig. 4 Case study no. 1: mild (Type 1) idiopathic toe walker. (a) Ankle kinematics. (b) Internal ankle moments (kinetics). (c) Ankle sagittal joint power (kinetics). (d) Representative foot progression angle. (e) Representative foot-floor angle. (f) Knee kinematics. The gray bands represent normative data from 34 healthy children aged 5 to 18 years within 1 SD of the mean

Idiopathic Toe Walking

b

Ankle Angle (Rep) 14.5

Toes Up-Down

Plant-Dors

a

1199

–8.7

–31.9 0.0

50.0

Foot-Floor Angle (Rep) 73.0

24.1

–24.9 0.0

100.0

% Gait Cycle

c

d

Ankle Moment (Rep)

Abs-Gen

Flex-Ext

1.5

0.5

–0.5 0.0

50.0

1.2

% Gait Cycle

e

50.0

100.0

% Gait Cycle

Knee Flexion - Extension (Rep)

f

65.7

Foot Progression 2.6

Ext-Int

Ext-Flex

100.0

Ankle Power (Rep)

3.5

–1.2 0.0

100.0

50.0 % Stance Phase

29.9

–21.2 –45.0

–6.0 50.0

0.0

100.0

0.0

Gastrocnemius

g

50.0

100.0

% Gait Cycle

% Gait Cycle

EMG Gastrocnemius (Right) 0.25 0.13 0.00 0.0

50.0

100.0

% Gait Cycle

Fig. 5 Case study no. 2: moderate (Type 2) idiopathic toe walker. (a) Representative ankle kinematics. (b) Representative foot-floor angle. (c) Representative internal ankle moments (kinetics). (d) Representative ankle sagittal joint power (kinetics). (e) Representative knee kinematics. (f) Representative foot progression angle. (g) Example of electromyography. The gray band represents normative data from 34 healthy children aged 5 to 18 years within 1 SD of the mean

rocker, and normal third ankle rocker formation. There is decreased ankle DF in swing bilaterally (Fig. 5a). Early heel rise is seen bilaterally with increased foot-floor angle initiated prior to 30% of the gait cycle (Fig. 5b). Internal ankle moments demonstrate the presence of an AM1 bilaterally, which are of equal or lesser value than AM2 (Fig. 5c). Sagittal ankle joint powers demonstrate early power generation corresponding to the early heel rise and decreased terminal stance phase power generation (Fig. 5d). Gait accommodations seen in this child include knee

1200

K. Davies et al.

hyperextension in late stance (Fig. 5e) and external foot progression angle (Fig. 5f). EMG shows preparatory gastrocnemius activation in late swing phase contributing to the preparatory positioning of the foot in equinus, in addition to activation throughout stance (Fig. 5g).

Case Study No. 3: Severe (Type 3) Idiopathic Toe Walking (Fig. 6) This is a 10-year-old male with severe ITW. On physical examination he has   gastrocnemius tightness with 10 and 20 ankle DF in knee extension on the right and left side, respectively. On video observation, he demonstrates a toe-toe weight acceptance pattern with brief heel contact and early heel rise bilaterally. There is an external foot progression angle bilaterally. Instrumented 3D gait analysis confirmed these findings. Ankle kinematics show an absent first and second rocker and decreased third rocker bilaterally. He is in ankle PF throughout stance phase and has decreased ankle DF in swing (Fig. 6a). The foot-floor angle also demonstrates PF throughout stance and early heel rise bilaterally occurring in early stance phase (Fig. 6b). Internal ankle moments demonstrate a predominant AM1 that is greater than AM2 bilaterally (Fig. 6c). Sagittal ankle joint powers show early power generation corresponding to the early heel rise and decreased terminal stance phase power generation (Fig. 6d). Gait accommodations seen in this child include knee hyperextension in late stance (Fig. 6e) and external foot progression angle, left greater than right (Fig. 6f). EMG shows preparatory gastrocnemius activation in late swing phase and activation throughout stance corresponding with early heel rise and terminal stance push-off (Fig. 6g). In comparison to the previous example of Type 2 ITW, this child shows increased severity with respect to his gastrocnemius contracture, predominant first ankle moment, and more severe external foot progression angle.

Summary Idiopathic toe walking is a condition of childhood characterized by a bilateral toe-toe gait pattern of unknown cause. Persistent toe walking is considered abnormal in typically developing children after 2–3 years of age and is a diagnosis of exclusion. Other neurologic and orthopedic conditions such as cerebral palsy, muscular dystrophy, peripheral neuropathy, and spinal cord abnormalities must be excluded prior to obtaining a diagnosis of ITW. The role of clinical gait analysis in children with ITW is to identify the condition and describe the necessary accommodations and impairments affecting their gait. It can be used to classify the severity of their condition and help identify children most likely to benefit from treatment. Children with a long-standing history of ITW typically present with decreased passive ankle DF ROM. Ankle kinematics often show a variable pattern with absent first rocker, early peak ankle DF in stance, and early heel rise occurring prior to 30% of the gait cycle. Swing phase ankle kinematics show normal ankle DF in early swing with

Idiopathic Toe Walking

1201

Ankle Angle (Rep)

a

72.8

Toes Up-Down

14.5 Plant-Dors

Foot-Floor Angle (Rep)

b

–9.6

–33.7

50.0

0.0

24.0

–24.9 0.0

100.0

% Gait Cycle

c

d

Ankle Moment (Rep)

Abs-Gen

Flex-Ext

100.0

Ankle Power (Rep) 3.5

1.6

0.5

–0.5

0.4

–2.7 0.0

50.0

100.0

0.0

% Gait Cycle

e

50.0

100.0

% Gait Cycle

Knee Flexion - Extension (Rep)

Foot Progression

f

65.7

2.6 Ext-Int

Ext-Flex

50.0 % Stance Phase

29.2

–19.9 –42.5

–7.2 0.0

50.0

0.0

100.0

Gastrocnemius

g

50.0

100.0

% Gait Cycle

% Gait Cycle

EMG Gastrocnemius (Left) 0.25 0.13

0.00 0.0

50.0

100.0

% Gait Cycle

Fig. 6 Case study no. 3: severe (Type 3) toe walker. (a) Representative ankle kinematics. (b) Representative foot-floor angle. (c) Representative internal ankle moments (kinetics). (d) Representative ankle sagittal joint power (kinetics). (e) Representative knee kinematics. (f) Representative foot progression angle. (g) Example of electromyography. The gray band represents normative data from 34 healthy children aged 5 to 18 years within 1 SD of the mean

decreased DF in late swing and poor ankle pre-positioning for initial foot contact. Ankle kinetics show an early internal ankle moment (AM1) and a diminished late internal ankle moment (AM2); in more severe cases, the early ankle moment (AM1) is greater than the late ankle moment (AM2). Pedobarography demonstrates decreased hindfoot pressure and medialization of the foot center of pressure secondary to an increased external foot progression angle. EMG is found to be similar in patients with cerebral palsy and other causes of toe walking, presenting with abnormal co-contraction and out-of-phase muscle activity with premature firing of gastrocnemius in swing as well as low-amplitude tibialis anterior firing throughout

1202

K. Davies et al.

stance and swing. Gait accommodations seen with ITW often include knee hyperextension in late stance and increased external foot progression angle. Three-dimensional gait analysis is useful for identifying the child with moderate toe walking who has multilevel gait accommodations not easily appreciated with observational gait assessment. Treatment options to address the gastrocnemius contracture include serial casting, casting with Botulinum Toxin, or casting following gastrocnemius recession or percutaneous/open TAL. After the initial intervention, it is important to follow immediately with ankle-foot orthoses for gait retraining and maintenance of ankle ROM in the growing child.

Cross-References ▶ 3D Dynamic Pose Estimation from Marker-Based Optical Data ▶ Clinical Gait Assessment by Video Observation and 2D Techniques ▶ The Conventional Gait Model - Success and Limitations ▶ EMG Activity in Gait: The Influence of Motor Disorders ▶ Functional Effects of Foot Orthoses ▶ Integration of Foot Pressure and Foot Kinematics Measurements for Medical Applications ▶ Interpreting Joint Moments and Powers in Gait ▶ Interpreting Spatiotemporal Parameters, Symmetry, and Variability in Clinical Gait Analysis

References Accardo P, Barrow W (2015) Toe walking in autism: further observations. J Child Neurol 30(5): 606–609 Accardo P, Morrow J, Heaney M et al (1992) Toe walking and language development. Clin Pediatr (Phila) 31:158–160 Alvarez C, De Vera M, Beauchamp R et al (2007) Classification of idiopathic toe walking based on gait analysis: development and application of the ITW severity classification. Gait Posture 26:428–435 Alvarez C, De Vera M, Chhina H et al (2008) Normative data for the dynamic pedobarographic profiles of children. Gait Posture 28:309–315 Baker R (2013) Measuring walking: a handbook of clinical gait analysis. MacKeith Press, London Barrow W, Jaworski M, Accardo P (2011) Persistent toe walking in autism. J Child Neurol 26:619–621 Bovens A, van Baak M, Vrencken J et al (1990) Variability and reliability of joint measurements. Am J Sports Med 18(1):58–63 Bowen T, Miller F, Castagno P et al (1998) A method of dynamic foot-pressure measurement for the evaluation of pediatric orthopaedic foot deformities. J Pediatr Orthop 18(6):789–793 Brunt D, Woo R, Kim H et al (2004) Effect of botulinum toxin type A on gait of children who are idiopathic toe-walkers. J Surg Orthop Adv 13(3):149–155 Burnett C, Johnson E (1971) Development of gait in childhood II. Dev Med Child Neurol 13(2): 207–215

Idiopathic Toe Walking

1203

Clark E, Sweeney J, Yocum A et al (2010) Effects of motor control intervention for children with idiopathic toe walking: a 5-case series. Pediatr Phys Ther 22:417–426 Crenna P, Redrizzi E, Andreucci E et al (2005) The heel-contact gait pattern of habitual toe walkers. Gait Posture 21(3):311–317 Cychosz C, Phisitkul P, Belatti D et al (2015) Gastrocnemius recession for foot and ankle conditions in adults: evidence-based recommendations. Foot Ankle Surg 21(2):77–85 DiGiovanni C, Kuo R, Tejwani N et al (2002) Isolated gastrocnemius tightness. J Bone Joint Surg 84(6):962–970 Eastwood D, Dennett X, Shield L et al (1997) Muscle abnormalities in idiopathic toe-walkers. J Pediatr Orthop Part B 6:215–218 Eastwood D, Memelaus M, Dickens D et al (2000) Idiopathic toe-walking: does treatment alter the natural history? J Pediatr Orthop 9:47–49 Engelbert R, Gorter J, Uiterwaal C et al (2011) Idiopathic toe-walking in children, adolescents and young adults: a matter of local or generalised stiffness? BMC Musculoskelet Disord 12:61 Engström P, Tedroff K (2012) The prevalence and course of idiopathic toe-walking in 5-year-old children. Pediatrics 130(2):279–284 Engström P, Bartonek A, Tedroff K et al (2013) Botulinum toxin A does not improve the results of cast treatment for idiopathic toe-walking. A randomized controlled trial. J Bone Joint Surg Am 95:400–407 Fox A, Deakin S, Pettigrew G et al (2006) Serial casting in the treatment of idiopathic toe-walkers and review of the literature. Acta Orthop Belg 72:722–730 Furrer F, Deonna T (1982) Persistent toe-walking in children: a comprehensive clinical study of 28 cases. Helv Paediatr Acta 37:301–316 Gage J (2004) A qualitative description of normal gait. In: Gage J (ed) The treatment of gait problems in cerebral palsy. Mac Keith Press, London, pp 42–70 Griffin P, Wheelhouse W, Shiavi R et al (1977) Habitual toe-walkers. a clinical and electromyographic gait analysis. J Bone Joint Surg Am 49(4):97–101 Grimston S, Nigg B, Hanley D et al (1993) Differences in ankle joint complex range of motion as a function of age. Foot Ankle 14(4):215–222 Hall J, Salter R, Bhalla S (1967) Congenital short tendo calcaneus. J Bone Joint Surg (Br) 49(4): 695–697 Hemo Y, Macdessi S, Pierce R et al (2006) Outcome of patients after achilles tendon lengthening for treatment of idiopathic toe walking. J Pediatr Orthop 26(3):336–340 Hicks R, Durinick N, Gage J (1988) Differentiation of idiopathic toe-walking and cerebral palsy. J Pediatr Orthop 8(2):160–163 Hill R (1995) Ankle equinus. Prevalence and linkage to common foot pathology. J Am Podiatr Med Assoc 85(6):295–300 Hirsch G, Wagner B (2004) The natural history of idiopathic toe-walking: a long-term follow-up of fourteen conservatively treated children. Acta Pediatr 93:196–199 Kalen V, Adler N, Bleck E (1986) Electromyography of idiopathic toe walking. J Pediatr Orthop 6:31–33 Karol L, Concha M, Johnston C (1997) Gait analysis and muscle strength in children with surgically treated clubfeet. J Pediatr Orthop 17(6):790–795 Kelly I, Jenkinson A, Stephens M et al (1997) The kinematic patterns of toe-walkers. J Pediatr Orthop 17(4):478–480 Le Cras S, Bouck J, Brausch S et al (2011) Cincinnati Children’s Hospital Medical Center: evidence-based clinical care guideline for management of idiopathic toe walking. Guideline 040, pp 1–17. Available via http://www.cincinnatichildrens.org/service/j/anderson-center/evi dence-based-care/occupational-therapy-physical-therapy/. Accessed 1 Aug 2016 McMulkin M, Baird G, Caskey P et al (2006) Comprehensive outcomes of surgically treated idiopathic toe walkers. J Pediatr Orthop 26(5):606–611 McMulkin M, Gordon A, Tompkins B et al (2016) Long term gait outcomes of surgically treated idiopathic toe walkers. Gait Posture 44:216–220

1204

K. Davies et al.

Ming X, Brimacombe M, Wagner G (2007) Prevalence of motor impairment in autism spectrum disorders. Brain and Development 29:565–570 Papariello S, Skinner S (1985) Dynamic electromyography analysis of habitual toe-walkers. J Pediatr Orthop 5:171–175 Perry J (1985) Normal and pathologic gait. In: Bunch W (ed) Atlas of orthotics, 2nd edn. C.V. Mosby, St. Louis, pp 76–111 Perry J (1992) Gait analysis: normal and pathological function. Slack Incorporated, Thorofare Pierz K, Õunpuu S (2013) Toe walking: how to know who to worry about. Paper presented at the 67th American Academy of Cerebral Palsy and Developmental Medicine Meeting, Milwaukee, 16–19 October 2013 Pomarino D, Ramirez Llamas J, Pomarino A (2016) Idiopathic toe walking tests and family predisposition. Foot Ankle Spec 20(10):1–6 Sawatzky B, Alvarez C, Beauchamp R et al (1999) Adaptations of gait on the contralateral limb in children with spastic hemiplegia. Paper presented at the 17th Congress of the International Society for Biomechanics, Calgary, 8–13 August 1999 Shulman L, Sala D, Chu M et al (1997) Developmental implications of idiopathic toe walking. J Pediatr 130(4):541–546 Sobel E, Caselli M, Velez Z (1997) Effect of persistent toe walking on ankle equinus. Analysis of 60 idiopathic toe walkers. J Am Podiatr Med Assoc 87(1):17–22 Stott N, Walt S, Lobb G et al (2004) Treatment for idiopathic toe walking: results at skeletal maturity. J Pediatr Orthop 24(1):63–69 Stricker S, Angulo J (1998) Idiopathic toe walking: a comparison of treatment methods. J Pediatr Orthop 18(3):289–293 Sutherland D, Olsen R, Cooper L et al (1980) The development of mature gait. J Bone Joint Surg Am 62(3):336–353 Tabrizi P, McIntyre W, Quesnel M et al (2000) Limited dorsiflexion predisposes to injuries of the ankle in children. J Bone Joint Surg 82-B(8):1103–1106 van Bemmel A, van de Graaf V, van den Bekerom M et al (2014) Outcome after conservative and operative treatment of children with idiopathic toe walking: a systematic review of the literature. Musculoskelet Surg 98:87–93 Van Kuijk A, Kosters R, Vugts M et al (2014) Treatment for idiopathic toe walking: a systematic review of the literature. J Rehabil Med 46:945–957 Walker J (1991) Musculoskeletal development: a review. Phys Ther 71(12):878–889 Westberry D, Davids J, Davis R et al (2008) Idiopathic toe walking: a kinematic and kinetic profile. J Pediatr Orthop 28(3):252–258 Williams C, Tinley P, Curtin M et al (2012) Vibration perception thresholds in children with idiopathic toe walking gait. J Child Neurol 27(8):1017–1021 Williams C, Michalitsis J, Murphy A et al (2013) Do external stimuli impact the gait of children with idiopathic toe walking? A study protocol for a within-subject randomized control trial. BMJ Open 3(3):e0002389 World Health Organization (2007) International classification of functioning, disability and health: children and youth version: ICF-CY. World Health Organization, Geneva

Gait Disorders in Persons After Stroke Johanna Jonsdottir and Maurizio Ferrarin

Abstract

Muscle weakness and motor control deficits associated with stroke result in fairly typical hemiparetic gait. This chapter will focus on persons with hemiparesis after stroke in the post-acute and chronic stage, their major movement disorders, and how these affect locomotion. Differences in spatiotemporal, kinematic, and kinetic parameters that commonly characterize hemiparetic gait in the subacute and chronic post-stroke periods will be reviewed in reference to characteristics of speed-matched gait of nondisabled persons. Keywords

Gait • Stroke • Gait analysis • Kinematics • Kinetics • Coactivation • Hemiparetic gait • Movement disorders • Speed-matched gait • Muscles • Weakness

Contents Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . State of the Art . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Importance of Gait Velocity in Hemiparetic Gait . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Characteristics of Hemiparetic Gait Independent of Gait Speed . . . . . . . . . . . . . . . . . . . . . . . . . . . Kinetics of Gait . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

1206 1207 1207 1208 1209

J. Jonsdottir (*) LaRiCE, Department of Neurorehabilitation, IRCCS Fondazione Don Carlo Gnocchi Onlus, Milan, Italy e-mail: [email protected] M. Ferrarin Biomedical Technology Department, IRCCS Fondazione Don Carlo Gnocchi Onlus, Milan, MI, Italy e-mail: [email protected] # Springer International Publishing AG, part of Springer Nature 2018 B. Müller, S.I. Wolf (eds.), Handbook of Human Motion, https://doi.org/10.1007/978-3-319-14418-4_61

1205

1206

J. Jonsdottir and M. Ferrarin

Neuromuscular Factors Affecting Gait . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Voluntary Changes in Gait Speed . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

1210 1213 1213 1214

Introduction Stroke is the most frequent cause of disability in industrialized countries, with hemiparesis the most common motor impairment affecting approximately 65% of stroke victims (WHO 2003). Stroke typically affects the upper motor neuron pathways, and hemiparesis following stroke is due to a unilateral disruption of descending neural pathways originating in the cortex or brain stem. The symptoms seen after stroke depend on the localization of the damage, but mostly a mixed pattern of muscle weakness and overactivity occurs, and in the period following stroke, recovery and compensatory strategies develop in the central and peripheral systems (Sheean and McGuire 2009). Persons after stroke may present different levels of motor and cognitive disorders depending on several factors, such as severity of stroke, time after stroke, and age (Teasell et al. 2005). Three bands/categories of stroke patients were proposed by Garraway (1985) based on severity of symptoms. An upper band of persons showing minimal deficits tend to recover well with minimal movement disorders that can be treated in a community or outpatient setting. A middle band of persons that have suffered moderately severe strokes, while presenting various motor and cognitive problems, recover markedly in most areas although remaining dependent on rehabilitation. Most of them become walkers and are discharged to the community after a period of rehabilitation. The lower band of persons are those who suffered a severe stroke and remained severely paretic, often with serious medical comorbidities which add to their stroke-related disability. This group most often does not achieve independent ambulation outside the house. The main focus in this chapter will be on mobility disorders in the above-defined upper-to-middle band of persons after stroke that are ambulatory with or without assistance. One of the characteristics of stroke is the functional change that occurs from the first hours after stroke to days and sometimes years after stroke. The damaged brain goes through recovery processes typical of the immediate post-stroke period of approximately 48 h, and then in the months following stroke, there is continued recovery through reorganization that occurs spontaneously and in response to therapy, as well as due to compensatory strategies developed to deal with daily activities. Even in the chronic stage after stroke, there is continued possibility of further functional improvements both through recovery and compensation (Teasell et al. 2005). This characteristic functional change in persons after stroke is reflected also in longitudinal improvements in kinematic and kinetic gait parameters although rarely do they return to pre-stroke values. Six months after stroke about 75–85% of persons

Gait Disorders in Persons After Stroke

1207

with resultant motor and sensorimotor impairment will have recovered the ability to walk at least 45 m without physical assistance (Patel et al. 2000). Mobility is achieved through interaction of various systems, of which the central nervous system and the peripheral neuromuscular and sensory systems play an important role. Standing and walking require (i) the retaining of the supporting of the body against gravity in order to prevent a fall and (ii) a propulsion force in order to move (Clark 2015). Problems at the neural, sensory, and/or muscular level, as well as problems with coordination among the systems, are all possible causes of gait deficits in persons after stroke. The resultant movement disorders, such as spasticity, ataxia, sensory, or proprioceptive deficits that are common in persons after stroke and evident on neurologic examination, all affect motor control and consequently gait performance (Snijders et al. 2007; Clark 2015). The majority of persons that have suffered a stroke show decreased gait velocity and abnormal gait kinematics and kinetics. Their gait is characterized by spatial and temporal asymmetries, with poor selective motor control, reduced weight bearing on the paretic limb, and reduced propulsive force generated by the paretic limb (Chen et al. 2005; Balasubramanian et al. 2007; Allen et al. 2011; Balaban and Tok 2014).

State of the Art Importance of Gait Velocity in Hemiparetic Gait Gait velocity is an important indicator of locomotor function and quality of life after stroke. Preferred (self-selected) velocity, which is the gait speed a person exhibits spontaneously when walking at ease, is considered an indicator of overall gait performance. It is commonly used to monitor locomotor performance and to evaluate the effects of rehabilitation in persons with neurological disorders (Perry et al. 1995). Elderly healthy individuals have a preferred speed of between 1.1 and 1.3 m/s (Turnbull et al. 1995; Lusardi et al. 2003) but can easily change their speed from slow to running. In general, walking speeds greater than 0.8 m/s indicate that a person can independently move around in the external environment (Perry et al. 1995; Lord et al. 2004). Being independent community walkers makes it more likely that a person can participate in community activities, including working. Gait speed of persons after stroke differs markedly from their healthy counterparts in that they tend to walk much slower, which considerably influences their participation in activities of daily life. Only about 40% of persons who recover walking abilities after stroke achieve community-walking velocities, while half of them will remain at gait speeds less than 0.5 m/s (Goldie et al. 2001; Macko et al. 2001; Lamontagne et al. 2007). Post-stroke persons are also generally more limited in their choice of gait speed; however, when asked some of them are able to increase their gait speed considerably or at least up to 0.2–0.3 m/s more than their spontaneous speed (Bayat et al. 2005; Jonsdottir et al. 2009). This indicates the availability of additional functional resources that are not commonly utilized in daily walking, maybe due

1208

J. Jonsdottir and M. Ferrarin

to a need to maintain balance, or divide attention between walking and various environmental requests. Along with reduced speed, studies of kinematic and kinetic characteristics of main features of hemiparetic gait of persons post-stroke through gait analysis have evidenced reduced cadence, stride length, and joint angular excursion, as well as asymmetries between the affected and non-affected leg in most gait parameters. Further, a reduction in force output from the affected leg and a general increased mechanical energetic cost have been evidenced (Roerdink and Beek 2011; Chen et al. 2005; Mulroy et al. 2003; Jonsdottir et al. 2009).

Characteristics of Hemiparetic Gait Independent of Gait Speed Walking velocity, however, tends to correlate with many of the measured temporal and spatial parameters of the gait cycle, and many of the gait deviation characteristics of hemiparetic gait are consistent with slower walking in nondisabled individuals. This underlines the importance of characterizing gait of persons post-stroke independent of gait speed (Chen et al. 2005; Jonsdottir et al. 2009; Wagenaar and Beek 1992; Rinaldi and Monaco 2013). This can be achieved by comparing the various gait parameters of the gait phases with nondisabled persons walking at similar speeds. Chen and colleagues (2005) studied differences in kinematic and kinetic parameters of speed-matched gait in five hemiparetic persons post-stroke and nondisabled controls walking on treadmill. They identified a large set of gait differences consistent with compensatory strategies related to impaired swing initiation in the affected limb and shortened single limb support on the affected limb. The impaired swing initiation was associated with inadequate propulsion of the leg during pre-swing, increased percentage of swing time of the affected leg, as well as reduced knee flexion at toe-off and mid-swing. This led the persons with stroke to compensate with leg circumduction or pelvis hiking during the swing phase. The shortened support time on the paretic limb instead was related to exaggerated propulsion of the non-affected limb during pre-swing to shorten its swing time. Other gait deficits independent of gait speed included asymmetric step length and increased step width. However, when speed-matched, spatiotemporal parameters cadence, stride time and stride length were similar between gait of persons poststroke and nondisabled persons and so were ankle plantar flexion angles at toe-off in both limbs. These findings of similar spatiotemporal parameters at matched speed are corroborated with findings from another study comparing the gait of 39 hemiparetic persons post-stroke with the gait of speed-matched healthy controls (Jonsdottir et al. 2009). The persons, all in the chronic phase after stroke and moderately affected, walked at their preferred speed and then as fast as possible for gait analysis. Their spatiotemporal parameters were compared with velocity dependent profiles of nondisabled persons. Algorithms were used to classify the resulting gait parameters as

Gait Disorders in Persons After Stroke

1209

reduced, normal, or increased (relative to two standard deviations from the mean of the nondisabled sample) at preferred and fast speed. At preferred speed cadence and stride length of the persons post-stroke tended to be similar to those of the non-disabled persons walking at matched speed, only about one quarter of the sample had increased cadence and/or reduced stride length, while the rest were within normal limits. During fast speed half of the sample had increased cadence and was within normal limits, while stride length remained similar with three quarters of the sample being within normal limits (Jonsdottir et al. 2009). Rinaldi and Monaco (2013) demonstrated similar findings from a group of persons poststroke that walked on treadmill. Their gait parameters were compared to those of nondisabled subjects walking at controlled matched speed, resulting in a reduction of differences in spatiotemporal parameters between the persons post-stroke and the nondisabled persons.

Kinetics of Gait Preferred gait speed is related to stride length, and increases in speed are associated with increased stride length and/or cadence. However, this always necessitates an associated production of mechanical work at the distal and proximal joints (Olney and Richards 1996; Jonsdottir et al. 2009). Persons post-stroke can, depending on their functional status, distribute this work production between the distal and proximal joints in different ways. Kinetic analysis of gait includes the measurement of ground reaction forces through which joints moment and power can be determined. Through knowledge of kinetics, it is possible to gain insight into the pathogenesis of walking impairment post-stroke. In both nondisabled and hemiparetic gait, the muscle groups that mainly contribute to the generation of energy for forward propulsion are the hip extensors in early stance, the plantar flexors at push-off, and the hip flexors at pre-swing (Olney et al. 1994; Parvatanemi et al. 2007). In nondisabled persons, faster walking speeds are associated with larger power generation bursts from the same muscle groups with the ankle plantar flexors providing about 75% of the force contribution (Winter 1990). In hemiparetic gait, the magnitude of power generation at both the hip and ankle is reduced with patterns of muscle activation, tending to differ in both timing and amplitude from normal values when both walk at self-selected speed (Winter 1990). However, when the joint power production of post-stroke persons is compared with that of healthy controls walking at matched speed, the most abnormal parameter is the power generation at the ankle, while power generation at the hip appears to be mostly in the range of that of healthy subjects (Jonsdottir et al. 2009; Hsu et al. 2003). It is likely that post-stroke persons compensate for poor power generation at the ankle by increasing the generation of power at the hip. According to Hsu et al. (2003) , the strength of the hip flexor muscles and the knee extensor muscles of the hemiparetic limb is the most important factor determining comfortable or fast walking speed. However, also plantar flexor muscles affected walking speed. When post-stroke persons are asked to walk at

1210

J. Jonsdottir and M. Ferrarin

maximum, or near maximum, gait velocity capacity, poor distal power generation appears to be compensated for either by increasing power generation proximally or by increasing the ankle power generation of the non-affected limb (Nadeau et al. 1999; Hsaio et al. 2016; Awad et al. 2015). These findings were corroborated by findings from Jonsdottir and colleagues (2009). When gait parameters of post-stroke persons walking at their preferred speed were compared to those of nondisabled persons walking at matched speeds, the most obvious difference was in positive work at the ankle with 64% of the hemiparetic sample showing significantly lower values than their healthy counterparts (see Fig. 1a from Jonsdottir et al. 2009). This difference became even larger at faster speed when over 80% of the hemiparetic sample produced significantly less work at the ankle. Instead, hip flexor positive work of post-stroke subjects was comparable to that of nondisabled subjects with 89% and 75% of the subjects demonstrating normal values during self-selected and fast gait, respectively (see Fig. 1b from Jonsdottir et al. 2009).

Neuromuscular Factors Affecting Gait Alterations in the normal activation patterns of typical hemiparetic gait include increased, early, delayed, and prolonged activation of muscles in the lower limb during gait activities as well as coactivation of agonist and antagonist muscles (Perry et al. 1995; Sheffler and Chae 2015). Muscle coactivation or timing problems at the knee and/or ankle joints are frequent in persons after stroke. Even in the less affected limb, excessive coactivation has been reported during both double support phases of gait. This general increase in coactivation in persons after stroke has been implicated to be an adaptive behavior to changed neuromuscular dynamics, namely, reduced plantar flexor strength that may be a limit on safe weight transfer during the double support phases (Lamontagne et al. 2000; Den Otter et al. 2007; Rosa et al. 2014). In order to compensate for the ankle plantar flexor weakness, the persons with stroke appear to increase coactivation. At the knee level, a characteristic of hemiparetic gait is a genu recurvatum or excessive knee extension in midstance caused by exaggerated activation or contracture of the plantar flexors and/or the quadriceps, hamstring, and/or quadriceps weakness. It has been demonstrated that when there is hyperextension of the affected leg during midstance, the plantar flexor produces inadequate power at terminal stance (Cooper et al. 2012). At the ankle level, foot drop is quite a common problem in hemiparetic gait and has been associated with insufficient strength of dorsiflexor muscles and/or coactivation of plantar flexors. Foot drop is noticeable during swing and may be the only manifestation of hemiparetic gait in less severely impaired persons after stroke. In more impaired post-stroke persons, this may become evident at initial foot

Gait Disorders in Persons After Stroke

1211

Fig. 1 Plots of ankle positive work (a) and hip positive work (b) vs. normalized gait velocity; individual patients’ values are reported at preferred velocity (circle) and max speed (triangle). The normal profile and its variability (_1 S.D. range as dash-dotted line) are also shown. The external dotted lines mark the bounds (_1.96 S.D.) for statistical analysis (Figure from Jonsdottir et al. 2009)

1212

J. Jonsdottir and M. Ferrarin

contact, so that the foot touches the ground in a flat position or even that the forefoot touches the ground first. The foot drop appears to be due to prolonged gastrocnemius-soleus activation that then leads to inadequate dorsiflexion during swing (Sheffler and Chae 2015). Another common problem during hemiparetic gait is equinovarus; this occurs when foot drop is associated with excessive plantar flexion and inversion. Prolonged activation of the tibialis muscles leads to initial contact occurring with the lateral border of the foot; weight bearing is thus shifted laterally and can lead to instability at the ankle and increased risk of falling (Sheffler and Chae 2015). Coactivation is particularly evident at the ankle during stance and at the knee during weight acceptance. It is likely that this increased muscle coactivation seen during post-stroke gait is a compensatory strategy to compensate for muscle weakness and postural instability. This is consistent with the suggestions of Martino et al. (2015) that the nervous system copes with unstable conditions, such as those associated to several pathologies, by prolonging the duration of basic muscle activity patterns. In support of this are the findings of Kitatani and colleagues that investigated the gait of a group of 44 ambulatory post-stroke persons and found that impaired balance ability and decreased ankle plantar flexor strength on the affected side predicted coactivation of the ankle muscles (Kitatani et al. 2016). This increase in ankle muscle coactivation during gait after stroke might thus be a compensatory strategy for impaired balance and ankle muscle weakness, as well as a sign of the disturbed motor control following a lesion to the neuromotor system (Kitatani et al. 2016). The plantar flexors appear to have a unique role during nondisabled gait, in that they are the only muscle group that can regulate angular momentum throughout the gait cycle and also appear essential in maintaining dynamic balance during walking (Neptune and McGowan 2011). Interesting findings by Honeine et al. (2013, 2014) highlighted the role of the triceps surae in nondisabled gait. Their results indicated that triceps activity during gait set the kinematics and kinetics of gait through counteracting the force of gravity. In the first study (2013), they demonstrated that modulation of triceps surae in both amplitude and duration allows the central nervous system to adapt step length and cadence. By setting the amplitude and duration of triceps surae activation, the inherent disequilibrium during gait is controlled, and consequently gait velocity is set. When subjects walked faster, both the braking action and triceps activity increased. In a second study, Honeine et al. (2014) demonstrated that the surge of triceps activity in faster walking is tied to increased requirement of the center of mass vertical braking action rather than an increase in propulsion. The central nervous system thus modulated step length and cadence through the control of triceps activity of the stance leg in the form of a braking action of the fall on the contralateral advancing leg. The authors speculated that in pathologies where interaction between brain, central pattern generators, motoneurons and sensory feedback is affected, such as after a stroke, the step length becomes inconsistent and asymmetric. If the controlled variable during changing of speed in healthy gait is the activity of triceps to break the fall on the advancing leg, it is likely that in most post-stroke persons, this

Gait Disorders in Persons After Stroke

1213

functional resource is not readily available, and thus their neuromotor system has to readapt its kinetics and kinematics to impose speed changes. Awad and colleagues (2015) further investigated the importance of ankle propulsion during longer distance walking. Forty four individuals walked over ground during gait analysis and did the 6 minutes walking test. The authors demonstrated that the paretic limb’s ability to generate propulsion was the determining factor in long distance walking function. The biomechanical variables during stance phase propulsion and trailing limb angle were related to distance walked, while swing phase parameters, paretic ankle dorsiflexion and knee flexion, and symmetry parameters, step length and swing time, did not influence long distance walking function.

Voluntary Changes in Gait Speed When it comes to persons post-stroke, a determinant characteristic of their gait is a deficit in muscle activation at the ankle; however, it appears that the capacity to modulate the timing of the force production may be mostly preserved. When asked to increase gait velocity from self-selected to fast, a typical strategy of persons after stroke is the more-than-normal advance of the ankle power onset of the affected leg, while there is little or no increase in positive work of the plantar flexors (Jonsdottir et al. 2009). As mentioned above, during gait this deficit in propulsion capacity can be compensated by more proximal propulsive mechanisms to achieve faster gait velocity (Neptune and McGowan 2011; Hsu et al. 2003; Beaman et al. 2010). According to Hsu et al. (2003) the strength of the hip flexor muscles and the knee extensor muscles of the hemiparetic limb are the most important factors determining comfortable or fast walking speed. Further, it has been demonstrated that the nonparetic plantar flexor muscles are among the primary mechanisms in speed modulation (Hsaio et al. 2016).

Summary The majority of persons with mild-to-moderate sequelae post-stroke have reduced gait velocity and abnormal gait kinematics and kinetics that can influence their quality of life, including reduced stride length, asymmetric step lengths, and long distance walking. Many of their gait deficits may be due to lack of torque-generating capacities of the leg muscles and incoordination of the neuromuscular system, as well as general instability. However, it is evident from the literature that comparison of gait characteristics of persons that have had stroke should be made with nondisabled persons walking at matched velocities, since many gait parameters are sensitive to speed. Some apparent abnormalities in hemiparetic gait disappear when speed is controlled for (Chen et al. 2005; Jonsdottir et al. 2009; Wagenaar and Beek 1992; Rinaldi and Monaco 2013). A determinant characteristic of the hemiparetic gait is a deficit in propulsion capacity at the affected ankle; however, it

1214

J. Jonsdottir and M. Ferrarin

appears that the capacity to modulate the timing of the force production may be mostly preserved. During gait this deficit in propulsion capacity can be compensated for by more proximal propulsive mechanisms or by an increase in propulsion of the plantar flexors on the contralateral side in order to maintain a self-selected gait velocity or to achieve faster gait velocity (Jonsdottir et al. 2009; Nadeau et al. 1999; Hsaio et al. 2016; Awad et al. 2015).

References Allen JL, Kautz SA, Neptune RR (2011) Step length asymmetry is representative of compensatory mechanisms used in post-stroke hemiparetic walking. Gait Posture 33(4):538–543. https://doi. org/10.1016/j.gaitpost.2011.01.004. Epub 2011 Feb 11 Awad LN, Binder-Macleod SA, Pohlig RT, Reisman D (2015) Paretic propulsion and trailing limb angle are key determinants of long-distance walking function after stroke. Neurorehabil Neural Repair 29(6):499–508. https://doi.org/10.1177/1545968314554625. Epub 2014 Nov 10 Balaban B, Tok F (2014) Gait disturbances in patients with stroke. PM R 6(7):635–642. https://doi. org/10.1016/j.pmrj.2013.12.017. Epub 2014 Jan 19 Balasubramanian CK, Bowden MG, Neptune RR, Kautz SA (2007) Relationship between step length asymmetry and walking performance in subjects with chronic hemiparesis. Arch Phys Med Rehabil 88(1):43–49 Beaman CB, Peterson CL, Neptune RR, Kautz SA (2010) Differences in self-selected and fastestcomfortable walking in poststroke hemiparetic persons. Gait Posture 31(3):311–316 Bayat R, Barbeau H, Lamontagne A (2005) Speed and temporal-distance adaptations during treadmill and overground walking following stroke. Neurorehabil Neural Repair 19(2):115–124 Chen G, Patten C, Kothari DH, Zajac FE (2005) Gait differences between individuals with poststroke hemiparesis and non-disabled controls at matched speeds. Gait Posture 22(1):51–56 Clark DJ (2015) Automaticity of walking: functional significance, mechanisms, measurement and rehabilitation strategies. Front Hum Neurosci 9:246 Cooper A, Alghamdi GA, Alghamdi MA et al (2012) The relationship of lower limb muscle strength and knee joint hyperextension during the stance phase of gait in hemiparetic stroke patients. Physiother Res Int 17:150–156 Den Otter AR, Geurts ACH, Mulder T, Duysens J (2007) Abnormalities in the temporal patterning of lower extremity muscle activity in hemiparetic gait. Gait Posture 25(3):342–352. https://doi. org/10.1016/j.gaitpost.2006.04.007 Garraway M (1985) Stroke rehabilitation units: concepts, evaluation, and unresolved issues. Stroke 16(2):178–181 Goldie PA, Matyas TA, Evans OM (2001) Gait after stroke: initial deficit and changes in temporal patterns for each gait phase. Arch Phys Med Rehabil 82(8):1057–1065 Honeine JL, Schieppati M, Gagey O, Do MC (2013) The functional role of the triceps surae muscle during human locomotion. PLoS One 8(1):e52943. https://doi.org/10.1371/journal.pone. 0052943. Epub 2013 Jan 16 Honeine J, Schieppati M, Gagey O, Do M (2014) By counteracting gravity, triceps surae sets both kinematics and kinetics of gait. Phys Rep 2:e00229. https://doi.org/10.1002/phy2.229 Hsiao H, Awad LN, Palmer JA, Higginson JS, Binder-Macleod SA (2016) Contribution of paretic and nonparetic limb peak propulsive forces to changes in walking speed in individuals poststroke. Neurorehabil Neural Repair 30(8):743–752. https://doi.org/10.1177/1545968 315624780. Epub 2015 Dec 31 Hsu AL, Tang PF, Jan MH (2003) Analysis of impairments influencing gait velocity and asymmetry of hemiplegic patients after mild to moderate stroke. Arch Phys Med Rehabil 84(8):1185–1193

Gait Disorders in Persons After Stroke

1215

Jonsdottir J, Recalcati M, Rabuffetti M, Casiraghi A, Boccardi S, Ferrarin M (2009) Functional resources to increase gait speed in people with stroke: strategies adopted compared to healthy controls. Gait Posture 29(3):355–359 Kitatani R, Ohata K, Sato S, Watanabe A, Hashiguchi Y, Yamakami N, Sakuma K, Yamada S (2016) Ankle muscle coactivation and its relationship with ankle joint kinematics and kinetics during gait in hemiplegic patients after stroke. Somatosens Mot Res 33(2):79–85. 2016 May 18 Lamontagne A, Fung J, McFadyen BJ, Faubert J (2007) Modulation of walking speed by changing optic flow in persons with stroke. J NeuroEng Rehabil 4:22. https://doi.org/10.1186/1743-00034-22 Lamontagne A, Malouin F, Richards CL (2000) Contribution of passive stiffness to ankle plantarflexor moment during gait after stroke. Arch Phys Med Rehabil 81(3):351–358. PMID:10724082 Lord SE, McPherson K, McNaughton HK, Rochester L, Weatherall M (2004) Community ambulation after stroke: how important and obtainable is it and what measures appear predictive? Arch Phys Med Rehabil 85:234–239 Lusardi MM, Pellecchia GL, Schulman M (2003) Functional performance in community living older adults. J Geriatr Phys Ther 26(3):14–22 Macko RF, Smith GV, Dobrovolny CL, Sorkin JD, Goldberg AP, Silver KH (2001) Treadmill training improves fitness reserve in chronic stroke patients. Arch Phys Med Rehabil 82:879–884 Martino G, Ivanenko YP, d’Avella A, Serrao M, Ranavolo A, Draicchio F, Cappellini G, Casali C, Lacquaniti F (2015) Neuromuscular adjustments of gait associated with unstable conditions. Neurobiologia 114(5):2867–2882. https://doi.org/10.1152/jn.00029.2015. Epub 2015 Sep 16 Mulroy S, Gronley J, Weiss W, Newsam C, Perry J (2003) Use of cluster analysis for gait pattern classification of patients in the early and late recovery phases following stroke. Gait Posture 18(1):114–125 Nadeau S, Gravel D, Arsenault AB, Bourbonnais D (1999) Plantarflexor weakness as a limiting factor of gait speed in stroke subjects and the compensating role of hip flexors. Clin Biomech (Bristol, Avon) 14(2):125–135 Neptune RR, McGowan CP (2011) Muscle contributions to whole-body sagittal plane angular momentum during walking. J Biomech 44(1):6–12. https://doi.org/10.1016/j.jbiomech.2010.08. 015. Epub 2010 Sep 15 Olney SJ, Griffin MP, McBride ID (1994) Temporal, kinematic, and kinetic variables related to gait speed in subjects with hemiplegia: a regression approach. Phys Ther 74(9):872–885 Olney SJ, Richards C (1996) Hemiparetic gait following stroke. Part I: characteristics. Gait Posture 4(2):136–148 Parvatanemi K, Olney SJ, Brouwer B (2007) Changes in muscle group work associated with changes in gait speed of persons with stroke. Clin Biomech 22:813–820 Patel AT, Duncan PW, Lai SM, Studenski S (2000) The relation between impairments and functional outcomes poststroke. Arch Phys Med Rehabil 81(10):1357–1363. PMID:11030501 Patterson KK, Gage WH, Brooks D, Black SE, McIlroy WE (2010) Changes in gait symmetry and velocity after stroke: a cross-sectional study from weeks to years after stroke. Neurorehabil Neural Repair 24(9):783–790 Perry J, Garrett M, Gronley JK, Mulroy SJ (1995; Jun) Classification of walking handicap in the stroke population. Stroke 26(6):982–989 Rinaldi LA, Monaco V (2013) Spatio-temporal parameters and intralimb coordination patterns describing hemiparetic locomotion at controlled speed. J NeuroEng Rehabil 10(1):53 Roerdink M1, Beek PJ (2011) Understanding inconsistent step-length asymmetries across hemiplegic stroke patients: impairments and compensatory gait. Neurorehabil Neural Repair 25(3): 253–258. https://doi.org/10.1177/1545968310380687. Epub 2010 Nov 1 Rosa MC, Marques A, Demain S, Metcalf CD (2014) Lower limb co-contraction during walking in subjects with stroke: a systematic review. J Electromyogr Kinesiol 24(1):1–10

1216

J. Jonsdottir and M. Ferrarin

Sheean G, McGuire JR (2009) Spastic hypertonia and movement disorders: pathophysiology, clinical presentation, and quantification. PM R 1(9):827–833. https://doi.org/10.1016/j. pmrj.2009.08.002 Sheffler LR, Chae J (2015) Hemiparetic Gait. Phys Med Rehabil Clin N Am 26(4):611–623 Snijders AH, van de Warrenburg BP, Giladi N, Bloem BR (2007) Neurological gait disorders in elderly people: clinical approach and classification. Lancet Neurol 6:63–74 Teasell R, Bayona NA, Bitensky J (2005) Plasticity and reorganization of the brain post stroke. Top Stroke Rehabil 12:11–26. https://doi.org/10.1310/6aum-etyw-q8xv-8xac Turnbull GI, Charteris J, Wall JC (1995) A comparison of the range of walking speeds between normal and hemiplegic subjects. Scand J Rehabil Med 27(3):175–182 Wagenaar RC, Beek WJ (1992) Hemiplegic gait: a kinematic analysis using walking speed as a basis. J Biomech 25(9):1007–1015 WHO (2003) World health report. World Health Organization, Geneva Winter DA (1990) Biomechanics and motor control of human movement. Wiley, New York

Hereditary Motor Sensory Neuropathy: Understanding Function Using Motion Analysis Sylvia Õunpuu and Kristan Pierz

Abstract

Hereditary motor sensory neuropathies, or Charcot-Marie-Tooth disease, represent a heterogeneous group of inherited neuropathies that are characterized by progressive wasting and resulting weakness of the distal muscles in the legs and arms. Lower extremities are typically initially effected and, as result, impact ambulation. At present there is no curative treatment available; therefore, treatment of gait issues is often sought to help with ambulation and activities of daily living. Computerized motion analysis techniques have improved our understanding of the various presentations of hereditary neuropathies and can assist in making optimal treatment decisions to improve gait. These presentations include three distinct ankle variations: excessive equinus (toe walking), cavo-varus (lateral border weight bearing), and flail foot (heel weight bearing) patterns. As each patient presents differently in terms of deformity specifics and severity, a detailed analysis that describes ankle/foot function during gait in terms of foot pressures, muscle activity, kinematics, and kinetics along with clinical examination information such as muscle strength and passive range of motion is very beneficial. Assessment of treatment outcomes from bracing to orthopedic surgery as well as disease progression which also varies person to person is necessary to develop evidence-based treatment indications and goals. Motion analysis can play a very important role in the assessment of inherited neuropathies on both an individual patient basis and in research with the ultimate goal of improving treatment outcomes.

S. Õunpuu (*) • K. Pierz Center for Motion Analysis, Division of Orthopaedics, Connecticut Children’s Medical Center, Farmington, CT, USA e-mail: [email protected]; [email protected]; [email protected] # Springer International Publishing AG, part of Springer Nature 2018 B. Müller, S.I. Wolf (eds.), Handbook of Human Motion, https://doi.org/10.1007/978-3-319-14418-4_62

1217

1218

S. Õunpuu and K. Pierz

Keywords

Hereditary neuropathies • Charcot-Marie-Tooth • Clinical gait analysis • Peak dorsiflexion • Joint kinematics • Joint kinetics • Ankle Abbreviations

AFO CMT EMG PLS

Ankle foot orthosis Charcot-Marie-Tooth Electromyography Posterior leaf spring orthosis

Contents Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . State of the Art . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Gait Function and Treatment Options . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Clinical Examination Findings in CMT . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Gait Findings: As Defined by Motion Analysis Parameters . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Management: Evaluation of Treatment Outcomes Using Motion Analysis . . . . . . . . . . . . . . . Future Directions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Cross-References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

1218 1219 1220 1220 1221 1228 1233 1235 1235

Introduction Hereditary motor sensory neuropathies, or Charcot-Marie-Tooth disease (CMT), represent a heterogeneous group of inherited neuropathies that are characterized by progressive weakness and resulting wasting of the distal muscles in the extremities (legs and arms). This group of neuropathies was described in detail with the tools available at the time in 1886 by neurologists Charcot, Marie, and Tooth who defined them as peroneal muscular atrophies of familial origin (Charcot 1886; Tooth 1886). These neuropathies are length dependent, meaning the longest nerves in the body are affected, and therefore, the lower extremities are typically initially affected in the course of disease progression. In terms of function, the resulting muscle weakness results in foot deformity such as flatfoot or high arch and gait issues such as ankle instability, clumsiness, and slow running. During standing and walking, ankle support is often needed and provided by orthoses, and in some cases walking becomes so difficult over time that wheelchair mobility is required. At present there is no curative treatment available, so treatment of gait issues can help provide more functional ambulation and improve/facilitate participation and activities of daily living. As with any complex movement disorder, CMT presents differently in each patient in terms of deformity, severity, and rate of disease progression. Comprehensive motion analysis techniques can provide objective documentation of (a) the pathomechanics, (b) disease progression, and (c) treatment outcomes. The goal of this chapter is to highlight what we have learned about the pathomechanics of gait in persons with CMT using comprehensive motion analysis techniques. The application of motion analysis is in its infancy for this gait pathology. The ultimate

Hereditary Motor Sensory Neuropathy: Understanding Function Using Motion. . .

1219

goal is to improve treatment outcomes and to understand prognosis for future function near the time of diagnosis through a better understanding of gait pathomechanics. Although progress has been made, there is much more to be done.

State of the Art The application of motion analysis in this clinical setting is relatively new in comparison to other gait disorders such as cerebral palsy and spina bifida, so our understanding of gait function in persons with CMT limited. Because CMT is a rare disease, the clinical application of motion analysis in this patient population is not as common, and research applications for the most part have focused on small diverse groups of patients that span a wide age range. However, over time this body of literature has grown, and clinicians have gained more understanding of the pathomechanics of gait. The presentation of CMT in classical textbooks generally describes a common set of impairments (weakness and contracture) that result in a specific gait pattern including excessive equinus in swing and increased hip flexion, pelvic hiking, and circumduction to clear the foot (Holmes and Hansen 1993). Clinical experience shows that many patients with CMT do not exhibit excessive equinus in swing and the associated proximal compensations. Also cavo-varus foot deformity is noted as common (Smith 2002); however, many patients with CMT have flail feet with pes valgus. It is clear from previous research that incorporates comprehensive motion analysis assessments that there are substantial variations in presentation and severity of impairments (Burns et al. 2005, 2006; Garcia et al. 1998; Vinci and Perelli 2002; Vinci et al. 2006) that result in a variety of gait pathologies (Don et al. 2007; Kuruvilla et al. 2000; Newman et al. 2007; Ramdharry et al. 2009). As a result, this latter group of authors divided CMT patients into groups based upon gait pathology. In many cases gait features often overlapped, that is, multiple features were present in a single patient. These studies mostly focused on adults with CMT, and as a result there is very little understanding of CMT gait pathology evolution during childhood when the impairments typically present. However, there are a few exceptions. The first signs of CMT were evaluated by Burns et al. 2009 and include issues with ability to toe walk and run and clumsy gait (Burns et al. 2009). Ferrarin (Ferrarin et al. 2012) divided gait pathology into three categories in a group of children and adolescents using motion analysis outcomes: (a) pseudo normal, (b) drop foot only, and (c) foot drop and push-off deficiency. In a group of similarly aged patients, Ounpuu (Ounpuu et al. 2013)focused on peak dorsiflexion in terminal stance as a way to differentiate gait in persons with CMT as this particular gait parameter is relevant to treatment strategy. In this study, patients were placed into three groups: (a) less than typical peak dorsiflexion, (b) typical peak dorsiflexion, and (c) increased peak dorsiflexion. Patients with CMT, however, rarely present with one issue at the ankle, so in groups (a) and (c), there was also simultaneous increased equinus, and in groups (b) and (c), there was delayed peak dorsiflexion in stance. There is, however, still work needed to further refine gait classifications in these patients, understand disease progression and implications of asymmetry and

1220

S. Õunpuu and K. Pierz

ultimately link gait issues to phenotype. Initial reports on CMT often depicted large asymmetries, however, it has been recently reported that asymmetry is not that common in CMT in terms of a wide variety of outcome measures including gait parameters (Burns et al. 2012). In that there is such variation in presentation, treatments need to be patient specific and based on a comprehensive patient examination including motion analysis especially at the level of the ankle and foot (Jani-Acsadi et al. 2015). There is, however, very little objective documentation of treatment outcomes in terms of gait function using comprehensive motion analysis techniques. As more centers incorporate comprehensive motion measurement techniques as part of the standard of care, this will become possible.

Gait Function and Treatment Options This section will include a description of the following for persons with CMT: (a) the current understanding of gait pathology in terms of gait analysis parameters and associated clinical examination findings and (b) the current standard of treatment for gait-related issues including therapies, orthoses, and orthopedic surgery. Example patient data will be incorporated to illustrate common findings.

Clinical Examination Findings in CMT Clinical examination findings which are collected using standard protocols for the assessment of strength, passive range of motion, and bony deformity are integral to the interpretation of motion analysis data so need to be a part of a full assessment of gait pathology. Clinical findings help to determine the primary impairments that lead to gait pathology. Persons with CMT typically have impairments that include weakness, contracture, and bony deformity which are further complicated by the disease progression (increased problems over time) as well as issues with reduced sensation and reflex response which also need to be assessed. For the context of this chapter with the focus on gait analysis, we will briefly discuss the primary impairments that impact gait in persons with CMT. The first manifestations of disease for most persons with CMT are distal limb weakness and muscle atrophy. Sensory loss and absent reflexes are also present (Thomas 1999). Comprehensive research on strength testing methods and comparison between controls and persons with CMT1A in terms of strength and passive range of motion findings have been completed by Burns and colleagues (Burns et al. 2005, 2009; Rose et al. 2010). Their detailed studies have led to some primary findings in children with CMT and cavus foot deformity which include reduced strength in all of the muscles of the feet. Initial deficits in young children in comparison with their age-matched peers start with weakness of the plantar flexors, dorsiflexors, and evertors of the ankle (Rose et al. 2010). Young children with CMT also have greater inversion-to-eversion and plantar flexion-to-dorsiflexion strength

Hereditary Motor Sensory Neuropathy: Understanding Function Using Motion. . .

1221

ratios and a high correlate between dorsiflexion range of motion and foot and ankle strength. Those with CMT also have significantly less passive ankle dorsiflexion range of motion. Strength assessments in adult persons with CMT have also documented weakness in the ankle dorsiflexors, plantar flexors, and evertors (Don et al. 2007; Newman et al. 2007). Strength and range of motion deficits result in gait pathology and should be linked to the interpretation of gait data to determine causes of gait pathology. This has been done in the majority of research involving motion analysis and CMT. Longitudinal data that would help to explain the pathogenesis and progression of this disease in terms of gait decline over time would be extremely helpful and is not yet available.

Gait Findings: As Defined by Motion Analysis Parameters Motion analysis has led to a more comprehensive understanding of the pathomechanics of gait in persons with CMT. The following section will describe CMT in the context of the following components of a comprehensive motion analysis: gait kinematics and kinetics, electromyography, and pedobarography. Each component provides a unique source of information not available using standard assessments and when combined provide a comprehensive view of the pathomechanics of gait in this population. Motion analysis also provides a unique opportunity to assess gait function objectively over time and to evaluate treatment outcomes such as orthoses and orthopedic surgery.

Kinematic and Kinetic Findings in CMT As mentioned above, the majority of the previous research focused on CMT and gait includes joint and segment kinematic and joint kinetic outcomes. These data provide an opportunity to better understand the pathomechanics of gait in CMT and help to make links between the primary pathology and compensations noted at proximal joints. It is clear from these works that CMT with its heterogeneous genetic makeup also has a heterogeneous impairment and associated functional presentation. Ankle function during gait in CMT is impacted depending on the impairments of weakness and contracture in the associated musculature. Therefore, presentation can be differentiated by the degree of peak dorsiflexion in terminal stance (which is impacted by both ankle plantar flexor range of motion and strength) and divided into three groups: (a) increased and delayed peak dorsiflexion, (b) delayed peak dorsiflexion, and (c) increased plantar flexion (Ounpuu et al. 2013). Similarly, ankle dorsiflexion in mid-swing and at initial contact is often impacted by ankle dorsiflexor strength but is also impacted by plantar flexor tightness that does not “allow” the anterior tibialis to effectively dorsiflex the ankle in a non-weight-bearing position (swing phase). So CMT commonly presents with excessive equinus in mid-swing and initial contact. Cavus foot deformity is also common but not well documented using current motion analysis methods. As it impacts the available length of the ankle plantar flexors, it is an important consideration for the overall function of the foot and ankle.

1222

S. Õunpuu and K. Pierz

Generally, the foot and ankle in CMT fall into three groups of impairments. • The flail foot with significant weakness in all ankle muscles that leads to increased and delayed peak dorsiflexion in stance and equinus in swing with medial and lateral instability resulting in pes planovalgus deformity over time (Fig. 1a–c). Although claw toes are common in earlier phases of the disease, in this group, the toe flexors are typically weak with no toe contact during stance. • The cavus foot deformity with lateral weight bearing during stance and delayed peak dorsiflexion in terminal stance consistent with ankle plantar flexor weakness and typically normal or only minimal increased equinus in swing (Fig. 2a–c). The cavus deformity limits available plantar flexor length which when at its end range may mask ankle plantar flexor weakness. • Equinus ankle deformity with increased equinus in stance and swing due to lack of plantar flexor range of motion (assessment of shank vs. plantar aspect of the foot) (Fig. 3a–c). This lack of passive range of motion may be a result of a 40

Plantar-Dorsiflexion

Dor deg Pla -40 2.0

Ankle Moment

Ext Nm/kg Flx -1.0 3.0

Ankle Power

W/kg

-2.0 25%

a

b

50% Gait Cycle

75%

c

Fig. 1 Example of the flail foot during (a) relaxed standing with significant weakness of the ankle musculature as evident in the (b) ankle sagittal plane kinematic, moment and power (three gait cycles) during stance and swing phases and in (c) foot pressures for the right foot. Increased and delayed peak dorsiflexion, reduced peak ankle plantar flexor moment and power generation in terminal stance and reduced pressures under the distal foot with a limited length of center of pressure path are evidence of ankle plantar flexor weakness. Increased plantar flexion in swing is evidence of ankle dorsiflexor weakness

Hereditary Motor Sensory Neuropathy: Understanding Function Using Motion. . .

40

1223

Plantar-Dorsiflexion

Dor deg Pla -40 2.0

Ankle Moment

Ext Nm/kg Flx -1.0 3.0

Ankle Moment

Gen W/kg Abs -2.0

a

b

25%

50%

Gait Cycle

75%

c

Fig. 2 Example of the cavo-varus foot during (a) relaxed standing with some evidence of plantar flexor weakness in the (b) ankle sagittal plane kinematic, moment and power (three gait cycles) during stance phase and evidence of lateral weight bearing in (c) foot pressures for the right foot. Weakness of the ankle plantar flexors is most evident in the delayed and increased peak ankle dorsiflexion in stance, however, this is not yet manifested at the joint kinetic level

combination of cavus and/or limited plantar flexor range of motion. The cavus deformity limits available plantar flexor length and may mask the presence of ankle plantar flexor weakness. The compensatory proximal gait findings depend on the ankle impairment. For those patients with increased equinus in swing and dorsiflexion in stance, compensatory steppage gait is common with increased hip flexion and in some cases circumduction to aid in clearance (Fig. 4). For those with increased ankle dorsiflexion in stance, increased knee and in some cases hip flexion occur. Because the kinematic and kinetic presentations vary from patient to patient, the treatment options need to be specific to the patient and are discussed in the Management section below.

Electromyographic Findings in CMT Muscle activity for those muscles that are used to support the ankle during gait in persons with CMT is often abnormal and is manifested in multiple ways including: (a) reduced activity, (b) no activity, and (c) atypical recruitment patterns. Lack of muscle activity during the appropriate phases of the gait cycle can result in ankle

1224

S. Õunpuu and K. Pierz

40

Plantar-Dorsiflexion

Dor deg Pla -40 2.0

Ankle Moment

Ext Nm/kg Flx -1.0 3.0

Ankle Power

Gen W/kg Abs -2.0

a

b

25%

50% 75% Gait Cycle

c

Fig. 3 Example of the equinus ankle during (a) relaxed standing with inability to dorsiflex above neutral and increased equinus documented in the (b) ankle sagittal plane kinematic, moment and power (three gait cycles) during stance and swing phases and in (c) foot pressures with the right forefoot and toes only in contact with the ground instance. Increased plantar flexion in stance and swing is evidence of plantar flexor tightness

lateral instability if the peroneals are involved, increased equinus in swing if the anterior tibial group is involved, and increased peak dorsiflexion in stance if the plantar flexors are involved. Abnormal muscle contractile patterns such as single motor unit recruitment and fasciculations (repetitive single motor unit contractions at rest) are also characteristic of demyelinated muscle that can be noted on the surface EMG signal. Dynamic electromyography (EMG) techniques can provide insight into which muscles are active during which phases of gait and assist in understanding potential for muscles to provide dynamic support when transferred. As in the varied kinematic presentation, one would expect that the muscle function for this patient group is also different from person to person as muscle activity is the driver of motion and stability. So EMG patterns for key ankle muscle groups can show minimal issues or provide evidence to explain why the ankle joint is unstable (Fig. 5a–c).

Pedobarography (Foot Pressures) Findings in CMT The analysis of foot pressures during gait provides an opportunity to understand where the peak pressures are under the foot and the path of the center of pressure during the stance phase of gait. High mean and peak pressures and pressure-time

Hereditary Motor Sensory Neuropathy: Understanding Function Using Motion. . .

a

b

Pelvic Tilt

40

20

1225

Pelvic Obliquity

Up

Ant

deg Pos

Dn

-20

-20

60

Hip Flexion-Extension

20

Hip Ab-Adduction

Add

Flx

deg Ext

Abd

-20

-20

80

Knee Flexion-Extension

25%

50%

75%

Gait Cycle

Flx

Ext -20 40

Plantar-Dorsiflexion

Dor deg Pla -40

25%

50%

75%

Gait Cycle

Fig. 4 Compensations for excessive equinus in swing typically include (a) increased hip flexion in the sagittal plane (left column second plot) and (b) circumduction and pelvic hiking in the coronal plane (right column)

1226

a) 5

S. Õunpuu and K. Pierz Right Side EMG Right Gastrocnemius

5

% Gait Cycle

Right Anterior Tibialis

5

% Gait Cycle

Right Peroneus Longus

5

% Gait Cycle

Right Peroneus Brevis

b) 5

% Gait Cycle

5

5

% Gait Cycle

Right Anterior Tibialis

5

Right Peroneus Longus*

5

-5

5

-5 5

Right Peroneus Brevis*

-5 5

-5

5

Right Gastrocnemius*

-5 5

Right Peroneus Longus

-5 5

% Gait Cycle

Right Anterior Tibialis*

Right Peroneus Brevis

-5 5

-5 5

Right Peroneus Longus*

-5 5

Right Peroneus Brevis*

-5 5 V -5

% Gait Cycle

c)

% Gait Cycle

5

Left Gastrocnemius*

% Gait Cycle

Left Anterior Tibialis

-5 5

% Gait Cycle

Left Anterior Tibialis*

% Gait Cycle

Right Peroneus Brevis**

% Gait Cycle

Left Peroneus Longus

-5 5

-5 5

% Gait Cycle

Left Peroneus Longus*

-5 5

V

V

V

-5

-5

% Gait Cycle

Left Peroneus Brevis

5

% Gait Cycle

Left Peroneus Brevis*

% Gait Cycle

-5

% Gait Cycle

Left Anterior Tibialis**

5

% Gait Cycle

Left Peroneus Longus**

% Gait Cycle

Left Peroneus Brevis**

V

V

V

Left Gastrocnemius**

V

V % Gait Cycle

5

-5

-5

% Gait Cycle

Right Peroneus Longus**

V

V

V

5

% Gait Cycle

Right Anterior Tibialis**

Left Side EMG Left Gastrocnemius

V

5

Right Gastrocnemius**

V % Gait Cycle

V

-5

% Gait Cycle

V % Gait Cycle

V % Gait Cycle

5

-5

5

% Gait Cycle

Right Peroneus Brevis**

V

V

-5

% Gait Cycle

Right Peroneus Longus**

V % Gait Cycle

-5

5

% Gait Cycle

Right Anterior Tibialis**

V % Gait Cycle

V % Gait Cycle

V -5

-5

-5

V % Gait Cycle

V

V -5

5

Right Gastrocnemius**

Right Side EMG Right Gastrocnemius

V -5

-5

V

V -5

5

% Gait Cycle

Right Anterior Tibialis*

V

V -5

-5

5 V

V

V -5

Right Gastrocnemius*

V

V -5

5

% Gait Cycle

-5

% Gait Cycle

Fig. 5 Example EMG data for three different patients with CMT show heterogeneity of this disease in terms of the primary muscles at the ankle with (a) minimal atypical findings except for the gastrocnemius which shows delayed onset and minimal activity consistent with the common finding of plantar flexor weakness and increased dorsiflexion in terminal stance, (b) minimal activity other than fasciculations of peroneus longus and (c) no EMG findings consistent with the flail foot

Hereditary Motor Sensory Neuropathy: Understanding Function Using Motion. . .

1227

Fig. 6 Example foot pressure plots highlight common patterns found in persons with CMT including (a) cavus deformity with no toe contact, (b) increased lateral weight bearing pressure, (c) inability to weight bear over the distal foot in any capacity with all pressures focused under the heel in comparison to the (d) typically developing reference. In very young patients who cannot cooperate with a clinical exam to evaluate strength, the foot pressure plot is an excellent tool to understand plantar flexor strength

integrals have been measured with pedobarography in persons with CMT; however, their relationship to foot pain is not clear (Burns et al. 2005; Crosbie et al. 2008). As with other measurements, there is a larger variety of presentations in terms of foot pressures in persons with CMT (Fig. 6). This assessment is of particular interest in children where skin changes may not yet provide adequate information to understand the impact of cavus and adductus deformities. When used in conjunction with radiographs of the foot, the impact of anatomical abnormalities can be better understood. Foot pressures also provide an assessment of ankle plantar flexor and toe flexor strength. When toe flexor strength is compromised, the ability of the last portion of push-off is compromised, and no toe contact is made with the floor (Fig. 6a, b). When ankle plantar flexors are compromised, there is reduced ability to weight bear over the distal foot. Making this assessment by observing the foot/ ankle is limited as while the flail foot may be in contact with the ground the center of pressure may not be able to move over the distal aspect of the foot (Fig. 6c). The foot pressure plot in some cases may be the only way to obtain adequate information regarding plantar flexor strength when a patient is too young or unable to understand directions in a strength assessment. Pre- versus postsurgical intervention to the foot

1228

S. Õunpuu and K. Pierz

can also be assessed and allow for a critical examination of the impact of bony and/or soft tissue surgery outcomes during gait.

Management: Evaluation of Treatment Outcomes Using Motion Analysis It is clear from the previous research and clinical experience that persons with CMT have a wide variety of presentations and therefore require different treatment strategies to provide optimum outcomes. Motion analysis is an excellent tool to assist in better understanding the pathomechanics with which to make more informed treatment and provide an opportunity to objectively evaluate treatment outcomes. If motion analysis is incorporated as the standard of care, it will be possible someday to provide improved care through evidenced-based medical practice. The following is a discussion of how motion analysis can be used to better define treatment and understand treatment outcomes.

