Molecular Imaging: Principles and Practice 1607950057, 9781607950059

Citation preview

MOLECULAR IMAGING Principles and Practice Ralph We is s le de r, MD, P HD Professor of Radiology and Systems Biology Harvard Medical School Director, Center for Systems Biology Massachusetts General Hospital Boston, Massachusetts

Brian D. Ro s s , P HD Professor of Radiology and Biological Chemistry Co-Director Center for Molecular Imaging University of Michigan Medical School Ann Arbor, Michigan

Alnawaz Re he mtulla, P HD Ruth Tuttle Freeman Research Professor, Department of Radiation Oncology and Radiology Co-Director Center for Molecular Imaging University of Michigan Medical School Ann Arbor, Michigan

Sanjiv S. Gambhir, MD, P HD Virginia & D.K. Ludwig Professor of Radiology and Bioengineering Director, Molecular Imaging Program at Stanford (MIPS) Director, Canary Center for Cancer Early Detection at Stanford Chief, Division of Nuclear Medicine Stanford University Stanford, California

2010 PEOPLE’S MEDICAL PUBLISHING HOUSE–USA SHELTON, CONNECTICUT

People’s Medical Publishing House–USA 2 Enterprise Drive, Suite 509 Shelton, CT 06484 Tel: 203-402-0646 Fax: 203-402-0854 E-mail: [email protected] © 2010 Ralph Weissleder, Brian D. Ross, Alnawaz Rehemtulla, and Sanjiv S. Gambhir All rights reserved. Without limiting the rights under copyright reserved above, no part of this publication may be reproduced, stored in or introduced into a retrie val system, or transmitted, in any form or by any means (electronic, mechanical, photocopying, recording, or otherwise), without the prior written per mission of the publisher. 09 10 11 12 13/PMPH/9 8 7 6 5 4 3 2 1 ISBN-13 978-1-60795-005-9 ISBN-10 1-60795-005-7 Printed in China by People’s Medical Publishing House of China Copyeditor/Typesetter: diacriTech; Cover Designer: Mary McKeon Sales and Distribution Canada McGraw-Hill Ryerson Education Customer Care 300 Water St Whitby, Ontario L1N 9B6 Canada Tel: 1-800-565-5758 Fax: 1-800-463-5885 www.mcgrawhill.ca

United Kingdom, Europe, Middle East, Africa McGraw Hill Education Shoppenhangers Road Maidenhead Berkshire, SL6 2QL England Tel: 44-0-1628-502500 Fax: 44-0-1628-635895 www.mcgraw-hill.co.uk

Brazil Tecmedd Importadora e Distribuidora de Livros Ltda. Avenida Maurilio Biagi 2850 City Ribeirao, Rebeirao, Preto SP Brazil CEP: 14021-000 Tel: 0800-992236 Fax: 16-3993-9000 Email: [email protected]

Foreign Rights John Scott & Company International Publisher’s Agency P.O. Box 878 Kimberton, PA 19442 USA Tel: 610-827-1640 Fax: 610-827-1671

Singapore, Thailand, Philippines, Indonesia, Vietnam, Pacific Rim, Korea McGraw-Hill Education 60 Tuas Basin Link Singapore 638775 Tel: 65-6863-1580 Fax: 65-6862-3354 www.mcgraw-hill.com.sg

Japan United Publishers Services Limited 1-32-5 Higashi-Shinagawa Shinagawa-ku, Tokyo 140-0002 Japan Tel: 03-5479-7251 Fax: 03-5479-7307 Email: [email protected]

Australia, New Zealand Elsevier Australia Locked Bag 7500 Chatswood DC NSW 2067 Australia Tel: +61 (2) 9422-8500 Fax: +61 (2) 9422-8562 www.elsevier.com.au

India, Bangladesh, Pakistan, Sri Lanka, Malaysia CBS Publishers 4819/X1 Prahlad Street 24 Ansari Road, Darya Ganj, New Delhi-110002 India Tel: 91-11-23266861/67 Fax: 91-11-23266818 Email:[email protected] People’s Republic of China PMPH Bldg 3, 3 rd District Fangqunyuan, Fangzhuang Beijing 100078 P.R. China Tel: 8610-67653342 Fax: 8610-67691034 www.pmph.com

Notice: The authors and pub lisher have made every effort to ensure that the patient care recommended herein, including choice o f drugs and dr ug dosages, is in accord with the accepted standard and practice at the time of pub lication. However, since research and re gulation constantly change clinical standards, the reader is urged to check the product infor mation sheet included in the package of each dr ug, which includes recommended doses, w arnings, and contraindications. This is particularly important with new or infrequently used dr ugs. Any treatment regimen, particularly one involving medication, involves inherent risk that must be weighed on a case-by-case basis against the benef its anticipated. The reader is cautioned that the pur pose of this book is to i nform and enlighten; the infor mation contained herein is not intended as, and should not be emplo yed as, a substitute for indi vidual diagnosis and treatment.

Acknowledgments The editors would like to acknowledge the extraordinary contributions of Tania Cunningham, Judy Schwimmer, and Melissa Carlson in the preparation of this te xt. Collaboratively, they assumed the responsibility for or ganization and completion of the chapters of this te xt for which the editors are g rateful. RW BDR AR SSG

iii

Contents Preface . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . xi Contributors . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . xiii

1

General Principles of Molecular Imaging . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1

Sanjiv S. Gambhir PART I: MOLECULAR IMAGING TECHNOLOGIES 2

Imaging of Structure and Function with PET/CT . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10

David W. Townsend 3

PET/MRI. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 29

Marcus D. Seemann 4

SPECT and SPECT/CT. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 40

Brian F. Hutton, Freek J. Beekman 5

Principles of Micro X-ray Computed Tomography . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 54

Shaun S. Gleason, Michael J. Paulus, Dustin Osborne 6

Small Animal SPECT, SPECT/CT, and SPECT/MRI . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 76

Neal H. Clinthorne, Ling-Jian Meng 7

Instrumentation and Methods to Combine Small Animal PET with Other Imaging Modalities . . . . . 99

Craig S. Levin 8

Functional Imaging Using Bioluminescent Markers . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 118

Christopher H. Contag 9

Optical Multimodality Technologies . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 139

Arion F. Chatziioannou 10

Fiber Optic Fluorescence Imaging. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 147

Rabi Upadhyay, Umar Mahmood 11

Fluorescence Tomography . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 160

Vasilis Ntziachristos

v

vi

Contents

12

Endomicroscopy . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 165

Seok (Andy) H. Yun, Charles P. Lin 13

Intravital Microscopy . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 176

Thorsten R. Mempel 14

Diffuse Optical Tomography and Spectroscopy . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 197

David R. Busch, Britton Chance 15

Ultrasound . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 225

F. Stuart Foster, Kevin Cheung, Emmanuel Cherin 16

Molecular Photoacoustic Tomography . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 237

Lihong V. Wang 17

Optical Projection Tomography . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 244

James Sharpe 18

Potential Roles for Retrospective Registration in Molecular Imaging . . . . . . . . . . . . . . . . . . . . . . . 262

Charles R. Meyer, Hyunjin Park, Bing Ma, Boklye Kim, Peyton H. Bland PART II: CHEMISTRY OF MOLECULAR IMAGING 19

Chemistry of Molecular Imaging: An Overview . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 277

Silvio Aime, Giovanni Battista Giovenzana, Enzo Terreno 20

Radiochemistry of PET . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 304

Henry F. VanBrocklin 21

Radiochemistry of SPECT: Examples of

99m

Tc and

111

In Complexes . . . . . . . . . . . . . . . . . . . . . . . . 327

Hank F. Kung 22

Nanochemistry for Molecular Imaging. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 337

Yun Xing, Jianghong Rao 23

Newer Bioconjugation Methods . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 353

Claude F. Meares 24

Targeted Antibodies and Peptides . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 362

Michael R. Lewis, Cathy S. Cutler, Silvia S. Jurisson 25

Hyperpolarized

13

C Magnetic Resonance Imaging—Principles and Applications . . . . . . . . . . . . . . 377

Jan Henrik Ardenkjær-Larsen, Klaes Golman, Kevin M. Brindle 26

Magnetic Resonance Imaging Agents . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 389

Elisenda Rodriguez Vargas, John W. Chen

Contents

27

vii

Optical Imaging Agents . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 405

Scott A. Hilderbrand 28

Ultrasound Contrast Agents . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 425

Mark A. Borden, Shengping Qin, Katherine W. Ferrara 29

Multimodality Agents . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 445

Weibo Cai, Xiaoyuan (Shawn) Chen 30

“Click Chemistry”: Applications to Molecular Imaging . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 471

Neal K. Devaraj, Ralph Weissleder 31

The “One-Bead-One-Compound” Combinatorial Approach to Identifying Molecular Imaging Probes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 481

Ruiwu Liu, Olulanu H. Aina, Ekama Onof iok, Kit S. Lam 32

Chemical Biology Approaches to Molecular Imaging . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 497

Stanley Shaw 33

Theranostics: Agents for Diagnosis and Therapy . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 509

Jason R. McCarthy 34

Magnetic Nanoparticles . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 523

Andrew Tsourkas, Lee Josephson 35

Fluorocarbon Agents for Quantitative Multimodal Molecular Imaging and Targeted Therapeutics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 542

Samuel A. Wickline, Ralph P. Mason, Shelton D. Caruthers, Junjie Chen, Patrick M. Winter, Michael S. Hughes, Gregory M. Lanza 36

Aptamers for Molecular Imaging . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 574

Bertrand Tavitian 37

Nonclinical Product Developmental Strategies, Safety Considerations, and Toxicity Profiles of Medical Imaging and Radiopharmaceuticals Products . . . . . . . . . . . . . . . . . . . 589

Sunday Awe, Siham Biade, Sally J. Hargus, Tushar Kokate, Adebayo Laniyonu, Yanli Ouyang PART III: MOLECULAR IMAGING IN CELL & MOLECULAR BIOLOGY 38

Overview of Molecular and Cell Biology . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 604

Harvey R. Herschman, Hidevaldo B. Machado 39

Systems Biology . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 628

Gregory Foltz, Leroy Hood 40

Protein Engineering for Molecular Imaging . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 644

Anna M. Wu

viii

41

Contents

Phage Display for Imaging Agent Development. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 660

Kimberly A. Kelly 42

Molecular Imaging of Gene Therapy . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 673

Maria Veronica Lopez, Qiana L. Matthews, David T. Curiel, Anton V. Borovjagin 43

Developing Diagnostic and Therapeutic Viral Vectors . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 689

Khalid Shah 44

Cell Voyeurism Using Magnetic Resonance Imaging . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 703

Naser Muja, Christopher M. Long, Jeff W. M. Bulte 45

Tumor Vasculature . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 726

Ambros J. Beer, Gang Niu, Xiaoyuan (Shawn) Chen, Markus Schwaiger 46

Imaging Hypoxia . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 756

Ashley A. Manzoor, Hong Yuan, Gregory M. Palmer, Benjamin L. Viglianti, Mark W. Dewhirst 47

Molecular Imaging of Protein–Protein Interactions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 781

Tarik F. Massoud, Ramasamy Paulmurugan, Pritha Ray, Abhijit De, Carmel Chan, Hua Fan-Minogue, Sanjiv S. Gambhir 48

Fluorescence Readouts of Biochemistry in Live Cells and Organisms . . . . . . . . . . . . . . . . . . . . . . 808

Roger Y. Tsien 49

Imaging of Signaling Pathways . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 829

Mahaveer S. Bhojani, Brian D. Ross, Alnawaz Rehemtulla PART IV: APPLICATIONS OF MOLECULAR IMAGING

Oncology: 50

Molecular and Functional Imaging of the Tumor Microenvironment . . . . . . . . . . . . . . . . . . . . . . . . . 844

Kristine Glunde, Robert R. Gillies, Michal Neeman, Zaver M. Bhujwalla 51

Novel MR and PET Imaging in the RT Planning and Assesment of Response of Malignant Gliomas . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 864

Christina Tsien 52

PET Diagnosis and Response Monitoring in Oncology . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 875

Rodney J. Hicks, Richard L. Wahl 53

Magnetic Resonance Spectroscopy Treatment Response and Detection . . . . . . . . . . . . . . . . . . . . 896

Sarah J. Nelson, John Kurhanewicz, Daniel B. Vigneron 54

Diffusion MRI: A Biomarker for Early Cancer Treatment Response Assessment . . . . . . . . . . . . . . 912

Brian D. Ross, Craig J. Galbán, Charles R. Meyer, Alnawaz Rehemtulla, Thomas L. Chenevert

Contents

ix

Cardiovascular: 55

Myocardial Metabolism . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 925

Heinrich R. Schelbert 56

Congestive Heart Failure . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 941

Antti Saraste, Marcus R. Makowski, Stephan Nekolla, Markus Schwaiger 57

Molecular Imaging of Atherosclerosis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 960

Farouc A. Jaffer, Peter Libby 58

Thrombosis and Embolism . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 980

Andrea J. Wiethoff, Elmar Spuentrup, René M. Botnar 59

Molecular Imaging of Stem Cells in Myocardial Infarction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 989

David E. Sosnovik, Joseph C. Wu CNS: 60

Central Nervous System Molecular Imaging . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1011

Dima A. Hammoud, Andreas H. Jacobs, Martin G. Pomper 61

Neuroreceptor Imaging: Applications, Advances, and Limitations . . . . . . . . . . . . . . . . . . . . . . . . . 1035

Rikki N. Waterhouse, Thomas Lee Collier 62

PET and SPECT Imaging of Neurodegenerative Diseases . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1060

Brian J. Lopresti, Victor L. Villemagne, Chester A. Mathis AUTOIMMUNE/IMMUNOLOGY 63

Molecular Imaging of Autoimmune Diseses . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1089

Alberto Signore, Marco Chianelli 64

Rheumatoid Arthritis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1108

Lars Stangenberg, Umar Mahmood 65

Autoimmune Diabetes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1130

Diane Mathis, Jason Gaglia 66

Imaging in Asthma . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1147

Mikael J. Pittet, Filip K. Swirski PART V: MOLECULAR IMAGING IN DRUG EVALUATION 67

Molecular and Functional Imaging in Drug Development . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1161

Nicholas van Bruggen, Bernard M. Fine, Markus Rudin

x

68

Contents

PET Imaging in Cancer Clinical Trials . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1179

David A. Mankoff 69

Magnetic Resonance Imaging in Clinical Trials . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1192

Jeffrey L. Evelhoch, Douglas L. Arnold, Edward A. Ashton, Barry T. Peterson, Deborah Burstein, Derek L. G. Hill, Chun Yuan 70

Imaging of Gene Therapy: Basis and Clinical Trials . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1214

Andreas H. Jacobs, Yannic Waerzeggers, E. Antonio Chiocca, June-Key Chung, Juri Gelovani PART VI: OTHER 71

Visualization . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1247

David S. Paik 72

Quantification of Radiotracer Uptake into Tissue . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1258

Michael M. Graham 73

Mining Genomic Data for Molecular Imaging Targets . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1271

Sylvia K. Plevritis 74

Pharmacokinetic Modeling . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1284

Sung-Cheng (Henry) Huang 75

Cost-Effectiveness Analysis/Economics of Probe Development . . . . . . . . . . . . . . . . . . . . . . . . . . 1290

Daniel C. Sullivan, Paula M. Jacobs 76

The Regulatory and Reimbursement Process for Imaging Agents and Devices . . . . . . . . . . . . . . 1299

John M. Hoffman Index . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1326

Preface Over the last decade, the f ield of molecular imaging of living subjects has e volved considerab ly and has seen spectacular advances in chemistr y, engineering and biomedical applications. In a relatively short period of time, comprehensive molecular imaging centers ha ve been established in the US, Europe, and Asia and are increasingly integrated into basic sciences and translational networks. Ne w in vestigators, collaborators, and students drawn into this multidisciplinar y f ield ha ve often expressed the desire and need for an authoritati ve te xtbook. This te xtbook w as designed precisel y to f ill this need. We have been fortunate to recruit over 160 leading authors contributing 76 chapters. Given the multidisciplinar y nature of the f ield, the book is broken into six different sections. Part 1 (Molecular Imaging Technologies) summarizes the dif ferent macro, meso and microscopic imaging technolo gies currently available. Part 2 (Chemistry of Molecular Imaging) is dedicated to re viewing chemical approaches to imaging probe designs for dif ferent types of imaging technologies. This section also contains chapters on the emerging f ield of nanomaterials, chemical biolo gy, and probe design as w ell as signal amplif ication strate gies. Part 3 (Molecular Imaging in Cell and Molecular Biology) contains chapters dedicated to protein engineering, vectors, and pathways. Part 4 (Applications of Molecular Imaging) summarizes the abo ve adv ances in dif ferent clinical disease entities. P art 5 (Molecular Imaging in Drug E valuation) i s d edicated t o i maging i n d rug

development, and P art 6 pro vides chapters on computation, bioinfor matics, and modeling. We hope that the organization of this large amount of information is logical, and we have worked hard to avoid redundancies among chapters. We ha ve also done our best to encourage the use of figures to illustrate concepts and to provide numerous molecular imaging examples. We have striven to make Molecular Imaging the most authoritative and ef fective resource a vailable for the student and newcomer at all levels. If we have succeeded, it is because of the hard work, knowledge, and devotion of our authors and their responses to our critiques. We are g rateful to our institutions and depar tments for the continuing support that has enabled this work. We are mindful of our families and students w ho tolerated our “limited bandwidths” necessary for the timely completion of this edition. We are also g rateful to BC Deck er, PMPH–USA, and the entire administrative staff at our centers for k eeping us in line. In particular we acknowledge the hard work by Tania Cunningham, Melissa Carlson, and Judy Schwimmer . We are optimistic that this book will contribute to the continuing education of a v ariety of professionals and will ultimately aid in the care of our patients to whom all of our efforts are dedicated. Ralph Weissleder, MD, PHD Brian D. Ross, P HD Alnawaz Rehemtulla, P HD Sanjiv S. Gambhir, MD, PHD

xi

Contributors SILVIO AIME, PHD [19] Molecular Imaging Center University of Torino Torino, Italy

AMBROS J. BEER, MD [45] Technische Universität München Klinikum rechts der Isar Munich, Germany

OLULANU H. AINA, D.V.M., PHD [31] UC Davis Cancer Center Division of Hematology/Oncology University of California, Davis Sacramento, CA

MAHAVEER S. BHOJANI, PHD [49] Center for Molecular Imaging University of Michigan Medical School Ann Arbor, MI

JAN HENRIK ARDENKJÆR-LARSEN, PHD [25] The Grove Center GE Healthcare Amersham, UK DOUGLAS L. ARNOLD, MD [69] Montreal Neurological Institute Montreal, Quebec, Canada EDWARD A. ASHTON, PHD [69] VirtualScopics, Inc. Rochester, NY

ZAVER M. BHUJWALLA, PHD [50] The JHU ICMIC Program The Sidney Kimmel Comprehensive Cancer Center Johns Hopkins University School of Medicine Baltimore, MD SIHAM BIADE, PHARMD, PHD [37] Office of Oncology Drug Products Office of New Drugs Division of Medical Imaging & Hematology Products Center for Drug Evaluation and Research U.S. Food and Drug Administration Silver Spring, MD

SUNDAY AWE, PHD, MBA [37] Office of Oncology Drug Products Office of New Drugs Division of Medical Imaging & Hematology Products Center for Drug Evaluation and Research U.S. Food and Drug Administration Silver Spring, MD

PEYTON H. BLAND, PHD [18] Center for Molecular Imaging University of Michigan Medical School Ann Arbor, MI

FREEK J. BEEKMAN, PHD [4] Section Radiation Detection and Medical Imaging Delft University of Technology Delft, The Netherlands

ANTON V. BOROVJAGIN, PHD [42] Institute of Oral Health Research University of Alabama at Birmingham School of Dentistry Birmingham, AL

MARK A. BORDEN, PHD [28] Columbia University New York, NY

xiii

xiv

Contributors

RENÉ M. BOTNAR, PHD [58] Division of Imaging Sciences, King’s College London London, UK KEVIN M. BRINDLE, DPHIL [25] Cambridge Research Institute Li Ka Shing Center University of Cambridge Cambridge, UK JEFF W. M. B ULTE, PHD [44] Division of MR Research Institute for Cell Engineering The Johns Hopkins University School of Medicine Baltimore, MD DEBORAH BURSTEIN, PHD [69] Beth Israel Deaconess Medical Center Boston, MA DAVID R. BUSCH, MS [14] University of Pennsylvania Philadelphia, PA WEIBO CAI, PHD [29] University of Wisconsin-Madison Madison, WI SHELTON D. CARUTHERS, PHD [35] Washington University in Saint Louis and Philips Healthcare St. Louis, MO CARMEL CHAN, PHD [47] Molecular Imaging Program at Stanford (MIPS) Stanford University School of Medicine Stanford, CA BRITTON CHANCE, PHD, SCD (CANTAB.), MD (HON) [14] Eldridge Reeves Johnson University Philadelphia, PA ARION F. CHATZIIOANNOU, PHD [9] Crump Institute for Molecular Imaging David Geffen School of Medicine at UCLA Los Angeles, CA

JOHN W. CHEN, MD, PHD [26] Center for Systems Biology Center for Molecular Imaging Research Massachusetts General Hospital Harvard Medical School Boston, MA JUNJIE CHEN, PHD [35] Cardiovascular Division Washington University School of Medicine St. Louis, MO XIAOYUAN (SHAWN) CHEN, PHD [29, 45] National Institute of Biomedical Imaging and Bioengineering (NIBIB) National Institutes of Health (NIH) Bethesda, MD THOMAS L. CHENEVERT, PHD [54] Center for Molecular Imaging University of Michigan Medical School Ann Arbor, MI EMMANUEL CHERIN, PHD [15] Sunnybrook Health Sciences Centre Toronto, Ontario, Canada KEVIN CHEUNG, MD [15] McMaster University Medical Centre Hamilton, Ontario, Canada. MARCO CHIANELLI, MD, PHD [63] University Medical Center Groningen University of Groningen, The Netherlands Regina Apostolorum Hospital Albano, Rome, Italy E. ANTONIO CHIOCCA, MD, PHD [70] Dardinger Center for Neuro-oncology and Neurosciences James Cancer Hospital/Solove Research Institute Ohio State University Medical Center Columbus, OH JUNE-KEY CHUNG, MD, PHD [70] Seoul National University Hospital Seoul National University College of Medicine Seoul, Korea

Contributors

NEAL H. CLINTHORNE, PHD [6] Division of Nuclear Medicine University of Michigan Ann Arbor, MI THOMAS LEE COLLIER, PHD [61] Advion Biosystems Inc Louisville, TN CHRISTOPHER H. CONTAG, PHD [8] Molecular Imaging Program at Stanford (MIPS) Stanford University School of Medicine Stanford, CA

HUA FAN-MINOGUE, PHD [47] Molecular Imaging Program at Stanford (MIPS) Stanford University School of Medicine Stanford, CA KATHERINE W. FERRARA, PHD [28] University of California, Davis Davis, CA BERNARD M. FINE, MD, PHD [67] BioOncology Genentech, Inc. South San Francisco, CA

DAVID T. CURIEL, MD, PHD [42] Division of Human Gene Therapy Gene Therapy Center University of Alabama at Birmingham Birmingham, AL

GREGORY FOLTZ, MD [39] Swedish Neuroscience Institute Swedish Medical Center Seattle, WA

CATHY S. CUTLER, PHD [24] Research Reactor Center University of Missouri-Columbia Columbia, MO

F. STUART FOSTER, PHD [15] Sunnybrook Health Sciences Centre University of Toronto Toronto, Ontario, Canada

ABHIJIT DE, PHD [47] Advanced Centre for Treatment, Research and Education in Cancer (ACTREC) Tata Memorial Centre Navi Mumbai, Maharashtra India

JASON GAGLIA, MD [65] Section on Immunology and Immunogenetics Joslin Diabetes Center Harvard Medical School Boston, MA

NEAL K. DEVARAJ, PHD [30] Center for Systems Biology Center for Molecular Imaging Research Massachusetts General Hospital Harvard Medical School Boston, MA MARK W. DEWHIRST, DVM, PHD [46] Duke University Medical Center Durham, NC JEFFREY L. EVELHOCH, PHD [69] Imaging Research Merck Research Laboratories West Point, PA

CRAIG J. GALBÁN, PHD [54] Center for Molecular Imaging University of Michigan Medical School Ann Arbor, MI SANJIV S. GAMBHIR, MD, PHD [1, 47] Molecular Imaging Program at Stanford (MIPS) Canary Center for Cancer Early Detection at Stanford Division of Nuclear Medicine Stanford University Stanford, CA JURI G. GELOVANI, MD, PHD [70] M.D. Anderson Cancer Center University of Texas Houston, TX

xv

xvi

Contributors

ROBERT J. GILLIES, PHD [50] Advanced Research Institute for Biomedical Imaging, Arizona Cancer Center Tucson, AZ GIOVANNI BATTISTA GIOVENZANA, PHD [19] University of Piemonte Orientale “Amedeo Avogadro” Novara, Italy SHAUN S. GLEASON, PHD [5] Image Science and Machine Vision Group Oak Ridge National Laboratory Oak Ridge, TN KRISTINE GLUNDE, PHD [50] JHU ICMIC Program The Sidney Kimmel Comprehensive Cancer Center Molecular Imaging Program The Johns Hopkins University School of Medicine Baltimore, MD KLAES GOLMAN, PHD [25] Imagnia AB Malmö, Sweden

HARVEY R. HERSCHMAN, PHD [38] Molecular Biology Institute Jonsson Comprehensive Cancer Center David Geffen School of Medicine at UCLA University of California, Los Angeles Los Angeles, CA RODNEY J. HICKS, MD, FRACP [52] The Peter MacCallum Cancer Centre Centre for Molecular Imaging and Translational Oncology The University of Melbourne East Melbourne, Victoria, Australia SCOTT A. HILDERBRAND, PHD [27] Center for Systems Biology Center for Molecular Imaging Research Massachusetts General Hospital Harvard Medical School Boston, MA DEREK L. G. H ILL, PHD [69] IXICO, Ltd. London, UK

MICHAEL M. GRAHAM, MD., PHD [72] Division of Nuclear Medicine University of Iowa Iowa City, IA

JOHN M. HOFFMAN, MD [76] Division of Nuclear Medicine and Molecular Imaging Huntsman Cancer Institute University of Utah, School of Medicine Salt Lake City, UT

DIMA A. HAMMOUD, MD [60] Radiology and Imaging Sciences Division of Neuroradiology National Institutes of Health Clinical Center Bethesda, MD

LEROY HOOD, PHD [39] Institute for Systems Biology Seattle, WA

SALLY J. HARGUS, PHD [37] Office of Oncology Drug Products Office of New Drugs Division of Medical Imaging & Hematology Products Center for Drug Evaluation and Research U.S. Food and Drug Administration Silver Spring, MD

SUNG-CHENG (HENRY) HUANG, DSC [74] Crump Institute for Molecular Imaging David Geffen School of Medicine at UCLA University of California, Los Angeles Los Angeles, CA MICHAEL S. HUGHES, PHD [35] Cardiovascular Division Washington University Medical School St. Louis, MO

Contributors

BRIAN F. HUTTON, MD, PHD [4] UCL and UCLH NHS Foundation Trust Institute of Nuclear Medicine London, U.K. ANDREAS H. JACOBS, MD [60, 70] Laboratory for Gene Therapy and Molecular Imaging MPI for Neurological Research Klinikum Fulda gAG Köln, Germany PAULA M. JACOBS, PHD [75] SAIC-Frederick Division of Cancer Treatment and Diagnosis Cancer Imaging Program National Cancer Institute Bethesda, MD FAROUC A. JAFFER, MD, PHD [57] Center for Molecular Imaging Research Cardiovascular Research Center and Cardiology Division Massachusetts General Hospital Harvard Medical School Boston, MA LEE JOSEPHSON, PHD [34] Division of Nuclear Medicine Massachusetts General Hospital Harvard Medical School Boston, MA SILVIA S. JURISSON, PHD [24] Department of Biomedical Engineering University of Missouri-Columbia Columbia, MO KIMBERLY A. KELLY, PHD [41] University of Virginia Charlottesville, VA BOKLYE KIM, PHD [18] Center for Molecular Imaging University of Michigan Medical School Ann Arbor, MI

TUSHAR KOKATE, PHD [37] Office of Oncology Drug Products Office of New Drugs Division of Medical Imaging & Hematology Products Center for Drug Evaluation and Research U.S. Food and Drug Administration Silver Spring, MD HANK F. KUNG, PHD [21] University of Pennsylvania, Philadelphia, PA JOHN KURHANEWICZ, PHD [53] University of California San Francisco San Francisco, CA KIT S. LAM, MD, PHD [31] UC Davis Cancer Center Division of Hematology/Oncology University of California Davis Sacramento, CA ADEBAYO LANIYONU, PHD [37] Office of Oncology Drug Products Office of New Drugs Division of Medical Imaging & Hematology Products Center for Drug Evaluation and Research U.S. Food and Drug Administration Silver Spring, MD GREGORY M. LANZA, MD, PHD [35] Cardiovascular Division Washington University Medical School St. Louis, MO CRAIG S. LEVIN, PHD [7] Molecular Imaging Program at Stanford (MIPS) Division of Nuclear Medicine Stanford University School of Medicine Stanford, CA MICHAEL R. LEWIS, M.S., P HD [24] University of Missouri-Columbia Columbia, MO

xvii

xviii

Contributors

PETER LIBBY, MD [57] Division of Cardiovascular Medicine Brigham and Women’s Hospital Harvard Medical School Boston, MA

MARCUS R. MAKOWSKI, MD [56] Imaging Sciences Division King’s College London St Thomas' Hospital London, UK

CHARLES P. L IN, PHD [12] Center for Systems Biology Wellman Center for Photomedicine Massachusetts General Hospital Harvard Medical School Boston, MA

DAVID A. MANKOFF, MD, PHD [68] University of Washington Seattle Cancer Care Alliance Seattle, WA

RUIWU LIU, PHD [31] UC Davis Cancer Center Division of Hematology/Oncology University of California, Davis Sacramento, CA CHRISTOPHER M. LONG, BS [44] The Johns Hopkins University School of Medicine Baltimore, MD MARÍA VERÓNICA LÓPEZ, PHD [42] Laboratory of Molecular and Cellular Therapy Leloir Institute Capital Federal, Buenos Aires, Argentina BRIAN J. L OPRESTI, BSC [62] PET Center University of Pittsburgh School of Medicine Pittsburgh, PA BING MA, PHD [18] Center for Molecular Imaging University of Michigan Medical School Ann Arbor, MI HIDEVALDO B. MACHADO, PHD [38] David Geffen School of Medicine University of California, Los Angeles Los Angeles, CA UMAR MAHMOOD, MD, PHD [10, 64] Division of Nuclear Medicine Massachusetts General Hospital Harvard Medical School Boston, MA

ASHLEY A. MANZOOR, BS [46] Duke University Medical Center Durham, NC RALPH P. MASON, PHD [35] Division of Advanced Radiological Sciences UT Southwestern Medical Center Dallas, TX TARIK F. MASSOUD, MD, PHD [47] Cambridge Cancer Center University of Cambridge Cambridge, UK CHESTER A. MATHIS, PHD [62] PET Facility University of Pittsburgh School of Medicine Pittsburgh, PA DIANE MATHIS, PHD [65] Department of Pathology Harvard Medical School Boston, MA QIANA L. MATTHEWS, PHD [42] Division of Human Gene Therapy Gene Therapy Center University of Alabama at Birmingham Birmingham, AL JASON R. MCCARTHY, PHD [33] Center for Systems Biology Center for Molecular Imaging Research Massachusetts General Hospital Harvard Medical School Boston, MA

Contributors

CLAUDE F. MEARES, PHD [23] University of California, Davis Davis, CA THORSTEN R. MEMPEL, MD, PHD [13] Center for Immunology and Inflammatory Diseases Center for Systems Biology Massachusetts General Hospital Harvard Medical School Boston, MA LING-JIAN MENG, PHD [6] Division of Nuclear Medicine University of Illinois Urbana Champaign, IL CHARLES R. MEYER, PHD [18, 54] Center for Molecular Imaging University of Michigan Medical School Ann Arbor, MI NASER MUJA, PHD [44] Division of MR Research Institute for Cell Engineering The Johns Hopkins University School of Medicine Baltimore, MD MICHAL NEEMAN, PHD [50] Weizmann Institute of Science Rehovot, Israel STEPHAN G. NEKOLLA, PHD [56] Klinikum rechts der Isar Technischen Universität München München, Bavaria, Germany SARAH J. NELSON, PHD [53] University of California, Berkeley Berkeley, CA GANG NIU, PHD [45] Molecular Imaging Program at Stanford (MIPS) Stanford University School of Medicine Stanford, CA

VASILIS NTZIACHRISTOS, MSC, PHD [11] Institute for Biological and Medical Imaging Technical University of Munich and Helmholtz Center Munich, Germany EKAMA ONOFIOK, BS [31] UC Davis Cancer Center Division of Hematology/Oncology University of California, Davis Sacramento, CA DUSTIN OSBORNE, PHD [5] Siemens Molecular Imaging Knoxville, TN YANLI OUYANG, MD, PHD, DABT [37] Office of Oncology Drug Products Office of New Drugs Division of Medical Imaging & Hematology Products Center for Drug Evaluation and Research U.S. Food and Drug Administration Silver Spring, MD DAVID S. PAIK, PHD [71] Richard M. Lucas Center Stanford University School of Medicine Stanford, CA GREGORY M. PALMER, PHD [46] Duke University Medical Center Durham, NC HYUNJIN PARK, PHD [18] Center for Molecular Imaging University of Michigan Medical School Ann Arbor, MI RAMASAMY PAULMURUGAN, PHD [47] Molecular Imaging Program at Stanford (MIPS) Stanford University School of Medicine Stanford, CA MICHAEL J. PAULUS, PHD [5] Siemens Molecular Imaging Knoxville, TN

xix

xx

Contributors

BARRY T. PETERSON, PHD [69] Division of Physiological Measurements Pfizer Global Research and Development New London, CT MIKAEL J. PITTET, PHD [66] Center for Systems Biology Center for Molecular Imaging Research Massachusetts General Hospital Harvard Medical School Boston, MA SYLVIA K. PLEVRITIS, PHD [73] Radiological Sciences Laboratory Stanford University School of Medicine Stanford, CA MARTIN G. POMPER, MD, PHD [60] Johns Hopkins Medical Institutions Baltimore, MD SHENGPING QIN, PHD [28] University of California, Davis Davis, CA JIANGHONG RAO, PHD [22] Molecular Imaging Program at Stanford (MIPS) Stanford University School of Medicine Stanford, CA PRITHA RAY, PHD [47] Advanced Centre for Treatment, Research and Education in Cancer (ACTREC) Tata Memorial Hospital Kharghar, Navi Mumbai, Maharastra, India ALNAWAZ REHEMTULLA, PHD [49, 54] Center for Molecular Imaging University of Michigan Medical School Ann Arbor, MI BRIAN D. ROSS, PHD [49, 54] Center for Molecular Imaging University of Michigan Medical School Ann Arbor, MI

MARKUS RUDIN, PHD [67] Institute for Biomedical Engineering, Electical Engineering and Information Technology University of Zürich and ETH Zürich Zürich, ZH, Switzerland ANTTI SARASTE, MD, PHD [56] Turku University Hospital, Main Hospital Turku, Finland HEINRICH R. SCHELBERT, MD, PHD [55] David Geffen School of Medicine at UCLA University of California, Los Angeles Los Angeles, CA MARKUS SCHWAIGER MD, PHD [3, 45, 56] Klinikum rechts der Isar Technischen Universität München München, Bavaria, Germany MARCUS D. SEEMANN, MD [3] Institute of Radiology and Nuclear Medicine University of Bochum, St. Josef-Hospital Bochum, Germany KHALID SHAH, PHD [43] Center for Molecular Imaging Research Massachusetts General Hospital Harvard Medical School Boston, MA JAMES SHARPE, PHD [17] Catalan Institute for Advanced Research and Education (ICREA) EMBL-CRG Systems Biology Unit Centre for Genomic Regulation, UPF Barcelona, Spain STANLEY SHAW, MD, PHD [32] Center for Systems Biology Massachusetts General Hospital Harvard Medical School Boston, MA

Contributors

ALBERTO SIGNORE, MD, PHD [63] II Faculty of Medicine University of Rome “Sapienza” Roma, Italy

DAVID W. TOWNSEND, PHD [2] PET and SPECT Development Singapore Bioimaging Consortium Singapore

DAVID E. SOSNOVIK, MD [59] Center for Molecular Imaging Research Division of Cardiology Massachusetts General Hospital Harvard Medical School Charlestown, MA

CHRISTINA TSIEN, MD [51] University of Michigan Medical Center Ann Arbor, MI

ELMAR SPUENTRUP, MD [58] University Hospital of Cologne Cologne, Germany LARS STANGENBERG, MD [64] Department of Surgery Massachusetts General Hospital Harvard Medical School Boston, MA DANIEL C. SULLIVAN, MD [75] Duke Comprehensive Cancer Center Duke University Medical Center Durham, NC FILIP K. SWIRSKI, PHD [66] Center for Systems Biology Center for Molecular Imaging Research Massachusetts General Hospital Harvard Medical School Boston, MA BERTRAND TAVITIAN, MD, PHD [36] CEA, Institut d’Imagerie Biomédicale Service Hospitalier Frédéric Joliot Laboratoire d’Imagerie Moléculaire Expérimentale and INSERM U803 Laboratoire d’Imagerie de l’Expression des Gènes Orsay, France ENZO TERRENO, PHD [19] Molecular Imaging Center University of Torino Torino, Italy

ROGER Y. TSIEN, PHD [48] Howard Hughes Medical Institute University of California, San Diego La Jolla, CA ANDREW TSOURKAS, PHD [34] School of Engineering and Applied Science University of Pennsylvania Philadelphia, PA RABI UPADHYAY, BS [10] Center for Systems Biology Center for Molecular Imaging Research Massachusetts General Hospital Harvard Medical School Boston, MA HENRY F. VANBROCKLIN, PHD [20] Radiopharmaceutical Research University of California San Francisco San Francisco, CA NICHOLAS VAN BRUGGEN, PHD [67] Biomedical Imaging Genentech, Inc. South San Francisco, CA ELISENDA RODRIGUEZ VARGAS, PHD [26] Center for Systems Biology Center for Molecular Imaging Research Massachusetts General Hospital Harvard Medical School Boston, MA BENJAMIN L. VIGLIANTI, PHD [46] Duke University Medical Center Durham, NC

xxi

xxii

Contributors

DANIEL B. VIGNERON, PHD [53] University of California San Francisco San Francisco, CA VICTOR L. VILLEMAGNE, MD [62] Austin Health The Mental Health Research Institute of Victoria University of Melbourne Melbourne, Victoria, Australia YANNIC WAERZEGGERS, MD [70] Laboratory for Gene Therapy and Molecular Imaging Max Planck Institute for Neurological Research Klaus-Joachim-Zülch Laboratories of the Max Planck Society Cologne, Germany RICHARD L. WAHL, MD, FACR [52] Nuclear Medicine/PET New Technology and Business Development Johns Hopkins University School of Medicine Baltimore, MD LIHONG V. WANG, PHD [16] Optical Imaging Laboratory Washington University St. Louis, MO RIKKI N. WATERHOUSE, PHD [61] Head, Cancer Imaging and Radiochemistr y Advanced Technology Global Pharmaceutical Discovery Abbott Abbott Park, IL RALPH WEISSLEDER, MD, PHD [30] Center for Systems Biology Center for Molecular Imaging Research Massachusetts General Hospital Harvard Medical School Boston, MA SAMUEL A. WICKLINE, MD [35] Cardiovascular Division Washington University School of Medicine St. Louis, MO

ANDREA J. WIETHOFF, PHD [58] Division of Imaging Sciences King’s College London Philips Healthcare London, UK PATRICK M. WINTER, PHD [35] Cardiovascular Division Washington University School of Medicine St. Louis, MO ANNA M. WU, PHD [40] Crump Institute for Molecular Imaging David Geffen School of Medicine at UCLA University of California, Los Angeles Los Angeles, CA JOSEPH C. WU, MD, PHD [59] Molecular Imaging Program at Stanford (MIPS) Stanford University School of Medicine Stanford, CA YUN XING, PHD [22] School of Engineering University of Dayton Dayton, OH CHUN YUAN, PHD [69] Vascular Imaging Laboratory University of Washington Seattle, WA HONG YUAN, PHD [46] Duke University Medical Center Durham, NC SEOK H. (ANDY) YUN, PHD [12] Wellman Center for Photomedicine Massachusetts General Hospital Harvard Medical Center Boston, MA

1 GENERAL PRINCIPLES

OF

MOLECULAR IMAGING

SANJIV S. GAMBHIR, MD, PHD

Molecular imaging (MI) of living subjects is an emerging field that aims to study molecular and cellular e vents in the intact living animal and human. These events can be as simple as location(s) of a specif ic population of cells or le vels of a gi ven protein receptor on the surf ace of cells. In addition, it is possib le to study more comple x events such as the interaction of tw o intracellular proteins, cellular metabolic flux, or transcription of a set of genes w hen one cell type comes into contact with another cell type. In contrast to molecular processes studied in intact cells outside the living subject (eg, with light microscopy techniques), it is much more dif ficult to longitudinally study the same processes in intact li ving subjects where most cells are located within deep tissues. It is the hope of many MI researchers that most of biology will eventually be ab le to be studied in the intact li ving subject instead of ha ving to remove tissues/cells for further analysis, as is no w commonly done. This will allow the study of simple and comple x processes w hile cells reside in their nati ve en vironment with all molecular feedback loops fully intact. The reasons for monitoring/imaging v arious molecular targets are usuall y related to characterizing a disease process that ma y cor relate with concentrations of one or more of these molecular tar gets. For e xample, the presence of relatively high levels of the somatostatin receptor type 2 (located on cell membranes) in the lungs of a subject may be indicative of the presence of cancer cells in the lung. This may then help guide medical management of the subject in w hich such a molecular signal is detected. Another v ery impor tant reason to study v arious cellular/molecular targets is to help dissect complex underlying biology. F or e xample, one might be ab le to study the migration of a specif ic subset of T-cells (or T l ymphocytes) into a tumor and subsequent acti vation of these T-cells by a T-cell receptor to better understand the details of the interaction of the tumor with the immune system.

An additional impor tant application of MI is in the process of dr ug disco very and v alidation, as w ell as in predicting and monitoring response to v arious types of therapy (see Chapters 51, “Novel MR and PET Imaging in the RT Planning and Assessment of Response of Malignant Gliomas,” 52, “PET Diagnosis and Response Monitoring in Oncolo gy,” 53, “MRS Treatment Response and Detection,” 54, “Dif fusion MRI: A Biomark er for Earl y Cancer Treatment Response, ” and 67, “Molecular and Functional Imaging in Dr ug Development”). Most of MI is perfor med b y introducing a molecular probe (e g, a small molecule) into the living subject, as will be detailed later. Since MI probes are often relati ves of pharmaceuticals and/or interact with the same molecular tar get(s) there are many important links between MI and phar macology (see Chapter 74, “Phar macokinetic Modeling”). The safety of subjects w hen using MI probes is of paramount impor tance. Ideally, one w ould not ha ve to introduce MI probes into the subject to quantitate the molecular targets of interest. However, since in most cases one must f irst introduce the MI probe, it is critical to ensure that this probe does not signif icantly per turb the living subject and certainly that it is not acutel y or chronically to xic to the subject (see Chapter 37, “Nonclinical Product Developmental Strategies, Safety Considerations and Toxicity Profiles of Medical Imaging and Radiopharmaceuticals Products”). MI often ser ves as an impor tant complement to more con ventional anatomical imaging (eg, computed tomo graphy [CT]). Together these techniques can help to impro ve disease management and understanding of biological processes of interest. The process of de veloping new strategies/assays for MI is illustrated in Figure 1 and will be referred to as the MI research chain. This is an iterati ve process, and if successful, leads to a useful MI assa y. Although what is shown is the translation of the assay for eventual clinical application, it is not al ways the case that e very assay is 1

2

MOLECULAR IMAGING: PRINCIPLES AND PRACTICE

Molecular Targets

Gene expression profiling Proteomics Systems biology

Chemistry

Organic synthesis of probes Combinatorial chemistry Radiochemistry Material science/Nanochemistry

Small/Large animal Models

Imaging

Molecular/Cell Biology

Cell delivery issues Cell efflux issues Cell-Cell interactions Spatial localization of target

Clinical Imaging Biodistribution Animal models of disease Probe Stability/Pharmacokinetics

Computer Modeling

Human pharmacokinetics/ Safety Clinical trials Outcomes assessment Drug efficacy

Multimodality animal imaging Pharmacokinetics Time-activity curves Individual variation studies Image Reconstruction/Quantitation Mathematical modeling Statistical analysis 3-D visualization

Figure 1. Molecular imaging research chain. The process of going from molecular targets to clinical molecular imaging is shown. This research chain is fundamental to the field and is critical for its success. Not all research in the field is intended for clinical translation and in some cases assay development terminates at the computer modeling stage. The process iterates based on lessons learned in a given application, as shown with the dotted line.

intended for clinical applications. Some assa ys may be only intended for animal models to ans wer fundamental biological questions. Man y MI assa ys are already in existence and y et more are being de veloped. Many MI probes and their uses are limited , and y et others ha ve found signif icant clinical use. Although the timeline for developing an assay and translating it for clinical use can be 3 to 7 years, this timeline is decreasing and hopefully can approach 1 to 3 y ears in the near future. To achieve a few well-used assays, the f ield has to be prepared to fail in its attempts at development and evolve from these failures. Each of the components of the MI research chain is reviewed next. The choice of molecular tar get(s) dri ves the entire research chain. Ideal molecular tar gets are present in multiple copies per cell. These are usually protein targets (100 to 1 million copies per cell) but can also be

messenger ribonucleic acid (mRN A, 50 to 1,000 copies per cell). Deo xyribonucleic acid (DN A) is not used because of both its low copy number (which makes it difficult to produce sufficient specific signal), and because imaging of DNA would not allow one to deter mine if a particular gene is being expressed, only that it is present. Clearly, the abundance and specif icity of the tar get for the disease process under study is critical to make the MI assay successful. One fundamental issue for MI is knowing w hich molecular tar gets are rele vant to study for a given set of biolo gical questions or for a gi ven disease management prob lem. In f act, e ven if MI w as ab le to interrogate concentrations of e very potential molecular target and various other events, it would be still difficult to kno w w hich e vents to actuall y monitor/image. It is currently not possible to highly-multiplex (detect multiple molecular targets of interest simultaneously) and as a

General Principles of Molecular Ima ging

HO O HO

OH HO [18]F

A

B

Figure 2. Anatomy of a molecular imaging probe. Most molecular imaging probes are made up of three primary components as shown in the left panel. These include (1) a chemical component that provides specificity for the molecular target of interest (shown in blue), (2) a component that will provide a signal which can be detected (shown in red), and (3) a linker (shown in gray scale), which may or may not be needed. Note that the three components are not shown to scale. Sometimes, the chemical specificity component is much larger than the signaling component (eg, a molecular imaging probe for PET) and at other times, it is much smaller than the signaling component (eg, a targeted microbubble molecular imaging probe in which the signaling component is a relatively large gas-filled microbubble for use with ultrasound). A specific molecular imaging probe is shown in the right panel. This molecular imaging probe is one of the most clinically successful MI probes to date ([18F]-2-fluoro-2-deoxyglucose [FDG]) and is discussed in detail in Chapter 20, “Radiochemistry of PET.” This MI probe has chemical specificity for the glucose transporters which transport it into cells as well as for the hexokinase type II enzyme which phosphorylates the molecule in the sixth position leaving it negatively charged and unable to diffuse out of cells. Furthermore, the phosphorylated FDG (FDG-6PO4) is unable to be further metabolized because it is not recognized by enzymes that normally metabolize glucose-6PO4. The signaling component is the fluorine-18, which is proton rich and decays by production of a positron. The positron annihilates with a nearby electron to produce two 511 keV gamma rays, which have a good probability of making it through tissues to be detected outside the subject.

result, selection of molecular tar gets is critical. Multiplexing is cur rently limited to 3 to 5 maximum molecular targets. If one relaxes the necessity to simultaneously measure the tar gets of interest, then one can perfor m serial measurements (e g, dail y) to deter mine le vels of additional targets. In most strategies, a MI probe must first be introduced into the li ving subject (e g, b y injection into the b loodstream). The “anatomy” of such a MI probe is sho wn in Figure 2 and is usually composed of a chemically specific component that interacts with the intended molecular target (eg, a protein), a signaling component that produces a signal that can hopefull y be detected , and a link er between the two components. MI probes are referred to by many names such as MI agents, imaging agents, tracers, radiopharmaceuticals, radiotracers, acti vatable or smar t probes, constitutively active probes, molecular detecti ves, and molecular spies. Even though a MI probe does provide “contrast” in an image, allo wing one to “see” molecular targets of interest relati ve to a backg round, the ter m

3

“contrast agent” is a poor ter m because most contrast agents refer to nonspecif ic agents that ma y ha ve poorl y defined molecular targets. MI probes can be broadl y categorized as constituti vely acti ve probes or acti vatable probes as illustrated in F igure 3. Radiolabeled probes produce their signal constantl y throughout as radioacti ve decay occurs and are therefore constituti vely acti ve. An activatable probe has the advantage of not producing signal until it interacts with its intended tar get(s), thus leading to a lo w backg round signal. MI probes are adv antageous because the y pro vide molecular specif icity, but the y are also the Achilles’ heel for the MI f ield because of re gulatory issues sur rounding the introduction of a no vel or foreign probe for human applications. It is unfor tunate that we currently do not have methods of “listening-in” on molecular events without first introducing MI probes. This would allow imaging of molecular/cellular processes without de veloping specif ic MI probes and w ould mark edly ease the re gulatory issues. There are a fe w exceptions to this limitation, such as magnetic resonance spectroscop y (MRS), w hich can “listen-in” on a fe w specif ic endo genous molecules (e g, choline) as discussed in Chapter 53, “MRS Treatment Response and Detection. ” Pro gress in this direction without compromising important parameters such as spatial and temporal resolution w ould be gamechanging for the f ield. Fur thermore, as the scientif ic understanding of nor mal and patholo gical processes continues to e volve, leading to a need to image dif ferent molecular tar gets, it is impor tant to ha ve strate gies that can allow development of MI probes for “ne w” molecular targets. Generalizab le strate gies such as engineered antibodies (see Chapter 40, “Protein Engineering for Molecular Imaging”) illustrate ho w MI probes can be de veloped relatively quickly as ne w molecular tar gets of interest are discovered. Chemistry helps drive the MI field as it is pivotal to developing the MI probes. Small molecules, peptides, aptamers, engineered proteins, and e ven more comple x nanoparticles are all possib le MI probes. The MI probe developmental chemistr y can often tak e many months. Ideally, the optimized chemistr y w ould allo w rapid synthesis of the MI probe with high purity so that it can be synthesized at all laborator y, clinical, and research sites that will each perform the imaging. In the cases of human translation, it is impor tant to also perfor m synthesis of the MI probe under good manuf acturing practice (GMP) guidelines. One may wonder while reading the chapters in this te xtbook about the potential to treat the disease of interest using the same probe that is being used to image the disease. If the MI probe can be constructed with a therapeutic component, in addition to the

4

MOLECULAR IMAGING: PRINCIPLES AND PRACTICE

Constitutively active probes

versus

Time

Activatable probes

Time

Figure 3. Two broad categories of molecular imaging probes. MI probes (red circle linked to a blue partial circle) are shown interacting with their intended blue circle targets, which in this case are located intracellularly and on the cell membrane. Other potential targets (for other MI probes specific for them) are shown as triangles and squares. Constitutively active probes (for PET/SPECT imaging and autoradiography) produce continuous signal, before and after interacting with their target(s), through the decay of the radioisotope. A time delay between injection of the probe and imaging helps to clear the nontrapped probes reducing background signal. Shown in the bottom left panel are some MI probes that produce signal even though they have not interacted with their blue circle target. Activatable probes produce signal only when they interact with their target(s) (eg, near-infrared fluorescent probes for optical imaging). These activatable probes can be thought of as either “light switches” (as in optical probes or hyperpolarized C-13 MR probes that turn from “off to on”) or “dimmer switches” (as in gadolinium MR probes that turn from “dull to bright”). A time delay between injection and imaging helps to achieve sufficient levels of activated probe at the target site where targets are typically enzymes. Background signal is inherently low in this category of MI probe because signal is only produced when the MI probe interacts with its intended targets. Shown in the bottom right panel are some MI probes that have not interacted with their blue circle targets and therefore do not produce signals.

signaling component, possib le link er, and the component specific for the molecular target(s) of interest, then the resulting “theranostic” could be used. In f act, in the field of Nuclear Medicine, the same probe that is used for imaging is often slightl y modif ied (by changing the radioisotope) and then ser ves as a therapeutic (e g, for tumor kill). Also, as discussed later on, the chemical development of multimodality probes enab les the performance of imaging on multiple imaging platfor ms that can span from small-to-lar ge animals and e ven humans. The MI probes can be tested in vitro (with extracts of cells) and eventually in intact cells in cell culture. Testing the probes with intact cells helps to better understand their ability to tra verse the cell membrane, the time

involved for tar geting and clearance from cells, and the potential for nonspecif ic interactions that would lead to increased background signal. Furthermore, use of standard molecular biology techniques allows the modulation of le vels of molecular tar gets (b y transducing the gene encoding for the protein tar get) to test the relationship between MI probe signal and le vels of molecular tar get. Additionally, repor ter genes can be tested and v alidated in cell culture and for such strate gies cell culture testing is critical (see also Chapters 38, “Overview of Molecular and Cell Biolo gy,” and 47, “Molecular Imaging of Protein-Protein Interactions”). Although testing in cell culture is useful, it does not help ans wer se veral critical issues for de veloping an MI assa y. These include i) ho w to deliver sufficient amounts of MI probe to the cells of

General Principles of Molecular Ima ging

interest when the cells reside deep within a li ving organism, ii) if the MI probe does not interact with the tar gets of interest, can it be cleared to reduce background signal, and iii) biodistribution and phar macokinetic issues related to deli very and clearance of the MI probe. F or these and other reasons, the next step is usually to test the MI probe in animal models. Small animal models are con venient to test these and other issues due to their relati vely low cost, highthroughput, ease of handling, etc. Mouse models can be set up with the molecular targets of interest. This can be done by implanting cells with the target(s) of interest or studying mice that spontaneously or otherwise develop the disease e xhibiting the cell/molecular tar get(s) of interest. Alternatively, murine models can be developed by introducing or deleting gene(s) of interest as is done in transgenic/knock-in mice and knock-out mice, respectively. Additionally, the le vels of the molecular target of interest can be manipulated using cur rent or novel pharmaceuticals. Sometimes, small animal models cannot properly reflect human disease so that lar ge animal (eg, porcine) models are used. For example, for many cardiovascular diseases, the porcine model ma y be more appropriate than a rodent model. F or studying neurological diseases, the primate brain ma y be the most appropriate. Also, high-resolution imaging of large animals (e g, rabbits) is also adv antageous for identifying hetero geneity of molecular e xpression within diseased tissue (e g, atherosclerotic plaques) which ma y be used for better e valuation of the MI probe’s ability to characterize disease in humans. Ev en if no animal model is fully reflective of the human disease, the use of animal models is also needed to test for potential to xicity of the MI probe in a li ving subject prior to translation to human studies. Additionally, in the case of MI probes that use a radioisotope for signal production, there is a need to deter mine radiation dosimetry to v arious or gans for e xtrapolation of the dosimetry to humans. The next key step is to test the animal model with the MI probe w hile imaging with the appropriate imaging instrument(s) of choice. Usuall y, the animal has to be anesthetized in order to be imaged unless the anesthesia may interfere with the process being studied or the temporal resolution (how quickly the imaging instr ument can “tak e a picture”) of the imaging instr ument(s) is extremely f ast. This allo ws one to tr uly e xamine the pharmacokinetics and biodistribution of the MI probe. It also allo ws one to optimize routes of administration (usually intravenous), mass of injected probe to achie ve desired signal, signal-to-backg round ratio, time to image

5

the animal, as w ell as se veral other impor tant f actors. Sometimes, the animal has to be imaged both for location(s) and concentrations of the MI probe as w ell as for anatomical information. By combining both the anatomical and molecular images, one can better understand underlying biological/pathological processes of interest. It would be ideal to ha ve a technology that could deter mine very lo w le vels of molecular tar get concentration (eg, picomolar or 10 –12 M), have the ability to follo w just a few cells instead of thousands to millions, have high spatial resolution (sub-millimeter), high temporal resolution (eg, milli-second), be lo w-cost, offer high-throughput, be fully quantitati ve, allo w inter rogation at all depths throughout the subject, and allow measurement of molecular tar gets located an ywhere in the subject and in an y location within the cell. Because no such single ideal imaging technology exists, it is often the case that based on the biolo gical questions being ask ed, the appropriate MI technology must be selected. It is thus impor tant that advocates of each MI technolo gy properl y acknowledge advantages and limitations of a gi ven strate gy and also acknowledge that sometimes a combination of technologies or an alter nate technolo gy is best suited for the biological/clinical question at hand. The next step is to look at quantitati ve issues in the animal model using the images acquired. Quantitation can be of v arious degrees ranging from minimal to absolute. One can attempt to quantitate the amount of MI probe at various sites including the tar get tissues and/or relate the signal from the MI probe back to absolute le vels of the molecular tar get(s) of interest. To perfor m absolute quantitation, it is usuall y necessar y to perfor m dynamic imaging in w hich serial images are tak en to characterize changes in the location(s) of the MI probe signal to produce time-activity curves (see also Chapter 4, “SPECT and SPECT/CT”). A key issue in almost e very molecular image is the requirement that le vels of the signal be ab le to be related back to levels, or even activity in the case of activatable probes, of the molecular target(s) of interest. In addition to quantitation, visualization strate gies can also be de veloped and v alidated so that one can displa y the molecular images in a way that makes it easier to interpret the findings. In man y but not all cases, the f inal steps are to translate the de veloped MI strate gy to clinical applications. This requires obtaining approval from appropriate agencies such as the F ood and Dr ug Administration (FDA) and the local inter nal review board (IRB). Radiation dosimetr y studies (if a radioisotope is used) and toxicology studies in preclinical models are therefore critical. The hope in the initial pilot clinical imaging

6

MOLECULAR IMAGING: PRINCIPLES AND PRACTICE

A

BB

C

D

Figure 4. From molecule to molecular image. The imaging of the location(s) and concentration of MI target(s) of interest starts with subject preparation and introduction of the MI probe into the subject (Panel A). An exception is in MRI where it is typically necessary to perform a pre-scan prior to injecting the agent to account for the “background” tissue signal which can vary greatly within a particular diseased tissue. However, ways of overcoming this limitation are being developed (eg, magnetic particle imaging (MPI)). The MI probe is usually introduced intravenously as shown, but can be introduced by almost any route (eg, orally). The total mass of MI probe introduced is a key issue and in some cases several different probes can be simultaneously injected (each intended for a different target). The next step is to allow the MI probe time to distribute throughout the living subject (Panel B). This allows enough time for the MI probe to reach its target(s) and if needed to clear from tissues where there is no target. This process can be done while the subject is within the imaging instrument or prior to placing the subject in the imaging instrument. Shown are the MI probes reaching their intracellular targets (blue spheres) where many but not all MI probes have bound. Subsequently, the detection of signal from the MI probes distributed throughout the body can now be performed (while the MI probe continues to clear). This may require placing the subject in an external field (Panel C) and/or exciting the distributed MI probes with energy (as in fluorescence optical imaging) so that they may provide a detectable signal. In the example shown, the probes are radioactive and do not require an external field, although one is being produced (as shown by the gray transducer emitting waves) to serve as a reminder of what is needed for different (eg, fluorescence optical imaging) strategies. The production of signal by the MI probes are then detected by one or more detectors placed around the subject. After collection of signal for a period of time, statistically sufficient data may be available to develop projection images (without depth information) and/or tomographic images (with full depth information) to visually represent the distribution of the MI probe throughout the field-of-view studied (Panel D). In the illustrated example, coronal PET images are shown and represent MI probe signal originating from a small lung tumor, the brain, and from the MI probe cleared by the kidneys into the bladder. In this case, low levels of MI signal throughout the rest of the body are also seen. Finally, repeat imaging to characterize changes in the biodistribution of the MI probe, with or without further mathematical modeling of the imaging data, to map biodistribution of the MI probe to the concentration of the molecular target(s) of interest can also be performed.

studies is that the MI probe beha ves similarl y to that observed in the animal model(s). Studying the biodistribution of the MI probe, signal to backg round in tar get tissue sites, signal in non-target tissues, and lack of toxicity in humans are all critical f irst steps. Studying the routes of elimination and metabolism of the MI probe is also usuall y impor tant. A f ailure to pro vide suf ficient signal to background at the target sites will usually lead to stopping further clinical trials. Quite often, the initial MI probe will not be optimal for clinical applications, but it will provide a lot of valuable information that will allow the MI research chain (see F igure 1) to be gin again. It could be that a ne w related molecular target(s) will be selected or more lik ely a new MI probe against the same molecular target(s) will be studied in detail to translate it for clinical applications. After fur ther clinical trials, reimbursement for the MI procedure by insurance carriers (including the Government) is key to help further its clinical use. A mathematical representation for all of MI of living subjects is given by Equation (1). Although highly simplified, this equation helps unify man y of the important characteristics of MI discussed throughout this textbook. The reader is also refer red to F igure 4, which shows how one goes from molecule to molecular image to help fur ther consolidate the infor mation presented. Signal Measured (x´, y´, z´, t, Δt) = Function ( Target Molecule Concentration (Δx, Δy, Δz, t), MI Probe Mass introduced into Subject, Pharmacokinetics of MI Probe, Field Properties Subject is Placed in, Output Signal from MI Probe, Signal Penetration Through Subject, Efficiency of Signal Detection by Imaging Instrument, Subject Preparation ) + Noise

Eq. (1)

Equation (1) needs signif icant discussion in order for the reader to understand v arious important details as outlined next. The signal measured from spatial coordinates x´, y´, z´ represents a measurement made using a physical detector, w hich is often but not al ways located outside the living subject. The detector itself usuall y occupies a

General Principles of Molecular Ima ging

volume but can effectively be considered a signal at some spatial coordinates x´, y´, z´. An imaging detector can also be placed within the li ving subject (e g, a catheter with an optical detector for intraoperative MI). The signal is often measured with man y detectors so that multiple signals emanating from all around the subject can be acquired, or alter natively the signal itself can be tagged with spatial information (eg, spatial encoding with gradients in MRI). This allows in many cases the tomographic reconstruction of images (with multiple vir tual imaging slices through the subject) to determine the spatial distribution of the injected molecular probe. The signal measured has to be collected for some time Δt to collect enough signal to obtain statistically useful data. Just as a conventional photo graphic camera that collects visib le light has to have its shutter open for a small Δt, so does a signal measurement for a MI instrument. This variable Δt determines the temporal resolution of the instr ument and therefore molecular processes that occur over a very rapid time scale relative to Δt will not be measurable or imaged by a given MI technique. The target molecule concentration in some volume element (Δx, Δy, Δz) at some time t is dependent on the underlying biolo gy and histor y of the li ving subject. For e xample, the concentration of a gi ven molecular receptor in the brain of a li ving subject may be dependent on exposure of the subject to a specif ic drug over the last 72 hours. The levels of a specif ic cell surf ace receptor ma y be dependent on transfor mation of the cell from a nor mal to a cancer cell. The parameter t is distinct from Δt, but in most cases it is assumed that during the time-period of signal acquisition ( Δt) the target molecular concentration does not change significantly. The MI probe mass introduced (eg, injected intravenously) is a critical f actor as one might e xpect. If a greater mass is injected , then potentiall y more probe can reach the molecular tar get(s) of interest. This can potentially mean a greater signal from the target molecule site(s) allo wing for shor ter acquisition times Δt. However, as greater masses of the probe of interest are injected, this leads to a greater potential for toxicity to the living subject or possib le signif icant per turbation to the subject. It is also possib le that the signal measured is a function of the concentration of more than one target molecule. For example, if a MI probe has to first be transported across the cell membrane and then bound to an intracellular receptor , then the signal of interest is a function of both the concentration of the transporter molecule and the intracellular receptor concentration.

7

PHARMACOKINETICS OF MI PROBE One of the key difficulties in imaging living subjects as compared to imaging cells e x vi vo is the inability to fully control the beha vior of the MI probe once introduced into the living subject. Whereas in ex vivo studies one might introduce an imaging probe and then simpl y wash away the excess, when studying living subjects no equivalent procedure e xists. Man y f actors including but not limited to chemical proper ties of the MI probe (eg, its lipophilicity), b lood flow, per meability, ability to cross the b lood-brain bar rier, routes of clearance (eg, renal vs hepatobiliar y), and metabolism influence the deli very and biodistribution of the MI probe. The pharmacokinetics of the MI probe is dependent on many of these f actors and can be mathematicall y modeled as detailed in Chapter 74, “Phar macokinetic Modeling.” This modeling is critical to e xtract quantitative infor mation re garding molecular tar get concentrations. It is important to note that it is often necessary to have multiple measurements over time (Signal Measured [x´, y´, z´, Δt1], Signal Measured [x´, y´, z´, Δt2], …, Signal Measured [x´, y´, z´, Δtn]) to be able to fully quantitate levels of target molecules in most cases. This might mean keeping the subject in the field-of-view of the detectors for a significant period of time to perform so-called “dynamic” imaging as opposed to a single measurement to obtain a “static” image.

PROPERTIES OF THE FIELD IN WHICH THE SUBJECT IS PLACED Magnetic, acoustic, and optical f ields in which the subject is placed can allow the signal to be created and/or detected. Magnetic f ields are critical in all for ms of MRI because without these f ields there w ould be no preferential alignment of protons within the body and no w ay to inter rogate the relaxation proper ties of those protons follo wing radiofrequency (RF) e xcitation. As such these f ields are also necessary to monitor changes in the relaxation properties of those protons following the accumulation of a nearby paramagnetic (eg, gadolinium-based) or superparamagnetic (eg, iron-o xide-based) MI probe (see also Chapters 3, “PET/MRI” 6, “Small Animal SPECT , SPECT/CT , and SPECT/MRI” and 54, “Dif fusion MRI: A Biomarker for Early Cancer Treatment Response Assessment”). In molecular ultrasound , the acoustic f ield is needed to deli ver energy to the MI probe (eg, a targeted microbubble) so that it ma y oscillate (due to a gas contained within the microbubble) in the deli vered acoustic f ield and produce a distinct acoustic signal, w hich can then be detected

8

MOLECULAR IMAGING: PRINCIPLES AND PRACTICE

(see also Chapter 28, “Ultrasound Contrast Agents”). Radionuclide-based methods do not require an e xternal field because the energy for the signal is “pre-contained” in the radioactive atoms that will produce the signal when they decay. In other cases such as optical fluorescence imaging, the MI probe has a fluorophore w hich must f irst absorb the appropriate w avelength of light in order for the MI probe to produce light of a dif ferent w avelength. In that case, energy has to be deli vered to the MI probe and is not “pre-contained” in the fluorophore.

detectors in order to be measured. If the detectors are outside the li ving subject, there is a cer tain probability that a gi ven signal will reach a specif ic detector. Unfortunately, it is rare for this probability to be high due tothe fact that man y signals do not penetrate through tissues. Some types of signal ha ve a much g reater probability of passing through tissues (e g, gamma ra ys for PET) than others (eg, visible light for optical imaging). The physical properties of the tissues through w hich the signal must penetrate determine the probability of detection.

OUTPUT SIGNAL FROM MI PROBE

EFFICIENCY OF SIGNAL DETECTION BY IMAGING INSTRUMENT

The output signal from the MI probe is k ey as it provides the means by which the spatial distribution of the molecular probe ma y be deter mined. The signaling component of the MI probe ma y consist of a radioacti ve atom (as in single photon emission CT) or a gas that responds with oscillations in an ultrasound f ield (as with microbubb lebased MI probes) (see Chapter 15, “Ultrasound”). Any portion of the physical spectrum as well as sound may be used to provide signal from the MI probe. Dif ferent portions of the physical spectrum will do better in penetrating through tissues, making some techniques more useful at all depths and others much more limited.There are several important issues related to the output signal. Ideall y, the MI probe w ould onl y produce the output signal after it finds and interacts with its intended molecular tar get(s). This would allow for a v ery low backg round signal and more accurate deter mination of the location(s) of the molecular tar get(s) of interest. Such a smar t probe (see Figure 3) is possib le for some MI strate gies (eg, fluorescence optical imaging) and not for others (e g, radionuclide-based strategies). Another key issue is w hether the output signal can be generated onl y once (e g, as with a radioactive atom) or multiple times (e g, a photoacoustic agent that can produce sound as long as it is e xcited and re-excited by light). Linked to this concept is the issue of whether the output signal requires e xcitation or not. It is also possible to de velop MI probes that pro vide multiple types of signals by having multiple signaling components (eg, for positron emission tomo graphy [PET] and optical (see Chapters 9, “Optical Multimodality Technologies,” and 29, “Multimodality Agents”)).

SIGNAL PENETRATION THROUGH SUBJECT The signal being produced b y the MI probe, w hether at the tar get molecule site(s) or not, needs to reach the

Even if a MI probe produces a signal that successfull y penetrates through the tissues of the li ving subject and reaches the detector, it may not be able to be stopped and actually detected b y the detector . In man y cases, a v ery small percentage of the total signal emitted is actuall y detected by the detectors. In PET , as little as 1 to 2% of the total signal produced may be detected. This efficiency of detection is a k ey variable since it will af fect the ability to deter mine v ariables such as Δt so that suf ficient signal can be detected for deter mination of the tar get molecule concentration. Some instr uments attempt to collect the signal from all around the subject b y using multiple detectors, w hereas others do not (sometimes to save costs), w hich leads to a fur ther decrease in o verall instrument efficiency.

SUBJECT PREPARATION It should not be underestimated that the manner in which the subject is prepared can signif icantly impact the measured signal. F or e xample, the ef fects of anesthesia can alter the MI probe phar macokinetics in the body . Shaving hair on the surf ace of a small li ving subject can significantly affect the detected signal in optical imaging. Of course, in some cases, one w ants a specif ic perturbation (eg, introduction of a drug to the subject) to produce a change in the le vels of molecular tar get(s) of interest, however, this is different from subject preparation per se. Keeping in mind these types of issues is key for quantitation and interpretation of the molecular images.

NOISE Noise in the measurement can be due to tw o primar y sources, random or str uctured. Random noise or statistical noise is directl y related to the detected number of

General Principles of Molecular Ima ging

signals from the MI probe. Structured noise refers to nonrandom v ariations in counting rates w hich adv ersely affect the inter pretation and anal ysis of the resulting images (eg, due to or gan/tissue motion, imaging instr ument non-uniformities).

CONCLUSION The f ield of MI of li ving subjects in volves man y biomedical disciplines for its progress to date and further continued evolution. These include physics and engineering for fundamental detector/instr ument design and construction, chemistr y and materials science for development of MI probes, molecular phar macology for optimized delivery and phar macokinetics of MI probes, cell/molecular biolo gy for understanding molecular targets of interest and for repor ter gene-based strategies, advances in genomics, proteomics and high-throughput screening technolo gies for disco very and v alidation of new molecular tar gets, mathematics and bioinfor matics for image reconstr uction and image/data modeling, and clinical medicine for applications of the strate gies for medical management. In addition, man y fields including immunology, microbiolo gy, and de velopmental biolo gy are using and helping to adv ance the v arious MI technologies. The MI f ield is limited b y the amount of currently a vailable trained indi viduals in the v arious sub-disciplines, and there continues to be a shor tage in several of the k ey areas needed to help e volve the f ield (eg, chemistry). Some key factors related to the optimism and caution for the f ield of MI of living subjects require a brief discussion. There continues to be optimism that the

9

molecular specificity provided by MI remains one of its key strengths. As biology and pathology are more full y understood at the molecular le vel, it remains lik ely that MI techniques will for m the basis for inter rogation of all living subjects. Some will argue that we may eventually not need MI because w e will be ab le to detect and treat disease with very specific drugs without requiring knowledge of the spatial localization of the disease. For example, a b lood test might detect a panel of protein biomarkers indicati ve of earl y cancer . Highl y specif ic drugs for this cancer could then be administered without any imaging, and serial b lood biomarker measurements would be used to monitor response of the subject to therapy. These types of in vitro assays have the adv antage o ver MI in that the y can be highl y multiple xed, potentially assa ying thousands of molecular e vents simultaneously. Ho wever, it ma y still be the case that some sites of disease are responding to treatment, whereas others are not; the b lood-based detection will not detect this. It is more lik ely that a combination of strategies, which include MI (e g, measuring blood protein biomark ers and perfor ming molecular/anatomical imaging), will lik ely play an increasing role in clinical disease management for the foreseeab le future. Ne vertheless, it is impor tant to objectively study and discuss the advantages and disadvantages of MI relative to nonimaging strategies as well as to objectively compare various imaging modalities with each other. The MI field is still in its inf ancy, and careful research should help usher in a ne w generation of technolo gies/assays that will continue to fundamentall y change our understanding of biolo gy and patholo gy and the clinical management of patients.

2 IMAGING OF STRUCTURE AND FUNCTION WITH PET/CT DAVID W. TOWNSEND, PHD

Historically, instr umentation for cross-sectional (tomographic) imaging of function, single photon emission computed tomography (SPECT) and positron emission tomography (PET) evolved along a path somewhat different f rom t hat o f a natomic i maging, c omputed tomography (CT) and magnetic resonance (MR) imaging (MRI), and the cor responding clinical studies w ere performed and interpreted separately in different medical departments, nuclear medicine and radiolo gy, respectively. Despite this se gregation, the use of combining anatomic and functional planar images w as e vident to physicians even in the 1960s, 1 preceding the invention of CT. The alignment of tomo graphic images is a comple x procedure owing to the lar ge number of de grees of freedom and without some common features, such as li ver and lung boundaries and boney structures, co-registration may be prob lematic. In addition to simple visual alignment, or the use of stereotactic frames that are undesirable or incon venient for diagnostic imaging, sophisticated image fusion software was developed from the late 1980s onwards.2 For (relatively) rigid parts, such as the brain, software can successfully align images from MR, CT, and PET, whereas in more flexible parts, such as the rest of the body, accurate alignment is more dif ficult owing to the lar ge number of possib le dif ferences (degrees of freedom) between the two data sets. Software fusion is also dependent on matching common features that are e xtracted either from the images (lung boundaries) or from the mark ers placed on the patient. Functional imaging modalities, such as PET and SPECT, often lack reliable anatomic cor relates and ha ve lower resolution and higher noise levels than CT or MR. One way to address the prob lems of software fusion is b y combining de vices (emission and transmission) rather than fusing the images post hoc, an approach that 10

has now coined the term hardware fusion. A combined or multimodality scanner , such as PET/CT , can acquire coregistered structure and function in a single study. The data are complementary allowing CT to accurately localize the functional abnormalities and PET to highlight the areas of abnor mal metabolism. A fur ther adv antage of combined instr umentation is that the anatomic images from CT can be used to impro ve quantitation of functional images through more accurate attenuation and scatter and par tial v olume cor rections; attenuation (absorption) and scatter of the annihilation photons occurs as the y interact with the tissue of the patient. These cor rections are impor tant to achie ve accurate and objective assessment of functional parameters, such as myocardial perfusion, tumor uptak e values, and dosimetry for treatment planning and monitoring response. Since the commercial introduction of PET/CT in 2001, adoption of the technolo gy has been rapid , particularly in oncology. Advances in CT and PET instrumentation ha ve been incor porated into the v ery latest PET/CT designs. A recent ar ticle3 offers an e xcellent overview of PET imaging, and it summarizes major advances in the instrumentation. This chapter will describe some of the early work that led to the commercial exploitation of PET/CT , mo ve forw ard to the cur rent designs, and subsequently discuss the impact of recent advances in CT and PET performance on these designs. Since photon attenuation (absor ption) is the parameter specifically measured by the CT scanner, an algorithm to derive correction factors from the CT images to compensate for the attenuation ef fects in PET images (CT -AC) will be proposed. The practical challenges to implement such an algorithm will also be addressed. The chapter will conclude with a brief re view of the clinical impact of PET/CT.

Imaging of Structure and Function with PET/CT

HISTORICAL CONCEPTS The origins of tomo graphic imaging in medicine date from the 1960s or e ven earlier, whereas fusion of tomographic images w as not investigated systematically until the late 1980s. 2 Following on from the earlier superposition of planar images, 1 the 1990s witnessed the de velopment of tw o principle approaches to image fusion: the software approach and the hardw are approach. The software approach attempts to align tw o image sets post hoc after they have been acquired from two different imaging modalities at tw o dif ferent times. In contrast, the hardware approach combines the instr umentation for tw o imaging modalities and thus acquires both modalities within the same reference frame thereb y ensuring as accurate alignment as possible.

Image Fusion with Software A thorough discussion of the topic is be yond the scope of this chapter. However, it is instructive to review some of the principles of softw are fusion and the e xtent to w hich they can be successful; a thorough re view of softw are fusion methods can be found in Ha wkes and colleagues, 4 and Slomka5 presents specific aspects of merging anatomic and molecular information and of ho w they relate to combined PET/CT design. Fusion of two image sets is achieved either by identifying common landmarks, b y f iducials that can then be aligned, or b y optimizing a metric based on image intensity values. Whatever the method, the number of possible de grees of freedom between the tw o image v olumes defines the comple xity of the subsequent mathematical transformation, the function that, when applied to one image set, converts it into the reference frame of the other image set. For distributions that do not involve a change in shape or size, rigid body transfor mations are adequate. When shears (or a nonisotropic dilation without shear) are in volved, an affine transfor mation comprising a linear transfor mation and translation is indicated. When there are no constraints on the deformation, a nonlinear transfor mation (warp) is used. Although methods involving the alignment of extracted features or fiducials6–8 have shown some success, at least for the brain, most cur rently used methods are intensity based , and images are coregistered by assessing the intrinsic information content. Metrics include intensity ratios 9 and mutual information.10 While such techniques have shown considerable success in aligning images of the brain obtained with CT, PET, SPECT, and MR, the y have been less successful for other parts of the body. Earlier clinical assessment in the lung11 and the pelvis 12 were disappointing, demonstrating a

11

local misalignment of the two image sets of the order of 5 to 8 mm, compared with a re gistration accuracy of better than 2 mm that can be achieved for the brain.13 A recent review14 suggests that softw are fusion can achie ve an accurac y of about half a pix el, or 2 to 3 mm, for some studies although clinical results from more recent generations of fusion software have not been particularly encouraging for example, in recurrent colorectal cancer.15 Software de velopment has nevertheless continued, as illustrated b y the recent pub lication of an automated w arping algorithm to align CT and PET images of the thorax. 16 Commercially a vailable softw are has considerably improved over the past fe w years both in the accurac y of the registration algorithms and in the sophistication of the user interface and displa y. As an e xample, Hermes Medical Solutions (Stockholm, Sw eden) of fers adv anced fusion software for man y clinical applications, including correction of misalignment errors for PET/CT scans, registration of PET/CT scans with MR, re gistration of longitudinal PET/CT studies, alignment of PET and MR scans in Alzheimer’s disease and other for ms of dementia, and registration of SPECT or PET myocardial perfusion studies with CT or MR of the hear t. Fusion software can also play an impor tant role in radiation therap y planning, where PET images are used to def ine the treatment plan.17,18 Fusion of the PET/CT study with the simulation CT scan that is used to plan the radiotherapy treatment can result in modif ication of that standard CT -only treatment plan in a signif icant percentage of cases, par ticularly for disease of the lung. Ho wever, despite considerab le progress, fusion softw are will probab ly ne ver compete with the simplicity and con venience of the core gistered studies acquired on a combined PET/CT scanner.

Multimodality Prototypes The pioneering w ork of the late Br uce Hase gawa and colleagues19,20 in the 1980s set the stage for the hardw are solution to image fusion. The aim of this w ork w as to design a de vice that could perfor m emission (radionuclide) and transmission (X-ray) tomography with the same detector that, in this case, w as se gmented high-purity germanium.20 Although this approach is attracti ve, it is difficult to design a detector that does not compromise performance for at least one of the tw o modalities. The work was signif icant, however, in that it highlighted the strengths of a single de vice that can perfor m both anatomic (CT) and functional (SPECT) imaging. 21 Of comparable significance was the use of the CT images to generate attenuation cor rection f actors (A CFs) for the

12

MOLECULAR IMAGING: PRINCIPLES AND PRACTICE

emission (SPECT) data.22 These factors are used to correct the SPECT images for the attenuation (absorption) of the photons as the y traverse the patient from the point of emission to the detector . The device was used for studies in phantoms, artificial constructions used to assess instrument performance, and in animals, par ticularly a study of myocardial perfusion in a porcine model. 23 However, recognizing the dif ficulty of building a detector that w ould operate optimall y for both CT and SPECT , Hase gawa turned to a dif ferent design that comprised a clinical SPECT gamma camera in tandem with a clinical singleslice CT scanner .24 The CT scanner (9800 Quick; GE Healthcare) was positioned in front of, and aligned with, a scintillation camera (600 XR/T ; GE Healthcare). The same bed was used to acquire both studies, and the images were registered by taking into account the axial displacement betw een the CT and the SPECT imaging f ields. After injection of the radiotracer and an uptake period, the patient was scanned first in the CT and subsequently in the SPECT scanner. The CT data were used to generate the SPECT ACFs. The combined de vice w as used to acquire a small number of clinical studies, such as those for quantitati ve estimation of radiation dosimetr y in patients with brain tumor.25 The proposal to combine PET with CT w as made in the early 1990s by Townsend and colleagues independently of the Hase gawa w ork. This concept originated from an earlier low-cost PET scanner design that comprised rotating banks of bismuth germanate (BGO) block detectors that w as de veloped b y Townsend and colleagues at the University of Geneva in 1991. The gaps between the banks of BGO detectors of fered the possibility to incorporate a different imaging modality, such as CT, within the PET scanner. Thus, the concept of PET/CT w as bor n with a 1991 proposal that the components of a CT scanner would be mounted in the gaps betw een the banks of BGO b lock detectors. The suggestion w as also made to use the CT images to generate the PET ACFs.26 The f irst prototype PET/CT scanner became operational in 1998,27 designed and built b y CTI PET Systems in Kno xville, TN (no w Siemens Molecular Imaging) and clinicall y e valuated at the University of Pittsburgh. The design incorporated a single-slice spiral CT scanner (Somatom AR.SP; Siemens Medical Solutions, F orchheim, Ger many) and a rotating ECAT ART scanner (CTI PET Systems). The PET detectors were mounted on the rear of the CT suppor t and the entire assembly rotated as a single unit (Figure 1). The data processing included an algorithm28 to scale the CT images from X-ra y ener gy to PET annihilation photon ener gy (511 keV) and generate the appropriate ACFs (see section “CT-Based ACFs”). Ov er 300 patients with cancer w ere

Figure 1. The first PET/CT prototype evaluated clinically at the University of Pittsburgh. The CT and PET components were mounted on a single rotating support and the data acquired from two separate consoles. The CT images were transferred to the PET console and then used for CT-based attenuation correction and localization.

scanned on the prototype, and the findings are presented in a series of publications.29–31 The results from the prototype demonstrated the impor tance of high-resolution anatom y accurately registered with functional data. The coregistered anatomy localized the functional abnor malities and clarified the equivocal situations, thus improving the accuracy and conf idence of the scan inter pretation. The use of a rapidly acquired, low-noise CT scan in place of a length y conventional PET transmission scan impro ved image quality and reduced scan time.

CURRENT PET/CT INSTRUMENTATION At the end of the 1990s, for a physician wishing to review fused images, the only real option was the software fusion techniques described above in the section “Image Fusion With Software.” Apart from the drawbacks of fusion software mentioned above, access to image data from dif ferent modalities w as f ar from a routine procedure, e ven with picture archiving and communication systems available. Thus, fusion imaging w as typically perfor med for, at most, only a small number of patients. Software fusion packages were, nevertheless, available on many imaging systems and par ticularly those used for radiation oncolhen GE ogy.32 This situation then changed in 1999 w Healthcare launched a dual-head scintillation camera combined with a lo w-power X-ra y tube and detectors,

Imaging of Structure and Function with PET/CT

called the Ha wkeye (GE Healthcare). 33,34 This design features tw o rectangular sodium iodide camera heads with a 350-W X-ra y tube. The Ha wkeye w as the f irst commercial scanner to of fer combined anatomic and functional imaging in a single unit. Less than 2 y ears after the f irst Hawkeye installation, PET/CT scanners incor porating clinical CT and clinical PET perfor mance became commerciall y available. The f irst commercial PET/CT scanner to be announced w as the Disco very LS (GE Healthcare) in early 2001. This was followed a few months later by the Biograph (Siemens Medical Solutions), and then, somewhat later b y the Gemini (Philips Medical Solutions). In the past 7 years, PET/CT designs from all vendors have evolved following the advances in CT and PET instr umentation that is described later . As to the present situation (2008), f ive v endors worldwide now offer PET/CT designs: GE Healthcare, Hitachi Medical, Philips Medical Systems, Toshiba Medical Corporation, and Siemens Medical Solutions. Cur rent designs of fered b y Siemens Molecular Imaging, GE Healthcare, and Philips Medical Systems are summarized in F igure 2. The specif ications and perfor mance of the PET components are v endor specif ic, with the Biograph H I-REZ TruePoint ( Figure 2 A; Si emens Medical Solutions) providing good spatial resolution in 3D with 4 mm × 4 mm × 20 mm lutetium oxyorthosilicate (LSO) cr ystals35; the original Bio graph design (Biograph Classic) w as based on 6.4 mm × 6.4 mm × 25 mm LSO detectors. The Bio graph is cur rently offered with 6-, 40-, and 64-slice CT scanners. The Discovery LS, the original PET/CT design from GE

B

A

Biograph 6, 40, 64

LSO 6.4 6.4 25 mm3 4 4 20 mm3 3D only (no septa) 6, 40, 64 slice CT 70 cm port 21.6 cm axial FOV 4.5 ns coincidence bed on rails

Discovery ST, STE, VCT, RX

BGO, LYSO 3 4.7 6.3 25 mm (BGO) 4.2 6.2 20 mm3(LYSO) 2D/3D (septa) 8, 16, 64 slice CT 70 cm port 15.7 cm axial FOV 11.7 ns coincidence dual-position bed

C

Gemini GXL, TF

GSO, LYSO 3 4 4 30 mm (GSO) 4 4 22 mm3(LYSO) 3D only (no septa) 6, 10, 16, 64 CT 71.7 cm port 18 cm axial FOV 6 ns coincidence bed support in tunnel

Figure 2. Current PET/CT scanner designs from three of the major suppliers of medical imaging equipment: A, the Siemens Biograph TruePoint, B, the GE Healthcare Discovery range, and C, the Philips Gemini series. See Table 1 for the physical properties of the different scintillators; LYSO is lutetium oxyorthosilicate (LSO) with a small percentage of yttrium.