Physical Therapy Physical therapy is a common component of treatment for patients with CMT with a focus on strengthening, maintaining range of motion, and balance training with the goal of maintaining mobility. It has been shown that strength training can benefit measures of muscle strength and other outcome measures such as walking velocity (Burns et al. 2009) as well as activities of daily living in adults with CMT (Chetlin et al. 2004); however, there is limited current knowledge on how strength training impacts gait function in terms of joint kinematics and kinetics. Understanding therapy outcomes at the joint level will help to explain why some therapies may be more successful than others. For example, understanding if strengthening of the plantar flexors improved ankle sagittal plane kinematics and kinetics would help to understand why there may be benefits in walking velocity. Motion analysis could also help with targeting specific therapies to provide the most functional benefits. Orthoses Orthoses are very effective treatment modalities for support of the ankle when muscle weakness is a primary clinical finding. In persons with CMT, ankle instability is a very common finding resulting from weakness of the ankle plantar flexors and the medial and lateral stabilizers (Mandarakas et al. 2013). Also, clearance problems may be present as a result of weakness of the anterior tibialis and long toe extensors. Motion analysis techniques can provide insight into the impact of orthoses on ankle function that includes not only kinematic function of the ankle but also kinetic function that can explain in some cases why patients are non-compliant with their orthoses. Walking velocity has been shown to improve with the application of orthoses (Phillips et al. 2011); however, the assessment of this valuable outcome measure on its own does not provide information about the impact of the orthosis at the joint level which is required to understand why an orthosis improves walking velocity and ultimately indications for orthosis specific design. Ramdharry

Hereditary Motor Sensory Neuropathy: Understanding Function Using Motion. . .

1229

(Ramdharry et al. 2012) incorporated comprehensive motion analysis and documented reduced (improved) excessive dorsiflexion in terminal stance and plantar flexion in swing with the application of an appropriately molded ankle foot orthosis (AFO). However, it is clear that in patients with CMT, one AFO design does not suit all due to the variation in presentation and degree of severity of pathology and the variety of orthosis design and purpose. Although the solid AFO design may provide excellent support for excessive ankle dorsiflexion in terminal stance due to weakness of the plantar flexors, in some cases it may overly restrict ankle movement and thus reduce the ankle power generation (Fig. 7a, b). For those persons with CMT who are more functional, this restriction in motion will have a negative impact on push-off especially during running. The posterior leaf spring orthosis (PLS) which has less support through trim lines posterior to the malleoli may result in less than adequate support in some patients but function as a solid AFO in others depending on brace stiffness, patient body weight, and plantar flexor strength which may be decreasing over time due to the disease progression.

40

Plantar-Dorsiflexion

40

Dor

Dor

deg

deg

Pla

Pla -40

-40

Ankle Moment

2.0

Ext

Nm/kg

Nm/kg

Flx

Flx -1.0

-1.0

Ankle Power

3.0

Ankle Power

3.0

W/kg

W/kg

-2.0

Ankle Moment

2.0

Ext

a

Plantar-Dorsiflexion

25%

50% 75% Gait Cycle

2.0

b

25%

50% Gait Cycle

75%

c

Fig. 7 Comparison of ankle sagittal plane kinematic, moment and power for (a) barefoot and (b) solid AFO walking for the right side in a youth with CMT. The solid AFO (c) can provide improved stability at the ankle in the sagittal plane by restricting excessive peak dorsiflexion in terminal stance, however, this is done at the expense peak ankle power generation which shows a decrease in the solid AFO. It is likely that this explains why some persons with CMT do not like wearing their solid AFOs especially when there are greater requirements for ankle push off such as in running

1230

S. Õunpuu and K. Pierz

b

a Peak Dorsiflexion (deg)

25

Barefoot Solid AFO

20

30

Barefoot PLS AFO

25 20

15

15 10

10

5 0

5 1

2

3

4

5

6

7

8

Number of Sides

9 10 11

0

1

2

3

4

5

6

7

8

9

10

Number of Sides

Fig. 8 Comparison peak ankle dorsiflexion during stance for barefoot versus AFO for the (a) solid AFO and (b) PLS AFO. The red line indicates typically developing peak dorsiflexion in stance. The large variation in values for peak dorsiflexion in terminal stance in both brace designs confirms the heterogeneous nature of CMT in terms of gait outcomes. In some cases, the brace design provided improvement and in others made no change and in other resulted in a worse outcome

We have recently reviewed AFO outcomes in children and adolescents with CMT and have shown that different AFOs (solid AFO vs. PLS), although intended to have differences in terms of function based upon their design, do not always operate as proposed. In many cases, the PLS design functions as a solid AFO and restricts most ankle motion, and in others the PLS does not provide adequate support and allows ongoing excessive ankle dorsiflexion. As for the solid AFO design, in some cases, it functions like a PLS and does not provide adequate support (Fig. 8a, b). It is also clear that hinged AFO designs, while supporting the ankle in swing and reducing excessive plantar flexion, do not provide adequate support in stance (Fig. 9a, b). Understanding orthosis function in persons with CMT using motion analysis has allowed us to clarify why some orthosis designs are more effective than others and to highlight the importance of understanding patient impairment and gait function during barefoot walking that needs to be supplemented with the AFO. Additional research is needed in this area to better match orthosis design with patient impairment. As well, there are technical considerations when using motion analysis techniques in the assessment of orthoses including marker placement and documentation of consistent ankle sagittal plane angles in both barefoot and orthosis conditions (Ounpuu 1996).

Orthopedic Surgery Orthopedic surgery may be required when patients with CMT have foot deformity and pain that limits walking and performing activities of daily living. Surgery to address the wide variety of issues varies from individual muscle procedures (lengthenings or transfers) to complex combinations of soft tissue and bony interventions. The majority of research evaluating surgical outcomes is based upon clinical examination findings alone and, in some cases, only postoperative assessments. Motion analysis provides

Hereditary Motor Sensory Neuropathy: Understanding Function Using Motion. . .

40

Plantar-Dorsiflexion

40

Dor

Dor

deg

deg

Pla

Pla

Ankle Moment

Ext

Nm/kg

Nm/kg

Flx

Flx -1.0

-1.0 Ankle Power

3.0

Gen

W/kg

W/kg

Abs

Abs 25%

50% Gait Cycle

Ankle Power

3.0

Gen

-2.0

Ankle Moment

2.0

Ext

a

Plantar-Dorsiflexion

-40

-40 2.0

1231

75%

b

-2.0

25%

50% Gait Cycle

75%

c

Fig. 9 Comparison of ankle sagittal plane kinematic, moment and power for (a) barefoot and (b) hinged AFO walking for the right side in a youth with CMT. Although the hinged AFO (c) can provide medial/lateral stability there is no restriction on peak ankle dorsiflexion in stance and therefore there is a continued reduction in peak ankle plantar flexor moment in stance and improvement in stability is not improved in the sagittal plane

an excellent opportunity to assist in surgical decision-making and to evaluate surgical outcomes objectively. This is possible in those clinical settings where the standard of care includes comprehensive motion analysis techniques both pre- and post-surgery. These assessments should include an evaluation of clinical impairments (muscle weakness, contracture, and bony deformity) which is integrated with gait findings documented with dynamic EMG, joint and segment kinematics, joint kinetics, and foot pressure data. This approach allows for the most informed treatment decisions and objective evaluation of outcomes when the patient has adequately recovered. This approach will allow new knowledge on a patient-by-patient basis and ultimately through research when this protocol is followed over time. Pairing treatment approaches with clear and specific indications at the level of impairment and function is needed. When interpreting postoperative data, it is important to also consider changes that might be due to the natural history of this progressive disorder. Motion analysis has provided some insight into treatment indications from both routine clinical use and research applications, but at this time, there is only one example of comprehensive motion analysis outcomes used to assess treatment in this

1232

S. Õunpuu and K. Pierz

patient population. It has been shown by Dreher (Dreher et al. 2014) that the transfer of the posterior tibialis tendon can reduce excessive equinus in swing which can lead to clearance issues and inappropriate prepositioning at initial contact. However, they also found reduced peak ankle plantar flexion at push-off. Joint kinetic data would provide additional insight into this finding related to plantar flexor function which is also confounded by the possibility of increased weakness over time. Orthopedic treatment outcomes have also been assessed using pedobarography with positive outcomes in terms of foot pressures; however, these measures do not correlate well with other aspects of gait (Metaxiotis et al. 2000). A study by Ward (Ward et al. 2008) evaluated 25 patients with CMT who underwent treatment for cavus foot including dorsiflexion osteotomy of the first metatarsal, transfer of the peroneus longus to the peroneus brevis, plantar fascia release, transfer of the extensor hallucis longus to the neck of the first metatarsal, and in selected cases transfer of the tibialis anterior tendon to the lateral cuneiform. Temporal and stride parameters were analyzed and revealed that those patients that had undergone the anterior tibialis tendon transfer spent less time in double-support stance phase. Additional outcomes of ankle kinematics and kinetics in the above study would have provided relevant information as to the causes of changes in temporal and stride parameters to further clarify treatment indications. This would have allowed confirmation of the surgical goals of addressing current deformity and ultimately long-term recurrence with objective measures of ankle function. Although these studies provide interesting findings, comprehensive motion analysis that includes both kinematic and kinetic outcomes would be very useful. Individual patient cases where motion analysis has been incorporated as the standard of care can also provide important knowledge especially if this methodology is followed over the long term. This is particularly important in rare diseases when it takes time to gather a large data base of patients to study. For example, motion analysis data for an individual patient has provided some insight into the possible impact of the plantar fascia release for the correction of cavus deformity. In a patient who is in excessive equinus in stance (toe walker) and swing due to limited dorsiflexion range of motion (plantar aspect of the foot in relation to the tibia), a plantar fascia release may provide sufficient reorientation of the foot anatomically to allow for increased passive dorsiflexion range to eliminate the excessive equinus in stance and swing (Fig. 10a, b). In a patient who has normal peak ankle dorsiflexion in terminal stance but a significant cavus deformity, a plantar fascia release to correct for the cavus and varus position may provide reorientation of the foot anatomically to allow for excessive ankle passive dorsiflexion range of motion and excessive dorsiflexion in terminal stance. This may lead secondarily to increased knee flexion in stance (Fig. 11). Unmasked ankle plantar flexor weakness may in part play a role in the downside of this procedure which may still be relevant to help with foot pain. Appreciating the complex relationship between maximum dorsiflexion range of motion, plantar flexor strength, patient body weight, and the extent of cavus deformity is all more possible with motion analysis. Learning from these individual cases is an important step toward the goal of improving treatment outcomes.

Hereditary Motor Sensory Neuropathy: Understanding Function Using Motion. . .

a 40

Plantar-Dorsiflexion

b 40

Dor

Dor

deg

deg

Pla

Pla

-40

-40

Ankle Moment

2.0

2.0

Ext

Ext

Nm/kg

Nm/kg

Flx

Flx

-1.0

1233

Plantar-Dorsiflexion

Ankle Moment

-1.0

Ankle Power

3.0

Ankle Power

3.0

Gen W/kg

W/kg

Abs -2.0 25%

50% Gait Cycle

75%

-2.0

25%

50% Gait Cycle

75%

Fig. 10 Comparison of the sagittal plane ankle kinematics, moments and powers for (a) pre and (b) several years post orthopaedic lengthening of the plantar fascia in an child with CMT. Three gait cycles on the right side are plotted for each condition. Improvements in ankle sagittal plane motion in stance and swing and ankle moment modulation due to a heel initial contact without reduction in ankle power generation are noted. Objective documentation of the outcome of this procedure using motion analysis can help to clarify surgical indications

Future Directions The application of comprehensive motion analysis techniques for clinical evaluation as part of the standard of care and in research settings for persons with CMT is relatively new compared to other complex gait pathologies such as cerebral palsy. As a result there is still much to learn and new developments needed to optimize motion measurement for this pathology. Future effort should focus on the following:

1234 Fig. 11 Comparison of the knee and ankle sagittal plane kinematics for pre (dashed) versus post (solid) orthopaedic surgical outcome for plantar fascia release. Increased peak ankle dorsiflexion and associated knee flexion may be due to multiple causes: (a) ongoing disease process of increasing weakness of the plantar flexors over time and/or (b) unmasking of existing ankle plantar flexor weakness through increased dorsiflexion range of motion

S. Õunpuu and K. Pierz

80

Knee Flexion-Extension

Flx

Ext

-20 40

Plantar-Dorsiflexion

Dor

deg

Pla

-40

25%

50% Gait Cycle

75%

(a) Application of a more comprehensive foot model is needed to better understand the complex relationship between cavus deformity, plantar flexor strength, and passive ankle dorsiflexion range of motion. The foot model needs to include measurement of the extent of cavus and may need to incorporate radiographic information to allow for the most accurate assessment. Understanding how the extent of cavus deformity impacts available plantar flexor length and masks weakness is critical for treatment decision-making at the ankle/foot in these patients. (b) The presentation of CMT is heterogeneous with varying clinical findings, severity, as well as disease progression. Establishing if there is any link between impairments and associated gait function with phenotype will provide a better ability to determine prognosis for future function at the time of diagnosis and lead to treatment guidelines based upon phenotype. (c) CMT is a progressive disease that typically results in increasing weakness and associated gait issues over time. Disease progression, however, is patient dependent and not a lot is known about expectations for decline for a given individual with CMT. Long-term natural progression studies are needed which hopefully can be linked eventually to phenotype. Achieving this goal is difficult as disease progression in many is not rapid so long-term studies are needed.

Hereditary Motor Sensory Neuropathy: Understanding Function Using Motion. . .

1235

(d) Treatment of gait issues for this patient population focuses on improving or maintaining gait function and reducing foot/ankle pain. There is very limited objective documentation of orthopedic and other treatment outcomes such as bracing in terms of comprehensive motion analysis. Systematic reviews of surgical outcomes are needed to assist in identifying specific treatment indications and expectations. This research is further complicated by disease progression which needs to be taken into account. (e) Finally, establishing a diagnosis of CMT is difficult, and in the initial phases, it is often confused with other diagnoses such as idiopathic toe walking to cavo-varus deformity. Initial steps include a detailed clinical assessment and should also include a family history. If there is a suspected CMT diagnosis, often genetic testing and nerve conduction tests are recommended to confirm a diagnosis and phenotype. These later tests are often not completed due to expense, and nerve conduction tests can be painful. Therefore, establishing functional biomarkers related to movement would be helpful in the initial stages of this disease to help correctly diagnosis and therefore treat appropriately common initial findings of cavo-varus foot deformity in some and toe walking in others. These biomarkers may include any or a combination of the following: ankle kinematic or kinetic variables, EMG signal analysis, and muscle impedance assessments. Additional investigation is needed to determine if these or other biomarkers can be established.

Cross-References ▶ 3D Dynamic Pose Estimation from Marker-Based Optical Data ▶ Assessing Pediatric Foot Deformities by Pedobarography ▶ Functional Effects of Foot Orthoses ▶ Interpreting Joint Moments and Powers in Gait ▶ Kinematic Foot Models for Instrumented Gait Analysis

References Burns J, Crosbie J, Hunt A, Ouvrier R (2005) The effect of pes cavus on foot pain and plantar pressure. Clin Biomech (Bristol, Avon) 20:877–882 Burns J, Crosbie J, Ouvrier R, Hunt A (2006) Effective orthotic therapy for the painful cavus foot: a randomized controlled trial. J Am Pediatr Med Assoc 96:205–211 Burns J, Raymond J, Ouvrier R (2009) Feasibility of foot and ankle strength training in childhood Charcot-Marie-Tooth disease. Neuromuscul Disord 19:818–821 Burns J, Ouvrier R, Estilow T, Shy R, Laura M, Eichinger K, Muntoni F, Reilly MM, Pareyson D, Acsadi G, Shy ME, Finkel RS (2012) Symmetry of foot alignment and ankle flexibility in paediatric Charcot-Marie-Tooth disease. Clin Biomech (Bristol, Avon) 27:744–747 Charcot JM (1886) Sue une forme particulaire d'atrophie musculaire progressive souvent familial debutant par let pieds et les jambes et atteingnant plus tard les mains. Rev Med Paris 6:97–138 Chetlin RD, Gutmann L, Tarnopolsky M, Ullrich IH, Yeater RA (2004) Resistance training effectiveness in patients with Charcot-Marie-Tooth disease: recommendations for exercise prescription. Arch Phys Med Rehabil 85:1217–1223

1236

S. Õunpuu and K. Pierz

Crosbie J, Burns J, Ouvrier RA (2008) Pressure characteristics in painful pes cavus feet resulting from Charcot-Marie-Tooth disease. Gait Posture 28:545–551 Don R, Serrao M, Vinci P, Ranavolo A, Cacchio A, Ioppolo F, Paoloni M, Procaccianti R, Frascarelli F, De Santis F, Pierelli F, Frascarelli M, Santilli V (2007) Foot drop and plantar flexion failure determine different gait strategies in Charcot-Marie-Tooth patients. Clin Biomech (Bristol, Avon) 22:905–916 Dreher T, Wolf SI, Heitzmann D, Fremd C, Klotz MC, Wenz W (2014) Tibialis posterior tendon transfer corrects the foot drop component of cavovarus foot deformity in Charcot-Marie-Tooth disease. J Bone Joint Surg Am 96:456–462 Ferrarin M, Bovi G, Rabuffetti M, Mazzoleni P, Montesano A, Pagliano E, Marchi A, Magro A, Marchesi C, Pareyson D, Moroni I (2012) Gait pattern classification in children with CharcotMarie-Tooth disease type 1A. Gait Posture 35:131–137 Garcia A, Combarros O, Calleja J, Berciano J (1998) Charcot-Marie-Tooth disease type 1A with 17p duplication in infancy and early childhood: a longitudinal clinical and electrophysiologic study. Neurology 50:1061–1067 Holmes JR, Hansen ST Jr (1993) Foot and ankle manifestations of Charcot-Marie-Tooth disease. Foot Ankle 14:476–486 Jani-Acsadi A, Ounpuu S, Pierz K, Acsadi G (2015) Pediatric Charcot-Marie-Tooth disease. Pediatr Clin N Am 62:767–786 Kuruvilla A, Costa JL, Wright RB, Yoder DM, Andriacchi TP (2000) Characterization of gait parameters in patients with Charcot-Marie-Tooth disease. Neurol India 48:49–55 Mandarakas M, Hiller CE, Rose KJ, Burns J (2013) Measuring ankle instability in pediatric Charcot-Marie-Tooth Disease. J Child Neurol 28:1456–1462 Metaxiotis D, Accles W, Pappas A, Doederlein L (2000) Dynamic pedobarography (DPB) in operative management of cavovarus foot deformity. Foot Ankle Int 21:935–947 Newman CJ, Walsh M, O’sullivan R, Jenkinson A, Bennett D, Lynch B, O’brien T (2007) The characteristics of gait in Charcot-Marie-Tooth disease types I and II. Gait Posture 26:120–127 Ounpuu S (1996) An evaluation of the posterior leaf spring orthosis using joint kinematics and kinetics. J Gerontol A Biol Sci Med Sci 16:378–384 Ounpuu S, Garibay E, Solomito M, Bell K, Pierz K, Thomson J, Acsadi G, Deluca P (2013) A comprehensive evaluation of the variation in ankle function during gait in children and youth with Charcot-Marie-Tooth disease. Gait Posture 38:900–906 Phillips MF, Robertson Z, Killen B, White B (2011) A pilot stud of a crossover trial with randomized use of ankle-foot orthoses for people with Charcot-Marie-Tooth disease. Clin Rehabil 26:534–544 Ramdharry GM, Day BL, Reilly MM, Marsden JF (2009) Hip flexor fatigue limits walking in Charcot-Marie-Tooth disease. Muscle Nerve 40:103–111 Ramdharry GM, Day BL, Reilly MM, Marsden JF (2012) Foot drop splints improve proximal as well as distal leg control during gait in Charcot-Marie-Tooth disease. Muscle Nerve 46:512–519 Rose KJ, Burns J, North KN (2010) Factors associated with foot and ankle strength in healthy preschool-age children and age-matched cases of Charcot-Marie-Tooth disease type 1A. J Child Neurol 25:463–468 Smith BG (2002) Hereditary sensory motor neuropathies. In: Fitzgerald RH, Kaufer H, Malkani AL (eds) Orthopaedics. Mosby, Missouri Thomas PK (1999) Overview of Charcot-Marie-Tooth disease type 1A. Ann N Y Acad Sci 883:1–5 Tooth HH (1886) The peroneal type of progressive muscular atrophy. MD thesis, University of Cambridge Vinci P, Perelli SL (2002) Footdrop, foot rotation, and plantarflexor failure in Charcot-Marie-Tooth disease. Arch Phys Med Rehabil 83:513–516 Vinci P, Serrao M, Pierelli F, Sandrini G, Santilli V (2006) Lower limb manual muscle testing in the early stages of Charcot-Marie-Tooth disease type 1A. Funct Neurol 21:159–163 Ward CM, Dolan LA, Bennett DL, Morcuende JA, Cooper RR (2008) Long-term results of reconstruction for treatment of a flexible cavovarus foot in Charcot-Marie-Tooth disease. J Bone Joint Surg Am 90:2631–2642

Motor Patterns Recognition in Parkinson’s Disease Pierpaolo Sorrentino, Valeria Agosti, and Giuseppe Sorrentino

Abstract

Parkinson’s disease (PD) is characterized clinically by main motor symptoms such as tremor at rest, rigidity, and bradykinesia that affect movements, including gait and postural adjustments. The diagnosis is based on the clinical recognition of these symptoms with the consequent high interrater variability. In order to perform an objective and early diagnosis, approaches that overcome the limitations inherent to clinical examination are needed. In the present work, we will describe several classical technological approaches, such as 3D motion analysis, to achieve an objective evaluation of the cardinal motor symptoms in PD. Furthermore, we will take into account the attempts to identify pathological patterns of integrated, more complex functions such as gait and posture. Finally, as future directions, we will discuss the machine learning approaches in the individuation of specific gait patterns in PD. Keywords

Parkinson’s disease • Movement pattern • Gait analysis • Machine learning • Clinical scales • Gait disorders • Postural instability

Contents Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1238 State of the Art . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1238 Objective Assessment of the Motor Cardinal Symptoms . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1239 P. Sorrentino Department of Engineering, University of Naples Parthenope, Naples, Italy e-mail: [email protected]; [email protected] V. Agosti (*) • G. Sorrentino Department of Motor Sciences and Wellness, University of Naples Parthenope, Naples, Italy Institute Hermitage-Capodimonte, Naples, Italy e-mail: [email protected]; [email protected] # Springer International Publishing AG, part of Springer Nature 2018 B. Müller, S.I. Wolf (eds.), Handbook of Human Motion, https://doi.org/10.1007/978-3-319-14418-4_64

1237

1238

P. Sorrentino et al.

Tremor . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Rigidity . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Bradykinesia . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Integrated Motor Functions: Gait and Postural Stability . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Technological Approaches to Gait and Posture Assessment . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Gait in PD . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Posture in PD . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Future Directions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

1239 1240 1241 1242 1243 1245 1246 1246 1248

Introduction Idiopathic Parkinson’s disease (PD) is a progressive multisystem neurodegenerative disease with a prevalence in industrialized countries estimated at 0.3% of the entire population and about 1% in people over 60 years of age (de Lau and Breteler 2006). Neuropathologically, PD is characterized by the degeneration of dopaminergic nigrostriatal neurons with deposits of insoluble polymers of alpha synuclein forming cytoplasmic inclusions called Lewy bodies. The diagnosis of PD is based on the clinical observation. The clinical symptoms are usually dominated by the motor disturbances. The main motor feature to diagnose PD is the slowness of voluntary movements; one additional symptom, either muscular rigidity, resting tremor, or postural instability, should be present. A number of nonmotor symptoms may also be present. However, nonmotor features have extremely low specificity, limiting their application in clinical diagnosis. The sole clinical evaluation may be source of possible interrater variability and mistakes. In order to perform an objective diagnosis, approaches that overcome the limitation inherent to clinical examination are needed. In this work, firstly we will discuss the methodological issues to take into account in order to use properly the available technological approaches. Furthermore, we will consider the attempts to identify pathological patterns of more complex functions that require the integration of many motor and cognitive skills, such as gait and posture. Finally, as future directions, we will discuss the machine learning approaches in the individuation of specific gait patterns in PD.

State of the Art In 1817, James Parkinson, in the article An Essay on the Shaking Palsy, first described the main clinical manifestations of the disease that now bears his name. In the work, based entirely on the personal visual observations of six patients, Parkinson stated “Involuntary tremulous motion, with lessened muscular power, in parts not in action and even when supported; with a propensity to bend the trunk forward, and to pass from a walking to a running pace: the senses and intellects being uninjured” (Walshe et al. 1961). As of today, except for a few new approaches (e.g., DaTscan) that have only a supportive value, the diagnosis of PD remains based

Motor Patterns Recognition in Parkinson’s Disease

1239

on the clinical observation. The disease is usually diagnosed by the first motor symptoms according to the criteria from the UK PD Brain bank (Hughes et al. 1992). The cardinal motor feature that physicians look at when diagnosing PD is the slowness of initiation of voluntary movements with progressive reduction in speed and amplitude of repetitive actions (bradykinesia); one additional symptom, either resting tremor, muscular rigidity, or postural instability, should be present (Hughes et al. 1992). However, differently from Parkinson’s opinion (the senses and intellects being uninjured), there are a number of nonmotor features that can be detected clinically or anamnestically earlier then the motor ones. Interestingly, nonmotor symptoms might show a much earlier onset (Schrag et al. 2015). However, nonmotor features have extremely low specificity, limiting their application in clinical practice. In the attempt to overcome the limitations linked to the subjective observation, clinical scales have been adopted, the Unified Parkinson’s Disease Rating Scale (UPDRS) part III being the most widely used (Gelb et al. 1999). Although currently considered the gold standard, clinical scales are flown on two different levels. Firstly, interrater variability is a source of possible mistakes and confounders (Shulman et al. 2016). On a more profound level, the detectability of clinical features depends on both the magnitude of the clinical features and the strength that the rater can feel. A proposed relationship between these two quantities can be estimated according to the Webener-Fetchner law (Dehaene 2003; Nieder and Miller 2003), whereby the intensity of the stimulus and the minimum perceivable variation would be related by a logarithmic relationship. This implies that a logarithmic variation of the stimulus would relate to an additive progression of the perception of the stimulus. Hence, in order to perform early diagnosis and stratification of patients, there is the necessity to have an objective and precise approach that overcomes the limitation inherent to clinical examination.

Objective Assessment of the Motor Cardinal Symptoms Tremor Tremor is usually defined as an involuntary, rhythmic, oscillatory movement of a body part (Daroff et al. 2016). Clinically, tremors are classically divided into rest, postural, and action tremors, according to its semiologic features. Typically, PD patients show rest tremor, characterized by the involvement of agonist and antagonist muscles, with a frequency between 4 and 6 Hz. Since the rating of tremor does not relate with the ratings of bradykinesia and rigidity (Martinez-Martin et al. 1994), it has been hypothesized that different pathophysiological mechanisms underlie these symptoms. Indeed, patients with tremorigen form of PD tend to have a milder course of the disease as compared to the ones with a rigid onset (Obeso et al. 2010). Hence, careful classification might yield direct information on specific central structures. Typically, PD shows the pattern of a central tremor, whose frequency is function of neither the limb properties (i.e., joint inertia, stiffness of the soft tissues) nor of the peripheral nervous system (loop–reflex time) (Elble and Deuschl 2011).

1240

P. Sorrentino et al.

Classically, the tremor is analyzed in order to help diagnosis, and its rating is part of various clinical scores used to monitor disease progression and to modify the therapy accordingly. In section III of the UPDRS, several items evaluate the tremor and its main characteristics. However, according to the above mentioned Webener–Fatchner law, the clinical evaluation of tremor is inherently bounded to the rater’s perception of the frequency and amplitude, and several attempts have been performed to achieve an objective evaluation of tremor. Classically used instruments are electromyography (EMG), accelerometers, gyroscopes, flexible angular sensors and goniometers, video and optoelectronic devices, force sensors, and wearable orthosis. First off, it has to be noticed how the choice of the device should be in accordance with the features that we want to capture and that the interpretation of the data should take into account the influence of the device on the measurements. For instance, devices that restrain the range of movement of the joint hinder the interpretation of amplitude while allowing for the simulation of isometric conditions. In this case, it has been showed that the frequency of the tremor is not influenced by the constraints, while the amplitude is. This would point to the idea that frequency is more directly related to the features of the tremor generators, while amplitude might by affected by the mechanics of the limb. The typical characteristic of the tremor signal that are studied in PD are: (a) frequency spectrum (by Fourier transform of the time series) (Timmer et al. 1996); (b) peak frequency (Gao 2004); (c) temporal fluctuation; (d) root mean square of linear acceleration; (e) approximate entropy; (f) shape of signal distribution; (g) wavelet coefficient; and (h) higher order statistics parameters (Thanawattano et al. 2015).

Rigidity The term rigidity defines an increase in muscle tone involving both agonist and antagonist muscles of the same joint (Daroff et al. 2016). In PD rigidity is one of the main clinical signs that concurs to impair the ability to move about freely. Although the underlying mechanisms are poorly understood, rigidity is a precise indicator of the residual motor function, and its detection and monitoring appears crucial since it is responsive to dopaminergic treatments. The part III of the UPDRS includes the rigidity evaluation. The neurologist performs manual maneuvers of passive flex-extension of the upper and lower limbs as well as of the trunk to estimate, subjectively, the rate of resistance or stiffness. However, similarly (may be more) to the tremor evaluation, interrater variability is a source of possible mistakes. To date, to assess muscle tone, surface EMG has been widely used, classically performed during dynamic flex/extension of the limb (Mortimer and Webster 1979; Xia et al. 2011). The methods to evaluate EMG signals are usually based on the analysis of the amplitude and spectrum (Nieminen and Takala 1996). However, given the inherent nonlinear nature of the EMG signals, great effort has been put into

Motor Patterns Recognition in Parkinson’s Disease

1241

the fractal-based dimensionality analysis as well as nonlinear time series analysis (Swie et al. 2005; Meigal et al. 2013). However, surface EMG signals are strongly influenced by the characteristics of the soft tissues. Indeed, the correlation between EMG characteristics and UPDRS-based evaluation of rigidity has shown not to be as strong as expected. In order to overcome these limitations, many studies have focused on purely biomechanical measures, such as peak torque, elastic coefficient, and mechanical impedance. Despite these techniques yielded improved correlations as compared to EMG-based analysis, they do not allow the distinction between the rigidity induced by neurologically mediated mechanisms from that derived by altered mechanical properties of the joints (Endo et al. 2009). Interestingly, Park et al. (2011) proposed a model that might allow the distinction between these two components of rigidity.

Bradykinesia Bradykinesia is defined as slowness of movement, often accompanied by other features such as hypokinesia (reduced amplitude of movement) (Daroff et al. 2016). Bradykinesia is a very complex sign, whose pathophysiological mechanisms are only partially understood. Furthermore, the patients present a general impoverishment of the movements including hypomimia, micrographia, and speech alterations (Berardelli et al. 2001). Concomitant causes that might lead to bradykinesia include: muscle weakness, rigidity, tremor, movement variability, and bradyphrenia. The severity of bradykinesia is classically measured with different tasks involving rapid and voluntary movements. In UPDRS III, the tasks for the upper limb include the finger tapping (patient taps thumb with index finger in rapid succession), the hand movement (patient opens and closes hands in rapid succession), and rapid alternating movement of hand (prono-supination of the hands). The tasks for the lower limbs include leg agility (patient taps heals on the ground in rapid succession picking up entire leg). In order to improve the reliability of such estimates, several studies investigated the use of accelerometers (Yokoe et al. 2009; Stamatakis et al. 2013) or magnetic coils (Kandori et al. 2004; Shima et al. 2008) in measuring finger tapping. Yokoe et al. (2009) analyzed 14 parameters of finger tapping and used a statistical approach called principal component analysis (PCA) to reduced them into three components, of which the first (velocity and amplitude related) and the third (rhythm related) could differentiate between patients and controls. Kandori et al. (2004) used magnetic coil to analyze the finger taping. After filtering the signal, they found out that the average amplitude of each waveform decreased proportionally to the Hoehn and Yahr staging system (Hoehn and Yahr 1967). However, both these clinical and instrumented measurement do not capture the complex and multifactorial nature of this clinical feature. Furthermore, these investigations are to be performed in advanced clinical settings and in a strictly controlled environment, hence they do not provide information on the movement of the patient in daily activity. A possible solution to bridge this gap might be the new

1242

P. Sorrentino et al.

technological applications such as wearable sensors tools, which allow us to check the patient daily movements. Focusing on two experimental tools and conditions, Salarian et al. (2007) experimented the use of gyroscopes in quantifying bradykinesia. In particular, they used a self-made micro-gyroscope unit, easy to wear and to use, to detect parameters of the movement and their correlation with the UPDRS III scores. The author showed that bradykinesia, as detected by the analysis of the entire recorded period, related with the UPDRS III bradykinesia subscore. Another interesting and more recent study analyzed the joint movement in the upper limb (finger and forearm) but also in the lower limb (ankle) by using a gyro-sensors system (Kim et al. 2015). In this study, 14 features were derived from the sensors and a stepwise multiple linear regression analysis was used to develop a model able to predict the bradykinesia score. Interestingly, the regression models were better than any single variable in predicting the clinical scores.

Integrated Motor Functions: Gait and Postural Stability According to Jacksonian’s model (Walshe et al. 1961), the nervous system is organized in three levels of increasing complexity. In this light, movements such as gait and postural adjustments may be considered as the result of the integration of the spinal cord (lower level), basal ganglia and motor cortex (middle level), and prefrontal cortex (higher level). Hence, from the functional stand point, gait and balance are the results of the integration of different mechanisms ranging from the automatic reflexes (spinal cord) to the cognitive functions (prefrontal cortex). Consequently, the recognition of motion patterns is a challenging issue since gait and posture can be influenced by the (mis)functioning of any of those structures. The gait describes human locomotion, or, according to what stated above, the way of walking. Since each individual has distinctive gait patterns it is useful to divide them into gait cycles, defined as the interval from the initial placement of the supporting heel on the ground to the moment when the same heel touches the ground again. Classically, the gait cycle is further divided into eight phases (Perry et al. 2010). This division is useful since each phase has a specific functional meaning, and this helps the clinical interpretation of the patterns. Hence, by evaluating individual gait patterns, we can determine specific weaknesses and/or tailor rehabilitation programs. Furthermore, when the gait is affected by injury or disease, it can show some features/patterns that are common to all affected individuals (Perry et al. 2010; Levine et al. 2012). The postural stability emerges from activation and integration of multiple sensory, motor, and cognitive systems. The complex patterns of coordinated activity that are needed to achieve postural stability are normally referred to as postural control (Massion 1992; Cappa et al. 2008). In other words, postural control could be defined as the ability to maintain stability of the body and its segments. Postural control is crucial to maintain the body’s orientation in the space, the body’s center of mass (CoM) over the base of support (BoS), and to stabilize the position of the head.

Motor Patterns Recognition in Parkinson’s Disease

1243

From the neuromechanical perspective, the central nervous system interprets the inputs from the various subsystems and organizes appropriate responses based on previous and current information (Enoka 2015). These responses could be either compensatory or anticipatory. The formers occur as reactions to external forces that displace the body’s CoM, and the latters anticipate internally generated destabilizing forces such as those induced, for instance, by raising arms or bending forward. Català et al. (2016) using a motion analysis system evaluated the dynamic stability control during gait in PD patients. They estimated the adaptability and the postural strategies to maintain balance during locomotion. Interestingly, they found that the compensatory responses (called “reactive” by the authors) are specifically impaired in PD, as opposed to anticipatory one. In order to explain the strict relationship between gait and postural stability, it is important to highlight that one of the three main tasks of the gait cycle is “to ensure the advancement maintaining dynamic stability.” Indeed, in order to achieve balance during static or dynamic tasks, a perfect integration between CoM position and velocity is required (Jian et al. 1993). This ability is often lost in neurological diseases (i.e., PD), predisposing to tripping and falling during walking (Ashburn et al. 2001).

Technological Approaches to Gait and Posture Assessment Gait parameters can be described as kinetic and/or kinematic data. On the one hand, the kinetic data is useful to describe the forces acting on the joints, the moments produced by the muscles crossing the joints, and the energy required during gait. The kinetic data usually results from a force plate, providing information about the components of the ground reaction force (GRF) that are in turn proportional to the magnitude and direction of the load applied to one or both feet when the body touches the ground. On the other hand, kinematic data helps us to study the timing and the magnitude of gait. In fact, kinematic data provides information about the phases of gait (spatiotemporal data) and/or about the three components of the articular range of motion (RoM). Kinematic data are usually derived from video analysis (two-dimensional data) or motion analysis (three-dimensional data, 3DMA) or even from special sensorized mats. On a practical standpoint, we can divide the devices that are eligible to study the gait in those that are to be used in specialized settings (i.e., laboratory) and those that can be easily used in a day to day context. The information that can be obtained in a controlled setting are better in quality and quantity, while the data from the living environment of the patients can provide valuable insight on the impact of the impairment (or the therapy) on the daily life of the patient. Integrated motion analysis systems allow to collect simultaneously kinematic, muscular, and kinetic data from stereophotogrammetric cameras, EMG, and force plate, respectively. Morris et al. (1999, 2005) by means of a motion analysis approach described the spatial and temporal dysfunctions in PD gait that is characterized by decreased velocity, shorter stride, increased cadence, stride length and

1244

P. Sorrentino et al.

stride time variability, altered heel strike, toe-off, and arm swing. Multiple studies on kinematic changes in PD showed decreased ankle RoM in pre-swing, decreased hip extension in pre-swing, and reduced power at ankle push off and hip pull off (Sofuwa et al. 2005; Švehlík et al. 2009; Roiz et al. 2010). 3DMA is a powerful approach in validating rehabilitation strategies as well. For instance, a significant improvement of both spatiotemporal (Vitale et al. 2012) and kinematic parameters (Agosti et al. 2016) has been demonstrated in PD patients following a global postural reeducation program. Despite the fact that the 3DMA remains the gold standard in the motor evaluation of PD (Sale et al. 2013), this technique has been criticized since it might induce biases due to the laboratory setting. In the experience of the authors, PD specific alterations, such as fatigue or altered perspiration (which makes it hard to attach the markers), may specifically bias the analysis of these patients. To overcome these problems, further laboratory instruments have been developed such as GAITRite® (McDonough et al. 2001). GAITRite® is a portable sensorized mat built for automated measurement of spatiotemporal gait parameters in daily ambulatory use. Nelson et al. (2002) used this tool to assess footfall patterns and selected gait characteristics in early PD. Recently, with the advent of new technological solutions, the use of wearable sensorized tools is gaining ground not only in the outpatient use but also for home monitoring. GAITRite ® was used to validate the measurement of these wearable systems. Indeed, Lin et al. (2016) proposed a new and useful portable monocular image system to track and analyze parkinsonian gait patterns using a centroid tracking algorithm (a system which can be easily used in locations with budget and space limitations). Lord et al. (2008) tested the validity of a five accelerometer portable system (the Vitaport Activity Monitor ® (VAM)) in assessing postural transitions and dynamic mobility during daily life activities in PD patients. In order to extend the analysis to the total body, Tzallas et al. (2014) presented a new wearable multi-sensor monitor unit composed by tri-axial accelerometers and gyroscopes connected wirelessly. This kind of device is a step toward a multiparameter approach to the monitoring of PD that might help to personalize the treatment and medication schedules according to comprehensive data generated in the daily life of the patient. Static stability has been classically investigated by a postural clinical evaluation or by a posturographic platform (Bloem et al. 2016). The UPDRS motor score has a task called “postural stability” whereby response to sudden posterior displacement is evaluated. Finally, dynamic postural stability is usually evaluated by means of a complex test, the Time Up and Go Test (TUG), which requires the ability of standing up from a chair, walking along 3 m, turning, and finally sitting down. An instrumented version (iTUG) of this scale was recently proposed on early-mild PD subjects using a single tri-axial accelerometer worn on the lower back (Palmerini et al. 2013). In the last few years, there is an increasing body of evidence, confirmed by systematic review (Hubble et al. 2015), that wearable sensors are useful systems not only in evaluating gait patterns but are also excellent screening tools for balance and postural instability.

Motor Patterns Recognition in Parkinson’s Disease

1245

Gait in PD Parkinsonian gait has a clearly identifiable pattern when the neurological involvement is advanced, including features such as festination and freezing of gait (FoG) (Grabli et al. 2012). The former expresses “the tendency to move forward with increasingly rapid, but ever smaller steps, associated with the Centre of gravity (CoG) falling forward over the stepping feet” (Bloem et al. 2004). The latter defines “the brief, episodic absence or marked reduction of forward progression of the feet despite the intention to walk” (Giladi and Nieuwboer 2008). The link between festination and FoG is still debated and is a topic of growing interest to understand gait impairment mechanisms in PD (Nutt et al. 2011). Further typical characteristics that might affect gait include a range of disorders globally known as axial motor impairments (such as lateral trunk flexion, aka PISA syndrome (PS), and camptocormia) (Doherty et al. 2011) as well as cognitive decline (Ebersbach et al. 2013). However, early detection of typical patterns, when the alterations are subtle, is a challenging issue. The picture becomes even more complex considering that PD involves compensatory mechanisms to maintain an effective gait (i.e., nonautomatic functioning can initially compensate for the misfunctioning of normally automated processes) (Ebersbach et al. 2013). Interestingly, some insight can be gained from the study of gait in carriers of genetic mutations causing PD. These patients carry in their genome a mutation that causes PD, yet being clinically healthy. These patients showed higher variability of the stride only during a challenging condition (either during walking at high speed or during a dual task). These results are intriguing since they might show that challenging tasks might unmask deficits by preventing the use of further cognitive resources (compensative mechanisms) (Mirelman et al. 2011). In this line of thinking, patients with Gaucher disease type 1, a condition that is classically nonneurological yet carrying a higher risk of developing PD, showed reduced amplitude of the RoMs of the lower limbs (Sorrentino et al. 2016), again suggesting that alterations may be present before clinical onset. Bonora et al. (2015) investigated gait initiation in PD. This is especially interesting, since the most common form of FoG, known as “start hesitation,” happens when the patient wishes to start walking. Furthermore, from a biomechanical point of view, the postural adjustments that anticipate the initiation of walking are fundamental to achieve stability. Using a system based on inertial sensors and a force plate, the authors showed that PD patients had reduced medio-lateral CoP displacement and concluded that the PD patients might have more difficulties in adjusting the imbalance phase during gait initiation. PS syndrome is defined as a “marked lateral tilt of the trunk, typically reducible by passive mobilization or when the patient lies down” (Doherty et al. 2011). A recent pilot observational cross-sectional study (Geroin et al. 2015), using both posturography and gait analysis data, estimated the impact of PS in PD patients. They showed that in PD patients with PS, as compared to PD patients without PS and age matched healthy controls, there was significantly greater body sway velocity in the antero-posterior and medial-lateral directions.

1246

P. Sorrentino et al.

Finally, mild cognitive impairment (MCI), that can be present in PD patients, may have an effect on gait patterns. For instance, comparable groups of PD patients with or without MCI showed different gait patterns, such as reduced step length and swing time. Furthermore, the worsening of gait parameters was more pronounced in PD patients with MCI than in patients without MCI (Amboni et al. 2012). A recent cross-sectional study (Kelly et al. 2015) demonstrated the association of specific motor dysfunctions with cognitive decline in a large PD cohort. The authors showed that the severity of motor symptoms of postural instability is linked to dysfunction in specific cognitive domains (executive function, memory, visuospatial function, and phonemic fluency).

Posture in PD Postural instability, a condition that is poorly responsive to levodopa, is a risk factor in predisposing PD patients to falls (Grabli et al. 2012). There are a number of clinical functional balance tests to evaluate postural stability, i.e., the Tinetti scale, the functional reach test, one leg stance test, the postural balance task on the UPDRS III, etc. (Mancini and Horak 2010). Clinical balance tests (especially when combined) allow a good prediction of fallers and nonfallers in the PD population (Jacobs et al. 2006) although with margin for improvement. Often the technological approach is useful to determine the underlying causes of the balance loss. However, when measuring in a clinical setting, it is important to bear in mind that consensus has been reached on some standard requirements (Scoppa et al. 2013). Static posturographic assessment showed greater postural imbalance in PD patients as compared to healthy controls (Doná et al. 2015). Further improvement can be obtained using dynamic posturography, which is the posturographic assessment during postural perturbations. It was found that in PD the forward bending induced by backward perturbation is impaired. Interestingly, after repetitive perturbation, the postural reaction of the PD patients improved as much as that of the healthy controls. This has implications for rehabilitation and is an example of the further insight achievable using dynamic posturography (Visser et al. 2010). Lastly, using an inertial–sensor based system, Rocchi et al. (2013) computed postural measures to classify PD subtypes. Furthermore, they showed that postural displacement measures might capture different compensative mechanisms.

Future Directions As we have mentioned, PD is a very heterogeneous disease, and in real world scenarios patients are likely to face a number of symptoms. Furthermore, the presence of a given set of symptoms has important implication for therapeutic purposes (i.e., patients with rigid form of PD will have a heavier disease load as well as a poorer and shorter response to drug therapy). Hence, a variety of statistical

Motor Patterns Recognition in Parkinson’s Disease

1247

techniques, that might be classified as machine learning, have been used in order to spot patterns of symptoms that might lead to diagnosis. Machine learning approaches include a bunch of techniques roughly involving the following steps: 1. Define the relevant features (either using prior knowledge or using data driven algorithms) 2. Select an appropriate algorithm to identify statistically consistent patterns among the selected features in a part of the population in analysis (training set) 3. Use the training set to identify classes (or to allocate elements in previously defined classes) 4. Test the performance by allocating new elements (test set) that the algorithm has never met before in accordance to the classification criteria defined on the training set It appears evident how such techniques can be used to exploit typical signs and symptoms of the disease in order to classify a subject, for instance, as affected or not affected. The most widely used devices to obtain data are accelerometer, gyroscopes and, less often, EMG recording. Once the signal has been obtained, it is important to choose how to analyze it. Various kind of parameters have been used to analyze time series, such as spectral features, skewness, curtosis, fractal features, entropy, and information theory–based metrics. Furthermore, a great effort has been put into the selection of the most effective algorithms in order to classify the features. Attempts have been made to apply neural networks, hidden Markov models, support vector machines, and decision trees. Daliri (2012) applied force-sensitive resistors (corresponding roughly to the force under the foot) to distinguish patients with Huntington’s disease, PD, and amyotrophic lateral sclerosis. Classical parameters used in gait analysis were retrieved, such as stride, swing, stance, and double support intervals. The features extracted of these parameters were the minimum, maximum, average, and the standard deviation. The most informative features were selected using a genetic algorithm, and then they were used in a support vector machine in order to classify the subjects. Various kernels were used, and the best results were obtained with the radial basis function kernel. Sensitivity, specificity, and accuracy were around 90% for all the diagnosis. In a subsequent study (Daliri 2013), the difference between the force measured in eight sensors placed underneath each foot was analyzed using a short time Fourier transform. Subsequently, a support vector machine was applied (with a kernel formed by the chi-square distance between the histogram of the selected features), obtaining an accuracy of roughly 91%. Lee and Lim (2012) used a database of features of the gait to implement a procedure to identify PD patients. The authors tried three different preprocessing approaches to obtain characteristic of the signals that might capture known features of the parkinsonian gait (i.e., shuffling gait). Then, using a wavelet-based feature extraction, they selected 40 features and used them as an input in a neural network with weighted fuzzy membership functions (NEWFM). The results varied according to the preprocessing methods, with an accuracy of around 75% for the three of them.

1248

P. Sorrentino et al.

Tahir and Manap (2012) compared artificial neural network and support vector machine in distinguishing gait pattern during self-selected speed walking, finding comparable results for the two methods. Yunfeng and Krishnan (2010) estimated the probability density functions (PDFs) of stride intervals. They found that the gait variability was increased in PD patients. The stride interval parameters were used in a support vector machine with polynomial kernel, and the results validated with the leave–one–out approach. The classification rate had an area under the receiving operator curve (AUC) of 0.952. Wahid et al. (2015) argued that true characteristics of the gait pattern might be modified by other factors affecting gait, such as age, height, body mass, and gender. To try to address this issue, they normalized gait parameters before using machine learning approaches. In this case, the random forest yielded the highest accuracy (92.6%) as compared to support vector machine (80.4%) and kernel Fisher discriminant (86.2%). Bächlin et al. (2009) used a set of accelerometers on shank, thigh, and waist in order to build a wearable device to identify FoG. They obtained a sensitivity of 73.1% and a specificity of 81.6%. Djuric-Jovicic et al. (2014) applied to the data obtained from an inertial sensor an algorithm technique to automatically detect and classify episodes of FoG. Tripoliti et al. (2013) explored how to combine a set of six accelerometers and two gyroscopes in order to classify episodes of FoG. The authors tried different algorithms (Naïve Bayes, Random Forests, Decision Trees, and Random Tree) and concluded that the best results were obtained applying a random forest to all the sensors (98.6 AUC). Interestingly, Cole et al. (2011) combined several accelerometers (forearm, thigh, and shin) and one EMG channel (shin) and set out to identify episodes of FoG. They applied a dynamic neural network and, when tested on experimental data from PD patients, the algorithm showed a specificity of 97% and sensitivity of 83%. All in all, it is possible to notice how machine learning techniques are able to capture features of the gait and exploit them to either diagnose subjects or classify symptoms. In the opinion of the authors, it comes to no surprise that very good results are achieved with random forests, given the fact that these algorithms are particularly prone to capture heavy nonlinearities such as those that are likely to arise between features of gait. Future direction might involve improved techniques of feature selection and/or better classification algorithms as well as the integration of more patient characteristics (anamnestic, clinical, imaging, gait analysis) into one classification algorithm. This will mimic always more the clinical process that goes on in the mind of every physician faced with a hard clinical diagnosis.

References Agosti V, Vitale C, Avella D et al (2016) Effects of Global Postural Reeducation on gait kinematics in parkinsonian patients: a pilot randomized three-dimensional motion analysis study. Neurol Sci 37:515–522. https://doi.org/10.1007/s10072-015-2433-5

Motor Patterns Recognition in Parkinson’s Disease

1249

Amboni M, Barone P, Iuppariello L et al (2012) Gait patterns in parkinsonian patients with or without mild cognitive impairment. Mov Disord 27:1536–1543. https://doi.org/10.1002/ mds.25165 Ashburn A, Stack E, Pickering RM, Ward CD (2001) A community-dwelling sample of people with Parkinson’s disease: characteristics of fallers and non-fallers. Age Ageing 30:47–52. https://doi. org/10.1093/ageing/30.1.47 Bächlin M, Roggen D, Tröster G et al (2009) Potentials of enhanced context awareness in wearable assistants for Parkinson’s disease patients with the freezing of gait syndrome. In: Proceedings – international symposium on wearable computers, ISWC, IEEE, 4–7 Sep - Linz, Austria, pp 123–130 Berardelli A, Rothwell JC, Thompson PD, Hallett M (2001) Pathophysiology of bradykinesia in Parkinson’s disease. Brain 124:2131–2146. https://doi.org/10.1093/brain/124.11.2131 Bloem BR, Hausdorff JM, Visser JE, Giladi N (2004) Falls and freezing of gait in Parkinson’s disease: a review of two interconnected, episodic phenomena. Mov Disord 19:871–884. https:// doi.org/10.1002/mds.20115 Bloem BR, Marinus J, Almeida Q et al (2016) Measurement instruments to assess posture, gait, and balance in Parkinson’s disease: critique and recommendations. Mov Disord. https://doi.org/ 10.1002/mds.26572 Bonora G, Carpinella I, Cattaneo D et al (2015) A new instrumented method for the evaluation of gait initiation and step climbing based on inertial sensors: a pilot application in Parkinson’s disease. J Neuroeng Rehabil 12:45. https://doi.org/10.1186/s12984-015-0038-0 Cappa P, Patanè F, Rossi S et al (2008) Effect of changing visual condition and frequency of horizontal oscillations on postural balance of standing healthy subjects. Gait Posture 28:615–626. https://doi.org/10.1016/j.gaitpost.2008.04.013 Catalá MM, Woitalla D, Arampatzis A (2016) Reactive but not predictive locomotor adaptability is impaired in young Parkinson’s disease patients. Gait Posture 48:177–182. https://doi.org/ 10.1016/j.gaitpost.2016.05.008 Cole BT, Roy SH, Nawab SH (2011) Detecting freezing-of-gait during unscripted and unconstrained activity. In: Proceedings of the annual international conference of the IEEE Engineering in Medicine and Biology Society, EMBS, IEEE, 30 Aug - 03 Sep - Boston, MA, USA, pp 5649–5652 Daliri MR (2012) Automatic diagnosis of neuro-degenerative diseases using gait dynamics. Meas J Int Meas Confed 45:1729–1734. https://doi.org/10.1016/j.measurement.2012.04.013 Daliri MR (2013) Chi-square distance kernel of the gaits for the diagnosis of Parkinson’s disease. Biomed Signal Process Control 8:66–70. https://doi.org/10.1016/j.bspc.2012.04.007 Daroff RB, Jankovic J, Mazziotta JC et al (2016) Bradley’s neurology in clinical practice, vols 1–2, 7th edn. Elsevier, London de Lau LM, Breteler MM (2006) Epidemiology of Parkinson’s disease. Lancet Neurol 5:525–535. https://doi.org/10.1016/S1474-4422(06)70471-9 Dehaene S (2003) The neural basis of the c law: a logarithmic mental number line. Trends Cogn Sci 7:145–147. https://doi.org/10.1016/S1364-6613(03)00055-X Djuric-Jovicic MD, Jovicic NS, Radovanovic SM et al (2014) Automatic identification and classification of freezing of gait episodes in Parkinson’s disease patients. IEEE Trans Neural Syst Rehabil Eng 22:685–694. https://doi.org/10.1109/TNSRE.2013.2287241 Doherty KM, van de Warrenburg BP, Peralta MC et al (2011) Postural deformities in Parkinson’s disease. Lancet Neurol 10:538–549 Doná F, Aquino CC, Gazzola JM et al (2015) Changes in postural control in patients with Parkinson’s disease: a posturographic study. Physiotherapy. https://doi.org/10.1016/j. physio.2015.08.009 Ebersbach G, Moreau C, Gandor F et al (2013) Clinical syndromes: Parkinsonian gait. Mov Disord 28:1552–1559. https://doi.org/10.1002/mds.25675 Elble R, Deuschl G (2011) Milestones in tremor research. Mov Disord 26:1096–1105. https://doi. org/10.1002/mds.23579

1250

P. Sorrentino et al.

Endo T, Okuno R, Yokoe M et al (2009) A novel method for systematic analysis of rigidity in Parkinson’s disease. Mov Disord 24:2218–2224. https://doi.org/10.1002/mds.22752 Enoka RM (2015) Neuromechanics of human movement. Human Kinetics, Champaign Gao JB (2004) Analysis of amplitude and frequency variations of essential and Parkinsonian tremors. Med Biol Eng Comput 42:345–349. https://doi.org/10.1007/BF02344710 Gelb DJ, Oliver E, Gilman S et al (1999) Diagnostic criteria for Parkinson disease. Arch Neurol 56:33. https://doi.org/10.1001/archneur.56.1.33 Geroin C, Smania N, Schena F et al (2015) Does the Pisa syndrome affect postural control, balance, and gait in patients with Parkinson’s disease? An observational cross-sectional study. Parkinsonism Relat Disord 21:736–741. https://doi.org/10.1016/j.parkreldis.2015.04.020 Giladi N, Nieuwboer A (2008) Understanding and treating freezing of gait in parkinsonism, proposed working definition, and setting the stage. Mov Disord 23:S423–S425. https://doi. org/10.1002/mds.21927 Grabli D, Karachi C, Welter M et al (2012) Normal and pathological gait: what we learn from Parkinson’s disease. J Neurol Neurosurg Psychiatry 83:979–985. https://doi.org/10.1136/jnnp2012-302263 Hoehn MM, Yahr MD (1967) Parkinsonism : onset, progression, and mortality. Neurology 17:427–442. https://doi.org/10.1212/WNL.17.5.427 Hubble RP, Naughton GA, Silburn PA, Cole MH (2015) Wearable sensor use for assessing standing balance and walking stability in people with Parkinson’s disease: a systematic review. PLoS One 10:1–22. https://doi.org/10.1371/journal.pone.0123705 Hughes AJ, Daniel SE, Kilford L, Lees AJ (1992) Accuracy of clinical diagnosis of idiopathic Parkinson’s disease: a clinico-pathological study of 100 cases. J Neurol Neurosurg Psychiatry 55:181–184. https://doi.org/10.1136/jnnp.55.3.181 Jacobs JV, Horak FB, Tran VK, Nutt JG (2006) Multiple balance tests improve the assessment of postural stability in subjects with Parkinson’s disease. J Neurol Neurosurg Psychiatry 77:322–326. https://doi.org/10.1136/jnnp.2005.068742 Jian Y, Winter D, Ishac M, Gilchrist L (1993) Trajectory of the body COG and COP during initiation and termination of gait. Gait Posture 1:9–22. https://doi.org/10.1016/0966-6362(93) 90038-3 Kandori A, Yokoe M, Sakoda S et al (2004) Quantitative magnetic detection of finger movements in patients with Parkinson’s disease. Neurosci Res 49:253–260. https://doi.org/10.1016/j. neures.2004.03.004 Kelly VE, Johnson CO, McGough EL et al (2015) Association of cognitive domains with postural instability/gait disturbance in Parkinson’s disease. Parkinsonism Relat Disord 21:692–697. https://doi.org/10.1016/j.parkreldis.2015.04.002 Kim J-W, Kwon Y, Yun J-S et al (2015) Regression models for the quantification of Parkinsonian bradykinesia. Biomed Mater Eng 26(Suppl 1):S2249–S2258. https://doi.org/10.3233/BME151531 Lee S-H, Lim JS (2012) Parkinson’s disease classification using gait characteristics and wavelet-based feature extraction. Expert Syst Appl 39:7338–7344. https://doi.org/10.1016/j.eswa.2012.01.084 Levine D, Richards J, Whittle M, Whittle M (2012) Whittle’s gait analysis. Churchill Livingstone/ Elsevier, Edinburgh Lin S-H, Chen S-W, Lo Y-C et al (2016) Quantitative measurement of Parkinsonian gait from walking in monocular image sequences using a centroid tracking algorithm. Med Biol Eng Comput 54:485–496. https://doi.org/10.1007/s11517-015-1335-2 Lord S, Rochester L, Baker K, Nieuwboer A (2008) Concurrent validity of accelerometry to measure gait in Parkinsons disease. Gait Posture 27:357–359. https://doi.org/10.1016/j. gaitpost.2007.04.001 Mancini M, Horak FB (2010) The relevance of clinical balance assessment tools to differentiate balance deficits. Eur J Phys Rehabil Med 46:239–248 Martinez-Martin P, Gil-Nagel A, Gracia LM et al (1994) Unified Parkinson’s disease rating scale characteristics and structure. Mov Disord 9:76–83. https://doi.org/10.1002/mds.870090112

Motor Patterns Recognition in Parkinson’s Disease

1251

Massion J (1992) Movement, posture and equilibrium: interaction and coordination. Prog Neurobiol 38:35–56. https://doi.org/10.1016/0301-0082(92)90034-C McDonough AL, Batavia M, Chen FC et al (2001) The validity and reliability of the GAITRite system’s measurements: a preliminary evaluation. Arch Phys Med Rehabil 82:419–425. https:// doi.org/10.1053/apmr.2001.19778 Meigal AY, Rissanen SM, Tarvainen MP et al (2013) Non-linear EMG parameters for differential and early diagnostics of Parkinson’s disease. Front Neurol. https://doi.org/10.3389/fneur.2013. 00135 Mirelman A, Gurevich T, Giladi N et al (2011) Gait alterations in healthy carriers of the LRRK2 G2019S mutation. Ann Neurol 69:193–197. https://doi.org/10.1002/ana.22165 Morris ME, McGinley J, Huxham F et al (1999) Constraints on the kinetic, kinematic and spatiotemporal parameters of gait in Parkinson’s disease. Hum Mov Sci 18:461–483 Morris M, Iansek R, McGinley J et al (2005) Three-dimensional gait biomechanics in Parkinson’s disease: evidence for a centrally mediated amplitude regulation disorder. Mov Disord 20:40–50. https://doi.org/10.1002/mds.20278 Mortimer JA, Webster DD (1979) Evidence for a quantitative association between EMG stretch responses and Parkinsonian rigidity. Brain Res 162:169–173. https://doi.org/10.1016/00068993(79)90768-6 Nelson AJ, Zwick D, Brody S et al (2002) The validity of the GaitRite and the functional ambulation performance scoring system in the analysis of Parkinson gait. NeuroRehabilitation 17:255–262 Nieder A, Miller EK (2003) Coding of cognitive magnitude: compressed scaling of numerical information in the primate prefrontal cortex. Neuron 37:149–157. https://doi.org/10.1016/ S0896-6273(02)01144-3 Nieminen H, Takala EP (1996) Evidence of deterministic chaos in the myoelectric signal. Electromyogr Clin Neurophysiol 36:49–58 Nutt JG, Bloem BR, Giladi N et al (2011) Freezing of gait: moving forward on a mysterious clinical phenomenon. Lancet Neurol 10:734–744. https://doi.org/10.1016/S1474-4422(11)70143-0 Obeso JA, Rodriguez-Oroz MC, Goetz CG (2010) Missing pieces in the Parkinson’s disease puzzle. Nat Med 16:653–661. https://doi.org/10.1038/nm.2165 Palmerini L, Mellone S, Avanzolini G et al (2013) Quantification of motor impairment in Parkinson’s disease using an instrumented timed up and go test. IEEE Trans Neural Syst Rehabil Eng 21:664–673. https://doi.org/10.1109/tnsre.2012.2236577 Park BK, Kwon Y, Kim JW et al (2011) Analysis of viscoelastic properties of wrist joint for quantification of parkinsonian rigidity. IEEE Trans Neural Syst Rehabil Eng 19:167–176. https://doi.org/10.1109/TNSRE.2010.2091149 Perry J, Burnfield JM, Cabico LM (2010) Gait analysis: normal and pathological function. Second Edition. SLACK Incorporated. Thorofare, NJ, USA Rocchi L, Palmerini L, Weiss A et al (2013) Balance testing with inertial sensors in patients with Parkinson’s disease: assessment of motor subtypes. IEEE Trans Neural Syst Rehabil Eng 22:1064–1071. https://doi.org/10.1109/TNSRE.2013.2292496 Roiz RDM, Cacho EWA, Pazinatto MM et al (2010) Gait analysis comparing Parkinson’s disease with healthy elderly subjects. Arq Neuropsiquiatr 68:81–86. https://doi.org/10.1590/S0004282X2010000100018 Salarian A, Russmann H, Wider C et al (2007) Quantification of tremor and bradykinesia in Parkinson’s disease using a novel ambulatory monitoring system. IEEE Trans Biomed Eng 54:313–322 Sale P, De Pandis MF, Vimercati SL et al (2013) The relation between Parkinson’s disease and ageing. Comparison of the gait patterns of young Parkinson’s disease subjects with healthy elderly subjects. Eur J Phys Rehabil Med 49:161–167 Schrag A, Horsfall L, Walters K et al (2015) Prediagnostic presentations of Parkinson’s disease in primary care: a case–control study. Lancet Neurol 14:57–64. https://doi.org/10.1016/S14744422(14)70287-X

1252

P. Sorrentino et al.

Scoppa F, Capra R, Gallamini M, Shiffer R (2013) Clinical stabilometry standardization. Basic definitions – acquisition interval – sampling frequency. Gait Posture 37:290–292. https://doi. org/10.1016/j.gaitpost.2012.07.009 Shima K, Tsuji T, Kan E et al (2008) Measurement and evaluation of finger tapping movements using magnetic sensors. Conf Proc IEEE Eng Med Biol Soc 1–8:5628–5631. https://doi.org/ 10.1109/IEMBS.2008.4650490 Shulman LM, Armstrong M, Ellis T et al (2016) Disability rating scales in Parkinson’s disease: critique and recommendations. Mov Disord. https://doi.org/10.1002/mds.26649 Sofuwa O, Nieuwboer A, Desloovere K et al (2005) Quantitative gait analysis in Parkinson’s disease: comparison with a healthy control group. Arch Phys Med Rehabil 86:1007–1013. https://doi.org/10.1016/j.apmr.2004.08.012 Sorrentino P, Barbato A, Del Gaudio L et al (2016) Impaired gait kinematics in type 1 Gaucher’s disease. J Parkinsons Dis 6:191–195. https://doi.org/10.3233/JPD-150660 Stamatakis J, Ambroise J, Crémers J et al (2013) Finger tapping clinimetric score prediction in Parkinson’ s disease using low-cost accelerometers. Comput Intell Neurosci. https://doi.org/ 10.1155/2013/717853 Švehlík M, Zwick EB, Steinwender G et al (2009) Gait analysis in patients with Parkinson’s disease off dopaminergic therapy. Arch Phys Med Rehabil 90:1880–1886. https://doi.org/10.1016/j. apmr.2009.06.017 Swie YW, Sakamoto K, Shimizu Y (2005) Chaotic analysis of electromyography signal at low back and lower limb muscles during forward bending posture. Electromyogr Clin Neurophysiol 45:329–342 Tahir NM, Manap HH (2012) Parkinson disease gait classification based on machine learning approach. J Appl Sci 12:180–185. https://doi.org/10.3923/jas.2012.180.185 Thanawattano C, Pongthornseri R, Anan C et al (2015) Temporal fluctuations of tremor signals from inertial sensor: a preliminary study in differentiating Parkinson’s disease from essential tremor. Biomed Eng Online 14:101. https://doi.org/10.1186/s12938-015-0098-1 Timmer J, Lauk M, Deuschl G (1996) Quantitative analysis of tremor time series. Electroencephalogr Clin Neurophysiol 101:461–468. https://doi.org/10.1016/0924-980X(96)94658-5 Tripoliti EE, Tzallas AT, Tsipouras MG et al (2013) Automatic detection of freezing of gait events in patients with Parkinson’s disease. Comput Methods Programs Biomed 110:12–26. https://doi. org/10.1016/j.cmpb.2012.10.016 Tzallas AT, Tsipouras MG, Rigas G et al (2014) PERFORM: a system for monitoring, assessment and management of patients with Parkinson’s disease. Sensors (Basel) 14:21329–21357. https:// doi.org/10.3390/s141121329 Visser JE, Oude Nijhuis LB, Janssen L et al (2010) Dynamic posturography in Parkinson’s disease: diagnostic utility of the “first trial effect”. Neuroscience 168:387–394. https://doi.org/10.1016/j. neuroscience.2010.03.068 Vitale C, Agosti V, Avella D et al (2012) Effect of global postural rehabilitation program on spatiotemporal gait parameters of parkinsonian patients: a three-dimensional motion analysis study. Neurol Sci 33:1337–1343. https://doi.org/10.1007/s10072-012-1202-y Wahid F, Begg RK, Hass CJ et al (2015) Classification of Parkinson’s disease gait using spatialtemporal gait features. IEEE J Biomed Health Inform 19:1794–1802. https://doi.org/10.1109/ JBHI.2015.2450232 Walshe FMR, Dejerine J, Ferrier D et al (1961) Contributions of John Hughlings Jackson to neurology. Arch Neurol 5:119–131. https://doi.org/10.1001/archneur.1961.00450140001001 Xia R, Powell D, Zev Rymer W et al (2011) Differentiation between the contributions of shortening reaction and stretch-induced inhibition to rigidity in Parkinson’s disease. Exp Brain Res 209:609–618. https://doi.org/10.1007/s00221-011-2594-2 Yokoe M, Okuno R, Hamasaki T et al (2009) Opening velocity, a novel parameter, for finger tapping test in patients with Parkinson’s disease. Parkinsonism Relat Disord 15:440–444. https://doi.org/ 10.1016/j.parkreldis.2008.11.003 Yunfeng W, Krishnan S (2010) Statistical analysis of gait rhythm in patients with Parkinson’s disease. IEEE Trans Neural Syst Rehabil Eng 18:150–158. https://doi.org/10.1109/tnsre.2009.2033062

Gait and Multiple Sclerosis James McLoughlin

Abstract

Gait analysis technology has led to an increased understanding of the changes in mobility that contribute to increasing disability in people with multiple sclerosis (MS). At all levels of the disease spectrum, changes in joint angles, moments, and muscle activation have been identified, which has lead to the investigation of key impairments that limit functional mobility in people with MS. Comprehensive three-dimensional gait analysis has characterized gait changes that worsen with increasing disability and deteriorate with walking-induced fatigue. This important ability to measure movement performance is resulting in a greater appreciation of the link between movement impairment during gait and its impact on daily functions. This knowledge will hopefully increase our ability to target specific aspects of gait with individualized physical therapy strategies. Keywords

Multiple sclerosis • Gait • Mobility limitation • Disability evaluation • Rehabilitation

Contents Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . State of the Art . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Comprehensive Gait Analysis in People with Multiple Sclerosis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Gait and MS Impairments . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Gait and Fatigue . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

1254 1254 1255 1255 1256

J. McLoughlin (*) Flinders University, Adelaide, Australia e-mail: james.mcloughlin@flinders.edu.au # Springer International Publishing AG, part of Springer Nature 2018 B. Müller, S.I. Wolf (eds.), Handbook of Human Motion, https://doi.org/10.1007/978-3-319-14418-4_65

1253

1254

J. McLoughlin

Some Targets for Therapeutic Intervention . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Cross-References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

1259 1260 1261 1262

Introduction Multiple sclerosis (MS) is a chronic degenerative condition that affects two million people worldwide (Trisolini et al. 2010) and is characterized by inflammation, demyelination, and axonal degeneration in the central nervous system. Symptoms of MS are heterogeneous and may include a number of problems associated with central nervous system dysfunction, such as optic neuritis, fatigue, weakness, spasticity, ataxia, tremor, numbness, paresthesia, bladder and bowel dysfunction, sexual dysfunction, speech difficulties, swallowing problems, and cognitive and emotional disturbances (Kurtzke 1983). People with MS often report difficulty with mobility as one of the most disabling aspects of living with the condition (Heesen et al. 2008; LaRocca 2011) with a significant impact on quality of life (Hemmett et al. 2004). Mobility-related problems in MS contribute to a considerable socioeconomic burden on society in terms of employment and healthcare utilization (Pike et al. 2012). Almost all people with MS will have some difficulty in walking due to a combination of symptoms, including poor balance (Sosnoff et al. 2011b), weakness (Yahia et al. 2011), spasticity (Sosnoff et al. 2011a; Sosnoff et al. 2010), and sensory loss (Cameron et al. 2008; Cattaneo and Jonsdottir 2009).

State of the Art Movement analysis technology has allowed both researchers and clinicians to further characterize some of the gait impairments that occur in people with MS. This information is increasing our understanding of disease impact and progression while enhancing our ability to target and measure therapeutic interventions aimed at improving safe mobility for people with MS. A number of studies have characterized the spatiotemporal differences in gait in people with MS compared to healthy controls. These studies have typically shown reductions in speed, cadence, and double limb support time in people with MS (Givon et al. 2009; Pilutti et al. 2012; Remelius et al. 2012; Sosnoff et al. 2012; Thoumie et al. 2005). Even in those people with MS who are fully ambulant with very minimal clinical disability, spatiotemporal gait changes are still present, demonstrating reductions in speed, cadence, step length, and increased step width and double limb support time (Sosnoff et al. 2012). These parameters are known to deteriorate further with higher disability levels as measured by the Expanded Disability Status Scale (EDSS) (Lizrova Preiningerova et al. 2015).

Gait and Multiple Sclerosis

1255

Comprehensive Gait Analysis in People with Multiple Sclerosis Two- and three-dimensional motion capture enables a more in-depth analysis of gait impairments and compensations that occur in people with MS and provides insight into disease progression and physical disability. Kinematic (movement) and kinetic (forces) gait parameters in people with MS typically show reductions in joint range of motion and moments (Kelleher et al. 2010a; Huisinga et al. 2012; Lizama et al. 2016). Reductions in power at the hip, knee, and ankle show a relationship with reduced levels of clinical disability as measured by the EDSS (Huisinga et al. 2012). Similarly, Kelleher et al. (2010a) also demonstrated that hip, knee, and ankle angles and propulsive force are reduced further in those people with lower ambulation levels (Kelleher et al. 2010a). Hip flexor moment is reduced as due to lower extensor and flexor powers at the end of stance and early swing phase (Huisinga et al. 2012). In early stance people with MS demonstrate a reduced knee extensor moment (Huisinga et al. 2012). This reduction in knee control has been shown to worsen with reduced mobility levels and may well be due to deterioration in a number of mechanisms, including knee extensor strength, coordination (Güner et al. 2015), and possibly hamstring spasticity limiting knee extension range in late swing phase (Kelleher et al. 2010a). Push-off power at the ankle, which provides most of the propulsive energy for walking (McGibbon 2003), has been shown to be reduced in people with MS (Huisinga et al. 2012). Reduced plantar flexion strength may be the major culprit, as reduced plantar flexion strength has been shown to be a major contributor to reduced walking capacity in people with MS (Wagner et al. 2014). Kinematic analysis of people with MS with very mild physical disability has found early reductions in ankle range of motion (Benedetti et al. 1999; Martin et al. 2006). Electromyographic (EMG) recordings also can provide useful information about individual muscle activity and timing. EMG recordings from gastrocnemius and tibialis anterior muscles during gait indicate muscle cocontraction in these muscles, leading to reduced range of motion and increased stiffness at the ankle. This is possibly a stabilization ankle strategy (Benedetti et al. 1999; Kelleher et al. 2010a). This pattern has also been detected in those people with MS without any clinical signs of pyramidal dysfunction (Martin et al. 2006).

Gait and MS Impairments Sensory tracts are often affected by demyelinating lesions in MS (Zackowski et al. 2009) resulting in reductions in important proprioceptive sensory information which is known to affect standing balance in people with MS (Cattaneo and Jonsdottir 2009; Cameron et al. 2008; Citaker et al. 2011); however, less is known on its effect on gait. Proprioceptive loss may well increase both flexor and extensor muscular compensations in order to maintain gait speed (Thoumie and Mevellec 2002; Rougier et al. 2007).

1256

J. McLoughlin

Cerebellar lesions resulting in a deterioration in coordination are also common in people with MS (Kutzelnigg et al. 2007) and appear to be a major contributor to poor balance and increasing disability (McLoughlin et al. 2015). Not surprisingly cerebellar pathology has been shown to relate to reduced balance (Prosperini et al. 2011; 2013a, b) and falls (Sosnoff et al. 2011). Cerebellar ataxia can contribute to some gait deficits such as increased step width (Givon et al. 2009). In people with MS, gait asymmetry and increased step width have been associated with increased risk of falls (Kasser et al. 2011). The overall influence of cerebellar ataxia on spatiotemporal gait deficits may be less when compared with other upper motor neuron (UMN) or pyramidal weakness (Thoumie et al. 2005; Kalron and Givon 2016). The UMN syndrome is characterized by varying degrees of velocity-dependent increases in muscle tone and weakness, both of which can alter the gait pattern observed in people with MS. Oversimplistic clinical spasticity scales used to measure increases in passive muscle tone are known to have major limitations and only show a relationship with changes in posture and balance rather than gait (Sosnoff et al. 2010, 2011). However, increases in muscle tone can restrict active movement and may well contribute to reductions in hip and knee range of motion in the gait cycle (Pau et al. 2015). Reduced strength is an important part of the UMN syndrome with weakness of the knee extensors and ankle plantar flexor linked to reduced mobility levels (Güner et al. 2015; Wagner et al. 2014). Despite a large heterogeneity and enormous mixed overlap of sensory, cerebellar, and pyramidal signs often present in people with MS, weakness stands out as a key impairment that contributes to many of the spatiotemporal, kinetic, and kinematic gait changes (Kalron and Givon 2016; Thoumie et al. 2005; Thoumie and Mevellec 2002). Overall, poor function at the ankle appears to be a major contributor to slow walking speeds in people with MS (Lizama et al. 2016) and may therefore present as a key therapeutic target for orthotic and physical therapies.

Gait and Fatigue The 1998 Multiple Sclerosis of Clinical Practice Guidelines define fatigue in MS as a “subjective lack of physical and/or mental energy that is perceived by the individual or caregiver to interfere with usual and desired activities” (for Clinical Practice Guidelines and Others 1998). People with MS often report fatigue as one of their most common and disabling symptoms (Bakshi 2003). Fatigue is known to impact heavily on employment levels (Hadjimichael et al. 2008; Smith and Arnett 2005) and is linked to reduced quality of life (Amato et al. 2001) in people with MS. Fatigue can occur in all MS subtypes, although is more pronounced in those people with a less stable natural history, accumulating impairment and reduced mobility levels (Hadjimichael et al. 2008). Some studies provide evidence for the link between motor fatigue and gait. Rate of foot taps and maximal force generation has been shown to be associated with walking speed and cadence in people with MS (Ng et al. 2004). Changes in corticospinal activity following walking-induced

Gait and Multiple Sclerosis

1257

fatigue have also been demonstrated in a small sample of people with MS; however, walking distance and fatigue levels were not recorded (Schubert et al. 1998). Mobility in MS can be objectively assessed using a number of measures including timed or distance measures of walking performance and is currently the gold standard in assessing accumulating disability levels in those people with MS with moderate to severe disease (Kurtzke 1983). Common clinical measures of mobility in people with MS include timed walks over 10 meters or 25 feet (Kieseier and Pozzilli 2012); however, more demanding tests of walking over 2 min (Gijbels et al. 2011) or 6 min (Gijbels et al. 2011; Motl et al. 2012b) may be more representative of functional mobility (Savci et al. 2005). Distance over 6 min of walking has been shown to correlate with fatigue on the physical subscale of the Modified Fatigue Impact Scale (Fisk et al. 1994). Together, mobility and fatigue have been recognized as important measures of functional capacity (Hutchinson et al. 2009) and major contributors to the physical impact of MS (Kehoe et al. 2014). In addition, there is potential in utilizing measures of mobility to also objectively measure motor fatigue (Burschka et al. 2012; Dobkin 2008; Phan-Ba et al. 2012). Walking measures that increase the energy expenditure are important when considering that energy cost with walking correlates with walking disability (Motl et al. 2010) and is associated with gait impairments, daily activity, and fatigue (Motl et al. 2012a). Preliminary evidence suggests that fatigue is associated with kinetic gait changes in people with MS, with reduced ankle power at push-off, reduced knee power absorption, and reduced peak knee extensor moments shown to be associated with increased levels of self-reported fatigue (Huisinga et al. 2011). One study of fourteen people with MS compared walking in the morning versus the afternoon when fatigue levels were significantly higher, yet found no change in spatiotemporal gait parameters (Morris et al. 2002). The six-minute walk test (6MWT) is a common mobility-related assessment that has been used as a measure of functional ambulation capacity (Savci et al. 2005). Recent research indicates that while the 6MWT can measure both walking and fitness in people with MS, it relates more strongly to walking performance rather than aerobic or muscular fitness (Sandroff et al. 2015). Performance of the 6MWT has been shown to produce marked increased levels in perceived fatigue (Barr et al. 2014). People with MS demonstrate reduced distance with the 6MWT, compared to healthy controls, which correlates with total and physical aspects of fatigue as measured by the Modified Fatigue Impact Scale (MFIS) (Goldman et al. 2008). 6MWT distance has also been associated with limitations in activities of daily living in people with MS (Karpatkin and Rzetelny 2012), which suggests that 6MWT performance may be an accurate measure of the capacity to undertake “real-world” daily activities. While there is no standardized method of measuring the 6MWT in people with MS, 6MWT protocols involving more frequent 180-degree turns have been shown to be more physically demanding when compared to square walkway protocols (Sandroff et al. 2014). It has become clear that both mobility and fatigue in MS are inextricably linked and should therefore be considered together. In 2007, a multidisciplinary consensus conference with the Consortium of MS Centres in the USA recommended that rehabilitation professionals need to improve their

1258

J. McLoughlin

understanding of gait and fatigue outcome measures used in MS rehabilitation (Hutchinson et al. 2009). Therefore, the investigation of activity-dependent fatigability in lower limb muscles following the 6MWT in people with MS remains an important research and clinical question. Further evidence suggests that following a period of walking, detrimental effects on spatiotemporal gait parameters are evident in people with MS with slow walking speed (below 0.82 m/s) (Feys et al. 2013). Only a few pilot studies have investigated the direct effect of walking-induced fatigue on kinematic and kinetic parameters of gait. A pilot study of walking at preferred walking speed to the point of exhaustion on a treadmill reported some increases in kinematic variability in selected patients (Sehle et al. 2014), whereas another study measuring gait variability found no kinematic or kinetic variability following a fatiguing walk in study of 20 people with MS (Crenshaw et al. 2006). For both of these studies, walking distance was not reported. McLoughlin et al. was the first controlled study to identify a number of fatigue-induced changes to kinematics and kinetic gait mechanics in a group of people with MS with mild to moderate disability (McLoughlin et al. 2016). Compared with healthy controls, the MS group had significantly reduced total 6MWT distance with a trend toward slowing down over the last 3 min. This reduced speed in the second half of the test supports the results of other studies that have found reduced speed in the last 1–5 min of the 6MWT (Dalgas et al. 2014; Gijbels et al. 2011). This gait slowing appears to coincide with steady-state aerobic metabolism, as Motl et al. (2012b) demonstrated in physiological measurements during the final 3 min of the 6MWT (Motl et al. 2012b). These findings suggest that a “deceleration index” may be a useful measure of walking capacity in people with MS (Phan-Ba et al. 2012). Another study showed that “decline” in walking speed, rather than “mean” walking speed over the 6MWT, had a stronger association with subjective fatigue (Burschka et al. 2012). While there has been some support for a shortened two-minute walk test to asses walking ability in MS (Gijbels et al. 2011), gait studies suggest a shortening of the 6MWT may result in the loss of important clinical information when assessing functional mobility and the relevant changes caused by walking-induced fatigue (McLoughlin et al. 2016). The finding of no change in spatiotemporal parameters following the 6MWT is consistent with a previous study, which only found reduced cadence following a 6MWT in a less able group of people with MS (Feys et al. 2013). It appears that moderately disabled people with MS are able to maintain their pre-walk spatiotemporal characteristics following the 6MWT, but kinematic and kinetic gait analysis reveals changes that are likely to be movement compensations in response to the walking-induced fatigue (McLoughlin et al. 2016). In particular, joint kinetic analyses show increases in moments and power absorption at the hip, knee, and ankle during stance phase – signs of fatigue-induced weakness (McLoughlin et al. 2014b) which are likely to further increase the energy cost of walking (Kuo et al. 2005). Increases in hip extensor moment in the less affected side suggests some capacity to increase the positive compensatory output to try and maintain speed, while an increase in hip flexor moment in the more affected side may be a compensation to assist toe clearance with increased fatigue (McLoughlin et al. 2016). People with MS are likely to increase the underlying

Gait and Multiple Sclerosis

1259

central corticomotor activity needed to produce these positive changes in joint kinetics, which could contribute further to perceived fatigue levels (Morgante et al. 2011; Thickbroom et al. 2006, 2008). A deeper understanding of the relationship between fatigue and mobility may provide opportunities to develop treatmentspecific strategies for both fatigue and mobility in people with MS.

Some Targets for Therapeutic Intervention One potential target for therapeutic intervention to improve walking capacity in people with MS is assisting ankle dorsiflexion during gait. Neurophysiological studies have shown that ankle dorsiflexor muscles are particularly susceptible to motor fatigue (Ng et al. 2004; Thickbroom et al. 2008) and may alter gait in people with MS (Martin et al. 2006; Benedetti et al. 1999). Kinematic analysis has demonstrated reduced dorsiflexion in the weaker leg at initial contact phase of gait following a fatiguing six-minute walk (McLoughlin et al. 2016). The consequences of reduced dorsiflexion can be serious as difficulties in toe clearance in the swing phase of gait can lead to trips, which are reported as the most common reasons for falling by people with MS (Matsuda et al. 2011). Ankle dorsiflexion can be assisted with commonly prescribed ankle-foot orthosis (AFOs) that comprise rigid or nonrigid articulated braces that support the foot from underneath to prevent foot drop. To date there is no evidence that AFOs can improve mobility in people with MS (Sheffler et al. 2008). A review of the effect of these devices found that rigid AFOs can compromise balance, while more flexible, nonrigid AFOs can facilitate dynamic balance (Ramstrand and Ramstrand 2010). In a crossover study involving people with MS, rigid and nonrigid AFOs improved static balance, but rigid AFOs impaired dynamic balance when walking (Cattaneo et al. 2002).One of the limitations of AFOs is that they limit not only the ankle and knee range of motion, but reduce the proprioceptive input via the sole of the foot, which may further exacerbate preexisting sensory deficits that limit balance ability (Cameron et al. 2008; Citaker et al. 2011). Another option to assist ankle dorsiflexion is the use of functional electrical stimulation (FES) to the pretibial muscles, which has been shown to assist ankle dorsiflexion during gait in people with MS (Scott et al. 2013). This form of FES has been shown to have an acute effect of speed and reduced physiological cost of walking during a five-minute walk around a 10-meter elliptical course (Paul et al. 2008). A small study, including only two people with MS with unilateral foot drop who used FES for 10 weeks, found improvements in walking speed and reduced physiological cost of walking over a 10-meter course (Taylor et al. 1999). Scheffler et al. found some improvement with stair ascent and descent in people with MS, with 9 out of the 11 participants preferring FES over an AFO and all preferring FES over no device (Sheffler et al. 2009). A more recent study examined the training effect of 12 weeks of FES use in nine people with MS and found improvement in ankle dorsiflexion in gait, increased speed over 10-meter walk test, and increased speed and reduced perceived exertion over the two-minute walk test

1260

J. McLoughlin

(van der Linden et al. 2014). This suggests that assisting ankle dorsiflexion may have benefits in terms of walking speed and reducing physiological cost of walking in people with MS. Another benefit of FES to the ankle dorsiflexors may be the potential to “retrain” the gait pattern via cortical neuroplasticity motor learning (Everaert et al. 2010). Prolonged use of FES for 8 weeks has shown some training effect to ankle and knee kinetics in stance phase in people with MS (Barr et al. 2016), which suggests that FES has potential as an adjunct to gait retraining therapy. From a practical viewpoint, however, FES devices are expensive and not easily accessible at present, which remains a significant barrier to their use for people with MS worldwide. An alternative (and perhaps more cost-effective) method of addressing the issues of restricted dynamic balance and physiological cost of walking may involve assisting ankle dorsiflexion with a dorsiflexion assist orthosis (DAO) (McLoughlin et al. 2014). The DAO dynamically assists ankle dorsiflexion with an elastic strap attached to the shoe that can assist pulling the foot toward dorsiflexion. This is a very simple and inexpensive device compared with custom-made AFOs and FES devices, and unlike AFOs the DAO does not interfere with the sensory interaction of the foot with the sole of the shoe. Maintaining and even enhancing the sensory feedback within the shoe remains an attractive target in people with MS, with textured (Dixon et al. 2014; Kelleher et al. 2010b) or contoured insoles (Ramdharry et al. 2006) showing promise for improving mobility and balance. The DAO is therefore an attractive option to assist dorsiflexion without interfering with sensory feedback via the sole of the foot. To date, this type of dorsiflexion assist orthosis (DAO) has not been evaluated with prolonged use in people with MS. When considering the issue of motor fatigue with MS, the known detrimental effects of AFOs on standing balance and the improved physiological cost of walking with the use of FES devices, both the DAO and FES, should be investigated in terms of reducing the physiological cost of walking and mitigating any effects of walking-induced fatigue on balance and lower limb muscle strength with prolonged walking. These devices may compliment individualized neurological physiotherapy focusing on ankle control and mobility, which has shown promise in improving gait and balance in MS (Davies et al. 2015).

Conclusion Gait analysis has provided useful information about the many sensorimotor impairments that contribute to declining mobility in people with multiple sclerosis. Spatiotemporal, kinematic, and kinetic measures have been able to capture the changes associated with accumulating disability over time. This type of research has identified useful therapeutic targets that have the potential to improve walking pattern and efficiency. It is hoped that this will lead to more meaningful improvements in functional mobility and associated fatigue-induced impairments. Future gait analysis will hopefully allow for more relevant subgrouping of gait presentations in MS and lead to specific interventions that are tailor-made for each individual.

Gait and Multiple Sclerosis

1261

Cross-References ▶ Assessing Clubfoot and Cerebral Palsy by Pedobarography ▶ Assessing the Impact of Aerobic Fitness on Gait ▶ Brain-Computer Interfaces for Motor Rehabilitation ▶ Clinical Gait Assessment by Video Observation and 2D Techniques ▶ Concussion Assessment During Gait ▶ Detecting and Measuring Ataxia in Gait ▶ Diagnostic Gait Analysis Use in the Treatment Protocol for Cerebral Palsy ▶ Effects of Knee Osteoarthritis and Joint Replacement Surgery on Gait ▶ Effects of Total Hip Arthroplasty on Gait ▶ EMG Activity in Gait: The Influence of Motor Disorders ▶ Foot and Ankle Motion in Cerebral Palsy ▶ Functional Capacity Evaluation and Quantitative Gait Analysis: Lower Limb Disorders ▶ Functional Dystonias ▶ Functional Effects of Ankle Sprain ▶ Functional Effects of Foot Orthoses ▶ Functional Effects of Shoes ▶ Gait and Multiple Sclerosis ▶ Gait Changes in Skeletal Dysplasia ▶ Gait Disorders in Persons After Stroke ▶ Gait During Real-World Challenges: Gait Initiation, Gait Termination, Acceleration, Deceleration, Turning, Slopes, and Stairs ▶ Gait Rehabilitation with Exoskeletons ▶ Gait Retraining for Balance Improvement ▶ Gait Scores: Interpretations and Limitations ▶ Hereditary Motor Sensory Neuropathy: Understanding Function Using Motion Analysis ▶ Idiopathic Toe Walking ▶ Impact of Scoliosis on Gait ▶ Influence of Prosthetic Socket Design and Fitting on Gait ▶ Integration of Foot Pressure and Foot Kinematics Measurements for Medical Applications ▶ Interpreting Ground Reaction Forces in Gait ▶ Interpreting Joint Moments and Powers in Gait ▶ Interpreting Spatiotemporal Parameters, Symmetry, and Variability in Clinical Gait Analysis ▶ Kinematic Foot Models for Instrumented Gait Analysis ▶ Measures to Determine Dynamic Balance ▶ Motor Patterns Recognition in Parkinson’s Disease ▶ Natural History of Cerebral Palsy and Outcome Assessment ▶ Next-Generation Models Using Optimized Joint Center Location ▶ Oxygen Consumption in Cerebral Palsy ▶ Prosthetic Foot Principles and Their Influence on Gait

1262

J. McLoughlin

▶ Skeletal Muscle Structure in Spastic Cerebral Palsy ▶ Slip and Fall Risk Assessment ▶ Slips, Trips, and Falls ▶ Spasticity Effect in Cerebral Palsy Gait ▶ Strength Related Stance Phase Problems in Cerebral Palsy ▶ Swing Phase Problems in Cerebral Palsy ▶ The Conventional Gait Model - Success and Limitations ▶ The Effects of Ankle Joint Replacement on Gait ▶ The Importance of Foot Pressure in Diabetes ▶ The Influence of Prosthetic Knee Joints on Gait ▶ Trunk and Spine Models for Instrumented Gait Analysis ▶ Upper Extremity Models for Clinical Movement Analysis ▶ Upper Extremity Movement Pathology in Functional Tasks ▶ Variations of Marker Sets and Models for Standard Gait Analysis ▶ Walking and Physical Activity Monitoring in Children with Cerebral Palsy

References Amato MP et al (2001) Quality of life in multiple sclerosis: the impact of depression, fatigue and disability. Mult Scler 7:340–344 Bakshi R (2003) Fatigue associated with multiple sclerosis: diagnosis, impact and management. Mult Scler 9:219–227 Barr C et al (2014) Walking for six minutes increases both simple reaction time and stepping reaction time in moderately disabled people with Multiple Sclerosis. Mult Scler Relat Disord 3(4):457–462 Barr CJ et al (2016) Orthotic and therapeutic effect of functional electrical stimulation on fatigue induced gait patterns in people with multiple sclerosis. Disabil Rehabil 0(0):1–13 Benedetti MG et al (1999) Gait abnormalities in minimally impaired multiple sclerosis patients. Mult Scler 5:363–368 Burschka JM et al (2012) An exploration of impaired walking dynamics and fatigue in multiple sclerosis. BMC Neurol 12:161 Cameron MH et al (2008) Imbalance in multiple sclerosis: a result of slowed spinal somatosensory conduction. Somatosens Mot Res 25:113–122 Cattaneo D, Jonsdottir J (2009) Sensory impairments in quiet standing in subjects with multiple sclerosis. Mult Scler 15:59–67 Cattaneo D et al (2002) Do static or dynamic AFOs improve balance? Clin Rehabil 16:894–899 Citaker S et al (2011) Relationship between foot sensation and standing balance in patients with multiple sclerosis. Gait Posture 34:275–278 Crenshaw SJ et al (2006) Gait variability in people with multiple sclerosis. Mult Scler 12:613–619 Dalgas U et al (2014) Aerobic intensity and pacing pattern during the six-minute walk test in patients with multiple sclerosis. J Rehabil Med 46:59–66 Davies BL et al (2015) Neurorehabilitation strategies focusing on ankle control improve mobility and posture in persons with multiple sclerosis. J Neurol Phys Ther 39(4):225–232 Dixon J et al (2014) Effect of textured insoles on balance and gait in people with multiple sclerosis: an exploratory trial. Physiotherapy 100(2):142–149 Dobkin BH (2008) Fatigue versus activity-dependent fatigability in patients with central or peripheral motor impairments. Neurorehabil Neural Repair 22:105–110 Everaert DG et al (2010) Does functional electrical stimulation for foot drop strengthen corticospinal connections? Neurorehabil Neural Repair 24:168–177

Gait and Multiple Sclerosis

1263

Feys P et al (2013) Spatio-temporal gait parameters change differently according to speed instructions and walking history in MS patients with different ambulatory dysfunction. Mult Scler Relat Disord 2(3):238–246. https://doi.org/10.1016/j.msard.2013.01.004 Fisk JD et al (1994) Measuring the functional impact of fatigue: initial validation of the fatigue impact scale. Clinical Infect Dis 18(Suppl 1):S79–S83 For Clinical Practice Guidelines, M.S.C. & Others (1998) Fatigue and multiple sclerosis: evidencebased management strategies for fatigue in multiple sclerosis: clinical practice guidelines. The Council. The Consortium of Multiple Sclerosis Centers, in conjunction with the Paralyzed Veterans of America (PVA). http://www.mscare.org/?page=practice_guidelines Gijbels D, Eijnde B, Feys P (2011) Comparison of the 2- and 6-minute walk test in multiple sclerosis. Mult Scler 17:1269–1272 Givon U, Zeilig G, Achiron A (2009) Gait analysis in multiple sclerosis: characterization of temporal-spatial parameters using GAITRite functional ambulation system. Gait Posture 29:138–142 Goldman MD, Marrie RA, Cohen JA (2008) Evaluation of the six-minute walk in multiple sclerosis subjects and healthy controls. Mult Scler 14:383–390 Güner S et al (2015) Knee muscle strength in multiple sclerosis: relationship with gait characteristics. J Phys Ther Sci 27(3):809–813 Hadjimichael O, Vollmer T, Oleen-Burkey M (2008) Fatigue characteristics in multiple sclerosis: the North American Research Committee on Multiple Sclerosis (NARCOMS) survey. Health Qual Life Outcomes 6:100 Heesen C et al (2008) Patient perception of bodily functions in multiple sclerosis: gait and visual function are the most valuable. Mult Scler 14:988–991 Hemmett L et al (2004) What drives quality of life in multiple sclerosis? QJM 97:671–676 Huisinga JM et al (2011) Is there a relationship between fatigue questionnaires and gait mechanics in persons with multiple sclerosis? Arch Phys Med Rehabil 92:1594–1601 Huisinga JM et al (2012) Gait mechanics are different between healthy controls and patients with multiple. J Appl Biomech 2012 Human Kinetics, p 2 Hutchinson B et al (2009) Toward a consensus on rehabilitation outcomes in MS: gait and fatigue. Int J MS Care 11:67–78 Kalron A, Givon U (2016) Gait characteristics according to pyramidal, sensory and cerebellar EDSS subcategories in people with multiple sclerosis. J Neurol. Available at: http://dx.doi.org/ 10.1007/s00415-016-8200-6 Karpatkin H, Rzetelny A (2012) Effect of a single bout of intermittent versus continuous walking on perceptions of fatigue in people with multiple sclerosis. Int J Care 14:124–131 Kasser SL et al (2011) A prospective evaluation of balance, gait, and strength to predict falling in women with multiple sclerosis. Arch Phys Med Rehabil. Available at: http://www.ncbi.nlm.nih. gov/pubmed/21840497 Kehoe M et al (2014) Predictors of the physical impact of multiple sclerosis following communitybased, exercise trial. Multiple Scler 10:1352458514549395 Kelleher KJ, Spence W et al (2010a) The characterisation of gait patterns of people with multiple sclerosis. Disabil Rehabil 32:1242–1250 Kelleher KJ, Spence WD et al (2010b) The effect of textured insoles on gait patterns of people with multiple sclerosis. Gait Posture 32:67–71 Kieseier BC, Pozzilli C (2012) Assessing walking disability in multiple sclerosis. Mult Scler J 18:914–924 Kuo AD, Donelan JM, Ruina A (2005) Energetic consequences of walking like an inverted pendulum: step-to-step transitions. Exerc Sport Sci Rev 33:88–97 Kurtzke JF (1983) Rating neurologic impairment in multiple sclerosis: an expanded disability status scale (EDSS). Neurology 33:1444–1452 Kutzelnigg A et al (2007) Widespread demyelination in the cerebellar cortex in multiple sclerosis. Brain Pathol 17:38–44 LaRocca NG (2011) Impact of walking impairment in multiple sclerosis. Patient 4:189–201

1264

J. McLoughlin

Lizama LEC et al (2016) The use of laboratory gait analysis for understanding gait deterioration in people with multiple sclerosis. Mult Scler J 22(14):1768–1776 Lizrova Preiningerova J et al (2015) Spatial and temporal characteristics of gait as outcome measures in multiple sclerosis (EDSS 0 to 6.5). J Neuroeng Rehabil 12(1):1–7 Martin CL et al (2006) Gait and balance impairment in early multiple sclerosis in the absence of clinical disability. Mult Scler 12:620–628 Matsuda PN et al (2011) Falls in multiple sclerosis. PM & R J Inj funct Rehabil 3:624–632, quiz 632 McGibbon CA (2003) Toward a better understanding of gait changes with age and disablement: neuromuscular adaptation. Exerc Sport Sci Rev 31(2):102–108 McLoughlin JV et al (2014) Dorsiflexion assist orthosis reduces the physiological cost and mitigates deterioration in strength and balance associated with walking in people with multiple sclerosis. Archives of physical medicine and rehabilitation. Available at: http://www.sciencedirect.com/ science/article/pii/S0003999314010703 McLoughlin JV et al (2014b) Six minutes of walking leads to reduced lower limb strength and increased postural sway in people with multiple sclerosis. NeuroRehabilitation 35(3):503–508 McLoughlin J et al (2015) Association of postural sway with disability status and cerebellar dysfunction in people with multiple sclerosis: a preliminary study. Int J MS Care 17(3):146–151 McLoughlin JV et al (2016) Fatigue induced changes to kinematic and kinetic gait parameters following six minutes of walking in people with multiple sclerosis. Disabil Rehabil 38(6):535–543 Morgante F et al (2011) Is central fatigue in multiple sclerosis a disorder of movement preparation? J Neurol 258:263–272 Morris ME et al (2002) Changes in gait and fatigue from morning to afternoon in people with multiple sclerosis. J Neurol Neurosurg Psychiatry 72:361–365 Motl RW et al (2010) Multiple sclerosis walking Scale-12 and oxygen cost of walking. Gait Posture 31:506–510 Motl RW, Sandroff BM et al (2012a) Energy cost of walking and its association with gait parameters, daily activity, and fatigue in persons with mild multiple sclerosis. Neurorehabil Neural Repair 26:1015–1021 Motl RW, Suh Y et al (2012b) Evidence for the different physiological significance of the 6-and 2-minute walk tests in multiple sclerosis. BMC Neurol 12:6 Ng AV et al (2004) Functional relationships of central and peripheral muscle alterations in multiple sclerosis. Muscle Nerve 29:843–852 Pau M et al (2015) Effect of spasticity on kinematics of gait and muscular activation in people with multiple sclerosis. J Neurol Sci 358(1–2):339–344 Paul L et al (2008) The effect of functional electrical stimulation on the physiological cost of gait in people with multiple sclerosis. Mult Scler 14:954–961 Phan-Ba R et al (2012) Motor fatigue measurement by distance-induced slow down of walking speed in multiple sclerosis. PloS One 7:e34744 Pike J et al (2012) Social and economic burden of walking and mobility problems in multiple sclerosis. BMC Neurol 12:94 Pilutti LA et al (2012) Further validation of multiple sclerosis walking scale-12 scores based on spatiotemporal gait parameters. Arch Phys Med Rehabil Prosperini L et al (2011) The relationship between infratentorial lesions, balance deficit and accidental falls in multiple sclerosis. J Neurol Sci 304:55–60 Prosperini L, Petsas N et al (2013a) Balance deficit with opened or closed eyes reveals involvement of different structures of the central nervous system in multiple sclerosis. Mult Scler J 20(1):81–90. https://doi.org/10.1177/1352458513490546 Prosperini L, Sbardella E et al (2013a) Multiple sclerosis: white and gray matter damage associated with balance deficit detected at static posturography. Radiology 268:181–189 Ramdharry GM et al (2006) De-stabilizing and training effects of foot orthoses in multiple sclerosis. Mult Scler 12:219–226 Ramstrand N, Ramstrand S (2010) AAOP state-of-the-science evidence report: the effect of anklefoot orthoses on balance—a systematic review. J Prosthet Orthot 22:P4–P23

Gait and Multiple Sclerosis

1265

Remelius JG et al (2012) Gait impairments in persons with multiple sclerosis across preferred and fixed walking speeds. Arch Phys Med Rehabil 93(9):1637–1642. https://doi.org/10.1016/j.apmr.2012.02.019 Rougier P et al (2007) What compensatory motor strategies do patients with multiple sclerosis develop for balance control? Rev Neurol 163:1054–1064 Sandroff BM et al (2014) Comparing two conditions of administering the six-minute walk test in people with multiple sclerosis. Int J MS Care 16:48–54 Sandroff BM, Pilutti LA, Motl RW (2015) Does the six-minute walk test measure walking performance or physical fitness in persons with multiple sclerosis? NeuroRehabilitation 37(1):149–155 Savci S et al (2005) Six-minute walk distance as a measure of functional exercise capacity in multiple sclerosis. Disabil Rehabil 27:1365–1371 Schubert M et al (1998) Walking and fatigue in multiple sclerosis: the role of the corticospinal system. Muscle Nerve 21:1068–1070 Scott SM et al (2013) Quantification of gait kinematics and walking ability of people with multiple sclerosis who are new users of functional electrical stimulation. J Rehabil Med 45:364–369 Sehle A et al (2014) Objective assessment of motor fatigue in multiple sclerosis: the Fatigue index Kliniken Schmieder (FKS). J Neurol 261(9):1752–1762 Sheffler LR et al (2008) Functional effect of an ankle foot orthosis on gait in multiple sclerosis: a pilot study. Am J Phys Med Rehabil 87:26 Sheffler LR et al (2009) Neuroprosthetic effect of peroneal nerve stimulation in multiple sclerosis: a preliminary study. Arch Phys Med Rehabil 90:362–365 Smith MM, Arnett PA (2005) Factors related to employment status changes in individuals with multiple sclerosis. Mult Scler 11:602–609 Sosnoff JJ, Shin S, Motl RW (2010) Multiple sclerosis and postural control: the role of spasticity. Arch Phys Med Rehabil 91:93–99 Sosnoff JJ et al (2011a) Influence of spasticity on mobility and balance in persons with multiple sclerosis. J Neurol Phys Ther 35:129–132 Sosnoff JJ et al (2011b) Mobility, balance and falls in persons with multiple sclerosis. PloS One 6: e28021 Sosnoff JJ, Sandroff BM, Motl RW (2012) Quantifying gait abnormalities in persons with multiple sclerosis with minimal disability. Gait Posture 36(1):154–156. https://doi.org/10.1016/j.gaitpost. 2011.11.027 Taylor PN et al (1999) Clinical use of the Odstock dropped foot stimulator: its effect on the speed and effort of walking. Arch Phys Med Rehabil 80:1577–1583 Thickbroom GW et al (2006) Central motor drive and perception of effort during fatigue in multiple sclerosis. J Neurol 253:1048–1053 Thickbroom GW et al (2008) Enhanced corticomotor excitability with dynamic fatiguing exercise of the lower limb in multiple sclerosis. J Neurol 255:1001–1005 Thoumie P, Mevellec E (2002) Relation between walking speed and muscle strength is affected by somatosensory loss in multiple sclerosis. J Neurol Neurosurg Psychiatry 73:313–315 Thoumie P et al (2005) Motor determinants of gait in 100 ambulatory patients with multiple sclerosis. Mult Scler 11:485–491 Trisolini M et al (2010) Global economic impact of multiple sclerosis. Multiple Sclerosis International Federation London, United Kingdom [Online] van der Linden ML et al (2014) Habitual functional electrical stimulation therapy improves gait kinematics and walking performance, but not patient-reported functional outcomes, of people with multiple sclerosis who present with foot-drop. PloS One 9:e103368 Wagner JM et al (2014) Plantarflexor weakness negatively impacts walking in persons with multiple sclerosis more than plantarflexor spasticity. Arch Phys Med Rehabil 95(7):1358–1365. https:// doi.org/10.1016/j.apmr.2014.01.030 Yahia A et al (2011) Relationship between muscular strength, gait and postural parameters in multiple sclerosis. Ann Phys Rehabil Med 54:144–155 Zackowski KM et al (2009) Sensorimotor dysfunction in multiple sclerosis and column-specific magnetization transfer-imaging abnormalities in the spinal cord. Brain J Neurol 132:1200–1209

Functional Dystonias Jessica Pruente and Deborah Gaebler-Spira

Abstract

Dystonia is a movement disorder characterized by involuntary muscle contractions resulting in twisting movements and abnormal postures. This movement disorder can cause significant impairments during functional tasks including gait, mobility, and reaching. Dystonia must be distinguished from other hypertonic movement disorders, spasticity, and rigidity, to guide treatment and management options. Several clinical measurement scales have been developed to identify dystonia and rate its severity; these can be easily adapted for use in motion analysis labs. Additionally, the use of motion analysis kinetics, kinematics, and surface EMG has increased in use for monitoring dystonia. This chapter will discuss the common etiologies of dystonia, clinical scales used for diagnosis and efficacy of treatments, and the role of instrumented gait analysis, kinetics, and kinematics in the evaluation of dystonia. Keywords

Barry-Albright Dystonia Scale • Co-contraction • Dystonia • Gait analysis • Hyperkinetic • Hypertonia asessment tool • Involuntary movements • Kinematics • Motion analysis • Movement disorder • Overflow muscle activation • Spasms

Electronic supplementary material: Supplementary material is available in the online version of this chapter at https://doi.org/10.1007/978-3-319-14418-4_70. J. Pruente (*) • D. Gaebler-Spira Shirley Ryan Ability Lab, Chicago, IL, USA e-mail: [email protected]; [email protected] # Springer International Publishing AG, part of Springer Nature 2018 B. Müller, S.I. Wolf (eds.), Handbook of Human Motion, https://doi.org/10.1007/978-3-319-14418-4_70

1267

1268

J. Pruente and D. Gaebler-Spira

Contents Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . State of the Art . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Pathology and Functional Effects of Dystonia . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Definitions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Etiologies . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Clinical Scales . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Kinetics/Kinematics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Summarizing . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

1268 1268 1269 1269 1270 1271 1272 1277 1277 1278

Introduction Dystonia is a movement disorder characterized by involuntary muscle contractions resulting in twisting movements and abnormal postures. Dystonia is often initiated or worsened by volitional activity and is associated with overflow muscle activation. Dystonia is highly variable but can impair a person’s function in walking, hand manipulation, and speech due to reduced quality and speed of voluntary movement (van der Kamp et al. 1989; Agostino et al. 1992; Inzelberg et al. 1995; Curra et al. 2000; Gregori et al. 2008). Utilization of motion analysis for gait or arm function is particularly helpful because dystonia is characterized as a movement disorder and also impacts muscle tone (Sanger 2015). The motion analysis laboratory captures data from various perspectives; this fosters integration of a methodical clinical examination, video review of gait, kinetics and kinematics, motion trajectories, muscle activation patterns, and co-contractions. In this chapter, we will review the definition of dystonia, the clinical tests discriminating dystonia from spasticity, severity rating scales for dystonia, etiologies, the concepts of co-contraction and dynamic motor control, as well as the kinetic and kinematic characteristics of functional dystonias. Though dystonia can be easily and commonly recognized during the review of motion analysis, little literature exists to provide a conclusive approach to reporting dystonia during motion analysis. Recommendations for inclusion of clinical scales, close observation of the video, as well as known gait characteristics of dystonia will be discussed for consideration. Motion analysis provides a unique quantifiable understanding of both motor control and the impact of dystonia on function.

State of the Art Information obtained through motion analysis provides insights into the pathology and functional effects of dystonia. In children with cerebral palsy, the use of motion analysis is frequent for planning interventions based on gait deviations or functional improvements for hand use. Though no published motion analysis criteria for dystonia exists, common findings reported between children and adults include a

Functional Dystonias

1269

high variability in step length and base of support. Surface electromyography data likewise suggests diagnosis of dystonia through co-contraction, overflow muscle activity, and increased muscle activity during volitional tasks. Further advancements in the use of formal motion analysis are needed to improve confidence in diagnosis and treatment, either surgical or medical, for dystonia.

Pathology and Functional Effects of Dystonia Definitions Dystonic syndromes are some of the more commonly observed movement disorders with an estimated prevalence of 2–50 per million in early-onset dystonia and 30–7,320 per million in late-onset dystonias (Carecchio et al. 2015). Secondary dystonia has been increasingly recognized in children with cerebral palsy. Since proper treatment of dystonia is available, it is crucial to be able to identify the movement disorder and the confounding hypertonic muscle abnormalities frequently coexisting in children with cerebral palsy. Dystonia is defined as a movement disorder in which involuntary sustained or intermittent muscle contractions cause twisting and repetitive movements, abnormal postures, or both. Dystonia falls into the category of hypertonic and hyperkinetic movement disorders, see Fig. 1. It is frequently worsened with voluntary activity and

Hypertonic

Hyperkinetic

Negative

Spasticity

Chorea

Weakness

Dystonia

Dystonia

Selective Motor Control

Rigidity

Athetosis

Ataxia

Myoclonus

Dyspraxia

Tremor Bradykinesia Tics Balance Stereotyples

Fig. 1 Positive and negative symptoms of hypertonia. Dystonia is associated with both hypertonic and hyperkinetic movements

1270

J. Pruente and D. Gaebler-Spira

may be reduced or absent when at rest. Various organization taxonomies have been proposed to categorize movement disorders. One such schema breaks disorders into those of tone, inhibition, execution, and planning of movements (Sanger 2003). Dystonia is therefore both a disorder of tone and of execution. Dystonia can have many clinical expressions; this includes dystonic spasms, tremor, repetitive movements, abnormal fixed postures, and hypertonia. An important distinction is delineating the difference between dystonia, spasticity, and rigidity. Spasticity refers to hypertonia in which resistance to movement increases with speed of stretch and varies with direction of movement. Rigidity is resistance to movement that does not depend upon movement speed or direction. Dystonia is more complex to define and includes resistance to joint movements that do not depend on speed, co-contraction of agonist, and antagonist muscle groups and is worsened by voluntary activity. Overflow movements may also suggest the presence of dystonia. Dystonia quite often presents in similar patterns despite a wide range of diagnoses and etiologies. One example is neck or back extension, variable scoliosis or kyphosis, ulnar wrist deviation and flexion, and finger flexion or extension (Sanger 2004). In the lower extremity, dystonic posturing often includes knee extension, plantarflexion, and inversion of the foot. Common dystonic syndromes include hand cramps, blepharospasm, torticollis, opisthotonus, and more generalized dystonias involving multiple extremities and the torso. Treatment and management of these various hyperkinetic and hypertonic movement disorders is very much dependent upon specific type. Certain medications, such as trihexyphenidyl, and surgical procedures, such as deep brain stimulation, have better success rates with dystonia.

Etiologies Early-onset dystonias refer to presentation prior to age 26. The dystonic movement disorders can be further divided into subsets of inherited and acquired dystonias. DYT1, also known as Oppenheim’s dystonia, is the most common of the inherited early-onset dystonias, with a worldwide frequency of 1:160,000 cases (Carecchio et al. 2015; de Carvalho Aguiar and Ozelius 2002). A GAG deletion in the TOR1A gene was identified as the cause of DYT1 dystonia. DYT1 is an autosomal dominant trait with reduced penetrance of about 30%. Clinical presentation typically occurs by age 12 with involvement of a single extremity; this generalizes to the remainder of extremities and the trunk in about 50% of cases within a few years. Treatment remains symptomatic with oral medications and in select refractory cases with deep brain stimulation. DYT 2 is an autosomal recessive trait of unknown genetic etiology. It is an early-onset dystonia; DYT6 dystonia is also an autosomal dominant inherited form of dystonia. This involves a variety of different mutations in the THAP1 gene. DYT6 dystonia also presents in childhood through adolescence. Clinical presentation typically involves oromandibular, craniocervical, or laryngeal dystonias.

Functional Dystonias

1271

The second major subset of dystonias are secondary or acquired dystonias. The most common etiology is related to acquired brain injuries. Dystonia is second only to post-traumatic tremor in the movement disorder sequelae of severe traumatic brain injury (Krauss and Jankovic 2002; Sanger 2015). It is thought to be related to involvement of the basal ganglia, caudate, and putamen, and there is some evidence to suggest cerebellar and thalamic involvement (Skogseid 2014). Traumatic and hypoxic brain injuries can lead to acquired dystonias in adults. The PAID syndrome, or paroxysmal autonomic instability and dystonia syndrome, occurs after hypoxic injuries, such as those sustained in cardiac arrest. In children, the most common cause of acquired dystonia is cerebral palsy, and in fact, dyskinetic cerebral palsy is the second largest CP type (Monbaliu et al. 2016). Dystonia may be related to hypoxic ischemic injuries or prematurity in these cases. Other etiologies include autoimmune disorders such as anti-N-methyl-D-aspartate receptor (NMDAR) encephalitis and autoimmune basal ganglia encephalitis (van Egmond et al. 2015). Dystonia can also be induced by a variety of drugs and toxins including levodopa, dopamine antagonists, selective serotonin reuptake inhibitors, buspirone, cocaine, monoamine oxidase inhibitors, carbon monoxide, manganese, and cyanide (Phukan et al. 2011). Inborn errors of metabolism may also cause dystonic movement disorders; if not treated early, this can be found in organic acidurias, glut-1 deficiency, and lysosomal storage disorders. A third category of dystonia refers to dystonia-plus syndromes. Dystonia-plus syndromes are dystonic syndromes that also include other neurological complaints most commonly parkinsonism or myoclonus. Examples of this include doparesponsive dystonia, rapid-onset dystonia parkinsonism, and myoclonus dystonia syndrome. These probably represent just 5% of childhood-onset dystonias.

Clinical Scales Appropriate diagnosis of dystonia can be difficult to achieve owing to the presence of coexisting movement disorders and spasticity in many of those affected. There are several clinical scales in use today to aid in the measurement of dystonia. Clinical scales were first utilized for primary dystonias. The Burke Fahn Marsden (BFM) scale, published in 1985, was the first utilized scale for rating generalized dystonia, hemidystonia, and segmental dystonia (Krystkowiak et al. 2007). The BFM dystonia scale evaluates the presence of dystonia in nine body regions, eyes, mouth, neck trunk, and right and left arms and legs. This scale identifies both provoking factors and severity factors in each region and includes a separate disability rating. Scores range from 0 to 120, with higher scores indicating more severe dystonia. This scale has demonstrated good inter- and intra-rater reliability. This scale remains in use in both clinical and research setting to track dystonia over time and in response to treatments. As secondary dystonias and available treatments have evolved, the need to discriminate dystonia from other hypertonic syndromes led to the development of additional measurement scales. The Barry–Albright Dystonia Scale was developed

1272

J. Pruente and D. Gaebler-Spira

to improve reliability and measurement of secondary dystonias (Barry et al. 1999; Pavone et al. 2013). Assessment of dystonia in this population can be more difficult secondary to presence of coexisting brain injury or cognitive impairment. This scale consists of assessing secondary dystonia in eight regions, eyes, mouth, neck, trunk, and the four extremities. Scores range from 0 to 32, and higher scores indicate more severe dystonia. Finally, the Unified Dystonia Rating Scale measures severity and duration of dystonia in 14 body regions, eyes and upper face, lower face, jaw and tongue, larynx, neck, trunk, shoulder/proximal arm, distal arm/hand, proximal leg, and distal leg/foot. This scale has high internal consistency and inter-rater reliability in primary dystonias (Goetz et al. 2008; Monbaliu et al. 2010). In children, assessment of movement disorders can be more difficult owing to difficulty with exam instructions. Several scales have been developed to address this group. The hypertonia assessment tool or HAT is one such scale. This scale was developed to help differentiate the different types of hypertonia; dystonia, spasticity, and rigidity (Jethwa et al. 2010; Pavone et al. 2013). The HAT is a seven-item clinical assessment tool designed for children aged 4–19. It involves a binary rating scale for three presentations of dystonia, two of spasticity and two of rigidity. This scale has fair inter-rater reliability and moderate test-retest reliability for dystonia. The Movement Disorder-Childhood Rating Scale was likewise designed for clinical evaluation of movement disorders in ages 4–19 (Battini et al. 2015). The movement disorder portion of this scale assesses dystonia at rest and during functional tasks in the eye/orbital region, face, neck, perioral, trunk, upper extremities, and lower extremities. The childhood rating scale includes assessment of motor function (including head control, sitting position, standing position, walking, reaching, grasping, and handwriting), oral/verbal function, self-care, and attention/alertness. This has been demonstrated to have high inter-rater reliability and high internal consistency. Dystonia can be difficult to measure accurately given its variable presentation, dynamic changes, and coexistence with other movement disorders. Newer therapeutic options including intrathecal baclofen and deep brain stimulation have increased the widespread utilization of these scales in order to ensure appropriate patient selection and to monitor response to treatment. New scales continue to be developed in response to the need for improved measurement of dystonia in response to treatments. The dyskinesia impairment scale was more recently developed and includes two subscales, for dystonia and choreoathetosis, and evaluates both duration and amplitude (Elegast Monbaliu et al. 2012). This was demonstrated to have good inter-rater reliability and internal consistency in initial studies (Fig. 2).

Kinetics/Kinematics While clinical scales can aid the clinician in diagnosis of dystonia, objective measurement using gait labs, kinetics, and kinematics also has an important role in dystonia assessment and determining therapeutic effect. Animal models were first used to characterize gait changes associated with dystonia. In a rat model of

Functional Dystonias

1273 HYPERTONIA ASSESSMENT TOOL (HAT)

HAT ITEM 1. Increased involuntary movements/postures of the designated limb with tactile stimulus of a distal body part

SCORING GUIDELINES (0=negative or 1=positive)

SCORE 0=negative 1=positive (circle score)

0= No involuntary movements or postures observed

0

1= Involuntary movements or postures observed

1

0= No involuntary movements or postures 2. Increased involuntary movements/postures with purposeful observed movements of a distal body part 1= Involuntary movements or postures observed

0

3. Velocity dependent resistance to stretch

4. Presence of a spastic catch

TYPE OF HYPERTONIA

1

0= No increased resistance noticed during fast stretch compared to slow stretch

0

1= Increased resistance noticed during fast stretch compared to slow stretch

1

0= No spastic catch noted

0

1= Spastic catch noted

1

0= Equal resistance not noted with bi-directional 5. Equal resistance to passive stretch movement during bi-directional movement of a 1= Equal resistance noted with bi-directional joint movement 0= No increased tone noted with purposeful 6. Increased tone with movement of movement a distal body part 1= Greater tone noted with purposeful movement 0= Limb returns (partially or fully) to original 7. Maintenance of limb position position after passive movement 1= Limb remains in final position of stretch

DYSTONIA

DYSTONIA

SPASTICITY

SPASTICITY

0 1

RIGIDITY

0 1

DYSTONIA

0 1

RIGIDITY

Fig. 2 The hypertonia assessment tool (Fehlings et al. 2010)

dystonia, decreased walking speed, increased hind limb spread, and increased step length ratio variability were consistent with dystonia. Of these, step length ratio variability was the most sensitive for detecting dystonia (Chaniary et al. 2009). Further data in rat models of dystonia and ataxia also demonstrate coactivation of muscle and similar changes to gait parameters (Scholle et al. 2010). Clinical studies in pediatrics are somewhat limited with the largest amount of data pertaining to upper extremity dystonias. Kinematics data have been collected in children with cerebral palsy during reach and grasp activities. Those with dystonia have slower movements during reaching and decreased coordination of movement. Different kinematics were obtained in the three CP subtypes, spastic, dystonic, and mixed, and may be useful in distinguishing between movement disorders (Butler et al. 2010; Kukke et al. 2015). The kinematic dystonia measure collects kinematic data during an upper extremity finger tapping task has been demonstrated to correlate with the Barry–Albright Dystonia Scale and may improve quantitative assessment of dystonia (Kawamura et al. 2012). A rest-tap test involving repeated tapping in one limb during assessment of the contralateral limb for dystonia demonstrates a different kinematic pattern compared to controls. This is characterized by overflow in the contralateral limb and increased joint excursion, see Fig. 3 (Gordon et al. 2006). Gait analysis in children that includes surface EMG suggests that co-contraction and increased resistance to external motion and slow velocities are present with dystonia (Lebiedowska et al. 2004). Dystonia presents in EMG data as an increased

1274

J. Pruente and D. Gaebler-Spira

Fig. 3 Dystonia kinematics. Displacement of the shoulder, elbow, and wrist joints over time during the rest-tap paradigm. (a) represents a control subject. (b) represents a cerebral palsy subject with low dystonia. (c) represents a cerebral palsy subject with high dystonia (Figure reproduced with permission from Pediatric Neurology, Can Spasticity and Dystonia Be Independently Measured in Cerebral Palsy (Gordon et al. 2006))

number of muscles responding during volitional activity but did not respond at rest or during quick stretch. This is in contrast to spasticity, which demonstrated brief bursts of EMG activity during quick stretch, but low activity levels during rest or volitional activity. A specific electromyographic protocol suggests that lower extremity assessment during rest, quick stretch, and five volitional tasks can detect different muscle activation patterns in spasticity and dystonia and might provide objective data for diagnosis (Beattie et al. 2016). Please see the example videos for two demonstrations of dystonic gait (online only). Figure 4 demonstrates EMG tracings during motion analysis that can help to distinguish dystonia from spasticity (Reference: personal communication with Dr. Jules Becher). Dystonia can be inferred by careful analysis of EMG patterns as well as by kinematic evidence of posturing of extremities during motion analysis. With respect to the electromyographic patterns, typically in dystonia, the raw or rectified tracing displays a peak and valley pattern or an inconsistent pattern of activation. This is in contrast to the muscle patterns of spasticity that are more consistent with the amplitude of the electromyographic signal remaining constant (see Fig. 4b). In adults, head and neck kinematics is useful for accurate description of severity of cervical dystonia as a baseline for treatment effects. Fastrack allows the extraction of kinematic information (i.e., posture, angular range of motion, movement times, angular velocity) about head deviations (Galardi et al. 2003; Jordan et al. 2000). As in dystonic gait, head and neck asymmetry or the symmetry index is increased for rotation and lateral flexion to those with dystonia compared to those unaffected (Boccagni et al. 2008). The use of lower extremity motion analysis for adults with dystonia has been used to evaluate and document effects of treatment for disorders characterized by dystonia. In a patient with dopamine-responsive dystonia, 3D motion analysis accurately

Functional Dystonias

a 2000

1275

L Anterior Tibialis

mV

–2000

L Gastrocnemius 2000

mV

–2000

L Rectus Femoris 2000

mV –2000

b

L Anterior Tibialis

1000 mV

–1000 1000

L Gastrocnemius

mV

–1000 1000

mV –1000

Fig. 4 (continued)

L Rectus Femoris

1276

J. Pruente and D. Gaebler-Spira

c 1000

L Anterior Tiabialis

mV

–1000

1000

L Gastrocnemius

mV

–1000 1000

L Rectus Femoris

mV

–1000

Fig. 4 EMG patterns obtained during motion analysis using surface EMG pads. (a) tracing shows the anterior tibialis, gastrocnemius, and rectus femoris in a child with mild dystonia. (b) tracing shows the same muscles in a more typical pattern for spasticity. (c) tracing shows an example of coexisting spasticity and dystonia

documented changes in gait with reduction of dystonia by medication. When the dystonia was reduced, the gait pattern demonstrated an increase in the walking speed, explained by a significant increase in step frequency and length. With improvement of dystonia, the asymmetry decreased, as did the step width. The gait analysis allowed clinicians to quantify the effects of dystonia on gait (Rebour et al. 2015). MAC for an adult patient with DYT-1 before and after DBS determined which involuntary movements were related to postural instability and gait disturbance. Neck and trunk markers add value and allow discrimination of the cervical dystonic posture, on balance or gait. Prior to DBS, posture and gait were asymmetrical and unstable. Functional body balance was controlled by changes of symptoms, with partial corrections of neck and spinal alignments in a static posture. The patient was better able to maintain the stability of center of mass and center of pressure. The neck angles remained abnormal with specific motions during gait compared to the spine while maintaining improved gait. Functional improvements of gait were captured by gait parameters including increasing of cadence (step rate) and walking speed, increased step length, reduction of a wide base, and extension of single support time and symmetry (Nakao 2011). These two case reports illustrate the power of motion analysis in capturing the effects of dystonia on gait rather than clinical measures of speed and distance tests.

Functional Dystonias

1277

Conclusion Motion analysis assists with planning surgical decisions and in establishing energy cost of walking. When dystonia is present on motion analysis data, this information informs the surgeon and could be useful with surgical decision-making. Whereas predictable outcomes are likely with spastic gait patterns, the child with dystonia will have more variation in surgical outcomes. Thus far, the published use of motion analysis in adults most often documents the treatment effects of various interventions. The physical therapy evaluation at the time of the motion analysis typically includes ROM, strength, and estimate of spasticity. By including the HAT and a severity rating of dystonia the clinical picture can then be associated with the biomechanics of gait. The video review portion of gait analysis is critical as dystonia will be apparent during gait with atypical trunk, arm, and hand postures identified. A severity rating can then be validated from the PT evaluation. Identifying dystonia by motion analysis can theoretically quantify and validate severity rating of dystonia (Sanger et al. 2010). Including the HAT and severity rating as a routine part of the clinical motion analysis is a good first step in validating the typical findings of asymmetry, variability, and EMG firing patterns reported in children with dystonia. With dystonia in the motion analysis lab, common findings reported in children and adults include a high variability in step length and base of support. Surface electromyography data likewise suggests diagnosis of dystonia through co-contraction, overflow muscle activity, and increased muscle activity during volitional tasks. Formal motion analysis has the potential to accurately quantify the motion deviation and determine changes of movement trajectories following treatment. Motion analysis for clinical and research practice promises quantifiable insight into the neural mechanisms of hypertonia and hyperkinetic syndromes during functional tasks.

Summarizing • Dystonia is a common hypertonic and hyperkinetic movement disorder that can have profound impact on gait and upper extremity function. • Diagnosis of dystonia relies upon visual observations of gait and functional tasks, dystonia severity rating scales, and regimented physical examination. • Distinguishing dystonia from other movement disorders, especially spasticity, informs treatment and management decisions. • Characteristic changes in motion analysis associated with dystonia include variable step length and base of support, muscle co-contraction, overflow muscle activity during functional tasks, and increased muscle activity during volitional tasks. • Motion analysis laboratories may play more of a role in the future in diagnosing dystonia, assessing treatment effects, and in surgical/treatment planning.

1278

J. Pruente and D. Gaebler-Spira

References Agostino R, Berardelli A, Formica A, Accornero N, Manfredi, M (1992). Sequential arm movements in patients with Parkinson’s disease, Huntington’s disease and dystonia. Brain 115 (Pt 5):1481–1495 Barry MJ, Van Swearingen JM, Albright AL (1999) Reliability and responsiveness of the BarryAlbright dystonia scale. Dev Med Child Neurol. 41(6), 404–411 Battini R, Olivieri I, Di Pietro R, Casarano M, Sgandurra G, Romeo DM, Cioni G (2015) Movement disorder-childhood rating scale: a sensitive tool to evaluate movement disorders. Pediatr Neurol 53(1):73–77. https://doi.org/10.1016/j.pediatrneurol.2015.02.014 Beattie C, Gormley M, Wervey R, Wendorf H (2016) An electromyographic protocol that distinguishes spasticity from dystonia. J Pediatr Rehabil Med 9(2):125–132. https://doi.org/10.3233/ PRM-160373 Boccagni C, Carpaneto J, Micera S, Bagnato S, Galardi G (2008) Motion analysis in cervical dystonia. Neurol Sci 29(6):375–381. https://doi.org/10.1007/s10072-008-1033-z Butler EE, Ladd AL, Lamont LE, Rose J (2010) Temporal-spatial parameters of the upper limb during a Reach & Grasp Cycle for children. Gait Posture 32(3):301–306. https://doi.org/ 10.1016/j.gaitpost.2010.05.013 Carecchio M, Zorzi G, Nardocci N (2015) Inherited isolated dystonia in children. J Pediatr Neurol 13(04):174–179. https://doi.org/10.1055/s-0035-1558863 Chaniary KD, Baron MS, Rice AC, Wetzel PA, Ramakrishnan V, Shapiro SM (2009) Quantification of gait in dystonic Gunn rats. J Neurosci Methods 180(2):273–277. https://doi.org/10.1016/j. jneumeth.2009.03.023 Curra A, Berardelli A, Agostino R, Giovannelli M, Koch G, Manfredi M (2000). Movement cueing and motor execution in patients with dystonia: a kinematic study. Mov Disord 15(1):103–112 de Carvalho Aguiar PM, Ozelius LJ (2002) Classification and genetics of dystonia. Lancet Neurol 1(5):316–325. https://doi.org/10.1016/s1474-4422(02)00137-0 Fehlings, D., Switzer, L, Jethwa, A, Mink, J, Macarthur, C, Knights, S, & Fehlings, T. (2010). Hypertonia assessment tool: user manual p.1–10 Galardi G, Micera S, Carpaneto J, Scolari S, Gambini M, Dario P (2003) Automated assessment of cervical dystonia. Mov Disord 18(11):1358–1367. https://doi.org/10.1002/mds.10506 Goetz CG, Nutt JG, Stebbins GT (2008) The unified dyskinesia rating scale: presentation and clinimetric profile. Mov Disord 23(16):2398–2403. https://doi.org/10.1002/mds.22341 Gordon LM, Keller JL, Stashinko EE, Hoon AH, Bastian AJ (2006) Can spasticity and dystonia be independently measured in cerebral palsy? Pediatr Neurol 35(6):375–381. https://doi.org/ 10.1016/j.pediatrneurol.2006.06.015 Gregori B, Agostino R, Bologna M, Dinapoli L, Colosimo C, Accornero N, Berardelli A (2008). Fast voluntary neck movements in patients with cervical dystonia: a kinematic study before and after therapy with botulinum toxin type A. Clin Neurophysiol. 119(2):273–280. https://doi.org/ 10.1016/j.clinph.2007.10.007 Inzelberg R, Flash T, Schechtman E, Korczyn AD (1995). Kinematic properties of upper limb trajectories in idiopathic torsion dystonia. J Neurol Neurosurg Psychiatry. 58(3):312–319 Jethwa A, Mink J, Macarthur C, Knights S, Fehlings T, Fehlings D (2010) Development of the hypertonia assessment tool (HAT): a discriminative tool for hypertonia in children. Dev Med Child Neurol 52(5):e83–e87 Jordan K, Dziedzic K, Jones PW, Ong BN, Dawes PT (2000) The reliability of the threedimensional FASTRAK measurement system in measuring cervical spine and shoulder range of motion in healthy subjects. Rheumatology (Oxford) 39(4):382–388 Kawamura A, Klejman S, Fehlings D (2012) Reliability and validity of the kinematic dystonia measure for children with upper extremity dystonia. J Child Neurol 27(7):907–913. https://doi. org/10.1177/0883073812443086 Krauss JK, Jankovic J (2002) Head injury and posttraumatic movement disorders. Neurosurgery 50(5): 927–939.