13

Healthcare, combined the Advance NXi PET scanner with a 4-, 8-, or 16-slice CT .36 The Disco very ST (Figure 2B; GE Healthcare) has 6.2 mm × 6.2 mm × 30 mm bismuth ger manate (BGO) detectors in combination with a 4-, 8-, or 16-slice CT scanner; unlik e the Discovery LS, the gantr y of the PET scanner matches the dimensions of the CT scanner . The higher resolution Disco very STE has 4.7 mm × 6.3 mm × 30 mm BGO detectors in combination with 8- or 16-slice CT scanners; the Disco very VCT is an STE conf igured with a 64-slice CT scanner . Although not listed of ficially, the Discovery RX is a research tomograph based on the scintillator LYSO (LSO with a small percentage of yttrium) with detector geometr y comparable to that of the STE, including the retractable septa. The Gemini GXL (Figure 2C; Philips Medical) comprises 4 mm (in plane) and 6 mm (axial) gadolinium o xyorthosilicate (GSO) detector pix els, 30 mm in depth; the Gemini is also an open design with the capability to ph ysically separate the CT and PET scanners for access to the patient. The Gemini GXL incor porates a 6- or 16-slice CT scanner. A recent addition to PET/CT designs is the Gemini TF, the f irst commercial time-of-flight (T OF)PET scanner .37 The Gemini TF has 4 mm × 4 mm × 22 mm LYSO detectors and is combined with a 16- or 64-slice CT scanner . All designs e xcept Discovery LS provide a 70-cm patient por t for both CT and PET. While Discovery and Gemini of fer standard PET transmission sources as an option, in practice, as mentioned above, most, if not all, institutions use CT-based attenuation correction because of the advantages of low noise and shor t scan times that f acilitate high patient throughput. The Gemini and Bio graph acquire PET data in 3D mode onl y, whereas the Disco very series incorporates retractab le septa and can acquire data in both 2D mode with the septa e xtended and the 3D mode with the septa retracted. Since 2001, numerous pub lications ha ve no w documented the benef its of PET/CT compared with PET and CT, with and without software fusion. A good review of the literature prior to September 2006 can be found in Czernin and colleagues. 38

ADVANCES IN PERFORMANCE FOR CT AND PET Multidetector CT Scanners Following the de velopment of single-slice spiral CT scanners in the earl y 1990s, 39 CT performance has experienced a re vival with the adv ent of multidetector

14

MOLECULAR IMAGING: PRINCIPLES AND PRACTICE

arrays (MDCT). This was accompanied by increases in X-ray power (from 30 kW up to 60 kW or g reater) and computer capacity for data processing and image reconstruction. Dual and 4-slice CT scanners f irst appeared around 1998 with scan times of 500 ms, followed b y 16-slice (2002) and more recentl y, 64-slice (2004) CT scanners. The increasing number of detector rows (slices) has been accompanied b y f aster rotation times so that state-of-the-ar t scanners can now achieve a full rotation in as little as 330 ms. Spatial resolution has improved from about 10 line pairs (Lp)/cm in 1990 up to 25 Lp/cm or better toda y, with a slice thickness less than 1 mm. A signif icant innovation that will contribute to increased CT perfor mance is the Straton X-ray tube. 40 In the Straton tube, the entire v acuum vessel including the anode and cathode rotates resulting in much more ef fective cooling and heat dissipation, which is a limitation of the con ventional tube. 41 Cooling rates 5 to 10 times higher than for con ventional tubes can be achie ved with the Straton tube that thus result in shor ter rotation times and f aster scans. Since weight is a problem for conventional X-ray sources, the dual-tube Def inition CT (Siemens Medical Solutions) is achie vable because of the Straton tube. After many years of slo w but steady pro gress, the past decade has seen signif icant adv ances in both hardw are and software for CT.

PET Scanners Muehllehner and Kar p3 offer an excellent review of progress in PET instr umentation, including a summar y of the physical performance of the new, fast scintillators recently introduced for PET. This section will summarize some of these advances as they relate to current PET/CT scanner performance.

introduction of new scintillators such as GSO42 and LSO,43 both doped with cerium, impro ved the perfor mance of PET scanners for clinical imaging. Both GSO and LSO have shorter decay times than BGO b y a f actor of 6 to 7, reducing system dead time and impro ving count rate performance, particularly at high activity levels in the field of view (FOV). The physical properties of these scintillators are compared in Table 1. Of even more importance for clinical imaging is the potential of f aster scintillators to decrease the coincidence timing window, thereby reducing the randoms coincidence rate. The increased light output of the ne w scintillators impro ves the ener gy resolution because the increased number of light photons reduces the statistical uncertainty in the energy measurement. However, other ph ysical ef fects contribute to the emission process, and the improvement in energy resolution is not a simple function of the number of light photons. The higher light output also increases the positioning accurac y of a block detector, allowing the b locks to be cut into smaller crystals, thereby improving spatial resolution. BGO, LSO, and GSO do not absorb moisture w hen exposed to air (ie, are not hygroscopic), thus facilitating the manufacture and packaging of the detectors. GSO is some what less r ugged and more dif ficult to machine than either BGO or LSO . LSO has an intrinsic radioacti vity of about 280 Bq/mL, with single photon emissions in the range 88 to 400 keV. Such a radioacti ve component is of little impor tance for coincidence counting at 511 keV, except maybe at very low emission count rates. Sensitivity

PET is intrinsically a 3D imaging methodology, replacing physical collimation required for single-photon imaging with the electronic collimation of coincidence detection. The f irst multi-ring PET scanners incor porated septa, lead, or tungsten annular shields mounted betw een the

New Scintillators for PET

In the 1970s, PET detectors sa w the transition from thallium-activated sodium iodide (NaI(Tl)) to BGO, a scintillator with higher density and photofraction. The photofraction is the fraction of incident annihilation photons that interact in the scintillator through the photoelectric effect; this is the desired process, in preference to Compton scattering that ma y involve multiple interaction points within the detector. While at least one PET scanner design continued to use NaI(Tl) until f airly recentl y, the majority of PET scanners installed during the 1990s w ere based on BGO block detectors. In the late 1990s, the

Table 1. PHYSICAL PROPERTIES OF PET SCINTILLATORS Property

NaI

Density (g/mL) 3.67 Effective atomic number 51 Attenuation length (cm) 2.88 Decay time (ns) 230 Photons/MeV 38,000 Light yield (%NaI) 100 Hygroscopic Yes

BGO

LSO

GSO

7.13 74 1.05 300 8,200 15 No

7.4 66 1.16 35–45 28,000 75 No

6.7 61 1.43 30–60 10,000 25 No

NaI = sodium iodide; BGO = bismuth germanate; LSO = lutetium oxyorthosilicate; GSO = gadolinium oxyorthosilicate.

Imaging of Structure and Function with PET/CT

detector rings. The purpose of the septa was to shield the detector rings from photons that scattered out of the transverse plane, thus restricting the use of electronic collimation to within the plane, a limitation that, while it mak es poor use of the radiation emitted from the patient, limits scattering and allows 2D image reconstruction algorithms to be used. The availability of BGO scanners from 1990 onwards with retractable septa encouraged the use of 3D methodology, at least for the brain, where the net increase of a factor f ive in sensitivity could be realized even with accompanying increases in both scatter fraction and randoms.44 The condition for whole body imaging is f ar less favorable, in part due to the presence of significant activity just outside the imaging FO V in most bed positions. Instead, par ticularly for lar ge patients, 2D imaging has been recommended although higher le vels of the injected biomarker, such as 2-deo xy-2-[F-18]fluoro-D-glucose (FDG), are required to obtain adequate count rates. This situation changed in the late 1990s with the introduction of LSO- and GSO-based scanners that could be operated with shor t coincidence time windo ws (4.5 to 6 ns) and higher energy thresholds (400 to 450 keV) compared with 10 to 12 ns and 350 k eV for a typical BGO scanner . Significantly impro ved w hole-body image quality has been achieved in 3D with a 10 mCi (370 MBq) injection of FDG. A recommended injected dose of 12 to 15 mCi corresponds to operation at peak noise equi valent count rate (NECR) for an LSO scanner in 3D.45 However, since the LSO and GSO scanners ha ve no septa and acquire data in 3D mode only, no comparison can be made with 2D operation. Within the past 2 to 3 y ears, a limited number of LYSO-based scanners with retractab le septa ha ve been evaluated in 2D and 3D operations, and recent publications suggest that 3D operation is no w prefer red o ver 2D operation.46–48 The sensitivity of a scanner can also be impro ved by the addition of more detector material. Planar sensiti vity can be increased b y extending the thickness of the scintillator. For example, a 50% increase in thickness (20 to 30 mm) results in a 40% increase in sensitivity. However, increasing the axial extent by 30% will result in a 78% increase in volume sensitivity (for 3D acquisition with no septa). The latter thus mak es more ef ficient use of the increased volume of LSO although there will also be an increase in the number of phototubes required (and hence increased cost). F ollowing an injection of a radioacti ve tracer, such as FDG, the patient recei ves a radiation dose from all annihilation photons, not just those emitted within the imaging FO V of the scanner . Therefore, the greater the axial co verage, the better use is made of the emitted radiation and the more ef ficient use is made of a

15

given volume of scintillator. For most PET/CT scanners, axial PET co verage is about 16 cm, with one design having an axial extent of 18 cm.37 The most recent design to be announced has an e xtended FOV covering 21.6 cm axially. The latter comprises over 32,000, 4 mm × 4 mm × 20 mm LSO pixels and images 109, 2-mm thick transaxial planes in a single position. Data acquisition is in fully 3D, and the scanner has a peak NECR of around 160 kcps.49,50

Signal-to-Noise

The availability of scintillators that are both fast and have sufficient density to stop a large fraction of incident photons (ie, high stopping power), such as LSO (and LYSO), has revived interest in PET TOF,51 interest that has been further stimulated by the announcement of the f irst commercial PET/CT with TOF, the Philips Gemini TrueFlight (TF).37 The principle of TOF PET is illustrated schematically in F igure 3; positron annihilation occurs in the patient at a distance d − d1 from one detector (Detector A) and d + d 1 from the other detector (Detector B). For photons traveling at the speed of light (c), the arrival time difference between the two photons at the detectors is 2d1/c. Photons originating from the center of the FO V (d1 = 0) obviously arrive in the detectors at the same time. Scanners with f ast scintillators and electronics can measure this time dif ference to within a cer tain resolution. For e xample, for a scanner with a coincidence timing resolution of 500 ps, the spatial uncer tainty on the position of the annihilation is 7.5 cm. This uncertainty is not sufficient to place the annihilation within a 2-mm voxel (and thereby eliminate reconstr uction), but it is superior to having no timing information at all and assigning equal probability to all v oxels along the line-of-response (Figure 3A). Instead , the most probab le location of the annihilation is at the center of the uncertainty distribution in F igure 3B . The TOF infor mation is incor porated directly into the reconstr uction algorithm leading to an improvement in signal-to-noise (SNR). The increase in SNR is proportional to ( D /δ d) , where D is the diameter of the activity distribution and δd is the spatial uncertainty. For a 40 cm diameter unifor m distribution and a 7.5 cm uncer tainty, the increase in SNR is a f actor of about 2.3. As the TOF resolution improves, the spatial uncertainty decreases and the SNR increases b y a lar ger factor. TOF PET w as f irst in vestigated in the earl y 1980s51 with scintillators that w ere fast but did not ha ve good stopping po wer for 511 k eV photons. Interest declined until the recent emer gence of scintillators that

16

MOLECULAR IMAGING: PRINCIPLES AND PRACTICE

B

A

Figure 3. A schematic illustrating positron emission tomography (PET) data acquisition with the incorporation of Timeof-Flight (TOF) reconstruction. By measuring the time difference between the arrival of the two annihilation photons, the position of the positron annihilation along the line-of-response (LOR) can be localized with an accuracy dependent on the precision of the temporal measurement: A, without TOF information, the annihilation is located with equal probability along the LOR and B, using TOF information the annihilation point can be localized to a limited range, for example, a 500 ps timing resolution corresponds to 7.5 cm FWHM.

are both f ast and sensiti ve. The new TOF PET scanners based on LSO or L YSO must demonstrate good timing resolution that is stable over time so as to a void frequent detector recalibration. While promising, the clinical impact of TOF PET has yet to be established, although it is anticipated to ha ve a role in the imaging of lar ge patients. This is because the lar ger the diameter of the activity distribution, the g reater the potential increase in SNR for a given time resolution. A more detailed review of the pub lished contributions to TOF development can be found in Muehllehner and Kar p.3 Reconstruction Algorithms

There has been signif icant progress during the past fe w years in image reconstruction methods through the introduction of statisticall y based algorithms into the clinical setting. Pre viously, one of the earliest and most widel y used 3D reconstr uction methods w as the reprojection algorithm (3DRP) based on a 3D e xtension of standard 2D filtered backprojection.52 While this algorithm works well for the lo wer noise en vironment of the brain, the quality for whole-body imaging is less than optimal, particularly when rod source ACFs are applied to lo w count emission data. F igure 4A sho ws a coronal image of a patient with a body mass index (BMI) of 25 reconstructed using 3DRP. Since CT-based ACFs have been applied, the quality is actuall y better than w ould have been obtained

with rod source ACFs. The development of Fourier rebinning (FORE) 53 was a breakthrough that enab led 3D data sets to be accuratel y rebinned into 2D data sets and then reconstructed in 2D with a statistically based expectation maximization (EM) algorithm. The adv antage of the rebinning step is that it accurately compresses a large 3D data set into 2D slices that can then be reconstructed separately using a 2D reconstr uction algorithm. Ev en so, it was not until the accelerated convergence achieved by the ordered-subset EM (OSEM) algorithm54 that those statistical methods became of clinical interest. While FORE and OSEM pro vide impro ved image quality compared with 3DRP, a fur ther advance was the incor poration of attenuation information directly into the reconstruction model in the for m of w eighting f actors: attenuationweighted (AW) OSEM.55 Figure 4B shows the same data set as in F igure 4A reconstr ucted with FORE and AWOSEM; the improved image quality is evident. Further improvement has been achieved by eliminating the rebinning step and implementing OSEM fully in 3D with corrections for randoms, scatter, and attenuation incorporated into the system model. 56,57 The result, again for the same data set, is shown in Figure 4C. Finally, in a recent development ter med high-def inition (HD) PET , the detector spatial response function has also been included in the reconstr uction model. 58 The point-spread function (PSF) v aries throughout the FO V o wing to the ob lique penetration of the detectors b y annihilation photons. By measuring this variability and then modeling the PSF, improved and near -uniform spatial resolution can be achieved throughout the FOV; the improvement can be seen by comparing F igure 4C with the PSF reconstr uction in F igure 4D; all reconstr uctions e xcept 3DRP are unsmoothed. The images in Figure 4 are reconstructed with clinical softw are pro vided b y a specif ic v endor (Siemens Molecular Imaging). Of course, most major vendors provide comparable software capable of producing clinical images of high quality . The VUE point algorithm (GE Healthcare) is an implementation of 3D OSEM that includes cor rections for randoms, scatter , and attenuation and also a z-axis smoothing.The Gemini TF (Philips Medical Systems) has TOF capability; and therefore, the TOF infor mation must be incor porated into the reconve struction.37 For their Gemini scanners, Philips ha implemented a distributed list-mode TOF algorithm (DLT) that is based on a TOF list-mode maximum likelihood approach de veloped b y P opescu and colleagues.59 They pre viously used a ro w-action maximum likelihood algorithm or RAMLA.60 The scatter correction algorithm requires modif ication to incor porate

Imaging of Structure and Function with PET/CT

A

B

C

D

17

Figure 4. A coronal section of an FDG-PET whole-body scan of a patient with a body mass index of 25 acquired in 3D mode with septa retracted and reconstructed using: A, 3D filtered backprojection algorithm with reprojection (3DRP) (7 mm Gaussian smooth), B, FORE + 2D OSEM (14 subsets, 2 iterations; no smoothing), C, 3D OP-OSEM (14 subsets, 2 iterations; no smoothing), and D, HD PET: 3D OSEM with point-spread function reconstruction (14 subsets, 2 iterations, no smoothing).

TOF infor mation. The g reatest outstanding ef fect on image quality and a challenge to reconstr uction algorithms is due mainly to the size of the patient, a signif icant problem given the cur rent levels of obesity among the US population.

CT-BASED ACFS For PET/CT, a reco gnized strength is the a vailability of CT images for attenuation correction of the PET data,28,61 eliminating the need for an additional, length y transmission scan. The use of the CT to generate ACFs not onl y reduces the scan time b y a signif icant amount but also results in more accurate ACFs. Since the attenuation values (µ) are energy dependent, the CT scan at a mean photon energy of ~70 keV must be scaled to PET (511 k eV) energy. The mean energy of a polychromatic X-ray beam is def ined as the ener gy of a monochromatic beam that would give the same linear attenuation as the polychromatic beam inte grated over energy.62 The polychromatic beam also results in beam hardening, the preferential interaction of lo wer energy photons as the beam traverses the body causing the mean ener gy to increase and the corresponding µ values to decrease.

Energy Scaling Algorithm for CT The attenuation of X-rays by tissue depends on the density and the ef fective atomic number (Z eff) of the

material. At these ener gies, the ph ysical processes b y which X-rays are attenuated are the photoelectric effect and Compton scattering. The photoelectric probability varies appro ximately as Z eff4 and scales as 1/E 3 with photon energy (E). The Compton scattering probability has little dependence on Zeff and decreases linearly with E. The linear attenuation coefficient for a given material is expressed by the sum of the two components: µ(E) = ρe{σc(E) + σph(E, Zeff)} where ρe is the electron density and σph and σc are the photoelectric and Compton cross-sections per electron, respectively. Electron density is the number of electrons per unit v olume, although w hen quantum mechanical effects are significant, it is the total probability of finding an electron within a unit v olume. Ho wever, at photon energies above about 100 keV in tissue, the photoelectric contribution is essentiall y ne gligible compared with the Compton contribution, and therefore, the e xpressions for the attenuation coefficient at X-ray energy Ex and gamma energy Eγ are: µ(Ex) = ρe{σc(Ex) + σph(Ex, Zeff)} µ(Eγ) = ρe σc(Eγ) As a consequence of the tw o separate contributions to µ(Ex), a single measurement of µ(Ex) will not uniquely determine µ(Eγ) because, for example, an increase in Z eff could of fset a decrease in ρe resulting in no change in µ(Ex). In general, therefore, a simple energy scaling of

18

MOLECULAR IMAGING: PRINCIPLES AND PRACTICE

µ(Ex) is insuf ficient to yield µ(Eγ). By restricting the problem to biolo gic tissues for w hich Z eff are all f airly comparable and noting that the contribution from σph is relatively small even at X-ray energies, changes in µ(Ex) are primarily due to changes in the electron density characteristic of the different tissue types and not to differgic ences in Z eff. Thus, for the limited range of biolo tissues, a single scaling f actor can be used to con vert µ(Ex) to µ(Eγ) for lung, li ver, fat, muscle, and other soft tissues. F or spongiosa and cor tical bone, ho wever, the same scale factor will not apply because of the significant calcium and phosphorous content of bone tissue that result in Z eff different from other tissues. This issue has been addressed 28 by se gmenting bone from soft tissue at a specif ic threshold and applying different scale f actors to the tw o different tissue classif ications, bone and nonbone, corresponding to different values of Zeff. Kinahan and colleagues28 adopted a threshold of 300 Hounsfield units (HU). Subsequentl y, Watson and colleagues62 proposed a mixture model in w hich all tissues with µ < µ(water) are treated as a mixture of air and w ater at v arious concentrations (k), while tissues with µ > µ(water) are treated as a mixture of water and cortical bone. Since this approach limits the composition to a single value for a given µ(Ex), a bilinear scaling function can be defined for biologic tissues, as sho wn in Figure 5. Recent pub lications on CT -based attenuation cor rection for PET also propose a break point at 0 HU ( µ value for w ater)63 although the most appropriate choice ma y be slightl y greater than zero because some soft tissues and b lood

µ(Eγ) (cm−1)

µmix

air–water

k . µcb (1

k) . µw

water–bone

µ(EX) (HU) Figure 5. The bilinear scaling function used to convert CT numbers (Hounsfield units) to linear attenuation values at 511 keV. The attenuation correction factors are generated by reprojecting the µ-map at 511 keV; w = water and cb = cortical bone; k is the concentration of the components of the mixture.

conform to the air–water mix but with densities greater than water for which a break point around 60 HU is more appropriate. The calibration function has been deri ved from phantom measurements and has also been v alidated with patient data. 64 The calibration of the CT scanner ensures that the soft tissue v alues (µ < 60 HU) are independent of the kVp setting of the X-ra y tube. This independence does not apply to bone-like tissue with µ > 60 HU, and therefore, different lines (slopes) are required for each kVp setting. 65 The CT is acquired before the emission data so the ACFs can be generated for the entire v olume. The CT images at ~70 keV are resampled to the spatial resolution of the emission data. Then, the images are scaled v oxel-byvoxel to the ener gy of the emission data b y applying the bilinear scaling function (see Figure 5). The scaled CT images are then forw ard projected to generate ACFs that match the sampling of the emission data. Since the introduction of the PET/CT scanner , CT -based attenuation correction has been a signif icant focus of research to address the v arious possib le ar tifacts. The follo wing sections will re view the status of this w ork and the outstanding challenges that remain for CT -based attenuation correction.

Artifacts Specific to CT-Based Attenuation Correction While the benefits of CT-based attenuation are now well known and documented , a number of challenges ha ve emerged as the technique has become more widel y adopted for PET/CT.66,67 There are tw o main concer ns: (1) the presence of materials in the patient with Z eff values that do not confor m to the basic assumptions in the bilinear model and (2) mismatch between the CT and PET due to patient respiration, cardiac motion, and bowel movement.68 Since the f irst commercial PET/CT installation in 2001, these issues have received considerable attention. Examples of the f irst concer n include metallic objects, 69,70 dental hardware,71 calcified lymph nodes, and intravenous72,73 and oral contrast. 74,75 Materials with high Zeff may even exceed the dynamic range of attenuation v alues measurab le b y CT , and se vere artifacts can be generated in the images. Of par ticular importance in the assessment of head and neck cancer is the presence of dental f illings.71 A number of metal artifact reduction techniques ha ve been in vestigated,76 including modif ied reconstr uction methods 77 and segmentation approaches 78 that can signif icantly reduce the artifacts.

Imaging of Structure and Function with PET/CT

Some characteristic artifacts associated with CT-based attenuation cor rection are sho wn in F igure 6. When slow regular (tidal) breathing is adopted for both CT and PET, respiration ef fects include an apparent displacement of the dome of the liver into the lower lobe of the right lung79 (Figure 6A) creating a cor responding region of apparent activity on the PET scan (ar row). A cur ved re gion of apparent low uptak e (photopenic) at the top of the li ver and spleen in the PET image (Figure 6B) is also observed in some studies. Although such ar tifacts ma y occur for any patient follo wing a tidal breathing protocol, 80 the documented incidence is much reduced for faster, higher performance CT scanners. F igure 6C is an e xample of a study acquired on a 6-slice CT scanner that sho ws no evidence of breathing ar tifacts or misre gistration. The clinical signif icance of respirator y artifacts has been studied for an earl y PET/CT design in a series of 300 patients81 and was found to result in around 2% of incorrect diagnoses.

enhance CT v alues without a cor responding change in density, and it is used in CT to enhance attenuation v alues in the v asculature b y increasing the photoelectric absorption compared with the blood. CT contrast results in a 40% change in attenuation at CT ener gies, whereas, at 511 k eV where the photoelectric ef fect is ne gligible, the presence of contrast has onl y a 2% ef fect or less on attenuation.82 However, if contrast-enhanced tissue pixels are misidentif ied as a w ater–bone mix, the scaling factor will be incorrect and the erroneously scaled pixels may generate ar tifacts in the PET image 83 (Figure 6D, top row). Tens of thousands of PET/CT scans ha ve now been perfor med in the presence of intra venous contrast and experience has shown that contrast administration does not generally cause a problem that could potentially interfere with the diagnostic value of PET/CT.72,84,85 This is lar gely due to the f act that intra venous contrast is fairly rapidly dispersed throughout the v ascular system. An exception may be the passage of the contrast bolus through a major v essel although e ven this does not always generate an artifact on the PET image (Figure 6D, bottom row). Optimized CT protocols ha ve been de veloped for the administration of intra venous contrast that avoid most of the issues. 86 A recent pub lication87 has documented the rate of tw o of the 100 patients studied , where an incor rect management decision w ould ha ve been made because of the use of noncontrast, lo w-dose

Intravenous Contrast

The use of intra venous contrast ma y be indicated w hen the CT scan is performed for clinical pur poses as opposed t o l ow-dose C T p erformed f or a ttenuation correction and localization only. Intravenous contrast contains iodine at concentrations high enough to

A

D

19

B

C

E

F

Figure 6. Potential image artifacts generated from CT-based attenuation correction: A, an artifact due to respiration in which the dome of the liver is displaced into the base of the right lung, B, curved photopenic areas above the liver and the spleen caused by CT and PET mismatch from respiratory movement of the diaphragm, C, an example of a well-registered study that is free of artifacts, D, the variable effects of intravenous contrast showing an artifact on the PET image (top row) due to a contrast bolus and the absence of an artifact on PET (bottom row), E, the effect of oral contrast where the presence of contrast in the GI tract does not cause an artifact on the PET image (arrow), and F, the effect of dental fillings on the CT and PET images.

20

MOLECULAR IMAGING: PRINCIPLES AND PRACTICE

CT acquired for localization, and CT -AC onl y. In one case, a cystoscopy was performed in a patient with l ymphoma to e valuate a mass near the b ladder missed b y both the CT and PET scans of the PET/CT ; the mass, seen on the diagnostic clinical CT, was worrisome for bladder cancer , but the c ystoscopy w as nor mal and the mass responded to treatment for lymphoma. The second management change also involved a patient with lymphoma, where a colonoscopy and surgical removal of the neoterminal ileum was required to diagnose nodules due to Crohn’s disease. The nodules w ere well seen on the diagnostic clinical CT but not on the CT or PET scans of the PET/CT. Oral Contrast

Oral contrast is administered to enhance the gastrointestinal tract and the distribution of the contrast material is some what v ariable, both in spatial distrib ution and level of enhancement. Modifications to the basic scaling algorithm ha ve been introduced to distinguish oral contrast-enhanced pix els from bone pix els. 82 As with intra venous contrast, there is little e vidence that the presence of oral contrast results in diagnostic errors of any signif icance.88 Figure 6E shows a patient imaged with oral contrast; enhancement of the colon on the CT image (left; ar rows) sho ws no cor responding artifactual uptake on the PET image (right). Ne vertheless, in some protocols, contrast CT is perfor med in addition to the lo w-dose CT for attenuation cor rection and localization, thereb y increasing the radiation dose to the patient. Ho wever, a lo w-dose w hole-body CT in addition to a clinical CT with contrast over a limited axial range (single PET bed position) ma y involve less radiation dose than a w hole-body clinical CT with contrast. Metal Implants

Dental ar tifacts can be cor rected on CT through the use of no vel reconstr uction techniques, 77 as sho wn in Figure 6F. The uncor rected (left) and cor rected (right) images for CT (top) and PET (bottom) demonstrate that although the reconstr uction algorithm signif icantly improves the CT image, it has v ery little impact on the PET image, v erifying that CT-based attenuation cor rection i s a ctually a r obust t echnique. While m etallic implants, such as ar tificial hip prostheses, can sometimes cause artifacts on CT, this appears to be due more to patient mo vement between the CT and the PET scan

than to the presence of prostheses per se , as demonstrated by Kaneta and colleagues. 89 Even so, it would be somewhat rare for the specif ic pathology under study to be located in the re gion af fected b y ar tifacts from the prosthesis. The nonattenuation corrected image is, in any case, available to resolve ambiguities. Respiratory Motion

Within the past 6 years, the most widely addressed issue related to CT-based attenuation cor rection has been respiratory motion90–93 and the artifacts created by the mismatch betw een CT and PET .94 Rotating 68Ge sources used in conventional PET scanners resulted in a transmission scan that a veraged patient respiration in a w ay compatible with the corresponding emission scan. The use of CT-AC suggests a number of dif ferent protocols that must be in vestigated to resolv e the mismatch problem. For example, the advent of fast, spiral CT scanners made breath-hold CT a reality although clinical images are typicall y acquired with full inspiration to separate lung structures. Such an expansion of the chest does not match a PET scan acquired with shallo w breathing and results in serious attenuation cor rection ar tifacts in the anterior chest w all. The appearance of ar tifacts due to respiratory motion and the spatial and temporal mismatch between CT and PET images has led to an intensive research initiative to identify the best respiratory protocol. A number of dif ferent protocols ha ve been investigated, including: • Continuous shallow breathing for both CT and PET 90 • CT acquired with limited breath-hold o ver diaphragm90 • Breath-hold CT acquired with par tial inspiration90 • Motion-averaged CT over many respiration cycles95,96 • Cine CT acquiring full breathing c ycle per slice 97 • Respiratory-gated CT; PET with shallow breathing98 • Deep inspiration breath-hold 99,100 • Breath-hold CT; gated PET 101,102 • Respiratory-gated CT and PET.103 Currently, the simplest and most widel y used protocol is shallo w breathing for both CT and PET .90 Early single- or dual-slice PET/CT designs showed a high incidence of breathing ar tifacts (F igures 6A and 6B) 80 but with the incorporation of fast MDCT into PET/CT scanners, the incidence of such ar tifacts has been considerably reduced. Ho wever, the CT images still do not exactly match the motion-averaged PET acquisitions

Imaging of Structure and Function with PET/CT

and protocols, such as slo w CT acquisition, ha ve also been in vestigated. The clinical signif icance of these attenuation cor rection ef fects continues to be debated , particularly with respect to lesions in the base of the lung and dome of the liver, where curved photopenic areas are observed (Figure 6B). Displacement of such lesions may result in incor rect localization or , w orse, a f ailure to identify them correctly leading to misdiagnosis. Shallow breathing during PET/CT has been sho wn to be inadequate for the comprehensi ve staging of lung cancer .104 Nevertheless, a signif icant percentage of studies acquired on even a 6-slice CT scanner sho w good registration with shallow breathing. Finally, two other effects can also influence the accuracy of CT-AC: the tr uncation of the transv erse FOV105 and the presence of scattered radiation. Truncation of the FOV arises because CT scanners typicall y have a 50 cm diameter FOV, whereas PET supports 60 cm. Simple software e xtrapolation techniques ha ve pro ved ef fective in extending the CT FOV to match that of PET, at least with accuracy adequate for CT -AC.106,107 Scatter is enhanced by imaging with the arms of the patient in the FOV. However, the short scan times achievable with state-of-the-art PET/CT allo w patients to easil y tolerate imaging with arms raised, reducing the effects due to increased scatter. The exception is head and neck cancer, where the patient is scanned with arms down. Despite the issues discussed above and rare opinions to the contrar y,108 CT-based attenuation cor rection has become the de facto standard for PET/CT although it can be affected by the ar tifacts described abo ve. The advantages, w hich include con venience and shor t acquisition times, largely outweigh the drawbacks. In a small number of studies, quantitati ve comparisons ha ve been made between ACFs generated from standard PET transmission scans and from CT 92,109,110; and although some differences in SUV values have been noted, nothing of diagnostic significance has been documented.

RADIATION DOSE CONSIDERATIONS Patient exposure to radiation from a PET/CT scan is both external from the CT scan and inter nal from the PET injected radionuclide.111

External Dose Dose assessment in CT is challenging and depends not only on the body re gion exposed but also on a v ariety of scan-specif ic parameters including tube potential

21

(kVp), the product of tube cur rent and e xposure time (in milliamp × seconds, mAs), slice collimation, and pitch.112 In addition, the dose also depends on certain technical features of the scanner , such as beam f iltration, beam shaping f ilter, geometry and the acquisition algorithm. Therefore, values for CT patient dose v ary considerably between centers and between scanners. The tendenc y is to o versimplify the situation b y not taking all of these factors into account. For whole body CT scans that extend from the level of the thyroid to the pubic symphysis, the effective CT dose Eext can be estimated approximately as: Eext = Γ CT · CTDIvol where Γ CT = 1.47 mSv/mGy is the dose coef ficient that relates the volume CT dose index CTDIvol to the effective dose. F or a typical set of clinical scan parameters, the CTDIvol is 13 mGy 113 resulting in a total effective wholebody dose of 19 mSv. However, many centers acquire the CT scan for attenuation cor rection and localization onl y, reducing the whole-body dose to as low as 3 mSv or less. In addition, there are a number of strate gies to make better use of the radiation, such as tube cur rent modulation and automatic exposure control.114,115

Internal Dose The internal radiation dose will depend upon the biodistribution and the ph ysical and biolo gic half-life of the biomarker. The dose is e xpressed as the radiation e xposure to the whole body and individually to the various organs. The critical organs are those that receive the maximum radiation dose. The ef fective dose Eint resulting from intra venous administration of a gi ven biomark er with activity A can be estimated from: Eint = Γ · A where Γ is a dose coef ficient computed for the adult hermaphrodite MIRD phantom. The only clinical biomarker of interest is FDG for w hich the dose coef ficient is 19 µSv/MBq,116 although a higher dose coef ficient of 29 µSv/MBq has also been published.117 The dose coefficient holds for standard patients with a body w eight of about 70 kg and is generic rather than patient specif ic since age, sex of patients, and individual pharmacokinetics are not taken into account. In fact, the radiation risk is somewhat higher for females and for younger patients when compared with males and older patients. Age and sex-specific dose coefficients can be found elsewhere.118

22

MOLECULAR IMAGING: PRINCIPLES AND PRACTICE

On the basis of the published value116 for the dose coefficient, the average whole-body dose for a typical 10 mCi (370 MBq) injection of FDG is 7 mSv . However, most biomarkers do not distribute unifor mly in the body , and the critical or gan with FDG, for e xample, is the b ladder due to excretion through the urinary system.

Total Radiation Dose The total ef fective dose for PET/CT is the sum of the internal and e xternal doses. F or a full y clinical CT and FDG-PET scan, the effective dose will be around 25 mSv. However, this can be reduced to 10 mSv or less w hen a low-dose CT is acquired for localization and attenuation correction only. In practice, the PET/CT dose to a specific organ will depend upon the exact protocol; for example, if the CT scan does not include the b ladder, the dose to the bladder wall will be due entirely to FDG. For a smaller patient imaged on a high-sensitivity scanner, a lower FDG dose can be used, potentially limiting the effective dose to 5 mSv or less. The worldwide average annual dose due to the natural radioactive background is 2.4 mSv.

THE CLINICAL ROLE OF PET/CT Prior to the introduction of PET/CT, essentially all multimodality clinical imaging w as based on softw are fusion techniques, 14 limited mainl y to the brain. The

introduction of the Ha wkeye scanner (GE Healthcare) in 1999, follo wed less than 2 y ears later b y the f irst commercial PET/CT scanner , has ir reversibly transformed the field of multimodality imaging. From 2001, the sales of PET -only scanners decreased to zero by 2006, completel y replaced b y PET/CT (F igure 7). Currently, in 2008, a w orldwide installed base of o ver 2500 units attests to the rapid adoption of the modality by physicians. The majority of this installed base is in routine clinical operation and there is, at least for oncolo gy, now a g rowing body of literature that suppor ts the accuracy of staging and restaging with PET/CT compared with either CT or PET acquired separately.38,119 Many of these pub lications are within the past 3 or 4 years, and they clearly document the significant impro vements in specif icity and to some e xtent also in sensiti vity, and especiall y in earl y detection of cancer recur rence.120 These improvements are incremental when compared with PET that alone demonstrates high le vels of sensiti vity and specif icity for a wide range of disease states. Impro ved accurac y has been documented for a v ariety of cancers including head and neck, 76,121 thyroid,122 lung,123–125 breast,126,127 esophageal, 128,129 colorectal,15,130 and melanoma. 131 There is also evidence that PET/CT improves accuracy in l ymphoma132 and solitar y pulmonar y nodules, 133,134 in spite of the f act that in l ymphoma the accurac y of PET alone is very high. 135

Figure 7. Shipments of PET and PET/CT scanners for the US market as recorded by the Nuclear Equipment Manufacturers Association (NEMA) for the period January 2002 to October 2007. Note that the figures (in $M) reflect the total revenue for all shipments from which the selling price and individual unit type cannot be determined. Shipments of PET-only scanners declined during this period to zero from January 2006 onwards. The overall market for PET or PET/CT remained fairly constant throughout this period, although since January 2007, with the reduction in reimbursement due to the introduction of the Deficit Reduction Act, sales have declined somewhat.

Imaging of Structure and Function with PET/CT

A

B

Figure 8. Two studies acquired on a Biograph 6 TruePoint TrueV PET/CT scanner. Transaxial sections are for PET (top row) and fused images (bottom row): A, a 50-year-old female patient with a diagnosis of pancreatic cancer (arrow). The images were acquired 94 min after injection of 10.3 mCi of 18FDG. The total scan duration was 10 min with acquisition of five bed positions at 2 min per position. The CT was acquired at 130 kVp and 180 mAs (Siemens CAREDose). B, a 58-year-old female patient with metastatic renal cell cancer (arrow) from an unknown primary. The images were acquired 110 min after injection of 9.7 mCi of 18FDG. The total scan duration was 15 min with acquisition of five bed positions at 3 min per position. The CT was acquired at 130 kVp and 180 mAs (Siemens CAREDose).

In summary, therefore, the impro vement in accurac y of PET/CT compared with PET or CT for staging and restaging is statisticall y signif icant and a verages 10 to 15% over all cancers.38 To illustrate typical state-of-the-art PET/CT scans, Figure 8 shows two studies acquired on a Biograph 6 TruePoint TrueV PET/CT (Siemens Molecular Imaging) with a 21.6 cm axial FO V. Figure 8A shows transaxial PET and fused images of a 50-y ear-old female patient with a diagnosis of pancreatic cancer . The images were acquired 94 minutes after injection of 10.3 mCi of 18 FDG. The total scan duration w as 10 minutes with acquisition of five bed positions at 2 minutes per position. The CT was acquired at 130 kVp and 180 mAs (Siemens CAREDose). The images demo nstrate intense focal uptake of 18FDG in a primar y neoplasm 3.4 × 2.6 cm in size that can be accurately located in the head of the pancreas (arrow). No FDG uptake was identified in any of the proximal nodes although the lik elihood of micrometastases w ould be high. F igure 8B sho ws a 58-y ear-old female patient with metastatic renal cell cancer from an unknown primary. The images were acquired 110 minutes after injection of 9.7 mCi of 18FDG. The total scan duration was 15 minutes with acquisition of five bed positions at 3 minutes per position. The CT w as acquired at 130 kVp and 180 mAs (Siemens CAREDose). The study

23

demonstrates a lar ge FDG a vid peripherall y-enhancing necrotic mass occup ying the anterior mid and lo wer left kidney. The mass is 10 cm in size and appears to in volve the lower pole collecting system (ar row). Another application for which PET/CT is also having an impact is that of radiotherap y treatment planning. Incorporation of FDG-PET images into therapy planning was already taking place prior to the introduction of PET/CT32 using softw are fusion techniques. 136,137 In some cases, the a vailability of the PET images led to a change in treatment plan by redefining the biologic target volume based on FDG uptak e. This w as par ticularly effective for the lung, 138 where reactive changes, such as a par tial or full collapsed lung (atelectasis), could be distinguished from malignanc y as a result of the differential uptak e of FDG. Reasonab le re gistration accuracy at the centimeter level could be achieved locally through the use of f iducials although the software fusion techniques were cumbersome and labor -intensive. From the inception, PET/CT pro vided more con venient and routine access to fused CT and PET images and earl y assessment of the consequences of using PET/CT in planning139–141 was encouraging. More recent surveys18,142 have confirmed the earlier conclusions. Molecular imaging with PET/CT is increasingl y being used to monitor response to therap y,143 for chemotherapy,144–147 for radiation therap y,148,149 and for combinations of each. 150 It has become increasingly evident that simple response e valuation criteria for solid tumors (RECIST) 151 based on anatomic measures of tumor size may not be adequate to accuratel y assess therapy response. The molecular signal is lik ely to be a more sensitive indicator as it reflects tumor metabolism rather than just tumor size. A metabolic change ma y be more suggestive of a response than a size change. The combination of ha ving both CT and PET for monitoring response offers a number of unique possibilities in spite of the technical difficulties associated with CT-based attenuation cor rection. Firstly, the anatomic and the functional volume of the tumor can be estimated, the former from CT measurements and the latter by summing all voxels with a Standardized Uptake Value (SUV) above a threshold that defines malignanc y. Therapy response can be assessed from changes in both these metrics or from a change in the total lesion glycolosis that is calculated as the product of the a verage SUV in the tumor and the v olume.152 The advantage of the CT is that an accurate measurement of tumor volume is available both before and after treatment. It is also helpful and more reliab le to def ine the tumor region-of-interest (R OI) directl y on the CT and then to transfer the same ROI onto the PET image. The boundary

24

MOLECULAR IMAGING: PRINCIPLES AND PRACTICE

of the tumor ma y be difficult to deter mine from the PET scan, par ticularly for metabolic responders as the lesion SUV decreases. The CT images ma y also be used to improve partial volume correction, such as by dividing the SUV from the PET image with a reco very coef ficient based on the spherical tumor diameter. Since tumors generally have a complex shape, a more sophisticated par tial volume cor rection method is desirab le.153 Thus, for both technical and practical reasons, PET/CT is continuing to successfully promote the use of PET for monitoring response to different forms of therapy. Cardiac PET/CT applications are in their inf ancy154 and ha ve recentl y encountered a number of dif ficulties. Obviously, the effects of cardiac and respiratory motion are critical for these studies. The problems of mismatch associated with CT -based attenuation cor rection discussed above are potentially more serious for cardiac studies than they are for oncolo gy in that all cardiac studies will be affected rather than just those w hole-body studies with lesions in cer tain sensitive regions, such as the lung. This misregistration results in w hat appears to be perfusion deficits in se gments of the hear t associated with the misalignment. A recent publication155 finds that up to 40% of cardiac PET/CT studies could be af fected by misregistration. A number of different strategies are being de veloped to address this issue, including (1) manual realignment of CT and PET , (2) acquiring a cine CT of the breathing motion and generating an a verage CT for attenuation correction, and (3) acquiring multiple CT scans to ensure at least one matches the PET scan as closel y as possib le. Obviously, the role of PET/CT in cardiolo gy has yet to be determined, but if a strong clinical demand e xists, it is to be expected that transient technical challenges, such as the misalignment issue, will ultimately be solved. A complete review of the status of PET/CT in cardiology can be found in DiCarli and Lipton. 156 Recently, the f irst human MR/PET design has been evaluated following the de velopment of MR-compatib le PET detectors. 157 The design, w hich comprises a PET detector ring inser ted into a 3-T MR, has been used for simultaneous MR and FDG-PET imaging of the human brain.158 While ultimately the aim is to develop a wholebody MR/PET that could f ind applications in oncolo gy and cardiology, cur rent devices are limited to the brain. There have been suggestions that MR/PET could eventually replace PET/CT ,159 although the question is reall y whether adding a PET inser t to MR will attract clinical applications a way from PET/CT . This seems unlik ely since both CT and MR have strengths in specif ic clinical areas, a situation that will probably not change because of the addition of PET . The realization of the full potential

of MR/PET must await the development of a satisfactory whole-body design.

CONCLUSION There is little doubt that, o ver the past 6 y ears, PET/CT has had a growing impact on clinical imaging and particularly s taging a nd r estaging d isease a nd m onitoring response to therap y. Although the technolo gy has been somewhat disruptive in the sense that it has brought together medical specialties that ha ve not traditionall y worked together, the overall impact has been positive. To meet the demand for cross-training of both the technologists w ho operate the de vices and the ph ysicians w ho interpret the studies, guidelines ha ve been pub lished160 and new standards established leading to a somewhat different situation today from the way radiology and nuclear medicine ha ve traditionall y functioned. This trend is likely to continue as other multimodality de vices reach the clinic, including SPECT/CT that w as introduced in 2004 and MR/PET that is cur rently under evaluation for brain imaging.

ACKNOWLEDGMENT This chapter is dedicated to the memor y of Professor Bruce Hase gawa from the Uni versity of Califor nia, San Francisco, a pioneer of the combined imaging of structure and function. He was a friend and colleague and his work was an inspiration to us all.

REFERENCES 1. Wagner HN. A personal histor y of nuclear medicine. London, UK: Springer; 2006. 2. Levin DN, Pelizzari CA, Chen GT, et al. Retrospecti ve geometric correlation of MR, CT, and PET images. Radiology 1988;169:817–23. 3. Muehllehner G, Kar p JS. P ositron emission tomo graphy. Phys Med Biol 2006;51:R117–37. 4. Hawkes DJ, Hill DL, Hallpik e L, Baile y DL. Core gistration of str uctural and functional images. In: Valk P, Bailey DL, Townsend DW, Maisey M, editors. P ositron emission tomo graphy: basic science and clinical practice. New York: Springer-Verlag; 2003. p. 181–98. 5. Slomka PJ. Software approach to mer ging molecular with anatomic information. J Nucl Med 2004;45 Suppl 1:S36–45. 6. Pelizzari CA, C hen GT , Spelbring DR, et al. Accurate threedimensional re gistration of CT , PET, and/or MR images of the brain. J Comput Assist Tomogr 1989;13:20–6. 7. Pietrzyk U , Herholz K, Heiss WD. Three-dimensional alignment of functional and morphological tomograms. J Comput Assist Tomogr 1990;14:51–9. 8. Mountz JM, Zhang B , Liu HG, Inampudi C. A reference method for correlation of anatomic and functional brain images: validation and clinical application. Semin Nucl Med 1994;24:256–71. 9. Woods RP, Mazziotta JC, Cher ry SR. MRI-PET re gistration with automated algorithm. J Comput Assist Tomogr 1993;17:536–46.

Imaging of Structure and Function with PET/CT

10. Maes F, Collignon A, Vandermeulen D, et al. Multimodality image registration by maximization of mutual infor mation. IEEE Trans Med Imaging 1997;16:187–98. 11. Wahl RL, Quint LE, Cieslak RD, et al. “Anatometabolic” tumor imaging: fusion of FDG PET with CT or MRI to localize foci of increased activity. J Nucl Med 1993;34:1190–7. 12. Hamilton RJ, Blend MJ, Pelizzari CA, et al. Using v ascular structure for CT -SPECT re gistration in the pelvis. J Nucl Med 1999; 40:347–51. 13. West J, Fitzpatrick JM, Wang MY, et al. Comparison and evaluation of retrospective intermodality brain image registration techniques. J Comput Assist Tomogr 1997;21:554–66. 14. Hutton BF, Braun M. Softw are for image re gistration: algorithms, accuracy, efficacy. Semin Nucl Med 2003;33:180–92. 15. Kim JH, Czer nin J, Allen-Auerbach MS, et al. Comparison betw een 18 F-FDG PET, in-line PET/CT, and softw are fusion for restaging of recurrent colorectal cancer. J Nucl Med 2005;46:587–95. 16. Slomka PJ, Dey D, Przetak C, et al. Automated 3-dimensional registration of stand-alone 18F-FDG whole-body PET with CT. J Nucl Med 2003;44:1156–67. 17. Dizendorf E, Cier nik IF , Baumer t B , et al. Impact of inte grated PETCT scanning on e xternal beam radiation treatment planning. J Nucl Med 2002;43:33P. 18. Gregoire V, Haustermans K, Geets X, et al. PET-based treatment planning in radiotherapy: a new standard? J Nucl Med 2007;48 Suppl 1:S68–77. 19. Hasegawa BH, Gingold EL, Reilly SM, et al. Description of a simultaneous emi ssion-transmission CT system. Proc SP IE 1990; 1231:50–60. 20. Hasegawa BH, Stebler B, Rutt BK, et al. A prototype high-purity germanium detector system with f ast photon-counting circuitr y for medical imaging. Med Phys 1991;18:900–9. 21. Lang TF, Hasegawa BH, Lie w SC, et al. Description of a prototype emission-transmission computed tomo graphy imaging system. J Nucl Med 1992;33:1881–7. 22. Hasegawa BH, Lang TF, Brown EL, et al. Object specif ic attenuation correction of SPECT with correlated dual-energy X-ray CT. IEEE Trans Nucl Sci 1993;NS-40:1242–52. 23. Kalki K, Blank espoor SC, Brown JK, et al. Myocardial perfusion imaging with a combined X-ra y CT and SPECT system. J Nucl Med 1997;38:1535–40. 24. Blankespoor SC, Wu X, Kalki K, et al.Attenuation correction of SPECT using X-ray CT on an emission-transmission CT system: myocardial perfusion assessment. IEEE Trans Nucl Sci 1996;43:2263–74. 25. Tang HR, Da Silv a AJ, Mattha y KK, et al. Neurob lastoma imaging using a combined CT scanner-scintillation camera and 131I-MIBG. J Nucl Med 2001;42:237–47. 26. Beyer T, Kinahan PE, Townsend DW, Sashin D. The use of X-ray CT for attenuation correction of PET data. Nuclear Science Symposium and Medical Imaging Conference, 1994. 1994 IEEE Conference Record 1994;4:1573–7. 27. Beyer T, Townsend DW, Brun T, et al. A combined PET/CT scanner for clinical oncology. J Nucl Med 2000;41:1369–79. 28. Kinahan PE, Townsend DW, Beyer T, Sashin D. Attenuation correction for a combined 3D PET/CT scanner. Med Phys 1998;25:2046–53. 29. Charron M, Be yer T, Bohnen NN , et al. Image anal ysis in patients with cancer studied with a combined PET and CT scanner . Clin Nucl Med 2000;25:905–10. 30. Kluetz PG, Meltzer CC, Villemagne VL, et al. Combined PET/CT imaging in oncolo gy. Impact on patient management. Clin Positron Imaging 2000;3:223–30. 31. Meltzer CC, Luk etich JD , F riedman D , et al. Whole-body FDG positron emission tomo graphic imaging for staging esophageal cancer comparison with computed tomo graphy. Clin Nucl Med 2000;25:882–7. 32. Caldwell CB, Mah K, Ung YC, et al. Observer variation in contouring gross tumor volume in patients with poorly defined non-small-cell

33.

34.

35.

36.

37.

38.

39.

40.

41.

42. 43.

44.

45.

46.

47.

48.

49.

50.

51. 52. 53.

25

lung tumors on CT : the impact of 18FDG-hybrid PET fusion. Int J Radiat Oncol Biol Phys 2001;51:923–31. Patton JA, Delbeke D, Sandler MP. Image fusion using an inte grated, dual-head coincidence camera with X-ra y tube-based attenuation maps. J Nucl Med 2000;41:1364–8. Bocher M, Balan A, Krausz Y, et al. Gamma camera-mounted anatomical X-ray tomography: technology, system characteristics and first images. Eur J Nucl Med 2000;27:619–27. Brambilla M, Secco C, Dominietto M, et al. Performance characteristics obtained for a ne w 3-dimensional lutetium o xyorthosilicatebased whole-body PET/CT scanner with the National Electrical Manufacturers Association NU 2-2001 standard. J Nucl Med 2005;46:2083–91. Mawlawi O, Podoloff DA, Kohlmyer S, et al. Performance characteristics of a ne wly developed PET/CT scanner using NEMA standards in 2D and 3D modes. J Nucl Med 2004;45:1734–42. Surti S, Kuhn A, Werner ME, et al. Performance of Philips Gemini TF PET/CT scanner with special consideration for its time-of-flight imaging capabilities. J Nucl Med 2007;48:471–80. Czernin J, Allen-Auerbach M, Schelbert HR. Improvements in cancer staging with PET/CT : literature-based e vidence as of September 2006. J Nucl Med 2007;48 Suppl 1:S78–88. Kalender WA, Seissler W, Klotz E, Vock P. Spiral volumetric CT with single-breath-hold technique, continuous transpor t, and continuous scanner rotation. Radiology 1990;176:181–3. Schardt P, Deuringer J, Freudenberger J, et al. New X-ray tube performance in computed tomography by introducing the rotating envelope tube technology. Med Phys 2004;31:2699–706. Kalender WA. Technical concepts. In: Computed tomo graphy: fundamentals, system technology, image quality, applications. Erlangen, Germany: Publicis Corporate Publishing; 2005. p. 47. Takagi K, Fukazawa T. Cerium-activated Gd2SiO5 single crystal scintillator. Appl Phys Lett 1983;42:43–5. Melcher CL, Schw eitzer JS. Cerium-doped lutetium o xy-orthosilicate: a fast, efficient new scintillator. IEEE Trans Nucl Sci 1992; 39:502–5. Townsend D W, Isoardi RA, Bendriem B . Volume imaging tomographs. In: Bendriem B , Townsend DW, editors. The theor y and practice of 3D PET. Dordrecht, The Netherlands: Kluwer Academic Publishers; p. 111–32. Watson CC, Casey ME, Bendriem B , et al. Optimizing injected dose in clinical PET by accurately modeling the counting-rate response functions specif ic to indi vidual patient scans. J Nucl Med 2005; 46:1825–34. Kemp BJ, Kim C, Williams JJ, et al. NEMA NU 2-2001 performance measurements of an LYSO-based PET/CT system in 2D and 3D acquisition modes. J Nucl Med 2006;47:1960–7. Lodge MA, Badawi RD, Gilbert R, et al. Comparison of 2-dimensional and 3-dimensional acquisition for 18F-FDG PET oncolo gy studies performed on an LSO-based scanner. J Nucl Med 2006;47:23–31. Strobel K, Rudy M, Treyer V, et al. Objective and subjective comparison of standard 2-D and fully 3-D reconstructed data on a PET/CT system. Nucl Med Commun 2007;28:555–9. Jakoby BW, Bercier Y, Watson CC, et al. Ph ysical perfor mance and clinical w orkflow of a ne w LSO HI-REZ PET/CT scanner . Nuclear Science Symposium Conference Record , 2006 IEEE 2006;5:3130–4. Townsend DW, Jakoby B, Long MJ, et al. Performance and clinical workflow of a new combined PET/CT scanner. J Nucl Med 2007; 48:437P. Budinger TF. Time-of-flight positron emission tomography: status relative to conventional PET. J Nucl Med 1983;24:73–8. Kinahan PE, Rodgers JG. Analytic 3D image reconstruction using all detected events. IEEE Trans Nucl Sci 1989;36:964–8. Defrise M, Kinahan PE, Townsend DW, et al. Exact and appro ximate rebinning algorithms for 3-D PET data. IEEE Trans Med Imaging 1997;16:145–58.

26

MOLECULAR IMAGING: PRINCIPLES AND PRACTICE

54. Hudson H, Larkin R. Accelerated image reconstruction using ordered subsets of projection data. IEEE Trans Med Imaging 1994; 13:601–9. 55. Comtat C, Kinahan PE, Defrise M, et al. F ast reconstr uction of 3D PET data with accurate statistical modeling. IEEE Trans Nucl Sci 1998;45:1083–9. 56. Comtat C, Bataille F, Michel C, et al. OSEM-3D reconstr uction strategies for the ECAT HRRT. Nuclear Science Symposium Conference Record, 2004 IEEE 2004;6:3492–6. 57. Liu X, Comtat C, Michel C, et al. Comparison of 3-D reconstr uction with 3D-OSEM and with FORE+OSEM for PET . IEEE Trans Med Imaging 2001;20:804–14. 58. Panin VY, Kehren F, Michel C, Casey M. Fully 3-D PET reconstr uction with system matrix derived from point source measurements. IEEE Trans Med Imaging 2006;25:907–21. 59. Popescu LM, Matej S, Lewitt RM. Iterative image reconstruction using geometrically ordered subsets with list-mode data. Nuclear Science Symposium Conference Record, 2004 IEEE 2004;6:3536–40. 60. Daube-Witherspoon ME, Matej S, Kar p JS, Le witt RM. Application of the ro w action maximum lik elihood algorithm with spherical basis functions to clinical PET imaging. IEEE Trans Nucl Sci 2001;48:24–30. 61. Kinahan PE, Hase gawa BH, Be yer T. X-ray-based attenuation correction for positron emission tomography/computed tomography scanners. Semin Nucl Med 2003;33:166–79. 62. Watson CC, Townsend D W, Bendriem B . PET/CT systems. In: Wernick M, Aarsvold J, editors. Emission tomo graphy. London, UK: Elsevier Science; 2004. p. 195–212. 63. Burger C, Goerres G, Schoenes S, et al. PET attenuation coefficients from CT images: e xperimental evaluation of the transfor mation of CT into PET 511-k eV attenuation coef ficients. Eur J Nucl Med Mol Imaging 2002;29:922–7. 64. Watson CC, Rappopor t V, Faul D, et al. A method for calibrating the CT-based attenuation cor rection of PET in human tissue. IEEE Trans Nucl Sci 2006;53:102–7. 65. Carney JP, Townsend D W, Rappopor t V, Bendriem B . Method for transforming CT images for attenuation cor rection in PET/CT imaging. Med Phys 2006;33:976–83. 66. Cohade C, Wahl RL. Applications of positron emission tomo graphy/computed tomography image fusion in clinical positron emission tomography-clinical use, interpretation methods, diagnostic improvements. Semin Nucl Med 2003;33:228–37. 67. Bockisch A, Be yer T, Antoch G, et al. P ositron emission tomo graphy/computed tomography-imaging protocols, ar tifacts, and pitfalls. Mol Imaging Biol 2004;6:188–99. 68. Nakamoto Y, Chin BB, Cohade C, et al. PET/CT : artifacts caused by bowel motion. Nucl Med Commun 2004;25:221–5. 69. Cohade C, Osman M, Marshall L,Wahl RL. Metallic object artifacts on PET-CT: clinical and phantom studies. J Nucl Med 2002;43:308P. 70. Goerres GW, Ziegler SI, Burger C, et al. Artifacts at PET and PET/CT caused b y metallic hip prosthetic material. Radiolo gy 2003; 226:577–84. 71. Kamel EM, Burger C, Buck A, et al. Impact of metallic dental implants on CT-based attenuation correction in a combined PET/CT scanner. Eur Radiol 2003;13:724–8. 72. Yau YY, Coel M, Chan WS, et al. Application of IV contrast in PETCT: does it reall y produce attenuation cor rection er ror. J Nucl Med 2003;44:272P. 73. Antoch G, F reudenberg LS, Be yer T, et al. To enhance or not to enhance? 18F-FDG and CT contrast agents in dual-modality 18 F-FDG PET/CT. J Nucl Med 2004;45 Suppl 1:S56–65. 74. Carney JP, Beyer T, Brasse D, et al. Clinical PET/CT scanning using oral CT contrast agents. J Nucl Med 2002;45:57P. 75. Cohade C, Osman M, Nakamoto Y, et al. Initial e xperience with oral contrast in PET/CT : phantom and clinical studies. J Nucl Med 2003;44:412–6.

76. Schafers KP, Raupach R, Beyer T. Combined 18F-FDG-PET/CT imaging of the head and neck. An approach to metal ar tifact cor rection. Nuklearmedizin 2006;45:219–22. 77. Lemmens C, Nuyts J, Faul D, et al. Suppression of metal streak ar tifacts in CT using a MAP reconstruction procedure. Nuclear Science Symposium Conference Record, 2006 IEEE 2006;6:3431–7. 78. Mirzaei S, Guerchaft M, Bonnier C, et al. Use of segmented CT transmission map to avoid metal artifacts in PET images by a PET-CT device. BMC Nucl Med 2005;5:3. 79. Sarikaya I, Yeung HW, Erdi Y, Larson SM. Respiratory artefact causing malpositioning of li ver dome lesion in right lo wer lung. Clin Nucl Med 2003;28:943–4. 80. Romer W, Chung M, Chan A, et al. Single-detector helical CT in PET-CT: assessment of image quality . AJR Am J Roentgenol 2004;182:1571–7. 81. Osman MM, Cohade C, Nakamoto Y, et al. Clinicall y signif icant inaccurate localization of lesions with PET/CT : frequency in 300 patients. J Nucl Med 2003;44:240–3. 82. Carney JP, Townsend DW. CT-based attenuation correction for PET/CT scanners. In: von Schulthess GK, editor. Molecular anatomic imaging: PET-CT and SPECT-CT integrated modality imaging. Philadelphia (PA): Lippincott Williams and Wilkins; 2006. p. 54–62. 83. Antoch G, F reudenberg LS, Egelhof T, et al. F ocal tracer uptak e: a potential ar tifact in contrast-enhanced dual-modality PET/CT scans. J Nucl Med 2002;43:1339–42. 84. Berthelsen AK, Holm S, Loft A, et al. PET/CT with intravenous contrast can be used for PET attenuation correction in cancer patients. Eur J Nucl Med Mol Imaging 2005;32:1167–75. 85. Mawlawi O, Erasmus JJ, Munden RF, et al. Quantifying the ef fect of IV contrast media on integrated PET/CT: clinical evaluation. AJR Am J Roentgenol 2006;186:308–19. 86. Brechtel K, Klein M, Vogel M, et al. Optimized contrast-enhanced CT protocols for diagnostic w hole-body 18F-FDG PET/CT : technical aspects of single-phase versus multiphase CT imaging. J Nucl Med 2006;47:470–6. 87. Gollub MJ, Hong R, Sarasohn DM, Akhurst T. Limitations of CT during PET/CT. J Nucl Med 2007;48:1583–91. 88. Otsuka H, Graham MM, Kubo A, Nishitani H. The effect of oral contrast on lar ge bo wel acti vity in FDG-PET/CT . Ann Nucl Med 2005;19:101–8. 89. Kaneta T, Takanami K, Wakayama Y, et al. High-density materials do not always induce artifacts on PET/CT: what is responsible for the difference? Nucl Med Commun 2007;28:495–9. 90. Beyer T, Antoch G, Blodgett T, et al. Dual-modality PET/CT imaging: the ef fect of respirator y motion on combined image quality in clinical oncology. Eur J Nucl Med Mol Imaging 2003;30:588–96. 91. Osman MM, Cohade C, Nakamoto Y, Wahl RL. Respirator y motion artifacts on PET emission images obtained using CT attenuation correction on PET-CT. Eur J Nucl Med Mol Imaging 2003; 30:603–6. 92. Papathanassiou D, Becker S, Amir R, et al. Respiratory motion artefact in the liver dome on FDG PET/CT: comparison of attenuation correction with CT and a caesium e xternal source. Eur J Nucl Med Mol Imaging 2005;32:1422–8. 93. Bacharach SL. PET/CT attenuation cor rection: breathing lessons. J Nucl Med 2007;48:677–9. 94. Beyer T, Rosenbaum S, Veit P, et al. Respiration ar tifacts in w holebody 18F-FDG PET/CT studies with combined PET/CT tomographs emplo ying spiral CT technolo gy with 1 to 16 detector rows. Eur J Nucl Med Mol Imaging 2005;32:1429–39. 95. Pan T, Mawlawi O, Nehmeh SA, et al. Attenuation correction of PET images with respiration-a veraged CT images in PET/CT . J Nucl Med 2005;46:1481–7. 96. Chi PC, Mawlawi O, Nehmeh SA, et al. Design of respiration a veraged CT for attenuation correction of the PET data from PET/CT. Med Phys 2007;34:2039–47.

Imaging of Structure and Function with PET/CT

97. Alessio AM, Kohlmyer S, Branch K, et al. Cine CT for attenuation correction in cardiac PET/CT. J Nucl Med 2007;48:794–801. 98. Pan T, Lee TY, Rietzel E, Chen GT . 4D-CT imaging of a v olume influenced by respiratory motion on multi-slice CT . Med Ph ys 2004;31:333–40. 99. Meirelles GS, Erdi YE, Nehmeh SA, et al. Deep-inspiration breathhold PET/CT: clinical f indings with a ne w technique for detection and characterization of thoracic lesions. J Nucl Med 2007;48:712–9. 100. Nehmeh SA, Erdi YE, Meirelles GS, et al. Deep-inspiration breathhold PET/CT of the thorax. J Nucl Med 2007;48:22–6. 101. Dawood M, Buther F, Lang N , et al. Respirator y gating in positron emission tomography: a quantitative comparison of different gating schemes. Med Phys 2007;34:3067–76. 102. Martinez-Moller A, Zikic D, Botnar RM, et al. Dual cardiac-respiratory gated PET: implementation and results from a feasibility study. Eur J Nucl Med Mol Imaging 2007;34:1447–54. 103. Nagel CC, Bosmans G, Dekk er AL, et al. Phased attenuation correction in respiration correlated computed tomography/positron emitted tomography. Med Phys 2006;33:1840–7. 104. Allen-Auerbach M, Yeom K, P ark J, et al. Standard PET/CT of the chest during shallo w breathing is inadequate for comprehensi ve staging of lung cancer. J Nucl Med 2006;47:298–301. 105. Beyer T , Bockisch A, K uhl H, Mar tinez MJ . Whole-body 18 F-FDG PET/CT in the presence of tr uncation artifacts. J Nucl Med 2006;47:91–9. 106. Carney JP, Townsend DW, Kinahan PE, et al. CT-based attenuation correction: the effects of imaging with the ar ms in the f ield of view [abstract]. J Nucl Med 2001;42:P56–7. 107. Mawlawi O, Pan T, Cody DD, et al. Evaluation of a new CT truncation cor rection algorithm for accurate quantif ication of PET/CT images [abstract]. J Nucl Med 2004;45:413P. 108. Zaidi H. Is radionuclide transmission scanning obsolete for dualmodality PET/CT systems? Eur J Nucl Med Mol Imaging 2007; 34:815–8. 109. Nakamoto Y, Osman M, Cohade C, et al. PET/CT : comparison of quantitative tracer uptak e between germanium and CT transmission attenuation-corrected images. J Nucl Med 2002;43:1137–43. 110. van Dalen JA, Visser EP, Vogel WV, et al. Impact of Ge-68/Ga-68-based versus CT-based attenuation cor rection on PET . Med Ph ys 2007; 34:889–97. 111. Brix G, Be yer T. PET/CT : dose-escalated image fusion? Nuklearmedizin 2005;44 Suppl 1:S51–7. 112. ICRP 87. Managing patient dose. In: Valentin J, editor. Computed tomography. Oxford, Ne w York: Pub lished for the Inter national Commission on Radiological Protection by Pergamon; 2000. 113. Brix G, Lechel U , Glatting G, et al. Radiation e xposure of patients undergoing whole-body dual-modality 18F-FDG PET/CT e xaminations. J Nucl Med 2005;46:608–13. 114. Kalender WA, Wolf H, Suess C, et al. Dose reduction in CT b y on-line tube cur rent control: principles and v alidation on phantoms and cadavers. Eur Radiol 1999;9:323–8. 115. Kachelriess M, Watzke O , Kalender WA. Generalized multidimensional adaptive f iltering for con ventional and spiral singleslice, multi-slice, and cone-beam CT. Med Phys 2001;28:475–90. 116. ICRP 53. Radiation dose to patients from radiophar maceuticals: addendum 2 to ICRP publication 53. Oxford, New York: Published for the Inter national Commission on Radiolo gical Protection b y Pergamon; 1999. 117. Deloar HM, Fujiwara T, Shidahara M, et al. Estimation of absorbed dose for 2-[F-18]fluoro-2-deoxy-D-glucose using whole-body positron emission tomo graphy and magnetic resonance imaging. Eur J Nucl Med 1998;25:565–74. 118. Hays MT, Watson EE, Thomas SR, Stabin M. MIRD dose estimate report no. 19: radiation absorbed dose estimates from 18F-FDG. J Nucl Med 2002;43:210–4. 119. Czernin J, Auerbach MA. Clinical PET/CT imaging: promises and misconceptions. Nuklearmedizin 2005;44 Suppl 1:S18–23.

27

120. Israel O , K uten A. Earl y detection of cancer recur rence: 18 F-FDG PET/CT can mak e a difference in diagnosis and patient care. J Nucl Med 2007;48 Suppl 1:S28–35. 121. Chen YK, Su CT , Ding HJ , et al. Clinical usefulness of fused PET/CT compared with PET alone or CT alone in nasophar yngeal carcinoma patients. Anticancer Res 2006;26:1471–7. 122. Palmedo H, Bucerius J, Joe A, et al. Integrated PET/CT in differentiated thyroid cancer: diagnostic accurac y and impact on patient management. J Nucl Med 2006;47:616–24. 123. Osman MM, Cohade C, Leal J, Wahl RL. Direct comparison of FDGPET and PET-CT imaging in staging and re-staging patients with lung cancer. J Nucl Med 2002;43:151P. 124. Lardinois D, Weder W, Hany TF, et al. Staging of non-small-cell lung cancer with inte grated positron-emission tomo graphy and computed tomography. N Engl J Med 2003;348:2500–7. 125. Cerfolio RJ, Bryant AS, Ojha B. Restaging patients with N2 (stage IIIa) non-small cell lung cancer after neoadjuvant chemoradiotherapy: a prospective study. J Thorac Cardiovasc Surg 2006;131:1229–35. 126. Fueger BJ , Weber WA, Quon A, et al. P erformance of 2-deoxy-2-[F-18]fluoro-D-glucose positron emission tomography and inte grated PET/CT in restaged breast cancer patients. Mol Imaging Biol 2005;7:369–76. 127. Tatsumi M, Cohade C, Mourtzikos KA, et al. Initial experience with FDG-PET/CT in the evaluation of breast cancer. Eur J Nucl Med Mol Imaging 2006;33:254–62. 128. Bar-Shalom R, Guralnik L, Tsalic M, et al. The additional v alue of PET/CT over PET in FDG imaging of oesophageal cancer . Eur J Nucl Med Mol Imaging 2005;32:918–24. 129. Guo H, Zhu H, XiY, et al. Diagnostic and prognostic value of 18F-FDG PET/CT for patients with suspected recurrence from squamous cell carcinoma of the esophagus. J Nucl Med 2007;48:1251–8. 130. Cohade C, Osman M, Leal J , Wahl RL. Direct comparison of 18 F-FDG PET and PET/CT in patients with colorectal carcinoma. J Nucl Med 2003;44:1797–803. 131. Reinhardt MJ, Joe AY, Jaeger U, et al. Diagnostic perfor mance of whole body dual modality 18F-FDG PET/CT imaging for N- and M-staging of malignant melanoma: experience with 250 consecutive patients. J Clin Oncol 2006;24:1178–87. 132. Freudenberg LS, Antoch G, Schutt P, et al. FDG-PET/CT in re-staging of patients with l ymphoma. Eur J Nucl Med Mol Imaging 2004;31:325–9. 133. Yi CA, Lee KS, Kim BT , et al. Tissue characterization of solitar y pulmonary nodule: comparati ve study betw een helical dynamic CT and integrated PET/CT. J Nucl Med 2006;47:443–50. 134. Kim SK, Allen-Auerbach M, Goldin J, et al. Accuracy of PET/CT in characterization of solitar y pulmonar y lesions. J Nucl Med 2007;48:214–20. 135. Alavi A, Mavi A, Basu S, F ischman A. Is PET-CT the only option? Eur J Nucl Med Mol Imaging 2007;34:819–21. 136. Yu JN, Fahey FH, Gage HD, et al. Intermodality, retrospective image registration in the thorax. J Nucl Med 1995;36:2333–8. 137. Mutic S, Dempsey JF, Bosch WR, et al. Multimodality image registration quality assurance for confor mal three-dimensional treatment planning. Int J Radiat Oncol Biol Ph ys 2001;51:255–60. 138. Bradley JD , Dehdashti F , Mintun MA, et al. P ositron emission tomography in limited-stage small-cell lung cancer: a prospective study. J Clin Oncol 2004;22:3248–54. 139. Schmucking M, Baum RP, Griesinger F, et al. Molecular whole-body cancer staging using positron emission tomography: consequences for therapeutic management and metabolic radiation treatment planning. Recent Results Cancer Res 2003;162:195–202. 140. Ciernik IF, Dizendorf E, Baumert BG, et al. Radiation treatment planning with an inte grated positron emission and computer tomography (PET/CT): a feasibility study. Int J Radiat Oncol Biol Phys 2003;57:853–63. 141. Esthappan J, Mutic S, Malyapa RS, et al. Treatment planning guidelines re garding the use of CT/PET -guided IMR T for cer vical

28

142.

143. 144.

145.

146.

147.

148.

149.

150.

151.

MOLECULAR IMAGING: PRINCIPLES AND PRACTICE

carcinoma with positi ve paraaor tic l ymph nodes. Int J Radiat Oncol Biol Phys 2004;58:1289–97. Yap JT, Townsend D W, Hall NC. PET -CT in IMR T planning. In: Mundt AJ, Roeske J, editors. Intensity modulated radiation therapy: a clinical perspective. Amsterdam, The Netherlands; Oxford, UK: Elsevier; 2004. p. 115–30. Weber WA, F iglin R. Monitoring cancer treatment with PET/CT : does it make a difference? J Nucl Med 2007;48 Suppl 1:S36–44. Wieder HA, Beer AJ, Lordick F , et al. Comparison of changes in tumor metabolic activity and tumor size during chemotherap y of adenocarcinomas of the esophago gastric junction. J Nucl Med 2005;46:2029–34. Pottgen C, Le vegrun S, Theegarten D, et al. Value of 18F-fluoro-2deoxy-D-glucose-positron emission tomography/computed tomography in non-small-cell lung cancer for prediction of patholo gic response and times to relapse after neoadjuvant chemoradiotherapy. Clin Cancer Res 2006;12:97–106. Nahmias C, Hanna WT, Wahl LM, et al. Time course of early response to chemotherap y in non-small cell lung cancer patients with 18F-FDG PET/CT. J Nucl Med 2007;48:744–51. de Geus-Oei LF, van der Heijden HF, Visser EP, et al. Chemotherapy response evaluation with 18F-FDG PET in patients with non-small cell lung cancer. J Nucl Med 2007;48:1592–8. Erdi YE, Rosenzweig K, Erdi AK, et al. Radiotherapy treatment planning for patients with non-small cell lung cancer using positron emission tomography (PET). Radiother Oncol 2002;62:51–60. Kong FM, F rey KA, Quint LE, et al. A pilot study of [ 18F]fluorodeoxyglucose positron emission tomo graphy scans during and after radiation-based therapy in patients with non small-cell lung cancer. J Clin Oncol 2007;25:3116–23. Weber WA, Wieder H. Monitoring chemotherap y and radiotherap y of solid tumors. Eur J Nucl Med Mol Imaging 2006;33 Suppl 1:27–37. Therasse P, Arbuck SG, Eisenhauer EA, et al. Ne w guidelines to evaluate the response to treatment in solid tumors. European

152.

153. 154.

155.

156. 157.

158. 159.

160.

Organization for Research and Treatment of Cancer , National Cancer Institute of the United States, National Cancer Institute of Canada. J Natl Cancer Inst 2000;92:205–16. Larson SM, Erdi Y, Akhurst T, et al. Tumor treatment response based on visual and quantitati ve changes in global tumor gl ycolysis using PET -FDG imaging. The visual response score and the change in total lesion glycolysis. Clin Positron Imaging 1999;2:159–71. Soret M, Bacharach SL, Buvat I. Partial-volume effect in PET tumor imaging. J Nucl Med 2007;48:932–45. Namdar M, Han y TF, K oepfli P, et al. Inte grated PET/CT for the assessment of coronar y artery disease: a feasibility study . J Nucl Med 2005;46:930–5. Gould KL, P an T, Lo ghin C, et al. F requent diagnostic er rors in cardiac PET/CT due to misregistration of CT attenuation and emission PET images: a def initive anal ysis of causes, consequences, and corrections. J Nucl Med 2007;48:1112–21. DiCarli M, Lipton M. Cardiac PET and PET/CT imaging. New York; London, UK: Springer; 2007. Schmand M, Burbar Z, Corbeil J , et al. Brain PET : f irst human tomograph for simultaneous (functional) PET and MR imaging. J Nucl Med 2007;48:45P. Schlemmer H, Pichler PJ, Wienhard K, et al. Simultaneous MR/PET for brain imaging: f irst patient scans. J Nucl Med 2007;48:45P. Zaidi H, Ma wlawi O, Or ton CG. P oint/counterpoint. Simultaneous PET/MR will replace PET/CT as the molecular multimodality imaging platform of choice. Med Phys 2007;34:1525–8. Coleman RE, Delbeke D, Guiberteau MJ, et al. Concurrent PET/CT with an integrated imaging system: intersociety dialogue from the joint working g roup of the American Colle ge of Radiolo gy, the Society of Nuclear Medicine, and the Society of Computed Body Tomography and Magnetic Resonance. J Nucl Med 2005; 46:1225–39.

3 PET/MRI MARCUS D. SEEMANN, MD

Whole-body imaging is cur rently a topic of g reat interest within the scientif ic community. For tumor staging, diagnostic imaging of the w hole body allo ws simultaneous evaluation of both the region of the primary tumor and the presence of metastases. This evaluation can be perfor med by using se veral dif ferent approaches that include molecular/biochemical imaging techniques such as planar scintigraphy, single photon emission computed tomo graphy (SPECT), positron emission tomo graphy (PET), or anatomic/morphologic imaging techniques such as conventional radio graphy, computed tomo graphy (CT), and magnetic resonance imaging (MRI). (Please also see Chapter 2 “Imaging of Str ucture and Function with PET/CT” and Chapter 4 “SPECT and SPECT/CT” for further discussion). Each of these imaging modalities has specific advantages as w ell as disadv antages in ter ms of, for example, sensitivity, specif icity, accuracy, radiation e xposure, costs, and image acquisition time. The fusion of molecular/biochemical with anatomic/mor phologic information can compensate for many disadvantages and therefore of fers se veral adv antages in comparison to using molecular/biochemical or anatomic/mor phologic diagnostic imaging techniques alone. In the scientif ic literature, the fusion of molecular/biochemical and anatomic/ morphologic information has been shown to improve diagnostic accurac y in identifying and characterizing malignancies, impro ve assessment of tumor stage, therapeutic response and tumor recur rence as compared to visual correlation of the images, allow discrimination of variable physiologic radiotracer uptak e (brain, th yroid gland , f at, striated muscle, myocardium, digestive tract, bone marrow, and genitourinary tract) that can mimic tumor or metastatic lesions from patholo gical uptak e, and help to a void potential f alse-positive inter pretations.1–6 Awareness of the hazards of radiation e xposure has prompted se veral investigators to focus on techniques that enab le w holebody scanning with lo w or no radiation e xposure. The

combination of the tw o e xcellent diagnostic imaging modalities, PET and MRI, into a single scanner of fers several advantages in comparison to using PET and MRI alone and impro ves diagnostic accurac y b y f acilitating the accurate cor relation and e valuation of molecular aspects and biochemical alterations of disease with e xact correlation to anatomic information and morphologic findings (Figure 1). Therefore, the expected diagnostic imaging value of a h ybrid w hole-body PET/MRI scanner is very high.

PET PET imaging relies primaril y on radiotracer uptak e changes for disease detection. PET f acilitates the assessment of molecular aspects and biochemical alterations of a wide v ariety of diseases that are fundamental in the detection of malignancies, evaluation of tumor and tumor stage, and assessment of therapeutic response and tumor recurrence.7 (Please also see Chapter 19 “Chemistr y of Molecular Imaging: An Overview,” Chapter 20 “Radiochemistry of PETm” and Chapter 38 “Ov erview of Molecular and Cell Biolo gy” for fur ther discussion). In general, the accelerated radiotracer acti vity occurs, and therefore can be seen, before anatomic/mor phologic structure changes can be depicted. The main advantage of PET is its high sensiti vity in identifying areas of cancerous in volvement at an earl y stage and distinguishing malignant lesions from benign lesions in most cases. Therefore, PET has become an accepted and v aluable diagnostic imaging tool for patients with cancer . By reducing the probability of overlooking involved areas, it influences the initial staging, restaging after chemotherapy or radiation treatment, and overall management of the disease. The main dif ficulty with PET is the lack of an anatomic/morphologic reference frame. The actual PET scanner allows a total scan range of up to 1,981 mm and 29

30

MOLECULAR IMAGING: PRINCIPLES AND PRACTICE

A A

B B

C C

D D

Figure 1. 63-year-old woman with cerebral, pulmonary, lymphogenic, osseous, and soft-tissue metastases of a malignant melanoma. Corresponding coronal whole-body 18F-labeled 2-fluoro-2-deoxy-D-glucose positron emission tomography (18F-FDG PET) image (CT-based attenuation correction) (A), whole-body T2-weighted turbo-short tau inversion-recovery (STIR) image (B), whole-body T2-weigthed half-fourier acquired single-shot turbo spin echo (HASTE) image (C), and whole-body T1-weighted Turbo Spin Echo (TSE) image (D).

therefore complete head-to-toe co verage.8 To eva luate molecular aspects and biochemical alterations, PET imaging can be perfor med with dif ferent radiotracers administered sequentially. The most commonl y used radiophar maceutical is a glucose analo g, 18F-labeled 2-fluoro-2-deo xy-D-glucose (18F-FDG). It relies on the detection of an increased rate of aerobic glycolysis. Since the American Food and Drug Administration (FD A) appro ved 18F-FDG in 1997 as a safe and ef fective radiophar maceutical for oncolo gic applications, and since the U .S. Health Care F inancing Administration (HCFA) authorized Medicare in 1998 to reimburse for 18F-FDG PET imaging for cer tain indications, 18F-FDG PET imaging has become an accepted and valuable sensiti ve diagnostic imaging technique for patients with cancer , because it can be used to spot areas of malignanc y and tumor g rowth and to assess tumor stage, therapeutic response, and tumor recur rence. (Please also see Chapter 76 “Re gulatory and Reimbursement Process for Imaging Agents and Devices” for further discussion). The cellular uptake of 18F-FDG molecules is a function of biochemical cell activity and associated with

increased cell tur nover. In most cancers, malignant cells are associated with increased biochemical activity. Therefore, increased uptake of 18F-FDG molecules can be used to spot areas of malignancy and tumor growth. In general, this accelerated biochemical acti vity occurs before anatomic/morphologic str ucture changes. Other imaging modalities, such as CT and MRI rel y primaril y on anatomic/morphologic str ucture changes for disease detection. Whole-body PET imaging with the radiolabeled glucose analo g 18F-FDG has gained widespread acceptance for the staging and restaging of cancer . Furthermore, 18F-FDG PET is a po werful tool for predicting chemotherapy response in cancer more accuratel y than conventional imaging methods. F or e xample, in locall y advanced breast cancer and esophageal cancer, it could be proven that the mean standardized uptak e values (SUV) decreased signif icantly e ven in the second w eek after beginning the f irst c ycle of primar y chemotherap y.9–11 The quantitative 18F-FDG measurement of the mean SUV of the malignancies sho wed a signif icant decrease of radiotracer uptake of 30%. The visual simulated 18F-FDG increase in metastases could be caused b y initial ef fects

PET/MRI

such as neo vascularization and inflammator y cell infiltration within the tumor boundaries and the sur rounding tissue. Thus, it is e xpected that 18F-FDG PET will be useful in reducing the costs of c ytotoxic therapy and the unnecessary side-ef fects of inef fective chemotherap y (Please also see Chapter 52 “PET Diagnosis and Response Monitoring in Oncology” and Chapter 72 “Quantif ication of Radiotracer Uptake into Tissue” for further discussion). 18 F-FDG PET has also been used for research pur poses, because it of fers molecular/biochemical images noninvasively, quantitatively, and repeatedly, not just in humans but also in small animals, using speciall y designed high-resolution small-animal scanning equipment. 12–14 (Please also see Chapter 6 “Small Animal SPECT , SPECT/CT and SPECT/MRI” and Chapter 7 “Instr umentation and Methods to Combine Small Animal PET with Other Imaging Modalities” for fur ther discussion). Neuroendocrine tumors are slo w-growing rare neoplasms of neuroectodermal origin that frequently express specific somatostatin receptors that can be of g reat value in the staging and treatment 15,16 of these tumors. Therefore, this type of tumor entity is imaged b y w ay of somatostatin receptors, and not with the radiolabeled glucose analog 18F-FDG. SPECT using radiolabeled analogs of somatostatin has become the standard in vivo imaging modality of choice in the identif ication and staging of neuroendocrine tumors and thus the most reliable tool for guiding therapy.17–21 Primary endocrine tumors are small neoplasms that arise predominantly in the gastrointestinal tract, pancreas, and lung. The most common metastatic sites are the li ver, l ymph nodes, bone, lung, and peritoneal ca vity. Numerous studies ha ve conf irmed higher sensitivity, specif icity, diagnostic accurac y, and positi ve predictive v alue of somatostatin receptor scintig raphy (SRS) (90%, 80%, 83%, and 100%, respecti vely) compared with mor phological imaging procedures (w hich have a sensitivity of between 50 and 70%, depending on the size and location of the lesion). 17–22 The sensitivity of SRS depends on the high v ariability in the e xpression of somatostatin receptors among the different tumor entities and even within dif ferent lesions in a single patient, 16,23 the limitations of the physical properties of SPECT imaging, and f inally the phar macokinetic characteristics of Indium-111 DTP A-octreotide as a SPECT tracer with high physiological uptake in li ver and spleen, and elimination by the kidne y and also at a lo w percentage b y the hepatobiliary system. (Please also see Chapter 21 “Radiochemistry of SPECT: Examples of 99mTC and 111IN Complexes” for further discussion). A promising somatostatin receptor ligand for PET imaging is [ 18F]FP-Gluc-TOCA (Nα-(1-deoxy-D-fructosyl)-Nε-(2-[18F]fluoropropionyl)Lys0-Tyr3-octreotate). [ 18F]FP-Gluc-TOCA showed a v ery

31

high affinity to human somatostatin receptor subtype (hsst) 2, a moderate affinity to hsst 5, a low affinity to hsst 4, and no affinity to hsst 1 and 3. Its lo w lipophilicity, low liver uptake, rapid renal elimination, and lo w intestinal activity, as well as its f ast and high tumor accumulation, pro vide excellent tumor-to-background ratios.24,25 In medical research, 18F-labeled 1- α-D-(5-fluoro5-deoxyarabinofuranosyl)-2-nitroimidazole ( 18F-fluoroazomycin arabinoside; 18F-FAZA) is a promising radioactive tracer for patients with cancer because it can be used nonin vasively in the detection of tumor hypoxia in humans, as w ell as in small animals. 26 As tumor h ypoxia has a major ne gative predicti ve v alue for local tumor pro gression, lik eliness of metastasis, and overall tumor prognosis in several types of human cancer, the presence of tumor tissue hypoxia is relevant in predicting pro gnosis and response to cur rent radiation treatment. 27 In addition, tumor cell h ypoxia has a negative ef fect on anticancer treatment, gi ven that hypoxic cells are 2 to 3 times more resistant to a single fraction of ionizing radiation than those with n ormal oxygenation le vels.28 (Please also see Chapter 4 6 “Hypoxia Imaging” for fur ther discussion).