Functional Dystonias

1279

Krystkowiak P, du Montcel ST, Vercueil L, Houeto JL, Lagrange C, Cornu P, ... Group S (2007) Reliability of the Burke-Fahn-Marsden scale in a multicenter trial for dystonia. Mov Disord 22(5):685–689. https://doi.org/10.1002/mds.21392 Kukke S, Curatalo L, de Campos A, Hallett M, Alter K, Damiano D (2015) Coordination of reachto-grasp kinematics in individuals with childhood-onset dystonia due to hemiplegic cerebral palsy. IEEE Trans Neural Syst Rehabil Eng. https://doi.org/10.1109/tnsre.2015.2458293 Lebiedowska MK, Gaebler-Spira D, Burns RS, Fisk JR (2004) Biomechanic characteristics of patients with spastic and dystonic hypertonia in cerebral palsy. Arch Phys Med Rehabil 85(6):875–880 Monbaliu E, Ortibus E, Roelens F, Desloovere K, Deklerck J, Prinzie P, ... Feys H (2010) Rating scales for dystonia in cerebral palsy: reliability and validity. Dev Med Child Neurol 52(6): 570–575. https://doi.org/10.1111/j.1469-8749.2009.03581.x Monbaliu E, Ortibus ELS, de Cat JOS, Dan B, Heyrman L, Prinzie P, ... Feys H (2012) The dyskinesia impairment scale: a new instrument to measure dystonia and choreoathetosis in dyskinetic cerebral palsy. Dev Med Child Neurol 54(3):278–283. https://doi.org/10.1111/ j.1469-8749.2011.04209.x Monbaliu E, de Cock P, Ortibus E, Heyrman L, Klingels K, Feys H (2016) Clinical patterns of dystonia and choreoathetosis in participants with dyskinetic cerebral palsy. Dev Med Child Neurol. 58(2):138–144. https://doi.org/10.1111/dmcn.12846 Nakao S (2011) Gait and posture assessments of a patient treated with deep brain stimulation in dystonia using three-dimensional motion analysis systems. J Med Investig 58:264–272 Pavone L, Burton J, Gaebler-Spira D (2013) Dystonia in childhood: clinical and objective measures and functional implications. J Child Neurol 28(3):340–350. https://doi.org/10.1177/ 0883073812444312 Phukan J, Albanese A, Gasser T, Warner T (2011) Primary dystonia and dystonia-plus syndromes: clinical characteristics, diagnosis, and pathogenesis. Lancet Neurol 10(12):1074–1085. https:// doi.org/10.1016/s1474-4422(11)70232-0 Rebour R, Delporte L, Revol P, Arsenault L, Mizuno K, Broussolle E, ... Rossetti Y (2015) Doparesponsive dystonia and gait analysis: a case study of levodopa therapeutic effects. Brain Dev 37(6): 643–650. https://doi.org/10.1016/j.braindev.2014.09.005 Sanger TD (2003) Pediatric movement disorders. Curr Opin Neurol 16(4):529–535. https://doi.org/ 10.1097/01.wco.0000084233.82329.0e Sanger TD (2004) Toward a definition of childhood dystonia. Curr Opin Pediatr 16:623–627 Sanger T (2015) Movement disorders in Cerebral Palsy. J Pediatr Neurol 13(04):198–207. https:// doi.org/10.1055/s-0035-1558866 Sanger TD, Chen D, Fehlings DL, Hallett M, Lang AE, Mink JW, ... Valero-Cuevas F (2010) Definition and classification of hyperkinetic movements in childhood. Mov Disord 25(11):1538–1549. https://doi.org/10.1002/mds.23088 Scholle HC, Jinnah HA, Arnold D, Biedermann FH, Faenger B, Grassme R, ... Schumann NP (2010) Kinematic and electromyographic tools for characterizing movement disorders in mice. Mov Disord 25(3):265–274. https://doi.org/10.1002/mds.22933 Skogseid IM (2014) Dystonia – new advances in classification, genetics, pathophysiology and treatment. Acta Neurol Scand Suppl 198:13–19. https://doi.org/10.1111/ane.12231 van der Kamp W, Berardelli A, Rothwell JC, Thompson PD, Day BL, Marsden CD (1989). Rapid elbow movements in patients with torsion dystonia. J Neurol Neurosurg Psychiatry. 52(9):1043–1049 van Egmond ME, Kuiper A, Eggink H, Sinke RJ, Brouwer OF, Verschuuren-Bemelmans CC, ... de Koning TJ (2015) Dystonia in children and adolescents: a systematic review and a new diagnostic algorithm. J Neurol Neurosurg Psychiatry 86(7):774–781. https://doi.org/10.1136/ jnnp-2014-309106

Part XVI Traumatic and Orthopedic Gait Disorders

Gait Changes in Skeletal Dysplasia William G. Mackenzie and Oussama Abousamra

Abstract

Skeletal dysplasias form a group of conditions that affect the musculoskeletal system resulting in gait changes. These conditions share some common gait patterns, such as short stride length, low gait velocity, and increased forward pelvic tilt. Increased pelvic rotation is also a common mechanism, used to increase the step length. However, specific gait deviations are found in each condition. Knee axis deviations in the coronal plane are well-described clinical features in these conditions. While bilateral knee varus is common in achondroplasia, bilateral knee valgus is a shared finding between diastrophic dysplasia and Morquio syndrome. Sagittal kinematic deviations in stance show normal knee flexion in achondroplasia and increased knee flexion values in diastrophic dysplasia and Morquio syndrome. Due to the different, multilevel gait deviations that follow the multilevel deformity found in skeletal dysplasia, instrumented, three-dimensional, gait analysis becomes a helpful tool to evaluate the individual aspects of the deformity and to approach the patient with an accordingly customized management plan. Keywords

Skeletal dysplasia • Achondroplasia • Diastrophic dysplasia • Morquio syndrome • Spondyloepiphyseal dysplasia

W.G. Mackenzie (*) Nemours A.I. duPont Hospital for Children, Wilmington, DE, USA e-mail: [email protected] O. Abousamra (*) Nemours Alfred I. duPont Hospital for Children, Wilmington, DE, USA e-mail: [email protected] # Springer International Publishing AG, part of Springer Nature 2018 B. Müller, S.I. Wolf (eds.), Handbook of Human Motion, https://doi.org/10.1007/978-3-319-14418-4_73

1283

1284

W.G. Mackenzie and O. Abousamra

Contents Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Gait Changes in Achondroplasia . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Gait Changes in Diastrophic Dysplasia . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Gait Changes in Morquio Syndrome . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Gait Changes in Spondyloepiphyseal Dysplasia . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

1284 1284 1285 1286 1287 1287 1287

Introduction Multiple conditions have been described under the term “skeletal dysplasia.” The pathophysiology of each condition has its effect on the anatomy and function of the musculoskeletal system. Since the affected body level, and the severity of the disease, can be different between these conditions, the total impact on the individual’s gait can be different as well. As in any other, gait disturbing, disorders, gait evaluation allows the managing team to address the deformity more accurately and tailor the surgical intervention accordingly. Gait studies in skeletal dysplasia have been scarce in the literature. Published studies have addressed the three-dimensional dynamic lower extremity deformities of children with different skeletal dysplasia conditions. Comparisons with dynamic measurements of typically developing children have been reported. This chapter reviews the available data that present gait changes in achondroplasia, diastrophic dysplasia, Morquio syndrome, and spondyloepiphyseal dysplasia.

Gait Changes in Achondroplasia Achondroplasia is the most common skeletal dysplasia. It is caused by a mutation of the gene encoding the fibroblast growth factor receptor 3 (FGFR3), and it is inherited as an autosomal dominant disorder. The extremities are affected more than the trunk, and long bones are affected more in the proximal portions of the extremities (Shirley and Ain 2009). Children with achondroplasia have shorter stride length and lower gait velocity when compared with typically developing children (stride length (achondroplasia, 67  18 cm; normal, 106  13 cm) and gait velocity (achondroplasia, 80  19 cm/s; normal, 122  18 cm/s)) (Inan et al. 2006). In order to increase the step length, children with achondroplasia show higher rotation at the pelvis level (21  10 ). Increased anterior pelvic tilt (21  6 ) and limited hip extension ( 10  9 ) are also noted and might be due to the increased lordotic curve seen in achondroplasia (Kopits 1988). In the coronal plane, varus malalignment of the knee (19  13 ) has been reported in children with achondroplasia (Inan et al. 2006). When knee varus

Gait Changes in Skeletal Dysplasia

1285

measurements were compared between gait analysis and radiographs of the same patients, poor correlation was found with less varus noted on the radiographs (16  9 ). This has been related to the pattern of weight bearing since the kinematic data were obtained in stance when the child was bearing weight on one leg, while the radiographs were taken with the child bearing weight on both legs (Inan et al. 2006). In the sagittal plane, knee flexion has been found normal across the gait cycle; however, radiographs of the same patients showed genu recurvatum (23.5  13.2 ) (Inan et al. 2006). This finding has been explained by the normal neuromuscular function in these children, which can prevent the recurvatum during walking. Conversely, reduced knee extension in midstance was also reported in adults with achondroplasia (mean age of 24.5  6.1 years), and this extension limitation remained after limb lengthening despite having a good range of motion (van der Meulen et al. 2008). In another group of adolescents with achondroplasia, knee hyperextension range was seen displaying relaxed hamstrings; however, there was still limited knee extension in midstance, which might be due to excessive ankle dorsiflexion (Egginton et al. 2006). Elevated internal valgus moments at the knee have been found compared with the general population and suggested as a good predictor of deformity progression and long-term degenerative disease (Inan et al. 2006). While tibial torsion kinematic measurements have showed a good correlation with the clinical measurements, using the angle between the transmalleolar axis and the long axis of the thigh, variable measurements have been reported between patients and even between limbs of the same patient. This variability necessitates a careful assessment in case tibial torsion is to be surgically addressed (Inan et al. 2006). Foot varus has been noted on foot pressure analysis. However, full passive flexibility was reported (Inan et al. 2006). Reduced plantar flexion at push off and increased dorsiflexion during swing were reported (Egginton et al. 2006) and suggested as a mechanism to facilitate foot clearance of the relatively long foot compared to the short leg (Egginton et al. 2006). In a different group of adults with achondroplasia who had limb lengthening, sagittal plane kinematics at the ankle were within one standard deviation of normal (van der Meulen et al. 2008). This finding suggests improved gait following limb lengthening (van der Meulen et al. 2008).

Gait Changes in Diastrophic Dysplasia Diastrophic dysplasia is a rare form of skeletal dysplasia, inherited as an autosomal recessive disorder. The genetic defect is in chromosome 5q that encodes diastrophic dysplasia sulfate transfer protein (Bayhan et al. 2015). The typical clinical picture is short stature with short trunk and short limbs (Poussa et al. 1991). Similar to children with achondroplasia, children with diastrophic dysplasia have shorter stride length and lower gait velocity when compared with typically

1286

W.G. Mackenzie and O. Abousamra

developing children (Bayhan et al. 2015) (stride length (diastrophic dysplasia, 52  16 cm; normal, 98  11 cm) and gait velocity (diastrophic dysplasia, 61  26 cm/s; normal, 115  6 cm/s)). Increased pelvic rotation (12.1  8.14 forward and 12.2  8.3 trailing), increased forward pelvic tilt (26.6  5.7 ), and increased forward trunk tilt (14.9  12.2 ) have also been reported for these children along with increased hip flexion and limited hip abduction (Bayhan et al. 2015). In the coronal plane, knee valgus malalignment (12.9  8.6 ) has been noted; however, these measurements did not correlate with radiographic knee valgus measurements (11.7  13 ), and they moderately correlated with the clinical valgus measurements (13  7 ) (Bayhan et al. 2015). In the sagittal plane, increased knee flexion has been reported with smaller range of motion for the knee flexion extension movement (lower delta knee motion: 24.9  6.9 ) (Bayhan et al. 2015). In this group of children, two subgroups were reported according to whether they have patellar dislocation or not. Average hip flexion and average knee flexion were both elevated in the subgroup with patellar dislocation (Bayhan et al. 2015). Therefore, the early management of the extensor mechanism malalignment, in addition to the valgus malalignment, was proposed as a possible effective treatment to slow deformity progression.

Gait Changes in Morquio Syndrome Morquio syndrome, or mucopolysaccharidosis IV, is a systemic lysosomal storage disorder (Dhawale et al. 2013). It is an autosomal recessive disorder caused by a deficiency of the lysosomal enzyme N-acetylgalactosamine-6-sulfate sulfatase (Tomatsu et al. 2011). The clinical picture in this syndrome includes a short trunk in addition to multiple limb deformities (Tomatsu et al. 2011). Children with Morquio syndrome also have shorter stride length and lower gait velocity when compared with typically developing children (stride length (Morquio, 72.3  7.9 cm; normal, 114.5  6.8 cm) and gait velocity (Morquio, 67.2  17.3 cm/s; normal, 126.3  14.7 cm/s)) (Dhawale et al. 2013). Increased forward pelvic tilt (19.3  2.9 ) and increased forward trunk tilt (5.9  8.9 ) have also been reported findings in these children. In addition, increased hip flexion and limited hip abduction have been described (Dhawale et al. 2013). In the coronal plane, knee valgus has been reported (22  11.2 ), and in this group of children, strong correlation was noted between kinematic knee valgus and radiographic knee valgus (24.7  9.6 ) (Dhawale et al. 2013). In the sagittal plane, increased knee flexion has been reported (19.3  15.5 ). Elevated internal varus moments at the knee have been found in response to knee valgus (Dhawale et al. 2013). Limited ankle dorsiflexion has been described compared with typically developing children (Dhawale et al. 2013), and pedobarography showed normal foot pressures (Table 1).

Gait Changes in Skeletal Dysplasia

1287

Table 1 Measurements of means  standard deviations in every condition in addition to the percentages of the measurements taken from the typically developing children Measurements Stride length (cm) Gait velocity (cm/s) Anterior pelvic tilt Knee varus/valgus

Achondroplasia 67  18 (63 %) 80  19 (66 %) 21  6 (162 %) 19  13 varus

Diastrophic dysplasia 52  16 (53 %) 61  26 (53 %) 26.6  5.7 (193 %) 12.9  8.6 valgus

Morquio syndrome 72.3  7.9 (63 %) 67.2  17.3 (53 %) 19.3  2.9 (127 %) 22  11.2 valgus

Gait Changes in Spondyloepiphyseal Dysplasia Spondyloepiphyseal dysplasia (SED) is a condition that primarily affects the vertebrae and the proximal epiphyses of the long bones (Anderson et al. 1990). The genetic defect has been localized on the COL2A1 gene coding for type II collagen (Anderson et al. 1990). The clinical findings include short trunk and short limbs with multiple deformities in the spine, chest, and hips. The authors have unpublished data for a group of children with SED congenita who had instrumented gait analysis before and after they underwent proximal femoral valgus osteotomy to address their coxa vara. The preoperative gait evaluation showed short stride length (70  18 cm) and reduced gait velocity (87  23 cm/s). Forward pelvic tilt was increased (34  9 ) as well as hip flexion (23  13 ). None of the parameters showed significant changes after the valgus osteotomy except forward pelvic tilt that reduced from (34  9 ) to (25  10 ). Gait deviation index (GDI) showed a statistically significant improvement (from 52  15 to 58  15); however, this change was within the normal, ten points, standard deviation of GDI.

Summary In skeletal dysplasia, specifically in the conditions discussed in this chapter, different gait changes are noted. Overall, people with skeletal dysplasia walk with shorter steps and slower speed than the general population. Increased forward pelvic tilt has been a common finding. Different malalignment patterns can be found at the knee level. Instrumented gait analysis can serve as a helpful tool to evaluate the multilevel malalignment profile in children with skeletal dysplasia and to customize the treatment based on their functional needs.

References Anderson IJ, Goldberg RB, Marion RW et al (1990) Spondyloepiphyseal dysplasia congenita: genetic linkage to type II collagen (COL2AI). Am J Hum Genet 46(5):896–901 Bayhan IA, Er MS, Nishnianidze T et al (2015) Gait pattern and lower extremity alignment in children with diastrophic dysplasia. J Pediatr Orthop. https://doi.org/10.1097/BPO. 0000000000000530

1288

W.G. Mackenzie and O. Abousamra

Dhawale AA, Church C, Henley J et al (2013) Gait pattern and lower extremity alignment in children with Morquio syndrome. J Pediatr Orthop B 22(1):59–62 Egginton R, Newman C, Walsh M et al (2006) Kinematic characteristics of Achondroplasia. Gait Posture. Elsevier 24(2):S249–S250 Inan M, Thacker M, Church C et al (2006) Dynamic lower extremity alignment in children with achondroplasia. J Pediatr Orthop 26(4):526–529 Kopits SE (1988) Thoracolumbar kyphosis and lumbosacral hyperlordosis in achondroplastic children. Basic Life Sci 48:241–255 Poussa M, Merikanto J, Ryöppy S et al (1991) The spine in diastrophic dysplasia. Spine 16(8):881–887 Shirley ED, Ain MC (2009) Achondroplasia: manifestations and treatment. J Am Acad Orthop Surg 17(4):231–241 Tomatsu S, Montano AM, Oikawa H et al (2011) Mucopolysaccharidosis type IVA (Morquio a disease): clinical review and current treatment: a special review. Curr Pharm Biotechnol 12(6):931–945 van der Meulen J, Dickens W, Burton M et al (2008) O052 Gait characteristics of achondroplasia following lower limb-lengthening. Gait Posture 28:S36–S37, Elsevier

Impact of Scoliosis on Gait Elizabeth A. Rapp and Peter G. Gabos

Abstract

Scoliosis is one of the most common orthopedic disorders in children and adolescents. The idiopathic classification has been studied at length in hopes of identifying factors contributing to the origin and progression of the disease. Gait analysis is frequently employed to analyze the balance and movement abnormalities associated with the disorder. While the majority of gait studies in scoliosis note some deviations from normal gait, specific conclusions are often based on weak or inconsistent evidence. The most widely reported findings include restricted motion of the pelvis and hip and an asymmetrical rotation of the trunk as well as general asymmetry between limbs. Additionally, energy cost and muscle activation are higher during scoliotic gait than in normal walking. These differences seem to improve with both orthotic and surgical treatment, although postoperative adolescents with idiopathic scoliosis still maintain a higher-energy cost of walking than their typically developing peers. Ultimately, the relationship between gait abnormalities and the origin or progression of the scoliotic curve remains unclear. The idea of a neurological dysfunction that contributes to both the spinal deformity and the gait deviation is predominantly rooted in theory. Still, future research into motor control and somatosensory function during gait may provide more insight into a neurological influence in the scoliosis population.

E.A. Rapp (*) University of Delaware, Newark, DE, USA e-mail: [email protected] P.G. Gabos (*) Nemours A.I. duPont Hospital for Children, Wilmington, DE, USA e-mail: [email protected] # Springer International Publishing AG, part of Springer Nature 2018 B. Müller, S.I. Wolf (eds.), Handbook of Human Motion, https://doi.org/10.1007/978-3-319-14418-4_68

1289

1290

E.A. Rapp and P.G. Gabos

Keywords

Gait • Idiopathic scoliosis • Spinal deformity • Scoliotic gait • Kinematics • Asymmetry • Kinetics • Dynamic balance • Proprioception • Muscle activity • Curve severity • Bracing • Spinal fusion • Energy cost

Contents Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . State of the Art . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Background . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Structural Differences . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Spatiotemporal Parameters . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Step Length . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Time in Stance . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Cadence . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Body Kinematics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Arms . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Spine and Trunk . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Pelvis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Distal Kinematics: Hip, Knee, and Ankle . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Lower Extremity Asymmetry . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Kinetics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Vertical Forces . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Horizontal Forces . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Moments and Powers . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Balance and Stability . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Muscle Activation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Relationship of Gait Parameters to Curve Severity . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Response to Treatment . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Response to Bracing . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Response to Surgery . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Summary and Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

1290 1291 1291 1292 1293 1294 1294 1295 1295 1295 1295 1296 1296 1297 1297 1297 1298 1298 1299 1299 1301 1301 1302 1302 1304 1305

Introduction Scoliosis is a widespread orthopedic condition and is the most prevalent type of spinal deformity in children and adolescents (Konieczny et al. 2013). The large majority of cases are idiopathic (scoliosis for which there is no obvious attributed cause), and the remainder are non-idiopathic, including those due to congenital abnormalities and those secondary to a neuromuscular disorder. Many of the underlying diseases that lead to non-idiopathic scoliosis can affect the musculoskeletal system and impair all aspects of mobility, including gait. Due to the confounding influences of these diseases, this chapter will focus only on the most prevalent classification (idiopathic scoliosis) in order to discuss the isolated contribution of the scoliotic deformity to gait patterns. This chapter will describe how scoliosis induces structural changes throughout the body, present evidence of gait patterns in patients with scoliosis, and discuss theories of how gait abnormalities relate to the pathoetiology of scoliosis.

Impact of Scoliosis on Gait

1291

State of the Art Despite the prevalence of idiopathic scoliosis, the factors contributing to the development of the disease are poorly understood. The current body of evidence offers little insight as to factors that contribute to the origin or progression of the deformity. The impact of scoliosis on gait parameters has been studied at length in an attempt to identify structural or neurological trends. The majority of studies note some deviations from normal gait, but weak or inconsistent results limit the substantiation of distinct conclusions. Consequently, these gait findings have limited application for diagnosis or treatment. Results from several studies regarding balance and postural control hint toward a potential neuromuscular influence. Future research will likely explore this direction, with the goal of identifying neurological factors in scoliosis that contribute to both the structural deformity and deviations from normal gait.

Background Literature suggests the incidence of adolescent idiopathic scoliosis to be between 0.47% and 5.2% of children between the ages of 10 and 18. Idiopathic scoliosis is more common in females than males, with this gender bias increasing substantially in more severe curves (Konieczny et al. 2013). Curvature is classified as thoracic, thoracolumbar, lumbar, or double major (Fig. 1), depending on the vertebral location of the apex of the curve. Thoracic curves are most common and are typically convex right, while lumbar curves are typically convex left. Thoracolumbar curves are also common, and their

Fig. 1 Curvature in scoliosis: (a) thoracic, (b) thoracolumbar, (c) lumbar, or (d) double major (Reproduced from Grivas et al. 2010 with permissions granted under the terms of the Creative Commons Attribution License: http://creativecommons.org/licenses/by/2.0)

1292

E.A. Rapp and P.G. Gabos

location and convexity can vary. Double or “S” curves can be the result of actual structural deformity or develop as compensatory spinal balance. Adolescents with idiopathic scoliosis complain of struggles with appearance, self-esteem, and in some cases, pain and functional deficits (Mayo et al. 1994). Curve progression is common without treatment: between 20% and 40% more likely with no treatment as opposed to even a conservative intervention such as bracing (Ascani et al. 1986). Adolescents with curves greater than 50 are at risk for curve progression throughout adult life. Curves with this degree of severity are typically recommended for surgical correction (Asher and Burton 2006). Patients do not commonly complain of difficulty walking; however, there is a clinically observed stiffness and asymmetry seen in scoliotic gait (Mahaudens et al. 2009). The majority of studies investigating gait in scoliosis have noted some abnormality in gait parameters, and thus impaired gait control appears to be moderately associated with idiopathic scoliosis (Schlösser et al. 2014). Nevertheless, there are conflicting findings with respect to some kinematic deviations, as well as opposing theories on the mechanism of these abnormalities. It is widely debated whether the observed gait pathology is a manifestation of the underlying deformity or whether it may be responsible for the causation or progression of the disease.

Structural Differences The three-dimensional spinal deformity in scoliosis creates a trunk distortion that alters the lateral position of the center of mass within the body (Fig. 2). A laterally displaced center of mass can in turn influence the direction of ground reaction forces, requiring atypical and asymmetric joint moments to maintain stability. Additionally, a deformity in the lumbar spine can alter pelvis orientation, inducing subsequent structural and alignment abnormalities that extend through the lower extremities. Scoliosis is associated with a multiplanar rotation deformity of the pelvis. In the transverse plane, the pelvis rotates to the right (clockwise). In anterior/ posterior radiographs, this results in the impression of the right ilium being wider than the left, i.e., a left to right ilium width ratio of less than one (Gum et al. 2007; Fig. 3). This pelvic torsion is the same rotational direction as a typical scoliotic thoracic curve and the opposite rotational direction of a typical lumbar curve. Furthermore, radiographic evidence also demonstrates increased pelvic and sacral obliquity with the segments commonly tilting toward the apex of the lumbar curve (Mahaudens et al. 2005; Schwender and Denis 2000). Both of these findings suggest that the pelvic orientation deformities are a compensatory response to the scoliotic curvature. Compensatory spinal curves are common in the coronal plane, and the radiographic evidence above indicates this response is neither limited to the spine nor to any one plane. All of the compensations are thought to be an attempt to retain balance.

Impact of Scoliosis on Gait

1293

Fig. 2 Drawing of right thoracic scoliosis. Center of mass shift can be noted by the gravitational vector skewed toward the right pelvis (Courtesy of University of Delaware Biomechanics and Movement Science Laboratories)

The deformity can extend all the way to the femur, with some patients showing increased femoral neck-shaft angles on both sides (Saji et al. 1995). While this result could be an extension of the postural stability compensations at the spine and pelvis, the bilateral occurrence of the effect leads some to suggest a neurological influence. Regardless of origin, these structural deformities can hinder the mechanism of walking. In normal gait, the trunk, pelvis, and limbs move in synchronized patterns to generate propulsion and maintain stability (Perry et al. 2010). Irregular segment rotation or alignment can disrupt coordination during gait, inducing compensatory changes in walking patterns and imposing excessive demands on muscles. Still, evidence and theory are mixed as to whether these structural changes are the cause of gait abnormalities or merely a side effect of a greater governing disorder.

Spatiotemporal Parameters Spatiotemporal parameters can provide a global measure of gait performance. In scoliotic gait, these measures are similar to those in normal walking, with a few marginal differences. In general, adolescents with idiopathic scoliosis do not show any deficits in walking speed. Still, while their overall speed is no different than their healthy peers, patients do exhibit differences in some of the components controlling the speed of ambulation (Schizas et al. 1998; Yang et al. 2013; Prince et al. 2010).

1294

E.A. Rapp and P.G. Gabos

Fig. 3 Standing position posterior-anterior radiograph of a patient with a right thoracic scoliosis illustrating the location of the hemi-pelvis landmarks, inferior ilium at the sacroiliac joint, and the anterior superior iliac spine and the lines necessary to measure the left and right hemi-pelvis coronal plane widths (Reproduced from Gum et al. 2007 with permission of Springer). Note the left/right (L/R) ilium width ratio of less than one, indicating a right-rotated pelvis

Step Length Irrespective of curve type, adolescents with idiopathic scoliosis have been shown to walk with a reduced step length (Kramers-de Quervain et al. 2004; Mahaudens et al. 2009; Syczewska et al. 2012; Mallau et al. 2007). This difference is generally small, but significant. Furthermore, while step length in normal gait is symmetrical across both sides, patients with scoliosis demonstrate bilateral differences: usually, the right step length is slightly shorter than the left (Yang et al. 2013). Recalling the prevalence of the right-rotated thoracic curve, right-rotated pelvis, and left-rotated lumbar curve, one might theorize that step length is related to the structural deformity. Surprisingly, there is no evidence of association between those observations – step length abnormalities occur in both left and right curves of all degrees of severity. Thus, any relationship remains theoretical.

Time in Stance Similar to step length, time in stance is marginally reduced in scoliosis (Mahaudens et al. 2009). The main difference from normal walking is an asymmetry between

Impact of Scoliosis on Gait

1295

sides, with longer time in stance phase on the right side, in line with the presentation of longer step length on the left (Yang et al. 2013).

Cadence In most cases, adolescents with idiopathic scoliosis walk with similar cadence to typically developing adolescents (Mahaudens et al. 2009; Yang et al. 2013). Patients with thoracolumbar curves, however, have been shown to walk with a slower cadence, theorized to be an attempt to increase postural stability (Chen et al. 1998). Still, differences in cadence and step length do not produce a clinically observable effect in patients, as evidenced by normal performance in walking speed.

Body Kinematics Kinematic analysis of gait in scoliosis can offer insight as to whether structural abnormalities translate to pathological motion. Beginning with the arms, the following section examines the motion of the limbs, trunk, and pelvis during the gait cycle and discusses how the scoliotic deformity might contribute to the observed patterns.

Arms A primary observation in scoliotic gait is an asymmetrical arm swing: patients walk with greater flexion and extension of both the shoulder and elbow on one side compared to the other (Kramers-de Quervain et al. 2004). While one might reasonably assume that the skewed arm motion occurs as a consequence of an irregularly shaped torso, this phenomenon is not seen on the same side across patients and does not appear to be related to the side or severity of the curve deformity (Kramers-de Quervain et al. 2004). In addition to arm swing asymmetry, frontal plane arm motion is slightly reduced in scoliosis (Mahaudens et al. 2009). However, while the unequal arm swing pattern is clinically detectable, the frontal plane differences are minimal. Motion is reduced by only a few degrees compared to healthy adolescents, which would not likely be noticeable to the naked eye.

Spine and Trunk Along with arm asymmetry, irregular motion of the trunk in the transverse plane is also a frequent observation during scoliotic walking. Adolescents with idiopathic scoliosis – particularly those with a thoracic curve – show asymmetric rotational progression of the trunk during the gait cycle. In normal gait, both shoulders alternately advance throughout the cycle, accompanied by a corresponding rotation of the trunk. In scoliosis, particularly patients with a right convex thoracic curve, the

1296 Fig. 4 Relative motion of the trunk versus the pelvis, reflecting torsional deformity (Adapted from Kramers-de Quervain et al. 2004 with permission of Springer)

E.A. Rapp and P.G. Gabos 40°

Trunk Versus Pelvis

20° 0° -20° -40° 0

20

40

60

80

100% Cycle

right shoulder remains more advanced relative to the line of progression. The asymmetrical trunk rotation can be seen as a torsion of the trunk relative to the pelvis or head. Throughout the gait cycle, patients maintain head orientation in the line of progression and rotate the pelvis side to side symmetrically, as seen in normal walking. Consequently, the right shoulder biasing toward the midline produces a right-forward rotation of the trunk throughout the gait cycle (Kramers-de Quervain et al. 2004; Fig. 4). One study involving patients with thoracolumbar and lumbar curves did not observe this asymmetrical trunk and pelvis pattern, and these authors suggested the torsion may only be a consequence of a thoracic deformity (Mahaudens et al. 2009). However, in general, the coordination of trunk and pelvis motion appears to be atypical and also more variable in scoliosis than observed in normal walking (Park et al. 2015).

Pelvis Motion of the pelvis during gait is reduced in scoliosis (Chen et al. 1998; Mahaudens et al. 2009; Park et al. 2015). Clinically, this presents as an observed “stiffness” during gait, and biomechanical analyses show that pelvic range of motion is restricted in all three planes. It is generally believed that these altered mechanics derive from the structural abnormalities of the pelvis and further theorized that the restricted range of motion is a direct consequence of the spinal deformity (Mahaudens et al. 2009; Kramers-de Quervain et al. 2004).

Distal Kinematics: Hip, Knee, and Ankle Hip motion is also slightly reduced in all three planes (Mahaudens et al. 2009); however, it is unknown whether this detail is related to restricted pelvic motion or the aforementioned femoral neck-shaft deformities. It is more commonly speculated that restrictions in hip motion may be an attempt at postural control and part of an overall strategy to maintain stability by limiting lower extremity range of motion in walking.

Impact of Scoliosis on Gait

1297

The kinematics of more distal segments are more variable in scoliosis, and there is not a strong consensus on whether knee and ankle motion deviate from normal patterns. There is some evidence of reduced knee motion within the sagittal plane; however, the mechanism and rationale for this effect are unclear. This phenomenon may be an extended consequence of the structural pelvic or femoral deformity, or it may be a stability strategy: an attempt to restrict range of motion. Nevertheless, reduced knee motion is not a consistent finding and does not typically bear consideration in the analysis of scoliotic gait.

Lower Extremity Asymmetry With the exception of arm swing and asymmetrical trunk rotation, evidence on the asymmetry of lower extremity kinematics is mixed. Most studies fail to find a statistically significant side-to-side asymmetry at specific lower extremity joints. Additionally, there does not appear to be any association between any observed kinematic asymmetry and the convex and concave sides of the curve (Mahaudens et al. 2009; Kramers-de Quervain et al. 2004). However, in scoliosis, frontal and transverse plane motion of opposite limbs during walking has been shown to be less correlated than in healthy adolescents (Yang et al. 2013). This suggests that asymmetry is not due to one particular joint, but rather the collective motion of all segments. Again, it is unclear whether this occurs due to an altered position of the center of mass or as part of a greater postural control strategy. The asymmetries, as well as the kinematic deviations, appear unrelated to the side and severity of the curve, challenging the hypothesis that gait abnormalities are an isolated result of the structural deformity. Still, theories of neurological dysfunction warrant further investigation, particularly in the arena of postural control and sensorimotor performance.

Kinetics The observations of altered kinematics during gait in scoliosis suggest corresponding alterations in the joint forces controlling movement. Kinetic analyses reveal abnormalities in muscle moments, as well as magnitudes and patterns of ground reaction forces. Patients demonstrate absolute deviations from healthy parameters as well as some bilateral asymmetries.

Vertical Forces Most research agrees that vertical ground reaction forces during gait in scoliosis are similar to those of healthy adolescents. Patients with scoliosis show normal peak and average ground reaction forces during the gait cycle. Additionally, there is no evidence of asymmetry of the vertical ground reaction forces of opposite limbs.

1298 Table 1 Correlation coefficient (CC) of the ground reaction force (GRF) (Reproduced from Yang et al. 2013 with permission of Springer)

E.A. Rapp and P.G. Gabos

FX FY FZ

CC Control 0.087 (0.02) 0.98 (0.00) 0.99 (0.00)

Scoliosis 0.75 (0.05) 0.97 (0.00) 0.99 (0.00)

P 0.039 ns ns

Values in brackets represent the standard error. P value is the t test result. FX, FY, and FZ represent the GRF in the medial/lateral, anterior/ posterior, and vertical forces, respectively. ns no significance

Still, while the force magnitudes are similar to those seen in healthy walking, there is some evidence of pattern differences. In one study that focused solely on vertical ground reaction forces, the majority of adolescents with scoliosis showed asymmetrical loading and unloading rates (Schizas et al. 1998). Previous data established that bilateral vertical ground reaction force differences in healthy walking were within 4% (Herzog et al. 1989), and the differences present within the scoliotic subjects exceeded this threshold. Nevertheless, these differences did not appear to be related to the severity or location of the scoliotic curve.

Horizontal Forces In contrast to the vertical forces, which mostly resemble normal walking, the horizontal ground reaction forces are notably different in scoliosis. Patients show a significant asymmetry in the anterior/posterior (Park et al. 2016) and medial/lateral (Giakas et al. 1996; Yang et al. 2013) component of the ground reaction force, as evidenced by lower correlations between right and left sides than the correlations observed in healthy subjects (Table 1). This finding aligns with the kinematic abnormalities seen during the gait cycle. Frontal and transverse plane asymmetry as seen in the trunk and pelvis deviates the center of mass from the midline of the body, inducing the observed bilateral differences in the medial/lateral component of the ground reaction force.

Moments and Powers Adolescents with idiopathic scoliosis appear to have similar peak joint moments and powers to their healthy peers; however, the bilateral differences in the medial/lateral ground reaction forces are accompanied by an asymmetry in the free moment. The free moment, or the reaction of the foot’s force acting about a vertical axis, is an indicator of torsional loading. This parameter is asymmetrical in scoliosis: the right side of patients with a right thoracic curve demonstrates a bias toward an external rotation moment (Kramers-de Quervain et al. 2004). This is the same direction of the structural pelvic torsion and the opposite direction of the asymmetrical trunk rotation

Impact of Scoliosis on Gait

1299

seen in the kinematics. An asymmetrical free moment is yet another finding that supports the idea of impaired control of motion about the vertical axis of the body.

Balance and Stability The kinematic and kinetic abnormalities in scoliosis all seem to suggest that stability may be compromised during walking. Standing and dynamic balance are of great interest in the scoliotic population, given the alterations in center of mass position resulting from the structural deformity and corresponding restrictions (i.e., stiffness) in joint motion. Measures of standing balance and proprioception in scoliosis demonstrate deficits in these parameters. Patients with scoliosis show differences between limbs in their abilities to reproduce a static knee angle, although the side of the limb with greater difficulty does not appear to be correlated with the side or severity of the spinal curve (Barrack et al. 1984). Additionally, in a standing postural control task, adolescents with idiopathic scoliosis exhibit significantly more sway in all directions (Chen et al. 1998). Dynamic stability, however, does not appear to differ from healthy adolescents (Mallau et al. 2007). As previously mentioned, walking speed is not slower in scoliosis. Furthermore, when confronted with a more difficult ambulation task, such as walking on a line or on a beam, adolescents with scoliosis utilize the same strategy as healthy walkers: they slow down. There is no difference in the rate or magnitude of this reduction of speed. Additionally, patients with scoliosis and healthy adolescents react similarly to an increase in difficulty of a dynamic balancing task by increasing horizontal angular dispersions, i.e., medial/lateral and anterior/ posterior limb motion.

Muscle Activation The uncertainty regarding a neurological influence on observed joint motion in scoliosis has prompted further investigation into motor function during gait. Abnormalities in proprioception and postural sway suggest impaired control, and the presence of altered kinetics and kinematics are sometimes theorized to be the result of neuromuscular dysfunction. For idiopathic scoliosis, where there is no known underlying cause of the deformity, analysis of muscle activation patterns may provide insight into the origin or progression of the scoliosis pathology. Adolescents with idiopathic scoliosis demonstrate increased duration of muscle activity throughout the gait cycle. The gluteus medius, quadratus lumborum, semitendinosus, and erector spinae muscles have been shown to be active for an additional 10% of the gait cycle on average, when compared to healthy walking (Mahaudens et al. 2009; Fig. 5). This phenomenon is observed for muscles on both the convex and concave sides (Mahaudens et al. 2005, 2009, 2010). Furthermore, the increased demand on the

1300

E.A. Rapp and P.G. Gabos Erector Spinae

Quadratus lumborum

Gluteus medius

Rectus femoris

Healthy Subject Cobb angle < 20° Cobb angle [20-40°] Cobb angle > 40°

Healthy Subject Cobb angle < 20° Cobb angle [20-40°] Cobb angle > 40° 0

20

40

60

80

100

Stride (%) Semitendinosus Healthy Subject Cobb angle < 20° Cobb angle [20-40°] Cobb angle > 40°

0

20

40 60 Stride (%)

80

100

Fig. 5 Typical trace of electromyographic activity of quadratus lumborum, erector spinae, gluteus medius, rectus femoris, and semitendinosus for a scoliosis patient from each scoliosis group compared to a normal subject, expressed as a function of normalized stride (expressed in %). The horizontal black bars represent the phasic activity of the muscles for the right side of the normal subject and convex side of patients. The horizontal gray bars represent the phasic activity of the muscles for the right side of the normal subject and concave side of patients (Reproduced from Mahaudens et al. 2009 with permission of Springer)

muscles raises the energetic cost of walking in scoliosis and decreases the efficiency of gait, as measured by total muscle work divided by the net energy cost (Mahaudens et al. 2010). It is unknown whether this prolonged activation is a result of neurological dysfunction or is simply a compensatory mechanism to maintain stability during

Impact of Scoliosis on Gait

1301

walking. If the latter, the abnormalities in muscle activity would be considered a result of the deformity rather than part of its cause. However, if this were the case, the duration of muscle activity would likely be related to the severity of the scoliotic curvature, and the effect would be exacerbated in more extreme curves. Instead, the prolonged activation is observed similarly in all levels of curve severity, even those with only minor scoliosis (Mahaudens et al. 2009). This finding reduces the likelihood that the gait pathology is entirely a consequence of the spine and pelvis deformities, again suggesting a neuromuscular contribution.

Relationship of Gait Parameters to Curve Severity Many studies have investigated the relationship of the gait pathology to the severity and type of scoliotic curve. With patients walking at a fixed speed, Mahaudens et al. found no significant relationship of any of the aforementioned gait abnormalities to the degree of curvature; however, they did observe trends suggesting transverse pelvic motion decreases with greater curve severity (2009). In another large study of patients with thoracolumbar curves, where patients were permitted to walk at a self-selected speed, several gait variables did appear to be related to curve severity. Knee flexion at initial contact increased with the degree of deformity, while knee range of motion decreased. Additionally, reductions in cadence and pelvic range of motion were more pronounced in patients with more severe curves (Syczewska et al. 2012). Evidence of the relationship between kinetic abnormalities and curve severity is mixed as well. Schizas et al. determined there was no association between ground reaction force asymmetry and degree of spinal deformity (1998). However, in a study that considered multiple types of curvature, Park et al. established relationships between ground reaction force asymmetry and the severity of spinal curvature and pelvic tilt (2016). Amid conflicting findings, this relationship of the gait pathology to curve severity continues to be of great interest for clinicians. If the gait pathology is related to curve severity, then correction of the deformity, such as surgical or therapeutic interventions, may influence gait outcomes.

Response to Treatment In theory, it seems likely that interventions for scoliosis would affect performance in walking, though one might argue these effects could be either beneficial or deleterious. The most common treatments, orthotic bracing and surgical fusion of the spine, are partially aimed at restoring structural symmetry. However, both bracing and surgery impose a rigidity on the torso, which could exacerbate the preexisting stiffness during gait. The following section discusses the effect of both types of treatment on motion of the spine and distal segments during walking.

1302

E.A. Rapp and P.G. Gabos

Fig. 6 Common bracing options for adolescents with idiopathic scoliosis: Boston brace (left) and Wilmington brace (right) (Courtesy of Dr. Peter Gabos, Nemours/ Alfred I. duPont Hospital for Children, Wilmington, DE)

Response to Bracing Bracing treatment is a common nonsurgical intervention for idiopathic scoliosis, but rarely results in any curve improvement. The treatment regimen typically requires the patient to wear a customized orthosis between 12 and 20 h a day. Various types of braces exist, some spanning the cervical spine to the sacrum, some shorter, some rigid, and some more flexible (Fig. 6). In the short term, both rigid and flexible braces appear to reduce motion of the hip and pelvis during gait (Wong et al. 2008). In contrast, after long-term orthotic treatment (6 months) and a substantial period between removing the brace and gait testing, treatment effects show increased pelvis and hip motion in the frontal plane (Mahaudens et al. 2014). As previously established, untreated scoliosis patients demonstrate restricted pelvis and hip motion compared to their healthy peers, and thus, an increase in these parameters represents progress toward a more normal pattern of walking. Additionally, a decrease in the abnormally high duration of activity of the erector spinae throughout the gait cycle is observed following long-term orthotic treatment (Mahaudens et al. 2014). In untreated patients, the excessive activity is theorized to provide a stiffening effect for balance, which may no longer be necessary following prolonged brace-wearing. Still, the energy cost of walking, which is elevated in scoliosis, does not appear significantly reduced after long-term bracing treatment.

Response to Surgery Surgical treatment in scoliosis is typically reserved for severe curves (those exceeding 50 ) (Weinstein et al. 2003). The most common technique, spinal fusion, involves the insertion of metal screws into vertebral pedicles and attachment of a rod spanning the length of the curve. This results in an immediate straightening and

Impact of Scoliosis on Gait

1303

Fig. 7 Example of a 14-year-old girl with a combined right thoracic and left lumbar (Lenke 2C-[R]) curve operated on with an all pedicle screw construct. (a) Preoperative and (b) last follow-up frontal radiographs; (c) preoperative and (d) last follow-up sagittal radiographs (Courtesy of Dr. Peter Gabos, Nemours/Alfred I. duPont Hospital for Children, Wilmington, DE)

de-rotation of the spine (Fig. 7) with patients typically returning to full activity by 6 months (Lehman et al. 2015). Fusion surgery imposes a restriction of spinal range of motion in all three planes, the extent of which depends on the number of vertebrae involved in the fusion (Danielsson et al. 2006; Engsberg et al. 2003). While the reduction in spinal range of motion could theoretically exacerbate the stiffness observed in the gait of untreated scoliosis patients, it is believed that the structural correction can reduce energy demands, thereby increasing muscle efficiency. Results vary by the type of scoliotic curve. Surgical correction of thoracic curves seems to have little effect on most gait variables. The main result is a reduction of transverse plane shoulder motion (Mahaudens et al. 2010), which essentially corrects the asymmetrical forward rotation of the trunk and shoulders described by Kramersde Quervain et al. (2004). For thoracolumbar and lumbar curves, frontal plane pelvis and hip motion increases postoperatively. These results are similar to the long-term effects of bracing and represent a normalization of the motion of these segments during gait (Mahaudens et al. 2010). Additionally, there is a reduction in lateral center of mass displacement (i.e., sway) during walking, demonstrating potential evidence of better dynamic stability postsurgery (Paul et al. 2014). Fusion surgery does not appear to effect significant changes in muscular work or muscle activation timing. There may be a trend toward reduction of energy cost; however, the differences are not significant. Even postsurgery, adolescents with

1304 4

Energy cost (J kg-1 m-1)

Fig. 8 Total energy cost. The mean (vertical bar chart) SD (vertical bars) are drawn in presurgery condition (white bar) and in postsurgery condition (gray-lined bar). The black bar represents the mean of norms (Adapted from Mahaudens et al. 2010 with permission of Springer)

E.A. Rapp and P.G. Gabos

2

0 AIS pre surgery

AIS post surgery

Norms

idiopathic scoliosis still walk with increased energy cost when compared to their healthy peers (Mahaudens et al. 2010; Fig. 8). Overall, while some abnormalities remain, treatment for scoliosis generally results in modified kinematics that better resemble motion in healthy walking, specifically increased motion of the hip and pelvis. Additionally, bracing treatment reduces excessive muscle activity, and surgical treatment appears to slightly improve efficiency of walking by reducing overall energy cost.

Summary and Conclusion Gait performance in scoliosis has been heavily researched, with most investigations reporting some abnormalities compared to healthy walking. Primary observations include reduced trunk and pelvic motion within the frontal and transverse planes, often reported clinically as a “stiffness” in ambulation. Various reports of asymmetry exist, the most consistent finding an uneven progression of the trunk in the transverse plane, with excessive forward rotation of the right shoulder throughout the gait cycle. There is some evidence of postural control deficits and asymmetries in lower limb kinematics and joint moments; however, these results vary throughout the literature. Muscle activity appears increased in duration throughout the gait cycle, and energy cost of walking is higher when compared to healthy adolescents. Treatment for scoliosis, both orthotic and surgical, has a positive effect on gait variables. Shoulder, pelvis, and hip motion improve toward a more normal pattern, and center of mass displacement is reduced. Furthermore, while overall levels are still higher than normal, treatment also alleviates some of the energy cost of walking for patients with scoliosis. Despite the extensive research, conclusions about how gait abnormalities relate to the origin or progression of scoliosis remain vague and often conflicted. Impaired gait does appear to be associated with idiopathic scoliosis. Still, the ideas that the gait pathology contributes to curve progression and that impaired gait and the spinal

Impact of Scoliosis on Gait

1305

deformity are both secondary to some underlying neurological disorder are still largely based in theory. Continued research into motor control and somatosensory function may provide more insight into neurological relationships between the deformity and gait performance. In the meantime, gait analysis can still provide a valuable assessment of global function and evaluation of therapeutic and surgical outcomes in scoliosis.

References Ascani E et al (1986) Natural history of untreated idiopathic scoliosis after skeletal maturity. Spine 11(8):784–789 Asher MA, Burton DC (2006) Adolescent idiopathic scoliosis: natural history and long term treatment effects. Scoliosis [Online] 1. Available at: http://www.ncbi.nlm.nih.gov/pmc/arti cles/PMC1475645/. Accessed 12 Feb 2015 Barrack RL et al (1984) Proprioception in idiopathic scoliosis. Spine (Phila Pa 1976) 9(7):681–685 Chen P-Q et al (1998) The postural stability control and gait pattern of idiopathic scoliosis adolescents. Clin Biomech 13(1):S52–S58 Danielsson AJ, Romberg K, Nachemson AL (2006) Spinal range of motion, muscle endurance, and back pain and function at least 20 years after fusion or brace treatment for adolescent idiopathic scoliosis: a case-control study. Spine (Phila Pa 1976) 31(3):275–283 Engsberg JR et al (2003) Prospective comparison of gait and trunk range of motion in adolescents with idiopathic thoracic scoliosis undergoing anterior or posterior spinal fusion. Spine (Phila Pa 1976) 28(17):1993–2000 Giakas G et al (1996) Comparison of gait patterns between healthy and scoliotic patients using time and frequency domain analysis of ground reaction forces. Spine (Phila Pa 1976) 21(19): 2235–2242 Grivas TB et al (2010) Brace technology thematic series: the dynamic derotation brace. Scoliosis [Online] 1. Available at: http://www.ncbi.nlm.nih.gov/pubmed/20858270. Accessed 18 Feb 2015 Gum JL et al (2007) Transverse plane pelvic rotation in adolescent idiopathic scoliosis: primary or compensatory? Eur Spine J 16(10):1579–1586 Herzog W et al (1989) Asymmetries in ground reaction force patterns in normal human gait. Med Sci Sports Exerc 21(1):110–114 Konieczny MR, Senyurt H, Krauspe R (2013) Epidemiology of adolescent idiopathic scoliosis. J Child Orthop 7:3–9 Kramers-de Quervain IA et al (2004) Gait analysis in patients with idiopathic scoliosis. Eur Spine J 13(5):449–456 Lehman RA et al (2015) Return to sports after surgery to correct adolescent idiopathic scoliosis: a survey of the Spinal Deformity Study Group. Spine J 15(5):951–958 Mahaudens P, Thonnard JL, Detrembleur C (2005) Influence of structural pelvic disorders during standing and walking in adolescents with idiopathic scoliosis. Spine J 5(4):427–433 Mahaudens P et al (2009) Gait in adolescent idiopathic scoliosis: kinematics and electromyographic analysis. Eur Spine J 18(4):512–521 Mahaudens P et al (2010) Gait in thoracolumbar/lumbar adolescent idiopathic scoliosis: effect of surgery on gait mechanisms. Eur Spine J 19(7):1179–1188 Mahaudens P et al (2014) Effect of long-term orthotic treatment on gait biomechanics in adolescent idiopathic scoliosis. Spine J 14(8):1510–1519 Mallau S et al (2007) Locomotor skills and balance strategies in adolescents idiopathic scoliosis. Spine (Phila Pa 1976) 32(1):E14–E22 Mayo NE et al (1994) The Ste-Justine adolescent idiopathic scoliosis cohort study. Part III: back pain. Spine (Phila Pa 1976) 19(14):1573–1581

1306

E.A. Rapp and P.G. Gabos

Park HJ et al (2015) Analysis of coordination between thoracic and pelvic kinematic movements during gait in adolescents with idiopathic scoliosis. Eur Spine J 25:385–393 Park YS et al (2016) Association of spinal deformity and pelvic tilt with gait asymmetry in adolescent idiopathic scoliosis patients: investigation of ground reaction force. Clin Biomech 36:52–57 Paul JC et al (2014) Gait stability improvement after fusion surgery for adolescent idiopathic scoliosis is influenced by corrective measures in coronal and sagittal planes. Gait Posture 40(4):510–515 Perry J, Burnfield JM, Cabico LM (2010) Gait analysis: normal and pathological function, 2nd edn. SLACK, Thorofare Prince F et al (2010) Comparison of locomotor pattern between idiopathic scoliosis patients and control subjects. Scoliosis [Online] 1. Available at: http://www.ncbi.nlm.nih.gov/pmc/articles/ PMC2938665/. Accessed 1 Mar 2015 Saji M, Upadhyay S, Leong J (1995) Increased femoral neck-shaft angles in adolescent idiopathic scoliosis. Spine (Phila Pa 1976) 20(3):303–311 Schizas CG et al (1998) Gait asymmetries in patients with idiopathic scoliosis using vertical forces measurement only. Eur Spine J 7(2):95–98 Schlösser TPC et al (2014) How “idiopathic” is adolescent idiopathic scoliosis? A systematic review on associated abnormalities. PLoS One [Online] 9(5). Available at: http://www.ncbi. nlm.nih.gov/pmc/articles/PMC4018432/. Accessed 14 Jan 2015 Schwender JD, Denis F (2000) Coronal plane imbalance in adolescent idiopathic scoliosis with left lumbar curves exceeding 40 degrees: the role of the lumbosacral hemicurve. Spine (Phila Pa 1976) 25(18):2358–2363 Syczewska M et al (2012) Influence of the structural deformity of the spine on the gait pathology in scoliotic patients. Gait Posture 35(2):209–213 Weinstein SL et al (2003) Health and function of patients with untreated idiopathic scoliosis: a 50-year natural history study. JAMA 289(5):559–567 Wong MS et al (2008) The effect of rigid versus flexible spinal orthosis on the gait pattern of patients with adolescent idiopathic scoliosis. Gait Posture 27(2):189–195 Yang JH, Suh SW et al (2013) Asymmetrical gait in adolescents with idiopathic scoliosis. Eur Spine J 22(11):2407–2413

Concussion Assessment During Gait Robert D. Catena and Kasee J. Hildenbrand

Abstract

The acute signs and symptoms (SS) of a concussion can vary widely between individuals. Clinicians currently use a variety of measures to diagnosis and manage both physical and cognitive SS associated with concussion. Balance is typically assessed using quick sideline measures in sports; however, researchers have found through more thorough assessments of dynamic balance during gait that SS may persist beyond those detected through typical assessment techniques. An appropriate gait assessment of concussion must be adequately complex to distinguish persistent balance deficits, but not so complex that healthy individuals would be challenged to maintain balance. A steady-state gait assessment may indicate conservative gait adaptations but will seldom yield distinct signs of continued dysfunction following concussion. Obstacle avoidance tasks demonstrate conservative gait adaptations long after other SS have resolved. Concussion typically results in balance deficits in divided attention dual-task paradigms even after a return to normal daily activities. Refinements of gait paradigms to be more specific and clinically useful define future advances in concussion assessment during gait. Keywords

Concussion assessment • Balance • Steady-state gait • Obstacle avoidance • Dual task • Balance deficits

R.D. Catena (*) Gait and Posture Biomechanics Lab, Washington State University, Pullman, WA, USA e-mail: [email protected] K.J. Hildenbrand Athletic Training Program, Washington State University, Pullman, WA, USA e-mail: [email protected] # Springer International Publishing AG, part of Springer Nature 2018 B. Müller, S.I. Wolf (eds.), Handbook of Human Motion, https://doi.org/10.1007/978-3-319-14418-4_69

1307

1308

R.D. Catena and K.J. Hildenbrand

Contents State of the Art . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . What is a Concussion? . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Current Clinical Considerations in Concussion Management . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Steady-State Gait Assessment . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Functional Gait Assessment . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Dual-Task Gait Assessment . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Future Directions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

1308 1308 1310 1311 1315 1316 1319 1319

State of the Art The state of concussion management and research has exploded over the last decade, with many different disciples using their unique lens to examine the issue. Clinicians may take the patient-centered perspective that focuses on specific symptoms, daily activities, and quality of life. Researchers use broad group comparisons to determine causation, correlations, diagnosis, and rehabilitative techniques. Each clinical and research discipline currently has its own criteria, which can make cross-discipline comparisons difficult. Historically, gait researchers have found significant differences when examining subjects with “mild” concussions versus “severe” concussion, while clinical professionals have now moved away from mild and severe dichotomization. Comparison of research between groups and between older and newer research is difficult with different groups occasionally changing the criteria by which they grade severity, measure symptoms, treat, and clear for activity. Gait analysis has the benefit of detecting changes from a concussion that will be present after other criteria have been used to return individuals to participation. Analysis of gait also allows treatment for patients to improve overall balance, movement efficiency, and overall quality of life. Currently many gait assessment techniques involve expensive and complex equipment, but strides are being made to use more widely available measures that can be implemented in a diagnostic or rehabilitative manner.

What is a Concussion? The Zurich Consensus Statement defines concussion as “a complex pathophysiological process affecting the brain, induced by traumatic biomechanical forces” (McCrory et al. 2013). The consensus statement goes further to indicate several common features that incorporate clinical, pathological, and biomechanical injury constructs, which can be helpful in discussing the nature of the concussion: 1. Concussion may be caused either by a direct blow to the head, face, or neck or a blow elsewhere on the body with an “impulsive” force transmitted to the head. 2. Concussion typically results in the rapid onset of short-lived impairment of neurologic function that resolves spontaneously.

Concussion Assessment During Gait

1309

3. Concussion may result in neuropathological changes, but the acute clinical symptoms largely reflect a functional disturbance rather than a structural injury. 4. Concussion results in a graded set of clinical symptoms that may or may not involve loss of consciousness. Resolution of the clinical and cognitive symptoms typically follows a sequential course. In a small percentage of cases, however, post-concussive symptoms may be prolonged. 5. No abnormality on standard structural neuroimaging studies is seen in concussion. The acute clinical symptoms of a concussion vary widely between individuals, complicating a diagnosis. This variation between individuals is one of the most difficult problems clinicians and researchers face. Concussions differ from other injuries because often the symptoms are vague and no clear indicators exist to determine if an injury occurred. Common indicators for a concussion are physical signs (loss of consciousness), behavioral changes (irritability), cognitive impairment (slowed reaction time), or sleep disturbances (drowsiness) (McCrory et al. 2009). Concussion as defined above, along with the common symptoms, is the information medical professionals must rely on to diagnose and guide treatment. Immediately after a head impact occurs, there is a release of the excitatory amino acid neurotransmitters glutamine and aspartate (Grady 2010). These molecules lead to a loss of cell membrane integrity in the brain, which increases the concentration of sodium ions and decreases the concentration of potassium ions within cells. These changes in ion concentration affect the brain cells’ pH and lead to an increase in calcium ion concentration. These disturbances in the ion concentration lead to cellular damage resulting in the death of the affected brain cells (Grady 2010). Upon cell death, these brain cells release cytokines. Cytokines are responsible for the body’s inflammatory response, and they upregulate inflammation upon release within cells. This increase in inflammation is what is observed in the brain following a concussion injury and is responsible for added damage to the affected brain region (Grady 2010). The brain’s response to a concussion can be thought of in two separate parts. Initially, there is cellular damage resulting from the changes in ion concentration, such as sodium and potassium. These changes are responsible for the acute symptoms of concussion such as headaches and dizziness (Grady 2010). The second part is the inflammatory response caused by the release of cytokines within brain cells after a concussion. The timeline of this inflammation is the reason concussion symptoms often worsen in the 6–24 h post-injury (Grady 2010). The concussions that cause these chemical reactions are commonly referred to as mild traumatic brain injuries (mTBIs). Mild traumatic brain injury is technically not considered synonymous with concussion by many experts, but mTBI is often seen in the literature as a replacement for concussion. Research on concussions and mTBI has rapidly increased over the last decade due to specific findings that concussions may lead to degenerative brain diseases and cognitive impairments later in life. Sports-related concussions (SRC) within an athletic population have been estimated between 1.1 and 1.9 million annually, while estimates range from 22.5% to

1310

R.D. Catena and K.J. Hildenbrand

52.7% of concussions likely not reported to a healthcare professional (Bryan et al. 2016). Evaluating for the incidence of concussions has an added challenge, as many athletes choose not to report their symptoms due to the restrictions from activity when suffering from a diagnosed concussion. Challenges in determining specific incidence rates of concussions also occur due to changes in the previous definition of SRC and difficulty in establishing appropriate reporting of SRC to coaches and healthcare professionals. Theye and Mueller (2004) state 20% of head injuries (>300,000) are sports-related concussions. A study has shown that emergency department (ED) visits for SRC have risen over 200% from 1997 to 2007, indicating a considerable increase in incidence (Schatz and Moser 2011). This growth can be attributed to both an increase in participation in sports and progression in knowledge of signs and symptoms of the condition, resulting in improved reporting of the condition (Register-Mihalik, et al. 2013). Another factor in the increase in reporting is the growth in awareness of concussions in the media, resulting in greater appreciation of its severity by the healthcare community.

Current Clinical Considerations in Concussion Management Certain licensed medical professionals such as athletic trainers in North America and medical doctors can both diagnose and determine qualifications for return to activity for those who have been afflicted with head trauma. One of the most widely used systems to diagnose concussions is the SCAT3 or Sport Concussion Assessment Tool – 3rd Edition (McCrory et al. 2013). The SCAT3 includes a system of questions that the afflicted person must answer. The person’s score at the end of the test is then compared to a baseline score or normative values if a baseline was not assessed. The baseline assessment should be conducted before the athlete experiences a concussion, since some athletes may already experience some symptoms naturally, such as having poor balance. The fewer the number of symptoms the athlete has, the lower the overall score on the SCAT3. Returning an athlete to activity requires a multifaceted approach with several measures of symptoms, cognitive function, balance, and ocular function all returning to baseline or “normed” values. The SCAT3 evaluates subjective symptoms as well, such as mood and nausea. Cognitive information is included, such as the current month, day, and year. The SCAT3 evaluates the athlete’s memory of sequences of words and numbers. Balance is also evaluated using the BESS test, or Balance Error Scoring System, where the athlete must balance on both feet, one foot, and then tandem stance. Another cognitive assessment tool commonly used to evaluate concussions is ImPACT or Immediate Post-Concussion Assessment and Cognitive Testing (Covassin et al. 2009). Many universities and some high schools pay to have access to this system, since it can track athletes’ records. Athletes should take the ImPACT as a baseline before the season begins to get a score, which is compared to normative ranges. Then, after an athlete sustains a concussion and his or her symptoms have subsided, the athlete retakes the ImPACT in order to gain an objective score of his or her cognitive function.

Concussion Assessment During Gait

1311

Even with these diagnostic tools, clinicians are expected to follow a strict progression when returning afflicted athletes to sport participation (McCrory et al. 2013). Return-to-play progression may not begin until the athlete’s symptoms have returned to baseline and subsided. At this point, an athlete can begin to move through stages of light exercise, then progress to sport-specific activities, noncontact training, and integration into practice, and finally return to play. However, 24 h must pass between each stage, and if the athlete experiences any reoccurrence of symptoms, he or she must return to the previous stage. The extreme care taken with traumatic brain injuries may seem extensive, but these strict regimens are necessary. If athletes return to sport participation with unresolved symptoms, the results can be irreversible, like with second impact syndrome (Cantu 1998). This syndrome occurs when an individual suffering postconcussive symptoms participates in activity and receives a second head trauma. At this point, major brain swelling occurs and death is a possible outcome with a 50% mortality rate. The cautious approach to returning an individual to activity is also used to minimize lingering symptoms past when normal resolution of symptoms should occur. Some medical professionals labeled this lingering of symptoms as postconcussion syndrome, though other professionals argue with the diagnosis. The important point for this discussion is that balance is often a symptom that can have lasting issues after other symptoms have resolved. Typically balance is clinically assessed quickly using a modified version of the BESS test or Romberg (standing on one leg with arms out to a T and eyes closed). Gait can be used to assess balance as well, but is not part of a typical concussion management program, especially within high school or college athletic programs. Specialized clinics for patients who have longer-lasting symptoms may use gait to detect differences beyond simple balance tests, and gait analysis can sometimes detect differences well beyond when an athlete may have returned to participation.

Steady-State Gait Assessment The most common form of gait assessment is a steady-state gait analysis. Gait requires a series of coordinating interjoint and interlimb movements to achieve movement of the center of mass while also balancing that center of mass appropriately to avoid a fall. Strength, reaction time, and coordination highlight some of the important motor components that are required during both of these two tasks in gait. Since balance incorporates many neuromuscular and neuropsychological components that can be affected by brain injury (Fig. 1), balance assessment is one of the most commonly performed and effective motor assessments following concussions. Balance deficits following concussion were widely recognized in the 1990s when about half of all surveyed concussed individuals (of all severity levels) reported dysfunction due to their injury in a 5 year post-TBI survey and physician assessment study appraising balance impairment (Hillier et al. 1997). This, along with some

1312

R.D. Catena and K.J. Hildenbrand

Fig. 1 A conceptual model of the postural control system (Reproduced from Maki and McIlroy 1996)

high-profile cases of concussions to children and famous athletes, paved the way for concussion balance research through the next decade. As indicated in Fig. 1 (Maki and McIlroy 1996), balance control can be modulated by a number of different factors. Concussion affects balance at the central nervous system level but can be measured through gait assessment of the mechanical output. Research findings are mixed for effects at any particular sensory system in this balance pathway. Some research indicates no particular deficit in any one sensory input to balance following traumatic brain injury, but still a level of imbalance corresponding with injury severity (Mrazik et al. 2000). Others suggest vestibular dysfunction resulting from concussion (Aligene and Lin 2013; Alsalaheen et al. 2010; Corwin et al. 2015; Fife and Kalra 2015; Murray et al. 2014). While sensory organization tests are a typical method for isolating the effect of vestibular dysfunction on balance, dysfunction can be evident through, and must be accounted for, in clinical tests involving balance such as a gait assessment. When performing gait, a kinematic asymmetry may be most easily detectable with unilateral vestibular dysfunction. Others have used more intricate measures of center of mass motion to detect dysfunction (Deshpande and Patla 2005). Stability, posture, and balance are sometimes considered synonymous. For our purposes, we will differentiate these terms to clearly define the research from here on. “Posture” is an instantaneous pose of the body and the many joint positions that create that pose. “Balance” is the instantaneous measure of the propensity to fall through measures of the body center of mass with respect to a fulcrum (center of pressure) or base of support (area contained within the feet). “Stability” represents the consistency of a repeated cyclical action, and in the case of postural stability, that

Concussion Assessment During Gait

1313

is the consistency of body motions at the joint level or of the whole body through a measure of the center of mass or more commonly measured center of pressure over a time period. These are referred to as “nonlinear analyses.” The interpretation of consistency versus randomness in movements in healthy human motion versus deficient motion is being explored by a number of research groups, and so the clinical applications of these nonlinear measurements are yet to be fully understood. We suggest that the reader consults additional sources such as Cavanaugh et al. (2005) about how nonlinear measurement techniques can provide clinically relevant information particular to different populations. Standing balance tests represent some of the earliest motor tests conducted in a concussed population to detect persistent balance deficits. More severe forms of brain injury will manifest symptoms in basic standing tests. Balance tests have indicated increased postural sway following concussions compared to healthy balance in clinical (Geurts et al. 1996; McCrea et al. 2003) and functional testing (Zhang et al. 2002). These tests will typically only detect immediate symptoms; however, individuals that have a history of concussive events can present persistent short-term (Gao et al. 2011; Quatman-Yates et al. 2015) and long-term (De Beaumont et al. 2011, 2013; Sosnoff et al. 2011) postural stability irregularities compared to healthy individuals. It typically takes more demanding motor tests to detect any balance symptoms from more mildly concussed individuals. In conducting a gait assessment, it is proper to first consider your goal. One goal may be to simply detect lingering motor (or other) performance symptoms that indicate that the concussed individual is not completely healthy (De Beaumont et al. 2011). Alternatively, your goal may be to detect particular lingering symptoms that could affect performance in a functional activity. Some forms of gait assessment may satisfy both of these goals but with reduced power. The more mildly concussed your group or individual is, the more precise you need your gait assessment to be. Secondly, you must develop an appropriate gait assessment considering the sensitivity and specificity needed to distinguish lingering symptoms for a particular level of concussion severity and duration since the concussive event. Since the persistence of symptoms is in question, the appropriate gait assessment must be robust enough to examine a range of likely symptomology. It includes an appropriate range of difficulties and considers how performance may be influenced by assessment duration, assessment time, and assessment environment. Balance is a measure of the center of mass with respect to the base of support. Most tests of balance in a clinical setting are either standing balance, which can be measured accurately without a direct measure of the center of mass, or involve costly tools that can better predict the center of mass motion. The challenge with a balance measure during gait is that a simpler measure, like how center of pressure is typically used in standing tests, doesn’t as accurately predict center of mass motion in this more dynamic task. The benefit of being able to measure balance during gait is that it provides the best indication of likelihood of a fall or other loss-of-balance injuries. Subsequent injury following concussions can be in the form of physical injury to the body (Brooks et al. 2016) and the potential for additional concussive events that compound the deleterious effects (CDC 1997).

1314

R.D. Catena and K.J. Hildenbrand

Balance in gait additionally highlights different components of balance in Fig. 1 and, to an amplified level, compared to standing balance tests. Volitional movements and feedforward corrections are more prevalent in gait than they are in standing; thus, there is more reliance on the cerebral cortex, the very component concussion affects. Coordination of the musculoskeletal linkage and processing visual and somatosensory information are also more changing in gait than in standing. Gait is a more unbalancing task than standing due to these factors and therefore may be a more appropriate test of balance for mildly concussed individuals or individuals further into their rehabilitation from a severe concussion. Even though balance is a direct measure of the center of mass with respect to the base of support, there is no perfect way to measure center of mass location in the human body during dynamic tasks. With a full-body motion capture system and previously established anthropometric models of segment inertial parameters, you can estimate the center of mass location changes during gait. Use of anthropometry to aid in determining the body center of mass doesn’t quite provide perfectly accurate information; however, the young adult concussed population (less so for adolescent concussed individuals) has plenty of anthropometry studies to validate their use, and there are methods by which you can optimize the data to better fit a specific individual (Pavol et al. 2002). A motion capture system can also track the base of support so that measures of the center of mass with respect to the base of support can measure balance. Without a motion capture system for identifying center of mass location, force plates can be used to estimate the center of mass motion. The projection of the center of mass onto the ground (center of gravity) can be used for several measures of balance and measured by double integration of force plate data (Zatsiorsky and King 1998). The problem with using the center of gravity is that it doesn’t account for three-dimensional motion of the center of mass, which can affect balance. Center of pressure alone (collected by force plate or pressure mat) has also been used to estimate gait balance. The center of gravity is encompassed and controlled by the center of pressure and ground reaction force during steady state gait, so center of pressure will always overestimate the actual motion of the center mass. Outside of this level of equipment sophistication, spatiotemporal measures of movement have been used in the past as measures of balance in gait but are several levels away from being true measures of balance. Research into balance assessment during gait following concussions began from several groups in the early 2000s. Mayo Clinic researchers used clinical assessments, standing posturography (with the sensory organization test), and gait analysis to compare assessment techniques following traumatic brain injury (TBI) (Basford et al. 2003; Kaufman et al. 2006). They demonstrated a high correlation between physical impairments described by the TBI participants and reduced scores on the sensory organization test designed to isolate vestibular input. Physical disability scores also correlated with reduced center of mass motion in the anterior/posterior direction. Functional disability index correlated with increased center of mass motion in the mediolateral direction. TBI participants also showed reduced gait A/P motion and increased gait M/L motion compared to healthy controls. This

Concussion Assessment During Gait

1315

analysis indicates the significance of a gait analysis in assessments of perceived impairments in a more severe TBI group. For the more mildly concussed adult, a simple gait test doesn’t clearly discern persistent balance deficits. Within 48 h of a concussive event, adults will display a slowed gait, but no indications of balance deficits (Parker et al. 2005). By a week post-concussion, single-task gait is indistinguishable to healthy adults just like typical neuropsychological tests (Parker et al. 2006). Concussed adolescents however are more likely to present with balance deficits in single-task gait (Howell et al. 2013b). It is important to note that high-impact athletes that may frequently encounter “subconcussive” blows to the head could present gait balance deficits even without a medical diagnosis of concussion (Parker et al. 2008). To accurately detect persistent balance symptoms in a clinical setting, it is important to consider baseline performance of your patient. In research, baseline information is ideal, but at minimum this previous research indicates that athletic participation is important to consider in group comparisons.