MRI MRI is an e xcellent anatomic/mor phologic imaging modality with a high anatomic resolution. Ov er the past decade, major impro vements in MRI ha ve occur red. Improvements in the hardw are of MRI machines ha ve facilitated the de velopment of f ast and ultraf ast pulse sequences as well as the development of phased-array multicoils allo wing for higher signal-to-noise ratio and thus higher anatomic resolution. In addition, introduction of navigator techniques has fur ther enhanced signal-to-noise ratio and spatial resolution allo wing for higher imaging matrices, resulting in improved diagnostic perfor mance in the detection of liver lesions.29 However, image blurring is one potential drawback of this technique if the navigator is incorrectly placed or if the patient has an ir regular breathing patter n. Then, the na vigator technique should be replaced with f ast or ultraf ast breath-hold techniques to avoid respirator y or motion-induced ar tifacts.30 However, breath-hold techniques result in poorer signal-to-noise and limited spatial resolution. In the past, MRI w as used only as a tool to image specif ic regions of the body, due to the prolonged imaging time, limited a vailability of scanning facilities, and extensive costs. The concept that MRI might become the ultimate whole-body imaging tool was initially proposed b y the MRI pioneers Damadian and Lauterbur.31,32 The de velopment of f ast and ultraf ast MRI techniques led to the possibility of rapid w hole-body

32

MOLECULAR IMAGING: PRINCIPLES AND PRACTICE

scanning. However, most in vestigators used the body coil for multistation data reception. 33,34 Thus, images suf fered from either low signal-to-noise ratio or poor spatial resolution.35–38 To overcome these limitations, other investigators examined patients on a sliding table platform with an integrated phased-array surface coil.39 Although this approach improves signal-to-noise, the distal limbs w ere not included.40 In addition, the adv antages of parallel acquisition techniques to achie ve higher spatial resolution or shorter acquisition times were not used. The new development of Total Imaging Matrix (TIM) allo ws, for the f irst time, perfor mance of high-resolution w hole-body co verage from head to toe within a single e xamination without the need for patient or surf ace coil repositioning, yielding excellent high-resolution image quality . The basic idea of TIM is the re volutionary matrix coil concept that allows the combination of 76 coil elements with up to 32 channels, a combination that enab les considerab le improvement in both acquisition speed and image quality . The actual MR scanner allo ws a total scan range of up to 2,050 mm and therefore complete head-to-toe co verage. Whole-body MRI is a promising diagnostic modality used for both diagnosis and management in man y diseases, and not onl y for patients with cancer . Ho wever, the optimal imaging protocol and patient preparation for the w holebody imaging ha ve still y et to be e valuated. P erforming MRI without the use of an intra venous paramagnetic contrast medium ma y include the possibility of o verlooking some hypervascular metastases that can only be seen in the hepatic ar terial-dominant contrast-enhanced phase. 41–43 However, Dromain and colleagues 42 found no signif icant difference betw een the hepatic ar terial-dominant contrast-enhanced T 1-weighted images and the f ast spin-echo T 2-weighted images in the detection of hepatic metastases from neuroendocrine tumors. The f ast spin-echo T 2-weighted sequence sho wed a signif icantly higher lesion-to-liver contrast-to-noise ratio than the hepatic ar terial-dominant contrast-enhanced T1-weighted sequence. For mor phologic imaging methods, the w eakness of l ymph node staging is the kno wn lack of reliab le criteria since assessment can be made onl y on the basis of size.44 Lymph nodes can onl y be staged as l ymph node metastases on the basis of mor phologic criteria lik e a nodular shape and a diameter of more than 10 mm (reference standard). Fur thermore, the dif ferentiation of abdominal lymph nodes and small intestine, without using oral contrast media, can be a prob lem. First of all, coronal sequences like T2-weighted turbo-short tau inversion-recovery (STIR), T1-weighted turbo spin echo (TSE) sequence, and T 2-weigthed half-F ourier acquired single-shot turbo spin echo (HASTE) sequence to image the w hole body,

can be perfor med. To a void respirator y motion-induced artifacts, the Turbo-STIR sequence as w ell as the T1-weighted TSE sequence in the chest and abdomen should be perfor med with breath holds. If necessar y, an additional T 1-weighted f ast lo w-angle shot (FLASH) sequence can be used to conf irm the presence of sk eletal metastases. The detection of osseous metastases in the ribs can be a problem. In addition, axial T2-weighted highresolution TSE sequences using prospecti ve acquisition with na vigator technique (respirator y gating to track diaphragmatic and cardiac mo vements) can be obtained from the chest, abdomen, and pelvis. This can be followed by high-resolution axial cross-sectional sequences to focus on the detected patholo gy, and f acultative techniques such as functional imaging, dif fusion and perfusion imaging, spectroscopy, and angiograpy. Finally, MRI is emer ging as a par ticularly adv antageous modality for molecular/ biochemical imaging. (Please also see Chapter 26 “MR Imaging Agents,” Chapter 34 “Magnetic Nanopar ticles,” Chapter 36 “ Aptamers for Molecular Imaging, ” and Chapter 44 “Cell Voyeurism using Magnetic Resonance Imaging” for fur ther discussion). P ossibly in conjunction with rational tar geted therapies, this technique could radically affect the practice of clinical diagnosis and therapy as these technologies continue to mature.45 MR imaging relies primarily on anatomic/mor phologic str ucture changes for disease and tumor detection. After chemotherap y, anatomic/morphologic imaging procedures allow the detection of changes in tumor size and v olume. Reduction of tumor volume as evidence of response to therapy requires a certain time dela y after initiation of therap y, and ma y be masked b y unspecif ic ef fects (e g, edema as a result of necrosis). Whole-body MRI protocols have to be optimized and adapted to allo w for better dif ferentiation of the primary tumor and possib le metastases, especiall y in the abdomen. In general, w hole-body MRI should be performed with intra venous contrast media and in oncolo gic patients in combination with oral contrast media.

PET/MRI Over the past decade, there ha ve been g reat technical improvements in PET and MRI. The combination of these two excellent diagnostic imaging modalities into a single scanner of fers se veral adv antages and impro ves diagnostic accuracy by facilitating the accurate re gistration of molecular aspects and biochemical alterations of disease with e xact cor relation to anatomic infor mation and morphologic findings. PET facilitates the evaluation of molecular aspects and biochemical alterations that are fundamental to the detection of malignanc y and

PET/MRI

assessment of tumor stage, therapeutic response, and tumor recurrence. In general, increased radiotracer accumulation can be seen before anatomic/morphologic structure changes. The primary difficulty with PET is the lack of an anatomic reference frame. The combination of PET with an anatomic/morphologic imaging modality such as CT or MRI can compensate for this disadv antage and offers se veral adv antages in comparison to using PET , CT, or MRI alone. Whole-body MRI produces lar ge amounts of image data, resulting in the possibility of overlooking subtle patholo gic f indings. Fur thermore, anatomic/morphologic imaging procedures do not allo w differentiation betw een viab le tumor tissue and f ibrotic scar tissue. 18F-FDG PET can dif ferentiate viable tumor tissue from atelectases and scars and is therefore helpful in planning radiotherap y of lung carcinoma. The posttherapeutic 18F-FDG PET/CT study can sho w, qualitatively and quantitati vely, decreased acti vity and v olume of all tumor lesions. The cellular uptake of the 18F-labeled glucose analog 2-fluoro-2-deoxy-D-glucose ( 18F-FDG) is a sensiti ve and v aluable mark er for biochemical alterations of cancer cells that is fundamental not onl y in the detection of a wide v ariety of malignancies, but also for prediction of neoadjuv ant chemotherap y response, and is more accurate than anatomic/morphologic imaging methods. A fur ther adv antage of PET is the sensiti ve detection of lymph node metastases of a size smaller than 10 mm that are not unequi vocally assessab le with anatomic/morphologic imaging and can be identif ied by PET, leading to an improvement of this morphologic reference standard. The w eakness of mor phologic l ymph node staging is the kno wn lack of reliab le criteria, since assessment can be made only on the basis of size.44 In the scientific literature, the simultaneous acquisition of coregistered molecular/biochemical and anatomic/mor phologic information has been sho wn to improve diagnostic accuracy of the cancer staging as compared to visual correlation of the images, allo ws the discrimination of v ariable physiologic radiotracer uptake (brain, thyroid gland, fat, striated muscle, m yocardium, digesti ve tract, bone marrow, and genitourinary tract) that can mimic metastatic lesions from patholo gical uptake, and helps to a void potential false-positive inter pretations.1–4,6 The fusion of PET with MRI can compensate for their separate disadvantages, and therefore of fers se veral combined adv antages in comparison to using PET or MRI alone. The expected diagnostic value of the combination of PET and MRI into a single w hole-body hybrid PET/MRI scanner is v ery high. In the case w here PET and MRI are not fused via a single scanner , the repositioning of the patient and time inter val betw een the scans mak es the

33

co-registration and fusion of separatel y obtained images difficult and inherentl y imprecise. 2 Whole-body PET/MRI is a v ery promising diagnostic modality for oncologic imaging and cancer screening in the decades to come, due to the considerably lower radiation exposure in contrast to PET/CT, and the high soft-tissue resolution of MRI (Figure 2). In the literature, 18F-FDG PET/CT and MRI in patients with dif ferent malignant diseases w ere compared. 46 The authors suggest the use of 18F-FDG PET/CT as a possib le first-line modality for whole-body tumor staging. However, they did not perform an image fusion of PET and MRI. The diagnostic value of whole-body imaging modalities PET, CT, MRI, and the image fusion of PET and CT (PET/CT) and PET and MRI (PET/MRI) in the detection of metastases of gastrointestinal neuroendocrine tumors was e valuated in a prospecti ve study .47 PET data w as acquired with a state-of-the-ar t high-count-rate lutetium oxyorthosilicate (LSO) detector Pico-3D full-ring PET scanner of a hybrid PET/CT from the base of the skull to the proximal thigh. As a radiopharmaceutical, a carbohydrate deri vatized F-18-labeled somatostatin-receptor ligand ([18F]FP-Gluc-TOCA = Nα-(1-deoxy-D-fructosyl)Nε-(2-[18F]fluoropropionyl)-Lys0-Tyr3-octreotate) w as used. CT data was acquired with the 16-slice CT scanner of the hybrid PET/CT using a venous-dominant contrastenhanced phase. F or an optimal assessment of the gastrointestinal tract, oral administration of diluted diatrizoate meglumine was perfor med beginning 1 hour before star ting the e xamination. MRI data w as acquired with a 1.5 Tesla w hole-body MRI scanner using the TIM technology.48 MRI w as perfor med with a coronal T2-weighted HASTE sequence, a coronal T2-weighted STIR sequence, a coronal T1-weighted TSE sequence, and a high-resolution axial T2-weighted TSE sequence with na vigator technique. F or the detection of li ver metastases, PET/MRI (100%) and MRI (98.2%) w ere most sensitive, whereas PET/CT (50.9%)( p < .001), PET (49.9%)( p < .001), and CT (37.1%)( p < .001) were significantly less reliab le (F igure 3). In this comparati ve study, MRI was the most sensiti ve imaging procedure in the detection of liver metastases. The main reason for this is the spatial resolution in the submillimeter area and the high soft-tissue contrast. MRI sho wed a lot of lesions with a diameter between 2 and 4 mm, which could not be seen in any other modality. PET, CT, and PET/CT underestimated the extension of liver metastases or even missed the metastatic disease, because of the lo w soft-tissue contrast in CT, the low spatial resolution and physiological tracer uptake in the normal liver tissue, and the dependency on the presence, type, and density of the

34

MOLECULAR IMAGING: PRINCIPLES AND PRACTICE

A

B

C C

D

E

F F

Figure 2. 63-year-old woman with malignant melanoma. The primary tumor in the left lower arm was surgically resected two years ago. Corresponding coronal centered view from head to diaphragm (A–C) showing a 18F-FDG PET image (A), a HASTE image (B), and a manual fused image of PET and MRI (PET/MRI) (C), with a cerebral metastasis and multiple soft-tissue metastases on the left lateral thoracic wall. Corresponding coronal centered view from the inlet of the thorax to inlet of the pelvis (D–F) showing a 18F-FDG PET image (D), a HASTE image (E), and a manual fused image of PET and MRI (PET/MRI) (F) with pulmonary and osseous metastases.

somatostatin receptors expressed by the tumor lesions as well as the tumor size w hen using PET. Metastases with a low somatostatin receptor density and necrotic metastases showed low to no tracer accumulation in PET. Nevertheless, PET seems to be ab le to gi ve additional information to MRI if there are prob lems with the na vigator technique. For the detection of l ymph node metastases, PET/CT (100%), PET/MRI (97.3%), and PET (91.9%) w ere most sensiti ve, w hereas CT (83.8%; p < .54) and MRI (64.9%; p < . 12) w ere considerab ly less reliable (F igures 4 and 5). PET w as the most sensiti ve imaging procedure in the detection of lymph node metastases. Six l ymph nodes (16.2%) (cer vical, n = 1; mediastinal, n = 2; retroperitoneal, n = 2; and intraperitoneal, n = 1) sho wed a clearl y increased tracer uptak e in PET but were smaller than the reference standard (10 mm) in CT and MRI. Three lymph nodes ( n = 8.1%) (retroperitoneal, n = 1 and intraperitoneal, n = 2) sho wed a size from 11 to 14 mm in CT but no increased tracer uptake in PET. The main adv antage of l ymph node staging with PET is the sensiti ve detection of serotonin-expressing small lymph nodes (< 10 mm), w hich are not assessab le

with morphologic imaging and can be identif ied by PET, leading to an impro vement of the sensiti vity. The detection rate of 100% for l ymph node staging in the combination of PET and anatomic/mor phologic imaging is not a realistic v alue, because it is possib le that indi vidual lymph node metastases sho wed no octreotate accumulation and also were not enlarged. The assessment of lymph node staging is dif ficult and unsatisf actory using morphologic imaging procedures onl y. The weakness of morphologic lymph node staging is due to theknown lack of reliable criteria, since assessment can be made only on the basis of size. 44,47 On the one hand, some lymph node metastases with a size smaller than 10 mm may not be detected, and lymph nodes with a size of g reater than 10 mm can also be caused b y inflammation. Using MRI, abdominal lymph node metastases w ere difficult to differentiate from small and lar ge intestine. F or the detection of osseous metastases, PET (100%), PET/CT (100%), and PET/MRI (100%) w ere m ost sensitive, whereas MRI (66.7%; p < .12) and CT (8.3%; p < .003) were less reliable (Figure 6). PET was the most sensitive imaging procedure in the detection of osseous

PET/MRI

A

B

C

D

E

F

G

H

35

Figure 3. 53-year-old male patient with multiple hepatic metastases of an intestinal neuroendocrine tumor. Corresponding axial Nα-(1deoxy-D-fructosyl)-Nε-(2-[18F]fluoropropionyl)-Lys0-Tyr3-octreotate ([18F]FP-Gluc-TOCA) PET image (A, H). T2-weighted STIR image (B), T2-weighted TSE image (C), HASTE image (D), T1-weighted TSE image (E), arterial-dominant contrast-enhanced T1-weighted TSE image (F), and venous-dominant contrast-enhanced T1-weighted TSE image (G). All MR images are shown with the manual fused image of PET (PET/MRI).

metastases. The metastases seen on CT w ere osteosclerotic. The differentiation of de generative changes of the vertebral column and osteosclerotic bone metastases can be a problem in anatomic/morphologic imaging methods. The missing osseous metastases in the MRI w ere localized in the ribs. The results from this comparati ve study suggested that only the combined use of molecular/ biochemical and anatomic/morphologic imaging procedures achie ved a cor rect tumor classif ication, and the combination of PET and MRI into a single scanner could be a v ery v aluable w hole-body diagnostic imaging tool

not onl y for endocrine tumors but also for oncolo gic tumor staging. 47 Schillaci and colleagues 20 compared the detection of abdominal metastases in SRS SPECT and anatomic/morphologic imaging procedures (ultrasound , CT, and/or MRI) in neuroendocrine tumors and suggested also that onl y the combined use of molecular/biochemical and anatomic/mor phologic imaging procedures achieved a correct classification. Planar scintigraphy, SPECT, and PET are the onl y clinically a vailable nonin vasive imaging techniques that can assess molecular aspects and biochemical

36

MOLECULAR IMAGING: PRINCIPLES AND PRACTICE

A

B

C

Figure 4. 62-year-old male patient with a lymph node metastases of an intestinal neuroendocrine tumor. Corresponding axial view from the inlet of the thorax showing a [18F]FP-Gluc-TOCA PET image (A), a T2-weighted TSE image (B), and the manual fused image of PET and MRI (PET/MRI) (C). PET identified a small lymph node metastases (5 × 6 × 8 mm).

A

B

C C

Figure 5. 62-year-old male patient with lymph node metastases of an intestinal neuroendocrine tumor. Corresponding axial view from the pelvis showing a [18F]FP-Gluc-TOCA PET image (A), a T1-weighted TSE image (B), and the manual fused image of PET and MRI (PET/MRI) (C).

A

B

C

Figure 6. 62-year-old male patient with an osseous metastasis of an intestinal neuroendocrine tumor. Corresponding axial view from the thorax showing a [18F]FP-Gluc-TOCA PET image (A), T2-weighted STIR image (B), and the manual fused image of PET and MRI (PET/MRI) (C).

alterations that are fundamental to cancer detection, cancer recur rence, and e valuation of therapeutic response using small amounts of radioacti ve labeled molecules in vi vo.49 PET techniques can assess the in vivo biodistribution of man y relevant radiopharmaceuticals and thus contribute significantly and distinctively

to the e valuation of tumors. 50 Small-animal imaging has gained increasing attention in recent y ears as an excellent in vivo evaluation method for molecular biology, oncolo gy, and neuroscience research. Smallanimal models of rats and mice are widel y used in biomedical research for mimicking and studying the

PET/MRI

human condition in health or disease, because of their genetic resemb lance to humans and the feasibility of gene transfer and gene modif ication.51,49 Similar to human h ybrid PET/CT , tumor -bearing small animals could be examined with human hybrid PET/MRI using special scanning and reconstr uction protocols, and this capability is also expected to contribute significantly to research with small-animal imaging. The investigation of cancer in small animals with h ybrid PET/MRI is probably one of the most challenging tasks in nuclear medicine since the str uctures of interest are almost in the same range as the maximum spatial resolution. Hybrid PET/MRI is considered to be par ticularly wellsuited for research pur poses, especially for the e valuation of tumor g rowth and g rowth inhibition f actors; development of new anti-tumor drugs and measuring of anti-tumor ef fects; and cancer treatment response to immunotherapy, chemotherap y and radiation therap y. (Please also see Chapter 67 “Molecular and Functional Imaging in Dr ug De velopment,” Chapter 68 “PET Imaging Clinical Trials,” and Chapter 69 “MR Imaging in Clinical Trials” for fur ther discussion). Cur rent interesting topics include implantation of human tumor cells in small animals and e valuation of tumor g rowth and growth inhibition factors; development of new antitumor dr ugs and measuring of anti-tumor ef fects; and cancer treatment response to immunotherap y, chemotherapy, and radiation therap y using molecular/ biochemical and anatomic/mor phologic parameters. The infor mation from h ybrid PET/MR imaging w ould be of interest for validation of oncologic research studies for the evaluation of new tumor tracers labeled with different positron emitters (F-18, C-11, N-13, and O-15). Combined systems pro vide the ability to accurately cor relate and e valuate biochemical and molecular aspects of cancers with anatomic infor mation and morphologic findings in human clinical routine examinations and are promising for in vi vo animal research. Hybrid PET/MRI scanners pro vide molecular and biochemical information noninvasively, quantitatively, and repeatedly. Hybrid PET/MRI is a highly valuable oncologic imaging modality with the feasibility of in vi vo imaging of tumor -bearing small animals for potential use in oncolo gy research using special scanning and reconstruction protocols. Due to the small dimensions of the str uctures to be imaged in animals, the smallanimal-specific protocols ha ve made it possib le for tumor h ybrid PET/MR images to be clear enough to resolve considerab le hetero geneity of tracer uptak e within the tumors with the help of thin-slice, high-quality MR images. The e xperience with human h ybrid

37

PET/MRI is e xpected to contribute signif icantly to research with small-animal imaging. 50 Although hybrid PET/CT has proved itself clinically as a highl y v aluable oncolo gic diagnostic modality , it might not be the ultimate diagnostic imaging technique since MRI offers several advantages compared with CT and therefore hybrid PET/CT will be in competition with hybrid PET/MRI. 52,53 In comparison to CT , MRI is not associated with radiation e xposure and has a much higher soft-tissue contrast. This has been sho wn to be advantageous in neuroradiological, musculoskeletal, cardiac, and oncolo gic (e g, detection and characterization of focal li ver lesions) applications. The lower radiation dose is important for pediatric applications and repeated imaging in oncolo gic patients. The higher soft tissue is also interesting in head and neck applications. MRI allows for additional techniques such as angio graphy (eg, tumor v asculature), functional MRI (e g, brain activation studies), dif fusion and perfusion techniques within one single examination, virtual endoscopic examinations (eg, bronchoscopy and colonography), and spectroscopy (eg, prostate cancer). The injection of iodinated contrast agents that are potentiall y nephroto xic is not necessary. Ne vertheless, there are limitations in the detection of li ver metastases due to ir regular breathing, the dif ferentiation of l ymph nodes, and the small and large intestine, and in the detection of osseous metastases. The main advantage of PET is the identification of cancerous lesions be yond the diagnostic sensiti vity of MRI, which allows the diagnosis of malignancies at an earlier stage. PET can be used to spot areas of malignancy, monitor tumor g rowth, predict response to therapy, and monitor therapeutic response and tumor recurrence. Malignancies with low or nor mal biochemical activity (eg, mucinous carcinomas, primary renal cell carcinoma, and prostate cancer) in the PET image component may show clearly positive or suspicious f indings in the MRI component of the h ybrid PET/MRI. MRI demonstrates metastases not found on PET . PET imaging is v ery sensiti ve but sometimes suf fers from specificity. MRI allo ws an e xact localization of the lesions, their differential diagnostic clarification, and the visualization of their patho gnomonic mor phologic appearance. A hybrid PET/MRI scanner can compensate for the disadv antages of using PET and MRI separatel y and therefore yields a clear impro vement of diagnostic accuracy by combining two already excellent modalities. The combination of w hole-body PET and w hole-body MRI into a single scanner of fers the ability for accurate registration of molecular aspects and biochemical alterations of a wide variety of diseases with e xact

38

MOLECULAR IMAGING: PRINCIPLES AND PRACTICE

correlation to anatomic/mor phologic f indings w hen using these two excellent modalities in oncolo gic applications. A h ybrid PET/MRI w ould be highl y a dvantageous in clinical practice in impro ving the diagnostic value of PET and MRI in identifying and characterizing malignancies and in tumor staging and assessment of therapeutic response and tumor recurrence. The MR images could be used for the attenuation cor rection, could pro vide high-quality anatomic details and the precise anatomic localization of increased radiotracer uptake of the PET imaging, and could be used for measurements of tumor lesions in distance and v olume. The PET images would be used for quantitative analysis of tracer accumulation in tumor lesions. 18F-FDG PET/CT has been sho wn to be v ery effective for early prediction of neoadjuv ant chemotherap y response in patients with metastatic tumor disease w ho underwent examination for therap y monitoring. Anatomic/morphologic imaging methods could re veal an increased necrosis of the primar y tumor and metastases and a progressive sclerosis of the l ytic osseous metastases that could be misinter preted as tumor pro gression. Anatomic/morphologic imaging procedures rel y on anatomic str ucture changes of disease and allo w the detection of changes in tumor size and v olume, but do not allow a differentiation between viable tumor tissue and f ibrotic scar tissue due to treatment. The morphologic criteria of reduction of tumor volume as evidence of response to therap y requires a cer tain time dela y after initiation of therap y and ma y be mask ed b y unspecific effects (eg, edema as a result of necrosis). Thus, it is expected that 18F-FDG PET will be useful in reducing the costs of c ytotoxic therap y and the unnecessary side ef fects of inef fective chemotherap y. Whole-body h ybrid PET/MRI is a v ery promising diagnostic modality for oncologic imaging and for use in cancer screening due to the considerably lower radiation e xposure in contrast to h ybrid PET/CT and the high soft-tissue resolution of MRI in contrast to CT.5,54 For the best pro gnostic stratif ication and to guide the most appropriate therapeutic approach, the use of the most sensitive imaging procedure to detect metastases and evaluate the number of metastases and their distribution should be recommended. An adequate diagnostic cer tainty for the detection of the e xtent of tumor metastases is not attained with any single imaging procedure. The whole-body hybrid PET/MRI scanner will be the state-of-the-ar t diagnostic imaging modality in oncologic applications in the decades to come and also possibly will be used in cancer screening and cardiovascular imaging.

REFERENCES 1. Blomqvist L, Torkzad MR. Whole-body imaging with MRI or PET/CT: the future for single-modality imaging in oncolo gy? JAMA 2003;290:3248–9. 2. Kluetz PG, Meltzer CC, Villemagne VL, et al. Combined PET/CT imaging in oncolo gy: impact on patient management. Clin Positron Imaging 2000;3:223–30. 3. Lardinois D, Weder W, Hany TF, et al. Staging of non-small-cell lung cancer with inte grated positron-emission tomo graphy and computed tomography. N Engl J Med 2003;348:2500–7. 4. Seemann MD . PET/CT : fundamental principles. Eur J Med Res 2004;9:241–6. 5. Seemann MD. Whole-body PET/MRI: the future in oncological imaging. Technol Cancer Res Treat 2005;4:577–82. 6. Shreve PD, Anzai Y, Wahl RL. Pitf alls in oncolo gic diagnosis with FDG PET imaging: physiologic and benign variants. Radiographics 1999;19:61–77. 7. Seemann MD. Diagnostic v alue of PET/CT for predicting of neoadjuvant chemotherapy response. Eur J Med Res 2007;12:90–1. 8. Seemann MD. Human PET/CT scanners: feasibility for oncological in vivo imaging in mice. Eur J Med Res 2004;9:468–72. 9. Biersack HJ, Bender H, P almedo H. FDG-PET in monitoring therap y of breast cancer . Eur J Nucl Med Mol Imaging 2004;31 Suppl 1:S112–7. 10. Brücher BL, Weber W, Bauer M, et al. Neoadjuv ant therap y of esophageal squamous cell carcinoma: response e valuation b y positron emission tomography. Ann Surg 2001;233:300–9. 11. Scheidhauer K, Walter C, Seemann MD. FDG PET and other imaging modalities in the primar y diagnosis of suspicious breast lesions. Eur J Nucl Med Mol Imaging 2004;31(Suppl 1):S70–9. 12. Chatziioannou AF. Molecular imaging of small animals with dedicated PET tomographs. Eur J Nucl Med 2002;29:98–114. 13. Myers R. The biolo gical application of small animal PET imaging. Nucl Med Biol 2001;28:585–93. 14. Seemann MD , Beck R, Zie gler S. In vi vo tumor imaging in mice using a state-of-the-ar t clinical PET/CT in comparison with a small animal PET and a small animal CT . Technol Cancer Res Treat 2006;5:537–42. 15. Kaltsas G, Rockall A, Papadogias D, et al. Recent advances in radiological und radionuclide imaging and therap y of neuroendocrine tumours. Eur J Endocrinol 2004;151:15–27. 16. Reubi JC, Kvols LK, Waser B, et al. Detection of somatostatin receptors in surgical and percutaneous needle biopsy samples of carcinoids and islet cell carcinomas. Cancer Res 1990;50:569–77. 17. Gibril F, Re ynolds JC, Doppman JL, et al. Somatostatin receptor scintigraphy: its sensiti vity compared with that of other imaging methods in detecting primar y and metastatic gastrinomas. Ann Intern Med 1996;125:26–34. 18. Krenning EP, Kw ekkeboom DJ , Bakk er WH, et al. Somatostatin receptor scintigraphy with [111In-DTPA-D-Phe1]- and [123I-Tyr3]octreotide: the Rotterdam e xperience with more than 1000 patients. Eur J Nucl Med 1993;20:716–31. 19. Lebtahi R, Cadiot G, Sarda L, et al. Clinical impact of somatostatin receptor scintigraphy in the management of patients with neuroendocrine gastroenteropancreatic tumors. J Nucl Med 1997;38:853–8. 20. Schillaci O, Scopinaro F, Angeletti S, et al. SPECT improves accuracy of somatostatin receptor scintig raphy in abdominal carcinoid tumors. J Nucl Med 1996;37:1452–6. 21. Schillaci O, Spanu A, Scopinaro F, et al. Somatostatin receptor scintigraphy in li ver metastasis detection from gastroenteropancreatic neuroendocrine tumors. J Nucl Med 2003;44:359–68. 22. Kwekkeboom D, Krenning EP, De Jong M. Peptide receptor imaging and therapy. J Nucl Med 2000;41:1704–13. 23. Patel YC. Somatostatin and its receptor family. Front Neuroendocrinol 1999;20:157–98.

PET/MRI

24. Meisetschlaeger G, Poethko T, Stahl A, et al. Gluc-Lys ([18F]FP)-TOCA PET in patients with SSTR-positi ve tumors: biodistribution and diagnostic e valuation compared with [111In]DTP A-octreotide. J Nucl Med 2006;47:566–73. 25. Wester H-J , Schottelius M, Scheidhauer K, et al. PET imaging of somatostatin receptors: design, synthesis and preclinicalevaluation of a novel 18F-labelled, carbohydrated analogue of octreotide. Eur J Nucl Med Mol Imaging 2003;30:117–22. 26. Piert M, Machulla HJ , Picchio M, et al. Hypo xia-specific tumor imaging with 18F-fluoroazom ycin arabinoside. J Nucl Med 2005; 46:106–13. 27. Nordsmark M, Hoyer M, Keller J, et al. The relationship between tumor oxygenation and cell proliferation in human soft tissue sarcomas. Int J Radiat Oncol Biol Phys 1996;35:701–8. 28. Harrison LB, Chadha M, Hill RJ , et al. Impact of tumor h ypoxia and anemia on radiation therap y outcomes. Oncolo gist 2002; 7:492–508. 29. Martin DR, Semelka RC. Imaging of benign and malignant focal liver lesions. Magn Reson Imaging Clin N AM 2001;9:785–802. 30. Gaa J, F ischer H, Laub G, Geor gi M. Breath-hold MR imaging of focal li ver lesions: comparison of f ast and ultraf ast techniques. Eur Radiol 1996;6:838–43. 31. Damadian R. F ield focusing n.m.r . (FONAR) and the for mation of chemical images in man. Philos Trans R Soc Lond B Biol Sci 1980;289:489–500. 32. Lauterbur PC. Pro gress in n.m.r . zeugmato graphy imaging. Philos Trans R Soc Lond B Biol Sci 1980;289:483–7. 33. Engelhard K, Hollenbach HP, Wohlfart K, et al. Comparison of wholebody MRI with automatic mo ving tab le technique and bone scintigraphy for screening for bone metastases in patients with breast cancer. Eur Radiol 2004;14:99–105. 34. Kavanagh E, Smith C, Eustace S. Whole-body turbo STIR MR imaging: controversies and avenues for development. Eur Radiol 2003; 13:2196–205. 35. Eustace S, Tello R, DeCarvalho V, et al. A comparison of whole-body turbo-STIR MR imaging and planar 99m Tc-methylene diphosponate scintigraphy in the examination of patients with suspected skeletal metastases. AJR 1997;169:1655–61. 36. Eustace S, Tello R, DeCar vallho V, et al. Whole-body turbo-STIR MRI in unknown primary tumor detection. J Magn Reson Imaging 1998;8:751–3. 37. Johnson KM, Leavitt GD, Kayser HW. Total-body MR imaging in as little as 18 seconds. Radiology 1997;202:252–6. 38. O Connell MJ, Hargaden G, Powell T, Eustace SJ. Whole-body turbo short tau inversion recovery MR imaging using a moving tabletop. AJR 2002;179:866–8. 39. Barkhausen J, Quick HH, Lauenstein T, et al. Whole-body MR imaging in 30 seconds with real-time tr ue FISP and a continuousl y rolling table platform: feasibility study. Radiology 2001;220:252–6.

39

40. Lauenstein TC, Freudenberg LS, Goehde SC, et al. Whole-body MRI using a rolling table platform for the detection of bone metastases. Eur Radiol 2002;12:2091–9. 41. Bader TR, Semelka RC, Chiu VC, et al. MRI of carcinoid tumors: spectrum of appearances in the gastrointestinal tract and li ver. J Magn Reson Imaging 2001;14:261–269. 42. Dromain C, de Baere T, Baudin E, et al. MR Imaging of hepatic metast ases caused b y neuroendocrine tumors: comparing four techniques. AJR 2003;180:121–8. 43. Paulson EK, McDermott VG, Keogan MT, et al. Carcinoid metastases to the li ver: role of triple-phase helical CT . Radiolo gy 1998; 206:143–50. 44. Bollen EC, Goei R, v an’t Hof-Grootenboer BE, et al. Interobser ver variability and accuracy of computed tomo graphic assessment of nodal status in lung cancer. Ann Thorac Surg 1994;58:158–62. 45. Weissleder R, Mahmood U . Molecular imaging. Radiolo gy 2001;219:316–33. 46. Antoch G, Vogt FM, Freudenberg LS, et al. Whole-body dual-modality PET/CT and w hole-body MRI for tumor staging in oncolo gy. JAMA 2003;290:3199–206. 47. Seemann MD, Meisetschlaeger G, Gaa J , Rummeny EJ. Assessment of the e xtent of metastases of gastrointestinal carcinoid tumors using whole-body PET, CT, MRI, PET/CT and PET/MRI. Eur J Med Res 2006;11:58–65. 48. Gaa J , Rummen y EJ , Seemann MD . Whole-body imaging with PET/MRI. Eur J Med Res 2004;9:309–12. 49. Schaefers KP. Imaging small animals with positron emission tomo graphy. Nuklearmedizin 2003;42:86–9. 50. Tatsumi M, Nakamoto Y, Traughber B, et al. Initial experience in small animal tumor imaging with a clinical positron emission tomo graphy/computed tomography scanner using 2-[F-18]Fluoro-2-deoxyD-glucose. Cancer Res 2003;63:6252–7. 51. Melder RJ, Brownell AL, Shoup TM, et al. Imaging of activated natural killer cells in mice b y positron emission tomo graphy: preferential uptake in tumors. Cancer Res 1993;53:5867–71. 52. Seemann MD, Schaefer JF, Englmeier K.-H. Virtual positron emission tomo graphy/computed tomo graphy-bronchoscopy: possibilities, advantages and limitations of clinical application. Eur Radiol 2007;17:709–715 (Epub 2006 August 15). 53. Seemann MD . Detection of metastases from gastrointestinal neuroendocrine tumors: prospecti ve comparison of 18F-TOCA PET, triple-phase CT and PET/CT. Technol Cancer Res Treat 2007;3: 213–20. 54. Seemann MD , Gaa J . Images in cardio vascular medicine. Cardiac metastasis: visualization with positron emission tomo graphy, computed tomo graphy, magnetic resonance imaging, positron emission tomography/computed tomography, and positron emission tomo graphy/magnetic resonance imaging. Circulation 2005;112:e329–30.

4 SPECT

AND

SPECT/CT

BRIAN F. HUTTON, PHD AND FREEK J. BEEKMAN, PHD

Today, the majority of clinical procedures using tracers to visualize specific tissue binding sites are performed with planar gamma-camera imaging, single photon emission computed tomo graphy (SPECT), and positron emission tomography (PET). Ev en after the recent e xplosive growth of clinical PET and PET/computed tomo graphy (CT) installations, the imaging of single-photon emitting radiopharmaceuticals with gamma cameras, both in planar mode or with SPECT , mak es up b y f ar the lar gest fraction of clinical nuclear medicine imaging procedures (approximately 95% in Europe in 2007). Man y clinically established tracers for gamma cameras are commerciall y available and used in imaging depar tments. The singlephoton emitting radionuclides used as labels for tracer molecules often have sufficiently long half-lives to allow for long-distance transportation or can be obtained on site via generator systems. Tracers for SPECT can often be readily prepared on site using commercial reagents and kits. Therefore, in contrast with PET , the infrastr ucture associated with cyclotron production is not required. Traditionally, SPECT has evolved using conventional planar systems mounted on some form of rotating assembly, designed for fle xibility of both planar and SPECT acquisition. There is a trend to ward application-specif ic systems, optimized for a specif ic purpose (heart, breast). Another trend in nuclear medicine is toward systems that provide molecular/functional images (PET, SPECT) registered with str uctural/anatomical images obtained with CT or magnetic resonance imaging (MRI). These are of crucial impor tance for research, diagnosis, and patient treatment. Multimodal approaches such as SPECT/CT and PET/CT have already proven to significantly enhance accuracy of diagnosis and patient management in man y cases. The main reason for this is that molecular processes that show up at the site of, for example, a tumor or infection process can be accuratel y localized in an anatomical framework and attributed to a specif ic tissue or an or gan. This combination of both modalities 40

presents added diagnostic information compared to either in isolation. Fur thermore, there is potential to enhance reconstruction and impro ve quantif ication of (local) amounts of radionuclides in the body using the combined information from multiple modalities. Like PET/CT, integrated SPECT/CT devices acquire both emission and transmission tomo graphy with the patient in the same position. Images can be readil y adjusted for differences in for mat and scanner geometr y to overlay the images. Grafting the high spatial resolution capabilities of toda y’s high-speed CT scanners with SPECT’s accurate def inition of disease processes v astly enhances anatomical mapping and localization, mo ving the new dual modality systems directly into a wider range of clinical applications. An impor tant additional adv antage is that with the re gistered CT scan, attenuation correction can be accuratel y applied, which greatly reduces the problems of distor tion and quantitati ve inaccuracies that typically occur with stand-alone SPECT. The goal of this chapter is to acquaint a broad readership with the principles of SPECT and SPECT/CT . In addition, we will attempt to place moder n SPECT in the perspective of past and future SPECT instr umentation and its clinical applications. A primer on the ph ysics of SPECT can be found in Cher ry and colleagues; 1 the reader interested in fur ther detail is refer red to Wernick and Aarsvold.2 A review of nuclear molecular imaging combined with str uctural/anatomical imaging can be found in the study by Cherry.3

IMAGING SINGLE-PHOTON EMITTING RADIONUCLIDES The Gamma Camera The gamma camera, almost al ways based on the so-called Anger principle,4 continues to be the main component of most commerciall y a vailable SPECT

SPECT and SPECT/CT

systems. These systems consist of one to four camera heads mounted on a gantr y so that the y can be rotated around the patient. The main consideration in the design of these systems in the past has been v ersatility of use since in man y applications SPECT has been considered complementary to planar imaging. This is still tr ue in some areas of clinical application (e g, bone scans) although in other areas (e g, myocardial perfusion and brain receptor studies) SPECT is the method of choice. Although dedicated SPECT systems ha ve been de veloped in research centers based on a full ring of detectors (like PET), the y lack v ersatility and compromise the need to position detectors close to the patients at all times; de velopment has therefore centered on adapting planar detectors for SPECT use. To understand the operation of SPECT , it is therefore helpful to have a general understanding of the Anger gamma camera. The basic components of the Anger gamma camera are illustrated in F igure 1. The detector is usually a single scintillation cr ystal, most times made out of sodium iodide (NaI) with thallium impurities of dimension typically 500 × 400 mm and 9.5 mm thick. Gamma photons interact with the cr ystal producing light that emanates in all directions from the point of interaction. The origin of the interaction is deter mined from the light distribution, w hereas the ener gy deposited is propor tional to the inte gral of the light produced. The position and ener gy are deter mined b y first converting the light to a measurab le signal, most times by means of a set of photomultiplier (PM) tubes; they also magnify the small signal that is produced in the photo-sensiti ve la yer at the entrance to the PM tubes. The f inal position and ener gy are then determined electronically (although nowadays this is usually a digital rather than analo gue calculation). Each detected photon therefore is assigned to an ener gy and detector surf ace position. Since photons that scatter lose energy, the energy information can be used to discriminate radiation scattered in the patient from photons that reach the detector without ha ving undergone interaction with tissue; the location simply allows photon counting for a f ine matrix of picture elements (pixels) that cor responds to the detector area. Due to various uncertainties in the detector system, the spatial resolution of the detector itself (intrinsic resolution) is typically 3 to 4 mm full width at half maximum v alue (FWHM) (relating to the spatial distribution of counts from a point source); the ener gy resolution for NaI is approximately 10% FWHM (in this case, relating to the energy distribution for a monoenergetic emitter).

Light guide

41

Scintillation crystal Collimator

C A PM tubes

B D

Figure 1. Cross section through patient and gamma camera. Emitted gamma quanta with different trajectories are shown. A, Ray that goes through collimator parallel with a hole, B, Ray that penetrates the collimator septum, C, Ray that is captured in collimator because angle deviates too much from hole direction and thus not detected, D, Ray that results in scintillation after a scatter event in the patient.

Improvements in the camera’ s intrinsic ener gy and spatial resolution are e xpected with alter natives to PM tubes, for example, using position-sensitive PM tubes or PM tubes with a higher quantum ef ficiency. In addition, several solid state alternatives to PM tubes such as silicon PMs,5 special charge-coupled devices (CCDs) lik e electron multipl ying CCDs, 6–10 avalanche photodiodes, and silicon drift diodes 11 are under development. See Pichler and Ziegler12 for a useful summary of some of these technologies. Also, se veral ne w scintillator materials ha ve been discovered or are under development (eg, LaBr3:Ce and LaCl:Ce), 13–15 which have signif icantly higher light output and density than NaI and other benef icial properties to improve gamma camera performance (see reviews by Moses and Shah 16). Scintillation cr ystals can also be improved by str ucturing the material in such a w ay that light spread is reduced , w hich is par ticularly impor tant when high-resolution light readout is possib le. Examples include monolithic cr ystals with g rooves, pix elated crystals, or tin y CsI needles that are g rown on a special substrate.9 In addition, alter natives to scintillation detectors are being de veloped such as solid state detectors (e g, cadmium zinc telluride [CZT]) that directly convert deposited γ-ray ener gy into electrical signals (see Vavrik and colleagues17 and summar y by Wagenaar18). The elimination of a light-conversion step results in good signal collection with ener gy resolution much impro ved compared with scintillation detectors (typicall y appro ximately 5% at

42

MOLECULAR IMAGING: PRINCIPLES AND PRACTICE

140 keV for medical systems that are based on pix elated detectors). This superior ener gy resolution pro vides improved discrimination of multiple radionuclides. An additional benef it is the small size of these detectors, which enab les g reater fle xibility in system design (see recent developments in the section below). A key element of the overall system is the collimator, which consists of a lead or tungsten “hone ycomb” for a parallel-hole collimator, designed so as to ideall y eliminate all photons that are not tra veling nor mal to the detector surf ace. The presence of a collimator limits the direction of the incoming photons. Without this, it becomes extremely hard to determine the origin of detected photons. To gi ve some perspecti ve, for a low-energy high-resolution collimator, the hole dimensions are approximately 35 mm long, 1.4 mm diameter with lead septa 0.15 mm thick. The collimator design largely deter mines not onl y the o verall spatial resolution of the system determined by the hole diameter and length but also the number of counts acquired per unit of activity or ef ficiency of the system. The collimator dimensions can be altered to def ine a dif ferent performance, but increasing ef ficiency b y increasing hole size will result in de gradation of resolution; there is usually a compromise between these two parameters. In addition, thickness of septa needs to be adjusted to minimize penetration w hile imaging higher ener gy radionuclides. Despite the man y de velopments in the Anger camera design since its inception, for most of the presently used commercial systems, the collimator remains the main component that deter mines o verall performance. The lo w sensiti vity means that studies must be acquired for a relati vely long time to accumulate sufficient counts. The only alternative would be to increase the administered acti vity, but this is limited by the radiation dose to the patient. Although parallelhole collimators are b y f ar the most commonl y used collimators, there are alter native systems that are f inding increasing application. Some of these are illustrated in Figure 2. Of par ticular interest are pinhole collimators that have proved ideal for preclinical imaging and various combinations of slits (eg, crossed slits 19 or slitslat collimators 20–22) that are being e xplored in combination with ne w detector designs. The main feature of the pinhole (and slit) collimator is that the acquired image is magnif ied, providing gains in both resolution and sensiti vity pro vided a small v olume placed close to the collimator is imaged , hence the attraction in imaging small or gans (e g, th yroid) and small animals (see Chapter 7 “microSPECT/CT/MRI”).

General Principles of SPECT Acquisition with Multihead Cameras SPECT Acquisition

The standard SPECT acquisition in volves rotation of the detectors with collimators around the patient to acquire a sufficient number of angular vie ws (“projections”) to enable reconstr uction of 3D v olume images. Although alternatives to rotation-based systems ha ve been built (eg, use of many pinholes with stationar y detectors, Rowe and colleagues 23), most principles of acquisition and reconstruction are similar . F or simplicity, the discussion will be limited to a rotating camera system. The cameras are kept as close as possib le to the patient during rotation often with the aid of automatic laser-based contouring control (eg, using a 180° orbit in the case of cardiac imaging). Provided the camera is perfectl y aligned with the axis of rotation, each row of the acquired projection pix els corresponds to a unique slice of acti vity and these projections are used for reconstr uction. Acquisition therefore involves simultaneous acquisition of multiple slices, with each projection angle being acquired for typically 10 to 40 seconds depending on the number of angles and total scan duration. This is quite unlik e PET w here data from all projection angles are acquired simultaneousl y, albeit for a smaller axial extent, or early model CT w here data were acquired very fast for a single slice (multi-slice acquisition on more recent CT systems has closer similarity to gamma camera acquisition). The projection-b y-projection acquisition of SPECT places limits on the ability to acquire dynamic studies although recently proposed designs have less limitations (see later subsection). Alternatively cardiac SPECT can be acquired using electrocardio graph (ECG) gating to evaluate cardiac w all motion, m yocardial thickness, and pump function (ejection fraction). There are some quality control (QC) considerations that are specif ic to SPECT ; these considerations mainl y address potential prob lems that occur with the type of acquisition system. Since indi vidual detectors are rotated around the patient, any focal non-uniformity on the detectors, especially near the center of the detector , will result in a visib le ring ar tifact on the reconstr ucted image. Therefore, careful attention to uniformity with regular QC checks is recommended. The rotation of a relatively heavy detector (due to lead collimator and shielding) in the past has resulted in prob lems of mechanical stability , usuall y reflected in some error in the mechanical – electrical alignment of the system. This er ror in the center of rotation (COR) can result in de gradation of resolution and hence

SPECT and SPECT/CT

43

L P

A

B

C

D

E

F

Figure 2.

Schematic of different collimators. A, parallel hole, B, fan-beam, C, cone-beam, D, pinhole, E, crossed slits, F, slit-slat.

the COR should be checked regularly. Correction normally involves a simple image shift, so the prob lem is easil y rectified. F ortunately, recent systems tend to use f airly sturdy gantries to ensure mechanical stability . Ho wever, note that system calibration becomes critical for pinhole systems where small er rors can be v ery much magnif ied; failure to accuratel y calibrate these systems can lead to extreme deterioration of image quality. Tomographic Reconstruction

There exist two main approaches to calculate the SPECT image volume from projection images (tomographic image reconstruction): (i) anal ytic methods that calculate the image by estimating the in verse of a for mula that represents the image for mation process or (ii) discrete methods that are in principle based on using a matrix representation

of the image for mation process. The image for mation process in this conte xt is ho w photons from positions (x, y, z) in the patient are mapped on projections (x’, y’). Image reconstr uction with anal ytical methods includes algorithms kno wn as “f iltered back-projection” for parallel and f an-beam collimators (Bar rett and Swindel24 and Tsui and F rey25) or for cone and pinhole geometries, methods lik e the F eldkamp algorithm. 26 In a very much simplif ied model of the emission process, the only knowledge of the source activity distribution that can be deduced from the measured projection (for a parallelhole collimator) is the line perpendicular to the detector at the point of interaction, along which the photon must have originated; the back-projection process ef fectively uses this assumption to allocate equal v alues to pix els that lie along this line. The super position of the line intensities determines the pixel intensities for the reconstructed slice.

44

MOLECULAR IMAGING: PRINCIPLES AND PRACTICE

Imagine first that there is only one radioactive point in the patient; then, in an ideal imaging system, this will result in a single spot on each projection. At the position of the radioactive source, the back projection results in a maximum, but around it, there is a signif icant amount of “reconstructed activity.” In the absence of noise, this blurring effect that occurs b y the summation of crossed lines can be exactly compensated with a special f ilter (referred to as a ramp f ilter). In practice, w here noise is present, some degree of local smoothing must also be applied. The advantage of anal ytical methods is their computational speed, but (i) they have limited robustness to quantum noise that is present in projections, (ii) the y usually do not compensate for image b lurring ef fects, and (iii) they are not really flexible enough to handle the complicated collimator geometries and detector placements that can be designed to e xtract additional information from the object being imaged. Therefore, most modern systems are equipped with iterati ve methods of reconstruction, such as the maximum lik elihood expectation maximization (ML-EM) algorithm (Lange and Carson27), and accelerated versions such as ordered subsets EM (OS-EM).28,29 ML-EM is a statistical algorithm since it tak es into account the characteristics of the Poisson noise in acquired projection pixels. In addition, the algorithms can incor porate models of image de gradation to compensate for these ef fects. Examples of effects being compensated b y iterati ve methods are nonuniform gamma ray attenuation, distance-dependent sensitivity, and scatter.30 Other factors such as spatiall y variant resolution and radiation penetration along the collimator hole edges can also be included 31,32 to

correct for these image de grading ef fects. As a result, statistical algorithms in general produce images with better resolution and ha ve better noise characteristics than analytic algorithms. Iterative Image Reconstruction

The task in iterative SPECT reconstruction is to calculate the 3D distribution of activity based on measured projections. Iterative reconstruction is based on the idea that the calculated 3D activity distribution is close to the real distribution when simulated projections based on the acti vity estimate closel y match projections acquired b y the SPECT camera. The process of iterative reconstruction is the iterative estimation of the best solution to this problem and in volves the repeated application of a set of operations (including the simulation of projections often referred to as forw ard projection) that pro gressively provides a solution that gets closer and closer to a correct estimate of the unkno wn activity distribution. F igure 3B shows the concept of this iterative updating. In practice, before the iterati ve calculations can star t the relationship between the object v oxels and measured projections has to be estab lished. F or a specif ic object voxel A i, one can estimate the probability that a photon will be detected at a specif ic detector pix el P j based on knowledge of the detector geometr y and lik elihood that photons will be emitted from the object (eg, patient); this probability is represented as an element of a matrix M ji. These elements need to be kno wn for each indi vidual voxel-pixel combination (this is often refer red to as the system model or system matrix). The entire set of

Projection space

Object space

Current estimate Ae M11 . A1

M12 . A2

......

M1V . AV

P1

M21 . A1

M22 . A2

......

M2V . AV

P2

A

MU2 . A2

......

MUV . AV

Estimated projection Pe “Compare” e.g. - or/ Measured projection P

Update

Object error map MU1 . A1

Simulation step M Ae

“Backprojection”

“Error” projection

PU

B

Figure 3. Frame (A): equations describing how activity in the object is mapped onto the projection images. Each projection pixel Pj measures the sum of contributions from all object voxels Ai, where Mji is the probability of photons being detected as defined by the system model (see text). The iterative scheme is shown in frame (B).

SPECT and SPECT/CT

measured projection pix els is represented b y a v ector P. The numbers M ji together with P determine the set of equations from which the activity distribution A has to be determined (see F igure 3A). During an iteration of the ML-EM algorithm, the actual estimate of A (which we call Ae) is used to generate an estimate of the projections, denoted with vector P e, simply by carrying out the summations, as presented in Figure 3A, but withAe instead of A. Next, ML-EM uses the ratio P/Pe to calculate errors in the projections from which an object er ror map is reconstructed (the dif ference (P − Pe) is used in alter native algorithms). The error map is then used to update Ae. The generation of a new Pe and the updating of A e often need to be repeated hundreds of times to obtain a good solution. Because of the requirement of man y iterations, acceleration methods to speed up the algorithms ha ve been developed. The OS-EM algorithm 27 is currently the most popular method. It updates the solution based on only a subset of projections rather than recalculating the update using all projections, w hich in volves much less computation. Overviews of the subject of iterati ve SPECT image reconstruction and information on how to car ry out comparisons and update steps during reconstr uction can be found in studies b y Hutton and colleagues 33 and Lalush and Wernick.34 The accurate deter mination of the matrix elements of A can be dif ficult, often requiring comple x calculations and/or measurements that are specific to each different SPECT device. An accurate match of the matrix elements and real detection probabilities has a criticall y important influence on the reconstructed image; the number of iterations and quality of image smoothing for noise suppression are also impor tant. With the de velopment of advanced algorithms that are cur rently available to reconstruct images from comple x geometries, images of superior resolution and quantitative accuracy can be produced.

QUANTITATIVE SPECT Most applications of SPECT rely on qualitative interpretation rather than absolute quantif ication; ho wever, the ultimate goal is to reduce or remove any artifacts and ideally to provide a map of the absolute acti vity concentration in the patient. A number of f actors tend to limit quantitative accurac y (as w ell as diagnostic quality): attenuation and scatter of photons in tissue, the limited resolution of SPECT , and presence of motion. Approaches e xist for the cor rection of all these f actors although there tends to be no standard approach, w hich leads to signif icant intersite v ariability in results. The objective of this section is to pro vide some insight to the

45

problems and an introduction to some of the a vailable approaches for correction. The reader is referred to more detailed coverage of this topic in the re views by Wernick and Aarsvold2 and Zaidi.35

Corrections for Attenuation The naïv e assumption that underlies f iltered backprojection (FBP) reconstruction is that the recorded counts simply reflect the line integrals for the activity distribution; of course in reality emitted photons interact with tissue and so are Compton scattered resulting in attenuation of the emitted radiation. It is con venient to distinguish attenuation as the reduction of measured counts compared with what would be expected in air versus scatter as an increase in the measured counts due to the inclusion of some scattered radiation that still falls in the selected energy window (misplaced compared to the primary events) (see Figure 1). The consequence of attenuation is there will be reduced number of photons originating from depth in tissue, reflected b y reduced reconstr ucted values central to the patient. The situation with non-unifor m attenuation is more complex as the v ariable degree of attenuation leads to inconsistencies in the recorded projections with at times quite serious ar tifacts that are not immediatel y obvious. For example, there is a tendency for increased counts to be reconstructed in areas of low attenuation with a reduction in counts in areas of higher attenuation. In practice, tw o distinct approaches are in widespread use. The Chang correction36 is an appro ximate cor rection, w hich can be sufficient for qualitati ve studies w here there is unifor m attenuation; it simpl y in volves estimating the a verage attenuation at each point in the subject and using this as a post-reconstruction cor rection f actor at the point concerned. Although modif ied Chang algorithms ha ve been developed to deal with non-unifor mities in attenuation, iterative reconstr uction that incor porates a measured attenuation map is preferred for these cases. The measured attenuation map is used to modify the system matrix that is used to estimate the probability of detection at a projection pixel given activity at a specif ic voxel (referred to as Mji in the section on reconstr uction above). Provided an attenuation map is a vailable, the implementation is relatively straightforw ard as the algorithm (e g, OS-EM) is identical except for the different system matrix. A number of techniques w ere de veloped to measure the attenuation in combination with SPECT . They originally involved use of an external radionuclide that was used as a transmission source, ef fectively recording a lo wquality transmission scan similar to CT . Various source configurations were used (see Bailey37 for a comprehensive

46

MOLECULAR IMAGING: PRINCIPLES AND PRACTICE

overview). Ho wever, the commercial implementations of most of these systems ha ve been tested and pro ved to be limited.38 Potentially interesting alter natives to these systems involve use of a scanning point source in combination with f an-beam collimators 39 or stationar y point sources combined with cone-beam collimators. 40 In the earl y nineties, Hase gawa and colleagues 41 proposed the use of CT for obtaining the transmission map, in effect the f irst suggested dual-modality system. The f irst commercial system, partly motivated by the need for attenuation cor rection for use in gamma camera coincidence imaging, w as released b y GE Healthcare in 1999 with other companies no w also of fering SPECT/CT systems. These pro vide transmission maps of higher quality than radionuclide methods although there can be prob lems due to motion ar tifacts in the thorax and mismatch of breathheld CT with emission data (see later section SPECT/CT).

Scatter Correction Photons interacting with tissue under go Compton scatter, with deflection and consequent loss of ener gy. Although energy discrimination is used to reduce the number of detected scatter photons, there are scattered photons that cannot be eliminated, predominantly for photons that have undergone a single interaction in tissue. They typicall y constitute 25 to 40% of detected photons. The spatial distribution of scatter from each point of acti vity is quite wide but is constrained to within the boundar y of the patient (unlike the situation in PET w here scatter e vents can lead to coincidence lines of response outside the body boundary). Man y methods for scatter cor rection ha ve been suggested (see re views35,42), but fe w are routinel y applied in practice. The most popular practical approach is the triple energy window method,43 where narrow energy windows are def ined close to the photopeak to estimate the scatter within the photopeak. The use of direct measurement overcomes the limitations of models that mak e simplifying assumptions regarding the scatter distribution, which is a distinct adv antage. A disadvantage is the additional noise propagated if scatter is subtracted from projections (subtraction of tw o noisy datasets results in a corrected image w here the noise amplitude is increased , while the image v alues in the dif ference image are smaller). The alter native is to simpl y add the measured scatter as part of the forward projection step during iterative reconstruction, which significantly reduces the noise. An approach that is gaining in popularity is to model scatter, which is computationally demanding especially if Monte Carlo modeling is used 44,45; ne vertheless, with

improved computer speed and optimized algorithms, these alternatives are now becoming practical. 46

Correction for Effects of Limited Resolution The limited resolution in SPECT not onl y results in blurred images but also a reduction in the observed/measured acti vity concentration for small objects (usually refer red to as the par tial volume effect). This is a direct consequence of the spreading of counts outside tissue boundaries due to poor resolution. Cor rections are therefore concerned with both improving resolution (image shar pness) and contrast and cor recting quantitative v alues. It is becoming quite common to include details of the collimator and detector b lur in the system matrix during iterati ve reconstr uction.47–50 This encourages a solution in w hich resolution is impro ved even if quantitati ve v alues are not full y reco vered. The side benef it of this approach is that reconstr ucted noise tends to ha ve improved characteristics, to the e xtent that reduced (“half ”) imaging time is being adv ocated. However, note that the f inal reconstr ucted data will still be affected b y some de gree of par tial v olume ef fects. The absolute correction of par tial volume effects remains difficult, and most approaches are relying on the availability of high-resolution anatomical data to accuratel y def ine boundaries for the area of uptake. Provided the anatomical data can be accuratel y registered with the emission data and the re gions of interest accuratel y segmented, cor rection for par tial volume losses can be applied on the basis of selected unifor m acti vity re gions51 or for indi vidual pixels.52 This has tended to limit cor rection to specif ic study types (e g, brain perfusion, m yocardial perfusion). Correction for lesions has mainly involved the assumption that lesion shape is spherical, in w hich case precalculated corrections can be applied. Attractive alternatives are iterative deconvolution algorithms that do not depend on the availability of anatomical data. 53 In all cases, an accurate estimate of the resolution must be available. Since resolution is position and object dependent (and depends on the number of iterations in the case of iterati ve reconstr uction), estimation of resolution is not tri vial (see the comprehensive review by Rousset and Zaidi 54).

Motion Correction It is some what ironic that gi ven the sophistication of instrumentation and reconstruction algorithms, correction for patient or or gan motion remains lar gely unsolv ed.

SPECT and SPECT/CT

Despite efforts to minimize motion with positioning aids, patient motion is hard to a void and cer tainly involuntary motion (eg, heart and lungs) cannot be avoided. It is common to use electronic gating to freeze c yclic motion at selected times relati ve to a measured signal (ECG for heart or chest displacement for lung). F or brain studies, the motion can be assumed to be rigid and so correction is more straightforward. This can be achieved using external motion measurement (eg, using optoelectronic or mechanical devices)55–57 where measured motion is incor porated as part of the reconstruction process either using an interiteration registration or once again modifying the system matrix.58,59 Data-driven correction for head motion during SPECT acquisition has also been demonstrated. 60,61 In some respects, the slow acquisition of SPECT can be useful in that it simpl y averages motion effects with a resultant b lurring but minimal ar tifacts. Ho wever, prob lems arise in re gistering SPECT data with data acquired more quickly, such as breath-held CT, since there will be a mismatch betw een the tw o studies. 62,63 Correction for this mismatch is the subject of cur rent research (see section SPECT/CT).

SPECT/CT We ha ve discussed the need for attenuation cor rection, especially in applications w here quantif ication is required. To accurately correct for attenuation, especially where there is non-unifor m attenuation, there is a need for a map of the attenuation coef ficients that can be derived from a CT scan. The first SPECT/CT system was designed specif ically for this pur pose,41 predating the introduction of PET/CT 64 by several years. A comparative study of ef fectiveness of attenuation cor rection clearly demonstrated the v alue of CT compared to the rather imprecise systems a vailable commerciall y using radionuclide transmission sources. 38 The a vailability of anatomical infor mation also aids g reatly in localization, especially in the case of tracers w here uptak e is highl y specific, for e xample, labeled monoclonal antibodies (Figure 4). The availability of the dual-modality system continues to f ind widening application to complement SPECT alone, although acceptance and application has been less dramatic than in the case of PET/CT, where use of CT in combination with Fluorine-18 labeled deoxyglucose (FDG) has had a significant impact on clinical diagnosis. In this case, full diagnostic CT systems are used as standard, whereas at the time of writing, there remains a range of specif ications for the CT component used in combination with SPECT (Table 1). The necessary specifications depend to some e xtent on the application

47

(eg, localization or simpl y attenuation cor rection where contrast detail is less necessar y) although clinical requirements remain some what undef ined (see Hamann and colleagues 65 for a useful comparison). Indeed , the question arises as to w hether a combined SPECT/CT gantry is required 66 or not; alternative means of combining data from SPECT and CT may be more cost effective and suf ficient, especiall y gi ven the indications for SPECT/CT studies; this ma y be achie ved either via software registration (eg, see Eberl and colleagues67) or using separate CT and SPECT gantries. 68 Guaranteeing re gistration between SPECT and CT is best achie ved through use of a combined gantr y, however, it must be noted that the studies are sequential, so there is still a possibility of movement betw een studies. Pro vided care is tak en in patient positioning with a similar bed and patient restraints, softw are re gistration should be possib le (although this assumes the patient is identical during both the e xaminations, w hich can seldom be guaranteed (eg, change in stomach contents, b ladder f illing). Although there are issues in guaranteeing alignment, the use of a single CT unit in combination with multiple cameras would appear to be an attractive solution. There are several considerations in using CT in combination with SPECT, which are outlined below.

Conversion of CT Numbers to Attenuation Coefficients Since CT depends directly on measuring electron density in tissue, the conversion from Hounsf ield Units to attenuation coefficients for a gi ven radionuclide is relati vely straightforward. Unfor tunately, attenuation coef ficients are energy dependent and consequently the required conversion is non-linear . A bilinear function w as suggested for application in PET studies69 and a similar approach is adopted for SPECT radionuclides. 70 A limitation in this conversion is that it results in signif icant over-estimation of the attenuation coef ficient for contrast materials; therefore, there is a tendenc y to acquire a lo w-dose CT for attenuation cor rection without contrast rather than correct for this aber ration (in fact to date contrast material is seldom used in conjunction with SPECT).

CT Artifacts There can be additional CT artifacts that not only influence the accurac y of attenuation cor rection but also interfere with CT inter pretation (see Bockisch and colleagues71). For example, the presence of metal can

48

MOLECULAR IMAGING: PRINCIPLES AND PRACTICE

Figure 4. Top: SPECT/CT system (Symbia; Siemens Healthcare) with two examples that clearly illustrates the value of fusion imaging to localize increased uptake of radioactive tracer in the foot (Images with permission from Siemens Healthcare). Bottom: low-dose CT, SPECT and SPECT/CT images (Hawkeye 4; GE Healthcare), which illustrate specific uptake of Iodine-123 labeled Metaiodobenzylguanidine (MIBG); without CT, exact localization is difficult. Note also serious artifacts due to bowel motion during acquisition; the artifacts in this case do not interfere with clinical diagnosis. (Images courtesy of University College Hospital, London.)

introduce streak artifacts in CT and even with the available correction techniques they are hard to avoid. A simple approach is to check the presence of suspicious areas on the non-cor rected reconstructions. Streak ar tifacts can also occur due to photon star vation, for example, if imaged with ar ms down using lo w exposure, or due to beam hardening, w hich results from presence of areas of high attenuation. Artifacts can also occur due to motion, w hether patient motion or in voluntary motion of or gans during data acquisition. A f ast acquisition period minimizes patient motion b ut renders the data more susceptib le to in voluntary motions that tend to be averaged in SPECT acquisition (eg, in the absence of gating, cardiac motion can cause ar tifacts on CT).

Registration Accuracy Since CT and SPECT studies are acquired sequentiall y, there is always possibility for misre gistration. A particular concer n in earl y scanners w as the mismatch that occurred due to change in bed height with bed extension; however, newer systems have improved bed supports that rectify this potential prob lem. It is still impor tant to

measure registration between SPECT and CT as par t of the overall QC procedures. An area of particular concern arises due to the practice of acquiring a f ast CT in the lung to minimize motion ar tifacts; the breath-held study freezes the image at a par ticular phase of respirator y motion, unlik e the SPECT study that is nor mally a veraged over a lengthy acquisition period. Although respiratory gating could be applied as in PET , this is not y et common practice for SPECT.

RECENT DEVELOPMENTS Although the clinical use of SPECT is dominated b y rotating conventional gamma cameras, there is increasing interest in developing systems that offer a better compromise between resolution and noise. The fast development of ultra-high-resolution systems for preclinical imaging and the a vailability of ne w detector designs ha ve par tly stimulated this renewed interest. It is clear that designing a system for a specif ic application provides potential for optimization, whereas the multihead planar/SPECT systems are primaril y designed for v ersatility. The motivation for the cur rent de velopments is being dri ven

SPECT and SPECT/CT

49

Table 1. SUMMARY OF SUPPLIER SPECIFICATIONS AT TIME OF WRITING Supplier

Model

No of Slices

Slice Width (mm)

kV

mA

Rotation Time (s)

GE

Hawkeye 4

Siemens

4

5

120, 140

1.0–2.5

23

23–57

Symbia T

1, 2, 6, 16

0.6–19

80, 110, 130

20–345

0.5–1.5

10–517

Philips

Precedence

6, 16

0.6–12

90, 120, 140

20–400

0.4–2.0

8–800

Philips

Brightview

140 (@1 mm)

0.3–2

120

12

60–960

5–80

mAs (360°)

Note that figures are indicative and may differ for the various options offered. Note that the exposure in milli-amp-seconds (mAs) is for a 360° acquisition and will be reduced when using helical scanning (or dose-reduction techniques).

foremost b y the desire to impro ve sensiti vity (or more precisely to reduce scan time) although ef fort is also being directed to impro ve resolution. There ha ve been several recent advances in instrumentation with development of ne w detector materials (with ref inement of the technology to impro ve stability and reduce cost), much emphasis on the search for a replacement of the con ventional PM tube (position-sensiti ve PM tubes, a valanche photodiodes, silicon photon multipliers, lo w noise CCDs), and novel system and collimator designs. To improve sensiti vity, it is clear that one requires more effective use of detectors with either a larger number of detectors sur rounding the or gan of interest or alternative novel approaches to maximize the acquisition counts. F or e xample, in cardiac imaging, there is renewed interest in collimator systems that acquire data only from the heart region using either multiple pinholes or multiple slant holes; in both cases, the objecti ve is to use standard lar ge detectors to acquire multi-angle data from a single position. 72 There are also no vel systems that are specif ically designed to impro ve sensiti vity (Figure 5). The CardiArc (Canton, USA) and MarC systems73 use slit-slat collimators (parallel-hole collimation in the axial direction with pinhole collimation in the transaxial direction). A set of slits is rotated during acquisition so as to acquire multi-angle projections. The system also per mits close positioning of the detector to the patient and impro ved patient comfor t. A further system with similar attention to patient comfort is the D-SPECT system from Spectr um Dynamics (Caesarea, Israel). 74 In this case, a set of CZT detectors is programmed so that each detector rotates on its o wn axis so as to acquire counts primaril y from the hear t region. In this case, ultra-high-sensitivity collimators are used to provide an overall gain of around 8 in the counts acquired from the hear t region as demonstrated in clinical studies. 72 Using this system, cardiac study acquisition is therefore reduced to 2 to 4 minutes. The system relies on proprietar y iterative reconstr uction algorithms

to achieve a reconstr ucted image quality similar to that obtained using a con ventional SPECT system. Fur ther cardiac systems proposed with e ven higher projected performance include use of multiple pinholes75 or multiple slant holes.76 The sensitivity to cardiac activity compared with con ventional dual-head gamma camera systems can be increased by a factor of 10. Finally, cone-beam collimation combined with point source at the focal point of the collimator also has a high potential to impro ve cardiac SPECT (Manglos and colleagues40). When combined with dual or triple detector large f ield-of-view cameras and asymmetric cone-beam designs,77–79 extremely high sensiti vity from the cardiac area can be combined with v ery efficient use of the transmission sources. In this w ay, both gated SPECT data and gated transmission data, as w ell as e xcellent attenuation cor rection, could be obtained on a v ersatile SPECT system. The impor tance of the reconstr uction algorithm should not be underestimated , especiall y in the case of iterative reconstruction. There is a trend toward including increasing complexity in the system model that describes the cor respondence betw een acti vity distribution and detected counts. In par ticular, inclusion of the collimator (and detector) b lurring as a function of source distance from the collimator is pro ving v ery useful in clinical practice. The e xtension of the system model to include collimator blurring results in images with impro ved contrast and superior noise characteristics. As a result, the acquisition time can be halv ed without an y detectab le worsening of image quality compared to con ventional reconstruction (without resolution modeling). Ho wever, note that this requires increased processing time (due to the need for more comple x computation and higher than normal number of iterations). There is also potential to use higher sensitivity collimators to capture further gains in acquisition time.80 Software that offers this approach is now available from all leading suppliers (and is now also being investigated for PET 81).

50

MOLECULAR IMAGING: PRINCIPLES AND PRACTICE

A

B

C

D

Figure 5. Some new developments: A, D-SPECT (images with permission from Spectrum Dynamics), B, CardiArc (images with permission from CardiArc), C, Cardiac U-SPECT (Reproduced with permission from Beekman and van der Have75), D, Brain U-SPECT (Reproduced with permission from Goorden et al.84).

In preclinical SPECT imaging (see Chapter 7), pinhole collimators have played a central role in achieving high-resolution magnif ication for small objects, providing a v ery ef fective means of achieving submillimeter resolution. Applying similar principles to human studies is not so straightforw ard as achie ving similar magnif ication for the lar ger object size w ould require prohibiti vely lar ge detectors. Ho wever, using detectors with high intrinsic resolution, there is scope for using pinhole or crossed-slit collimators without magnification.82 In fact, using a high number of pinhole apertures with minification can pro vide significantly improved resolution and sensitivity over pinhole systems with con ventional detectors. 82–84 Dedicated systems have been developed for heart and breast imaging. Brain imagers with high resolution can also be de veloped, and the impro vement o ver presentl y a vailable systems

depends very much on the detector resolution a vailable. Demand for the latter tends to be increasing with the introduction of ne w radiophar maceuticals that of fer diagnostic promise (e g, studies of am yloid deposition and various receptor systems).

GENERAL DISCUSSION There are v arious pros and cons for the use of SPECT in comparison with studies perfor med using PET . SPECT uses single-photon emitting radionuclides that tend to have longer half-lives and consequently are more easily distributed than positron-emitting radionuclides. The radionuclides ha ve emissions at dif ferent ener gies so that simultaneous measurement of multiple tracers is possib le, unlike PET where the detected radiation always has energy of 511 keV. Also, specific binding in SPECT tracers can be

SPECT and SPECT/CT

51

Table 2. COMPARISON OF VARIOUS PARAMETERS FOR SPECT VERSUS PET SPECT

PET

Intrinsic resolution

3–4 mm

2–4 mm

Reconstructed resolution

10–15 mm (brain 8 mm)

5–8 mm

Sensitivity

0.03% (dual head)

3.0% (3D)

Energy resolution

~10%

~15–20%

Dual radionuclides

Yes

No

Attenuation (thorax)

× 10

× 20

Scatter

0.35 (140 keV)

2D ~0.15; 3D ~0.45

Random coincidences*

N/A

Proportional to square of singles event rate*

Count-rate (@10% loss)

> 250 k (total)

< 100 k (true coincidences*)

Contributors to patient exposure

Mainly gamma/long half-life

Positron/short half-life

Staff exposure

Low < 2 mSv/yr

Can be high ~4 mSv/yr

SPECT = single-photon emission computed tomography; PET = positron emission tomography. Note that numbers are indicative only and are intended to illustrate differences between the two technologies. *Readers should refer to Chapter 3 “PET/MRI instrumentation” for PET parameter definitions and further details of PET performance.

much higher than those labeled with positron emitters. SPECT still has poorer resolution than PET as cur rently used clinically (as opposed to small-animal systems where the reverse is true). The need for collimation results in poor sensitivity compared to PET although the gap in sensiti vity is nar rowing with the newer technology now available. It should, of course, be remembered that w hat matters is the relati ve reconstr ucted resolution-noise trade-of f, not always directly predicted from the relative counts. Despite criticisms in the past that SPECT w as non-quantitati ve, careful attention to all ph ysical cor rections permits quantification to a high degree of accuracy.85,86 What is limited is the typicall y longer acquisition time often required for SPECT, so patient motion can be problematic and the minimum length of acquisition for dynamic studies is usuall y limited (although less prob lematic with some ne wer designs). Attenuation factors are smaller for SPECT , as is the radiation dose per unit acti vity (for pure gamma emitters). Table 2 summarizes the comparison of human dualhead SPECT and PET general proper ties. Note that the figures presented are representati ve of general use and do not necessarily reflect recent developments in instrumentation (e g, time-of-flight PET or high sensiti vity dedicated cardiac SPECT).

CONCLUSIONS SPECT continues to be widel y used to complement planar nuclear medicine studies, being the method of choice in an increasing number of applications (eg, heart,

brain). The ease of access to single-photon emitting radionuclides with good imaging proper ties and the wide range of suitab ly labeled radiophar maceuticals guarantees continued use. Continuing de velopments in instrumentation and reconstruction and the availability of combined SPECT/CT units suggest an increasing utility in both clinical practice and research.

REFERENCES 1. Cherry SR, Sorenson J A, Phelps ME. Ph ysics in nuclear medicine. USA: Philadelphia, Pennsylvania. Elsevier Health Sciences; 2003. 2. Wernick MN , Aarsvold JN , editors. Emission tomo graphy: the fundamentals of SPECT and PET . San Die go (CA): Else vier; 2004. p. 270–92. 3. Cherry SR. Multimodality in vi vo imaging systems: twice the po wer or doub le the troub le? [re view]. Annu Re v Biomed Eng 2006;8:35–62. 4. Anger HO. Scintillation camera. Rev Sci Instrum 1958;29:27–33. 5. Dolgoshein B, Balagura V, Buzhan P, et al. Status repor t on silicon photo-multiplier de velopment and its applications. Calice/SiPM Collaboration. Nucl Instr um Methods Ph ys Res A 2006; 563:368–76. 6. de Vree GA, Westra AH, v an der Ha ve F, et al. Photon counting gamma camera based on an electron-multipl ying CCD . IEEE Trans Nucl Sci 2005;52:580–8. 7. Beekman FJ , de Vree GA. Photon-counting v ersus an inte grating CCD-based gamma camera: impor tant consequences for spatial resolution. Phys Med Biol 2005;50:N109–19. 8. Meng LJ , Clinthor ne NH, Skinner S, et al. Design and feasibility study of a single photon emission microscope system for small animal I-125 imaging. IEEE Trans Nucl Sci 2006;53:1168–1178. 9. Nagarkar VV, Shestak ova I, Ga ysinskiy V, et al. A CCD-based detector for SPECT. IEEE Trans Nucl Sci 2006;53:54–8. 10. Miller BW, Bradford Barber H, Bar rett HH, et al. Single-photon spatial and ener gy resolution enhancement of a columnar

52

11.

12.

13.

14.

15.

16.

17.

18.

19.

20.

21.

22. 23.

24.

25.

26. 27. 28.

29. 30. 31.

32.

MOLECULAR IMAGING: PRINCIPLES AND PRACTICE

CsI(Tl)/EMCCD gamma-camera using maximum-lik elihood estimation. Proc SPIE, 6142, 61421T, 2006. Fiorini C, Longoni A, Perotti F, et al. A monolithic ar ray of silicon drift detectors coupled to a single scintillator for gammaray imaging with sub-millimeter position resolution. Nucl Instrum Methods Phys Res A 2003;512:265–71. Pichler BJ, Ziegler SI. Photodetectors. In: Wernick MN, Aarsvold JN, editors. Emission tomo graphy: the fundamentals of SPECT and PET. San Diego (CA): Elsevier; 2004. p. 270–92. van Loef EVD , Dorenbos P , v an Eijk CWE, et al. High-ener gyresolution scintillator: Ce3+ activated LaCl3. Appl Phys Lett 2000; 77:1467–8. van Loef EVD, Dorenbos P, van Eijk CWE, et al. High-ener gy-resolution scintillator . Ce 3+ activated LaBr 3. Appl Ph ys Lett 2002; 79:1573–5. Dorenbos P, de Haas JTM, v an Eijk CWE. Gamma ra y spectroscopy with a circle divide 19 × 19 mm(3) LaBr3 : 0.5% Ce3+ scintillator. IEEE Trans Nucl Sci 2004;51:1289–96. Moses WW, Shah KS. P otential for RbGd 2Br7:Ce, LaBr 3:Ce, LaBr3:Ce, and LuI 3:Ce in nuclear medical imaging. Nucl Instr um Methods Phys Res A 2005;537:317–20. Vavrik D, Jakubek J , Visschers J, et al. F irst tests of a Medipix 1 pixel detector for X-ra y dynamic defectoscop y. Nucl Instr um Methods Phys Res A 2002;487:216–23. Wagenaar DJ. CdTe and CdZnTe semiconductor detectors for nuclear medicine imaging. In: Wernick MN, Aarsvold JN, editors. Emission tomography: the fundamentals of SPECT and PET. San Diego (CA): Elsevier; 2004. p. 270–92. Huang Q, Zeng GL. An anal ytical algorithm for sk ew-slit imaging geometry with nonunifor m attenuation cor rection. Med Ph ys 2006;33:997–1004. Walrand S, Jamar F , de Jong M, P auwels S. Ev aluation of no vel whole-body high-resolution rodent SPECT (Lino view) based on direct acquisition of lino gram projections. J Nucl Med 2005; 46:1872–80. Rogers WL, Clinthorne NH, Shao L, et al. SPRINT II: a second generation single photon ring tomography. IEEE Trans Med Imaging 1988;7:291–7. Metzler SD, Accorsi R, Novak JR, et al. On-axis sensitivity and resolution of a slit-slat collimator. J Nucl Med 2006;47:1884–90. Rowe RK, Aarsvold JN, Barrett HH, et al. A stationary hemispherical SPECT imager for three-dimensional brain imaging. J Nucl Med 1993;34:474–80. Barrett HH, Swindell W. Radiological imaging. The theory of image formation, detection, and processing. New York: Academic Press; 1981. Tsui BMW, F rey EC. Analytical image reconstr uction methods in emission computed tomography. In: Zaidi H, editor . Quantitative analysis in nuclear medicine imaging. Ne w York: Springer; 2006. p. 82–106. Feldkamp LA, Da vis LC, Kress JW. Practical cone-beam algorithm. J Opt Soc Am A 1984;1:612–9. Lange K, Carson R. EM reconstr uction algorithms for emission and transmission tomography. J Comput Assist Tomogr 1984;8:306–16. Hudson HM, Larkin RS. Accelerated image reconstr uction using ordered subsets of projection data. IEEE Trans Med Imaging 1994;13:601–9. Hutton BF, Hudson HM, Beekman FJ . A clinical perspective of accelerated statistical reconstruction. Eur J Nucl Med 1997;24:797–808. Tsui B, Frey E, LaCroix K, et al. Quantitati ve myocardial perfusion SPECT. J Nucl Cardiol 1998;5:507–22. Staelens S, de Wit TC, Beekman FJ . Fast hybrid SPECT simulation including ef ficient septal penetration modeling (SP-PSF). Ph ys Med Biol 2007;52:3027–43. Frey EC, Tsui BMW. Collimator-detector response compensation in SPECT. In: Zaidi H, editor. Quantitative analysis in nuclear medicine imaging. New York: Springer; 2006. p. 141–66.

33. Hutton BF, Nuyts J , Zaidi H. Iterati ve reconstr uction methods. In: Zaidi H, editor . Quantitative analysis in nuclear medicine imaging. New York: Springer; 2006. p. 107–40. 34. Lalush DS, Wernick MN. Iterative image reconstruction. In: Wernick MN, Aarsvold JN, editors. Emission tomo graphy: the fundamentals of PET and SPECT. San Diego (CA): Academic Press; 2004. p. 443–72 35. Zaidi H, editor . Quantitati ve anal ysis in nuclear medicine imaging. New York: Springer; 2006. 36. Chang LT. A method for attenuation cor rection in radionuclide computed tomography. IEEE Trans Nucl Sci 1978;25:638–43. 37. Bailey DL. Transmission scanning in emission tomo graphy. Eur J Nucl Med 1998;25:774–87. 38. O’Connor MK, K emp B . A multicenter e valuation of commercial attenuation compensation techniques in cardiac SPECT using phantom models. J Nucl Cardiol 2002;9:361–76. 39. Beekman FJ , Kamphuis C, Hutton BF , v an Rijk PP . Half-f anbeam collimators combined with scanning point sources for simultaneous emission-transmission imaging. J Nucl Med 1998; 39:1996–2003. 40. Manglos SH, Bassano DA, Thomas FD, Grossman ZD. Imaging of the human torso using cone-beam transmission CT implemented on a rotating gamma camera. J Nucl Med 1992;33:150–156. 41. Lang TF, Hasegawa BH, Soo Chin L, et al. Description of a prototype emission-transmission computed tomo graphy imaging system. J Nucl Med 1992;33:1881–7. 42. Buvat I, Benali H, Todd-Pokropek A, Di Paola R. Scatter correction in scintigraphy; the state-of-the-art. Eur J Nucl Med 1994;21:675–94. 43. Ogawa K, Ichihara T, Kubo A. Accurate scatter correction in single photon emission CT. Ann Nucl Med Sci 1994;7:145–50. 44. Beekman FJ, de Jong HWAM, van Geloven S. Efficient fully 3D iterative SPECT reconstruction with Monte Carlo based scatter compensation. IEEE Trans Med Imaging 2002;21:867–77. 45. De Beenhouwer J, Staelens S, Vandenberghe S, Lemahieu I. Acceleration of GA TE SPECT simulations. Medical Ph ysics 2008; 35:1476–85. 46. Xiao J, de Wit C, Staelens S, Beekman FJ . Evaluation of 3D Monte Carlo-based scatter cor rection for Tc99m cardiac perfusion SPECT. J Nucl Med 2006;47:1662–9. 47. Hutton BF, Lau YH. Application of distance-dependent resolution compensation and post-reconstr uction f iltering for m yocardial SPECT. Phys Med Biol 1998;43:1679–93. 48. Tsui BMW, Frey EC, Zhao X, et al. The importance and implementation of accurate 3D compensation methods for quantitati ve SPECT. Phys Med Biol 1994;39:509–30. 49. Pretorius PH, Kingm MA, P anm T-S, et al. Reducing the influence of the partial volume effect on SPECT activity quantification with 3D modelling of spatial resolution in iterative reconstr uction. Ph ys Med Biol 1998;43:407–20. 50. Zeng GL, Gullberg GT, Bai C, et al. Iterati ve reconstruction of Fluorine-18 SPECT using geomotric point response correction. J Nucl Med 1998;39:124–30. 51. Rousset O, Ma Y, Evans A. Cor rection for par tial volume effects in PET: principle and validation. J Nucl Med 1998;39:904–11. 52. Muller-Gartner H, Links J, Prince J, et al. Measurement of radiotracer concentration in brain grey matter using positron emission tomography: MRI-based cor rection for par tial volume effects. J Cereb Blood Flow Metab 1992;12:571–83. 53. Tohka J, Reilhac A. Deconvolution-based partial volume cor rection in Raclopride-PET and Monte Carlo comparison to MR-based method. Neuroimage 2008;39:1570–84. 54. Rousset O, Zaidi H. Correction for partial volume effects in emission tomography. In: Zaidi H, editor . Quantitative analysis in nuclear medicine imaging. New York: Springer; 2006. p. 236–71. 55. Lopresti BJ, Russo A, Jones WF, et al. Implementation and performance of an optical motion tracking system for high resolution brain PET imaging. IEEE Trans Nucl Sci 1999;46:2059–67.