Functional Gait Assessment Physical obstacles that must be negotiated during daily gait include curbs, stairs, unstable surfaces, traffic, and a variety of other obstacles. Our ability to reorient our attention to these particular obstacles factors into gait performance (Catena et al. 2009b). Concussed individuals demonstrate deficits in spatially orienting attention in both auditory and visual tasks (Breton et al. 1991; Cremona-Meteyard et al. 1992; Daffner et al. 2000; Halterman et al. 2006). In this process, we must disengage, shift, and reengage attention (Posner 1980) through unique neuronal pathways. Broad posterior parietal lobe damage has been linked to disengagement of attention (Posner et al. 1984). The superior parietal gyrus is linked to shifting attention (Vandenberghe et al. 2001). The intraparietal sulcus is involved in shifting and refocusing attention (Yantis et al. 2002) but also the superior colliculus and lateral pulvinar when distractions are present (Posner and Petersen 1990). Obstacle avoidance tasks are a typical functional activity added to gait assessment to increase the balance complexity and make the task more indicative of everyday hazards to injury. The complexity of the obstacle crossing task can be modulated to the expected ability of your population by making the task more or less physically demanding (Chou et al. 2004) or more perceptually demanding (Baker and Cinelli 2014). Compared to clinical assessment techniques of balance, such as the Berg Balance Test, an obstacle crossing task was better in distinguishing TBI individuals from healthy controls with slower gait velocities, increased obstacle clearance, and decreased stride lengths indicating the TBI group adopted a cautious gait during obstacle crossing (McFadyen et al. 2003). Others have shown similar results from an obstacle crossing task (Fait et al. 2013; Martini et al. 2011; Vallee et al. 2006). Measures of whole body center of mass motion of TBI patients during the crossing of several different obstacle heights show that participants with TBI and healthy controls have similar gait patterns during unobstructed walking, indicating that

1316

R.D. Catena and K.J. Hildenbrand

normal level walking may not be as sensitive in detecting long-term changes in dynamic balance (Chou et al. 2004). On the other hand, obstructed walking resulted in slowed gait velocities and shorter stride lengths (indicating more cautious gait) and increased mediolateral swaying motion (indicating a lack of balance control) 2 years after a TBI (Chou et al. 2004). The balance effects of an obstacle crossing during gait are essentially equivalent between more mildly concussed individuals within 48 h of injury and healthy individuals (Catena et al. 2007a), as both groups tend to be taxed with the challenge of maintaining balance during obstacle crossing depending on the obstacle height. However, the potential for a trip is higher after a recent concussion (indicated by lower foot clearances and higher trip rates) for individuals that also have deficits in spatially orienting attention (Catena et al. 2009b). Similar to long-term performance in TBI individuals, mildly concussed individuals adopt a more conservative obstacle crossing strategy (indicated both in balance and obstacle clearance measures) as concussion symptoms subside several weeks following injury compared to their 48 h performance (Catena et al. 2009a) and compared to healthy individuals (Catena et al. 2009a; Sambasivan et al. 2015). On the other hand, balance deficits are more likely to be elicited by gait tasks that require a cognitive reaction to a suddenly presented perturbation (Powers et al. 2014).

Dual-Task Gait Assessment While broad neuropsychological tests have become the standard method for assessing persistent concussive symptoms for a clinical return-to-activity decision (Resch et al. 2013), it is important to note that widely used neuropsychological clinical tests don’t measure all cognitive deficits to the same degree (Choe and Giza 2015) and broad tests of cognition don’t always present the same findings, as there are specific tests that focus on particular cognitive components (Keightley et al. 2009). Nevertheless, there is the potential to refine current cognitive testing to become an even better measure of persistent symptoms. In doing so, the chance that a patient is involved in another deleterious subsequent concussive event is reduced. One area to improve current testing is to focus on enhancing testing methodologies that correlate with (to potentially predict) motor performance as concussion-induced motor deficiencies could result in subsequent injuries (Brooks et al. 2016; Herman et al. 2015). Unlike many neurological pathologies, concussions don’t present any consistently localized cognitive symptoms. Instead, axonal injury is diffuse, and so are cognitive symptoms. Cognitive processing distribution throughout the brain obviously leads to questions about the probability of any one component affected by a single biomechanical force with an intricate direction, magnitude, and point of application. Cognitive deficits could include (but not necessarily limited to or necessarily include in specific individuals) executive dysfunction, slower reaction times, decreased focus, reduced working memory, reduced attention capacity, and inability to shift attention.

Concussion Assessment During Gait

1317

Gait is not an automated task in which no attentional resources are needed. Decreased attention capacity has been eluded to as a major determinant in reduced gait performance following a concussion (Catena et al. 2011). No matter the general theory of divided attention to which you prescribe, there is an abundance of evidence to suggest an interaction between gait performance and cognition through attention in healthy individuals (de Bruin and Schmidt 2010; Hegeman et al. 2012; Lajoie et al. 1993; Szturm et al. 2013). Accomplishing a dual task, with reasonable success in both simultaneously performed tasks, is even less automatic when challenged by a deficit that can affect performance of either task (Brown et al. 1999; Vaportzis et al. 2015; Yardley et al. 2001). Concussion, directly affecting cognition and neurophysiology, challenges an individual to complete both simple and more complex dualtask scenarios (Bernstein 2002; De Monte et al. 2005; Tapper et al. 2016; Vilkki et al. 1996). Dual-task gait research has provided even more scientific evidence of divided attention deficits. Executive control over cognitive processes allows individuals to achieve goals by planning, focusing, and coordinating actions. Executive dysfunction has been consistently reported as deficient following concussion (Hart et al. 2005; Howell et al. 2013a; Moore et al. 2016; Serino et al. 2006; Tapper et al. 2016). In particular, sustained attention [primarily controlled in the right frontal areas (Posner and Petersen 1990; Sturm et al. 1999; Wang et al. 2005; Wilkins et al. 1987)] on gait performance is important for populations at risk of fall, injury, or re-injury due to a fall. This is particularly important in gait when balance has been compromised. Sustained attention does not seem to be deficient shortly after concussion when balance is compromised (Halterman et al. 2006; van Donkelaar et al. 2005, 2006), but there is some research that indicates a positive relationship between concussion and lapses in attention long after a reported concussion occurred (Killgore et al. 2016; Pontifex et al. 2012). In sustaining attention, executive control involves resolving conflicting information. Concussed individuals experience conflict resolution deficits up to a month post-injury (Chan 2002; Chan et al. 2003; Halterman et al. 2006; Larson et al. 2011; Moore et al. 2014). The anterior cingulate cortex seems to be primarily responsible for conflict resolution in such tasks (Posner and Rothbart 1998; Swick and Jovanovic 2002) and more specifically the mid-dorsal region (Swick and Jovanovic 2002). The dorsal prefrontal cortex has also been linked to the actual selection response in more difficult tasks (MacDonald et al. 2000). Including a cognitive component to balance tests is one alternative to provide increased task complexity to tease out milder symptoms or concussion symptoms over longer duration. Distractions clearly play a role when assessing balance deficits following concussion (Rahn et al. 2015). Cognitive performance is correlated to balance performance following concussion (Alsalaheen et al. 2016). And through attention, cognitive and balance performance interacts to diminish the performance of either, or both, following a concussion (Catena et al. 2007a). Immediate balance deficits are commonly observed in a dual-task paradigm (Sosnoff et al. 2008). Month-long dual-task balance deficits are occasionally evidenced in the literature as well (Dorman et al. 2015). Cognition may also interact with other motor

1318

R.D. Catena and K.J. Hildenbrand

components, or even specific motor components used in balance, following concussion (Brown et al. 2015). Gait with a simultaneously performed cognitive task (dual task) has become the functional paradigm of interest in the concussion research over the last decade. These types of paradigms have been described as most similar to real-world conditions (Cock et al. 2003; Weerdesteyn et al. 2003) when we are often performing cognitive processing along with gait. In performing a dual-task analysis, the cognitive task can be modulated to involve or exclude particular sensory and cognitive tracts along with modulation of task complexity to fit your population. This is on top of the modulation that can be made to the gait task as described for obstacle crossing above. As such, there are a wide variety of paradigms described throughout the research literature similar to the wide variety of dual-task situations faced daily. Consider that concussion is a diffuse injury that can affect multiple areas of the brain, when picking the correct dual-task paradigm. Differences between TBI patients and healthy individuals are mixed when combining obstacle avoidance with a cognitive secondary task compared to just singletask obstacle crossing (Chiu et al. 2013; Martini et al. 2011; McFadyen et al. 2009; Vallee et al. 2006). Cognitive tasks may interfere with motor performance through peripheral sensory distraction or through central nervous system attention division in dual-task paradigms. Peripheral sensory distractions, for example, when visual cognition tracts are tasked simultaneously with an inherently visual motor task like obstacle crossing, typically want to be avoided as they are a challenge regardless of injury or injury severity (Fait et al. 2013). Along with picking the correct mode of cognitive task, it is important to consider the correct amplitude of cognitive complexity. While a visual cognition and obstacle crossing task may present as too complex for even healthy individuals, some tasks like a simple reaction time test during gait may present as too simple for mild concussion patients (Catena et al. 2007b). A continuous choice mental task has typically shown to be best at discriminating between symptomatic and asymptomatic individuals. A dual-task paradigm that includes steady-state gait and a variety of continuous mental tasks results in few changes for healthy individuals but reduced performance on cognitive tasks and increased spatial-temporal gait variability for severe TBI patients, similar to individuals after stroke and subarachnoid hemorrhage (Haggard et al. 2000). Following mild TBI, spatial-temporal variables only seem to be sensitive to conservative control of dual-task gait immediate after injury (Parker et al. 2005) and don’t seem to be sensitive enough to detect any lingering single-task gait performance deficits several weeks after concussion (Howell et al. 2013b; Parker et al. 2006; Sambasivan et al. 2015). However, continued balance deficits measured by center of mass motion can still be detected several weeks following a concussion using a dual-task paradigm (Parker et al. 2006). There is evidence to suggest that balance deficits are even more apparent following an adolescent concussion (Howell et al. 2015a) and can be prolonged by a return to activity too soon (Howell et al. 2015b, c; Parker et al. 2008).

Concussion Assessment During Gait

1319

Future Directions Published research is skewed toward positive results. It is not clear how likely it is for a concussed individual to present with deficits in a gait assessment. There is immediate need for epidemiological research to uncover how likely concussions are to cause gait deficits. This research could also account for other standard assessment techniques and, in doing so, better inform clinicians about a comprehensive assessment for concussion management. Ideally clinicians could accurately measure all cognitive and motor deficits following a concussion both quickly and cost-effectively. Unlike cognitive testing, gait assessment is neither quick nor cheap. Individual-specific results are required in a clinical setting, so baseline information and multiple testing post-concussion are crucial for diagnostic comparisons. Future research needs to advance gait assessment techniques so that they can be implemented in a quick and cost-effective manner. Mobile technology, through inertial measurement units (IMUs) or force platforms, could provide a route as technology advances for more cost-effective precise measurement advances. Research is also exploring augmented reality, which may refine our ability to provide more realistic dual-task scenarios during gait in the lab or in the clinic. Simultaneously, researchers need to direct efforts toward correlations of attention components to gait performance to refine dual-task paradigms. Concussions are a neurophysiological phenomenon currently tested via cognitive assessment, but motor deficits are often as crucial to a return to normal activity as cognitive deficits as they result in subsequent injuries (Brooks et al. 2016; Herman et al. 2015). If research advances in predicting gait performance from cognitive assessment results, more costly and time-consuming motor tests may be avoided altogether, but as of now, the perfect cognitive test to make this motor performance prediction has yet to be created. A team approach is crucial in both developing new measurement techniques and refining current techniques. The partnership between a clinician and a gait researcher in writing this chapter highlighted to us the importance for both to be equally involved in future collaborative research into concussion assessment during gait. One such difficulty is the use of terminology between the fields and roles. Clinicians have abandoned the idea of “grading” concussions or labeling them as mild versus severe, while researchers designing experiments may continue to divide subjects into those suffering from “mild” and “severe” concussions. This difference in terminology can make translation of research results to clinical practice more difficult. Work must continue with both sides collaborating together in a common language with the patients’ health and well-being as the focus.

References Aligene K, Lin E (2013) Vestibular and balance treatment of the concussed athlete. NeuroRehabilitation 32:543–553. https://doi.org/10.3233/nre-130876 Alsalaheen BA et al (2010) Vestibular rehabilitation for dizziness and balance disorders after concussion. J Neuro Phys Ther 34:87–93. https://doi.org/10.1097/NPT.0b013e3181dde568

1320

R.D. Catena and K.J. Hildenbrand

Alsalaheen BA, Whitney SL, Marchetti GF, Furman JM, Kontos AP, Collins MW, Sparto PJ (2016) Relationship between cognitive assessment and balance measures in adolescents referred for vestibular physical therapy after concussion. Clin J Sport Med 26:46–52. https://doi.org/ 10.1097/jsm.0000000000000185 Baker CS, Cinelli ME (2014) Visuomotor deficits during locomotion in previously concussed athletes 30 or more days following return to play. Physiol Reports 2. https://doi.org/10.14814/ phy2.12252 Basford JR et al (2003) An assessment of gait and balance deficits after traumatic brain injury. Arch Phys Med Rehabil 84:343–349 Bernstein DM (2002) Information processing difficulty long after self-reported concussion. J Int Neuropsychol Soc 8:673–682 Breton F, Pincemaille Y, Tarriere C, Renault B (1991) Event-related potential assessment of attention and the orienting reaction in boxers before and after a fight. Biol Psychol 31:57–71 Brooks MA, Peterson K, Biese K, Sanfilippo J, Heiderscheit BC, Bell DR (2016) Concussion increases odds of sustaining a lower extremity musculoskeletal injury after return to play among collegiate athletes. Am J Sports Med 44:742–747. https://doi.org/10.1177/0363546 515622387 Brown LA, Shumway-Cook A, Woollacott MH (1999) Attentional demands and postural recovery: the effects of aging. J Gerontol 54A:M165–M171 Brown JA, Dalecki M, Hughes C, Macpherson AK, Sergio LE (2015) Cognitive-motor integration deficits in young adult athletes following concussion. BMC Sports Sci Med Rehabil 7:25. https://doi.org/10.1186/s13102-015-0019-4 Bryan MA, Rowhani-Rahbar A, Comstock RD et al (2016) Sports-and recreation-related concussions in US youth. Pediatrics 138(1), e20154635 Cantu RC (1998) Second-impact syndrome. Clin Sport Med 17(1):37–44 Catena RD, van Donkelaar P, Chou LS (2007a) Altered balance control following concussion is better detected with an attention test during gait. Gait Posture 25:406–411 Catena RD, van Donkelaar P, Chou LS (2007b) Cognitive task effects on gait stability following concussion. Exp Brain Res 176:23–31. https://doi.org/10.1007/s00221-006-0596-2 Catena RD, van Donkelaar P, Chou L-S (2009a) Different gait tasks distinguish immediate vs. longterm effects of concussion on balance control. J Neuroeng Rehabil 6:25–25 Catena RD, van Donkelaar P, Halterman CI, Chou LS (2009b) Spatial orientation of attention and obstacle avoidance following concussion. Exp Brain Res 194:67–77 Catena RD, van Donkelaar P, Chou LS (2011) The effects of attention capacity on dynamic balance control following concussion. J Neuroeng Rehabil 8:8. https://doi.org/10.1186/1743-0003-8-8 Cavanaugh JT, Guskiewicz KM, Stergiou N (2005) A nonlinear dynamic approach for evaluating postural control: new directions for the management of sport-related cerebral concussion. Sports Med 35:935–950 CDC (1997) Sports-related recurrent brain injuries – United States. Morb Mortal Wkly Rep 46:224–227 Chan RC (2002) Attentional deficits in patients with persisting postconcussive complaints: a general deficit or specific component deficit? J Clin Exp Neuropsychol 24:1081–1093 Chan RC, Hoosain R, Lee TM, Fan YW, Fong D (2003) Are there sub-types of attentional deficits in patients with persisting post-concussive symptoms? A cluster analytical study. Brain Injury 17:131–148 Chiu SL, Osternig L, Chou LS (2013) Concussion induces gait inter-joint coordination variability under conditions of divided attention and obstacle crossing. Gait Posture 38:717–722. https:// doi.org/10.1016/j.gaitpost.2013.03.010 Choe MC, Giza CC (2015) Diagnosis and management of acute concussion. Semin Neurol 35:29–41. https://doi.org/10.1055/s-0035-1544243 Chou LS, Kaufman KR, Walker-Rabatin AE, Brey RH, Basford JR (2004) Dynamic instability during obstacle crossing following traumatic brain injury. Gait Posture 20:245–254. https://doi. org/10.1016/j.gaitpost.2003.09.007

Concussion Assessment During Gait

1321

Cock J, Fordham C, Cockburn J, Haggard P (2003) Who knows best? Awareness of divided attention difficulty in a neurological rehabilitation setting. Brain Inj 17:561–574 Corwin DJ, Wiebe DJ, Zonfrillo MR, Grady MF, Robinson RL, Goodman AM, Master CL (2015) Vestibular deficits following youth concussion. J Pediatr 166:1221–1225. https://doi.org/ 10.1016/j.jpeds.2015.01.039 Covassin T, Elbin RJ III, Stiller-Ostrowski JL et al (2009) Immediate post-concussion assessment and cognitive testing (ImPACT) practices of sports medicine professionals. J Athl Train 44(6): 639–644 Cremona-Meteyard SL, Clark CR, Wright MJ, Geffen GM (1992) Covert orientation of visual attention after closed head injury. Neuropsychologia 30:123–132 Daffner KR et al (2000) The central role of the prefrontal cortex in directing attention to novel events. Brain 123:927–939 De Beaumont L et al (2011) Persistent motor system abnormalities in formerly concussed. J Athl Train 46:234–240 De Beaumont L, Tremblay S, Henry LC, Poirier J, Lassonde M, Theoret H (2013) Motor system alterations in retired former athletes: the role of aging and concussion history. BMC Neurol 13:109. https://doi.org/10.1186/1471-2377-13-109 de Bruin ED, Schmidt A (2010) Walking behaviour of healthy elderly: attention should be paid. Behavior Brain Funct 6:59. https://doi.org/10.1186/1744-9081-6-59 De Monte VE, Geffen GM, May CR, McFarland K, Heath P, Neralic M (2005) The acute effects of mild traumatic brain injury on finger tapping with and without word repetition. J Clin Exp Neuropsychol 27:224–239. https://doi.org/10.1080/13803390490515766 Deshpande N, Patla AE (2005) Dynamic visual-vestibular integration during goal directed human locomotion. Exp Brain Res 166:237–247. https://doi.org/10.1007/s00221-005-2364-0 Dorman JC, Valentine VD, Munce TA, Tjarks BJ, Thompson PA, Bergeron MF (2015) Tracking postural stability of young concussion patients using dual-task interference. J Sci Med Sport 18:2–7. https://doi.org/10.1016/j.jsams.2013.11.010 Fait P, Swaine B, Cantin JF, Leblond J, McFadyen BJ (2013) Altered integrated locomotor and cognitive function in elite athletes 30 days postconcussion: a preliminary study. J Head Trauma Rehabil 28:293–301. https://doi.org/10.1097/HTR.0b013e3182407ace Fife TD, Kalra D (2015) Persistent vertigo and dizziness after mild traumatic brain injury. Ann N Y Acad Sci 1343:97–105. https://doi.org/10.1111/nyas.12678 Gao J, Hu J, Buckley T, White K, Hass C (2011) Shannon and Renyi entropies to classify effects of mild traumatic brain injury on postural sway. PLoS One 6, e24446. https://doi.org/10.1371/ journal.pone.0024446 Geurts AC, Ribbers GM, Knoop JA, Limbeek JV (1996) Identification of static and dynamic postural instability following traumatic brain injury. Arch Phys Med Rehabil 77:639–644 Grady M (2010) Concussion in the adolescent athlete. Curr Probl Pediatr Adolesc Health Care 40(7):154–169 Haggard P, Cockburn J, Cock J, Fordham C, Wade D (2000) Interference between gait and cognitive tasks in a rehabilitating neurological population. J Neurol Neurosurg Psychiatry 69:479–486 Halterman CI, Langan J, Drew A, Rodriguez E, Osternig LR, Chou LS, van Donkelaar P (2006) Tracking the recovery of visuospatial attention deficits in mild traumatic brain injury. Brain 129:747–753 Hart T, Whyte J, Kim J, Vaccaro M (2005) Executive function and self-awareness of “real-world” behavior and attention deficits following traumatic brain injury. J Head Trauma Rehabil 20:333–347 Hegeman J, Weerdesteyn V, van den Bemt B, Nienhuis B, van Limbeek J, Duysens J (2012) Dualtasking interferes with obstacle avoidance reactions in healthy seniors. Gait Posture 36:236–240. https://doi.org/10.1016/j.gaitpost.2012.02.024 Herman DC, Zaremski JL, Vincent HK, Vincent KR (2015) Effect of neurocognition and concussion on musculoskeletal injury risk. Curr Sports Med Rep 14:194–199. https://doi.org/10.1249/ jsr.0000000000000157

1322

R.D. Catena and K.J. Hildenbrand

Hillier SL, Sharpe MH, Metzer J (1997) Outcomes 5 years post-traumatic brain injury (with further reference to neurophysical impairment and disability). Brain Inj 11:661–675 Howell D, Osternig L, Van Donkelaar P, Mayr U, Chou LS (2013a) Effects of concussion on attention and executive function in adolescents. Med Sci Sports Exerc 45:1030–1037. https:// doi.org/10.1249/MSS.0b013e3182814595 Howell DR, Osternig LR, Chou LS (2013b) Dual-task effect on gait balance control in adolescents with concussion. Arch Phys Med Rehabil 94:1513–1520. https://doi.org/10.1016/j.apmr.2013. 04.015 Howell DR, Osternig LR, Chou LS, Chou LS (2015a) Adolescents demonstrate greater gait balance control deficits after concussion than young adults. Am J Sports Med 43:625–632. https://doi. org/10.1177/0363546514560994 Howell DR, Osternig LR, Chou LS (2015b) Return to activity after concussion affects dual-task gait balance control recovery. Med Sci Sports Exerc 47:673–680. https://doi.org/10.1249/ mss.0000000000000462 Howell DR, Osternig LR, Christie AD, Chou LS (2015c) Return to physical activity timing and dual-task gait stability are associated 2 months following concussion. J Head Trauma Rehabil 31(4):262–268. https://doi.org/10.1097/htr.0000000000000176 Kaufman KR, Brey RH, Chou LS, Rabatin A, Brown AW, Basford JR (2006) Comparison of subjective and objective measurements of balance disorders following traumatic brain injury. Med Eng Phys 28:234–239. https://doi.org/10.1016/j.medengphy.2005.05.005 Keightley M et al (2009) Paediatric sports-related mild traumatic brain injury. BMJ Case Rep. https://doi.org/10.1136/bcr.06.2008.0148 Killgore WD, Singh P, Kipman M, Pisner D, Fridman A, Weber M (2016) Gray matter volume and executive functioning correlate with time since injury following mild traumatic brain injury. Neurosci Lett 612:238–244. https://doi.org/10.1016/j.neulet.2015.12.033 Lajoie Y, Teasdale N, Bard C, Fleury M (1993) Attentional demands for static and dynamic equilibrium. Exp Brain Res 97:139–144 Larson MJ, Farrer TJ, Clayson PE (2011) Cognitive control in mild traumatic brain injury: conflict monitoring and conflict adaptation. Int J Psychophysiol 82:69–78. https://doi.org/10.1016/j. ijpsycho.2011.02.018 MacDonald AW 3rd, Cohen JD, Stenger VA, Carter CS (2000) Dissociating the role of the dorsolateral prefrontal and anterior cingulate cortex in cognitive control. Science 288:1835–1838 Maki BE, McIlroy WE (1996) Postural control in the older adult. Clin Geriatr Med 12:635–658 Martini DN, Sabin MJ, DePesa SA, Leal EW, Negrete TN, Sosnoff JJ, Broglio SP (2011) The chronic effects of concussion on gait. Arch Phys Med Rehabil 92:585–589. https://doi.org/ 10.1016/j.apmr.2010.11.029 McCrea M et al (2003) Acute effects and recovery time following concussion in collegiate football players: the NCAA Concussion Study. JAMA 290:2556–2563 McCrory P, Meeuwisse W, Johnston K (2009) Consensus statement on concussion in sport 3rd international 3rd international conference on concussion in sport held in Zurich, November 2008. Clin J Sport Med 19(3):185–200 McCrory P, Meeuwisse WH, Aubry M et al (2013) Consensus statement on concussion in sport: the 4th international conference on concussion in sport held in Zurich, November 2012. Brit J of Sport Med 47(5):250–258 McFadyen BJ, Swaine B, Dumas D, Durand A (2003) Residual effects of a traumatic brain injury on locomotor capacity: a first study of spatiotemporal patterns during unobstructed and obstructed walking. J Head Trauma Rehabil 18:512–525 McFadyen BJ, Cantin JF, Swaine B, Duchesneau G, Doyon J, Dumas D, Fait P (2009) Modalityspecific, multitask locomotor deficits persist despite good recovery after a traumatic brain injury. Arch Phys Med Rehabil 90:1596–1606. https://doi.org/10.1016/j.apmr.2009.03.010 Moore RD, Hillman CH, Broglio SP (2014) The persistent influence of concussive injuries on cognitive control and neuroelectric function. J Athl Train 49:24–35. https://doi.org/10.4085/ 1062-6050-49.1.01

Concussion Assessment During Gait

1323

Moore DR, Pindus DM, Raine LB, Drollette ES, Scudder MR, Ellemberg D, Hillman CH (2016) The persistent influence of concussion on attention, executive control and neuroelectric function in preadolescent children. Int J Psychophysiol 99:85–95. https://doi.org/10.1016/j.ijpsycho. 2015.11.010 Mrazik M, Ferrara MS, Peterson CL, Elliott RE, Courson RW, Clanton MD, Hynd GW (2000) Injury severity and neuropsychological and balance outcomes of four college athletes. Brain Inj 14:921–931 Murray NG, Ambati VN, Contreras MM, Salvatore AP, Reed-Jones RJ (2014) Assessment of oculomotor control and balance post-concussion: a preliminary study for a novel approach to concussion management. Brain Inj 28:496–503. https://doi.org/10.3109/02699052.2014.887144 Parker TM, Osternig LR, Lee HJ, Donkelaar P, Chou LS (2005) The effect of divided attention on gait stability following concussion. Clin Biomech (Bristol, Avon) 20:389–395. https://doi.org/ 10.1016/j.clinbiomech.2004.12.004 Parker TM, Osternig LR, Van Donkelaar P, Chou LS (2006) Gait stability following concussion. Med Sci Sports Exerc 38:1032–1040. https://doi.org/10.1249/01.mss.0000222828.56982.a4 Parker TM, Osternig LR, van Donkelaar P, Chou LS (2008) Balance control during gait in athletes and non-athletes following concussion. Med Eng Phys 30:959–967. https://doi.org/10.1016/j. medengphy.2007.12.006 Pavol MJ, Owings TM, Grabiner MD (2002) Body segment inertial parameter estimation for the general population of older adults. J Biomech 35:707–712 Pontifex MB, Broglio SP, Drollette ES, Scudder MR, Johnson CR, O'Connor PM, Hillman CH (2012) The relation of mild traumatic brain injury to chronic lapses of attention. Res Q Exerc Sport 83:553–559. https://doi.org/10.1080/02701367.2012.10599252 Posner MI (1980) Orienting of attention. Q J Exp Psychol 32:3–25 Posner MI, Petersen SE (1990) The attention system of the human brain. Annu Rev Neurosci 13:25–42 Posner MI, Rothbart MK (1998) Attention, self-regulation and consciousness. Phil Trans R Soc Lond B Biol Sci 353:1915–1927 Posner MI, Walker JA, Friedrich FJ, Rafal RD (1984) Effects of parietal injury on covert orienting of attention. J Neurosci 4:1863–1874 Powers KC, Kalmar JM, Cinelli ME (2014) Dynamic stability and steering control following a sportinduced concussion. Gait Posture 39:728–732. https://doi.org/10.1016/j.gaitpost.2013.10.005 Quatman-Yates CC, Bonnette S, Hugentobler JA, Mede B, Kiefer AW, Kurowski BG, Riley MA (2015) Postconcussion postural sway variability changes in youth: the benefit of structural variability analyses. Pediatr Phys Ther 27:316–327. https://doi.org/10.1097/pep.0000000000000193 Rahn C, Munkasy BA, Barry Joyner A, Buckley TA (2015) Sideline performance of the balance error scoring system during a live sporting event clinical. J Sport Med 25:248–253. https://doi. org/10.1097/jsm.0000000000000141 Register-Mihalik JK, Guskiewicz KM, McLeod TC et al (2013) Knowledge, attitude, and concussion-reporting behaviors among high school athletes: a preliminary study. J Athl Train 48(5):645–653 Resch J et al (2013) ImPact test-retest reliability: reliably unreliable? J Athl Train 48:506–511. https://doi.org/10.4085/1062-6050-48.3.09 Sambasivan K, Grilli L, Gagnon L (2015) Balance and mobility in clinically recovered children and adolescents after a mild traumatic brain injury. J Pediatr Rehabil Med 8:335–344. https://doi. org/10.3233/prm-150351 Schatz P, Moser R (2011) Current issues in pediatric sports concussion. Clin Neuropsychol 25(6):1042–1057 Serino A, Ciaramelli E, Di Santantonio A, Malagù S, Servadei F, Làdavas E (2006) Central executive system impairment in traumatic brain injury. Brain Inj 20:23–32 Sosnoff JJ, Broglio SP, Ferrara MS (2008) Cognitive and motor function are associated following mild traumatic brain injury. Exp Brain Res 187:563–571. https://doi.org/ 10.1007/s00221-008-1324-x

1324

R.D. Catena and K.J. Hildenbrand

Sosnoff JJ, Broglio SP, Shin S, Ferrara MS (2011) Previous mild traumatic brain injury and postural-control dynamics. J Athl Train 46:85–91. https://doi.org/10.4085/1062-6050-46.1.85 Sturm W et al (1999) Functional anatomy of intrinsic alertness: evidence for a fronto-parietalthalamic-brainstem network in the right hemisphere. Neuropsychologia 37:797–805 Swick D, Jovanovic J (2002) Anterior cingulate cortex and the stroop task: neuropsychological evidence for topographic specificity. Neuropsychologia 40:1240–1253 Szturm T, Maharjan P, Marotta JJ, Shay B, Shrestha S, Sakhalkar V (2013) The interacting effect of cognitive and motor task demands on performance of gait, balance and cognition in young adults. Gait Posture 38:596–602. https://doi.org/10.1016/j.gaitpost.2013.02.004 Tapper A, Gonzalez D, Roy E, Niechwiej-Szwedo E (2016) Executive function deficits in team sport athletes with a history of concussion revealed by a visual-auditory dual task paradigm. J Sports Sci 1–10 https://doi.org/10.1080/02640414.2016.1161214 Theye F, Mueller KA (2004) “Heads up”: concussions in high school sports. Clin Med Res 2(3):165–171 Vallee M, McFadyen BJ, Swaine B, Doyon J, Cantin JF, Dumas D (2006) Effects of environmental demands on locomotion after traumatic brain injury. Arch Phys Med Rehabil 87:806–813. https://doi.org/10.1016/j.apmr.2006.02.031 van Donkelaar P, Langan J, Rodriguez E, Drew A, Halterman C, Osternig LR, Chou L-S (2005) Attentional deficits in concussion. Brain Inj 19:1031–1039 van Donkelaar P, Osternig LR, Chou L-S (2006) Attentional and biomechanical deficits interact after mild traumatic brain injury. Exerc Sport Sci Rev 34:77–82 Vandenberghe R, Gitelman DR, Parrish TB, Mesulam MM (2001) Functional specificity of superior parietal mediation of spatial shifting. Neuroimage 14:661–673 Vaportzis E, Georgiou-Karistianis N, Churchyard A, Stout JC (2015) Effects of task difficulty during dual-task circle tracing in Huntington’s disease. J Neurol 262:268–276. https://doi.org/ 10.1007/s00415-014-7563-9 Vilkki J, Virtanen S, Surma-Aho O, Servo A (1996) Dual task performance after focal cerebral lesions and closed head injuries. Neuropsychologia 34:1051–1056 Wang J, Rao H, Wetmore GS, Furlan PM, Korczykowski M, Dinges DF, Detre JA (2005) Perfusion functional MRI reveals cerebral blood flow pattern under psychological stress. Proc Natl Acad Sci U S A 102:17804–17809 Weerdesteyn V, Schillings AM, van Galen GP, Duysens J (2003) Distraction affects the performance of obstacle avoidance during walking. J Mot Behav 35:53–63 Wilkins AJ, Shallice T, McCarthy R (1987) Frontal lesions and sustained attention. Neuropsychologia 25:359–365 Yantis S, Schwarzbach J, Serences JT, Carlson RL, Steinmetz MA, Pekar JJ, Courtney SM (2002) Transient neural activity in human parietal cortex during spatial attention shifts. Nat Neurosci 5:995–1002 Yardley L, Gardner M, Bronstein A, Davies R, Buckwell D, Luxon L (2001) Interference between postural control and mental task performance in patients with vestibular disorder and healthy controls. J Neurol Neurosurg Psychiatry 71:48–52 Zatsiorsky VM, King DL (1998) An algorithm for determining gravity line location from posturographic recordings. J Biomech 31:161–164 Zhang L, Abreu BC, Gonzales V, Huddleston N, Ottenbacher KJ (2002) The effect of predictable and unpredictable motor tasks on postural control after traumatic brain injury. NeuroRehabilitation 17:225–230

Functional Effects of Ankle Sprain Ilona M. Punt and Lara Allet

Abstract

Ankle sprain is one of the most common sports-related injuries and can lead to recurrences and chronic ankle instability (CAI). In the acute phase, ankle sprain patients experience mostly pain, limited ankle mobility, and reduced ankle muscle strength. CAI patients have a history of their ankle “giving way” and/or “feeling unstable,” after at least one significant ankle sprain. They continue to suffer from pain and impaired performance during functional tasks. Both acute ankle sprains and CAI have a negative influence on daily life activities such as walking, sportsrelated activities such as jump landings, as well as on patients’ perception of health and function. Functional deficits should be carefully assessed for appropriate clinical decision making and to propose the most suitable, individualized (physiotherapeutic) intervention. Acute ankle sprains are first treated according to the rest, ice, compression, and elevation (RICE) protocol. Nonsteroidal antiinflammatory drugs may also be recommended for pain management. A short period of immobilization by means of a lower leg cast can facilitate rapid decrease in pain and swelling. Afterward, functional exercise therapy is recommended.

I.M. Punt (*) Department of Epidemiology, Maastricht University, CAPHRI, Maastricht, The Netherlands Department of Physical Therapy, University of Applied Sciences of Western Switzerland, Carouge, Switzerland e-mail: [email protected] L. Allet Department of Physical Therapy, University of Applied Sciences of Western Switzerland, Carouge, Switzerland Department of Community Medicine, Geneva University Hospitals and University of Geneva, Geneva, Switzerland e-mail: [email protected] # Springer International Publishing AG, part of Springer Nature 2018 B. Müller, S.I. Wolf (eds.), Handbook of Human Motion, https://doi.org/10.1007/978-3-319-14418-4_72

1325

1326

I.M. Punt and L. Allet

In the case of CAI, patients should wear external ankle support during sporting activities to reduce the risk of recurring sprains and undergo exercise therapy including balance and muscle strengthening exercises. New technologies could be implemented in future rehabilitation programs in order to offer athletes greater flexibility in terms of training time and more varied, sports-related, exercises at home. Keywords

Ankle sprain • Chronic ankle instability • Clinical exam • Gait • Balance • Jump • Patient reported outcome measures • Treatment

Contents State of the Art . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Definition . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Symptoms and Functional Deficits Related to an Ankle Sprain . . . . . . . . . . . . . . . . . . . . . . . . . . . Risk Factors . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Epidemiology . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Health-Care Costs Related to Ankle Sprain . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Assessments Needed for Proper Decision Making . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Clinical Exam . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Specific Tests . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Functional Tests . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Patient Reported Outcome Measures (PROMs) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Treatment Modalities . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Future Directions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

1326 1326 1327 1327 1327 1328 1328 1328 1330 1331 1335 1335 1336 1337

State of the Art Definition Ankle sprain is defined as a partial or complete tear of the ligaments of the ankle due to sudden stretching. The most common mechanism causing lateral ankle sprain is excessive and explosive inversion and some degree of plantar flexion of the rear-foot on the tibia (Balduini and Tetzlaff 1982) during gait, cutting maneuvers during sports, jump landings, or stepping off an uneven surface (Bullock-Saxton et al. 1994; Hertel 2008; Wikstrom et al. 2006). In particular, athletes playing indoor/ court sports (i.e., basketball, volleyball, tennis), field-based sports (i.e., soccer), or long-distance running have an increased risk of ankle sprain injuries (Doherty et al. 2014b; Nery et al. 2016). After an ankle sprain injury, the lateral ligaments of the ankle are the most frequently injured, in particular the anterior talofibular ligament (ATFL), followed by injuries to the calcaneal fibular ligament (CFL) (Martin et al. 2013). The severity of an ankle sprain can be graded as follows:

Functional Effects of Ankle Sprain

1327

• Grade I: mild damage to the fibers of the ligament without ligamentous laxity of the affected joint • Grade II: partial tear of the ligament with abnormal laxity of the ankle joint • Grade III: complete rupture of the ankle ligament (Birrer et al. 1999) Assessing the grade of an ankle sprain is important to make the appropriate decision about future treatment strategy. The more severe the grade, the more time the patient will need to fully recover.

Symptoms and Functional Deficits Related to an Ankle Sprain Ankle sprain patients experience in the acute phase mostly pain, limited ankle mobility, and reduced ankle muscle strength. These symptoms negatively influence daily life activities such as gait, balance performance, and sports-related activities such as jump landings (Aiken et al. 2008; Hertel 2000; Rose et al. 2000). Persons with chronic ankle instability (CAI) have a history of their ankle “giving way” and/or “feeling unstable” after at least one significant ankle sprain that was associated with inflammatory response. These individuals experience pain and demonstrate impaired performance during functional tasks (Mcgovern and Martin 2016). CAI patients also find activities like (single-leg) balance performance, gait, and sports-related activities difficult to perform.

Risk Factors Risk factors and mechanisms which potentially contribute to recurrent ankle sprains include altered intrinsic body functions such as decreased proprioception in the ankle ligaments, muscle weakness, limited range of motion of the ankle joint, and extrinsic factors such as inappropriate footwear (Van Rijn et al. 2008; Mckeon and Hertel 2008a). A history of ankle sprain is in itself a risk factor for a re-sprain and may lead to mechanical or functional instability resulting in CAI (Van Rijn et al. 2008; Hertel 2002).

Epidemiology Lateral ankle sprain injury is the most common sports-related acute injury and occurs predominantly in persons aged 15–19 years (Fong et al. 2007; Hootman et al. 2007; Waterman et al. 2010). The incidence rate is 11.55 per 1000 exposures (Doherty et al. 2014b). Despite various treatment modalities, persons with a history of ankle sprain are known to present higher risk for re-spraining their ankle and developing mechanical or functional instability, resulting in chronic ankle instability (CAI) (Van Rijn et al. 2008; Hertel 2002). Up to 34% of patients report recurrent ankle sprains during the first year after the initial injury (Van Rijn et al. 2008). Up to 74% of all ankle sprain patients also experience residual symptoms such as pain, swelling,

1328

I.M. Punt and L. Allet

peroneal muscle weakness, or neuromuscular dysfunctions (Hertel 2000), all of which make patients susceptible for further injury and negatively influence activities of daily life (ADL) and sport performance.

Health-Care Costs Related to Ankle Sprain Ankle sprains lead to high direct and indirect health-care costs (Nazarenko et al. 2013; Verhagen et al. 2005). In the United States (US), the estimated costs are between 318 and 914 US dollars per acute ankle sprain (Nazarenko et al. 2013). In the Netherlands, the costs are estimated to be 360 euros per sprain (Verhagen et al. 2005). Another Dutch study calculated direct health-care costs of patients visiting the emergency department after a ligamentous ankle injury. These costs are estimated to be 684 euros per injury (De Boer et al. 2014). Costs increase with patients’ age, in particular for ambulance care, home care and rehabilitation. Since the introduction of new guidelines in the Netherlands, patients with minor injuries are able to visit the general practitioner 24 h a day, 7 days a week. Consequently, fewer ankle sprain patients visited the emergency department (De Boer et al. 2014).

Assessments Needed for Proper Decision Making Whenever a patient visits a medical doctor or physical therapist after suffering an ankle injury, the physician and/or therapist inquires about the accident occurrence and performs a clinical exam. In particular, the physician and/or physical therapist inquire about perceived pain intensity and if the patient has a history of “giving way” and/or “feelings of instability” which might indicate CAI. The degree of swelling is measured, and the remaining range of motion (ROM) and muscle strength evaluated. Physicians/therapists then check for functional deficits such as deficits in balance, gait, and sports-related tasks (e.g., jump landings). Patient reported outcome measures (PROMs), assessing patients’ perception of their health and function, complete the examination.

Clinical Exam A clinical exam serves to develop an individualized evidence-based rehabilitation plan that supports recovery while decreasing the risk of reinjury (Mcgovern and Martin 2016). During the clinical exam, the physician or physical therapist inquires about pain intensity during rest and activity, observes the patient in the standing and lying down positions, and screens for swelling, hematoma, bruising, and deformity. The physician or therapist also checks postural issues such as overpronation of the foot or difficulty in putting the foot on the floor. The entire ankle joint is then palpated to assess skin temperature (which can increase due to acute inflammation) and swelling and ascertain whether the ligaments are painful to touch. Then the

Functional Effects of Ankle Sprain

1329

therapist measures the degree of edema, range of motion (ROM), and muscle strength and checks for functional deficits.

Swelling The gold standard to measure edema is the water displacement method (Mawdsley et al. 2000; Mckay et al. 2001), but this method may be too time-consuming for efficient clinical use. However, Mawdsley et al. (2000) and Watson et al. (2008) showed recently that the figure-of-eight method is valid when assessing ankle edema in a clinical setting and that the inter-rater reliability of this method is excellent (ICC > 0.99, SEM of  0.2 cm). The therapist wraps a tape measure around standardized anatomical landmarks near the ankle, and the distance provides a circumferential estimate of volume (Fig. 1). Range of Motion (ROM) Ankle mobility can be assessed actively and passively in the classical way using a goniometer. However, measuring dorsiflexion in standing simulates the ROM required for functional tasks. This is particularly relevant because the torques applied to the ankle during weight bearing are clearly greater than in non-weight-bearing conditions and the resulting measurement may be more indicative of the range available for functional activities (Bennell et al. 1998; Bohannon et al. 1989). To measure the ankle dorsiflexion range during weight bearing, the participant stands on an apparatus consisting of a horizontal footplate attached to a vertical board. Participants align the big toe and heel of the test leg over a line marked along the center of the footplate. Participants are instructed not to lift the test heel, which is checked by the examiner who gently palpates for lifting while the participant moves his knee forward into a lunge position until the patella touches the midline of the vertical board (Bennell et al. 1998; Bohannon et al. 1989). To prevent forward movement of the big toe as the knee moves forward over the foot, a block is placed in front of the big toe. The distance (in cm) from the vertical board to the big toe is measured. Fig. 1 The figure-of-eight method for measuring ankle edema

1330

I.M. Punt and L. Allet

Muscle Strength Classical manual muscle testing in a sitting position is indicated to test strength in dorsiflexion, inversion, inversion with dorsiflexion, eversion, and eversion with plantar flexion. However, using a handheld dynamometer could further improve the measuring of muscle strength. The strength of the gastrocnemius and soleus can also be tested in a standing position. To test the gastrocnemius, the patient stands on the test limb with the knee extended. Patients may place one or two fingers on a table to assist with balance. The patient actively raises his heel from the floor 20 consecutive times without resting through full range of plantar flexion. To test the soleus the same procedure is used; only the patient stands on the test limb with his knee slightly flexed (Hislop et al. 2013; Spink et al. 2010).

Specific Tests After these classical tests, specific ankle tests are used to assess the integrity of the ligaments or the damage sustained. • Anterior drawer test: used to assess the ATFL. The patient sits with his knee flexed in order to relax the gastrocnemius and soleus muscles. The ankle is in 10 plantar flexion. The heel is held and forcefully pulled forward with one hand, while the other hand applies proximal counterpressure (Fig. 2). The test is positive when the injured ankle shows severe anterior subluxation compared to the noninjured ankle (Welck et al. 2015). • Talar tilt test: used to test the CFL. The patient is positioned with his knee flexed. The heel is grasped and the talus tilted into varus (Fig. 3). A normal degree of tilt is 0–23 . The injured side is compared to the noninjured ankle (Welck et al. 2015).

Fig. 2 Anterior drawer test

Functional Effects of Ankle Sprain

1331

Fig. 3 Talar tilt test

• External rotation test: used to assess the syndesmosis. The leg is stabilized proximally to the ankle joint while grasping the plantar aspect of the foot and rotating the foot externally. The test is positive when painful (Alonso et al. 1998). • Squeeze tests: also used to assess the syndesmosis. The fibula and tibia are compressed at midcalf. The test is positive when it elicits pain (Alonso et al. 1998; Welck et al. 2015).

Functional Tests Physical therapists should use functional tests to assess how the patient moves and how the ankle injury affects balance, walking, and jumping. The severity of the ankle injury dictates which tests are selected.

Balance Performance Balance deficits have been found to be present after acute ankle sprains (Mckeon and Hertel 2008a) for up to 1 week both in the injured ankle and in the noninjured ankle (Evans et al. 2004). Balance performance can be assessed using clinical tests or laboratory analyses using force platforms. A frequently used dynamic clinical test is the Star Excursion Balance Test (SEBT), a series of single-limb squats using the non-stance limb to reach maximally to touch a point along one of eight designated lines on the ground (Fig. 4) (Gribble et al. 2012). The SEBT showed that acute ankle sprain patients have a shorter anterior reach/leg length compared to healthy controls (Pourkazemi et al. 2016; Akbari et al. 2006). Force plate data are most often characterized by the analysis of the trajectory of the center of pressure (COP). Parameters that are frequently chosen to assess balance performance are COP range, length, and speed. The review of Mckeon and Hertel (2008a) showed that COP range, length, and speed of the injured ankle were increased after an acute lateral ankle sprain compared to healthy controls.

1332

I.M. Punt and L. Allet

Anterior

Anteriorlateral

Anteromedial

Lateral

Medial

Posteromedial

Posterolateral Posterior

Fig. 4 Reach direction for left ankle stance of the Star Excursion Balance Test (SEBT). Directions are labeled based on the reach direction from the stance limb

Although patients significantly improve postural control (e.g., COP range, length, as well as speed) during the first 4 weeks after an ankle sprain (Evans et al. 2004; Hertel et al. 2001), they frequently present residual functional deficits (muscle strength, mobility) and impaired postural control after this period (Hertel et al. 2001; Holme et al. 1999). Genthon et al. (2010) showed that ankle sprain patients present asymmetric balance in bipedal stance during the first 10 days after injury. From day 10 to day 30, bipedal balance improved and returned to normal after 30 days (Genthon et al. 2010). However, deficits may become more evident while balancing on one leg (Mckeon and Hertel 2008a; Rozzi et al. 1999; Wikstrom et al. 2009a). The study of Hertel et al. (2001), for example, showed that during a singleleg balance test, COP length and speed were greater in the injured ankle compared to the noninjured ankle for up to 4 weeks after injury. However, both parameters significantly improved 4 weeks after injury compared to the day after the injury (Hertel et al. 2001). Using the noninjured leg as a reference is not recommended to estimate residual deficit because it is assumed that central neural changes after unilateral lateral ankle trauma affect motor control of both extremities (Holme et al. 1999). The meanings differ as regard to the balance impairments for CAI patients. Dynamic clinical tests (i.e., SEBT) in CAI patient showed that presenting deficits in ankle dorsiflexion ROM leads to difficulties with the anterior reach direction of the SEBT (Basnett et al. 2013; Munn et al. 2010; Arnold et al. 2009). Metaanalyses studying force plate data concluded that ankle instability leads to

Functional Effects of Ankle Sprain

1333

impaired balance performance (Arnold et al. 2009; Munn et al. 2010). These metaanalyses further stated that it remains unclear whether these differences in balance preexisted or were the consequence of the ankle injury (Arnold et al. 2009). Furthermore, no definitive conclusion could be made based on a systematic review from Mckeon and Hertel (2008a). They compared the COP performance achieved with the injured ankle of CAI patients with the COP performance of healthy controls or the one achieved with the uninjured ankle (Mckeon and Hertel 2008a).

Gait Performance Temporal-Spatial Acute ankle sprain patients demonstrated disturbed gait parameters such as slower gait speed, shorter step length, shorter single support time, as well as disturbed symmetry for single support time (Crosbie et al. 1999; Punt et al. 2015). For example, healthy persons demonstrate on average a walking speed of 1.29  0.17 m/s, while ankle sprain patients only walk 1.12  0.25 m/s 4 weeks after the initial injury (Punt et al. 2015). Punt et al. (2015) showed that decreased walking speed was correlated to increased pain levels and deficits in dorsiflexion muscle strength measured with a handheld dynamometer. In contrast, CAI patients show similar gait speed compared to healthy age-matched controls (Monaghan et al. 2006).

Kinematics For normal gait, which is one of the most frequent activities, a minimum ankle dorsiflexion of 10 has been shown to be necessary (Riener et al. 2002). In acute ankle sprain patients, Crosbie et al. (1999) showed that the degree of maximum passive dorsiflexion of the ankle measured was correlated with gait speed, step length, and symmetry for single support time. However, Punt et al. (2015) found no difference between ankle sprain patients and healthy controls for maximum dorsiflexion during the stance phase of gait while walking at a self-selected walking speed. In contrast, Punt et al. (2015) found that the maximum plantar flexion was reduced on the injured side of patients ( 14.2  7.9) compared to healthy controls ( 18.7  8.0). Doherty et al. (2015a) described similar findings when comparing the injured ankle with the noninjured side. In addition, the timing of maximum plantar flexion was delayed at the injured ankle compared with that of the healthy group (Punt et al. 2015). Furthermore acute ankle sprain patients demonstrated increased ankle inversion with a greater inversion moment (Delahunt et al. 2006; Doherty et al. 2015a; Monaghan et al. 2006). While walking with a similar gait speed, CAI patients showed more ankle inversion compared to healthy controls (Monaghan et al. 2006; Delahunt et al. 2006). In addition, CAI patients inverted the ankle at a rate of 0.5 rad/s around heel strike, while healthy controls slowly everted their ankle at a rate of 0.1 rad/s (Monaghan et al. 2006).

1334

I.M. Punt and L. Allet

Kinetics Ankle sprain patients demonstrated lower maximum concentric dorsiflexion power compared with the healthy controls. They also demonstrated lower maximum eccentric plantar flexion power compared with the healthy subjects. Furthermore, the maximum moment was lower in the ankle sprain group compared with the healthy group (Punt et al. 2015). These findings indicate that patients with an ankle sprain use a more conservative and secure gait pattern, characterized by slow self-selected walking speed. This might have been even more marked if patients had been tested with stricter requirements, such as faster walking speed, running, or irregular walking surfaces. CAI patients showed an evertor moment directly after heel strike, whereas healthy controls showed an inventor moment while walking at a similar gait speed. Joint power also differed between these two groups after heel strike. CAI patients showed concentric power generation, while healthy controls showed eccentric power generation (Monaghan et al. 2006).

Sports-Related Tasks Previous studies showed that ankle sprain patients displayed reduced ankle plantar flexion while performing a bipedal drop vertical jump 2 weeks, but also 6 months, after the initial injury, compared to healthy controls (Doherty et al. 2014a, 2015b). Similar findings were found when patients performed a single-leg jump 4 weeks after the injury compared to a control group (Allet et al. 2016). Thus, perhaps the altered movement pattern of ankle sprain patients represents a security strategy to avoid recurrences. Increased dorsiflexion of the ankle brings the joint into a more closed-packed position that protects the lateral ligament complex more efficiently. This evasive movement avoids the typical ankle sprain injury mechanism, a combination of inversion and plantar flexion of the ankle (Balduini and Tetzlaff 1982). Nevertheless, abnormal foot positioning at initial contact might lead to faulty neuromuscular preprogramming of ankle joint movement, thereby, in the long term, contributing to the development of CAI (Hertel 2008). Alternatively, dorsiflexion could be a precaution behavior of ankle sprain patients. Ankle sprain patients may not jump as high as healthy persons, and therefore their toes would not be able to clear the floor. Studies including CAI patients showed divergent results ranging from increased ankle dorsiflexion (Caulfield and Garrett 2002) to decreased dorsiflexion when patients performed a single-leg drop landing (Ashtonmiller et al. 1996). Studies investigating neuromuscular control mechanisms in CAI patients also reported reduced activation of the peroneus muscle before initial contact during a single-leg drop landing compared to healthy subjects (Caulfield et al. 2004; Delahunt et al. 2006). CAI patients, whose landing phase while running and stop-jump maneuvers was evaluated, showed a more inverted ankle, reduced muscle co-contraction, and decreased dynamic stiffness in the ankle joint during landing phase compared to healthy controls (Lin et al. 2011).

Functional Effects of Ankle Sprain

1335

Patient Reported Outcome Measures (PROMs) Patient reported outcome measures (PROMs) are instruments assessing patients’ perception of their health and function. Ankle PROMs typically include questions addressing pain, mobility, function, and quality of life. PROMs are important but should supplement, rather than replace, existing measures of quality and performance. Several validated PROMs are available for foot and ankle disorders (Eechaute et al. 2007; Weel et al. 2015; Martin and Irrgang 2007). The most commonly used PROMs related to foot and ankle disorders are the following: • Ankle Joint Functional Assessment Tool (AJFAT): it contains five impairment items (pain, stiffness, stability, strength, rolling over), four activity-related items (walking on uneven ground, cutting when running, jogging, and descending stairs), and one overall quality item. The maximal total score of the AJFAT is 40 points, and the minimum is 0 points (Rozzi et al. 1999). • Functional Ankle Ability Measure (FAAM): this self-report questionnaire consists of a 21-item, activities of daily living (ADL) subscale and an 8-item sportsrelated subscale. The final score ranges from 0 to 100 for ADL as well as sports, higher scores indicating higher levels of function (Martin et al. 2005). • Foot and Ankle Disability Index (FADI): this 34-item questionnaire is divided into two subscales: the FADI and the FADI Sport. The FADI contains 4 painrelated items and 22 activity-related items. The FADI Sport contains eight activity-related items. The scores of the FADI and FADI Sport are then transformed into percentages (Hale and Hertel 2005). • Functional Ankle Outcome Score (FAOS): this 42-item questionnaire is divided into five subscales, pain, other symptoms, ADL, sport and recreation function, and foot and ankle-related quality of life. Final scores are then transformed into ratings from 0 to 100 (worst to best score) (Roos et al. 2001). According to Eechaute et al. (2007), the FADI and the FAAM are considered as the most appropriate self-report tools to quantify functional disabilities in patients with CAI. CAI patients demonstrate worse FADI and FADI Sport scores compared to healthy controls (Wikstrom et al. 2009b).

Treatment Modalities Appropriate management of lateral ankle sprain is vital for a successful recovery. Acute ankle sprains can be managed initially according to the rest, ice, compression, and elevation (RICE) protocol, and the use of nonsteroidal anti-inflammatory drugs may be recommended for pain management (Kerkhoffs and Van Dijk 2013). A short period ( 2.5) at the right limb femoral neck and the L2 to L4 spine NA

1

1

SCI

SCI

Observation measures: knee joint angle

Observation measures: knee joint angle

recording of subjects’ walking and walking speed Outcome measures: EMG

(continued)

Absence of a randomized clinical trial design comparing exoskeleton use to conventional gait training The exoskeleton Absence of a products a randomized standing-up and a clinical trial sitting motion design support systems comparing for completely exoskeleton paraplegic use to patients conventional gait training

The exoskeleton products a standing-up and a sitting motion support systems for completely paraplegic patients

exoskeleton use to conventional gait training

Gait Rehabilitation with Exoskeletons 1453

Experimental trial

Zeilig et al. (2012))

ReWalk

Exoskeleton HAL

1. 16–70 years of age 2. Weight < 100 kg 3. Height from 155–200 cm 4. Complete motor impairment C7–C8 or T1–T12 5. At least 6 months since injury 6. Regular user of a RGO or therapeutic standing frame

a

6

Sample inclusion Sample criteriaa size 1. Hemiparesis 22 resulting from unilateral ischemic or hemorrhagic stroke 2. Time since stroke onset of < 6 months

NA not available: in the study inclusion criteria were not reported All the measures’ abbreviations are explained in the Table 2 b See Table 2 for the measure reference details

Israel

Study design Controlled clinical trial

Studies in alphabetic order Country Watanabe Japan et al. (2014)

Table 1 (continued) Sample type Measuresb Stroke Clinical measures: isometric muscle strength Observation measures: maximum walking speed Outcome measures: FAC, TUG, 6 MWT, SPPB, and FM-LE SCI Observation measure: distance walked in 6 minutes Outcome measures: 10 MWT and TUG The exoskeleton increases mobility and functional abilities and decreases the risk of secondary injuries

Strengths

Absence of a randomized clinical trial design comparing exoskeleton use to conventional gait training

Weaknesses

1454 S. Federici et al.

Gait Rehabilitation with Exoskeletons

1455

extension, trunk flexion and extension, ankle dorsi-/plantar flexion) and lower extremity circumferences • Outcome measures (standardized tests, neurophysiological, or neuroimaging techniques), to evaluate the effectiveness of a treatment • Observation measures, to observe the subject’s functional performance while wearing the exoskeleton – such as knee joint angle or torque, a video recording of a subject walking, hip and knee joint angle trajectories during walking, the time practicing, and distance walked in a certain time period. Table 2 provides details of the indicators and the studies in which they were used. In the following subsections, we discuss 16 brands of exoskeletons experimentally tested in the scientific literature, of which five have already been commercialized (HAL, Ekso, ReWalk, Indego, and Walking Assist Device) and 11 are not yet commercialized (Vanderbilt lower limb exoskeleton, Human Universal Load Carrier (HULC), MindWalker exoskeleton, advanced reciprocating gait orthosis (ARGO), eLEGS, X1 robotic exoskeleton (MINA), ATLAS, ABLE system, Tibion Bionic Technologies, Bionic Leg, and H2 robotic exoskeleton). Hybrid Assistive Limb (HAL) The HAL exoskeleton was tested with paraplegic subjects with a SCI at the T10 level, complete or incomplete, or having hemiparesis after a stroke (Aach et al. 2013, 2014; Kawamoto et al. 2010; Nilsson et al. 2014; Sczesny-Kaiser et al. 2013; Tsukahara et al. 2009, 2010; Watanabe et al. 2014). Clinical, observational, and outcome measures administered to investigate the effectiveness of HAL varied. For example, Sczesny-Kaiser et al. (2013) used functional magnetic resonance imaging and electromyography (EMG) to evaluate cortical excitability and plastic changes after a three-month period of treadmill training supported by HAL. Taken together, the results of these studies guaranteed the system’s safety (Nilsson et al. 2014) and its effectiveness. Paraplegic patients gained significant increases in over ground walking functional abilities (Aach et al. 2013, 2014), and a larger knee angle was measured during leg flexion (Kawamoto et al. 2010). In addition, diagnostic imaging displayed an augmented paired pulse inhibition of somatosensory evoked potentials in both hemispheres following median nerve stimulation at the wrist. There was also a reduced somatosensory cortex activation of the activated area in both hemispheres after tactile stimulation of the index finger (Sczesny-Kaiser et al. 2013). Finally, even a gait training program with the single-leg version of HAL could facilitate independent walking more efficiently than conventional gait training (Watanabe et al. 2014). ReWalk The ReWalk exoskeleton was tested with paraplegic patients with complete SCIs, at the C7-T12 and the T1-T12 level. By using the ReWalk, paraplegic patients were able to walk independently, supervised by one person (Raab et al. 2016), and to achieve a level of walking proficiency that was close to that needed for limited community ambulation in an urban setting (Asselin et al. 2015; Benson et al. 2016; Esquenazi et al. 2012;

Abbreviation TUG

10 MWT

6 MWT

/

BBS

/

Measure Timed up and go test

10-meter walk test

6-minute walking test

Knee joint angle or torque

Berg balance scale

Walking speed

6

6

7

8

9

Frequency 10

Observation measures

Outcome measures

Observation measures

Outcome measures

Outcome measures

Measure type Outcome measures

Study where the measure was administered Aach et al. (2013, 2014), Bishop et al. (2012), Bortole et al. (2015), Farris et al. (2014), Nilsson et al. (2014), Quintero et al. (2012), Stein et al. (2014), Watanabe et al. (2014), Zeilig et al. (2012) Aach et al. (2013, 2014), Bishop et al. (2012), Esquenazi et al. (2012), Farris et al. (2014), Hartigan et al. (2015), Nilsson et al. (2014), Stein et al. (2014), Zeilig et al. (2012) Aach et al. (2014), Bishop et al. (2012), Bortole et al. (2015), Esquenazi et al. (2012), Farris et al. (2014), Hartigan et al. (2015), Stein et al. (2014), Watanabe et al. (2014) Bortole et al. (2015), Farris et al. (2011, 2012), Kawamoto et al. (2010), Strausser and Kazerooni (2011), Tsukahara et al. (2009, 2010) Bishop et al. (2012), Bortole et al. (2015), Nilsson et al. (2014), Stein et al. (2014) Li et al. (2015), Raab et al. (2016) Aach et al. (2013), Farris et al. (2011), Nilsson et al. (2014), Spungen et al. (2013), Talaty et al. (2013), Watanabe et al. (2014)

Table 2 Summary of the measures adopted in the studies reviewed ordered by largest to smallest frequency in use

/

Downs et al. (2013)

/

Reybrouck (2003)

Peters et al. (2014)

Measures’ references Podsiadlo and Richardson (1991)

1456 S. Federici et al.

2

2

1 1

FM-LE

/

BI

FAC

/

/

WISCI II

/

/

/

CAFE 40

Fugl-Meyer assessment for lower extremity

Video recording of subject’s walking

Barthel index

Functional ambulation category

Isometric muscle strength (hip/knee/ trunk flexion and extension, ankle dorsi /plantar flexion) Time practicing

Walking index for SCI II

Ashworth scale

Assistive devices used during ambulation Blood pressure

California functional evaluation 40

1

1

2

2

2

4

4

4

EMG

Electromyography

4

/

Distance walked in a certain time period

Observation measures Outcome measures Outcome measures Observation measures Observation measures Outcome measures

Outcome measures Clinical measures

Observation measures Outcome measures

Outcome measures

Outcome measures

Observation measures

Stein et al. (2014)

Kolakowsky-Hayner et al. (2013)

Kolakowsky-Hayner et al. (2013)

Aach et al. (2014)

Kolakowsky-Hayner et al. (2013), Strausser and Kazerooni (2011) Aach et al. (2013, 2014)

Nilsson et al. (2014), Watanabe et al. (2014) Talaty et al. (2013), Watanabe et al. (2014)

Aach et al. (2013), Esquenazi et al. (2012), Kolakowsky-Hayner et al. (2013), Zeilig et al. (2012) Li et al. (2015), Sczesny-Kaiser et al. (2013), Sylos-Labini et al. (2014), Talaty et al. (2013) Bortole et al. (2015), Li et al. (2015), Nilsson et al. (2014), Watanabe et al. (2014) Ikehara et al. (2011), Li et al. (2015), Strausser et al. (2010), Talaty et al. (2013) Bortole et al. (2015), Nilsson et al. (2014)

(continued)

Fung et al. (1997)

/

Ditunno et al. (2013) Pandyan et al. (1999) /

/

Cuesta-Vargas and Perez-Cruzado (2014) Mehrholz et al. (2007) /

/

Park and Choi (2014)

Mohseni Bandpei et al. (2014)

/

Gait Rehabilitation with Exoskeletons 1457

1 1

1 1

/

DGI

EFAP

FES(S)

5XSST

FIM

fMRI

HR

/

/

LEMS

NIHSS

Emory functional ambulation profile

Falls efficacy scale, Swedish version

Five times sit-to-stand test

Functional independence measure

Functional magnetic resonance imaging

Heart rate

Kinematic magnitudes and exchanged forces Lower extremity circumference

Lower extremity motor score

National Institutes of Health Stroke scale

1

1

1

1

1

1

1

1

Frequency 1

Abbreviation S-COVS

Measure Clinical outcome variable scale, Swedish version Conversation with the user during walking Dynamic gait index

Table 2 (continued) Measure type Outcome measures Observation measures Outcome measures Outcome measures Outcome measures Outcome measures Outcome measures Outcome measures Observation measures Observation measures Clinical measures Outcome measures Outcome measures Nilsson et al. (2014)

Aach et al. (2014)

Aach et al. (2014)

Belforte et al. (2001)

Asselin et al. (2015)

Sczesny-Kaiser et al. (2013)

Nilsson et al. (2014)

Stein et al. (2014)

Nilsson et al. (2014)

Stein et al. (2014)

Raab et al. (2016)

Neuhaus et al. (2011)

Study where the measure was administered Nilsson et al. (2014)

Yang et al. (2014)

Shin et al. (2011)

/

/

/

Buxton (2013)

Hellstrom et al. (2002) Whitney et al. (2005) Saji et al. (2015)

Wolf et al. (1999)

Tinetti (1986)

Measures’ references Andersson and Franzen (2015) /

1458 S. Federici et al.

1 1

1 1 1

/

/

SF-36

SPPB

/

/

/

/

/

VAS

Romberg’s test

Rotations and applied torques for each joint Short form health survey

Short physical performance battery

Skin evaluation

Spasticity

Step length

Time and level of assistance needed to transfer into and on device Time response of the angles and electric currents of each joint Visual analogue scale

1

1

1

1

1

1

/

Pain level

1

VO2

Oxygen uptake

Observation measures Observation measures Outcome measures Observation measures Outcome measures Outcome measures Observation measures Observation measures Observation measures Observation measures Observation measures Outcome measures Nilsson et al. (2014)

Mori et al. (2006)

Kolakowsky-Hayner et al. (2013)

Kolakowsky-Hayner et al. (2013)

Kolakowsky-Hayner et al. (2013)

Kolakowsky-Hayner et al. (2013)

Watanabe et al. (2014)

Raab et al. (2016)

Belforte et al. (2001)

Stein et al. (2014)

Kolakowsky-Hayner et al. (2013)

Asselin et al. (2015)

Reed and Van Nostran (2014)

/

/

/

/

McHorney et al. (1993) Stookey et al. (2014) /

Agrawal et al. (2011) /

/

/

Gait Rehabilitation with Exoskeletons 1459

1460

S. Federici et al.

Spungen et al. 2013), for example, for a distance of 100 m (Zeilig et al. 2012), with a fundamentally symmetrical gait (Talaty et al. 2013). Daily use of the exoskeleton seemed to increase activity energy expenditure, but this would be expected to have positive cardiopulmonary and metabolic benefits. The level of effort required to use the ReWalk exoskeleton system to ambulate appears to be acceptable and, as such, could be envisioned as a device that people with SCI would use in their daily lives (Asselin et al. 2015). Moreover, quality of life, mobility, risk of falling, motor skills, and control of bladder and bowel functions were improved after robot-assisted gait training (Raab et al. 2016). Nonetheless, the presence of skin aberrations was unexpectedly high, and the use of the exoskeleton generally did not meet subjects’ expectations in terms of perceived benefits and impact on quality of life (Benson et al. 2016). Vanderbilt Lower Limb Exoskeleton The Vanderbilt lower limb exoskeleton was developed by a team of engineers of the Vanderbilt University, at Nashville in Tennessee, chaired by H. A. Quintero (Farris et al. 2011, 2012, 2014; Quintero et al. 2012). Their clinical trials involved paraplegic patients with motor and sensory complete SCIs at the T10 level. Findings showed the Vanderbilt lower limb exoskeleton system assisted paraplegic patients to perform gait activities faster (Farris et al. 2012, 2014), with knee and hip joint amplitudes similar to those observed in non-SCI walking (Farris et al. 2011). Human Universal Load Carrier (HULC) The HULC was developed by H. Kazerooni and his team at Ekso Bionics in the USA. Given that the HULC is designed to assist able-bodied individuals by powering knee movements only in extension, just one study on its effectiveness met our inclusion criteria (Strausser et al. 2010). In this study, double-acting hydraulic cylinders replaced the single acting ones of the HULC, providing powered flexion and extension at the knee. Likewise, bracing used in the exoskeleton was augmented to support a patient with limited leg and torso muscle control. This clinical trial involved paraplegic patients with a motor complete SCI at the T5-T10 level. The study’s purpose was to discover whether the development of an exoskeleton for medical use would facilitate an active life in paraplegic people, so reducing the occurrence of secondary complications. The results confirmed that the HULC exoskeleton, when readapted for medical use, is able to safely increase mobility for those who are unable to walk unaided. MindWalker Exoskeleton The MindWalker was developed by a European consortium coordinated by M. Ilzkovitz and funded by the European Commission in 2009. Only one study (Sylos-Labini et al. 2014) involved participants with an SCI at the T7-L1 level. The aim of this study was to quantify the level of muscle activity in a sample of intact and injured patients while they walked with MindWalker. The measures used

Gait Rehabilitation with Exoskeletons

1461

were EMG activity, joint angles, and torques. The results showed that, in SCI patients, EMG activity of the upper limb muscles was augmented, while activation of the leg muscles was minimal. Contrary to expectations, however, in the neurologically intact subjects, EMG activity of the leg muscles was similar, or even greater, during exoskeleton-assisted walking compared to normal over ground walking. In addition, significant variations in the EMG waveforms were found in different walking conditions; the most variable pattern was observed in the hamstring muscles. Overall, the results are consistent with a nonlinear reorganization of the locomotor output when using the robotic stepping devices. Advanced Reciprocating Gait Orthosis (ARGO) Only one study tested the ARGO, produced by RSLSteeper, with paraplegic patients (Belforte et al. 2001). This clinical trial involved one participant with a motor complete SCI at the T3 level, enrolled in Italy. The design and construction of the ARGO, and experimental testing to assist locomotion in paraplegic subjects when using the ARGO, were described. Findings showed the device is ideal for the rehabilitation stage, given the structure’s modularity and the extremely flexible means used to regulate gait characteristics. Walking Assist Device The Walking Assist Device was tested by a team of engineers led by T. Ikehara. One of their studies (Ikehara et al. 2011) involved two participants with motor paralysis, enrolled in Japan. The results of the experimental study showed that the device could reproduce the power of kicking motions at the ankle joints when controlled by the hybrid system. Exoskeleton Lower Extremity Gait System (eLEGS) Only one clinical trial (Strausser and Kazerooni 2011) was carried out on this system. This study involved five participants with no leg motion due to SCI or ataxia, recruited in the USA. The authors tested whether the eLEGS was intuitive and easy to learn and use. The measures used were knee angle and time practicing. The results showed that the eLEGS’ human machine interface was easy to learn; all five subjects were able to quickly learn how to use it. Indeed, the walking performance of the five participants displayed an increased time between the heel off and the step, as compared to the able-bodied user. However, this time was decreased in the experienced user, while the user with no experience with the device had an average time of 0.859 sec, the experienced user was able to reduce the lag time to 0.590 s. X1 Robotic Exoskeleton (MINA) The team of the Institute for Human and Machine Cognition led by P. Neuhaus and the NASA Johnson Space Center jointly developed MINA (Neuhaus et al. 2011). The team carried out a clinical trial involving two participants with motor complete SCIs at the T10 level, recruited in the USA. The recruitment inclusion criteria stipulated an American Spinal Injury Association Impairment Scale of either an A (complete) or B (incomplete) and a Walking Index for Spinal Cord Injury (WISCI) of level  9 (ambulates with a walker, with braces, and no physical

1462

S. Federici et al.

assistance, 10 m). The paper presents a clinical and rehabilitative evaluation of the MINA exoskeleton. The measures used qualitatively evaluated the cognitive effort required to use MINA: researchers could talk with the subjects while they walked with MINA. In addition, static standing balance stability was tested by having the subjects catch and throw a ball while standing on both legs and using one crutch for balance. Findings provided evidence that MINA currently facilitates paraplegics’ walking mobility at speeds of up to 0.20 m/sec. In addition, MINA is not physically taxing and requires little cognitive effort, allowing the user to converse and maintain eye contact while walking. ATLAS The Center for Automation and Robotics in Spain developed the prototype of the ATLAS exoskeleton to help a quadriplegic child to walk. Its development and main features were tested by D. Sanz-Merodio and colleagues (Sanz-Merodio et al. 2012). This clinical trial involved one participant, a Spanish girl aged 8 years, affected by quadriplegia. Experiments validated a good controlled performance in following the gait pattern given by the parameterized trajectory generator. ABLE The ABLE system was designed and tested in Japan by Y. Mori and collaborators at Ibaraki University, in the Department of Intelligent Systems Engineering (Mori et al. 2006). This clinical trial involved one participant with motor paralysis, recruited in Japan. Mori and his team developed a standing style transfer system for a person with disabled legs. They proposed a new motion technique and compared it to their previous system (Mori et al. 2004). The measures used were the time response of the angles of the joints and electric currents, corresponding with the torque of each joint. The subject succeeded in standing up; a large arm force was needed in the beginning, but was not needed afterward. Tibion Bionic Technologies In 2013, Tibion Bionic Technologies was acquired by AlterG that now produces the exoskeleton tested by L. Bishop of Columbia University in 2012 (Bishop et al. 2012). This clinical trial involved one participant with a motor incomplete SCI at the C5-C6 level, enrolled in USA. Study outcomes suggested that the use of this device, during a physical therapy program for an individual with incomplete SCI, is practical and useful when used in addition to the standard training. Bionic Leg The Bionic Leg, produced by the Californian AlterG, is a powered knee orthosis for patients with unilateral neurologic or orthopedic conditions, tested by the team of the Department of Rehabilitation and Regenerative Medicine, Columbia University College of Physicians and Surgeons, led by J. Stein (Li et al. 2015; Stein et al. 2014). These clinical trials involved hemiparesis patients after a CVA. Subjects were required to be independent in household ambulation (with or without the use of unilateral assistive devices and with or without the use of ankle–foot orthoses).