SPECT and SPECT/CT

56. Goldstein SR, Daube-Witherspoon M, Green MV, Eidsath A. A head motion measurement system suitab le for emission computed tomography. IEEE Trans Med Imaging 1997;16:17–27. 57. Fulton R, Hutton BF , Braun M, et al. Use of 3D reconstr uction to correct for patient motion in SPECT . Ph ys Med Biol 1994; 39:563–74. 58. Hutton BF, Kyme AZ, Lau YH, et al. A hybrid 3D reconstruction/registration algorithm for cor rection of head motion in emission tomography. IEEE Trans Nucl Sci 2002;49:188–94. 59. Kyme AZ, Hutton BF, Hatton RL, et al. Practical aspects of a datadriven motion cor rection approach for brain SPECT . IEEE Trans Med Imaging 2003;22:722–9. 60. Feng B , Gif ford HC, Beach RD , et al. Use of three-dimensional Gaussian interpolation in the projector/backprojector pair of iterative reconstr uction for compensation of kno wn rigid-body motion in SPECT. IEEE Trans Med Imaging 2006;25:838–44. 61. Lamare F, Ledesma Carbayo MJ, Reader AJ, et al. Respiratory motion correction in 4D PET/CT : comparison of implementation methodologies for incor poration of elastic transfor mations in the reconstruction system matrix. IEEE Nucl Sci Symp Conf Record 2006. pp. 2365–69. 62. McQuaid SJ, Hutton BF. Sources of attenuation-cor rection ar tefacts in cardiac PET/CT and SPECT/CT. Eur J Nucl Med Mol Imaging 2008;35:1117–23. 63. Tonge CM, Ellul G, Pandit M, et al. The value of registration correction in the attenuation cor rection of m yocardial SPECT studies using lo w resolution computed tomo graphy images. Nucl Med Commun 2006;27:843–52. 64. Beyer T, Townsend DW, Brun T, et al. A combined PET/CT scanner for clinical oncology. J Nucl Med 2000;41:1369–79. 65. Hamann M, Aldridge M, Dickson J , et al. Ev aluation of a lo wdose/slow-rotating SPECT -CT system. Ph ys Med Biol 2008; 53:2495–508. 66. Beekman FJ, Hutton BF. Multi-modality imaging on track [editorial]. Eur J Nucl Med Mol Imaging 2007;34:1410–4. 67. Eberl S, Kanno I, Fulton RR, et al. Automated interstudy image re gistration technique for SPECT and PET . J Nucl Med 1996; 37:137–45. 68. Bailey DL, Roach PJ, Bailey EA, et al. De velopment of a cost ef fective modular SPECT/CT scanner . Eur J Nucl Med Mol Imaging 2007;34:1415–26. 69. Kinahan PE, Townsend DW, Beyer T, Sashin D . Attenuation cor rection for a combined 3D PET/CT scanner . Med Ph ys 1998; 25:2046–53. 70. Blankespoor SC. Attenuation cor rection of SPECT using X-ray CT on an emission-transmission CT system: m yocardial perfusion assessment. IEEE Trans Nucl Sci 1996;43:2263–74.

53

71. Bockisch A, Be yer T, Antoch G. P ositron emission tomo graphy/ computed tomography–imaging protocols, ar tifacts, pitfalls. Mol Imaging Biol 2004;6:188–99. 72. Patton JA, Slomka PJ, Germano G, Ber man DS. Recent technolo gic advances in nuclear cardiology. J Nucl Cardiol 2007;14:501–13. 73. Chang W, Liang H, Liu J . Design concepts and potential performance of MarC-SPECT–a high-perfor mance cardiac SPECT system [abstract]. J Nucl Med 2006;47:P190–1. 74. Gambhir SS, Berman DS, Ziffer J, et al. A novel high sensitivity rapid acquisition single photon molecular imaging camera. J Nucl Med 2008. [in press 2009] 75. Beekman FJ, van der Have F. The pinhole: gateway to ultra-high resolution three-dimensional radionuclide imaging. Eur J Nucl Med Mol Imaging 2007;34:151–61. 76. Xu J, Liu C, Wang Y, et al. Quantitati ve rotating multise gment slanthole SPECT mammo graphy with attenuation and collimator reponse compensation. IEEE Trans Med Imaging 2007;26:906–16. 77. Kamphuis C, Beekman FJ. A feasibility study of offset one-beam collimators for combined emission transmission brain SPECT: a feasibility study. IEEE Trans Nucl Sci 1998;45:1250–4. 78. Beekman FJ. Apparatus for making tomo graphic images. Inter national number WO97/43667, US6324258, European patent EP1007989, and Australian patent AU730166B. 1997. 79. Li J, Jaszczak RJ, Van Mullekom A, et al. Half-cone beam collimation for triple-camera SPECT systems. J Nucl Med 1996;37:498–502. 80. Kacperski K, Hutton BF. Optimal parallel hole collimator for cardiac imaging with iterati ve reconstr uction and resolution reco very. Proc full y three-dimensional image reconstr uction in radiolo gy and nuclear medicine 2007;174–7. 81. Panin VY, Kehren F, Michel C, Casey M. Fully 3-D PET reconstr uction with system matrix derived from point source measurements. IEEE Trans Med Imaging 2006;25:907–21. 82. Rentmeester MCM, van der Have F, Beekman FJ. Optimizing multipinhole SPECT geometries using an anal ytical model. Phys Med Biol 2007;52:2567–81. 83. Rogulski MM, Barber HB , Bar rett HH, et al. Ultra-high-resolution brain SPECT imaging: simulation results. IEEE Trans Nucl Sci 1993;40:1123–9. 84. Goorden MC, Rentmeester MCM, Beekman FJ . Theoretical analysis of multi-pinhole brain SPECT. [Submitted] 85. Willowson K, Baile y DL, Baldock C. Quantitati ve SPECT reconstruction using CT -derived cor rections. Ph ys Med Biol 2008; 53:3099–112. 86. Iida H, Eberl S, Kim K-M, et al. Absolute quantitation of myocardial blood flow with 201Tl abd dynamic SPECT in canine: optimisation and validation of kinetic modelling. Eur J Nucl Med Mol Imaging 2008;35:896–905.

5 PRINCIPLES OF MICRO X-RAY COMPUTED TOMOGRAPHY SHAUN S. GLEASON, PHD, MICHAEL J. PAULUS, PHD, AND DUSTIN OSBORNE, PHD

Like its clinical counterpart, X-ray computed tomography (CT), high-resolution X-ray micro-computed tomo graphy (micro-CT) is a widel y used modality for imaging anatomy in li ving specimens. In this chapter , w e re view the basic physics of micro-CT systems designed for highresolution studies of laboratory animals, the mathematical principles used to de velop reconstr ucted images, the k ey factors that deter mine image quality , and some of the commonly used applications for this technolo gy. Anatomic infor mation pro vided with micro-CT technology is v aluable in molecular imaging applications in at least two specif ic areas: (1) the anatomy provides a physical context or “map” that shows where in the body molecular events are taking place and (2) there are molecular events that ha ve a direct impact on anatomic str uctures that can be imaged using micro-CT.

film cassette to acquire projection images. 3 The X-ra y film was subsequently processed and digitized , providing data sets with sufficient resolution (~150 microns) to reconstruct images of small-animal or gans. By 1984, high-resolution X-ray detector technology had improved, and Burstein and colleagues 4 reported an ~50 mm resolution image of a mouse thorax obtained using a 90 kilovolt potential (kVp) X-ra y source and a 512-element linear ar ray of X-ray detectors. During this period, con ventional CT systems w ere also used to simultaneously image multiple mice 5 with relati vely low resolution (> 800 mm) but v ery high throughput (eight mice at a time, 9.6 seconds per image). In 1987, Flanner y and colleagues 6 brought X-ra y microtomography into a new era with the introduction of

BACKGROUND The typical conf iguration of earl y small-animal computed tomography (CT) systems is shown schematically in Figure 1, where the subject to be imaged is placed on a rotating stage betw een an X-ra y source and a tw odimensional (2-D) X-ra y detector , and typicall y hundreds of 2-D projection vie ws are acquired as the subject rotates. A three-dimensional (3-D) tomographic image v olume is then reconstr ucted using a computer algorithm. Although the principles of X-ra y CT ha ve been understood since the early work of Nobel Laureates Cormack1 and Hounsf ield,2 the de velopment of useful micro-CT systems required that se veral technolo gical advances take place f irst. In the earl y 1980s, the a vailable electronic X-ra y detectors did not ha ve suf ficient spatial resolution to generate useful images of rodents, so some of the early developers used a translating X-ray

54

projection subject on rotating stage x-ray source Figure 1. Schematic diagram of micro-CT system with rotating specimen. The X-ray source generates an X-ray beam that passes through a subject mounted to a rotating stage, and the radiograph, or projection, is captured on the opposite side by an X-ray detector.

Principles of Micro X-ray Computed Tomography

a 3-D imaging system using a 2-D detector consisting of a phosphor plate opticall y coupled to a charge-coupled detector (CCD) ar ray. To acquire a large number of X-ray photons in each micropix el (~2.5 µm × 2.5 µm), these in vestigators used a synchrotron X-ra y source beam line in place of the con ventional X-ray tube. During this time, the F ord Motor Company Research Laboratories also de veloped a 3-D microtomo graphy system for industrial applications using a laborator y X-ray tube for the source and an image intensifier screen coupled to a video readout. As a part of this effort, the scanner was used to study the subchondral (ie, directly under the cartilage) bone architecture in guinea pigs with osteoarthritis,7 human cancellous (ie, spongy , porous) bone, 8 and trabecular (ie, lattice, or f ine matrix) bone str ucture.9 A fundamental contribution of the F ord g roup w as the development of a ne w 3-D “cone-beam” image reconstruction algorithm, that is, the F eldkamp algorithm, which remains one of the most widel y used v olumetric reconstruction algorithms. 10 Rather than treating the X-ray beam as a collection of parallel “f ans,” the F eldkamp algorithm takes into account the diverging, conical nature of the X-ra y beam in its geometric model and is therefore a more geometricall y accurate reconstr uction technique. In the 1990s, a number of groups11–25 developed microtomography systems for high-resolution specimen analysis. Most of these systems used CCD-based detector ar rays, micro-focus X-ra y tubes, and had reconstructed image resolutions betw een 20 and 100 microns. The majority of the studies perfor med using these instr uments focused on high-density tissue,

x-ray source

55

such as bone or teeth, for w hich magnetic resonance imaging (MRI) is less successful. There was signif icant work in the 1990s on mouse genotyping, and se veral genotypes of interest resulted in phenotypes that include bone mark ers, hence micro-CT w as considered a v aluable tool in this effort. For in vi vo small-animal studies, par ticularly lar ge population studies, the scanner conf iguration sho wn in Figure 1 can be cumbersome because the subject must be confined in a rotating carrier designed to prevent soft-tissue or gan motion. Most commerciall y a vailable microCT systems designed for li ve animal studies use the configuration shown in Figure 2, where the X-ray source and detector rotate about a fixed animal pallet. These systems ha ve become highl y sophisticated; using detector elements with up to 16 million pixels, X-ray sources with focal spots less than 10 microns, and the ability to scan a whole mouse in less than 1 minute.

BASIC PHYSICS At i ts s implest, a m icro-CT s ystem c onsists o f a n X-ray source and an X-ra y detector that generate, as the y rotate about a specimen, 2-D projection images.These key components determine the characteristics of the acquired image data and are briefly described here.

X-ray Source The X-ray source typically used in micro-CT scanners (F igure 3) consists of an e xternal high v oltage supply, a f ilament, electron optics, and an anode encapsulated in an evacuated glass envelope. The filament produces a cloud of electrons when heated by an electric current. The electrons are accelerated away from the filament by a potential difference of typically 20 to 150 kV X-rays Cathode (2 )

Be window Anode (1 ) (metal target)

Subject Electrons

2

x-ray detector array Figure 2. Schematic diagram of micro-CT system with X-ray source and detector rotating about a stationary subject.

1

Figure 3. Schematic diagram of an X-ray source. A large voltage potential is placed between the anode and cathode. The cathode then emits a beam of electrons that travel and strike the metal target (anode). When they strike the anode, an X-ray beam is generated and travels out of the housing through a Be window .

56

MOLECULAR IMAGING: PRINCIPLES AND PRACTICE

and focused b y the electron optics to a f ine (5 to 50 micron) focal spot on the anode. The electrons decelerate and stop in the anode material (typically tungsten), converting their kinetic ener gy to heat and X-ra ys. Typically, less than 1% of the electron kinetic ener gy is converted to X-rays. The X-rays then typically travel out of the housing through a beryllium window. Beryllium is a metal that is both highl y transmissi ve to X-ra ys and strong enough to pro vide a ph ysical bar rier between the vacuum inside the X-ray source and the external environment. An aluminum filter is normally placed over the exit window to remo ve lo w-energy X-ra ys. The remo val of low-energy X-ray is called “hardening” the X-ra y beam. This is important because most of the low-energy X-rays are absorbed in the body causing additional dose to the subject, and the y can produce streaking ar tifacts around X-ray opaque objects, such as bones. The characteristic energy profile of the emitted X-ray flux is shown in Figure 4. The maximum X-ray energy in the spectrum is given by the expression as follows:

Emax = q × V ,

(1)

where V is the X-ray source operating voltage and q is the fundamental electron char ge. Thus, for e xample, a tube biased with 100 kVp will produce X-ra ys with a maximum ener gy of 100 k eV. The low-energy shape of the

spectrum is deter mined b y the thickness of the aluminum f ilter, with thick er f ilters remo ving more of the lo w-energy spectr um. The characteristic ener gy peaks are associated with the anode material; tungsten anode X-ray tubes produce characteristic energy peaks at 59.3 keV and 67.2 k eV. Note that the a verage energy of the spectrum is typically 30 to 40% of the peak energy.26 The main disadv antage associated with a broad ener gy, polychromatic X-ra y source is that lo wer ener gy (soft) X-rays are preferentially absorbed in the subject. This preferential absor ption adds nonuseful dose to the subject and creates an erroneous relationship between the thickness of a material and the ef fective attenuation coefficient. The result of this is a “cupping” or “beamhardening” artifact in the reconstructed image where the calculated attenuation coef ficient is lo wer near the center of the subject. To overcome these problems, a purely monoenergetic X-ray beam can be generated b y a synchrotron (par ticle accelerator), w hich is a lar ge, expensive, and generally inaccessible device typically used in highly specialized research f acilities. Approximately, monoenergetic beams can be generated using a more standard polychromatic X-ray source and then f iltering the beam to remo ve X-ra ys at undesired ener gies. This will help minimize an y beam-hardening ar tifacts. The disadvantage of using a highl y f iltered beam is the

Simulated X-ray Spectra (Tungsten Anode) Constant Current 5 1 mA

2.00E 1 05 1.80E 1 05

Relative Counts per second

1.60E 1 05 1.40E 1 05 1.20E 1 05

80 kVp 110 kVp

1.00E 1 05

140 kVp 170 kVp (ext) 200 kVp (ext)

8.00E 1 04 6.00E 1 04 4.00E 1 04 2.00E 1 04 0.00E 1 00 0

Figure 4.

20

40

60

80

100 120 Bias Voltage (kVp)

140

160

180

Calculated spectra of a tungsten anode X-ray source as a function of operating voltage.

200

Principles of Micro X-ray Computed Tomography

57

amount of remaining X-ra y flux will be relati vely low, and this will result in long scan times and/or noisy images due to a low system signal-to-noise ratio.

designed using micro-focus X-ra y sources with focal spot sizes ranging from a fe w microns up to ~50 µm to minimize the geometric unsharpness.

Geometric Unsharpness

Power Limitations

An impor tant perfor mance specif ication of the X-ra y source is the focal spot size A. As shown in Figure 5, when X-rays emitted from a focal spot are used to cast an image of an object located at a distance D1 from the focal spot onto an image plane located at a distance D1 + D2 from the focal spot, the resultant image has a magnification m given by the expression:

As pre viously noted , the v ast majority of the ener gy deposited by accelerated electrons onto the X-ra y source anode is dissipated as heat. The maximum power that may be applied to a micro-focus X-ra y source is, therefore, limited by the rate at w hich heat can be removed from the X-ray source target. Flynn and colleagues 25 noted that for stationary tar gets with small focal spot sizes, the heat dissipation is predominantly radial and approximately proportional to the focal spot diameter. They observed that the maximum power for a stationary-target micro-focus X-ray source appro ximately follo ws the empirical relationship expressed as:

( D1 + D2 ) m= . D1

(2)

The blur B associated with the focal spot size is gi ven by the expression:

A D2 B= = A( m − 1). D1

(3)

Note that the b lur in Equation 3 is a measure of the unsharpness in the image plane. It is typical practice to normalize the b lur b y the magnif ication w hen def ining the geometric unsharpness Ug to relate the image error to the geometry of the object being imaged:

U g = A (1 − 1 / m).

(4)

From Equation 4, it is e vident that an unmagnif ied image (ie, the limit in which an object of zero thickness is imaged while in contact with the image plane, and m = 1) will have a geometric unshar pness of 0, w hile a highl y magnif ied image will have a geometric unshar pness approaching the focal spot size. High-resolution micro-CT scanners are

Focal Spot

A

D1

OBJECT PLANE

D2

B

B

IMAGE PLANE

Figure 5. Schematic diagram showing blur (B) due to the finite size of the X-ray source focal spot (A). D1 is the distance from the focal spot to the object and D2 is the distance from the object to the image plane.

Pmax ≈ 1.4 ( A)0.88 ,

(5)

where Pmax is the maximum X-ra y tube po wer in Watts and A is the focal spot size in microns. A survey of specifications from a number of X-ra y source suppliers shows Flynn’s relationship to hold for most commerciall y available X-ra y tubes. F or e xample, a 5- µm focal spot source will typically have a maximum po wer rating of 6, while a 50-µm focal spot source will typically have a maximum power rating of 45 W. Because the total X-ra y flux emitted by an X-ray source is a function of applied external voltage V and the anode cur rent I, this power ceiling (Power = I·V) imposes an upper limit on the available X-ray flux, w hich in tur n imposes a lo wer limit on scan times due to the f act that the lo wer the X-ra y flux in the beam, the longer the X-ra y detector must dwell to collect enough signal to then generate lo w-noise image reconstructions. Most commerciall y available micro-CT scanners have minimum scan times on the order of 1 minute. In clinical CT scanners, this power limitation is alleviated by using rotating anode sources and by pulsing the X-ray source. Rotating anode sources replace the f ixed anode with a disk rotating at speeds in e xcess of 3000 rpm, thereby spreading the heat over a much larger area because the area of the tar get e xposed to the electron beam is al ways changing, allo wing better dissipation of heat a way from the tar get. Pulsed sources limit heat production to only the short periods of time during which image data is being acquired to minimize the heat load. Neither of these approaches has found wide use in commercial micro-CT scanners to date. Even minor wobble in a rotating anode can broaden the focal spot size,

58

MOLECULAR IMAGING: PRINCIPLES AND PRACTICE

thereby increasing the geometric unshar pness, and although pulsed micro-focus X-ra y sources are under development, in some cases using recently developed carbon nanotube emitters, 27–29 these de vices ha ve also not yet been widely adopted by commercial micro-CT manufacturers due to lack of general a vailability and some uncertainty about their long-term stability and reliability.

Optimal X-ray Energy For a Poisson-statistics limited detection system, in which a f inite number of X-ra ys are emitted b y the X-ray source, an optimal X-ra y ener gy e xists for best contrast resolution in CT studies.30 Micro-CT data sets are typically Poisson-statistics limited due to the limited X-ray flux emitted b y the X-ra y source and the small detector element sizes. Following the work of Grodzins,30 if a unifor m cylinder of attenuating material is scanned , the contrast resolution limit in the reconstr ucted CT image may be expressed as:

noise = signal

2 D exp(µ D )

)

3

N (Δ x µ

2

,

(6)

where noise is the statistical variation in the reconstructed image, signal is the cor rect v alue for the reconstr ucted image, D is the diameter of the c ylinder, µ is the energydependent X-ra y attenuation coef ficient of the c ylinder material (discussed in greater detail below), N is the number of photons emitted b y the X-ra y source during the study, and Δx is the detector element spacing. At low energies, where µ is large, the contrast resolution is limited by the small number of X-rays penetrating the subject (ie, no X-rays reach the detector). At higher energies, where µ is small, the contrast resolution is limited by the small number of X-rays absorbed in the subject (ie, all of the X-rays pass through the subject and reach the detector). Equation 6 reaches an optimal (minimum) v alue at the ener gy for which µ = 2/D. Flanner y and colleagues 6 have repor ted similar conclusions. Assuming an animal’s gross attenuation characteristics are appro ximately that of w ater, for a 3-cm (mou se-sized) phantom the o ptimal ener gy is approximately 25 keV and for a 5-cm (rat-sized) phantom the optimum energy is approximately 30 keV.

Beam Hardening

energy dependent, especiall y at the lo w X-ra y ener gies preferred for small-animal studies. As the X-ra y beam passes through the subject, the lo wer-energy X-rays are preferentially absorbed near the surf ace, causing the image to be artificially brighter near the edges of the subject. This is the w ell-documented beam-hardening ar tifact.26,31–33 Prefiltration of the X-ra y beam to reduce the lower energy, or “soft,” X-rays can reduce the ar tifact by making the beam more monochromatic, and a number of algorithms ha ve been de veloped to par tially cor rect for beam hardening.31,32 Nonetheless, the effect is difficult to eliminate completely.

X-ray Detector Most commercially available preclinical micro-CT scanners use lar ge area (> 50 cm 2) 2-D detectors with geometries that result in relati vely lar ge ( ≥ 5°) X-ray cone-beam angles. The choice of the cone-beam architecture (2-D detector array) over the fan-beam architecture (approximately 1-D detector array) is preferred in clinical systems because it is primaril y driven by a need to more efficiently collect the low X-ray flux produced by micro-focus X-ray sources. The most widely used detector design employs a CCD detector coupled via f iber optic taper to a phosphor screen (Figure 6). In this design, light produced when a gadolinium oxysulphide (GOS) or thallium-doped cesium iodide (CsI:Tl) phosphor screen on the f ace of the detector is focused onto a smaller CCD detector through the fiber-optic taper. This design takes advantage of the maturity of CCD technolo gy to produce detectors with lo w electronic noise, lar ge (> 107) pixel counts, and e xcellent stability. The method is limited , however, by the cost and size limitations of the fiber-optic taper, by light loss in tapers with lar ge minif ication f actors, and b y a need to correct for distortions induced by the f iber-optic taper.

X-ray

CCD Fiber Optic Taper

The pol ychromatic X-ra y spectr um leads to a second important c onsideration, b eam h ardening. As n oted above, the X-ra y attenuation coef ficient is strongl y

Phosphor Screen

Figure 6. Schematic diagram of an X-ray detector consisting of a charge-coupled device (CCD) coupled via fiber-optic taper to a phosphor screen.

Principles of Micro X-ray Computed Tomography

Other commercial systems use detectors consisting of 2-D complementar y metal-o xide-semiconductor (CMOS) sensor ar rays coupled directl y to X-ra y scintillating materials that con vert X-ra ys into visib le light. CMOS sensor technolo gy has the adv antage of suppor ting the f abrication of sensors with lar ger surf ace areas than CCD sensors. Ho wever, because of the relati ve immaturity of the technolo gy, CMOS detectors tend to have smaller (~10 6) pixel counts, more pixel defects, and poorer signal-to-noise ratios. These shortcomings are balanced, however, by the elimination of the need for a fiberoptic taper. Without a fiber-optic taper, these detectors tend to be less expensive, more compact, and do not require geometric distortion correction. For both detector systems, the spatial resolution is determined b y the characteristics of the scintillating layer and the size of the pix el elements. When used to image a standard line-pair phantom for resolution characterization, widel y used scintillators pro vide resolutions betw een 10 and 20 line pairs per millimeter (25–50 micron resolution). Detector element sizes range from 10 to 50 microns. Together these parameters set the typical intrinsic resolution of widel y used X-ra y detectors between 25 and 70 microns at the f ace of the detector. The resolution of the acquired images can be impro ved significantly through magnif ication. Recall the e xpression for the scanner magnif ication f actor (Equation 1). The unsharpness related to the detector resolution depends linearly on the magnif ication factor,

U d = U D / m,

(7)

where Ud is the magnified unsharpness due to the detector in the image plane and UD is the intrinsic detector unsharpness at the face of the detector. The overall unsharpness U in the image is given by the expression:

U = (U g2 + U d2 )1/ 2 ,

(8)

U = U D [1 / m 2 + (1 − 1 / m)2 A2 / U D2 ]1/ 2 .

(9)

or

Evaluation of Equation 9 shows that at low magnification (m ≈ 1) the image unshar pness is equal to the intrinsic detector unshar pness, w hile at high magnif ication the image unshar pness is appro ximately gi ven b y the focal spot size. The use of magnif ication to impro ve image resolution comes with a linear penalty in f ield-of-view (FOV):

FOV = Detector Size / m.

(10)

59

Thus, the choice of magnification factor is always a compromise between the desired resolution and the maximum size of the animal to be imaged. Frequently, the X-ray source and detector are placed on computer -controlled movable stages to allow the operator to select the magnification factor and FOV on a scan-by-scan basis.

MICRO-CT RECONSTRUCTION Commercially a vailable micro-CT systems typicall y acquire data using a step-and-shoot technique that results in a collection of 2-D X-ra y projection images, each acquired at a dif ferent rotation angle relati ve to the subject. These individual projections contain only 2-D information, but because the y are acquired tomo graphically (radially around the subject), the y can be processed to create a 3-D image volume of the subject. The process of converting the 2-D tomo graphic projections into a 3-D volume is termed image reconstruction. This section presents a summary of the most commonly used reconstruction techniques for micro-CT and also introduces the typical computational platfor ms on w hich micro-CT reconstruction algorithms are implemented. For the purposes of this discussion, it is assumed that a third-generation CT scanner is used , which means that a point-source X-ray source is used. Fur thermore, a 2-D X-ray detector geometr y is assumed , w hich is the most common conf iguration used in micro-CT systems, and this results in a “cone-beam” X-ray geometry. A cylindrical volume is typicall y reconstr ucted, where the axis of the cylinder is aligned with the CT system’s axis of rotation. The diameter of the cylinder is equal to the transaxial dimension of the X-ra y detector di vided b y the CT system’s magnification factor. The length of the c ylinder depends on the type of trajector y, which is described in more detail below. Before reconstruction, the projection data are corrected for backg round or “dark” signal and are nor malized by a “bright-field” projection or a projection image acquired without any attenuating material between the X-ray source and the detector . The dark-f ield projection characterizes charge accumulation in the detector typically due to thermal phenomena unrelated to the image acquisition, and the bright-field projection characterizes the nonuniform system response due to variations in detector sensitivity and X-ray source flux density . F or a single 2-D projection p in a tomographic data set, the nor malization follo ws the expression:

( p − pdark ) pnorm = , ( pbright − pdark )

(11)

60

MOLECULAR IMAGING: PRINCIPLES AND PRACTICE

where pdark is the dark-f ield reference projection, pbright is the bright-field reference projection, and pnorm is the resultant nor malized projection. Each nor malized projection has pixel values ranging from 1.0 down to nearly 0, where values of 1.0 represent re gions in w hich the X-ray flux is un-attenuated (ie, no material in the X-ra y path), and v alues approaching 0 represent regions in which the majority of the X-ray flux is absorbed. Each normalized 2-D projection is a measure of the attenuation of the X-ra y flux along the lines of response between the X-ra y focal spot and each pix el in the projection image. F or a gi ven line of response j in the normalized projection pnorm, the X-ra y flux attenuation is described by the Beer–Lambert Law: − µ ( l ) dl pnorm , j = e ∫ ,

(12)

where µ(l ) is the spatially varying attenuation coefficient of the imaged subject along the path l of the line of response. The goal of any X-ray CT reconstruction algorithm is to use the set of projections acquired at multiple angles to calculate µ for each voxel in the image volume. There are a v ariety of techniques that can be used to reconstruct the nor malized projection data. The optimal image reconstruction approach will v ary depending on the geometry of the CT scan and the application at hand. In general, the categories of reconstruction algorithms are characterized by the trajectory used during data acquisition and the computational methods used to generate the images. The sections below f irst outline the dif ferent acquisition trajectories (ie, path of X-ray source/detector around subject) currently in use or under investigation followed by a description of the two general classes of reconstr uction algorithms.

A circular orbit is the simplest of all the possib le orbits, and the resulting data set are also the most straightforward to reconstruct. The disadvantage of the circular orbit with a cone-beam CT system is that the projections suffer from incomplete data, causing errors in the resulting cylindrical reconstruction. Assuming a 2-D square X-ray detector, the theoretical volume of support that can be reconstructed correctly with a circular orbit is a sphere. When data from a circular orbit scan are used to reconstr uct a cylinder, as is typical practice, the image data that lies within the cylinder but outside of the “ideal sphere” are subject to error. The e xtent of the er ror is dependent on the cone angle of the system (ie, the angle of the X-ray axial direction of the CT scanner) and the type of algorithm used.

Helical Orbit To generate a complete or suf ficient set of data in a cone-beam X-ra y CT system, Tuy33 has sho wn that every plane w hich passes through the imaging FO V must also cut through the orbit of the focal point (ie, the center of the 2-D detector) at least once. Although a circular orbit does not satisfy this condition, a helical orbit can meet this requirement. In a helical orbit scan, data are acquired w hen the subject is translated axiall y, that is, in parallel with the CT system’ s axis of rotation, whereas the X-ray source and the detector rotate around the subject (F igure 7). In addition to meeting Tuy’s requirement for data suf ficiency, a helical orbit can reduce ring ar tifacts frequently found in images recon-

Acquisition Trajectories The geometry of the micro-CT scanner hardware and the flexibility of the motion control system (hardw are and software) will def ine the type of trajector y that can be used during the scan. All micro-CT systems must include an ability to rotate the X-ra y source and detector relati ve to the subject. If the motion hardw are and softw are also allow the axial position of the subject to be changed during the micro-CT scan, then the operator has more flexibility in ter ms of def ining the trajector y of the data acquisition. Three trajectories are described below.

Circular Orbit If the X-ray source and detector rotate about the subject and the subject remains axiall y stationary throughout the scan, the data set are acquired with a circular orbit.

Figure 7. The blue arc defines the helical orbit of the X-ray source/detector about the subject. The red line defines the axis of rotation of the computed tomography (CT) scanner. The helical orbit is achieved by spinning the source/detector while moving the animal parallel to the axis of rotation.

Principles of Micro X-ray Computed Tomography

structed from a circular orbit scan. A number of researchers have recently described various methods for reconstructing helical CT data. 34–38

Nontraditional Orbits One alternative approach to acquire a theoreticall y complete data set via helical scan is to acquire data in a circle-plus-line orbit. This trajectory starts with the standard circular orbit described previously, acquiring a full set of X-ray projections as the source and the detector rotate 360 degrees about the subject. Subsequentl y, another set of projections are acquired by holding the source and the detector stationar y (no rotation) w hile translating the subject axially, taking multiple projections along this line (Figure 8). Examples of reconstructions using this type of trajectory are given by Noo and colleagues.39 A second nontraditional trajectory is accomplished when the X-ray source tra vels along a saddle trajector y relati ve to the subject. Although this is less practicall y implemented , and therefore less commonl y used than the pre viously described trajectories, it does have advantages for axially truncated projection data, w hich has been described in detail by several researchers.40–42

Classes of Algorithms Two discrete classes of reconstr uction algorithms are widely used to generate micro-CT images. Analytic algorithms calculate the image data directl y using defined single-pass mathematical methods. These

61

algorithms have the advantage of relative simplicity and speed but tend to produce noisier reconstr ucted images. In contrast, iterati ve reconstr uction algorithms use complex models of the acquisition system, repeatedl y estimate solutions, and compare the estimate with a modeled “ideal” solution until predefined convergence criteria are met. Iterati ve algorithms typicall y produce superior images but can take much longer to execute.

Analytic Reconstruction Algorithms Analytic algorithms, such as those based on f iltered backprojection (FBP), 43 model the tomo graphic data acquisition mathematically via the Radon transform. These models are based on theoreticall y ideal and continuous acquisition conditions. Of course, these assumptions are subject to error: the data acquired are discrete and are collected using nonideal scanners. Because analytic algorithms are mathematically based, they are relatively easy to implement and also have relatively short reconstruction times. One of the main disadvantages of the analytic algorithms is that they do not account for noise in the acquisition system and data. This noise results in lo w image quality f or a pplications t hat h ave l ow s ignal-to-noise ratios, such as nuclear medicine studies (eg, positron emission tomography [PET] and single photon emission computed tomography [SPECT]). Fortunately, X-ray micro-CT scanners can achie ve high signal-to-noise ratios because the X-ra y sources used can generate lar ge numbers of X-ray photons. It is for this reason that analytic algorithms based on FBP have been frequently implemented to reconstruct data from a micro-CT system.Two categories of analytic algorithms ha ve been implemented for micro-CT reconstruction: approximate reconstruction algorithms and exact reconstruction algorithms. Approximate Reconstruction Algorithms

Figure 8. Circle-plus-line orbit. The black circle defines a circular orbit of the X-ray source/detector, and the black line represents movement of the source/detector along the axis of rotation along which additional projections are acquired. The red line defines the axis of rotation.

Approximate algorithms make mathematical assumptions regarding the data acquisition so that the reconstr uction can be implemented easil y and/or in a computationall y efficient manner. The widely used F eldkamp10 algorithm falls into the category of approximate analytic algorithms. For a micro-CT scanner that uses a circular orbit, the Feldkamp algorithm generates a geometricall y cor rect reconstruction only for the center slice of the c ylindrical volume. The geometric accurac y of the reconstr uction diminishes for slices that are not in the center , and the geometric ar tifacts increase as the slices mo ve axiall y away from the center for the FO V. So, for lar ge cone angles, the Feldkamp algorithm suffers from increasingly

62

MOLECULAR IMAGING: PRINCIPLES AND PRACTICE

severe geometric distor tions (F igure 9). F or small cone angles (e g, less than 8 de grees), the ar tifacts are small enough to be ne gligible for most CT applications. The advantages of the F eldkamp algorithm include ease of implementation and f ast reconstr uction. Development of methods to impro ve the speed and FO V coverage of the Feldkamp algorithm is a focus of ongoing research. Recently, progress has also been made in generalizing the Feldkamp algorithm to suppor t cone-beam helical orbits, further increasing its usefulness. 35 The speed and ease of implementation adv antages typicall y f ar outw eigh the problems, and therefore the F eldkamp algorithm is almost universally included in a micro-CT reconstr uction tool set. Another recentl y de veloped g roup of appro ximate analytic reconstr uction algorithms that is potentiall y useful for alter native trajectories with possib ly limited views is the so-called “chord-based” class of algorithms. This class of algorithms has been demonstrated to be useful w hen scanning constraints (e g, size of subject, limitation of motion control system) limit the scanner’ s ability to perform a standard trajectory. Also, this class is useful for re gion-of-interest reconstr uctions w here an entire standard trajectory is not needed to co ver a particular region within the body of the subject. A treatment of chord-based algorithms and their noise characteristics is presented by Xia and colleagues. 44 Exact Reconstruction Algorithms

Mathematicians ha ve de voted a considerab le deal of effort in the quest to de velop e xact cone-beam reconstruction algorithms. Although the resultant methods are theoretically valuable and elegant, they have thus f ar suffered from the implicit assumption of ideal conditions for data acquisition. The assumptions of ideal data are

Figure 9. Disk phantom (left) and Feldkamp reconstruction from circular orbit (right). Although the phantom is a stack of parallel discs, the reconstructed image shows a distortion of the disc shapes (nonparallel lines) that increases toward the edges of the cone-beam (top and bottom of image). Images courtesy of QRM website .

of course in valid, and hence, substantial ar tifacts are common w hen reconstr ucting real-w orld data with theoretically exact algorithms.45 The two most commonly referenced exact algorithms were originally developed by Katsevich46 and Grangeat. 47 Exact cone-beam reconstruction remains an area of acti ve research, and ne w variations of both of these algorithms ha ve been recently developed.48,49

Iterative Reconstruction Algorithms An alter native to anal ytic reconstr uction is the use of algorithms based on iterati ve techniques. Iterati ve algorithms take into account the discrete, or digital, nature of the data and generall y represent the data as v ectors and the acquisition process as a matrix. The typical flo w of an iterative reconstruction algorithm is to (1) generate an estimated image solution, (2) forw ard project the solution to generate a calculated set of “pseudo-projections,” (3) compare the “pseudo-projections” with the acquired projections, (4) adjust the estimated image solution based upon this comparison, and (5) repeat the process until a predef ined convergence criteria is met. The most pressing challenge in iterati ve reconstr uction of microCT data lies with numerical dif ficulties sur rounding matrix in version of lar ge data sets and with the time required to complete the iterations. Generally speaking, iterative algorithms are far more computationally intensi ve than anal ytic algorithms and require substantiall y more computing resources. F ortunately, with the continued increase in performance and decreasing cost of com puters, iterati ve techniques are becoming more practical and have recently become standard in clinical nuclear medicine, w here the reconstructed v olumes are relati vely small. Micro-CT image volumes are much larger than clinical nuclear medicine image v olumes because of the relati vely high spatial resolution (ie, larger number of voxels per unit volume), however, and for this application statistical iterative algorithms remain a computational challenge. For example, a typical micro-CT detector might have an array of 512 by 512 imaging elements, or pix els, and a typical micro-CT scan might consist of 360 projections o ver 360 de grees. The probability system matrix that is required for a statistical reconstr uction technique, such as maximum likelihood e xpectation maximization (ML-EM), 50 for a reconstructed image volume of 512 × 512 × 512 would be hundreds of terab ytes (w here 1 terab yte = 1 trillion bytes). Such an o verwhelming prob lem can be made more manageable by taking advantage of geometric symmetries, only loading subsections of the system matrix,

Principles of Micro X-ray Computed Tomography

and di viding up the prob lem across man y computing nodes,51 but the challenge remains daunting. One area of study where iterative reconstruction methods may offer sufficient advantage to justify the use of an iterative algorithm for micro-CT would be in cases where a very low X-ray dose is required. Lo w-dose studies yield data sets with poor statistics, and anal ytic reconstructions of statistically poor data typically have significant artifacts. Iterative statistical algorithms ha ve been sho wn to of fer a clear advantage in producing quantitatively accurate image data from data with poor statistics. 52

Computational Platforms X-ray micro-CT data sets are generall y considered to be large relati ve to clinical applications. Some commercially a vailable micro-CT systems use detectors with up to 16 me gapixels, w hich can produce reconstr ucted image volumes greater than 64 gigavoxels in size. Ev en standard micro-CT systems typically produce reconstructed image v olumes on the order of hundreds of megavoxels in size. These lar ge v olumes are computationally challenging, e ven for f ast anal ytic techniques, such as the Feldkamp algorithm. With no optimization, it can take several hours to reconstruct a typical micro-CT data set on a standard laborator y computer using the Feldkamp algorithm. F or this reason, a tremendous amount of research has been perfor med to accelerate micro-CT reconstruction.

Cluster-Based Reconstruction Platforms As computer costs ha ve decreased, it has become af fordable to assemb le a lar ge number of netw orked computers to share the reconstruction load. These collections of computers are commonly called clusters and are useful to speed up any computing tasks that are highly parallelizable. Clusters are most commonly configured with the LINUX operating system, but Windows™-based clusters are becoming more pre valent as w ell. P arallel implementations of CT reconstruction algorithms for these platfor ms are becoming common because of the highly parallelizable nature of the CT reconstr uction problem. An example of a message passing interf ace (MPI)-based cluster implementation of the Feldkamp algorithm with intelligent focus-of-attention FOV support is presented in the study b y Gregor and colleagues.53 The advantages of clusters are that they are relatively inexpensive, easily upgradeable, and are fle xible in that they can be used for a multitude of computationall y challenging tasks. The disadvantages of clusters are their size and relatively large power requirements.

63

Graphics Processing Units- and Field Programmable Gate Arrays-Based Reconstruction Platforms Specialized hardw are can also be used to accelerate reconstruction algorithms for micro-CT. Again, due to its highly parallelizable nature, the F eldkamp algorithm has been implemented on many specialized hardware platforms from array processors, to f ield programmable gate arrays (FPGAs) 54 and graphics processing units (GPUs). Due to their low cost, high availability, and applicability to CT (and other modality) reconstr uction, GPUs ha ve been a tar get platform for a large number of researchers and companies that are de veloping accelerated reconstruction technology55–57 GPUs, which are typically used for 2-D and 3-D graphics rendering applications, are also well-suited to implement the forward- and back-projectors used in image reconstr uction. A forward-projector takes the cur rent representation of the 3-D v olume and projects it onto the 2-D projection space, and a back-projector does the opposite—it tak es a 2-D projection and back-projects (“smears”) it back across the 3-D v olume. Finally, m any r econstruction e xperts a re t aking advantage of the latest multi-CPU , multi-core PC platforms to accelerate CT reconstruction algorithms by writing C-code that is efficiently multithreaded.58

The Hounsfield Unit As previously noted, the objective of any reconstruction algorithm is to calculate the attenuation coef ficient of the tissue in each voxel of the image volume. For practical reasons, the attenuation coef ficient has historicall y been reported in Hounsfield Units (HUs) by normalizing the measured attenuation coef ficients to the attenuation of w ater and scaling the result b y a f actor of 1000 expressed as:

(µ − µ w ) HU = × 1000. µw

(13)

The HU , also called the CT number , of fers tw o advantages over the ra w attenuation coef ficient. F irst, by nor malizing the v alues of µ obtained with a gi ven scanner to the v alues of µw obtained with the same system, some reduction in measurement v ariation between scanners is achie ved. Second, b y scaling the normalized parameter b y 1000, it becomes possib le to report the quantity as an inte ger rather than as a floating point number , reducing f ile sizes, and increasing computational efficiency.

64

MOLECULAR IMAGING: PRINCIPLES AND PRACTICE

From Equation 13, it is evident that regions where µ = 0 (ie, air) will have HU values of −1000, while regions where µ = µw will have HU v alues of 0. This calibration should be repeatable for studies acquired with a given system using a specific X-ray source operating voltage and a specific X-ray filter. However, because attenuation coefficients are strongly energy dependent, any change in a system that changes the X-ra y spectr um will af fect the HU calibration. In practice, it is typically necessary to have a separation calibration for each X-ray tube setting used. Furthermore, e ven w hen a system is properl y calibrated the measured HUs are ener gy dependent. F or example, in a clinical CT scanner operating with a mean X-ray energy of 60 to 70 k eV, measured HUs for cor tical bone are typically ~1000. In contrast, a micro-CT scanner operating with mean X-ra y energy of 25 to 30 k eV will repor t HUs for cor tical bone g reater than 2000. Thus, to compare data from two different scanners, both systems must have acquired the data using similar X-ra y source settings and the X-ra y sources must produce similar spectra. As a practical matter, to perform the HU calibration, a water phantom or w ater equivalent phantom is necessary. This phantom can be any sort of uniform cylindrical object made of a low-density material and filled with distilled water. A centrifuge tube works nicely for this application with small-bore animal scanners. The tube is placed into the scanner FO V and the calibration scan is performed using the same protocol that will be used with the animal. The critical constraint is that the X-ray source voltage and the X-ray filter thickness must be the same for the calibration scan and the animal study.