Gait Rehabilitation with Exoskeletons

1463

The study of Stein et al. (Li et al. 2015, Stein et al. 2014) was designed to test how the Bionic leg restores mobility in stroke survivors in their living environments, while Li et al. (2015) aimed to demonstrate the training effects of a 3-week robotic leg orthosis and to investigate possible mechanisms of the sensorimotor alterations and improvements by using gait analysis and EMG. Outcomes suggested that robotic therapy for ambulatory stroke patients with chronic hemiparesis and using a robotic knee brace resulted in only modest functional benefits, in comparison with a group receiving only exercise intervention, for example, without using the Bionic Leg (Stein et al. 2014). The 3-week training period, using the wearable orthoses, improved the participants’ gait performance and improved muscle activation and walking speed (Li et al. 2015). Ekso One study tested the Ekso exoskeleton (Kolakowsky-Hayner et al. 2013). The study involved participants with a motor complete SCI. The main inclusion criteria for participants stipulated a body size compatible with the Ekso suit. The main goal of the study was to evaluate the feasibility and safety of using Ekso to aid ambulation in a group of individuals with SCI who had completed their initial SCI rehabilitation. Secondarily, training effects with the progressive use of Ekso were evaluated in terms of time tolerated, distance traveled, and assistance needed. Outcomes suggest that Ekso is safe for those with a complete thoracic SCI in a controlled environment, in the presence of experts. Ekso may eventually enhance mobility in those without volitional lower extremity function. There appears to be a training effect in the device. Indego The Indego exoskeleton was tested by the team of Virginia C. Crawford Research Institute, Shepherd Center, Atlanta, Georgia (Hartigan et al. 2015). This study was conducted to evaluate mobility outcomes for 16 SCI subjects with injury levels (ranging from C5 complete to L1 incomplete) after five gait training sessions with a powered exoskeleton. The primary goal was to characterize the ease of learning and usability of the system. Outcome measures of the study included the 10-meter walk test (10 MWT) and the 6-minute walk test (6 MWT). Results highlighted that the average walking speed was 0.22 m/sec for persons with C5–6 motor complete tetraplegia, 0.26 m/sec for T1–T8 motor complete paraplegia, and 0.45 m/sec for T9–L1 paraplegia. Distances covered in 6 min averaged 64 m for those with C5–C6, 74 m for T1–T8, and 121 m for T9–L1. Tetraplegic and paraplegic patients learned to use the Indego exoskeleton quickly and could manage a variety of surfaces. The walking speeds and distances achieved also indicated that some individuals with paraplegia could quickly become limited community ambulators using this system. H2 Robotic Exoskeleton The Exo-H2 has been the result of many years of research in the Grupo de Bioingeniería of the Spanish National Research Council (CSIC), who has conceded an exclusive license to Technaid S.L. for the design, manufacturing, and commercial exploitation of the system.

1464

S. Federici et al.

Bortole et al. (2015) evaluated the safety and usability of the H2 robotic exoskeleton for gait rehabilitation in three hemiparetic stroke patients across 4 weeks of training per individual (approximately 12 sessions). Results showed that the training was well tolerated and that H2 appears to be safe and easy to use. The system is robust and safe when applied to assist a stroke patient performing an over ground walking task.

Conclusions In this chapter, we have reviewed the clinical effectiveness in rehabilitation of various types of active, powered, and wearable lower limb exoskeletons used to facilitate and rehabilitate paraplegic patients’ gait disorders resulting from serious CNS lesions due to, for example, SCIs or CVAs. The literature review revealed that the exoskeletons subjected to the highest number of studies were the HAL (Aach et al. 2013, 2014; Kawamoto et al. 2010; Nilsson et al. 2014; Sczesny-Kaiser et al. 2013; Tsukahara et al. 2009, 2010; Watanabe et al. 2014), the ReWalk (Asselin et al. 2015; Benson et al. 2016; Esquenazi et al. 2012; Raab et al. 2016; Spungen et al. 2013; Talaty et al. 2013; Zeilig et al. 2012), and the Vanderbilt lower limb exoskeleton (Farris et al. 2011, 2012, 2014; Quintero et al. 2012). By qualitatively analyzing the results for each type of exoskeleton, it was found that the rehabilitative use of an exoskeleton to restore gait disorders is safe and practical (Bishop et al. 2012; Bortole et al. 2015; Kolakowsky-Hayner et al. 2013; Nilsson et al. 2014), is not physically exhausting, and requires only a little cognitive (Neuhaus et al. 2011) or energetic (Asselin et al. 2015) effort. In addition, it is easy to learn (Strausser and Kazerooni 2011), can increase mobility and functional abilities, and decreases the risk of secondary injuries (Aach et al. 2013, 2014; Asselin et al. 2015; Hartigan et al. 2015; Kolakowsky-Hayner et al. 2013; Raab et al. 2016; Strausser et al. 2010; Zeilig et al. 2012), as well as allowing restoration of a gait pattern comparable to normal over ground walking (Esquenazi et al. 2012; Farris et al. 2011; Spungen et al. 2013; Sylos-Labini et al. 2014). Among these positive attributes, Benson et al. (2016) stressed a negative one, claiming that the use of exoskeleton is characterized by the presence of a high rate of skin aberrations In addition to the advantages pointed out by the literature review, we found that the exoskeleton can be considered as an ecological device, replacing wheelchairs for many hours at a time; it enables patients who cannot walk to regain a degree of walking mobility and to retard the onset of a wide range of secondary disabilities associated with the long-term use of wheelchairs. The exoskeleton can improve the autonomy of the patient, who is enabled to walk independently, simply by wearing it. No other rehabilitative or therapeutic techniques and technologies provide such an extraordinary potential for autonomy. Nevertheless, there are still some limitations in the rehabilitative use of an exoskeleton. First, the wearability criteria are too restrictive; its use is limited to people with specific values of height and weight (Esquenazi et al. 2012; Nilsson et al. 2014;

Gait Rehabilitation with Exoskeletons

1465

Spungen et al. 2013; Sylos-Labini et al. 2014). Second, it requires very complex and specialized training to use the exoskeleton autonomously at home. Third, it is still an extremely expensive device, hardly covered by private or public healthcare systems. For instance, the National Health Services in Europe generally support the use of the exoskeleton for a rehabilitation program in specialized medical centers, but never for a private individual’s use at home or in the workplace. A final limitation is the scarcity of experimental designs based on evidence that demonstrates the effectiveness of the exoskeleton compared to other rehabilitative techniques and technologies. Only two studies that adopted a randomized clinical trial design compared exoskeleton use to conventional gait training (Stein et al. 2014; Watanabe et al. 2014); furthermore, the results of these two studies are contradictory. Finally, user experience with an exoskeleton in daily life activities generally did not meet subjects’ expectations in terms of perceived benefits and impact on quality of life (Benson et al. 2016).

Cross-References ▶ Brain-Computer Interfaces for Motor Rehabilitation ▶ Gait During Real-World Challenges: Gait Initiation, Gait Termination, Acceleration, Deceleration, Turning, Slopes, and Stairs ▶ Gait Retraining for Balance Improvement ▶ Measures to Determine Dynamic Balance ▶ Slip and Fall Risk Assessment

References Aach M, Meindl R, Hayashi T, Lange I, Geßmann J, Sander A, Nicolas V, Schwenkreis P, Tegenthoff M, Sankai Y, Schildhauer TA (2013) Exoskeletal neuro-rehabilitation in chronic paraplegic patients – initial results. In: Pons JL, Torricelli D, Pajaro M (eds) Converging clinical and engineering research on neurorehabilitation. Springer, Berlin, pp 233–236. https://doi.org/ 10.1007/978-3-642-34546-3_99 Aach M, Cruciger O, Sczesny-Kaiser M, Hoffken O, Meindl RC, Tegenthoff M, Schwenkreis P, Sankai Y, Schildhauer TA (2014) Voluntary driven exoskeleton as a new tool for rehabilitation in chronic spinal cord injury: a pilot study. Spine J 14(12):2847–2853. https://doi.org/10.1016/J. Spinee.2014.03.042 Agrawal Y, Carey JP, Hoffman HJ, Sklare DA, Schubert MC (2011) The modified Romberg balance test: normative data in US adults. Otol Neurotol 32(8):1309–1311. https://doi.org/10.1097/ MAO.0b013e31822e5bee Andersson P, Franzen E (2015) Effects of weight-shift training on walking ability, ambulation, and weight distribution in individuals with chronic stroke: a pilot study. Top Stroke Rehabil [Epub ahead of print]. https://doi.org/10.1179/1074935715Z.00000000052 Asselin P, Knezevic S, Kornfeld S, Cirnigliaro C, Agranova-Breyter I, Bauman WA, Spungen AM (2015) Heart rate and oxygen demand of powered exoskeleton-assisted walking in persons with paraplegia. J Rehabil Res Dev 52(2):147–158. https://doi.org/10.1682/JRRD.2014.02.0060 Belforte G, Gastaldi L, Sorli M (2001) Pneumatic active gait orthosis. Mechatronics 11(3):301–323. https://doi.org/10.1016/S0957-4158(00)00017-9

1466

S. Federici et al.

Benson I, Hart K, Tussler D, van Middendorp JJ (2016) Lower-limb exoskeletons for individuals with chronic spinal cord injury: findings from a feasibility study. Clin Rehabil 30(1):73–84. https://doi.org/10.1177/0269215515575166 Bishop L, Stein J, Wong CK (2012) Robot-aided gait training in an individual with chronic spinal cord injury: a case study. J Neurol Phys Ther 36(3):138–143. https://doi.org/10.1097/ NPT.0b013e3182624c87 Bortole M, Venkatakrishnan A, Zhu F, Moreno JC, Francisco GE, Pons JL, Contreras-Vidal JL (2015) The H2 robotic exoskeleton for gait rehabilitation after stroke: early findings from a clinical study. J Neuroeng Rehabil 12:54. https://doi.org/10.1186/s12984-015-0048-y Buxton RB (2013) The physics of functional magnetic resonance imaging (fMRI). Rep Prog Phys 76(9):096601. https://doi.org/10.1088/0034-4885/76/9/096601 Chaigneau D, Arsicault M, Gazeau JP, Zeghloul S (2008) LMS robotic hand grasp and manipulation planning (an isomorphic exoskeleton approach). Robotica 26(2):177–188. https://doi.org/ 10.1017/S0263574707003736 Cuesta-Vargas AI, Perez-Cruzado D (2014) Relationship between Barthel index with physical tests in adults with intellectual disabilities. SpringerPlus 3(543). https://doi.org/10.1186/2193-18013-543 Dickstein R, Levy S, Shefi S, Holtzman S, Peleg S, Vatine J-J (2014) Motor imagery group practice for gait rehabilitation in individuals with post-stroke hemiparesis: a pilot study. NeuroRehabilitation 34(2):267–276. https://doi.org/10.3233/NRE-131035 Ditunno JFJ, Ditunno PL, Scivoletto G, Patrick M, Dijkers M, Barbeau H, Burns AS, Marino RJ, Schmidt-Read M (2013) The walking index for spinal cord injury (WISCI/WISCI II): nature, metric properties, use and misuse. Spinal Cord 51(5):346–355. https://doi.org/10.1038/ sc.2013.9 Downs S, Marquez J, Chiarelli P (2013) The berg balance scale has high intra- and inter-rater reliability but absolute reliability varies across the scale: a systematic review. J Physiother 59 (2):93–99. https://doi.org/10.1016/S1836-9553(13)70161-9 Esquenazi A, Talaty M, Packel A, Saulino M (2012) The ReWalk powered exoskeleton to restore ambulatory function to individuals with thoracic-level motor-complete spinal cord injury. Am J Phys Med Rehabil 91(11):911–921. https://doi.org/10.1097/PHM.0b013e318269d9a3 Farris RJ, Quintero HA, Goldfarb M (2011) Preliminary evaluation of a powered lower limb orthosis to aid walking in paraplegic individuals. IEEE Trans Neural Syst Rehabil Eng 19(6): 652–659. https://doi.org/10.1109/TNSRE.2011.2163083 Farris RJ, Quintero HA, Goldfarb M (2012) Performance evaluation of a lower limb exoskeleton for stair ascent and descent with paraplegia. In: 34th Annual International Conference of the IEEE Engineering in Medicine and Biology Society: EMBC 2012, San Diego, 28 Aug–1 Sep 2012. pp 1908–1911. https://doi.org/10.1109/EMBC.2012.6346326 Farris RJ, Quintero HA, Murray SA, Ha KH, Hartigan C, Goldfarb M (2014) A preliminary assessment of legged mobility provided by a lower limb exoskeleton for persons with paraplegia. IEEE Trans Neural Syst Rehabil Eng 22(3):482–490. https://doi.org/10.1109/ TNSRE.2013.2268320 Fondazione Santa Lucia (2015) Maratona di roma 2015, un esoscheletro hi-tech per tornare a correre. http://www.hsantalucia.it/modules.php?name=News&file=article&sid=989. Accessed 15 May 2015 Fung S, Byl N, Melnick M, Callahan P, Selinger A, Ishii K, Devins J, Fischer P, Torburn L, Andrade C-K (1997) Functional outcomes: the development of a new instrument to monitor the effectiveness of physical therapy. Eur J Phys Rehab Med 7(2):31–41 Hartigan C, Kandilakis C, Dalley S, Clausen M, Wilson E, Morrison S, Etheridge S, Farris R (2015) Mobility outcomes following five training sessions with a powered exoskeleton. Top Spinal Cord Inj Rehabil 21(2):93–99. https://doi.org/10.1310/sci2102-93 Hellstrom K, Lindmark B, Fugl-Meyer A (2002) The falls-efficacy scale, Swedish version: does it reflect clinically meaningful changes after stroke? Disabil Rehabil 24(9):471–481. https://doi. org/10.1080/09638280110105259

Gait Rehabilitation with Exoskeletons

1467

Herr H (2009) Exoskeletons and orthoses: classification, design challenges and future directions. J Neuroeng Rehabil 6(1):1–9. https://doi.org/10.1186/1743-0003-6-21 Ikehara T, Nagamura K, Ushida T, Tanaka E, Saegusa S, Kojima S, Yuge L (2011) Development of closed-fitting-type walking assistance device for legs and evaluation of muscle activity. In: IEEE International Conference on Rehabilitation Robotics: ICORR 2011, Zurich, 29 Jun–1 Jul 2011. pp 1–7. https://doi.org/10.1109/ICORR.2011.5975449 Kao P-C, Lewis CL, Ferris DP (2010) Invariant ankle moment patterns when walking with and without a robotic ankle exoskeleton. J Biomech 43(2):203–209. https://doi.org/10.1016/j.jbiomech.2009. 09.030 Kawamoto H, Taal S, Niniss H, Hayashi T, Kamibayashi K, Eguchi K, Sankai Y (2010) Voluntary motion support control of robot suit HAL triggered by bioelectrical signal for hemiplegia. In: 32nd Annual International Conference of the IEEE Engineering in Medicine and Biology Society: EMBC 2010, Buenos Aires, 31 Aug–4 Sep 2010. pp 462–466. https://doi.org/ 10.1109/IEMBS.2010.5626191 Kolakowsky-Hayner SA, Crew J, Moran S, Shah A (2013) Safety and feasibility of using the EksoTM bionic exoskeleton to aid ambulation after spinal cord injury. J Spine S4(3):1–8. https:// doi.org/10.4172/2165-7939.S4-003 Lee HS, Song J, Min K, Choi Y-S, Kim S-M, Cho S-R, Kim M (2014) Short-term effects of erythropoietin on neurodevelopment in infants with cerebral palsy: a pilot study. Brain Dev 36 (9):764–769. https://doi.org/10.1016/j.braindev.2013.11.002 Li L, Ding L, Chen N, Mao Y, Huang D, Li L (2015) Improved walking ability with wearable robotassisted training in patients suffering chronic stroke. Biomed Mater Eng 26:S329–S340. https:// doi.org/10.3233/bme-151320 McHorney CA, Ware JE, Raczek AE (1993) The MOS 36-item short-form health survey (SF-36): II. Psychometric and clinical tests of validity in measuring physical and mental health constructs. Med Care 31(3):247–263 Mehrholz J, Wagner K, Rutte K, Meissner D, Pohl M (2007) Predictive validity and responsiveness of the functional ambulation category in hemiparetic patients after stroke. Arch Phys Med Rehabil 88(10):1314–1319. https://doi.org/10.1016/j.apmr.2007.06.764 Mohseni Bandpei MA, Rahmani N, Majdoleslam B, Abdollahi I, Ali SS, Ahmad A (2014) Reliability of surface electromyography in the assessment of paraspinal muscle fatigue: an updated systematic review. J Manipulative Physiol Ther 37(7):510–521. https://doi.org/ 10.1016/j.jmpt.2014.05.006 Mooney LM, Rouse EJ, Herr HM (2014) Autonomous exoskeleton reduces metabolic cost of human walking. J Neuroeng Rehabil 11(151):2–5. https://doi.org/10.1186/1743-0003-11-151 Moreno J, Ama A, Reyes-Guzmán A, Gil-Agudo Á, Ceres R, Pons J (2011) Neurorobotic and hybrid management of lower limb motor disorders: a review. Med Biol Eng Comput 49(10): 1119–1130. https://doi.org/10.1007/s11517-011-0821-4 Mori Y, Takayama K, Zengo T, Nakamura T (2004) Development of straight style transfer equipment for lower limbs disabled “able”. J Robot Mechatron 16(5):456–463 Mori Y, Okada J, Takayama K (2006) Development of a standing style transfer system “able” for disabled lower limbs. IEEE/ASME Trans Mechatronics 11(4):372–380. https://doi.org/10.1109/ TMECH.2006.878558 Nef T, Riener R (2012) Three-dimensional multi-degree-of-freedom arm therapy robot (ARMin). In: Dietz V, Nef T, Rymer WZ (eds) Neurorehabilitation technology. Springer, London, pp 141–157 Neuhaus PD, Noorden JH, Craig TJ, Torres T, Kirschbaum J, Pratt JE (2011) Design and evaluation of mina: a robotic orthosis for paraplegics. In: IEEE International Conference on Rehabilitation Robotics: ICORR 2011, Zurich, 29 Jun–1 Jul 2011. pp 1–8. https://doi.org/10.1109/ ICORR.2011.5975468 Nilsson A, Vreede KS, Haglund V, Kawamoto H, Sankai Y, Borg J (2014) Gait training early after stroke with a new exoskeleton – the hybrid assistive limb: a study of safety and feasibility. J Neuroeng Rehabil 11(92):1–10. https://doi.org/10.1186/1743-0003-11-92

1468

S. Federici et al.

Pandyan AD, Johnson GR, Price CI, Curless RH, Barnes MP, Rodgers H (1999) A review of the properties and limitations of the Ashworth and modified Ashworth scales as measures of spasticity. Clin Rehabil 13(5):373–383 Park EY, Choi YI (2014) Psychometric properties of the lower extremity subscale of the Fugl-Myer assessment for community-dwelling hemiplegic stroke patients. J Phys Ther Sci 26(11): 1775–1777. https://doi.org/10.1589/jpts.26.1775 Peters DM, Middleton A, Donley JW, Blanck EL, Fritz SL (2014) Concurrent validity of walking speed values calculated via the GAITRite electronic walkway and 3 meter walk test in the chronic stroke population. Physiother Theory Pract 30(3):183–188. https://doi.org/10.3109/ 09593985.2013.845805 Podsiadlo D, Richardson S (1991) The timed “up & go”: a test of basic functional mobility for frail elderly persons. J Am Geriatr Soc 39(2):142–148 Quintero HA, Farris RJ, Goldfarb M (2012) A method for the autonomous control of lower limb exoskeletons for persons with paraplegia. J Med Devices 6(4):1–6. 10.1115/1.4007181 Raab K, Krakow K, Tripp F, Jung M (2016) Effects of training with the ReWalk exoskeleton on quality of life in incomplete spinal cord injury: a single case study. Spinal Cord Ser Cases 1(1):15025. https://doi.org/10.1038/scsandc.2015.25 Rahman T, Sample W, Jayakumar S, King MM, Wee JY, Seliktar R, Alexander M, Scavina M, Clark A (2006) Passive exoskeletons for assisting limb movement. J Rehabil Res Dev 43(5):583–590. https://doi.org/10.1682/JRRD.2005.04.0070 Reed MD, Van Nostran W (2014) Assessing pain intensity with the visual analog scale: a plea for uniformity. J Clin Pharmacol 54(3):241–144. https://doi.org/10.1002/jcph.250 Reybrouck T (2003) Clinical usefulness and limitations of the 6-minute walk test in patients with cardiovascular or pulmonary disease. Chest 123(2):325–327. https://doi.org/10.1378/ chest.123.2.325 Saji N, Kimura K, Ohsaka G, Higashi Y, Teramoto Y, Usui M, Kita Y (2015) Functional independence measure scores predict level of long-term care required by patients after stroke: a multicenter retrospective cohort study. Disabil Rehabil 37(4):331–337. https://doi.org/ 10.3109/09638288.2014.918195 Sanz-Merodio D, Cestari M, Arevalo JC, Garcia E (2012) A lower-limb exoskeleton for gait assistance in quadriplegia. In: IEEE International Conference on Robotics and Biomimetics: ROBIO 2012, Guangzhou, 11–14 Dec 2012. pp 122–127. https://doi.org/10.1109/ ROBIO.2012.6490954 Sczesny-Kaiser M, Höffken O, Lissek S, Lenz M, Schlaffke L, Nicolas V, Meindl R, Aach M, Sankai Y, Schildhauer TA, Tegenthoff M, Schwenkreis P (2013) Neurorehabilitation in chronic paraplegic patients with the HAL® exoskeleton – preliminary electrophysiological and fMRI data of a pilot study. In: Pons JL, Torricelli D, Pajaro M (eds) Converging clinical and engineering research on neurorehabilitation. Springer, Berlin, pp 611–615. https://doi.org/ 10.1007/978-3-642-34546-3_99 Shin JC, Yoo JH, Jung TH, Goo HR (2011) Comparison of lower extremity motor score parameters for patients with motor incomplete spinal cord injury using gait parameters. Spinal Cord 49(4): 529–533. https://doi.org/10.1038/sc.2010.158 Spungen AM, Asselin P, Fineberg DB, Kornfeld SD, Harel NY (2013) Exoskeletal-assisted walking for persons with motor-complete paraplegia. In: STO Human Factors and Medicine Panel (HFM) Symposium, Milan, 15–17 Apr 2013 Stein J, Bishop L, Stein DJ, Wong CK (2014) Gait training with a robotic leg brace after stroke: a randomized controlled pilot study. Am J Phys Med Rehabil 93(11):987–994. https://doi.org/ 10.1097/PHM.0000000000000119 Stookey AD, Katzel LI, Steinbrenner G, Shaughnessy M, Ivey FM (2014) The short physical performance battery as a predictor of functional capacity after stroke. J Stroke Cerebrovasc Dis 23(1):130–135. https://doi.org/10.1016/j.jstrokecerebrovasdis.2012.11.003 Strausser KA, Kazerooni H (2011) The development and testing of a human machine interface for a mobile medical exoskeleton. In: IEEE/RSJ International Conference on Intelligent Robots and

Gait Rehabilitation with Exoskeletons

1469

Systems: IROS 2011, San Francisco, 25–30 Sep 2011. pp 4911–4916. https://doi.org/10.1109/ IROS.2011.6095025 Strausser KA, Swift TA, Zoss AB, Kazerooni H (2010) Prototype medical exoskeleton for paraplegic mobility: first experimental results. In: ASME 2010 Dynamic Systems and Control Conference: DSCC 2010, Cambridge, MA, 12–15 Sep 2010. ASME, pp 453–458. https://doi. org/10.1115/DSCC2010-4261 Suzuki K, Kawamura Y, Hayashi T, Sakurai T, Hasegawa Y, Sankai Y (2005) Intention-based walking support for paraplegia patient. In: IEEE International Conference on Systems, Man and Cybernetics: SMC 2005, Waikoloa, 10–12 Oct 2005. pp 2707–2713 Vol. 2703. https://doi.org/ 10.1109/ICSMC.2005.1571559 Sylos-Labini F, La Scaleia V, d’Avella A, Pisotta I, Tamburella F, Scivoletto G, Molinari M, Wang S, Wang L, van Asseldonk E, van der Kooij H, Hoellinger T, Cheron G, Thorsteinsson F, Ilzkovitz M, Gancet J, Hauffe R, Zanov F, Lacquaniti F, Ivanenko YP (2014) EMG patterns during assisted walking in the exoskeleton. Front Hum Neurosci 8(423):1–12. https://doi. org/10.3389/fnhum.2014.00423 Talaty M, Esquenazi A, Briceño JE (2013) Differentiating ability in users of the ReWalk™ powered exoskeleton: an analysis of walking kinematics. In: IEEE International Conference on Rehabilitation Robotics: ICORR 2013 Seattle, 24–26 Jun 2013. pp 1–5. https://doi.org/10.1109/ ICORR.2013.6650469 Tinetti ME (1986) Performance-oriented assessment of mobility problems in elderly patients. J Am Geriatr Soc 34(2):119–126. https://doi.org/10.1111/j.1532-5415.1986.tb05480.x Tsukahara A, Hasegawa Y, Sankai Y (2009) Standing-up motion support for paraplegic patient with robot suit hal. In: IEEE International Conference on Rehabilitation Robotics: ICORR 2009, Kyoto, 23–26 Jun 2009. pp 211–217. https://doi.org/10.1109/ICORR.2009.5209567 Tsukahara A, Kawanishi R, Hasegawa Y, Sankai Y (2010) Sit-to-stand and stand-to-sit transfer support for complete paraplegic patients with robot suit HAL. Adv Robotics 24(11):1615–1638. https://doi.org/10.1163/016918610X512622 Watanabe H, Tanaka N, Inuta T, Saitou H, Yanagi H (2014) Locomotion improvement using a hybrid assistive limb in recovery phase stroke patients: a randomized controlled pilot study. Arch Phys Med Rehabil 95(11):2006–2012. https://doi.org/10.1016/J.Apmr.2014.07.002 Whitney SL, Wrisley DM, Marchetti GF, Gee MA, Redfern MS, Furman JM (2005) Clinical measurement of sit-to-stand performance in people with balance disorders: validity of data for the five-times-sit-to-stand test. Phys Ther 85(10):1034–1045 Wolf SL, Catlin PA, Gage K, Gurucharri K, Robertson R, Stephen K (1999) Establishing the reliability and validity of measurements of walking time using the Emory functional ambulation profile. Phys Ther 79(12):1122–1133 Yang N, Zhang B, Gao C (2014) The baseline NIHSS score in female and male patients and shorttime outcome: a study in young ischemic stroke. J Thromb Thrombolysis 37(4):565–570. https://doi.org/10.1007/s11239-013-0986-9 Zeilig G, Weingarden H, Zwecker M, Dudkiewicz I, Bloch A, Esquenazi A (2012) Safety and tolerance of the ReWalk™ exoskeleton suit for ambulation by people with complete spinal cord injury: a pilot study. J Spinal Cord Med 35(2):96–101. https://doi.org/10.1179/ 2045772312Y.0000000003

Brain-Computer Interfaces for Motor Rehabilitation Rüdiger Rupp

Abstract

Injuries of the central nervous system such as stroke or spinal cord injury are a major cause for severe motor impairments. The inability to activate muscles voluntarily and the associated loss of ambulatory or manipulation skills constitute a substantial handicap in the patients’ life. This results in a tremendously reduced quality of life and represents a severe barrier for social and professional integration. Therefore, rehabilitation aims at achieving the greatest amount of autonomy in everyday life by application of restorative or compensatory therapeutic approaches. While restorative strategies are based on principles of motor learning and aim at the recovery of the original function, compensatory therapies are applied in cases where recovery is unlikely to happen and are often based on the use of assistive technology. Brain-computer interfaces (BCIs) are an emerging technology that measure brain activities and translate them into control signals for a variety of assistive devices. BCIs may contribute to both rehabilitative and compensatory therapeutic strategies. While most of the BCI-controlled assistive technology such as communication devices, robot arms, or neuroprostheses based on functional electrical stimulation focus on the compensation of a lost function, there are only a few, however, very promising examples on the successful use of rehabilitative BCIs for restoration of grasping or walking. Although assistive and rehabilitative BCIs seem to be a valuable component of motor rehabilitation programs, more end user studies are needed to reveal the full potential of BCIs for better participation and improved quality of life.

R. Rupp (*) Spinal Cord Injury Center – Experimental Neurorehabilitation, Heidelberg University Hospital, Heidelberg, Germany e-mail: [email protected] # Springer International Publishing AG, part of Springer Nature 2018 B. Müller, S.I. Wolf (eds.), Handbook of Human Motion, https://doi.org/10.1007/978-3-319-14418-4_67

1471

1472

R. Rupp

Keywords

Brain-computer interface • Brain-machine interface • Motor rehabilitation • Restoration • Recovery • Compensation • Robotic arm • Neuroprosthesis • Functional electrical stimulation • Grasp • Exoskeleton • Ambulation

Contents Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . State of the Art . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Noninvasive Brain-Computer Interfaces . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Invasive Brain-Computer Interfaces . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Applications of BCIs in Motor Rehabilitation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Brain-Computer Interfaces for Control of Assistive Devices . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Brain-Computer Interfaces for Movement Restoration . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

1472 1475 1475 1480 1483 1483 1489 1494 1496

Introduction Among the consequences of injuries or diseases of the central nervous system (CNS) are impairments of motor functions such as grasping, trunk stability, or walking. The limited ability to voluntarily initiate movements up to the complete loss of motor functions represents a major barrier for leading an independent, self-determined life and results in a substantial decrease of the perceived quality of life. People with restrictions of motor functions are often not fully integrated in the society resulting in a relevant economic burden on the social system. Stroke is one of the most prevalent neurological conditions worldwide and one of the leading causes of motor impairment (Warlow et al. 2008). Strokes can be classified into ischemic and hemorrhagic. Ischemic strokes, which form the majority of strokes, are caused by an interruption of the blood supply of parts of the brain, while hemorrhagic strokes result from the rupture of a blood vessel and the associated bleeding. In Europe every year 1.1 Mio. first strokes occur, from which 70% survive the first year, 80% with a hemiparesis/-plegia, i.e., impaired to no upper and lower extremity function on one side of the body (Truelsen et al. 1997). Estimates suggest that ~80% of all stroke survivors with upper limb motor deficits do not fully regain the function of the affected limb (Hendricks et al. 2002). Moreover, the economic burden of stroke is high and is likely to increase in the future (Feigin et al. 2009). In Europe, an estimated number of 330,000 people are affected by the consequences of spinal cord injury (SCI) with 11,000 new injuries per year (Ouzký 2002; van den Berg et al. 2010). Approximately 50% of the SCI population are tetraplegic due to injuries of the cervical spinal cord with paralyses of the lower as well as the upper extremities. The bilateral loss of the grasp function severely limits the affected individuals’ ability to live independently (Anderson 2004; Snoek et al. 2004) and retain gainful employment postinjury (NSCISC 2015). In industrial countries, there

Brain-Computer Interfaces for Motor Rehabilitation

1473

is trend toward a higher percentage of people with incomplete SCI and with high cervical injuries (McCaughey et al. 2016). Additionally, in the western world, the percentage of nontraumatic SCI due to spinal infections, ischemia, or degenerative spine diseases increases over the last decades resulting in more incomplete injuries with preserved sensory and/or motor functions below the level of injury (Exner 2004). The aim of rehabilitation of neurological patients is to provide them with as much autonomy as possible. A strong focus is put on the ability to perform activities of daily living such as dressing/undressing, personal hygiene, eating, computer operation, or mobility in indoor and outdoor environments. In this context, motor rehabilitation is based on two fundamental therapeutic principles, namely, restoration and compensation/substitution. Restoration or recovery stands for the reacquisition of elemental motor patterns by motor learning. In the absence of reacquisition, functional improvements are achieved by adaptation of remaining (compensation) or integration of alternative (substitution) motor elements. The term recovery of motor performance is defined as the restoration of elemental motor pattern present prior to CNS injury meaning that a given task is performed using the same end effectors and joints in the same movement patterns typically used by able-bodied individuals. Although the neurobiological basis for motor recovery is not yet fully understood, it is generally accepted that the structural and functional reorganization of neural networks at different levels of the CNS largely contributes to it. This ability of the CNS for reorganization is called neuroplasticity. In chronic stroke, the present view of brain reorganization is that overuse of the unimpaired contralesional and underuse of the ipsilesional hemisphere leads to increased inhibition of the ipsilesional hemisphere by the contralesional hemisphere. This inhibition is thought to block excitatory reorganization of the intact ipsilesional areas and block recovery of the affected motor system (Ward and Cohen 2004). This hypothesis is supported by the positive effects of constraint-induced movement therapy (CIMT) in chronic stroke (Mark and Taub 2004). In this therapy, physical restraint of the healthy limb for an extended period of time forces the patient to use the paretic limb and increases excitatory neural activity in the lesioned hemisphere. Therefore, modern stroke rehabilitation aims at assisting or inducing the reorganization of neural circuits to achieve motor recovery. Motor compensation is defined as the appearance of new motor patterns resulting from the adaptation of remaining motor elements or substitution, meaning that functions are taken over, replaced, or substituted by technical aids or assistive devices. The decision, whether a restorative or compensatory/substitutional therapeutic strategy is applied, depends in general on several factors, but time after injury and severity of the injury are the most important ones (Fig. 1). The potential for spontaneous motor recovery is highest during the first 6 months after SCI or stroke (Langhorne et al. 2011; Curt et al. 2004). In patients with initially severe impairments or complete SCI, the potential for recovery is much lower than in patients with moderate impairments or motor incomplete lesions. The resulting commonly accepted therapeutic guideline states that in less severe and more acute CNS injuries, the therapeutic focus is on restoration of the original function. This does not mean

1474

R. Rupp

Fig. 1 Therapeutic strategy in relation to severity and time after CNS lesion

that intensive restorative therapies do not lead to any functional improvement in the chronic stage, but the degree of the improvement is likely much lower than in the subacute stage. Over the last two decades, the use of technology in the rehabilitation of motor impairments has constantly increased both in restorative therapies based on motor learning approaches as well as for compensation or substitution of permanently restricted or lost motor functions. In a restorative setting, body weight-supported treadmill training with or without the help of motorized gait orthoses helps to achieve a high number of task repetitions – one of the key principles of motor learning – and represent an established therapy for improving ambulatory functions (Dietz and Fouad 2014). Passive or active arm support systems help patients to accomplish specific upper extremity motor tasks and thereby improving their motor skills (Prochazka 2015). If lower extremity function is lost, wheelchairs provide some degree of mobility. In recent times, active mobile exoskeletons allow individuals with no or almost no residual muscle functions in the lower extremities but stable trunk to ambulate at least over short distances (Rupp et al. 2015b). Grasp neuroprostheses on the basis of functional electrical stimulation (FES) alone or in combination with an active orthosis or robot arms with several degrees of freedom compensate for the loss of upper extremity function (Rupp et al. 2015a). Recent advances in neurotechnology have led to the development of braincomputer interfaces (BCIs, also named brain-machine interfaces (BMIs)) that might become an important component of new technological strategies that strive to overcome severe CNS-injury-related motor impairments. First, BCIs may bypass

Brain-Computer Interfaces for Motor Rehabilitation

1475

nonfunctional corticospinal pathways to allow for brain control of FES devices or high-dimensional robotic arms to assist in daily life activities (assistive BCI). Second, BCIs could facilitate CNS and in particular brain neuroplasticity (see chapter ▶ “Observing and Learning Complex Actions: On the Example of Guitar Playing”), thus aiming at movement restoration by enhancing motor learning and motor recovery. If a BCI is used in the context of restoration of motor functions, it is often referred to as rehabilitative BCI, although the term “restorative BCI” would be more precise (Soekadar et al. 2015; Chaudhary et al. 2016).

State of the Art BCIs are technical systems that provide a direct connection between the human brain and a computer (Wolpaw et al. 2002). These systems are able to detect thoughtmodulated changes in electrophysiological brain activity and transform the changes into control signals (Fig. 2). A BCI system consists of five sequential components: (1) signal acquisition, (2) feature extraction, (3) feature translation, and (4) classification output, which interfaces to an output device, and is providing (5) feedback to the user. These components are controlled by an operating protocol that defines the onset and timing of operation, the details of signal processing, the nature of the device commands, and the oversight of performance (Shih et al. 2012). Although all implementations of BCIs build upon the same basic components, they differ substantially in regard to the degree of invasiveness, their complexity in terms of hard- and software components, the underlying physiological mechanisms, and their basic mode of operation (cue based, synchronous vs. asynchronous).

Noninvasive Brain-Computer Interfaces Noninvasive BCIs obviously represent the first choice in end-user applications due to their ease of application. Although a variety of different data acquisition methods can be used for setup of a BCI, e.g., near infrared spectroscopy (NIRS), functional magnetic resonance imaging (fMRI), and magnetoencephalography (MEG), most of the practically used noninvasive systems rely on brain signals that are recorded by electrodes on the scalp (electroencephalogram, EEG). EEG-based BCI systems are based on relatively inexpensive, small equipment, which is commercially available, and therefore offer the possibility of everyday use in the clinical rehabilitation environment and at end users’ homes. Depending on the residual capabilities of the end user and the requirements of the application in terms of intuitiveness, training time, or speed, a variety of EEG signals have been used as noninvasive measures of brain activity. The most common are signals such as event-related potentials (ERPs), steady-state responses mostly visual evoked potentials (SSVEP), and frequency oscillations particularly of sensorimotor rhythms (SMRs). For rehabilitative BCIs, mainly the motor-imagery- or

1476

R. Rupp

Fig. 2 General framework of brain-computer interfaces for motor rehabilitation: Invasive BCI approaches (right) include the measurement of local field potentials (LFPs) and electrocorticography (ECoG). Noninvasive BCI approaches (left) are mainly based on EEG. Brain signals are processed to extract features relevant to the aim of the BCI and then classified using a translational algorithm to construct a control signal that drives the BCI. BCIs can be classified as assistive to substitute for a loss of motor functions or as rehabilitative to enhance neural recovery

motor-execution-based modulation of SMRs registered with electrodes over the motor cortex is used. Depending on the underlying signal, EEG-based BCIs can be categorized into endogenous, asynchronous and exogenous, synchronous systems. Synchronous BCIs depend on the electrophysiological activity evoked by external stimuli and do therefore not require intensive training. The most common synchronous BCI is

Brain-Computer Interfaces for Motor Rehabilitation

1477

Table 1 A overview of the most common practical types of EEG-based assistive BCIs together with their minimal number of electrodes, a qualitative estimation of typical training times, and their typical accuracy and bit rate

BCI SMR (2-class) P300 SSVEP

Parameter Minimal (typical) number of electrodes 4 (10) + 1 reference + 1 ground 3 (9) + 1 reference + 1 ground 6 + 1 reference + 1 ground

Training time Weeks to months Minutes to 1 nm/kg. In hip arthritis, difficulty with stair climbing has been consistently reported, due to the weakness of the major hip muscles leading to reduced muscle moments (Fig. 7) and an altered power generation (Fig. 8). Power generation is seen to reduce significantly in hip arthritis which would lead to compensation strategies to achieve the range of motion required to initiate stair climbing. The inability of muscles to produce enough power for the increased range of motion required during stair climbing is compensated by a reduced stair climbing speed in these patients. All of the above results in patients adopting altered angular and loading strategies when both ascending and descending stairs, which may lead to bilateral asymmetries.

Effects of Total Hip Arthroplasty on Gait

1517

Fig. 8 Hip joint power during ascend and descend of stairs for healthy adults (gray), patients with hip arthritis (black), and patients following 1-year total hip arthroplasty (red). Results are from the Motion Analysis Laboratory, at Mayo Clinic, Rochester, MN

Following THA, joint muscle moment and power are seen to improve significantly, returning the power generation to near normal requirements for stair climbing. However, the improvement is not comparable to a healthy hip. Note that the transverse pain moments, though significantly lower compared to the other two planes, could have some effect due to the altered pattern during both ascending and descending stairs.

Future Work Research suggests that investigating the extent of remaining preoperative gait patterns at each joint would help improve rehabilitation, by reducing preoperative gait abnormalities as early as possible. Large-scale studies are therefore needed, assessing all lower extremity joints both pre- and postoperatively for both short- and long-term THA outcomes. There is also a need for a simplified objective gait score, assessing the severity of gait alterations in patients with hip arthritis which would assist clinicians in planning treatments, accordingly.

Cross-References ▶ 3D Dynamic Pose Estimation from Marker-Based Optical Data ▶ 3D Dynamic Probabilistic Pose Estimation from Data Collected Using Cameras and Reflective Markers ▶ Gait Parameters Estimated Using Inertial Measurement Units ▶ Interpreting Spatiotemporal Parameters, Symmetry, and Variability in Clinical Gait Analysis ▶ Three-Dimensional Human Kinematic Estimation Using Magneto-Inertial Measurement Units

1518

S. Chopra and K.R. Kaufman

References Akiyama K, Nakata K, Kitada M, Yamamura M, Ohori T, Owaki H, Fuji T (2016) Changes in axial alignment of the ipsilateral hip and knee after total hip arthroplasty. Bone Joint J 98-B:349–358 Andriacchi TP, Andersson GB, Fermier RW, Stern D, Galante JO (1980) A study of lower-limb mechanics during stair-climbing. J Bone Joint Surg Am 62:749–757 Barrett WP, Turner SE, Leopold JP (2013) Prospective randomized study of direct anterior vs postero-lateral approach for total hip arthroplasty. J Arthroplast 28:1634–1638 Beaulieu ML, Lamontagne M, Beaulé PE (2010) Lower limb biomechanics during gait do not return to normal following total hip arthroplasty. Gait Posture 32:269–273 Bejjani FJ, Lockett R, Pavlidis L (1992) Videofluoroscopy system for in vivo motion analysis. Google Patents Ben-Galim P, Ben-Galim T, Rand N, Haim A, Hipp J, Dekel S, Floman Y (2007) Hip-spine syndrome: the effect of total hip replacement surgery on low back pain in severe osteoarthritis of the hip. Spine (Phila Pa 1976) 32:2099–2102 Centers For Disease Control and Prevention (2010) National Hospital Discharge Survey: 2010 table, procedures by selected patient characteristics – number by procedure category and age Della Croce U, Cappuzzo A, Kerrigan DC (1999) Pelvis and lower limb anatomical landmark calibration precision and its propagation to bone geometry and joint angles. Med Biol Eng Comput 37:155–161 Evenson KR, Buchner DM, Morland KB (2012) Objective measurement of physical activity and sedentary behavior among US adults aged 60 years or older. Prev Chronic Dis 9:E26 Fiorentino NM, Kutschke MJ, Atkins PR, Foreman KB, Kapron AL, Anderson AE (2016) Accuracy of functional and predictive methods to calculate the hip joint center in young non-pathologic asymptomatic adults with dual fluoroscopy as a reference standard. Ann Biomed Eng 44:2168–2180 Foucher KC, Wimmer MA (2012) Contralateral hip and knee gait biomechanics are unchanged by total hip replacement for unilateral hip osteoarthritis. Gait Posture 35:61–65 Hagstromer M, Oja P, Sjostrom M (2007) Physical activity and inactivity in an adult population assessed by accelerometry. Med Sci Sports Exerc 39:1502–1508 Harding P, Holland AE, Delany C, Hinman RS (2014) Do activity levels increase after total hip and knee arthroplasty? Clin Orthop Relat Res 472:1502–1511 Horstmann T, Listringhaus R, Haase GB, Grau S, Mundermann A (2013) Changes in gait patterns and muscle activity following total hip arthroplasty: a six-month follow-up. Clin Biomech (Bristol, Avon) 28:762–769 Lin BA, Thomas P, Spiezia F, Loppini M, Maffulli N (2013) Changes in daily physical activity before and after total hip arthroplasty. A pilot study using accelerometry. Surgeon 11:87–91 Louriuro A, Mills PM, Barrett RS (2013) Muscle weakness in hip osteoarthritis: a systematic review. Arthritis Care Res 65:340–352 Lubbeke A, Zimmermann-Sloutskis D, Stern R, Roussos C, Bonvin A, Perneger T, Peter R, Hoffmeyer P (2014) Physical activity before and after primary total hip arthroplasty: a registry-based study. Arthritis Care Res 66:277–284 Madsen MS, Ritter MA, Morris HH, Meding JB, Berend ME, Faris PM, Vardaxis VG (2004) The effect of total hip arthroplasty surgical approach on gait. J Orthop Res 22:44–50 OECD (2015) Health at a glance 2015: OECD indicators (summary). OECD Publishing, Paris Perry J, Burnfield J (2010) Gait analysis: normal and pathological function. J Sports Sci Med 9:353 Queen RM, Butler RJ, Watters TS, Kelley SS, Attarian DE, Bolognesi MP (2011) The effect of total hip arthroplasty surgical approach on postoperative gait mechanics. J Arthroplast 26:66–71 Queen RM, Newman ET, Abbey AN, Vail TP, Bolognesi MP (2013) Stair ascending and descending in hip resurfacing and large head total hip arthroplasty patients. J Arthroplast 28:684–689

Effects of Total Hip Arthroplasty on Gait

1519

Queen RM, Appleton JS, Butler RJ, Newman ET, Kelley SS, Attarian DE, Bolognesi MP (2014) Total hip arthroplasty surgical approach does not alter postoperative gait mechanics one year after surgery. PM R 6:221–226; quiz 226 Queen RM, Attarian DE, Bolognesi MP, Butler RJ (2015) Bilateral symmetry in lower extremity mechanics during stair ascent and descent following a total hip arthroplasty a one-year longitudinal study. Clin Biomech (Bristol, Avon) 30:53–58 Reardon K, Galea M, Dennett X, Choong P, Byrne E (2001) Quadriceps muscle wasting persists 5 months after total hip arthroplasty for osteoarthritis of the hip: a pilot study. Intern Med J 31:7–14 Roberts A (2010) Gait analysis: normal and pathological function (2nd edition). Bone Joint J 92-B(8):1184 Shrader W, Bhowmik-Stoker M, Jacofsky MC, Jacofsky DJ (2009) Gait and stair function in total and resurfacing hip arthroplasty: a pilot study. Clin Orthop Relat Res 467:1476–1484 Tao W, Liu T, Zheng R, Feng H (2012) Gait analysis using wearable sensors. Sensors (Basel, Switzerland) 12:2255–2283 Umeda N, Miki H, Nishii T, Yoshikawa H, Sugano N (2009) Progression of osteoarthritis of the knee after unilateral total hip arthroplasty: minimum 10-year follow-up study. Arch Orthop Trauma Surg 129:149–154 Van Den Bogert AJ, Read L, Nigg BM (1999) An analysis of hip joint loading during walking, running, and skiing. Med Sci Sports Exerc 31:131–142 Watelain E, Dujardin F, Babier F, Dubois D, Allard P (2001) Pelvic and lower limb compensatory actions of subjects in an early stage of hip osteoarthritis. Arch Phys Med Rehabil 82:1705–1711 Weber T, Al-Munajjed AA, Verkerke GJ, Dendorfer S, Renkawitz T (2014) Influence of minimally invasive total hip replacement on hip reaction forces and their orientations. J Orthop Res 32:1680–1687 Wesseling M, Meyer C, Corten K, Simon JP, Desloovere K, Jonkers I (2016) Does surgical approach or prosthesis type affect hip joint loading one year after surgery? Gait Posture 44:74–82 Yoshimoto H, Sato S, Masuda T, Kanno T, Shundo M, Hyakumachi T, Yanagibashi Y (2005) Spinopelvic alignment in patients with osteoarthrosis of the hip: a radiographic comparison to patients with low back pain. Spine (Phila Pa 1976) 30:1650–1657

Effects of Knee Osteoarthritis and Joint Replacement Surgery on Gait Cheryl L. Hubley-Kozey and Janie Astephen Wilson

Abstract

Knee joint osteoarthritis (OA) is the most common condition that is managed with knee joint replacement surgery. Gait has provided a model to study knee OA processes as knee joint function is altered in the presence of knee OA including joint level angular motions, joint level moments, and muscle activation patterns during walking. In the pre-total joint arthroplasty state, there is significant joint damage and severe symptoms (pain) that manifest as altered musculoskeletal function during walking. These alterations in joint and muscle function during gait can be accurately quantified with state-of-the-art surface motion capture systems, force plates, biomechanical modeling, and surface electromyography to capture muscle activation patterns. This chapter provides an overview of the gait biomechanics and EMG studies related to (i) knee joint OA processes and (ii) knee joint replacement surgery. The summary and conclusions provide ideas on future directions suggesting that human movement studies need to be better integrated in clinical musculoskeletal practice to provide the quantitative measures of function that can inform clinical decision making.

C.L. Hubley-Kozey (*) School of Physiotherapy, Faculty of Health Professions, Dalhousie University, Halifax, NS, Canada School of Biomedical Engineering, Faculty of Medicine, Dalhousie University, Halifax, NS, Canada Nova Scotia Health Authority, Halifax, NS, Canada e-mail: [email protected]; [email protected] J. Astephen Wilson School of Biomedical Engineering, Faculty of Medicine, Dalhousie University, Halifax, NS, Canada Nova Scotia Health Authority, Halifax, NS, Canada e-mail: [email protected] # Springer International Publishing AG, part of Springer Nature 2018 B. Müller, S.I. Wolf (eds.), Handbook of Human Motion, https://doi.org/10.1007/978-3-319-14418-4_82

1521

1522

C.L. Hubley-Kozey and J. Astephen Wilson

Keywords

Knee Arthroplasty • Knee Osteoarthritis • Gait Biomechanics • Joint Function • Muscle Function • Electromyography

Contents Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . State of the Art . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Gait Features Associated with Knee OA Processes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Changes in Gait Metrics Associated with Knee OA Presence and Severity . . . . . . . . . . . . . . Knee Joint Angles and Knee OA . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Knee Joint Moments and Knee OA . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Surface Electromyography and Knee OA . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Gait Metrics Associated with Predicting Progression Outcomes . . . . . . . . . . . . . . . . . . . . . . . . . . Gait Metrics and Joint Replacement Surgery . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Knee Flexion Angles and Total Knee Arthroplasty Surgery . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Knee Joint Moments and Total Knee Arthroplasty . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . EMG and Post-Total Joint Arthroplasty . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Predicting TKA Outcomes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Summary and Future Directions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Cross-References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

1522 1524 1524 1524 1525 1528 1531 1535 1536 1537 1539 1541 1543 1544 1545 1545

Introduction Knee joint osteoarthritis (OA) is the most common condition that is managed with knee joint replacement surgery. OA is characterized by an imbalance in the normal coupling of articular cartilage degradation and synthesis, with biomechanical factors related to joint loading thought to trigger the biochemical responses leading to degradation of the cartilage. While articular cartilage degradation is the hallmark of OA, the OA disease process affects the entire joint including cartilage, bone, muscles, nerves, and ligaments (Brandt et al. 2006). The knee is the most common lower limb joint affected by OA, with knee OA the most common cause of pain and disability in older adults. Over the past 40 years researchers have used walking (gait) as a model to understand the relationship between joint level biomechanics and OA processes in vivo, because walking is the activity of daily living that has the highest frequency of occurrence and provides significant cumulative load to the knee joint (Andriacchi et al. 2004). Joint level biomechanics of the knee joint show that knee joint motion and joint loading are altered with OA. In particular, data from joint moments has improved our understanding of the link between joint loading and OA outcomes. The majority of OA gait studies have focused on joint kinematics and kinetics, with only a few studies examining muscle function during walking using surface electromyography (EMG). EMG provides complimentary information to joint biomechanics as the muscles surrounding the knee joint play a significant role in joint loading, motion, and stability. Given the limitations of using net resultant joint

Effects of Knee Osteoarthritis and Joint Replacement Surgery on Gait

1523

moments to represent true joint contact forces during gait, researchers are also incorporating EMG data into biomechanical models to develop more physiological models of knee joint contact forces in OA (Manal et al. 2015). Numerous studies have examined joint level biomechanics and muscle activation of healthy gait. This chapter focuses on gait alterations with knee OA severity and the effect of standard of care knee joint arthroplasty surgery on walking gait. There are two progressive knee OA processes that take place. One is the worsening of symptoms (e.g., pain, stiffness), and the other is worsening of the structural damage (e.g., loss of cartilage resulting in joint space narrowing, formation of osteophytes, changes in subchondral bone). These two processes are not always linked as there is a discordance reported between severity of pain and structural damage (Barker et al. 2004). Currently there is no cure for knee OA, with management focused on pain relief primarily through pharmaceutical interventions. There are also guidelines supporting exercise and physical activity and biomechanical interventions that aim to alter joint loading (bracing and heel wedges). While these management approaches have been shown to reduce pain, there is minimal evidence to show that they slow the degradation rate of the joint structures. The treatment for end-stage OA is joint replacement surgery, and clinical decisions for joint replacement are based on the presence and severity of both severe pain/symptoms and structural damage (Gossec et al.). Relevant to this chapter is that both pain and structural damage influence joint function with gait metrics providing a quantitative method to assess changes in OA severity and the functional effect of interventions for OA, both surgical and conservative. Gait metrics have been identified that are predictive of structural progression outcomes, clinical progression to total knee arthroplasty (TKA) surgery, and TKA outcomes. Figure 1 illustrates this reciprocal relationship where knee joint pathology (structural and functional impairment) associated with knee OA processes affect movements that are necessary for daily living and how these movements can impact joint pathological processes. This chapter presents current knowledge of the changes in gait biomechanics and muscle

KNEE OSTEOARTHRITIS

Impacts Pathology

STRUCTURE/ FUNCTIONAL IMPAIRMENT

Impacts Movement

ACTIVITY LIMITATIONS/ ALTERATIONS

Fig. 1 People with knee osteoarthritis (OA) have structural damage to joint tissues and pain that impacts mobility, such as walking. How they walk can also impact pain and pathology progression. This figure depicts the reciprocal relationship where joint pathology affects movement and movement can affect joint pathology. # Cheryl Hubley-Kozey 2017

1524

C.L. Hubley-Kozey and J. Astephen Wilson

activity related to (i) knee joint OA processes and (ii) knee joint replacement surgery. The goal is to provide an overview of the state of knowledge on the key biomechanical and muscle activation features relevant to the clinical management of knee OA and to stimulate interest in advancing knee OA gait research.

State of the Art Biomechanics has evolved over the past 50 years to permit sensitive and accurate quantitative measures of joint motion, joint moments, and muscle activation. The key tools are motion tracking sensors and embedded force plates in a laboratory setting, with modeled three-dimensional (3D) joint-level kinematics and joint-level kinetics based on inverse dynamics equations the most common data reported. In addition, some laboratories include sensing of muscle activation through surface EMG as the muscles are substantial contributors to joint loading, motion, and stability. Various forms of accelerometer and GPS technologies are being integrated to study the frequency of loading and advances in wearable motion sensing technologies, and signal processing approaches will help advance our understanding of joint level mechanics in real-world settings. Other technologies such as real-time feedback of walking mechanics, video fluoroscopy to examine bone/joint contact dynamics, and perturbation tools to stress gait have added to our laboratory-based measurements. These technologies and advances in analytical approaches including multivariate modeling and pattern recognition are evolving and will likely play a stronger role in the future.

Gait Features Associated with Knee OA Processes Knee OA affects knee joint function and subsequently overall function such as walking, rising from a chair, stair climbing, and stair descent. Gait studies have reported changes in walking velocity, 3D angular kinematics, and kinetics and muscle activation features in the presence of OA and with increasing OA severity of symptoms and structural damage. In addition, emerging evidence is showing that gait metrics are predictive of knee OA outcomes such as structural and clinical progression (which includes reduced function, worsening of symptoms, and progressive joint structural damage). The effect of OA progression processes on gait is important to understanding the functional implications of clinical progression and the pre-TKA surgery state.

Changes in Gait Metrics Associated with Knee OA Presence and Severity Stride characteristics, in particular walking velocity, have been considered as good indicators of overall musculoskeletal function. Early studies showed that in the

Effects of Knee Osteoarthritis and Joint Replacement Surgery on Gait

1525

presence of knee OA, individuals walked at slower self-selected walking speeds (Kaufman et al. 2001; Andriacchi et al. 1977; Teixeira and Olney 1996; Landry et al. 2007) than asymptomatic controls, indicative of reduced overall function. In later studies (Landry et al. 2007; Mundermann et al. 2004) where more mild to moderate knee OA participants were compared to asymptomatic controls, no differences in self-selected walking speed were found between groups. Indeed higher OA pain severity was associated with slower self-selected walking speed (Astephen Wilson et al. 2011). Given that walking speed can impact many knee joint level biomechanics and EMG measures, some researchers have had participants walk at a set speed (Hurwitz et al. 2002) or within a speed range (Childs et al. 2004) to remove the effect of walking speed differences. Researchers have also studied the effects of different walking velocities on knee joint level outcomes, clearly demonstrating an impact (Landry et al. 2007; Mundermann et al. 2004; Zeni and Higginson 2009). However, walking at slower speeds can misrepresent the mechanical load most often placed on the joint in vivo (Mundermann et al. 2004), and having people walk at a speed other than their self-selected speed may introduce variability associated with performing an unnatural movement. Others have included speed as a covariate when looking at OA differences in joint mechanics (Kaufman et al. 2001; Lewek et al. 2004; Zeni and Higginson 2009). However the value of such approaches has been questioned given that walking speed is part of the disease process and using it as a covariate may wash out part of the disease effects (Astephen Wilson 2012). This highlights that results of gait studies should be interpreted in light of the protocol. While differences among studies exist in protocol as well as in the demographics of the groups examined, the most consistent finding is that those with severe knee OA on average walk slower than asymptomatic controls and often slower than those with moderate knee OA (Zeni and Higginson 2009; Hubley-Kozey et al. 2009; Thorp et al. 2006). Other stride characteristics such as step length and step times walking speed is a good descriptive measure of overall measure of function, but has minimal diagnostic value related to knee OA as it does not provide information specific to the knee joint and knee muscle function. Three-dimensional joint level angular motions and moments provide a more specific assessment of joint function.

Knee Joint Angles and Knee OA While OA can occur in the medial, lateral, and patellofemoral compartments of the knee joint, the medial compartment is most commonly affected and hence the majority of gait metrics identified to date are specific to the medial compartment knee OA. Fig. 2 provides knee angle and net external resultant moment waveforms for illustrative purposes on asymptomatic controls, a moderate medial compartment knee OA group, and a severe medial compartment knee OA group that is within 1 week prior to receiving a total joint replacement with descriptions of these groups in previous publications (Landry et al. 2007; McKean et al. 2007; Astephen et al. 2008a). 3D angular motions provide a description of the dynamic joint-level

1526

b

Knee Flexion Angle

Knee Flexion Moment 0.6

80 ASYM (n=188) OA (n=239) Severe (n=98)

60 40

Progressive Severe

20 0

a 0

b

c

20

c

0.4

Moment (Nm/kg)

Angle (degrees)

a

C.L. Hubley-Kozey and J. Astephen Wilson

0

OA

–0.2

d

40 60 Percent Gait Cycle

Severe

0.2

Progressive a

–0.4 80

0

100

d

Knee Adduction Angle

b 20

c 40 60 80 Percent Stance Phase

100

Knee Adduction Moment 0.6

10

OA Moment (Nm/kg)

Angle (degrees)

8 6 4 2

0.4 0.2 0

0 a

–2 0

20

a

–0.2 60

80

0

100

f Moment (Nm/kg)

10

5

0 20

40 60 80 Percent Stance Phase

d 60

80

100

100

Knee Rotation Moment

0.2

0.15 0.1 Severe

0.05 0 –0.05

0

40

Percent Stance Phase

Knee Rotation Angle

15

c

b 20

Percent Stance Phase

e

Angle (degrees)

40

0

a 20

b 40 60 80 Percent Stance Phase

100

Fig. 2 Mean waveforms from the Dynamics of Human Motion laboratory, Dalhousie University for a group of asymptomatic adults over 40 years old (black), a moderate medial compartment knee OA group (blue), and a clinically severe OA, pre-total knee arthroplasty (TKA) group (red). (a) Knee flexion angle (KFA), (b) Net external knee flexion moment (KFM), (c) Knee adduction angle (KAA), (d) Net external knee adduction moment (KAM), (e) Knee rotation angle (KRA), and (f) Net external knee rotation moment (KRM). Positive indicates flexion, adduction, and internal rotation and negative indicates extension, abduction, and external rotation. Small letters are points where typical peak measures are taken for (a) KFA a = KFA at heel strike, b = early stance peak KFA, c = late stance KFA minimum, and d = swing phase peak; for (b) KFM a = early stance knee extension moment (KEM) peak, b = stance phase peak KFM, c = late stance KEM peak; (c) KAA a = mid-stance value, (d) KAM a = early stance peak knee abduction moment, b = first peak KAM, c = mid-stance KAM, and d = peak late stance KAM and (f) a = early stance peak external KRM and b = late stance peak internal KRM. Arrows indicate differences in amplitude measures with progressive meaning differences among all three groups, OA meaning the OA groups both

Effects of Knee Osteoarthritis and Joint Replacement Surgery on Gait

1527

function during gait. A typical knee flexion angle (KFA) waveform for an asymptomatic group is found in Fig. 2a illustrating that the knee is extended (straight) at heel contact followed by a gradual increase in flexion during weight acceptance, with the knee extending toward late stance followed by the largest flexion angle (around 60 degrees) during swing phase. There are several discrete measures often reported including KFA at heel contact, peak KFA, minimum late stance phase KFA (extension), the KFA range/excursion during stance (b–c in Fig. 2a), the peak KFA during swing. The most consistent finding in early studies was less KFA range during stance phase of the gait cycle (Baliunas et al. 2002; Schnitzer et al. 1993) in those with knee OA compared to controls. This decreased KFA range during stance is referred to as a “stiff knee” gait. Childs (2004) later reported a lower KFA range and less extension during late stance in those with knee OA. In contrast Kauffman’s OA participants walked at a slower speed (1.09 vs. 1.17 m/s) than asymptomatic controls, and he found no differences in peak KFA between groups when using walking speed as a covariate (Kaufman et al. 2001). Discrete features require a priori knowledge as to what features are important, and they do not capture the dynamic nature of the waveform through the gait cycle. Principal component analysis (PCA) is a technique that has been used to capture dynamic features over the entire gait waveform, including overall magnitude, differences operators, and phase shifts. It has the benefit of not requiring a priori knowledge in its selection of independent features of variation, and has been used to examine dynamic waveforms associated with knee OA gait (Deluzio and Astephen 2007; Astephen et al. 2008b; Federolf et al. 2013; Landry et al. 2007; Smith et al. 2004; Brandon et al. 2013). Using PCA on mild to moderate knee OA participants who walked at similar speeds to controls showed no differences in KFA waveform features based on PCA (Landry et al. 2007); however, McKean et al. (2007) reported on a similar moderate knee OA cohort, and found that women with OA had smaller KFA excursions during stance than men with OA (McKean et al. 2007). Sex effects for KFA were reported for OA gait mechanics in the two studies that examined them (McKean et al. 2007; Kaufman et al. 2001). Recent studies have shown that severity of OA can impact which KFA variables are different. The most common assessment of OA severity was based on measures of structural damage using the Kellgren and Lawrence (KL) radiographic grading scale (Kellgren and Lawrence 1957). Alterations in KFA metrics were reported among structural severity grades (Zeni and Higginson 2009). However, relevant to this chapter is that both structural damage and pain can influence joint function, and clinical severity is more relevant to TKA given that surgical decisions are based on

ä Fig. 2 (continued) differ from ASYM, and severe means just the severe group differs from the other two groups. The X-axis is percent of the gait cycle from heel contact to second heel contact on the tested leg for panel A only whereas all other panels are percent of stance phase from heel contact to toe off on the tested leg. # Cheryl Hubley-Kozey 2017

1528

C.L. Hubley-Kozey and J. Astephen Wilson

severity of symptoms and structural damage (Gossec et al. 2011). Fig. 2a illustrates a systematic decrease in early KFA when loading the joint with progression of clinical severity, whereas only the severe OA group had lower KFA than the other groups during swing phase based on both discrete and PCA features (Astephen et al. 2008a, b). Less well studied are the frontal (knee adduction angle, KAA) and transverse plane angles (knee rotation angle, KRA) illustrated in Fig. 2c and e; although both are deemed important to normal joint function, measures are not as reliable. KFA features have been shown to be reliable between days (ICCs 0.77–0.90) for discrete and PCA features for OA participants (Robbins et al. 2013), but the KAA and KRA featured were not, and suffer from the influence of kinematic cross-talk (Piazza and Cavanaugh 2000). Only the KAA mid-stance had an ICC of 0.74. Collectively, walking speed and motion features are descriptive measures of overall function and knee joint function, but do not provide direct information on the mechanical loading environment of the joint. Hence, these measures can be used for screening and monitoring change in clinical status, but are limited in their diagnostic content.

Knee Joint Moments and Knee OA Three-dimensional net resultant joint moments provide information related to OA knee joint function specific to joint loading magnitude, duration, and pattern. The OA gait literature can be confusing as inconsistencies exist with some papers reporting net internal moments and others reporting net external moments, both not always clearly stated. Net external resultant moments are the net sum of all external (gravitation, inertial, reaction forces) moments acting around the joint in each plane, which in turn are balanced by and therefore also represent the net internal moment created by all active muscles, joint contact, and passive tissue forces crossing the joint. Net resultant moments are often equated to joint loading, but it is important to recognize that they do not represent forces that are specific to any particular joint tissue including articular contact or specific muscles. Muscle forces are not easily measured, and due to redundancy and synergy of muscles spanning each joint, the contribution of individual muscles forces to a resultant joint force or moment at present cannot be easily estimated. All moments discussed and illustrated in this chapter are net external joint moments. Numerous discrete measures of dynamic moments during gait are often reported, illustrated in Fig. 2b, d, and f, making it difficult to compare between studies. Difficulties also exist in comparing amplitude values across studies as there are differences in amplitude units reported including nonnormalized values in Newton meters (Nm), amplitude-normalized to body mass (Nm/Kg) and normalized to body weight times height (Nm BW*height), and at times with units not provided. Moments therefore should be interpreted within the context of the normalization used, and if an absolute value is important such as in calculating a cumulative load, for example (Maly 2008), then nonnormalized units (Nm) may be most appropriate. Other applications may require normalization to determine relative changes.

Effects of Knee Osteoarthritis and Joint Replacement Surgery on Gait

1529

Interestingly, for some comparative findings the normalization technique used did not change the relationship of the differences between diagnostic groups (McKean et al. 2007; Hatfield et al. 2015a). The external knee adduction moment (KAM) in the frontal plane (Fig. 2d) and knee flexion moment (KFM) in the sagittal plane (Fig. 2b) are the two most commonly reported moments in knee OA studies. Fewer studies report on the external transverse plane moments related to internal and external rotation, referred to as the knee rotation moment (KRM). Between day reliability for KAM discrete and PCA measures are excellent, with most measures having ICCs greater than 0.90 (Birmingham et al. 2007; Robbins et al. 2013), and all but the initial stance peak KFM had good to excellent ICCs for a moderate OA group (Robbins et al. 2013). The peak and overall KRM during stance had very good reliability (ICC > 0.84).), but with ICCs for other features less than 0.5. The external KAM provides a measure of the ratio of the load between the medial and lateral compartment of the knee joint and is the most reported variable in knee OA gait studies, particularly for medial compartment knee OA. The typical dynamic waveform for the external KAM of asymptomatic adults includes a small knee abduction (negative) moment just after heel contact, followed by a rapid rise to a peak early in stance, then a dip at mid-stance associated with unloading the medial compartment, and then another peak mid to late stance (see Fig. 2d). This typical asymptomatic waveform has a bimodal, positive (indicating more load through medial compartment) pattern that is consistent with the vertical ground reaction force. The most reported feature of KAM is the initial peak KAM which has been shown to be influenced by body mass and walking velocity (Mundermann et al. 2004). Other measures include the mid-stance minimum, the peak in late stance and more recently, KAM impulse (the integral of the nontime normalized KAM during stance), and overall magnitude and shape features captured using PCA. The impulse captures the overall magnitude and duration of the KAM and is thought to represent medial compartment loading exposure, a different feature from the peak KAM (Thorp et al. 2006; Bennell et al. 2011). The KAM impulse is highly correlated to the overall magnitude feature from PCA, but very different waveform shapes can result in the same impulse value (Hatfield et al. 2015a). Early studies reported higher peak KAMs in knee OA compared to controls (Baliunas et al. 2002) even when all participants walked at a set speed of 1 m/s (Hurwitz et al.) Static alignment from mechanical axis measures was the best predictor of peak KAM but only accounted for 50% of the variance, emphasizing the need to assess frontal plane dynamics (Hurwitz et al.). Mundermann et al. (2004) reported similar peak KAM between moderate OA and asymptomatic controls consistent with the overall PCA magnitude features reported by Landry et al. (2007) as illustrated in Fig. 2b. In both studies, the peak values were increased with higher walking speed (Landry et al. 2007; Mundermann et al. 2004). The overall magnitude of KAM during stance was a feature that did not differ with increasing clinical severity (Astephen et al. 2008b) or increasing KL grade (Zeni and Higginson 2009), but was associated with a global measure of structural severity from radiographs (Astephen Wilson et al. 2011).

1530

C.L. Hubley-Kozey and J. Astephen Wilson

The second peak of KAM mid to late stance has been reported to be similar between asymptomatic controls and those with OA and between clinical OA severities (Astephen et al. 2008a) although a late stance difference between moderate severity (KL 3) and controls was reported and is consistent with Fig. 2d (Thorp et al. 2006). However, it has been noted that some asymptomatic adults (39%) and those with OA (52%, 30%) had no distinct second peak KAM (Hurwitz et al. 2002; Robbins et al. 2013), questioning the reliability of this feature. PCA models have captured a feature related to the late stance peak relative to the early stance peak, again with no reported differences with OA (Landry et al. 2007; McKean et al. 2007). The mid-stance minimum KAM has not been well studied, although some have identified the mid-stance value as important to differentiate clinical OA severity (Astephen et al. 2008a) and radiographic severity (Thorp et al. 2006) consistent with an early study that showed mid-stance KAM was better at differentiating between OA and controls than the peak (Weidenhielm et al. 1994). A feature of the KAM representing the mid-stance value captured with PCA was shown to be higher with clinical symptoms of OA compared to a group with the same structural evidence of OA but no symptoms (Astephen Wilson et al. 2016), supporting that the inability to unload is important in the clinical (symptom) manifestation of OA. The sagittal plane external knee flexion moment (KFM) represents the overall loading at the joint, and is counterbalanced primarily by the internal knee extensor muscle moment assuming minimal knee flexor muscle moments. A typical asymptomatic external KFM is depicted in Fig. 2c including an initial knee extension (negative) moment (KEM) of short duration just after heel strike, followed by a gradually increasing KFM that peaks early in stance during weight acceptance and gradually decreases to become a KEM that peaks mid to late stance, then decreases in magnitude toward foot off. In asymptomatic, the early KFM and mid-late stance KEM are often biphasic with approximately equal magnitude peaks. Differences in KFM have been reported between asymptomatic and those with knee OA as illustrated in Fig. 2b, with the most common feature the initial peak KFM with the early peak KEM, the mid-stance KEM, and the KFM-KEM stance range (b–c in Fig. 2b) also reported. The initial peak KFM is typically reduced in knee OA, and this was thought to reflect reduced knee quadriceps strength or a quadriceps avoidance gait. Decreased peak mid-stance KEM was found with knee OA (Baliunas et al. 2002; Kaufman et al. 2001), which was only apparent in those with severe knee OA and not more moderate levels (Astephen et al. 2008a). A PCA feature capturing the KFM-KEM range showed that moderate OA was associated with less difference between early stance KFM and later stance KEM compared to asymptomatic (McKean et al. 2007; Astephen et al. 2008b) with this difference influenced by sex. Women with OA had less range than women without OA, but with no difference between men with and without OA (McKean et al. 2007). The net external transverse plane knee rotation moment (KRM) typically follows a biphasic pattern in asymptomatic individuals, with a small external KRM (negative) in early stance, changing to a gradually increasing internal KRM during later stance (See Fig. 2f). McKean et al. (2007) using PCA found less early stance

Effects of Knee Osteoarthritis and Joint Replacement Surgery on Gait

1531

external KRM with OA, again with a sex effect, where women with OA had less early stance external KRM compared to asymptomatic women. Astephen et al. (2008a) also found that only those with severe clinical OA just prior to TKA had reduced late stance internal KRM (Fig. 2f). With only a few studies, the KRM changes are not as well understood.

Surface Electromyography and Knee OA Fewer studies have examined changes in the lower limb muscle activation during gait in those with knee OA, with the majority of surface EMG OA gait studies published within the past 15 years. There are differences among studies with respect to the muscles examined, the variables reported, and the amplitude normalization approaches used. Early studies included representative muscles from different lower limb functional muscle groups (Childs et al. 2004; Simon et al. 1983) whereas more recent studies recognized the importance of examining medial and lateral muscle pairs at the knee joint given the asymmetric distribution of the disease among compartments (Hubley-Kozey et al. 2006; Lewek et al. 2004; Zeni et al. 2010; Rudolph et al. 2007). Measures reported include onset and offset times and amplitude features, with more recent studies examining the relative activation of muscle pairs acting around the knee joint using the co-contraction index (CCI) (Rudolph et al. 2007) and dynamic waveform features using PCA (Brandon et al. 2013; Hubley-Kozey et al. 2006). Despite these differences there are consistent trends illustrating altered muscle activation patterns that help put the biomechanics alterations described above into context. Typical waveforms are depicted in Fig. 3 for medial and lateral gastrocnemius, vasti and hamstring muscle pairs for adult asymptomatic controls, with amplitude values normalized to maximum effort voluntary isometric contractions (values in % MVIC) to provide a physiological reference (Hubley-Kozey et al. 2006). These waveforms are consistent with numerous studies of healthy adult gait and the OA waveforms show qualitative differences from asymptomatic. One of the early EMG OA gait studies showed that the duration of muscle activation was significantly longer for the vastus lateralis, medial hamstring, tibialis anterior, and medial gastrocnemius muscles for those diagnosed with unilateral symptomatic knee OA based on clinical symptoms and radiographic evidence versus age-matched (62 years) asymptomatic controls during controlled walking at speeds between 1.12 and 1.34 m/s (Childs et al. 2004). The longer vastus lateralis duration was related to prolonged activation during stance, the medial gastrocnemius was earlier onset, and both the hamstring and tibialis anterior muscles had earlier onset and more prolonged activity. They also reported more agonist antagonist coactivation based on higher CCI for the OA group compared to controls for the vasti lateralis:medial hamstring pair (28% vs. 15%) and for the tibialis anterior:medial gastrocnemius pair (20% vs. 11%) (Childs et al. 2004) based on EMG amplitude normalized to MVIC. Their OA group was heterogeneous with respect to compartment affected although the majority of the participants did have some medial compartment involvement.

1532

C.L. Hubley-Kozey and J. Astephen Wilson

a 60

Severe

Severe

40

20

0

0

a 20

c

b 40

80

d Muscle Activity (%MVIC)

Muscle Activity (%MVIC)

VL EMG Progressive

40 30 20 10 0 0

20

80

20 a 0

20

60

80

100

80

100

80

100

60 VM EMG Severe

50 40 30 20 10 0

20

b 40

60

Percent Gait Cycle

f

LH EMG

b 40

Percent Gait Cycle

100

60

MH EMG 60

ASYM (n=186) OA (n=230) Severe (n=95)

50 40

Muscle Activity (%MVIC)

Muscle Activity (%MVIC)

Severe 40

0 60

Percent Gait Cycle

e

60

100

60

b 40

Severe

0 60

Percent Gait Cycle

50

MG EMG

80 Muscle Activity (%MVIC)

Muscle Activity (%MVIC)

b

LG EMG

80

Progressive

30 20 10 b

0 0

20

40

50 40 30

Severe

20 10 0

60

Percent Gait Cycle

80

100

0

20

b 40

60

Percent Gait Cycle

Fig. 3 Ensemble average surface EMG waveforms from the Dynamics of Human Motion laboratory, Dalhousie University for asymptomatic adults over 40 years old (black), a moderate medial compartment knee OA group (blue), and a clinically severe, pre-total knee arthroplasty group (red) as in Fig. 2. A and B are the lateral (LG) and medial gastrocnemius (MG), C and D are the vastus lateralis (VL) and vastus medialis (VM), and E and F are the Lateral (LH) and medial hamstring (MH) muscles. Amplitudes are normalized to maximum voluntary isometric contractions (% MVIC). Small letters are phases where typical amplitude measures are taken: a = early stance amplitudes and b = mid-stance amplitudes. In (a) dotted arrows = difference between a and b and in (b) the horizontal arrow = a phase shift both captured from PCA. CCI values are typically calculated pre-heel strike to peak KAM which can differ for different populations. Arrows indicate differences in measures between groups with progressive meaning differences among all three groups, OA meaning the OA groups both differ from ASYM, and severe means just the severe group differs from the other two groups. The X-axis is percent of the gait cycle from heel contact to second heel contact on the tested leg. # Cheryl Hubley-Kozey 2017

Effects of Knee Osteoarthritis and Joint Replacement Surgery on Gait

1533

While all participants had a minimum KL grade of 2, the distribution of severity was also not presented. As with the biomechanics trends above certain differences in EMG are associated with compartment involved and disease presence whereas others systematically change with severity levels. The differences in waveforms for a group with mild to moderate medial compartment knee OA who had a relatively high level of function, mild to moderate radiographic structural changes (KL 2 and 3 grades) and were not TKA surgical candidates are subtle compared to the asymptomatic group, whereas the differences in the severe medial compartment knee OA group who were TKA candidates (severe symptoms, structural damage, and poor function) was profound (Fig. 3) (HubleyKozey et al. 2006, 2009). The moderate OA groups walked slower (1.3 m/s) than the asymptomatic group (1.4 m/s), but both walked at relatively fast speeds confirming the high function level for the OA group (Hubley-Kozey et al. 2006). Using PCA, there was a slight phase shift with the medial gastrocnemius coming on later and to lower amplitude in the OA group compared to asymptomatic controls (Fig. 3b). The vastus laterals had higher overall activity compared to the asymptomatic group but the vastus medialis did not (Fig. 3c and d). Most notable was the higher overall lateral hamstring activity and higher mid-stance activity compared to asymptomatic controls and to the medial hamstring of the OA group, whereas no medial lateral differences were found in asymptomatic controls (Fig.3e and f). These selective medial-lateral differences suggest a response to the higher medial joint load, perhaps pain or a response to structural changes associated with medial joint space narrowing. Others reported medial/lateral site-specific differences of those with predominant medial compartment knee OA, including higher vastus medialis:medial gastrocnemius CCI (17% vs. 11%) compared to asymptomatic for those who exhibited medial joint laxity and were scheduled for a high tibial osteotomy surgery (Lewek et al. 2004). Lewek’s group was more severe than the moderate OA group in the previous study as participants were surgical candidates and the medial joint laxity explains the higher medial site coactivation in their OA group (Lewek et al. 2004). Indeed studies show that muscle activation patterns differ across structural and clinical severity (Hubley-Kozey et al. 2009; Rutherford et al. 2013; Rutherford et al. 2011; Zeni et al. 2010). Fig. 3 illustrates that some of the changes among clinical severity groups were found to be progressive and some changes were specific to the severe OA group only based on PCA features and CCI (Hubley-Kozey et al. 2009). Progressive increases in coactivation were found for the vastus lateralis and lateral hamstring (Fig. 3c and e) whereas only the severe clinical group had higher values for vastus medialis and medial hamstring (Fig. 3d and f). While self-selected walking speed was different among the three OA groups, a later study showed that muscle activation differences still existed when the three diagnostic groups were matched based on similar waking speeds, and changes were not just higher overall amplitude that can be explained by strength deficits along the severity spectrum (Rutherford et al. 2011). The medial gastrocnemius muscle had a phase shift to the right where it

1534

C.L. Hubley-Kozey and J. Astephen Wilson

came on later with a smaller difference between early and mid-stance amplitude in both gastrocnemius sites in the clinical severe group compared to both moderate OA and controls (Fig. 3a and b). Structural severity is a component of clinical severity and similar progressive trends along the structural severity spectrum were found for quadriceps and hamstrings mean activation amplitudes (Zeni et al. 2010) and for PCA features across the KL grade severity spectrum for groups walking at similar speeds (Rutherford et al. 2013). In the latter study, what was most interesting is that the progressive trends were not present for the lateral gastrocnemius and medial hamstrings muscles indicating that the lateral hamstring systematic increase found was specific to changes in structural severity which includes medial compartment narrowing and not pain (Rutherford et al. 2013). All of the differences in the above studies cannot be related directly to strength differences among groups only, as there were no strength differences among groups in some of the studies and differences were not always systematic throughout the gait cycle or between medial and lateral muscle pairs. Differences in muscle activation features identified between clinical severity and structural severity are not surprising as altered biomechanical features were reported with increased pain (Henriksen et al. 2006, 2007) and differences were found between two groups with similar structural severity where the symptomatic group had higher KAM features except for the initial peak KAM and a lower early stance KFM compared to those who were asymptomatic (Astephen Wilson et al. 2016). The symptomatic group had higher EMG activity for three quadriceps sites and the lateral hamstring indicative of higher coactivity. Furthermore different biomechanical and muscle activation features were significantly correlated with radiographic severity (KAM magnitude and body mass index) versus symptom (pain) severity (gait speed, lateral gastrocnemius, and medial hamstring activation) (Astephen Wilson et al. 2011). These pain severity findings are consistent with the changes in the lateral gastrocnemius and medial hamstring (Fig. 3a and f) being influenced more by pain for the clinical severe group (i.e., a more general and prolonged coactivation of all the knee joint muscles) than in the moderate OA group. The need for amplitude normalization is particularly important in OA research as the amplitude of the EMG signal is affected by the volume conducting properties between the muscle and the electrode at the skin surface. This can be problematic in the knee OA populations given the high prevalence of adiposity, which makes between groups and muscle comparisons difficult without a standard reference. While difficult and time-consuming normalization to maximal effort contractions demonstrates very good to excellent reliability for both CCI and PCA features (Hubley-Kozey et al. 2013). Furthermore several studies have reported CCI values for knee OA gait and when amplitudes were normalized to MVIC and similar equations used, the findings are consistent across studies (Childs et al. 2004; Lewek et al. 2004; Rudolph et al. 2007; Hubley-Kozey et al. 2009). Where differences exist among these studies they can be explained by the use of different time frames for calculating the index (Zeni et al. 2010) or different OA severity of the samples (Lewek et al. 2004). The CCI provides a relative measure of activity between pairs, but the limitation is that the CCI does not have a unique solution as

Effects of Knee Osteoarthritis and Joint Replacement Surgery on Gait

1535

different patterns can give similar values (Hubley-Kozey et al. 2009), thus presenting the waveforms would help with interpretation. The literature supports that we can reliably measure muscular alterations in knee OA, findings can be compared among studies with similar protocols, and these measures are responsive to changes in joint pathology and pain severity. Collectively these EMG studies provide additional information that can assist in interpreting the joint level biomechanics alterations during walking associated with presence and severity of knee OA. The higher EMG activity of both hamstrings and quadriceps muscles in severe knee OA does not support that the lower peak KFM is related to a quadriceps avoidance gait. Also the prolonged hamstring and quadriceps activity during mid-stance is consistent with the higher KAM mid-stance value and the lower KFM-KEM range indicative of an altered unloading pattern of the knee joint and a “stiff gait” pattern. Furthermore they highlight the limitation of reporting net external joint moments only to understand changes in muscle function and in joint loading during walking in the OA population.

Gait Metrics Associated with Predicting Progression Outcomes Evidence from longitudinal studies supports that walking mechanics at baseline can predict OA progression outcomes, and this work has led to the development of biomechanical targets with a number of intervention studies aimed at altering external KAM features to manage medial compartment knee OA. Most studies have focused on structural progression assessed through imaging techniques such as radiographs or magnetic resonance imaging (MRI). Specifically, at baseline a higher peak KAM was found for those who have progressive OA changes in joint structures at follow-up (Chehab et al. 2014), and this was found to be predictive of structural progression (Chang et al. 2015; Miyazaki et al. 2002). In a shorter followup period (Bennell et al. 2011), the peak KAM was not predictive of structural changes based on MRI measures; however, a higher KAM impulse was predictive of structural progression consistent with a recent longer term study (Chang et al. 2015). These short-term findings suggest that the KAM impulse may be more sensitive to early structural changes than the peak KAM. Higher external KFM peaks were also found at baseline for those who progressed structurally, but the changes were in the tibial cartilage (Chehab et al. 2014), and not the femoral cartilage, the latter finding consistent with a recent study that also used MRI features (Chang et al. 2015). Progression of pain is a key factor in clinical progression, with few longitudinal studies on progression of OA pain. Amin et al. (2004) found higher baseline KAM peaks for different tasks in older adults who developed pain at follow-up, with 13% higher KAM peaks during walking. A recent paper showed that at baseline those with moderate knee OA who went on to TKA surgery within a 7-year follow-up had a smaller knee KFM-KEM range, higher overall KAM magnitudes, and smaller differential between early and mid-stance KAM (all captured with PCA), and two discrete variables including higher peak KAM and KAM impulse than those who did not have surgery at follow-up (Hatfield et al. 2015a, b). A prediction model was

1536

C.L. Hubley-Kozey and J. Astephen Wilson

developed for risk of TKA, and for the first time a mid-stance unloading feature related to stiff knee gait (KFM-KEM range) was included along with the KAM magnitude feature consistent with the clinical OA severity literature previously presented. In summary, the cross-sectional gait studies consistently show that joint level biomechanics and muscle activation features are altered in knee OA gait with differences between men and women for some variables, and certain variables are progressive, some are OA only, whereas others are specific to increased OA severity levels. These alterations are significant for those with severe clinical knee OA who are candidates for TKA surgery demonstrating a measurable decline in joint function. Most studies report discrete measures from biomechanical and muscle activation waveforms capturing either a peak value at an instance in time or a summary variable such as an impulse, mean, or co-contraction index with PCA capturing dynamic features that differentiate among severity groups not captured by discrete values alone. The differences in variables associated with structural and symptom/ clinical severity are expected as pain alone can alter gait mechanics. The evidence supports that specific biomechanical and muscle activation measures have value as screening and monitoring tools for OA severity as well as assessing change in joint function related to clinical status and interventions. Clinical decision making is presently based on self-report of symptoms and static imaging related to joint damage. The predictive potential of specific gait variables from the longitudinal studies is encouraging as it has led to biomechanical interventions; however, these interventions have focused primarily on one variable, the KAM magnitude. Understanding the magnitude of these alterations in those with severe knee OA is important to perhaps develop a threshold for functional decline on which to base decisions on who will best benefit from TKA surgery.

Gait Metrics and Joint Replacement Surgery A primary goal of joint replacement surgery is to improve overall function by reducing pain and replacing the damaged joint tissues with an artificial joint. There are numerous small variations of implants, but in general the standard of care design is similar among implants, and differences are primarily based on material choices for bearing surfaces (most common are metal femoral component on a polyethylene tibial tray), ligament preservation (cruciate retaining designs versus posterior stabilized), as well as some small design criteria aimed at creating more natural kinematics or minimizing polyethylene wear particles (medial pivot, mobile bearing designs). TKA surgery is considered a highly successful surgery and its success is primarily related to its ability to reduce pain. There is, however, a reported 20% dissatisfaction rate with TKA surgery, where satisfaction for pain and function are similar (pain 72–86%, function 70–84%) (Bourne et al. 2010). However, many (33%) patients still report functional limitations at 1 year post-TKA (Franklin et al. 2008). From the previous section, knee joint motion, moments, and muscle activation patterns along with walking speed indicate significant knee joint functional

Effects of Knee Osteoarthritis and Joint Replacement Surgery on Gait

1537

deficits in those with clinically severe knee OA compared to asymptomatic controls and those with moderate knee OA. We know that the artificial joint differs from the “real/native” joint and that it is unlikely that joint function could return to that of a typical healthy knee. Gait studies have examined the effect of TKA surgery on knee joint mechanics and muscle activation patterns providing a quantitative assessment of joint function post-TKA surgery, and there is evidence that gait patterns can predict outcomes related to implant stability and pain following TKA surgery. Early studies compared post-TKA gait to values for asymptomatic controls. In general, at 1 year post-TKA, self-selected walking speed remained slower in the joint replacement group compared to asymptomatic (Benedetti et al. 2003; Yoshida et al. 2008), whereas (Wilson et al. 1996; Simon et al. 1983) no difference was found compared to asymptomatic controls with a longer follow-up time (greater than 3 years). However, walking speeds for the asymptomatic controls in the latter two studies were slower than typical for an asymptomatic cohort (100% increase) and fatal to 1.5 million (
25% increase) (Haagsma et al. 2016). The Global Burden of Disease (GBD) has developed a metric called the DALY (disability-adjusted life years), which quantifies the years of life lost and disability years due to injury, disease, violence, etc. (where an increase in the DALY corresponds to an increase in death and injury rates). From 1990 to 2013, DALY values have decreased in rich nations (the most substantial being – 67% in Asia Pacific and – 61% in Western Europe), while overall changes in DALY values were not significant in developing nations indicating essentially no change. Four-wheeled injuries significantly increased in South Asia (+22%) and sub-Saharan Africa (+20%) (Haagsma et al. 2016). These epidemiological data indicate enhanced safety design across major automobile manufacturers as well as improved road conditions and traffic management. However, governments in developing nations have virtually ignored road and traffic conditions. Injuries incurred from traffic crashes are a significant cause of morbidity and mortality. The continued burden of traffic crash injuries involves rigorous experimental and forensic biomechanics to better understand injury mechanisms. Traffic crash biomechanics have the end goal of adapting road and vehicle design to everchanging traffic conditions.

Current Approaches and Data Analysis Injury Severity Quantification A variety of methods have been developed for quantifying injury severity over the years, but it remains a difficulty to standardize an objective method for injury severity. Currently, injury severity metrics are based on consensus as well as historical data. There remains a need for precision in injury severity quantification. In research, defining injury risk relies heavily on the Abbreviated Injury Scale (AIS) for both quantifying historical injury severities as well as model development (e.g., probability of death score; PODS). Injury severity scores are an important measure for biomechanical research used for retrospective analyses that enable engineering advancements with the aim to minimize injury risk from traffic-related incidents. AIS The Abbreviated Injury Scale (AIS) of the Association for the Advancement of Automotive Medicine (AAAM) is a numerical rating system for quantifying injury severity in motor vehicle crashes (MVCs). AIS is a proprietary classification system and its use requires trained personnel to properly code a victim. As a result, not all traffic-related incident victims are coded according to the AIS. Currently, AIS is in its sixth revision, and it is the most widely used metric for coding injury severity.

2366

B.D. Goodwin et al.

The AIS code consists of a 6-digit value followed by a single-digit value (i.e., 123456.1). The first “pre-dot” number identifies the body region, type of anatomical structure, its specific structure, and level (Gabler et al. 2015). The second component of the score has a range between 0 and 6 corresponding to no injury up to the most severe injury, respectively. High AIS ratings indicate injuries with high mortality rates, as it is essentially a threat-to-life scale. The AIS is implemented for all body regions, each with different methods for quantifying injury severity. For example, AIS level for a vertebral body compression fracture depends on the height of the vertebra after injury, and the AIS level for a skeletal chest injury is a function of the number of sustained rib fractures (as well as the presence of a flail chest for AIS  3). AIS levels also apply to soft tissue damage, which depend heavily on the organ that was damaged.

ISS, NISS, and PODS The Injury Severity Scale (ISS) is a measure of the probability of survival. The ISS value is simply calculated by finding the maximum AIS values from each body region (eight body regions total: head, neck, face, chest, abdomen, spine, extremity, and external) then summing the squares of the top three maximum AIS values. This calculation gives the ISS metric a range of 1 to 75. The goal of the ISS level is to quantify the overall bodily injury severity. The New Injury Severity Score (NISS) incorporates two modifications of the ISS: (1) the sum of the squares of the top three ISS scores no matter the body region and (2) if any AIS score is 6, then NISS is automatically set to 75. ISS is more widely used, but NISS has been shown to better predict hospitalization durations and organ failure (Balogh et al. 2000, 2003; Gabler et al. 2015). The Probability of Death Score (PODS) is an estimate of probability of survival. PODSa is the same estimate while accounting for victim age. The two probabilities are quantified as follows. PODS ¼ PODSa ¼

ex 1 þ ex

(1)

for PODS x ¼ 2:2ðAIS1 Þ þ 0:9ðAIS2 Þ  11:25 þ C

(2)

x ¼ 2:7ðAIS1 Þ þ 1:0ðAIS2 Þ þ 0:06ðAGEÞ  15:4 þ C

(3)

for PODSa

Where C = 0.764 for cars, and AIS1 and AIS2 are the first and second highest AIS values. The equations above have been derived from an empirical data fit (Somers 1983a, b). The advantage of PODS over ISS is that it describes real historical data and its value has an explainable meaning by virtue of its basis in probability.

Injury Mechanisms in Traffic Accidents

2367

Human Injury Probability Curve A human injury probability curve (HIPC) is an estimate of the injury risk with respect to a defined variable called the injury criterion. Real-world forensic data provides knowledge of possible injuries. Safety engineering aims to minimize risk of these known injuries. Quantifying injury risk is left to biomechanics research for identifying injury mechanisms, establishing injury criteria, designing PMHS experiments, and quantifying human tolerance through statistical analyses. Injury criteria are the independent variables for which injury risk is estimated, and these depend on both the method for processing experimental PMHS data (Yoganandan et al. 2014b) and the statistical methodology for estimating risk (Parr et al. 2013; Yoganandan et al. 2014a, 2015a). Local criteria such as stress and strain or global criteria such as force and moment are often selected based on principles of failure mechanics (Fig. 1). In this sense, an injury risk of 50% means that one-half of the considered population has a tolerance limit lower than the corresponding value of the independent variable (or injury criterion). Injury risk estimates or HIPCs do not translate well across differing loading conditions or body regions. As a result, many injury risk curves are necessary to assess whole body risk under well-defined conditions. For example, risk of femur fracture under 3-point bending does not translate to fracture risk during axial loading of the femur. Human tolerance or risk of injury (c.f. Fig. 1) is computed using statistical techniques such as survival analysis. A survival analysis approach lends itself to injury risk estimation because of the nature of datasets to which the HIPC is fit (McKay and Bir 2009; Parr et al. 2013; Yoganandan et al. 2013b). Datasets that contain human injury and noninjury experiments contain various types of censored (uncertain) data, which represent injurious and noninjurious experiments. A survival analysis has been commonly carried out for clinical studies where the time of

Fig. 1 Example of a human injury probability curve where impact force is the independent variable for estimating probability/risk of injury

2368

B.D. Goodwin et al.

survival needs to be estimated following diagnosis of a fatal disease. The time of survival is considered censored when the date of death is unknown, and the survival analysis incorporates statistical techniques to account for datasets with censored data (Pintar et al. 1998; Funk et al. 2002; Yoganandan et al. 2014a; Petitjean et al. 2015; Yoganandan et al. 2015a). For HIPC generation, an injury criteria variable is used for estimating risk instead of time of survival.

Explaining Mechanisms Through Forensic Data Retrospective studies rely heavily on traffic crash, diagnoses, and patient outcomes. Databases of field data are managed by government departments and fall into one of three categories (Gabler et al. 2015): (1) fatal crash data, (2) details of crash investigations, and (3) higher-level data of all fatal and nonfatal crashes. This chapter will give a brief overview of popular databases that are managed in the United States (Gabler et al. 2015), and the reader is encouraged to seek other sources for more in-depth descriptions of internationally available tools. CIREN Although international databases have been developed, the Crash Injury Research Engineering Network (CIREN; managed by National Highway Traffic Safety Administration) and National Automotive Sampling System (NASS) contain vast amounts of forensic data. Multiple CIREN centers in the United States collected crash and medical data from as early as 1996. CIREN has enabled detailed retrospective analyses from medical and biomechanical perspectives (Augenstein et al. 2000; Augenstein and Diggs 2003; Kirk and Morris 2003). For example, Yoganandan et al. (2009) used CIREN data to survey a database of over 1,800 traffic crash victims and found that diffuse axonal injury occurred in crashes where victims experienced contact-induced blunt force head trauma (Yoganandan et al. 2009). NASS The National Automotive Sampling System (NASS) is a database containing detailed characteristics of more than 5,000 crashes each year. This database contains data from crashes with a range of severities from no injury to minor to fatal, which involved cars, light trucks, vans, and/or sport utility vehicles. The NASS database is populated by NHTSA crash investigators who document data and evidence from crash sites through forensics, photography, interviews, and hospital injury severity codes. FARS The Fatality Analysis Reporting System (FARS) is a comprehensive database of all traffic related fatalities in the United States. This database includes traffic crashes of all vehicle types including bicyclist and pedestrian fatalities. FARS contains less specific data than the NASS, which includes over 400 data elements per crash compared to about 175 elements in FARS. For a crash to be included in the FARS database, at least one fatality must occur within 30 days of the incident. Injuries are

Injury Mechanisms in Traffic Accidents

2369

coded using the coarse KABCO scale (killed, incapacitated, moderate injury, complaint of pain, or property damage only). FARS has been maintained by NHTSA since 1975 and contains data from 30,000 to 40,000 fatal accidents.

State of the Art Elucidating injury mechanisms from MVCs requires robust experimentation and data analysis in applied impact mechanics. Generally, exploratory research within the field of biomechanics is rare since the field has evolved to focus heavily on applied biomechanics for injury prevention/minimization. As a result, analytical techniques for estimating injury risk have improved significantly (Pintar et al. 1998; Petitjean and Trosseille 2011; Takhounts et al. 2013). New PMHS experimental methodologies are consistently published to estimate biomechanical responses under simulated impact conditions. Multiple PMHS experiments are performed with the aim to generate a mean response or response corridor with high biofidelity. Injury and noninjury responses are then used for deriving injury criteria, which enables injury risk estimation under defined loads. Since PMHS specimens are unidentical, data censoring approaches are necessary to estimate risk for the average human. The survival analysis can handle censored data types. It is a fairly new approach to estimate injury risk along with a confidence level of that risk, which has been called the human injury probability curve (HIPC) (cf. Sec. Human Injury Probability Curve). The field of applied biomechanics has advanced significantly where experimental setups can closely simulate MVCs including impacts from rollover (Lessley et al. 2014), oblique (Yoganandan et al. 2015b), and cyclist/pedestrian (Matsui and Oikawa 2015) collisions. High-speed videography when integrated with digital image correlation (DIC) (Anuta 1970; Chu et al. 1985) provides the means to capture widespread strain (and strain rate) in large-scale applications (e.g., vehicle deformation) or small-scale experiments (e.g., bone or ligament strain) under highly dynamic or quasistatic loading conditions (McCormick and Lord 2010; Begonia et al. 2015).

Head and Neck Head and Brain Brain Injury It is generally accepted that traumatic brain injury (TBI) results from sudden movement of the head (usually caused by blunt impact), resulting in both linear and angular acceleration. Studies show that these two modes of acceleration contain fundamental differences in terms of their injury mechanisms. Diffuse axonal injury (DAI) is the most common TBI where sharp, dynamic head movements directly cause neuron damage leading to neurologic sequelae. A study from 2009 found that a survey of 1,823 cases contained brain traumas to

2370

B.D. Goodwin et al.

67 (41 adults) vehicle occupants that were coded as DAI (3.6%; no crash resulted in more than one DAI) (Yoganandan et al. 2009). Within the adult sample size, 33 were lateral crashes (80%), and all DAI occurrences involved head contact loads. Hardy et al. (2001) were able to show that pure angular acceleration (20–25 ms duration) exposes white matter to tensile strains resulting in damaged axon fibers (Hardy et al. 2001), while Anderson et al. (2003) observed that DAI severity was correlated with peak linear and angular accelerations in the sheep model (Anderson et al. 2003). Neural damage occurred ipsilateral and contralateral to the impact, which suggests that brain matter undergoes shear and/or tensile stresses due to its abrupt displacement relative to the skull. Axonal stretching under angular acceleration still requires further validation as an injury mechanism, but it is the most reasonable hypothesis for DAI. Linear acceleration subjects the brain to a pressure wave beginning with compression where the cells nearest to the impact experience the highest compressive forces. An effective fluid percussion device has been used in dogs to show that a pressure wave causes concussion (Gurdjian et al. 1955). These waves varied in duration and pressure from 1 to 46 ms and 34.5 to 345 kPa, respectively. VandeVord et al. (2012) were able to show that externally applied hyperbaric blasts cause significant neurotrauma in rats, which lead to cognitive deficits from diffuse glial cell damage (Vandevord et al. 2012). Short duration (5–10 ms) angular accelerations from blunt impact can also rupture veins or arteries causing acute subdural hematoma (ASDH) resulting in cell damage (Davceva et al. 2012).

Acute Subdural Hematoma ASDH is one of the most deadly injuries having a mortality rate of greater than 50% across studies. The ischemia resulting from vascular damage causes neuronal damage, and head trauma patients die or become seriously disabled (Wilberger et al. 1991). ASDH is found in victims of blunt force trauma and is the most grave of injuries due to its high incidence (30%), high mortality (60%), and injury severity (common Glasgow Coma scores range 3–5) (King 2015). Data seems to point to a 50% risk of ASDH subsequent to experiencing a severe TBI (>9 Glasgow Coma score) based on a study from 17 Austrian centers, which documented the injuries and outcomes of 360 patients (Leitgeb et al. 2012). Among those patients with an ASDH, 47% died in the hospital, 19% survived with “unfavorable” outcome, and 32% survived with a “favorable” outcome. Bridging vein rupture has been viewed as the injury mechanism for ASDH. However, as King points out, this mechanistic explanation seems unsound since fluid mechanics principles are violated by virtue of the adhesive resistance at the interface of the dura and arachnoid (King 2015). Blood would have to enlarge the space between dura and arachnoid before it is visible through medical imaging. Blood would flow into the superior sagittal sinus if it were to flow in the path of least resistance (and it will), which implies that there is insufficient pressure for an ASDH to form at the dura-arachnoid interface. While the evidence is strong for ASDH formation following bridging vein rupture (Gennarelli and Thibault 1982; Depreitere et al. 2006), this mechanism is difficult to defend in terms of fluid mechanics. Four

Injury Mechanisms in Traffic Accidents

2371

hypotheses have been proposed by King (2015): (1) almost all ASDHs from impacts result from cortical arterial rupture at the dura-arachnoid interface where arachnoid border cells need to separate to tear cortical arteries, (2) radial separation of the border cell layers occur when under dynamic skull deformation from direct impacts, (3) angular acceleration subjects the border cell layer to shearing deformation causing cortical artery rupture, and (4) the bridging vein rupture is only correlated to the events surrounding the formation of ASDH but does not cause it (King 2015). The first hypothesis would provide a sensible explanation for ASDH formation, the second provides a reason as to why ASDH forms at sites remote from the impact since skull deformation does not occur only locally, the third is more unlikely due to the “slow” motion of the brain surface when exposed to angular accelerations under 10,000 rad/s2 at the head center of gravity, and the fourth is the logical consequence if the bridging vein rupture hypothesis is false.

Cervical Injury Epidemiology Traumatic spinal cord injuries (TSCIs) from MVCs represent the majority of spinal cord trauma cases per capita on an international level (Haagsma et al. 2016). With the exception of poor regions and regions that contain with highly dense populations (Tropical Latin America, South Asia, Oceania, and Eastern Europe; regions per World Health Organization), traffic-related incidents involving four-wheeled vehicles, motorcycles, bicycles, or pedestrians account for approximately 50% or more of documented TSCI cases on a per region basis (Sekhon and Fehlings 2001; Middleton et al. 2012; Lee et al. 2014). TSCI cases from traffic-related incidents in developed regions remains either stable or reduced, while underdeveloped regions fail to constrain the rise in TSCI cases due to poor traffic conditions and diminished vehicle safety standards (Haagsma et al. 2016). Neck Pain and Whiplash Neck pain is the most commonly reported symptom following a rear end vehicular collision. Evidence suggests that 50% of victims of whiplash report pain 1 year after the injury, where greater initial pain, number of symptoms, and degree of debilitation predicted recovery rates (Carroll et al. 2009). Nociceptors are nerve endings that act as pain receptors, and they have a high stimulus threshold to action potential initiation compared to other nerve endings. The signals from nociceptors are sent to the spinal cord and brain by which pain is perceived. Intervertebral discs often become enflamed from whiplash injury, which has the effect of lowering the stimulus threshold of surrounding nociceptors causing an increased sensitivity to pain. The consequent pain sensations will then be amplified under reduced loading conditions giving the perception of chronic pain. Additionally, nociceptors become more concentrated and numerous around the degenerated region in the soft tissue. Within the intervertebral space, nociceptors

2372

B.D. Goodwin et al.

reside in facet capsules, spinal ligaments, tendons, and muscles (Deng et al. 2000). Barnsley et al. (1995) and Bogduk and Marsland (1988) provide clinical evidence of cervical facet pain in patients with neck pain following whiplash (Bogduk and Marsland 1988; Barnsley et al. 1995). A number of hypotheses have been proposed for the precise mechanism of whiplash that leads to chronic pain after initial injury. Anatomical complexities of the neck make it difficult to converge on a hypothesis without conflicting premises (Siegmund et al. 2001, 2008). Yang and Begeman (1996) proposed a shear force hypothesis where forces are transferred up the cervical spine to the occipital condyles (Yang and Begeman 1996). The torso is thrust forward and its momentum pulls the head forward subjecting the intervertebral space to shear especially at lower levels where the facet angle is farther from the vertical. In this case, the pain is attributed to straining facet capsules. Deng et al. (2000) as well as Lu et al. (2005) confirm the hypothesis that painful signals are received from nociceptors surrounding the facet capsule and ligament resulting from shear forces in the neck (Deng et al. 2000; Lu et al. 2005). Their PMHS and animal experiments show that the neck is subjected to compression, tension, shear forces, flexion, and extension throughout the duration of the impact. The shear force hypothesis has been further corroborated through computational models (Stemper et al. 2004; Panzer et al. 2011; Fice and Cronin 2012). Literature seems to indicate that the mechanism of injury discussed here is quite definitive. Close attention can be given to the engineering of methods to best prevent whiplash injury unless compelling evidence is put forth that disconfirms the shear force hypothesis.

Cervical Spine Injury Since the most severe injuries of the cervical spine involve vertebral body fractures or dislocations, this section is limited to an overview of vertebral body anatomy of the cervical spine. Additional information regarding anatomical detail is left to other sources (Nightingale et al. 2015). The cervical spine is comprised of seven vertebrae where the occipital condyle (OC) sits on top of C1 (OC-C1 joint) and C7 rests on top of T1 (C7-T1 joint) (Fig. 2). The cervical spine has a slight curve, called lordosis, during nominal posture. The upper cervical spine (UCS; OC to C2) is anatomically distinct from the rest of the cervical spine. Technically, C1 does not have a vertebral body but is a ring with distinguishably larger facets (Fig. 2). The UCS is capable of a greater range of motion than the rest of the spine. Damage to the UCS is almost unsurvivable due to the vulnerability of the spinal cord in the UCS. As a result, data on UCS fractures is sparse since victims are not rushed to the hospital (Yoganandan et al. 1989). The C1 vertebra (also known as the “atlas”) is susceptible to a multipart fracture, and fatalities are common with a four-part fracture (Jefferson 1919). Vertebral bodies in the lower cervical spine (LCS; C3 to T1) are uniformly shaped and have lesser range of motion than those of the UCS. Burst fractures, dislocations, and fracture dislocations are common to the lower cervical spine vertebrae, which include fractures of the cortical bone, endplate fractures, and loss of disk height. Injuries to the spinal cord are less common with fractures in the LCS compared to the UCS (Fig. 3).

Injury Mechanisms in Traffic Accidents

2373

Fig. 2 Cervical vertebral bodies

POSTERIOR

ANTERIOR

Facets Capsular ligament

Intertransverse ligament

Anterior longitudinal ligament Anterior one-half annulus fibrosus Interspinous and supraspinous ligaments Ligamentum flavum

Fig. 3 Vertebral body with ligaments

Posterior one-half annulus fibrosus

Posterior longitudinal ligament

2374

B.D. Goodwin et al.

Much consideration has been devoted to clinical classifications of neck injuries (Allen et al. 1982; Torg 1985; Yoganandan et al. 1989; Myers and Winkelstein 1995), but the complexity of the cervical spine leads physicians and researchers to necessarily base classifications on inference and anecdotal evidence (Nightingale et al. 2015). Relatively speaking, the neck is strong during compression, but headfirst impacts from small drop heights (0.5–2 m) have produced a variety of neck injuries in PMHS experiments (Nusholtz et al. 1981; Yoganandan et al. 1986). Neck compression injuries arise from conditions where the victim lands headfirst such as vehicle occupant ejection, motorcyclist ejection, or vehicle rollover. The fracture level and type are a function of the buckling mode during injury, which is subject to the initial orientation of the spine (Nusholtz et al. 1981, 1983), or the degree of neck lordosis (Culver et al. 1978). Compression injury mechanisms are further complicated by the decoupling response of the head and neck (Yoganandan et al. 1991). Yoganandan et al. (1991) were among the first to use whole cervical spines under compressive loads and suggested that the neck behavior during buckling influences injury. Vertical impact tests on well-constrained head and neck specimens without lordosis (straightened necks where the occipital condyle was approximately concentric with T1) were performed by Pintar et al. (1995), which produced a wide range of injuries, and the complex neck behavior revealed no single metric as an injury predictor (Pintar et al. 1995). Additionally, Nightingale et al. (1996a, b, 1997) performed a variety of experiments with differing postures and impact angles and quantified compressive failure loads (Nightingale et al. 1996a, b, 1997). A wide range of injuries were produced including midlevel burst fractures, odontoid fractures, and Hangman’s fractures. Midlevel fractures are best explained in terms of the complex kinematics of first and second order buckling of the neck. Neck tension-extension injuries in the cervical spine are much less common than compression injuries but are more likely to have higher injury severity and have a fatal outcome. Etiologically, extension injuries have been found in victims who were not wearing a seat belt or were too close to the airbag when it deployed. Tensionextension injuries are very difficult to reproduce since neck musculature has a substantial influence on the presence and/or nature of the injury itself (Chancey et al. 2003). The maximum tensile tolerance of the neck essentially is increased by 19% when the occupant anticipates the injurious event by activating neck muscles. Regardless, the ligamentous cervical spine will fail at lower loads, but the neck tensile tolerance is increased during muscle activation. Some PMHS experiments fail to reproduce injuries observed in real-life traffic accidents due to the relaxed musculature of the cadaver. It remains to be seen how each parameter that defines the initial conditions prior to injury affect the characteristics of the resulting injury. However, the body of work on this topic suggests that injury traits are influenced by the orientation of the neck, head, and subject-specific anatomical characteristics (e.g., degree of lordosis) immediately before injury.

Injury Mechanisms in Traffic Accidents

2375

Thorax Introduction Chest injury ranks just behind head injury in overall number of fatalities in traffic accidents (Services UDoHaH 2007). During a vehicle collision, the thorax is exposed to vehicle interior components including restraint systems, each of which poses varying risks. Thorax injuries are common in frontal and side collisions as well as their oblique counterparts. A retrospective analysis by Nirula and Pintar (2008) shows that the incidence of severe chest injury ( AIS 3) was 5.5% and 33% in NASS and CIREN, respectively (Nirula and Pintar 2008). The steering wheel, door panel, armrest, and seat were all identified as contact points with substantial risk of severe injury to the thorax. The reader is encouraged to look to other sources for an overview of thorax anatomy in the context of injury mechanisms (Cavanaugh and Yoganandan 2015). The following survey of injury mechanisms will refer to various regions of the thorax with very little review of anatomy basics.

Traumatic Rupture of the Aorta Background Though traumatic rupture of the aorta (TRA) does not happen frequently, it has the highest fatality rate in traffic crashes of all injuries to the thorax. Crash data from 1998 to 2006 shows that TRAs occurred in approximately 1% of all vehicle occupants in traffic-related incidents, but was the cause of 21% (8%) of all fatalities (Shkrum et al. 1999). In the occurrence of a TRA, there is only a brief time-window where the injury can be treated before it becomes fatal. Bertrand et al. (2008) found that TRAs were twice as common in occupants involved in side impact (2.4%) compared to frontal impact (1.1%) in the United Kingdom (Bertrand et al. 2008). The difference in TRA incidence from side impact happened after the advent of air bag and seat belt restraints to prevent frontal crash injuries. Anatomy The aorta extends from the base of the left ventricle of the heart at the aortic root. The ascending arc of the aorta is relatively flexible while the descending aorta is secured to the thoracic spine via the pleural reflection. The peri-isthmic region is between the anastomosis of the left subclavian artery and the descending aorta. The peri-isthmus is the most common place for TRAs to originate (Fig. 4). Mechanism Tears in the aorta have occurred in the peri-isthmic region for an estimated 94% of all TRAs (Katyal et al. 1997). Past PMHS studies struggled to reproduce TRAs until Hardy et al. (2006) was able to demonstrate that TRAs can be induced in PMHS

2376

B.D. Goodwin et al.

Fig. 4 Heart anatomy of interest

specimens through longitudinal quasistatic loading (Hardy et al. 2006). This study was also the first to orient the specimen to place the diaphragm, heart, and aorta in an anatomically consistent way with a healthy human. Interestingly, the aorta was subjected to tension without having to induce whole body acceleration. Circumferential tears were found to be almost ubiquitous among the observed TRAs. Both the intima (innermost layer) and media (middle, thicker layer) of the aorta were found to be sensitive to tearing (Cammack et al. 1959). The accepted mechanism of injury for TRA is tension in the aorta, which apparently develops immediately after (not during) direct chest impact. Hardy et al. (2008) used radio-opaque markers to track the response of the peri-isthmus region while acquiring high-frequency x-ray, and they point out specific catalysts of TRA: (1) deformation of the thorax, (2) elongation (longitudinal stretch) of the aorta, and (3) the tethering of the aorta to the spine, which promotes stretching (Hardy et al. 2008). Here, PHMS specimens were impacted in different modes: shoveling, side impact, submarining, and combined. All impact modes caused TRAs by transverse tears and two oblique tears.

Thoracic Spine Injuries Injuries to the thoracic spine have predominantly originated from traffic-related incidents (Robertson et al. 2002b; Leucht et al. 2009), which could be a surprising statistic since it is difficult to imagine how the thoracic spine could be exposed to such forces in a vehicle passenger in a seated position. In the United States, 8% of spine fractures from MVCs are AIS  3, and 9% of all thoracic spine injuries are

Injury Mechanisms in Traffic Accidents

2377

Fig. 5 Spine load in PMHS during horizontal deceleration from a frontal MVC

AIS  3 (Wang et al. 2009). In a retrospective data analysis from the United Kingdom, almost 23,000 patients were surveyed for motorcycle and MVC victims (Robertson et al. 2002a). Spinal injuries were present in 126 (11.2%) motorcyclists and 383 (14.1%) car occupants. The thoracic region was the most common spine injury in motorcyclists (54.8%; n = 126), while thoracic spine injuries in car occupants were present in 26.6% (n = 383) of spinal injury cases. Similarly, Pintar et al. (2012) analyzed CIREN and NASS data from MVCs and found the dorsal spine to be particularly vulnerable despite public awareness and seatbelt use in the United States (Pintar et al. 2012) (Fig. 5).

2378

B.D. Goodwin et al.

Compression Fracture Perhaps nonintuitively, injuries related to spine compression are often observed in frontal and rear-end crashes or in the presence of abrupt decelerations/accelerations. Begeman et al. (1973) simulated frontal crash responses in PHMS using a sled (Begeman et al. 1973). Three-point seat belts (with shoulder strap) were found to promote spine compression compared to a single lap restraint. Begemen et al. found that a deceleration of about 15 G’s results in over 600 lbs of shoulder belt tension and a spine load of about 900 lbs. Additionally, it was found that the axial loads are augmented when the body is held erect by the shoulder restraint. Yang and King (1995) hypothesized that the shoulder strap restrains the upper body before the forward momentum of the body can be sufficiently diminished (King and Yang 1995). The combined force from the forward momentum and the upper restraint acts on the thoracic spine forcing it into a straightened or lordotic posture, which produces dangerous loads on more caudal segments. Furthermore, the asymmetric 3-point belt restraint will also load the spine asymmetrically exposing vertebral bodies to concentrated loading conditions. Forensic clinical studies provide a real-world risk of spine compression injury from frontal crashes (Ball et al. 2000). Burst fractures between L1 and T12 levels were found in 80% of patients where 3-point seat belts were fastened while in only 25% of patients wearing a lap belt. Nonetheless, a recent retrospective study (Pintar et al. 2012) points out that despite the frequent occurrence of thoracolumbar fractures in frontal impacts, the precise injury mechanism remains elusive. Flexion Distraction Injury Flexion-distraction injuries (or Chance fracture (Chance 1948)) have been found in vehicle occupants involved in MVCs, and they are not limited to the thoracic spine. For example, Chance fractures have been reproduced through airbag deployment in PMHS experiments (Cheng et al. 1982). The thoracolumbar region seems to be especially exposed to the possibility of Chance fractures during a MVC. The mechanism of injury involves fracture initiation in the posterior aspect of the neural arch and continues anteriorly (Fig. 6). The spine experiences hyperflexion followed by a distraction (Stemper et al. 2015). This injury has been especially common in occupants wearing only a lap belt or an improperly secured 3-point restraint. A Canadian study analyzed medical data of eight children involved in an MVC wearing lap or 3-point restraints (Santschi et al. 2005). Of the five children wearing a lap belt, four experienced flexion-distraction fractures to the lumbar spine and three were permanently paralyzed. Flexion-distraction fractures were accompanied with intra-abdominal injuries, and this coincidence seems to be a pattern (LeGay et al. 1990). The high incidence of spinal injury and intra-abdominal injuries that result from wearing a lap belt restraint was apparently enough motivation for the United States to prohibit the lap belt design in all cars sold in the United States since September of 2007 (NHTSA 2005).

Injury Mechanisms in Traffic Accidents

2379

Fig. 6 Flexion-distraction injury (Chance fracture)

Intra-abdominal Injuries Anatomy The abdomen is conventionally described through nine regions: (R/L) hypochontriac, (R/L) lumbar, (R/L) iliac, epigastric, umbilical, and hypoastric (Fig. 7). This section will focus on anatomical features pertinent to traffic accident injury, but for a detailed description of abdomen anatomy, the reader is encouraged to look to other sources (Hardy et al. 2015). The lower ribs offer protection from blunt trauma to the upper abdomen, and the anterolateral abdominal wall (skin, subcutaneous tissue, muscle, fascia, and parietal peritoneum) provides additional protection to the abdominal viscera. Soft tissues and organs (Fig. 8) are tethered through the mesentery, which is made up of two layers of peritoneum either between organs or tethering viscera to the abdominal wall. A certain amount of movable freedom is granted to abdominal organs based on the length of the tethering mesentery, which protects from robust vibrations and abrupt accelerations. The stomach, small intestine, large intestine, and gallbladder are classified as hollow or membranous organs, and these organs are especially prone to serosal tears and perforations in MVCs (Shinkawa et al. 2004). Abdominal organs that are classified as solid include the liver, spleen, kidneys, and pancreas. Solid organs tend toward increased injury severity in traffic crashes, and the liver and spleen are

2380

B.D. Goodwin et al.

Fig. 7 The abdominal region

the most frequently injured abdominal organs in MVCs (Klinich et al. 2010; Hardy et al. 2015). The liver is highly vascular and fluid-filled, and it is the largest internal organ in the body located in the upper right region of the abdomen. The liver is vulnerable to blunt trauma and bone fragments from a rib fracture can puncture, and depending on the magnitude of the impact, rupture or burst. The spleen is located in the upper left region of the abdomen, and it receives protection from the lower rib cage from blunt impact. Spleen rupture is the most common injury in traffic crashes.

Background The abdomen is a highly complex region of the body where the effects of different loads or impacts on individual organs are difficult to quantify (Yoganandan et al. 2000). A study that analyzed NASS data from 1993 to 1998 found that approximately half of 129,269 abdominal injuries resulted from front-end collisions (Yoganandan et al. 2000). AIS-6 abdominal injuries were present in 94 vehicle occupants and occurred from right (22%) and frontal (78%) impacts. Of the abdominal trauma cases, 31% were spleen injury, 30% were liver injury, and ~33% of all injuries were coded AIS  3. Frontal collision MVCs yielded injuries most common in the liver (39%), spleen (29%), and digestive organs (11%). Other common injuries (>AIS-2) were of the kidney, diaphragm, arteries, and urogenital systems. The abdomen is especially vulnerable to injury during side and oblique impacts, and

Injury Mechanisms in Traffic Accidents

2381

Fig. 8 Soft tissues and organs of the abdomen

the state of automobile safety has progressed to curtail abdominal injuries through side air bags and side curtain air bags (Baur et al. 2000; Yoganandan et al. 2007). Due to anatomical location, the spleen or the liver has been the most seriously injured from far-side impacts (Augenstein et al. 2000).

Mechanisms Ball et al. (2000) point out that there is a high probability of intra-abdominal injury requiring laparotomy in patients injured while wearing only a lap belt compared to a 3-point restraint (
60% vs. 25%) (Ball et al. 2000). Incidence of life-threatening intra-abdominal trauma from wearing a lap belt during vehicle collision has been reported in approximately 50% of flexion-distraction injury patients (Gertzbein and Court-Brown 1988; LeGay et al. 1990; Green et al. 1991). Anderson et al. (1991) reported a high rate of abdominal trauma associated with crashes where passengers were wearing only lap belts (Anderson et al. 1991). Thirteen of 20 patients (65%) required laparotomy, and eight of nine younger patients (>16y.o.) had lifethreatening intra-abdominal injuries. Anderson et al. note that children appear more likely to experience abdominal injury from lap belts in the event of a collision.

2382

B.D. Goodwin et al.

Other literature is in agreement with the study by Ball et al. (2000), which found that the most common injuries were small-bowel perforations and serosal tears. A PMHS study from 2015 found that specimens experience lateral abdominal deflections >50 mm in a vehicle with side and curtain airbags exposed to an oblique side-impact collision at approximately 25 km/h (Yoganandan et al. 2015b). These abrupt deformations of the thorax resulted in fractures to the four lowest ribs, which increases the risk of liver or spleen puncture and dramatically increases the internal pressure of abdominal organs facilitating rupture. Arm placement modestly influences the nature of injury during side impact to vehicle occupants, but it seems to mainly affect the number of rib fractures and their locations (Kemper et al. 2008). Injury criteria for the abdomen have been extensively studied where almost every measurable mechanical response has been investigated for use as a predictor of injury for the abdomen (Yoganandan et al. 2001). First and perhaps foremost, correlations have been identified between the presence of injury (and severity) and the amount of abdominal compression (Melvin et al. 1973; Stalnaker et al. 1973; Viano et al. 1989; Lamielle et al. 2008; Hardy et al. 2015). The most common organs (or regions) of the abdomen that have been injured during abdominal compression are upper and lower abdomen, liver, spleen, jejunum-ileum, and pancreas. Second, both the nature of the injury and its severity have been found to be considerably sensitive to the change in velocity during impact or impactor speed (Melvin et al. 1973; Kroell et al. 1981; Yoganandan et al. 2000). This velocity criteria was then taken a step further and combined with compression (or V*C), which was called the abdominal injury criteria (AIC) (Rouhana et al. 1985). It was found that the severity of abdominal injury was correlated well with AIC. Other variations of this criteria have been proposed, such as Vmax  Cmax (Stalnaker and Ulman 1985) and VC(t)max (called the viscous tolerance criterion for thoracic impact) (Viano and Lau 1983), each having good correlation with injury severity under various impact conditions. Kroell et al. (1986) investigated the relationships between velocity, compression, and heart rupture trauma from blunt impact to porcine subjects, and they found a better correlation between Vmax  Cmax over VC(t)max for probability of injury and AIS  4 (Kroell et al. 1986). Lastly, force or the manner in which force is applied influences the nature and severity of injury (Stalnaker et al. 1973; Trollope et al. Trollope et al. 1973; Haffner et al. 1996). Talantikite et al. (1993) performed 25 pendulum impact experiments on excised human livers at various speeds, and they found that peak forces (as measured from the pendulum) greater than 500 N yielded injuries AIS = 3 or greater (Talantikite et al. 1993). In this same study, whole body experiments appeared to have a tolerance threshold of 4.4 kN, which caused deflection of half the abdomen. Studies have also been carried out using variants of the force injury criteria such as impact energy, Fmax  Cmax, and pressure. Organ pressure correlations to injury seem to be limited to the liver (Foster et al. 2006), kidney (Rouhana et al. 1985), and lower abdomen (Miller 1989). Sparks et al. (2007) were able to show that internal pressure is a reasonable predictor of abdominal injury but found that the product of peak internal pressure and the peak time-derivative of pressure was the best predictor of liver injury (Sparks et al. 2007). These data

Injury Mechanisms in Traffic Accidents

2383

suggest that there is a 50% risk of injury with a vascular pressure of 64 kPa (9.28 psi). The abdomen seems to have a wide variety of injury mechanisms, which vary with organ tissue. Consequently, mechanisms of abdominal injury have been simplified through individual injury criteria.

Pelvis Pelvic fractures are considered severe injuries, and pelvic injury is often accompanied by damage to other organs. A high incidence of pelvis injuries exists among traffic crash victims (Ooi et al. 2010). The mortality rate of occupants in MVCs with pelvic fractures seems to indicate that pelvis fractures are accompanying injuries; i.e., pelvis injuries seem to be only correlated to incidence of death (Gokcen et al. 1994; Petrisor and Bhandari 2005). However, evidence suggests that fracture related hemorrhaging is associated with a high risk of death following pelvic disruption (Gabbe et al. 2011).

Anatomy The pelvis is a ring structure that supports the flexible spinal column and transmits the stress of weight bearing to the lower extremities. It is surrounded by a complex arrangement of muscles providing a thick compliant covering over the pelvis. The pelvis is stronger in the vertical and longitudinal directions to bear loads during walking and running. Conversely, it is weaker along the lateral direction due to small pubis bones at the front of the pelvic girdle. Four bones form the pelvis. Two innominate bones form the anterior and lateral walls of the ring, and the sacrum and coccyx make up the posterior wall. Each innominate bone consists of three fused segments, the ilium, ischium, and pubis. The fusion occurs around a cup-shaped articular cavity called the acetabulum, or the socket part of the hip joint. The ilium is divided into the large wing-like ala and the body of ilium forming the superior part of the acetabulum. The anterior superior iliac spine (ASIS) and the posterior superior iliac spine (PSIS) are bony prominences at the anterior and posterior extremities of the iliac crest. Sacroiliac joints attach the sacrum to the ilia through weight bearing synovial joints. The articular surfaces of the sacroiliac joint contain irregular depressions that interlock the two bones allowing limited joint motion. The ischium forms the lower lateral part of the innominate bone, and it constitutes the posterior third of the acetabulum cup. The lowest portion of the body is the ischial tuberosity, which supports the upper torso in a seated posture. The pubic bone is irregularly shaped and contains a body and two rami: the superior and inferior pubic rami. The superior ramus extends from the body to the mid-sagittal plane, where it articulates with the corresponding ramus on the opposite

2384

B.D. Goodwin et al.

side. The joint formed by the two superior rami is called pubic symphysis, which is a slightly movable joint containing a cartilaginous disc between the two bones. The inferior pubic rami join each other through the pubic symphysis. There are two major ligament groups in the pelvic region: (1) the ligaments surrounding the vertical load bearing sacroiliac joints, and (2) the ligaments passing between the sacrum and ischium. The former consists of anterior sacroiliac ligaments, posterior sacroiliac ligaments, and interosseous ligaments. The latter consists of sacrospinous ligaments and sacrotuberous ligaments.

Fracture Types Fracture types are classified according to the energy input to the pelvis as well as the direction of the force impact (Linnau et al. 2007). Pennal and Tile (1980) classified pelvic fractures based on the presumed force vector direction (Pennal et al. 1980), and Young and Burgess (1990) refined these classifications (Burgess et al. 1990). Pelvis fractures have been generally characterized according to the presumed loading vector as either anteroposterior compression (APC), lateral compression (LC), vertical shear (VS), or combined mechanical injury (CM) (Pennal et al. 1980). Each of these injury types has a commonly used associative severity scale from I to III, where III is considered unstable or most severe (e.g., APC-III or LC-III). APC fractures in the anterior aspect typically demonstrate pubic symphysis diastasis or a vertical fracture pattern of rami and posterior injuries defined by subsets. APC-I shows slight widening of pubic symphysis and one SI joint but have intact anterior and posterior SI joint ligaments. APC-II results in injury widening SI joint anteriorly affecting anterior SI ligaments and ipsilateral sacrotuberous and sacrospinous ligaments with intact posterior SI ligaments. Stability of APC-II depends on various ligaments involved. APC-III represents complete separation of hemipelvis from pelvic ring with rupture of anterior and posterior SI ligaments. APC-III is again treated as severely unstable and involves disruption of all sacroiliac joint ligaments (Young and Resnik 1990). LC pelvic injuries are caused through impact to the proximal femur that transfers a load to the iliac crest. Lateral compression fractures occur as the pelvis rotates toward the midline along the impact vector. The ilium rotates medially while its posterior is attached to the sacrum through posterior sacroiliac (SI) ligaments (Burgess et al. 1990). A secondary crush is common within the hemipelvis on the impact side, which causes an internal rotation; i.e., the contralateral hemi pelvis rotates externally causing an “open-book fracture.” This injury is considered LC-III or severely unstable since the compressive force also affects the contralateral side by way of an external rotation of the anterior pelvis (Young and Resnik 1990). VS injury results in a diastasis of the symphysis or a vertical fracture pattern of rami anteriorly, and it vertically displaces the hemipelvis (Burgess et al. 1990; Manson et al. 2010). Most of these fractures result from a fall from height and are uncommon in MVCs.

Injury Mechanisms in Traffic Accidents

2385

Mechanisms The majority of experimental PMHS work has been done for pelvis injury mechanisms through direct lateral or frontal impacts to the pelvis (Yoganandan et al. 2013a). Therefore, this discussion of injury mechanisms focuses primarily on those from side, oblique, and frontal impacts in MVCs.

Lateral Impact Lateral impact injury mechanisms are pertinent to side or oblique impacts in MVCs. During side impact MVCs, the impact is often delivered in a direct manner from the intruding door. Historically, PMHS studies have exposed whole cadavers to impacts along the lateral aspect of the greater trochanter of the femur. Pubic rami were found to experience substantial strain during lateral impact (Molz et al. 1997). According to Bonquet et al. (Bouquet et al. 1998), a 50% injury risk of AIS  2 is incurred when the pelvis deflection (or change in maximum lateral width) reaches 46 mm (Bouquet et al. 1998). While the risk of injury can be developed based on test measurements, the nature of the injury depends on the properties of the impactor. Using 20 unembalmed male cadavers, Nusholtz and Kaiker (1986) found that the nature of the injury depends on the rigidity of the impactor (Nusholtz and Kaiker 1986). In absence of padding, pelvic damage occurred at or around the acetabulum whereas the pubic area was prone to fractures from impact from a rigid device. Schiff et al. (2008) studied 728 lateral impact crashes that resulted in pelvic fractures and compared them to 5710 control cases involved in lateral impact without suffering any pelvic injuries (Schiff et al. 2008). The authors found statistically significant factors including occupant (1) age, (2) sex, (3) weight, and (4) vehicle type: (1) occupants 65 years and above had 70% increased risk, (2) nonpregnant females had 60% increased risk, (3) underweight occupants had 80% increased risk, and (4) occupants of vans had 80% decreased risk. Magnitude of lateral intrusion had the strongest relation to pelvic fracture. Cavanaugh et al. (1990) laterally impacted 12 whole unembalmed cadavers in a way where the energy was directed through the greater trochanter (Cavanaugh et al. 1990). The impact velocities ranged from 6.7 to 10.5 m/s. Injuries occurred to inferior and superior left pubic rami and the left sacroiliac joint. Pelvic injury tolerance was found between 8 kN and 10.6 kN. A peak force of 7.98 kN yielded 25% probability of injury. A peak compression of 32.6% corresponded to a 25% risk of fracture. Similar to intra-abdominal injury studies (cf. Sec. 0), the best injury criterion was related to Vmax  Cmax (peak velocity times peak compression). Likewise, Viano et al. (1989) performed a series of pendulum (23.4 kg) impact tests on unembalmed cadavers with impact velocities of 4.5, 6.7, and 9.4 m/s where the force vector was directed through the torso (Viano et al. 1989). They found that peak pelvic acceleration and pelvic deformation were not reliable measures, but the ratio of pelvic deformation to pelvic width correlated well with pubic rami fracture, which was the only type of injury observed. The tolerance level for 25% probability of serious injury to the pelvis was found to be 27% of pelvic compression, based on the entire width of the pelvis.

2386

B.D. Goodwin et al.

Impact energy that is delivered through the greater trochanter may result in substantial kinetic energy, which moves the contralateral hemipelvis against the center console of the vehicle. Tencer et al. (2007) analyzed pelvis accelerations and reviewed CIREN for pelvic injuries of occupants in vehicles with and without a center console. They found a higher incidence of pelvis fracture in MVCs with center consoles (Tencer et al. 2007). The abrupt deceleration from the center console contributes to injuries in the contralateral hemipelvis. The maximum and minimum accelerations for (1) fixed seat without console, (2) fixed seat with console, and (3) movable seat without console were (1) 28.5 g and 3.3 g, (2) 24.8 g and 10.5 g, and (3) 15.3 g and 3.8 g, respectively. A 50% reduction in primary pelvic acceleration was observed for vehicles without a center console.

Frontal/Rear Impact Occupants involved in head-on collisions in MVCs are exposed to a high risk of a frontal load on the pelvis. Compared to lateral impacts, frontal impacts indirectly load the pelvis where energy is transmitted from the knee, which is impacted by the dashboard (King 2001). Generally, injury mechanisms differ between front and rear vehicle occupants. A patella fracture is often accompanied by posterior hip dislocation in front-seat passengers, which have a tendency to sit with their knees together and hips flexed at 90 when the knee impacts the dashboard (Markham 1972). Rear side passengers thrown forward on impact may suffer anterior dislocation of the hip joint where higher energy loads can give rise to sacroiliac joint separation. This injury pattern for rear seat passengers is due to their tendency to sit with knees and hips flexed at 90 degrees and with hips in abduction with a slight external rotation (Markham 1972). Salzar et al. (2006) conducted a frontal impact study using an isolated hemi-pelvis model to study the posture-dependency of resulting fractures of the acetabulum and proximal femur (Salzar et al. 2006). Peak forces varied with the acetabular support area, and the highest sustained forces occurred under abduction and extension of the specimen (i.e., having the largest contact area) while adduction and flexion failed under lesser forces (i.e., having the least contact area). Interestingly, Masson et al. (2005) found the critical quasistatic load to strongly depend on sex by testing seven male and five female isolated pelves from embalmed cadavers (Masson et al. 2005). Quasistatic loads were applied in the anteriorposterior direction to the symphysis through a rectangular rigid plate. The sustainable force sustained before collapse ranged from 556 to 3981 N with average peak loads for female equal to 1053 N and for males 2501 N. The force-displacement corridors indicated that the female pelvic ring was more fragile due to the larger retro-pubic angle when controlling for bone density. Though a majority of pedestrians are struck laterally, frontal loading to the pelvis can occur when a pedestrian is hit from the front or back. After analyzing 1014 cases, 20.7% of the cases had injuries from anteroposterior compression (Eastridge and Burgess 1997). Pedestrian injury patterns are primarily influenced by front-end vehicle design. Simms et al. (2006) reviewed accident data and simulated pedestrian-vehicle interactions between sport utility vehicles (SUVs) and cars

Injury Mechanisms in Traffic Accidents

2387

(Simms and Wood 2006). The probability of pelvis injury was 2.5 times as likely in pedestrians when struck by a vehicle with an elevated bumper, as is common in SUVs.

Lower Extremities Anatomy The lower extremities include the femur head, acetabulum joint, thigh, femur, knee, patella, tibia, fibula, ankle, and foot. It consists of 29 distinct bone types, 72 articulating surfaces, 30 synovial joints, greater than 100 ligaments, and 30 muscle attachments (Crandall et al. 1996). The knee is made up of the patella, femoral condyles, and knee ligaments. The thigh comprises the supracondylar, shaft, and subtrochanteric region of the femur. The hip consists of the femoral head, neck, and acetabulum (the hip socket). Femur anatomy is included in the knee, thigh, and hip. The patella bone (or knee cap) hovers over the anterior knee, which acts as a mechanism to aid in leg extension and knee protection. Major knee ligaments include the posterior cruciate ligament (PCL), anterior cruciate ligament (ACL), medial collateral ligament (MCL), and lateral collateral ligament (LCL). The reader is encouraged to seek out other literature for in-depth explanations of ligament anatomy and functionality if so desired (Rupp 2015). The femur is a strong bone that spans the length of the thigh, and it consists of medial and lateral condyles at its distal head that articulate along the superior end of the tibia. The proximal end of the femur forms the trochanteric region from which the femoral neck sprouts allowing rotation about the hip. From a CIREN database study, from 1997 to 2003, the femoral shaft (31.5%) and the acetabulum (21.6%) were the most commonly reported injuries within the knee-thigh-hip arrangement (Rupp 2006). The tibia and fibula are attached by articulating surfaces at its proximal and distal ends. The ankle joint contains the talocrural joint, talocalcaneal joint, talocalcaneonavicular joint, and transverse tarsal joint. The hindfoot rotates in three planes of motion through internal and external rotation (transverse plane), dorsiflexion and plantarflexion (sagittal plane), and inversion and eversion (coronal plane) (Salzar et al. 2015). A common misconception is that the hindfoot is attached to the leg through hinge joints while the actual motion of the hindfoot is not contained within any single plane. This particularly applies to flexion of the hindfoot joints, which is not an isolated movement but is linked to inversion and eversion.

Mechanisms The knee-thigh-hip complex injuries appear to be the most commonly injured region of the lower extremities in MVC, and they are also the most severe lower extremity

2388

B.D. Goodwin et al.

injury due to the high risk of arterial rupture (Kuppa et al. 2003). The lower extremities are the most frequently injured in MVCs where
25% of these injuries are toward the knee due to the presence of knee bolsters. Knee injuries often have are associated with high DALY values due to long recovery times. Knee bolster stiffness varies across vehicles and many are sufficiently padded transmitting the critical loads to the femur and hip (Salzar et al. 2015).

Knee Injury Knee injuries make up about 16% of all injuries to the lower extremities (Chang et al. 2008). Database studies indicate that few knee injuries during lateral impacts have been found to be AIS > 2, and as a result, little attention has been given to lateral loading mechanisms throughout literature (Kuppa and Wang 2001; Rupp et al. 2002). Patellar fractures are the most common knee injury and frequently occur in frontal impacts when the knee interacts with the dashboard or, in the case of rear seat passengers, the seats in front of them. Fracture occurs after critical compression force between the external impactor and the femoral condyles. Evidence suggests that risk of injury increases especially in drivers who have a contracted knee extensor muscle for braking (or bracing) immediately prior to the MVC (Atkinson et al. 1998). The patella is also susceptible to a tension load when it impacts a surface obliquely, which subjects the patella to a sliding motion and can cause a patellar ligament avulsion. Femur Injury About 39% of all knee-thigh-hip injuries are thigh fractures (Chang et al. 2008). Injury to the knee is often accompanied by fractures to the femoral condyles because the patella is pushed into the intracondylar notch, which disunites the condyles (Powell et al. 1975). Femoral shaft fractures are the most frequent, and they are commonly the product of axial compression. The curvature of the femoral shaft has a weakening effect on its axial strength. Axial loads tend to induce a posterior-toanterior bending moment, especially from knee loading applied medially to the femoral condyle (Viano and Khalil 1976; Viano and Stalnaker 1980; Rupp et al. 2003; Ivarsson et al. 2009). The mechanisms of injury of the trochanteric femur due to axial loading remain somewhat elusive, but it has been hypothesized that the mechanism is a combination of tension within the greater trochanter, the anchoring of the femoral head to the acetabulum, and the moment arm of the femoral neck (Rupp 2006). In fact, evidence suggests that fractures within the femoral neck occur through similar mechanisms (Rupp et al. 2003). Occupants often brace for the impending collision causing muscle contraction. PMHS experiments designed to reproduce loads caused by MVCs have produced injuries that are somewhat inconsistent with data collected by CIREN. The relaxed nature of PMHS specimen yields knee, distal femur, and hip injuries from frontal MVCs, while the CIREN database indicates that the midshaft of the femur is the most common injury (Rupp 2006). Chang et al. (2008) proposed that muscle activation preloads the femur prior to impact increasing the bending moment applied to the femur (Chang et al. 2008).

Injury Mechanisms in Traffic Accidents

2389

Hip Injury Fourty-five percent of knee-thigh-hip injuries are hip fractures and dislocations from frontal impacts where fracture of the acetabulum is most frequent (Rupp et al. 2003; Chang et al. 2008). There are many parameters that influence hip response and susceptibility to injury including hip posture during knee loading as well as interactions between the trochanter and components of the vehicle interior. Differences in the geometry of the acetabulum between men and women also have been shown to affect the extent and nature of fracture (Wang et al. 2004; Holcombe et al. 2011). Similarly, body mass and the center of gravity of the upper body prior to impact have a considerable affect on the hip response since it is coupled to the abdomen and lower extremity masses (Rupp et al. 2008). Approximately 70% of AIS > 2 injuries to the lower extremities from lateral impacts in nearside occupants during MVCs are hip or pelvis injuries compared to the 9% of injuries that are to either the thigh or knee (Banglmaier et al. 2003). Far side occupants are exposed to different loading conditions and only 38% of AIS > 2 injuries are to the hip or pelvis. Lower Leg Injury The vast majority of lower extremity injuries are to the knee-thigh-hip, and injuries to the lower leg complex rarely result in life-threatening circumstances. As vehicular safety standards have improved and the number of airbag equipped vehicles increased, severe injuries to the lower leg have increased since critical body regions are more protected than the lower leg. Additionally, prior to the advent of airbags, lower leg injuries were less relevant and generally undocumented in the event of a life threatening or fatal injury (Salzar et al. 2015). The effects of bending moments on the lower leg have been extensively studied, and its tolerance depends on the loading rate, direction, and magnitude (Schreiber et al. 1998; Kerrigan et al. 2004; Ivarsson et al. 2005; Yoganandan et al. 2014a). Focal lateral loads directed toward the midshaft region of the tibia generated injuries that indicated the general location of the fracture was relatively independent of the force direction (e.g., lateral-medial vs. anterior-posterior) (Rabl et al. 1996). Ankle Injury Ankle injuries are the most severe lower leg injuries and are common in frontend MVCs. Interactions between the lower leg and the break/accelerator pedal have been attributed to approximately 25% of inversion or eversion (Xversion; rotation about the X-axis; rotation about the anterior-posterior axis) related injuries (Morris et al. 1997). Lower leg injuries depend upon axial preloads or preflexion (e.g., through the Achilles tendon), and injuries are likely more severe when lower leg muscles are activated to hurriedly and forcefully depress the break pedal, which preloads the foot in dorsiflexion (Funk et al. 2001). It is commonly understood that injuries to the ankle occur when the foot is forced into Xversion. PMHS experiments that employed axial preloads followed by forced Xversion produced (medial and lateral) ligament tears, malleolar fractures, and tibial osteochondral fractures (Funk et al. 2002). Petit et al. (1997) were able to simulate

2390

B.D. Goodwin et al.

tension in the Achilles tendon while quasistatically loading the foot into dorsiflexion and consistently observed medial malleolus fractures and calcaneofibular ligament tears (Petit et al. 1997). Rudd et al. (2004) conducted PMHS experiments that simulated the force applied to the foot under dorsiflexion when the driver was depressing the break pedal (Rudd et al. 2004). This setup provided an appropriate simulation of the leg in the event of a front-end MVC. Bony fractures at the ankle joint were present in 11 of 20 specimens, ligament ruptures in 4 specimens, and osteochondral damage in almost all specimens. Rudd et al. (2004) used surface acoustic sensors to detect fracture timing, and they point out that the fractures to the ankle do not necessarily align with the peak force of the dynamic load. Instead, the ankle seems to continue to bear higher load after initial fractures until the fracture is catastrophic.

Future Directions Despite the fact that biomechanics has largely evolved to applied biomechanics, considerable debate surrounds many accepted mechanisms. The current debate calls for further validation and repeated experiments. However, objections to accepted injury mechanisms are difficult to validate since funding opportunities for experimental replication are scarce. Albert King comments on this problem, “. . . [those] who review research proposals . . . are disinclined to reverse a previous opinion and tend to disapprove any proposal that will reverse this opinion” (King 2015). However, advancements in instrumentation will continue to allow researchers to peer into the biomechanical response with more detail. As long as funding opportunities in applied biomechanics remain available. As transportation technology evolves, first-world economies will continue to demand more sophisticated tests from the biomechanics community. However, the complexity of tests often precedes scientific understanding. As a result, injury explanations can contain much speculation, but PMHS tests will indeed indicate which injuries are likely to occur under defined loading conditions. Finally, the anthropomorphic test devices (ATDs) have provided insight into the biomechanical response with acceptable biofidelity (Baudrit and Trosseille 2015). The Hybrid III ATD was developed and has been used since the early 1990s (Mertz 1993). As vehicle and traffic technology advances, the body will be exposed to new and variable loads in the event of an MVC. Recently, Danelson et al. evaluated the biofidelity of the Hybrid III during vertical loading, and they concluded that there is a need for further developments to improve biofidelity (Danelson et al. 2015). The Hybrid III was designed specifically to mimic a seated automobile occupant, but the divergence from the PMHS response was sizable. Similarly, Foreman et al. point out the need for a biofidelic ATD response from pedestrian hits (Foreman et al. 2015). Injuries with complex mechanisms (especially soft tissue injuries) are difficult to predict from the ATD response. Future work is necessary to better estimate the human response through ATDs, and a vast amount of work has been carried out by

Injury Mechanisms in Traffic Accidents

2391

way of biofidelity studies, response corridor generation techniques, and PMHS experimentation to pave the way for future improvements.

Cross-References ▶ Applications in Forensic Biomechanics ▶ Biomechanical Forensics in Pediatric Head Trauma ▶ Cross-Platform Comparison of Imaging Technologies for Measuring Musculoskeletal Motion ▶ Estimation of the Body Segment Inertial Parameters for the Rigid Body Biomechanical Models Used in Motion Analysis ▶ Expert Opinion and Legal Considerations ▶ Time Series Analysis in Biomechanics ▶ Scaling and Normalization ▶ Vehicle Occupants in Traffic Accidents

References Allen B Jr, Ferguson R, Lehmann TR, O’brien RP (1982) A mechanistic classification of closed, indirect fractures and dislocations of the lower cervical spine. Spine (Phila Pa 1976) 7:1–27 Anderson PA, Henley MB, Rivara FP, Maier R V (1991) Flexion distraction and chance injuries to the thoracolumbar spine. J Orthop Trauma 5:153–160 Anderson RWG, Brown CJ, Blumbergs PC et al (2003) Impact mechanics and axonal injury in a sheep model. J Neurotrauma 20:961–974. https://doi.org/10.1089/089771503770195812 Anuta PE (1970) Spatial registration of multispectral and multitemporal digital imagery using fast Fourier transform techniques. IEEE Trans Geosci Electron 8:353–368 Atkinson P, Atkinson T, Haut R, Eusebi C, Maripudi V, Hill T, Sambatur K (1998) Development of injury criteria for human surrogates to address current trends in knee-to-instrument panel injuries (No. 983146). SAE Technical Paper Augenstein J, Diggs K (2003) Performance of advanced air bags based on data William Lehman Injury Research Center and new NASS PSUs. Annu Proc Assoc Adv Automot Med 47:99–101 Augenstein J, Perdeck E, Martin P et al (2000) Injuries to restrained occupants in far-side crashes. Ann Adv Automot Med 44:57–66 Ball ST, Vaccaro AR, Albert TJ (2000) Injuries of the thoracolumbar spine associated with restraint use in head-on motor vehicle accidents. Spinal Disord 13:297–304 Balogh ZJ, Offner PJ, Moore EE, Biffl WL (2000) NISS predicts postinjury multiple organ failure better than the ISS. J Trauma Acute Care Surg 48:624–628 Balogh ZJ, Varga E, Tomka J et al (2003) The new injury severity score is a better predictor of extended hospitalization and intensive care unit admission than the injury severity score in patients with multiple orthopaedic injuries. J Orthop Trauma 17:508–512. https://doi.org/ 10.1097/00005131-200308000-00006 Banglmaier RF, Rouhana SW, Beillas P, Yang KH (2003) Lower extremity injuries in lateral impact: a retrospective study. Ann Proc Assoc Adv Automot Med 47:425–444 Barnsley L, Lord S, Wallis B, Bogduk N (1995) The prevalence of chronic cervical zygapophysial joint pain after whiplash. Spine (Phila Pa 1976) 20:20–26 Baudrit P, Trosseille X (2015) Proposed method for development of small female and midsize male thorax dynamic response corridors in side and forward oblique impact tests. Stapp Car Crash J 59:177–202

2392

B.D. Goodwin et al.

Baur P, Lange W, Messner G et al (2000) Comparison of real world side impact/rollover collisions with and without thorax airbag/head protection system: a first field experience study. Ann Proc Assoc Adv Automot Med 44:187–201 Begeman PC, King AI, Prasad P (1973) Spinal loads resulting from -Gx acceleration. In: Proceedings of the 17th stapp car crash conference, Coronado, CA, pp 343–360 Begonia MT, Dallas M, Vizcarra B et al (2015) Non-contact strain measurement in the mouse forearm loading model using digital image correlation (DIC). Bone 81:593–601. https://doi.org/ 10.1016/j.bone.2015.09.007 Bergen GS, Chen LH, Warner M (2008) Injury in the United States; 2007 chartbook Bertrand S, Cuny S, Petit P et al (2008) Traumatic rupture of thoracic aorta in real-world motor vehicle crashes. Traffic Inj Prev 9:153–161. https://doi.org/10.1080/15389580701775777 Bogduk N, Marsland A (1988) The cervical zygapophysial joints as a source of neck pain. Spine (Phila Pa 1976) 13:610–617 Bouquet R, Ramet M, Bermond F et al (1998) Pelvis human response to lateral impact. In: Proceedings of the 16th international technical conference on the enhanced safety of vehicles, Windsor, pp 1665–1686 Burgess AR, Eastridge BJ, Young JWR, Ellison TS, Ellison Jr PS, Poka A, et al (1990) Pelvic ring disruptions: effective classification system and treatment protocols. J Trauma Acute Care Surg 30:848–856 Cammack K, Rapport RL, Paul J, Baird WC (1959) Deceleration injuries of the thoracic aorta. AMA Arch Surg 79:244–251. https://doi.org/10.1001/archsurg.1959.04320080080010 Carroll LJ, Holm LW, Hogg-Johnson S et al (2009) Course and prognostic factors for neck pain in Whiplash-Associated Disorders (WAD). Results of the bone and joint decade 2000-2010 Task force on neck pain and its associated disorders. J Manip Physiol Ther 32:S97–S107. https://doi. org/10.1016/j.jmpt.2008.11.014 Cavanaugh JM, Yoganandan NA (2015) Thorax injury biomechanics. In: Yoganandan N, Nahum AM, Melvin JW (eds) Accidental injury: biomechanics and prevention, 3rd edn. Springer, New York, pp 332–334 Cavanaugh JM, Walilko TJ, Malhotra A, Zhu Y, King AI (1990) Biomechanical response and injury tolerance of the pelvis in twelve sled side impacts (No. 902305). SAE Technical Paper Chance GQ (1948) Note on a type of flexion fracture of the spine. Br J Radiol 21:452 Chancey VC, Nightingale RW, C a VE et al (2003) Improved estimation of human neck tensile tolerance: reducing the range of reported tolerance using anthropometrically correct muscles and optimized physiologic initial conditions. Stapp Car Crash J 47:135–153 Chang C-Y, Rupp JD, Kikuchi N, Schneider LW (2008) Development of a finite element model to study the effects of muscle forces on knee-thigh-hip injuries in frontal crashes. Stapp Car Crash J 52:475 Cheng R, Yang KH, Levine RS, King AI, Morgan R (1982) Injuries to the cervical spine caused by distributed frontal load to the chest. SAE Paper #821155 Chu TC, Ranson WF, Sutton MA (1985) Applications of digital-image-correlation techniques to experimental mechanics. Exp Mech 25:232–244 Crandall J, Martin P, Bass C et al. (1996) Foot and ankle injury: the roles of driver anthropometry, footwear, and pedal controls. Paper presented at: 40th Annual Proceedings of the Association for the Advancement of AutomotiveMedicine Culver RH, Bender M, Melvin JW (1978) Mechanisms, tolerances, and responses obtained under dynamic superior-inferior head impact. Ann Arbor 7:103 Danelson KA, Kemper AR, Mason MJ et al (2015) Comparison of ATD to PMHS Response in the under-body blast environment. Stapp Car Crash J 59:445–520 Davceva N, Janevska V, Ilievski B et al (2012) The occurrence of acute subdural haematoma and diffuse axonal injury as two typical acceleration injuries. J Forensic Legal Med 19:480–484. https://doi.org/10.1016/j.jflm.2012.04.022 Deng B, Begeman PC, Yang KH et al (2000) Kinematics of human cadaver cervical spine during low speed rear-end impacts. Stapp Car Crash J 44:171–188

Injury Mechanisms in Traffic Accidents

2393

Depreitere B, Van Lierde C, Vander SJ et al (2006) Mechanics of acute subdural hematomas resulting from bridging vein rupture. J Neurosurg 104(6):950. https://doi.org/10.3171/jns.2006.104.6.950 Eastridge BJ, Burgess AR (1997) Pedestrian pelvic fractures: 5-year experience of a major urban trauma center. J Trauma Acute Care Surg 42:695–700 Fice JB, Cronin DS (2012) Investigation of whiplash injuries in the upper cervical spine using a detailed neck model. J Biomech 45:1098–1102. https://doi.org/10.1016/j.jbiomech.2012. 01.016 Foreman JL, Joodaki H, Forghani A et al (2015) Whole-body response for pedestrian impact with a generic sedan buck. Stapp Car Crash J 59:401–444 Foster CD, Hardy WN, Yang KH et al (2006) High-speed seatbelt pretensioner loading of the abdomen. Stapp Car Crash J 50:27–51 Funk JR, Crandall JR, Tourret LJ, MacMahon CB, Bass CR, Khaewpong N, Eppinger RH (2001) The effect of active muscle tension on the axial injury tolerance of the human foot/ankle complex (No. 2001-06-0074). SAE Technical Paper Funk JR, Srinivasan SCM, Crandall JR et al (2002) The effects of axial preload and dorsiflexion on the tolerance of the ankle/subtalar joint to dynamic inversion and eversion. Stapp Car Crash J 46:245–265. doi:2002-22-0013 [pii] Gabbe BJ, De Steiger R, Esser M et al (2011) Predictors of mortality following severe pelvic ring fracture: results of a population-based study. Injury 42:985–991 Gabler HC, Weaver AA, Stitzel JD (2015) Automotive field data in injury. In: Yoganandan N, Nahum AM, Melvin JW (eds) Accidental injury: biomechanics and prevention, 3rd edn. Springer, New York, pp 33–47 Gennarelli TA, Thibault LE (1982) Biomechanics of acute subdural hematoma. J Trauma 22:680–686 Gertzbein SD, Court-Brown CM (1988) Flexion-distraction injuries of the lumbar spine. Clin Orthop Relat Res 227:52–60 Gokcen EC, Burgess AR, Siegel JH et al (1994) Pelvic fracture mechanism of injury in vehicular trauma patients. J Trauma Acute Care Surg 36:789–796 Green DA, Green NE, Spengler DM, Devito DP (1991) Flexion-distraction injuries to the lumbar spine associated with abdominal injuries. J Spinal Disord Tech 4:312–318 Gurdjian ES, Webster JE, LH R (1955) Observations on the mechanism of brain concussion, contusion, and laceration. Surg Gynecol Obstet 101:680–690 Haagsma JA, Graetz N, Bolliger I et al (2016) The global burden of injury: incidence, mortality, disability-adjusted life years and time trends from the Global Burden of Disease study 2013. Inj Prev 22:3–18. https://doi.org/10.1136/injuryprev-2015-041616 Haffner MP, Sances S, Kumaresan S et al (1996) Response of human lower thorax to impact. Ann Proc Assoc Adv Automot Med 40:33–43 Hardy WN, Foster CD, Mason MJ, Yang KH, King AI, Tashman S (2001) Investigation of head injury mechanisms using neutral density technology and high-speed biplanar x-ray. Stapp Car Crash J 45:337–368 Hardy WN, Shah CS, Kopacz JM et al (2006) Study of potential mechanisms of traumatic rupture of the aorta using insitu experiments. Stapp Car Crash J 50:247–266 Hardy WN, Shah CS, Mason MJ et al (2008) Mechanisms of traumatic rupture of the aorta and associated peri-isthmic motion and deformation. Stapp Car Crash J 52:233–265. doi:2008-22-0010 [pii] Hardy WN, Howes MK, Kemper AR, Rouhana SW (2015) Impact and injury response of the abdomen. In: Yoganandan N, Nahum AM, Melvin JW (eds) Accidental injury: biomechanics and prevention, 3rd edn. Springer, New York, pp 373–434 Holcombe S, Kohoyda-Inglis C, Wang L et al (2011) Patterns of acetabular femoral head coverage. Stapp Car Crash J 55:479–490 Ivarsson BJ, Kerrigan JR, Lessley DJ, Drinkwater DC, Kam CY, Murphy DB, . . . Kent RW (2005) Dynamic response corridors of the human thigh and leg in non-midpoint three-point bending (No. 2005-01-0305). SAE Technical Paper Ivarsson BJ, Genovese D, Crandall JR et al (2009) The tolerance of the femoral shaft in combined axial compression and bending loading. Stapp Car Crash J 53:251

2394

B.D. Goodwin et al.

Jefferson G (1919) Fracture of the atlas vertebra. Report of four cases, and a review of those previously recorded. Br J Surg 7:407–422 Katyal D, McLellan BA, Brenneman FD et al (1997) Lateral impact motor vehicle collisions: significant cause of blunt traumatic rupture of the thoracic aorta. J Trauma 42:769–772 Kauvar DS, Wade CE (2005) The epidemiology and modern management of traumatic hemorrhage: US and international perspectives. Crit Care 9(Suppl 5):S1–S9. https://doi.org/10.1186/cc3779 Kemper AR, McNally C, E a K et al (2008) The influence of arm position on thoracic response in side impacts. Stapp Car Crash J 52:379–420. https://doi.org/10.4271/811007 Kerrigan JR, Drinkwater DC, Kam CY et al (2004) Tolerance of the human leg and thigh in dynamic latero-medial bending. Int J Crashworthiness 9:607–623 King AI (2001) Fundamentals of impact biomechanics: part 2-biomechanics of the abdomen, pelvis, and lower extremities. Annu Rev Biomed Eng 3:27–55 King AI (2015) Introduction to and applications of injury biomechanics. In: Narayan Y, Nahum AM, Melvin JW (eds) Accidental injury: biomechanics and prevention, 3rd edn. Springer, New York, pp 7–14 King AI, Yang KH (1995) Research in biomechanics of occupant protection. J Trauma 38:570–576 Kirk A, Morris A (2003) Side airbag deployments in the UK – initial case reviews. In: Proceedings of the 18th international technical conference on enhanced safety of vehicles. Nagoya, pp 1–8 Klinich KD, Flannagan CAC, Nicholson K et al (2010) Factors associated with abdominal injury in frontal, farside, and nearside crashes. Stapp Car Crash J 54:73–91 Kroell CK, Pope ME, Viano DC et al (1981) Interrelationship of velocity and chest compression in blunt thoracic impact. In: 25th Stapp car crash conference proceedings. SAE Technical Paper 811016, Warrendale, PA, pp 547–580 Kroell C, Allen SD, Warner CY, Perl T (1986) Interrelatinoship of velocity and chest compression in blunt thoracic impact to Swine II. In: 30th Stapp car crash conference proceedings SAE Technical Paper 861881, Warrendale, PA, pp 99–121 Kuppa S, Wang J, Haffner M, Eppinger R (2001) Lower extremity injuries and associated injury criteria. In 17th ESV Conference, Paper (No. 457) Kuppa S, Fessahaie O et al (2003) An overview of knee-thigh-hip injuries in frontal crashes in the United States. Natl Highw Traffic Saf Adm ISSI. https://doi.org/10.1017/CBO9781107415324.004 Lamielle S, Vezin P, Verriest JP et al (2008) 3D deformation and dynamics of the human cadaver abdomen under seatbelt loading. Stapp Car Crash J 52:267 Lee BB, Cripps RA, Fitzharris M, Wing PC (2014) The global map for traumatic spinal cord injury epidemiology: update 2011, global incidence rate. Spinal Cord 52:110–116. https://doi.org/ 10.1038/sc.2012.158 LeGay DA, Petrie DP, Alexander DI (1990) Flexion-distraction injuries of the lumbar spine and associated abdominal trauma. J Trauma 30:436–444 Leitgeb J, Mauritz W, Brazinova A et al (2012) Outcome after severe brain trauma due to acute subdural hematoma. J Neurosurg 117:324–333. https://doi.org/10.3171/2012.4.JNS111448 Lessley DJ, Riley P, Zhang Q et al (2014) Occupant kinematics in laboratory rollover tests: PMHS response. Stapp Car Crash J 58:251 Leucht P, Fischer K, Muhr G, Mueller EJ (2009) Epidemiology of traumatic spine fractures. Injury 40:166–172. https://doi.org/10.1016/j.injury.2008.06.040 Linnau KF, Blackmore CC, Kaufman R et al (2007) Do initial radiographs agree with crash site mechanism of injury in pelvic ring disruptions? A pilot study. J Orthop Trauma 21:375–380 Lu Y, Chen C, Kallakuri S et al (2005) Neurophysiological and biomechanical characterization of goat cervical facet joint capsules. J Orthop Res 23:779–787. https://doi.org/10.1016/j.orthres.2005.01.002 Manson T, O’Toole RV, Whitney A et al (2010) Young-Burgess classification of pelvic ring fractures: does it predict mortality, transfusion requirements, and non-orthopaedic injuries? J Orthop Trauma 24:603–609 Markham DE (1972) Anterior dislocation of the hip and diastasis of the contralateral sacro-iliac joint – the rear-seat passenger’s injury? Br J Surg 59:296–298 Masson C, Baque P, Brunet C (2005) Quasi-static compression of the human pelvis: an experimental study. Comput Methods Biomech Biomed Engin 8:191–192

Injury Mechanisms in Traffic Accidents

2395

Matsui Y, Oikawa S (2015) Risks of serious injuries and fatalities of cyclists associated with impact velocities of cars in car-cyclist accidents in Japan. Stapp Car Crash J 59:385–400 McCormick N, Lord J (2010) Digital image correlation. Mater Today 13:52–54. https://doi.org/ 10.1016/S1369-7021(10)70235-2 McKay BJ, Bir CA (2009) Lower extremity injury criteria for evaluating military vehicle occupant injury in underbelly blast events. Stapp Car Crash J 53:229–249 Melvin JW, Stalnaker RL, Roberts VL, Trollope ML (1973) Impact injury mechanisms in abdominal organs. In: 17th Stapp car crash conference. Society of Automotive Engineers, Oklahoma City Mertz HJ (1993) Anthropomorphic test devices. In: Accidental injury. Springer, New York, pp 66–84 Middleton JW, Dayton A, Walsh J et al (2012) Life expectancy after spinal cord injury: a 50-year study. Spinal Cord 50:803–811 Miller M (1989) The biomechanical response of the lower abdomen to belt restraint loading. J Trauma Acute Care Surg 29:1571–1584 Molz FJI V, George PD, Go LS et al (1997) Simulated automotive side impact on the isolated human pelvis: Phase I: development of a containment device. Phase II: analysis of pubic symphysis motion and overall pelvic compression. In: Stapp car crash conference proceedings, Warrendale, PA, pp 75–89 Morris A, Thomas P, Taylor AM, Wallace WA (1997) Mechanisms of fracture in ankle and hindfoot injuries to front seat car occupants- an in-depth accident data analysis. Stapp Car Crash Conf 41:181–192. https://doi.org/10.4271/973328 Myers BS, Winkelstein BA (1995) Epidemiology, classification, mechanism, and tolerance of human cervical spine injuries. Crit Rev Biomed Eng 23(5–6):307–409 NHTSA (2005) Federal motor vehicle safety standards; Occupant crash protection. Docket No. NHTSA-04-18726 Nightingale RW, McElhaney JH, Richardson WJ et al (1996a) Experimental impact injury to the cervical spine: relating motion of the head and the mechanism of injury. J Bone Jt Surg Am 78:412–421 Nightingale RW, McElhaney JH, Richardson WJ, Myers BS (1996b) Dynamic responses of the head and cervical spine to axial impact loading. J Biomech 29:307–318 Nightingale RW, McElhaney JH, Camacho DL, Kleinberger M, Winkelstein BA, Myers BS (1997) The dynamic responses of the cervical spine: buckling, end conditions, and tolerance in compressive impacts (No. 973344). SAE Technical Paper Nightingale RW, Myers BS, Yoganandan NA (2015) Neck injury biomechanics. In: Yoganandan NA, Nahum AM, Melvin JW (eds) Accidental injury: biomechanics and prevention, 3rd edn. Springer, New York, pp 259–308 Nirula R, Pintar FA (2008) Identification of vehicle components associated with severe thoracic injury in motor vehicle crashes: a CIREN and NASS analysis. Accid Anal Prev 40:137–141. https://doi.org/10.1016/j.aap.2007.04.013 Nusholtz GS, Kaiker PS (1986) Pelvic stress. J Biomech 19:1003–1014 Nusholtz GS, Melvin JW, Huelke DF, Alem NM, Blank JG (1981) Response of the cervical spine to superiorinferior head impact (No. 811005). SAE Technical Paper Nusholtz GS, Huelke DE, Lux P, Alem NM, Montalvo F (1983) Cervical spine injury mechanisms (No. 831616). SAE Technical Paper Ooi CK, Goh HK, Tay SY, Phua DH (2010) Patients with pelvic fracture: what factors are associated with mortality? Int J Emerg Med 3:299–304 Panzer MB, Fice JB, Cronin DS (2011) Cervical spine response in frontal crash. Med Eng Phys 33:1147–1159. https://doi.org/10.1016/j.medengphy.2011.05.004 Parr JC, Miller ME, Pellettiere JA, Erich RA (2013) Neck injury criteria formulation and injury risk curves for the ejection environment: a pilot study. Aviat Sp Environ Med 84:1240–1248. https:// doi.org/10.3357/ASEM.3722.2013 Pennal GF, Tile M, Waddell JP, Garside H (1980) Pelvic disruption: assessment and classification. Clin Orthop Relat Res 151:12–21

2396

B.D. Goodwin et al.

Petit P, Portier L, Foret-Bruno J-Y et al (1997) Quasistatic characterization of the human foot-ankle joints in a simulated tensed state and updated accidentological data. In: Proceedings of the international research council on the biomechanics of injury conference, Hannover, Germany, pp 363–376 Petitjean A, Trosseille X (2011) Statistical simulations to evaluate the methods of the construction of injury risk curves. Stapp Car Crash J 55:411 Petitjean A, Trosseille X, Yoganandan N, Pintar FA (2015) Normalization and scaling for human response corridors and development of injury risk curves. In: Yoganandan N, Nahum AM, Melvin J (eds) Accidental injury: biomechanics and prevention, 3rd edn. Springer, New York, pp 769–792 Petrisor BA, Bhandari M (2005) (i) Injuries to the pelvic ring: Incidence, classification, associated injuries and mortality rates. Curr Orthop 19:327–333 Pintar FA, Yoganandan N, Voo L, Cusick JF, Maiman DJ, Sances A (1995) Dynamic characteristics of the human cervical spine (No. 952722). SAE Technical Paper Pintar FA, Yoganandan NA, Voo L (1998) Effect of age and loading rate on human cervical spine injury threshold. Spine (Phila Pa 1976) 23:1957–1962 Pintar FA, Yoganandan NA, Maiman DJ et al (2012) Thoracolumbar spine fractures in frontal impact crashes. Ann Adv Automot Med 56:277–283 Powell WR, Ojala SJ, Advani SH, Martin RB (1975) Cadaver femur responses to longitudinal impacts (No. 751160). SAE Technical Paper Rabl W, Haid C, Krismer M (1996) Biomechanical properties of the human tibia: fracture behavior and morphology. Forensic Sci Int 83:39–49 Robertson A, Branfoot T, Barlow IF, Giannoudis PV (2002a) Spinal injury patterns resulting from car and motorcycle accidents. Spine (Phila Pa 1976) 27:2825–2830. https://doi.org/10.1097/01. BRS.0000035686.45726.0E Robertson A, Giannoudis P V, Branfoot T et al (2002b) Spinal injuries in motorcycle crashes: patterns and outcomes. J Trauma 53:5–8. https://doi.org/10.1097/00005373-200207000-00002 Rouhana SW, Lau IV, Ridella SA (1985) Influence of velocity and forced compression on the severity of abdominal injury in blunt, nonpenetrating lateral impact. J Trauma 25:490–500 Rudd R, Crandall J, Millington S et al (2004) Injury tolerance and response of the ankle joint in dynamic dorsiflexion. Stapp Car Crash J 48:1–26. doi:2004-22-0001 [pii] Rupp JD (2006) Biomechanics of hip fractures in frontal motor vehicle crashes. Phdthesis, PhD dissertation, The University of Michigan, Ann Arbor Rupp JD (2015) Knee, thigh, and hip injury biomechanics. In: Yoganandan NA, Nahum AM, Melvin JW (eds) Accidental injury: biomechanics and prevention, 3rd edn. Springer, New York, pp 471–497 Rupp JD, Reed MP, Van Ee CA et al (2002) The tolerance of the human hip to dynamic knee loading. In: 46th Stapp car crash conference, Ponte Verdra Beach, FL Rupp JD, Reed MP, Jeffreys TA, Schneider LW (2003) Effects of hip posture on the frontal impact tolerance of the human hip joint. Stapp Car Crash J 47:21 Rupp JD, Miller CS, Reed MP et al (2008) Characterization of knee-thigh-hip response in frontal impacts using biomechanical testing and computational simulations. Stapp Car Crash J 52:421 Salzar RS, “Dale” Bass CR, Kent R et al (2006) Development of injury criteria for pelvic fracture in frontal crashes. Traffic Inj Prev 7:299–305 Salzar RS, Lievers BW, Bailey AM, Crandall JR (2015) Leg, foot, and ankle injury biomechanics. In: Yoganandan NA, Nahum AM, Melvin JW (eds) Accidental injury: biomechanics and prevention, 3rd edn. Springer, New York, pp 499–547 Santschi M, Echavé V, Laflamme S et al (2005) Seat-belt injuries in children involved in motor vehicle crashes. Can J Surg 48:373–376 Schiff MA, Tencer AF, Mack CD (2008) Risk factors for pelvic fractures in lateral impact motor vehicle crashes. Accid Anal Prev 40:387–391 Schreiber P, Crandall J, Hurwitz S, Nusholtz GS (1998) Static and dynamic bending strength of the leg. Int J Crashworthiness 3:295–308

Injury Mechanisms in Traffic Accidents

2397

Sekhon LH, Fehlings MG (2001) Epidemiology, demographics, and pathophysiology of acute spinal cord injury. Spine (Phila Pa 1976) 26:S2–12. https://doi.org/10.1097/00007632-200112151-00002 Services UDoHaH (2007) Injury in the United States: 2007 Shinkawa H, Yasuhara H, Naka S et al (2004) Characteristic features of abdominal organ injuries associated with gastric rupture in blunt abdominal trauma. Am J Surg 187:394–397 Shkrum M, McClafferty K, Green R et al (1999) Mechanisms of aortic injury in fatalities occurring in motor vehicle collisions. J Forensic Sci 44:44–56 Siegmund GP, Myers BS, Davis MB et al (2001) Mechanical evidence of cervical facet capsule injury during whiplash: a cadaveric study using combined shear, compression, and extension loading. Spine (Phila Pa 1976) 26:2095–2101 Siegmund GP, Davis MB, Quinn KP et al (2008) Head-turned postures increase the risk of cervical facet capsule injury during whiplash. Spine (Phila Pa 1976) 33:1643–1649. https://doi.org/ 10.1097/BRS.0b013e31817b5bcf Simms CK, Wood DP (2006) Pedestrian risk from cars and sport utility vehicles-a comparative analytical study. Proc Inst Mech Eng Part D J Automob Eng 220:1085–1100 Somers RL (1983a) The probability of death score: a measure of injury severity for use in planning and evaluating accident prevention. Accid Anal Prev 15:259–266. https://doi.org/10.1016/ 0001-4575(83)90050-7 Somers RL (1983b) The probability of death score: an improvement of the injury severity score. Accid Anal Prev 15:247–257. https://doi.org/10.1016/0001-4575(83)90049-0 Sparks JL, Bolte JH, Dupaix RB et al (2007) Using pressure to predict liver injury risk from blunt impact. Stapp Car Crash J 51:401–432 Stalnaker RL, Ulman MS (1985) Abdominal trauma – review, response, and criteria. In: 29th Stapp car crash conference proceedings, pp 1–16 Stalnaker RL, McElhaney JH, Roberts VL, Trollope ML (1973) Human torso response to blunt trauma. In: King WF, Mertz HJ (eds) Human impact response: measurement and simulation. Springer US, Boston, pp 181–199 Stemper BD, Yoganandan NA, Pintar FA (2004) Validation of a head-neck computer model for whiplash simulation. Med Biol Eng Comput 42:333–338 Stemper BD, Pintar FA, Baisden JL (2015) Lumbar spine injury biomechanics. In: Yoganandan N, Nahum AM, Melvin JW (eds) Accidental injury: biomechanics and prevention, 3rd edn. Springer, New York, pp 451–470 Takhounts EG, Craig MJ, Moorhouse K et al (2013) Development of brain injury criteria (BrIC). Stapp Car Crash J 57:243 Talantikite Y, Brun-Cassan F, Le coz J, Tarriere C (1993) Abdominal protection in side impact. Injury mechanisms and protection criteria. Proc Int Res Counc Biomech Inj Conf 21:131–144 Tencer AF, Kaufman R, Huber P et al (2007) Reducing primary and secondary impact loads on the pelvis during side impact. Traffic Inj Prev 8:101–106 Tile M, Pennal GF (1980) Pelvic disruption: principles of management. Clin Orthop Relat Res 151:56–64 Torg JS (1985) Epidemiology, pathomechanics, and prevention of athletic injuries to the cervical spine. Med Sci Sports Exerc 17:295–303 Trollope ML, Stalnaker RL, JH ME, Frey CF (1973) The mechanism of injury in blunt abdominal trauma. J Trauma 13:962–970 Vandevord PJ, Bolander R, Sajja VSSS et al (2012) Mild neurotrauma indicates a range-specific pressure response to low level shock wave exposure. Ann Biomed Eng 40:227–236. https://doi. org/10.1007/s10439-011-0420-4 Viano DC, Khalil TB (1976) Investigation of impact response and fracture of the human femur by finite element modeling (No. 760773). SAE Technical Paper Viano DC, Lau VK (1983) Role of impact velocity and chest compression in thoracic injury. Aviat Sp Environ Med 54:16–21 Viano DC, Stalnaker RL (1980) Mechanisms of femoral fracture. J Biomech. https://doi.org/ 10.1016/0021-9290(80)90356-5

2398

B.D. Goodwin et al.

Viano DC, Lau IV, Asbury C et al (1989) Biomechanics of the human chest, abdomen, and pelvis in lateral impact. Accid Anal Prev 21:553–574. https://doi.org/10.1016/0001-4575(89)90070-5 Wang SC, Brede C, Lange D et al (2004) Gender differences in hip anatomy: possible implications for injury tolerance in frontal collisions. Annu Proc Assoc Adv Automot Med 48:287–301 Wang MC, Pintar F, Yoganandan N, Maiman DJ (2009) The continued burden of spine fractures after motor vehicle crashes. J Neurosurg Spine 10:86–92. https://doi.org/10.3171/SPI.2008.10.08279 Wilberger JE, Harris M, Diamond DL (1991) Acute subdural hematoma: morbidity, mortality, and operative timing. J Neurosurg 74:212–218. https://doi.org/10.3171/jns.1991.74.2.0212 Yang KH, Begeman PC (1996) A proposed role for facet joints in neck pain in low to moderate speed rear end impacts part I: biomechanics. In: 6th injury prevention through biomechanics symposium, Wayne State University, Detroit, MI, pp 59–63 Yoganandan NA, Sances Jr A, Maiman DJ et al (1986) Experimental spinal injuries with vertical impact. Spine (Phila Pa 1976) 11:855–860 Yoganandan N, Haffner M, Maiman DJ, Nichols H, Pintar FA, Jentzen J, . . . Sances A (1989) Epidemiology and injury biomechanics of motor vehicle related trauma to the human spine (No. 892438). SAE Technical Paper Yoganandan NA, Pintar FA, Sances Jr A et al (1991) Strength and kinematic response of dynamic cervical spine injuries. Spine (Phila Pa 1976) 16:S511–S517 Yoganandan NA, Pintar FA, Gennarelli TA, Maltese MR (2000) Patterns of abdominal injuries in frontal and side impacts. Ann Proc Assoc Adv Automot Med 44:17–36 Yoganandan N, Pintar FA, Maltese MR (2001) Biomechanics of abdominal injuries. Crit Rev™ Biomed Eng 29(2) Yoganandan NA, Pintar FA, Zhang J, Gennarelli TA (2007) Lateral impact injuries with side airbag deployments-A descriptive study. Accid Anal Prev 39:22–27. https://doi.org/10.1016/j.aap.2006.05.014 Yoganandan NA, Gennarelli TA, Zhang J et al (2009) Association of contact loading in diffuse axonal injuries from motor vehicle crashes. J Trauma 66:309–315. https://doi.org/10.1097/ TA.0b013e3181692104 Yoganandan NA, Humm JR, Pintar FA (2013a) Force corridors of post mortem human surrogates in oblique side impacts from sled tests. Ann Biomed Eng 41:2391–2398. https://doi.org/10.1007/ s10439-013-0847-x Yoganandan N, Stemper BD, Pintar FA et al (2013b) Cervical spine injury biomechanics: applications for under body blast loadings in military environments. Clin Biomech 28:602–609. https:// doi.org/10.1016/j.clinbiomech.2013.05.007 Yoganandan NA, Arun MWJ, Pintar FA, Szabo A (2014a) Optimized lower leg injury probability curves from postmortem human subject tests under axial impacts. Traffic Inj Prev 15(Suppl 1): S151–S156. https://doi.org/10.1080/15389588.2014.935357 Yoganandan N, Arun MWJ, Pintar FA (2014b) Normalizing and scaling of data to derive human response corridors from impact tests. J Biomech 47:1749–1756. https://doi.org/10.1016/j. jbiomech.2014.03.010 Yoganandan NA, Arun MWJ, Pintar FA, Banerjee A (2015a) Lower leg injury reference values and risk curves from survival analysis for male and female dummies: meta-analysis of postmortem human subject tests. Traffic Inj Prev 16(Suppl 1):S100–S107. https://doi.org/10.1080/15389588.2015.1015118 Yoganandan NA, Humm JR, Pintar FA et al (2015b) Oblique loading in post mortem human surrogates from vehicle tests using chestbands. Stapp Car Crash J 59:1–22 Young JW, Resnik CS (1990) Fracture of the pelvis: current concepts of classification. AJR Am J Roentgenol 155:1169–1175 Zhu F, Dong L, Jin X et al (2015) Testing and modeling the responses of Hybrid III crash-dummy lower extremity under high-speed vertical loading. Stapp Car Crash J 59:521–536

Vehicle Occupants in Traffic Accidents Garrett A. Mattos

Abstract

Occupant motion and injury response in motor vehicle crashes is dictated by the forces applied to the human body in combination with the relative motion between occupant and vehicle. These responses are complex and dependent on many factors including those relating to crash, vehicle, and occupant characteristics. The study of occupant response is an important step in improving vehicle design and crashworthiness. This chapter provides an overview of the important aspects of occupant response for the four main crash modes: frontal, side, rollover, and rear. The general characteristics of occupant kinematic response are discussed with respect to each crash mode. Typical injury patterns and mechanisms are identified and their relationship to occupant motion and interaction with the crash environment is highlighted. Keywords

Crash mode • Injury patterns • Occupant kinematics • Serious injury

Contents Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . State of the Art . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Frontal . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Injury Patterns . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Side . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Injury Patterns . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Rollover . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Injury Patterns . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

2400 2401 2401 2402 2405 2406 2408 2409

G.A. Mattos (*) Transport and Road Safety (TARS) Research Centre, University of New South Wales, Sydney, NSW, Australia e-mail: [email protected] # Springer International Publishing AG, part of Springer Nature 2018 B. Müller, S.I. Wolf (eds.), Handbook of Human Motion, https://doi.org/10.1007/978-3-319-14418-4_94

2399

2400

G.A. Mattos

Rear . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Injury Patterns . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Countermeasures . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Cross-References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

2411 2411 2413 2414 2414 2415

Introduction Traffic accidents generally involve one or more discrete impact events that result in global acceleration and local deformation of the vehicle structure. The specific characteristics of the vehicle response ultimately affect the motion of its occupants. This chapter will explore the relationships between the characteristics of vehicle crashes and occupant kinematics and injury patterns. Significant work has been accomplished, historically, to study occupant motion and predict injury response in motor vehicle crashes. Early testing programs utilized live animals or human volunteers to investigate occupant motion under subinjurious loading conditions. The use of post mortem human subjects (PMHS) to explore the response of the human body goes back to the nineteenth century and continues to this day. In the mid-1900s, anthropomorphic test devices (ATDs) began to be developed which could replicate human kinematic response under specific loading conditions. Recently, finite element (FE) modeling has been used to investigate the effects of a wide range of select crash and vehicle factors on occupant response outcomes by performing large amounts of simulations that could not practically be performed in a physical setting. Over the past 10 years, significant efforts have been made to create highly accurate FE models of the human body. These models allow for even greater control and variability of occupant anthropometry and morphology. Crash investigations and epidemiological studies provide the road safety community with the evidence base necessary to proceed with safety recommendations, design changes, or research projects to investigate the causes and mechanisms of pertinent injuries. These studies utilize large nationally representative databases or small subsets of very detailed crash reconstructions to identify the characteristics, and distribution of injuries occurring on the road and their findings translate into work focused on remedying deficiencies. This process is continuous and ongoing as the patterns of injury shift over time due to the changes in vehicle and road design and the implementation of improved safety features. The following will provide an overview of the current state of knowledge of the general characteristics of occupant kinematics, injury patterns, and injury sources for the four main crash configurations. Current capabilities in mitigating injury severity through the use of countermeasures and design choice will also be discussed. Much of which is provided below is available because of the aforementioned types of research that has been conducted and is available in the literature. For brevity and clarity, this chapter will focus on adult occupants of light passenger vehicles that are restrained with three-point seatbelts and involved in single-event crashes, unless otherwise noted. Injuries are described in general terms and their severity is defined

Vehicle Occupants in Traffic Accidents

2401

using the Abbreviated Injury Scale (AIS) which rates injury from 1 to 6 based on the risk of mortality (AAAM 2008).

State of the Art Improvements in experimental methods, human surrogates, and data acquisition technologies continue to provide opportunities to study occupant response with more detail than before. For example, with regard to experimental methods, crash testing facilities are continuing to push the envelope in crash severity. Frontal crashes are being conducted at reduced amounts of overlap and considerations are being made to evaluate the effect of automated emergency braking on occupant response prior to and during the crash. Recently, for the first time, full-scale rollover tests have been performed using PMHS to study occupant kinematics and injury response. Improvements in FE model design, such as the definition of active musculature, and in computing power have allowed for more complex and realistic replications of the occupant behavior. Physical testing has similarly improved with the development of advanced acquisition systems that allow for investigators to visualize and accurately measure occupant movement. These include high-speed x-ray systems that can record the motion of the skeletal system during impacts and multicamera motion tracking systems that provide three-dimensional occupant motion measurements that are accurate to a fraction of a millimeter.

Frontal Frontal crashes account for over half of all crashes involving passenger vehicles. They occur when the front plane of a forward-traveling vehicle strikes another vehicle or fixed object. These impacts can involve the entire face or just a fraction of overlap of the struck vehicle’s front plane. The vehicle and occupant kinematics are dependent on the amount of overlap, defined by the amount of the frontal plane of the struck vehicle that is impacted by the striking vehicle or object. In full frontal crashes, the vehicle decelerates almost purely along its longitudinal axis as the crushzone structures of the vehicle frame deform and absorb the crash energy, while generally preventing intrusion into the occupant compartment. This impact scenario results in the occupants moving directly forward and into their seatbelts and forward airbags. The combination of seatbelt, airbags, and crush tubes aims to allow the occupant to ride down the crash event under a survivable deceleration. As the amount of overlap decreases, greater lateral and rotational motions begin to come into play, especially when the impact is applied at an oblique angle. Smalloverlap crashes, in which less than one-quarter of the front structure is impacted, often do not engage the crush-zone structure of the vehicle frame. In this scenario, the vehicle sustains major damage that can extend from the forward wheel to the passenger footwell, and even to the B-pillar in severe crashes. While the vehicle

2402

G.A. Mattos

Fig. 1 Occupant and vehicle motion in small-overlap frontal crash

decelerates along its longitudinal axis, it also moves laterally and rotates, as shown in Fig. 1. The multidirectional vehicle response in small-overlap and oblique frontal crashes causes the occupants to move toward the impact on the front corner of the vehicle. Moving in this angled direction, forward and lateral relative to the vehicle, reduces the effectiveness of the airbags or prevents engagement altogether. A common phenomenon that occurs in frontal crashes, especially to rear-seated occupants, is known as submarining. Defined as insufficient restraint of the pelvis by the lap belt, submarining occurs when the lap belt fails to effectively engage the pelvis due to a combination of poor belt geometry and improper placement. For ideal performance, the lap belt is designed to be placed across the pelvis which is able to withstand the forces applied by the lap belt in a frontal crash. During submarining, the occupant’s pelvis moves forward while the lap belt rides up onto the abdomen. This can result in severe loading of the abdomen and lumbar spine, increased excursion of the lower extremities, and decreased displacement of the head.

Injury Patterns Injuries sustained in frontal crashes are primarily due to direct impact with the vehicle interior or loading on the body by the seatbelt or seatpan. The pattern of injuries is dependent on the crash scenario and occupant seat position, both relative to the location of impact and differentiated by front or rear location as demonstrated in Fig. 2. Small-overlap crashes tend to be more severe in terms of injury outcome, vehicle structural deformation, and occupant response. Children seated in front seats experienced significantly increased injury risk due to the mismatch between their size and injury tolerance and the performance of front seat restraints and airbags which are optimized for the 50th percentile male. The use of seatbelts greatly reduces the risk of injury to all body regions, especially the head and thorax, in low- and high-speed frontal crashes (Viano and Parenteau 2010).

Vehicle Occupants in Traffic Accidents

2403

Fig. 2 Distribution of seriously (AIS3+) injured occupants in large-overlap (left) and small-overlap (right) frontal impact crashes by body region sustaining serious injury (Hallman et al. 2011)

Head injury patterns differ between large- and small-overlap frontal crashes. With a decrease in the amount of overlap comes an increase in the likelihood of the head impacting components not protected by airbags, such as the A-pillar or center instrument panel. This results in a disparity of head injuries which is highlighted in field data. Occupants in large-overlap frontal crashes generally experience impacts between their head and the steering wheel or dash panel airbag directly in front of their seat. Impacts with the frontal airbag generally have a lower incidence of minor head injuries, such as bruising, and a greater incidence of moderate (AI2+) injuries involving a loss of consciousness (Hallman et al. 2011). The interaction between an occupant’s head and rigid interior components is more common in small-overlap frontal crashes and results in a higher rate of skull fractures and brain tissue damage. Moderate and severe head injuries are relatively rare for rear seat occupants. This is due to the combined effect of the vehicle’s structural performance and the rear seat occupant’s kinematics. Structural intrusion rarely reaches the rear seat occupants and, thus, limits their exposure to impacts with intruding objects. Secondly, rear seat occupants are more likely to experience submarining resulting in reduced upper torso and head excursion and limiting their exposure to direct impact. Spinal injuries, specifically those involving the posterior aspects of the cervical, i.e., upper, vertebra, occur at twice the rate in small-overlap versus large-overlap frontal crashes due to the oblique nature of the event (Hallman et al. 2011). The mechanism of these types of fractures is head impact with the A-pillar resulting in compression-extension of the cervical spine. Injuries to the lumbar, i.e., lower,

2404

G.A. Mattos

spine are likely caused by seatpan loading of the pelvis that compresses the spine (Pintar et al. 2014). Chest injuries on the lower end of the injury severity spectrum frequently sustained in frontal crashes include skin contusions and abrasions and sternum fractures directly related to seatbelt loading. The most common serious (AIS3+) thoracic injury, also most often attributed to belt loading, is a unilateral lung contusion in higher severity crashes. Small-overlap impacts have a higher rate of bilateral lung contusions than large-overlap crashes, while both crash configurations have similar rates of injuries involving multiple uni- and bilateral rib fractures (Hallman et al. 2011). Thoracic injury risk, specifically to the lungs and heart, is increased in severe impact events that result in the occupant bottoming-out the airbag and directly impacting the steering column (Chen and Gabler 2014). For rear-seated occupants in frontal crashes, the thorax is the most commonly injured body region and occurs almost exclusively due to seatbelt loading (Beck et al. 2016). The balance between practicality and performance in seatbelt design has resulted in almost universal use of three-point seatbelt systems that create asymmetric chest loading in frontal impacts that is suboptimal for thoracic injury mitigation. However, the overall benefits of seatbelt use far outweigh any minor deficiencies in their design. Unbelted drivers are significantly more likely than belted drivers to impact the steering wheel or front dash in a frontal crash. Such impacts greatly increase the risk of serious thoracic injuries such as aortic, heart, and liver lacerations (Chen and Gabler 2014). A factor that can increase the severity of front belt-restrained driver and passenger injuries in a frontal impact is the presence of an unrestrained rear seat occupant. The rear seat occupant can load the rear of the front seat increasing the deformation of the driver or passenger’s thorax. Abdomen injuries are relatively rare in frontal crashes for front seatbeltrestrained occupants and account for approximately 5% of all AIS3+ injuries. For frontal occupants, they are primarily caused by interaction between the occupant and the steering wheel or lap belt (Reichert et al. 2013). Rear seat occupants, especially adolescents, are at a much higher risk of sustaining abdomen injuries due to their increased risk of submarining. Occupants that experience submarining are subjected to loading of the abdomen by the lap belt which can result in upper abdominal injuries as well as concomitant fractures of the lower rib cage. Upper extremity injuries in frontal crashes commonly consist of fractures of the hand, radius, and ulna. The incidence of injury to the outboard extremity is increased in small-overlap crashes due to its interaction with intruding components and exposure to crash forces. Airbag deployment increases the risk of upper extremity injury, especially to the hand and forearm for drivers that have their hands on the steering wheel (Jernigan et al. 2005). The lower extremities constitute the most commonly injured body region in frontal crashes, and for large-overlap crashes, they are the most frequently injured body region at the moderate (AIS2+) and serious (AIS3+) levels. Moderate (AIS2+) lower extremity injuries sustained in frontal crashes most often involve the pelvis, femur, knee, and tibia in the form of fractures (Hallman et al. 2011). The oblique occupant kinematics resulting from small-overlap frontal crashes increases pelvis

Vehicle Occupants in Traffic Accidents

2405

loading while decreasing loading of the feet, as compared to large-overlap crashes. These injuries are sustained due to footpan loading caused by intrusion or from interaction between the knee and instrument panel as the occupant moves forward relative to the vehicle.

Side Side impact crashes are generally defined as planar crashes between two vehicles, or between one vehicle and a fixed object, in which the primary direction of force is within 45 of the lateral axis of the vehicle. While multivehicle crashes, in which the front of one vehicle impacts the side of another, are more common than those involving a single vehicle, i.e., fixed object collisions, the resulting occupant kinematics is similar. This crash configuration is dominated by a lateral acceleration of the vehicle, but generally also involves a longitudinal acceleration component. Typical impact scenarios include intersection crashes, in which both the struck and striking vehicles are moving, and single-vehicle crashes, in which loss of control events lead to a side impact with a fixed object. Occupants involved in side impact crashes can be classified by their seat position relative to the impacted side of the vehicle. Those seated on the struck-side are identified as near-side occupants, while those seated on the nonstruck side are identified as far-side occupants. The kinematic and injury response differs for near- and far-side occupants. The kinematic response of the near-side occupant is dominated by the intrusion of the adjacent door and B-pillar. The term “intrusion” is used here to describe relative motion between the vehicle’s nominal shape and its deformed shape. In the case of a side-impact against a stationary fixed object such as a tree, the intruding structure does not necessarily move relative to the global reference frame, e.g., surrounding environment, since it is also stationary against the fixed object. In a vehicle-tovehicle side impact crash, the intruding structure moves relative to both the global and the local vehicle reference frame. At impact, the struck side of the vehicle begins to intrude inward while the vehicle is simultaneously accelerated laterally. This results in a compounded relative lateral motion between the occupant and the vehicle interior, resulting in impact between the occupant and the intruding structure at essentially the preimpact speed. This impact typically begins with the pelvis and progresses with time upward through to the upper thorax and head (Yoganandan et al. 2015). Near-side (also referred to as struck side) occupant responses may be modulated by the deployment of torso side airbags and side air curtains and can be affected by the position of the upper extremity. As the event proceeds, the occupant will unload the vehicle’s struck side interior and begin to move toward the center of the vehicle. At this point, the occupant’s inboard pelvis will decelerate against the lap belt and center console, if one exists, and the upper torso may displace toward the far-side of the vehicle, depending on the performance of the shoulder belt.

2406

G.A. Mattos

Unlike the response of the near-side occupant, the initial response of the far-side occupant is driven by the vehicle’s lateral acceleration, rather than its deformation, and is heavily influenced by the geometry and performance of the restraint system. As the vehicle accelerates laterally, the occupant is displaced toward the struck side. The displacement of the pelvis is more effectively mitigated than the displacement of the torso. The geometry of the lap belt and the additional support provided by the center console, if one exists, limits pelvis displacement to 100–300 mm. The shoulder belt is the sole source of torso restraint, and its effectiveness depends on its ability to remain in position over the shoulder. Increased belt slip, defined as the amount the shoulder belt moves off of its nominal position, reduces the effectiveness of the restraint. Depending on the amount of belt slip, the head can displace laterally up to 732 mm (Forman et al. 2013). As the head and torso move across the center of the vehicle, they interact with the near-side occupant, if there is one, and the deformed vehicle structure. By the time the far-side occupant reaches the struck side, the deformation event is generally over. The upper extremities, which are essentially unrestrained, often flail toward the struck side.

Injury Patterns Injury characteristics, risk, and patterns differ between near- and far-side occupants as indicated in Fig. 3. The proximity of near-side occupants to the impact zone increases their risk of injury for a given crash due to the direct loading that occurs between their body and the door. Head injuries have been found to almost exclusively result from direct contact and most often with the roof rail, B-pillar, or striking vehicle/fixed object (Yoganandan et al. 2010). Serious head injuries more frequently involve the brain than the skull, which is likely the result of an airbag’s ability to mitigate skull fractures more effectively than brain injuries (Yoganandan et al. 2010). Depending on the severity of the crash and the performance of the side air curtain, if present, the near-side occupant’s head may come into direct contact with the impacting face of the striking vehicle or fixed object. The head of the far-side occupant is prone to impacting a wide range of interior components, often on the struck-side of the vehicle, and the adjacent occupant, if there is one, due to its relatively unrestrained nature (Gabler et al. 2005). Spine injuries are somewhat rare in side impact crashes, though they comprise a significant portion of severe injuries and are often cited as a main cause of death. They can result from head impacts which load the neck axially and laterally (McIntosh et al. 2007). Soft tissue and joint injuries are also possible under inertial loading in side impacts. Thoracic injuries sustained by far-side occupants typically result from impact with the seatback, seatbelt, and vehicle interior (Gabler et al. 2005). Far-side occupants are likely to impact near-side seats that have deformed into the far-side occupant’s trajectory. For near-side occupants, structural intrusion into the occupant space is believed to be a causal factor in producing chest injuries (Pintar

Vehicle Occupants in Traffic Accidents

2407

Fig. 3 Distribution of seriously (AIS3+) injured near-side (left) and far-side (right) occupants in side impact crashes by body region sustaining moderate serious (AIS3+) injury (Rupp et al. 2013)

et al. 2007). As the impact direction becomes more oblique, chest injury severity tends to increase due to the reduced injury tolerance of the ribcage to oblique loading. Rib fractures and lung contusions dominate the injury profile, with the vast majority of injuries located on the struck side of the body. While aortic injury is rare in side impact crashes, sustained by less than