Micro-CT Protocol Considerations One of the most impor tant advantages of an y small-animal in vivo imaging technology is the ability to continue to study the animal after an imaging procedure. F or this reason, it is critical that the imaging protocol has a negligible effect on the health of the animal and the outcome of the e xperiment. The anesthesia, radiation dose, and contrast media must be carefull y selected to ensure that the study is minimally invasive.

anesthetic agents include isoflurane, k etamine/xylazine, pentobarbital, and tribromoethanol. The current trend is for most imaging studies to be perfor med using isoflurane. Detailed re views of v arious anesthesia protocols for small animals may be found elsewhere.59–61

Radiation Dose Radiation dose to the animal is an impor tant concer n when designing a micro-CT protocol. A number of authors have examined the dose delivered to a mouse or rat during a micro-CT study both empiricall y and using numerical simulations. 11,62–67 Reported dose v alues range from 1 to 15 centig ray (cGy) per scan, with typical v alues of less than 5 cGy per scan. Although the dose le vel for a single scan typicall y f alls belo w the detectable limit for a ph ysiological response, with multiple scans the accumulated dose may have a deleterious effect on an experimental study. The delivered dose depends strongly on the X-ray source settings, exposure times, and number of projections included in the data acquisition.

Contrast Media In clinical studies, iodine or barium contrast media are typically administered orally, intravenously (IV), or rectally to enhance the measured CT numbers of various organs or tissues.26 Similar protocols ha ve been de veloped for smallanimal studies, although protocol design is complicated b y the rapid clearance of most clinical contrast agents in mice and rats and b y the relatively long scan times of most preclinical micro-CT systems. In most cases, some use of a contrast-enhancing agent is required to dif ferentiate between soft tissue organs in a micro-CT study. Frequently used protocols include intraperitoneal (IP) or IV injection of a nonionic w ater-soluble iodine contrast medium (eg, Amersham Omnipaque™-300), IV injection of slo wly clearing b lood pool contrast agents (e g, ART F enestra VC™) and liver imaging agents (e g, ART F enestra LC™), and oral deli very of clinical gastrointestinal imaging agents (eg, barium sulf ate). The increase in CT number pro vided b y contrast media is appro ximately expressed as:

Anesthesia

ΔCT Number µ ʹ′contrast media ≈ , C µ ʹ′

High-resolution in vivo imaging requires that the subject be immobilized during the scan; this is typicall y accomplished b y anesthetizing the animal. Commonl y used

where C is the contrast medium concentration in the tissue (mg/mL) and µʹ′ is the density-normalized attenuation coefficient (µʹ′ = µ/density cm2/g).

(14)

water

Principles of Micro X-ray Computed Tomography

IMAGING APPLICATIONS We now tur n to a re view of some of the more widel y used applications for micro-CT. In general, micro-CT is the tool of choice for studies requiring high-resolution anatomic images of tissues with high relati ve contrast. For e xample, micro-CT is ideall y suited for bone studies due to the high contrast between calcified tissue and soft tissue as w ell as for lung studies due to the high contrast betw een air and lung soft tissue. MicroCT scans of soft tissue or gans are also frequentl y performed, but these studies typically require the use of contrast-enhancing agents. In most cases, MRI is the preferred tool for anatomic imaging of soft tissue. Micro-CT is also frequentl y used to pro vide an anatomic reference for PET or SPECT studies and as a source for attenuation coef ficients to suppor t attenuation and scatter cor rection in PET and SPECT . MicroCT is an attracti ve tool for generating anatomic reference images because the hardw are is relati vely inexpensive and easy to use and scan times can be short, allowing the imager to devote most of the study time to the PET or SPECT acquisition. Because micro-CT intrinsically measures tissue attenuation coef ficients, it is well suited for generating maps for attenuation correction, although, as described belo w, some w ork is required to scale the micro-CT attenuation coef ficients to the energies of the PET or SPECT γ-rays.

Whole-Body Imaging with Micro-CT Perhaps the most commonl y used micro-CT imaging protocol is the whole-body scan. Many commercially available micro-CT scanners can acquire w hole-mouse images in a single orbit and can acquire w hole-rat images with multiple bed positions. Whole-mouse data sets typically have resolutions of 50 to 100 microns, are acquired in 1 to 5 minutes, and deliver ~5 cGy dose to the animal. A representative w hole-mouse image is shown in F igure 10. Two applications for w hole-body data sets are se gmentation of v arious anatomic or gans and structure for numerical anal ysis and calculation of attenuation coefficients in suppor t of PET and SPECT image reconstr uctions. These applications are discussed briefly below.

Whole-Body Tissue Segmentation Although the image in Figure 10 is useful for visualizing the anatom y of a laborator y animal, the reconstr ucted data set does not intrinsically provide quantitative values

65

for the sizes and v olumes of anatomic str uctures. To extract these values from the image data, it is necessar y to se gment the image v olume into dif ferent tissues of interest. If the tissue of interest has high contrast with its surroundings (ie, bone or lung), the se gmentation can often occur automaticall y b y simpl y setting g rayscale thresholds. If the image is noisy or if contrast is poor , a more manual se gmentation is required. This can be tedious and time consuming. Assume an investigator wishes to deter mine the volume of the left kidne y in F igure 11A. In this case, the mouse was given an IP injection of a clinical w ater-soluble iodinated contrast agent before the study. The contrast agent clears through the kidneys, providing good contrast between the kidne ys and the sur rounding tissue. With sufficient contrast, it may be possible to use an automated region-growing se gmentation tool to deter mine the boundaries of the kidney. In this case, the investigator identifies a point within the region of interest and defines upper and lo wer g rayscale thresholds, w hich encompass the range of v alues found in the kidne y. The algorithm then searches for boundaries between the kidney and surrounding tissue in three dimensions. When the data have sufficient contrast to suppor t this process, se gmentation can be accomplished very quickly. Frequently, the contrast betw een the str ucture of interest and the sur rounding tissue is not adequate for currently available region growing algorithms to reproducibly se gment soft tissue or gans. In this case, the quasi-manual tools are typically used, where the operator steps through the slices in the volume and draws the organ boundaries, sometimes with the aid of an edgedetecting tool. F igure 11B sho ws a slice with the kidney boundar y def ined. Once the 2-D boundaries ha ve been def ined in each slice (for e xpediency often the boundaries are manually defined in every fifth or tenth slice and an interpolation algorithm is used to fill in the gaps), the or gan volume has been def ined. Figure 11C shows a rendering of the kidne y surf ace follo wing segmentation. Assuming that the voxel size of the reconstructed image is w ell known (this is almost al ways the case), the segmented or gan surf ace ma y be used to calculate the volume of the or gan. If the density of the or gan is known (most soft tissue has a density close to that of water, 1 g/cm 3), then the w eight of the or gan ma y also be determined. In man y cases, the te xture of the image data within the surface may also be analyzed to characterize the health of the organ.67,68 With the use of contrast agents, most soft or gans can be segmented in a whole-body data set. Figure 12 shows a

66

MOLECULAR IMAGING: PRINCIPLES AND PRACTICE

A

B

Figure 10. Typical whole-mouse micro-computed tomography (CT) images. In each case, the animal was given an intraperitoneal injection of water-soluble iodinated contrast agent before the CT scan. The images were reconstructed from a tomographic data set acquired with the following paramaters: A, 360 projections (256 × 384 pixels), 360 degrees, X-ray with 80 kVp and 500 µA, reconstructed on 150 µm isotropic voxel grid; B, 512 projections (512 × 768 pixels), 360 degrees, X-ray with 80 kVp and 500 µA, reconstructed on a 84 µm voxel grid.

representative data set with multiple organ segmentations. Volumetric image se gmentation is an area of intense research focus for both clinical and preclinical applications. Ne w tools will cer tainly become a vailable to accelerate and to reduce the user inter vention in the segmentation process.

Whole-Body Imaging for PET and SPECT Attenuation Correction PET and SPECT are both emission tomo graphy technologies in w hich the patient or animal is injected with a γ-ray emitting isotope link ed to a tar geting compound designed to bind to a specif ic tissue or process of interest. As the isotope deca ys, the emitted γ-rays are detected and used to reconstr uct image v olumes in a manner similar to that used for micro-CT image

formation. The γ-rays must pass through tissue along varying path lengths before e xiting the subject and, like the X-ra ys used in micro-CT studies, are subject to attenuation. Unlike micro-CT, where the X-ray attenuation is the source of the image data, photon attenuation in PET and SPECT studies are a source of er ror that should be cor rected. Micro-CT pro vides a means for performing this correction. In PET , the γ-ray emitting isotope is actuall y a positron emitter. The positron quickly annihilates with an electron to produce two 511 keV γ-rays traveling in opposite directions. The PET scanner looks for γ-rays detected at the same time (coincident pairs) and assigns a line of response to each coincident pair . Each e vent that contributes to a PET image, therefore, is associated with tw o γ-rays that in combination travel entirely through the animal. The probability that one of the tw o γ-rays in the

Principles of Micro X-ray Computed Tomography

67

A

Figure 12. Segmented whole mouse. The lungs are shown in blue, the kidneys are brown, the heart is red, and the carotid artery in the neck is pink.

B

C

Figure 11. Micro-computed tomography (CT) image of mouse kidneys. A, raw image, (B) single slice segmentation, and (C) segmented kidney volume.

coincident pair will be attenuated is, therefore, the same as the probability that a photon emitted on one side of the animal will be attenuated as it passes through the animal to be detected on the other side. As previously described, during a micro-CT study an X-ray source and detector rotate about the object acquiring 2-D projections at v arious angles around the object. The ra w CT data are then passed to a reconstr uction program that transfor ms the 2-D projections into a 3-D map of the attenuation coef ficients ( µ-map) for the animal. Once the data is reconstr ucted, it is possib le to then forward project the CT image to create an attenuation sinogram or a map of the probability of attenuation for a number o f ph oton p aths t hrough t he a nimal. After correcting for dif ferences between the ener gy of the CT X-rays and the PET γ-rays, a raw PET data set (PET sinogram) may be attenuation cor rected b y scaling the PET sinogram b y the attenuation sino gram. This process is shown schematically in Figure 13. Unlike P ET c oincident p airs, t he p robability a SPECT γ-ray will be attenuated is dependent upon the depth at w hich the γ-ray is emitted. A number of approximate methods exist to partially correct for SPECT γ-ray attenuation, but the most widely used method today is to incor porate measured attenuation coef ficients into the probability matrix of an iterati ve reconstr uction algorithm. Once again, micro-CT data ma y be used to generate this attenuation data. If photon attenuation coef ficients w ere not ener gy dependent, it w ould be possib le to use micro-CT data sets directly to pro vide PET and SPECT attenuation

68

MOLECULAR IMAGING: PRINCIPLES AND PRACTICE

Figure 13. Schematic diagram showing the data flow for computed tomography (CT)-based positron emission tomography (PET) image reconstruction.

Tissue, Soft (ICRU – 44)

104

m/r men /r

m/r or men /r , cm2/g

103 102 101 100

102 1 102 2 102 3

102 2

102 1 100 Photon Energy. MeV

101

102

Figure 14. Energy dependent attenuation coefficients for photons passing through water.

correction. Unfor tunately, µ-values are strongl y ener gy dependent as sho wn in F igure 14. Preclinical micro-CT images are typically acquired using X-ray source voltages of around 80 kVp, producing an X-ra y flux with a mean energy of appro ximately 30 keV. In contrast, PET isotopes emit γ-rays with energies of 511 keV, while SPECT isotopes emit γ-rays with energies typically ranging from 100 keV to 300 keV. In their seminal w ork on combined PET -CT imaging systems, Kinahan and colleagues68 identified three methods

to cor rect for the ener gy dependence of the attenuation coefficients. The first approach is to simply scale the attenuation coefficient by a cor rection f actor. The scaling approach estimates the attenuation image at 511 k eV b y multiplying the CT image b y the ratio of attenuation coefficients of w ater at CT and PET ener gies. A single “effective” energy is chosen to represent the CT spectr um. This is the easiest method to implement, but the assumption of a linear relationship betw een photon ener gy and attenuation coefficient is only valid over a narrow range of energies. The scaling method tends to yield er roneous values for materials with higher atomic Z v alues, such as bone. The second approach is to segment the image into up to four dif ferent tissue types (typicall y bone, soft tissue, fat, and lung) and then assign kno wn attenuation coef ficients to each tissue type. This approach tends to be more accurate and has the added adv antage of being noiseless, but it can produce ar tifacts at the abr upt boundaries between tissue types. A signif icant problem, however, is that some soft tissue regions will have continuously varying densities that ma y not be accuratel y represented by a discrete set of segmented values, such as, for example, the lungs, where the density varies by as much as 30%. 69,70 The third approach is a hybrid of the preceding methods. The attenuation image at 511 k eV is estimated b y first using a threshold to separate out the bone component of the CT image and then using separate scaling factors for the bone and nonbone components. This

Principles of Micro X-ray Computed Tomography

69

method is easier to implement than the full se gmentation method and yields superior results to the simple scaling method.

Oncology Oncology research is one of the most acti ve f ields using laboratory animal imaging, and micro-CT has been used to measure tumor v olume and as a suppor ting tool for PET and SPECT studies. 70–80 Micro-CT is particularly useful in the study of lung tumors, w here the natural contrast between the air -filled lung and solid mass of the tumor makes it possib le to readil y identify and measure the v olume of developing tumors and to track response to therapeutic compounds.72,75,76 In these studies, respiratory gating is typically used to minimize blur due to respiratory motion. Micro-CT is also a valuable tool for bone tumor studies,73,78 where the natural contrast of sk eletal tissue suppor ts highcontrast imaging. Xeno graft studies also benef it from micro-CT scans w here the high-resolution images pro vide greater measurement accuracy of tumor volume than traditional assays. When used in conjunction with PET or SPECT studies, micro-CT provides anatomic localization of observed radiotracer uptak e79 and a µ-map for attenuation cor rection. The anatomic reference pro vided b y micro-CT has proven to be par ticularly important when highly specific tracers are used; the absence of a general uptake deprives the SPECT or PET image of an intrinsic anatomic reference, necessitating the use of a reference image to accuratel y identify the anatomic locus of uptake (Figure 15). As noted above, it is impor tant to monitor the deli vered dose w hen imaging tumor de velopment and response to therapeutic intervention to avoid confounding the observed results.

Lung Imaging For some time, CT has been a staple in clinical lung imaging due to the natural contrast that exists between the air and the tissue in the lungs. This makes CT an e xcellent choice for e xamining the comple x and inter twined structures of the pulmonar y system. In laborator y animal studies, high-resolution in vitro imaging allo ws an investigator to e xamine the highl y detailed str uctures of excised lung tissue, 80,81 while in vivo imaging allows for detailed imaging of the lung function. 82–84 The versatility of micro-CT gi ves the in vestigator a wide range of control and tools for use in analyzing pulmonary disease and for research into the function of the lung.

Figure 15. Positron emission tomography (PET)-computed tomography (CT) image (courtesy of Dr. Jamey Weichert, University of Wisconsin, Madison).

In vitro micro-CT imaging is typically performed using very high-resolution acquisition protocols. Dedicated in vitro scanners can provide image resolutions on the order of 5 µm. Because the lung or section of the lung has been excised, longer scan times and higher doses are of lesser concern. The X-ray source voltage is typically set between 60 kVp and 70 kVp with minimal f iltration to maximize the low energy component of the X-ra y spectrum, thereby maximizing the absor ption of X-ra ys in the soft tissue structures of the sample. The high-resolution images are useful for vie wing minute str uctures lik e the parench yma and for mapping of the bronchial netw orks of the lung. These images can also be used for studying pulmonar y diseases, such as emphysema and pulmonary fibrosis. In vivo imaging systems can provide images with resolutions between 10 µm and 15 µm in li ving animals and can also pro vide a considerab le deal of insight into the function of the lung. With the use of physiological monitoring systems to pro vide respiratory gating signals to the scanner, it is possible to image the mechanics of the

70

MOLECULAR IMAGING: PRINCIPLES AND PRACTICE

lung as the motion mo ves from full inspiration to full expiration. Although in vi vo imaging studies generall y use protocols that deli ver lo wer doses than in vitro systems, the dose is still high with v alues ranging from 15 cGy85 to 1 Gy83 depending on the desired resolution. Several methods of respirator y gating are typicall y used during in vi vo studies. The simplest method is to detect the respirator y c ycle of an anesthetized animal using a pneumatic sensor 83,86,87 or an optical or infrared sensor.88 The respirator y gating signal is used to trigger the projection acquisition, significantly reducing artifacts due to respiratory motion. This method has the advantage of ease of implementation but typically does not remove all of the motion-related blurring. A second method is to acquire a relati vely large number of ungated projections and then to use image analysis to retrospectively sor t the projections into se veral compar tments associated with dif ferent phases of the respirator y cycle.84,86 This method is the simplest to implement because it requires no gating apparatus during the scan, but it typically does not remove all of the motion-related blurring and it car ries with it the burden of e xcess dose to the animal received during the acquisition of unused projections. The third and largely preferred method is to place the animal on a v entilator and control the respirator y cycle.82–85 By imposing a breath-hold during the acquisition of each projection, blurring due to respiratory motion is vir tually eliminated , enab ling researchers to tak e full adv antage of the resolution of micro-CT scanners (Figure 16).

A

B

Bone Imaging Due to the high relati ve density of sk eletal tissue, bone imaging has long been a key application for high-resolution micro-CT.13,89–92 Key areas of interest include the direct measurement of bone mineral density,93 analysis of trabecular bone str ucture and calculation of trabecular mor phometric parameters, 90 tracking and anal ysis of bone tumor formation,78 and evaluation of bone implant materials. 94 Bone mineral density is typically measured by calibrating the scanner using a tissue equivalent phantom with calibrated reference materials. The measured CT numbers associated with a range of reference materials are recorded and the image data are then calibrated using the derived scale factor. In trabecular bone studies, the trabecular netw ork is typicall y se gmented from the sur rounding cor tical bone before anal ysis. The trabecular lattice is then numerically anal yzed to yield a set of mor phometric parameters including relati ve bone v olume, bone v olume/tissue volume, bone surface to volume ratio, bone

Figure 16. Micro-computed tomography (CT) lung images: A, micro-CT image of an excised mouse lung and B, a transaxial image from an in vivo micro-CT study (courtesy of Dr. Eric Hoffman, University of Iowa).

Principles of Micro X-ray Computed Tomography

surface area/bone v olume, trabecular w all thickness, trabecular wall spacing, trabecular number, and trabecular patter n f actor. These numerical f actors ser ve as indicators of bone quality (Figure 17). Bone metastases are associated with a number of cancers, breast cancer in par ticular, and are readil y detected as regions of lower density in micro-CT studies. Micro-CT has been used to detect and measure the progression of bone metastases in mouse models. 78,95 Optimization of the interface between bone tissue and implants is an area of increasing focus as joint replacements become widel y used in clinical care. The most widely used implant material, titanium, is suf ficiently radiolucent to be studied using micro-CT . A number of groups have used micro-CT to investigate animal response to sur gical methods, bone cement for mulations and implant materials in animal models. 94,96,97 A representative titanium implant study is shown in Figure 18.

FUTURE DIRECTIONS As micro-CT systems are becoming more of a standard tool used in the f ields of genetic, phar maceutical, and disease research, users w ant to use them to routinel y scan lar ge numbers of animals as quickly as possible. This will require new micro-CT developments to enable rapid scanning. The major components that limit the speed at which one can scan using micro-CT are (1) the X-ray source, (2) the X-ray detector, and (3) the rotating gantry. In addition, researchers

A

Figure 17.

71

would lik e to ha ve the ability to perfor m dynamic CT studies on animals, but the relatively slow speed of microCT as compared with the rapid rate of ph ysiologic processes in mice make high resolution (< 50 µm), dynamic micro-CT a challenge to achieve. There is a minimum dw ell time required for each projection acquired during a micro-CT scan to allo w enough X-ra y signal to reach the inte grating X-ray detector . If the po wer of the X-ra y source (and hence the X-ra y flux) can be increased , then this dw ell time can be reduced , and hence, reduce the o verall scan time. There are very powerful X-ray sources available (ie, those used in clinical CT scanners), but the challenge is to increase the po wer while maintaining a small X-ra y focal spot (eg, < 50 µm) to preserve the required micro-CT spatial resolution. Ne w X-ra y source de velopments that increase po wer w hile maintaining small focal spot size will benef it the micro-CT user . Lastl y, pulsed X-ra y sources (e g, cold cathode) impro ve one’s ability to perform gated micro-CT studies w here the hear tbeat or the respiration of the subject is used to trigger the image acquisition at a precise point in the cardiac or respirator y cycle. If the X-ray source can be pulsed (rapidly turned on and off), then there is no longer a need to use a mechanical shutter to control the timing of the X-ra y exposure. Another component that limits overall micro-CT scan time is the readout speed of the X-ra y detector. For example, a 4096 × 4096 CCD-based X-ray detector can take > 1 second to complete the full data readout, which can account

B

Trabecular Structure; in the head of a mouse femur (A) and a vertical section along the length of a mouse femur (B).

72

MOLECULAR IMAGING: PRINCIPLES AND PRACTICE

A

B

Figure 18. Titanium implant study. Micro-computed tomography (CT) volume rendering (left) and CT slice (right) of a titanium implant in the femur of a mouse.

for substantial time if the micro-CT study has hundreds of projections. If someone is interested in high-resolution studies over a large FOV, then a detector with a large number of imaging elements (pixels) is required, and detector readout can become a significant factor affecting overall scan time. Ne w detectors with f ast, lo w-noise, multipor t readout will improve the speed at which a micro-CT system can scan the subject at high resolution. If one can impro ve the X-ra y source and detector as outlined, then the X-ra y source and detector could be rapidly rotated around the subject while collecting highquality projections. At this point, the limiting f actor in micro-CT scan time can become the speed at w hich the rotating gantry can spin. High-resolution micro-CT systems with slip ring technolo gy would then become a more critical development in this f ield. Slip rings are used in clinical CT scanners because clinical systems require that the gantry be rotated at high speed through man y 360-degree orbits during a study . A slip ring system allo ws the gantr y to continuously rotate because it per mits removal of hardwired cables that can limit the total rotation of the micro-CT gantry. The slip ring interf ace provides a connection path for both power and data transfer, so cables are not required. Once the micro-CT system has a high po wer, highresolution X-ra y source, a f ast detector , and a slip ring

gantry architecture, spiral CT scanning of animals becomes possible. High-speed spiral CT has man y advantages over circular orbit cone-beam CT in ter ms of impro ved image quality and the ability to capture dynamic events in the body. Lastly, the issue of data management in micro-CT is a current challenge. A single micro-CT study of a mouse with 50 µm resolution will create an image volume of ~2 GB. In a high-throughput situation where 20 mice may be scanned in a day, 40 GB of data will be generated per day. After 1 month of scanning, the user has piled up 1200 GB of image data. This creates a substantial data storage problem, but even more importantly, a tremendous data management problem. Researchers w ould lik e to be ab le to quickl y search their micro-CT image database to f ind images of specific mice, images of multiple mice with similar anatomic features, or a collection of longitudinal studies on a single animal, just to name a few examples. The data management tools used by most micro-CT researchers fall short of providing such capabilities. Note that these data management challenges are not limited to the CT modality but are compounded w hen a user also has micro v ersions of PET, MR, and SPECT, for example, and needs to manage multiple image data modalities across their animal population. New software applications are needed that provide efficient ways to store and retrieve useful information from

Principles of Micro X-ray Computed Tomography

the potentially massive micro-CT (PET, MRI, SPECT, etc) image database. Such an application will require innovative developments in automated image analysis, feature extraction, classif ication and image database inde xing strategies for rapid search, and retrieval capabilities.

REFERENCES 1. Cormack A. Computed tomo graphy: some histor y and recent de velopments. J Appl Phys 1963;34:2722–7. 2. Hounsfield G. Diagnosis and management decisions in infections of the deep fascial spaces of the head and neck utilizing computerized tomography. Br J Radiol 1973;46:1016–22. 3. Kujoory MA, Hillman BJ , Bar rett HH. A low-cost computed tomographic scanner. Invest Radiol 1980;15:148–54. 4. Burstein P, Bjorkholm RC, Chase RC, Se guin FH. The lar gest and smallest X-ra y computed tomo graphy systems. Nucl Instr um Methods Phys Res 1984;221:207–12. 5. Miller GG, Dawson DT, Battista JJ. Compute tomographic assessment of radiation induced damage in the lung of nor mal and wr 2721 protected laf1 mice. Int J Radiat Oncol Biol Phys 1986;12:1971–5. 6. Flannery BP , Deckman HW , Rober ge WG, D'Amico KL. Threedimensional X-ray microtomography. Science 1987;237:1439–44. 7. Leyton MW, Goldstein SA, Goulet RW, et al. Examination of subchondral bone architecture in experimental osteoarthritis by microscopic computed axial tomography. Arthritis Rheum 1988;31:1400–5. 8. Feldkamp LA, Goldstein SA, P arfitt AM, et al. The direct e xamination of three-dimensional bone architecture in vitro b y computed tomography. J Bone Miner Res 1989;4:3–11. 9. Kuhn JL, Goldstein SA, F eldkamp LA, et al. Ev aluation of a microcomputed tomography system to study trabecular bone str ucture. J Orthop Res 1990;8:833–42. 10. Feldkamp L, Da vis LC, Kress JW . Practical cone-beam algorithm. J Opt Soc Am 1984; Issue 6, pp. 612–19. 11. Paulus M, Sari-Sar raf H, Gleason SS, et al. A new X-ray computed tomography system for laborator y mouse imaging. IEEE Trans Nucl Sci 1999;3:558–64. 12. Paulus MJ , Gleason SS, K ennel SJ , et al. High resolution X-ray computed tomo graphy: an emer ging tool for small animal cancer research. Neoplasia 2000;2:62–70. 13. Rüegsegger P, Koller B, Müller R. A microtomographic system for the nondestructive evaluation of bone architecture. Calcif Tissue Int 1996;58:24–9. 14. Stenström M, Olander B , Carlsson CA, et al. The use of computed microtomography to monitor mor phological changes in small animals. Appl Radiat Isot 1998;49:565–70. 15. Stenström MOB, Carlsson CA, Carlsson GA, Håkanson R. Methodologic aspects of computed microtomo graphy to monitor the development of osteoporosis in gastrectomized rats. Acad Radiol 1995;2:785–91. 16. Müller R, K oller B, Hildebrand T, et al. Relationship betw een twodimensional and three-dimensional bone architecture in predicting the mechanical strength of the pig mandib le. Technol Health Care 1996;4:113–9. 17. Müller RR, Rüe gsegger P. Micro-tomographic imaging for the nondestructive evaluation of trabecular bone architecture. Stud Health Technol Inform 1997;40:61–79. 18. Johnson RH, Hu H, Ha worth ST, et al. Feldkamp and circle-and-line cone-beam reconstruction for 3-d micro-CT of vascular networks. Phys Med Biol 1998;43:929–40. 19. Corrigan NM, Chavez AE, Wisner ER, Boone JM. A multiple detector ar ray helical X-ra y microtomo graphy system for specimen imaging. Med Phys 1999;26:1708–13.

73

20. Boone JM, Alexander GM, Seiber t J A. A fluoroscop y-based computed-tomography scanner for small specimen research. Invest Radiol 1993;28:539–44. 21. Anderson P, Levinkind M, Elliot JC. In vitro of mechanisms of de- and remineralization of dental enamel. Arch Oral Biol 1998;43:649–56. 22. Reid BK, Cunningham IA. High resolution X-ra y imaging at soft diagnostic ener gies using a silicon photodiode ar ray. Med Ph ys 1993;20:1497–504. 23. Holdsworth DW, Drangova M, Fenster A. A high-resolution xrii-based quantitative volume ct scanner. Med Phys 1993;20:449–62. 24. Holdsworth DW, Drango va M, Schulenber g K, F enster A. A tab letop ct system for high-resolution v olume imaging. SPIE 1990; 1231:239–45. 25. Flynn MJ, Hames SM, Reimann D A, Wilderman SJ. Microfocus X-ray sources for 3-d microtomo graphy. Nucl Instr umMethods Phys Res Sect A 1994;353:312–5. 26. Sprawls P . The ph ysical principles of medical imaging, 2nd ed. Rockville: Aspen publications; 1993, Vol 2, Madison, WI: Medical Physics Publishing; 2000. 27. Cheng Y, Zhang J, Lee YZ, et al. Dynamic X-ray radiography using a carbon nanotube f ield emission X-ra y source. Re v Sci Instr um 2004;10:3264. 28. Zhang J, Yang G, Cheng Y, et al. Stationar y scanning X-ra y source based on carbon nanotube f ield emitters. Appl Ph ys Lett 2005;86:184104–6100. 29. Liu Z, Yang G, Lee Yueh Z, et al. Carbon nanotube based microfocus field emission X-ray source for microcomputed tomography. Appl Phys Lett 2006;89:103–11. 30. Grodzins L. Optimum ener gies for X-ra y transmission tomo graphy of small samples: applications of synchrotron radiation to computerized tomo graphy. Nucl Instrum Methods Phys Res 1983; 206:541–5. 31. Stonestrom JP, Alvarez RE, Maco vski A. A frame work for spectral artifact cor rections in X-ra y CT . IEEE Trans Biomed Eng 1981;28:128–41. 32. Duerinckx AJM, Macovski A. Polychromatic streak ar tifacts in computed tomography images. J Comput Assist Tomogr 1978;2:481–7. 33. Tuy HK. An inversion formula for cone-beam reconstr uction. SIAM J Appl Math 1983;43:546–52. 34. Yan M, Zhang C. Tilted plane feldkamp type reconstruction algorithm for spiral cone beam ct. Med Ph ys 2005;32:3455–67. 35. Defrise M, Noo F, Kudo H. A solution to the long-object prob lem in helical cone-beam tomography. Phys Med Biol 2000;45:623–43. 36. Liu V, Lariviere NR, Wang G. X-ray micro-CT with a displaced detector ar ray: application to helical cone-beam reconstr uction. Med Phys 2003;30:2758–61. 37. Yu H, Wang G. Studies on implementation of the katsevich algorithm for spiral cone-beam ct. J Xray Sci Technol 2004;12:97–116. 38. Noo F, Pack J, Heuscher D. Exact helical reconstruction using native cone-beam geometries. Phys Med Biol 2003;48:3787–818. 39. Noo F, Hoppe S, Dennerlein F, et al. A new scheme for view-dependent data differentiation in f an-beam and cone-beam computed tomo graphy. Phys Med Biol 2007;52:5393–414. 40. Pack JD, Noo F. Cone-beam reconstr uction using 1 d f iltering along the projection of-lines. Inverse Probl 2005;21:1105–20. 41. Zabic S, Hoppe S, Dennerlein F, et al. Resampling density v alues on r-lines into density v alues on a car tesian g rid. IEEE 2006; Proceedings of the IEEE Nuclear Science Symposium, M06–449. 42. Yang H, Li M, Koizumi K, Kudo H. Exact cone beam reconstr uction for a saddle trajectory. Phys Med Biol 2006;51:1157–72. 43. Herman G. Image reconstructions from projections: the fundamentals of computerized tomography. New York: Academic Press; 1980. 44. Xia D, Yu L, Sidky EY, et al. Noise proper ties of chord-image reconstruction. IEEE Trans Med Imaging 2007;26:1328–44. 45. Lee S, Cho G, Wang G. Artifacts associated with implementation of the grangeat formula. Med Phys 2002;29:2871–80.

74

MOLECULAR IMAGING: PRINCIPLES AND PRACTICE

46. Katsevich A. Theoretically exact filtered backprojection-type inversion algorithm for spiral ct. SIAM J Appl Math 2002;62:2012–26. 47. Grangeat P. Mathematical framwork of cone beam 3-D reconstruction via the first derivative of the radon transform. In: Louis AK, Natterer F, editors. Mathematical methods in tomo graphy. Vol 1497. Berlin, Germany: Springer; 1991. p. 66–97. 48. Lee SW, Cho G, Wang G. Artifacts associated with implementation of the grangeat formula. Med Phys 2002;29:2871–80. 49. Deng J, Yu H, Ni J , et al. A parallel implementation of the katse vich algorithm for 3-d CT image reconstruction. J Super Comput 2006; 38:35–47. 50. Shepp LA, Vardi Y. Maximum lik elihood reconstr uction in positron emission tomography. IEEE Trans Med Imaging 1982;1:113–22. 51. Benson TM, Gregor J. Framework for iterative cone-beam micro-CT reconstruction. IEEE Trans Nucl Sci 2005;52:1335–40. 52. Elbakri IA, Zhang Y, Chen L, et al. Statistical reconstr uction for quantitative ct applications. IEEE Nucl Sci Symp 2003; 4:2978–80. 53. Gregor J, Gleason SS, P aulus MJ, Cates J. Fast feldkamp reconstruction based on focus of attention and distributed computing. Int J Imaging Syst Technol 2002;12:229–34. 54. Stsepankou D, Kornmesser K, Hesser J, Manner R. FPGA-acceleration of cone-beam reconstruction for the X-ray CT. In: IEEE international conference on f ield-programmable technolo gy. 2004. p. 327–30. 55. Xu F and Mueller K. GPU-Real-T ime 3D Computed Tomographic Reconstruction Using Commodity Graphics Hardware, Physics in Medicine and Biology, 2007;52:3405–3419. 56. Okitsu Y, Ino F , and Hagihara K. Accelerating Cone Beam Reconstruction Using the CUD A-Enabled GPU , HiPC 2008, LNCS 5374, P. Sadayappan et al. (Eds.), Springer -Verlag Berlin Heidelberg, 2008, pp. 108–19. 57. Sharp GC, Kandasamy N, Singh H, F olkert M. GPU-based architectures for CT reconstruction and registration. Phys Med Biol 2007; 52:5771–83. 58. Zeng K, Bai E, and Wang G. A Fast CT Reconstruction Scheme for a General Multi-Core PC, Inter national Jour nal of Biomedical Imaging, 2007. 59. Flecknell PA. Anaesthesia of animals for biochemical research. Br J Anaesth 1993;71:885–94. 60. Flecknell PA. Laboratory animal anaesthesia. San Die go, CA: Academic Press; 1996. 61. Gardner D , Da vis J A, Weina PJ , Theune B . Comparison of tribromoethanol, ketamine/acetylpromazine, telazol/xylazine, pentobarbital, and metho xyflurane anesthesia in hsd:Icr mice. Lab Anim Sci 1995;45:199–204. 62. Obenaus A, Smith A. Radiation dose in rodent tissues during microCT imaging. J Xray Sci Technol 2004;12:241–9. 63. Taschereau R, Chow C, Chatziioannou AF. Current status and new horizons in monte carlo simulations of dose from micro-CT imaging procedures in a realistic mouse phantom. Med Ph ys 2006; 33:216–24. 64. De Clerck N, Meurrens K, Weiler H, et al. In vivo high-resolution X-ray microtomography for liver and spleen tumor assessment in mice. Neoplasia 2004;6:374–9. 65. Boone J , Velazquez O , Cher ry SR. Small-animal X-ra y dose from micro-CT. Mol Imaging 2004;3:149–58. 66. Carlson S, Classic KL, Bender CE, Russell SJ. Small animal absorbed radiation dose from serial micro-computed tomography imaging. Mol Imaging Biol 2007;2:78–82. 67. Gleason SS, Sari-Sarraf H, Abidi MA, et al. A new deformable model for analysis of X-ray CT images in preclinical studies of mice for polycystic kidne y disease. IEEE Trans Med Imaging 2002; 10:1302–9. 68. Kinahan P E, Townsend D W, B eyer T, S ashin D . Attenuation correction for a combined 3-d pet/ct scanner . Med Ph ys 1998; 10:2046–53.

69. Robinson P, Kreel L. Pulmonar y tissue attenuation with computed tomography: comparison of inspiration and expiration scans. J Comput Assist Tomogr 1979;3:740–8. 70. Weichert JP, Moser AR, Weber SM, et al. Radioiodinated NM404-a universal tumor imaging agent? Acad Radiol 2005; Issue 5, p. S58–S59. 71. Badea CT, Hedlund LW, De Lin M, et al. Tumor imaging in small animals with a combined micro-CT/micro-DSA system using iodinated con ventional and b lood pool contrast agents. Contrast Media Mol Imaging 2006;1:153–64. 72. Chang C, Jan ML, F an KH, et al. Longitudinal e valuation of tumor metastasis by an fdg-micropet/microct dual-imaging modality in a lung carcinoma-bearing mouse model. Anticancer Res 2006; 26:159–66. 73. Mouchess M, Sohara Y, Nelson MD Jr , et al. Multimodal imaging analysis of tumor pro gression and bone resor ption in a murine cancer model. J Comput Assist Tomogr 2006;30:525–34. 74. Pickhardt PJ, Halberg RB, Taylor AJ, et al. Microcomputed tomography colonography for polyp detection in an in vivo mouse tumour model. Proc Natl Acad Sci U S A 2005;102:3419–22. 75. Cody DD, Nelson CL, Bradle y WM, et al. Murine lung tumor measurement using respirator ygated micro-computed tomo graphy. Invest Radiol 2005;40:263–9. 76. Kennel SJ , Da vis IA, Branning J , et al. High resolution computed tomography and MRI for monitoring lung tumor g rowth in mice undergoing radioimmunotherapy: correlation with histology. Med Phys 2000;27:1101–7. 77. Weber SM, P eterson KA, Durk ee B, et al. Imaging of murine li ver tumor using micro-CT with a hepatocyte-selective contrast agent: accuracy is dependent on adequate contrast enhancement. J Sur g Res 2004;119:41–5. 78. Winkelmann CT, Figueroa SD, Rold TL, et al. Microimaging characterization of a b16-f10 melanoma metastasis mouse model. Mol Imaging 2006;5:105–14. 79. Chow PL, Stout DB , Komisopoulou E, Chatziioannou AF. A method of image re gistration for small animal, multi-modality imaging. Phys Med Biol 2006;51:379–90. 80. Langheinrich AC, Leithäuser B , Greschus S, et al. Acute rat lung injury: feasibility of assessment with micro-CT. Radiology 2004; 233:165–71. 81. Sera T, Uesugi K, Yagi N. Localized mor phometric defor mations of small airways and alveoli in intact mouse lungs under quasi-static inflation. Respir Physiol Neurobiol 2005;147:51–63. 82. Hoffman E, Chon D . Computed tomo graphy studies of lung v entilation and perfusion. The Proceedings of the American Thoracic Society 2005. pp. 492–98. 83. Namati E, Chon D, Thiesse J, et al. In vivo micro-CT lung imaging via a computer-controlled intermittent iso-pressure breath hold (IIBH) technique. Phys Med Biol 2006;51:6061–75. 84. Badea C, Hedlund LW, Johnson GA. Micro-CT with respirator y and cardiac gating. Med Phys 2004;31:3324–9. 85. Cavanaugh D, Johnson E, Price RE, et al. In vi vo respirator y-gated micro-CT imaging in small-animal oncolo gy models. Mol Imaging 2004;3:55–62. 86. Ford NL, Nik olov HN , Norle y CJ , et al. Prospecti ve respirator ygated micro-CT of free breathing rodents. Med Ph ys 2005; 32:2888–98. 87. Hsu LL, Schimel DM. Computed tomo graphy imaging of lungs in mouse models of human disease—advancing the computing interfaces with physiology. In: 17th IEEE symposium on computerbased medical systems (CBMS'04). 2004. p. 385. 88. Gum F, Agaoglu D, Schild R, et al. Phantom v erification of respiratory gating system for lung and liver. Int J Radiat Oncol Biol Phys 2006;66:S605. 89. Hildebrand T, Rüegsegger P. A new method for the model-independent assessment of thickness in three-dimensional images. J Microsc 1997;185:67–75.

Principles of Micro X-ray Computed Tomography

90. Laib A, Barou O , Vico L, et al. 3-d micro-computed tomo graphy of trabecular and cortical bone architecture with application to a rat model of immobilisation osteoporosis. Med Biol Eng Comput 2000;38:326–32. 91. Engelke K, Karolczak M, LutzA, et al. Micro-CT: technology and applications for assessing bone str ucture. Radiologe 1999;39:203–12. 92. Bagi CM, Hanson N , Andresen C, et al. The use of micro-CT to evaluate cortical bone geometry and strength in nude rats: correlation with mechanical testing, pqct and dxa. Bone 2006; 38:136–44. 93. Schmidt C, Priemel M, K ohler T, et al. Precision and accurac y of peripheral quantitative computed tomography (pqct) in the mouse skeleton compared with histolo gy and microcomputed tomo graphy (micro-CT). J Bone Miner Res 2003;18:1486–96. 94. Akca K, Sarac E, Baysal U, et al. Micro-morphologic changes around biophysically-stimulated titanium implants in ovariectomized rats, head & face medicine. Head Face Med 2007;3.

75

95. Li X, Zanzonico P, Ling CC, O'Dono ghue J. Visualization of experimental lung and bone metastases in li ve nude mice b y X-ra y micro-computed tomo graphy. Technol Cancer Res Treat 2006; 5:147–55. 96. Xing Z, Hasty KA, Smith RA. Administration of pamidronate alters bone-titanium attachent in the presence of endotoxin-coated polyethylene par ticles. J Biomed Mater Res B Appl Biomater 2007;83:354–8. 97. Otsuki B, Takemoto M, Fujibayashi S, et al. Pore throat size and connectivity deter mine bone and tissue ing rowth into porous implants: three-dimensional micro-CT based str uctural anal yses of porous bioacti ve titanium implants. Biomaterials 2006; 27:5892–900. 98. Gleason, S., Osborne, D. Study of CT-Based Attenuation Correction Factors for microPET using kVp Dependent HU Calibration. Joint Molecular Imaging Conference: Pro vidence, RI 2007; 07-A-1153-SMI

6 SMALL ANIMAL SPECT, SPECT/CT, SPECT/MRI

AND

NEAL H. CLINTHORNE, PHD AND LING-JIAN MENG, PHD

Single photon emission computed tomo graphy (SPECT) of small animals is becoming an increasingl y impor tant tool in biomedical research. As in positron emission tomography (PET), it is a functional imaging technique that for ms a three-dimensional image of the distrib ution of a radiotracer injected into a small animal.And like PET, the trend has been to ward the de velopment of SPECT instruments that combine either X-ray computed tomography (CT) or magnetic resonance imaging (MRI). In contrast to PET, however, counting efficiency is generally lower requiring longer imaging times or higher injected doses, w hereas spatial resolution—depending on the needs of the e xperiment—can either be w orse or signif icantly better than PET. Although it has often been claimed that PET is quantitati ve while SPECT is not, there is no reason that SPECT images cannot be quantitati ve if appropriate calibrations and cor rections are made. Moreover, at resolutions in the deep submillimeter re gime, SPECT likely has the advantage over PET in quantitative accuracy. PET images are b lurred b y the f act that positrons tra vel some distance before annihilating, and although “resolution reco very” methods can be used to reduce the resulting par tial v olume ef fects (see Section “Accounting for Degradations: Advanced SPECT Reconstruction”), the y signif icantly increase noise in reconstructed images. A key advantage is that SPECT tends to be easier for many laboratories to integrate. Although many dedicated small animal SPECT instruments are commercially available, often gamma cameras for clinical imaging that have been decommissioned from the hospital can be modif ied for small animal SPECT . Fur thermore, preparation of SPECT tracers is typicall y simpler than PET . Man y radiotracers used in patient imaging are readil y usab le in small animals, eliminating the need for adv anced 76

radiochemical synthesis and the c yclotron typicall y required in PET. The pur pose of this chapter is to pro vide an overview of small animal SPECT applications and instrumentation. Because of the e xplosion of both research and commercial acti vity in SPECT , SPECT/CT, and SPECT/MRI devices for small animals, this chapter has been organized to provide some history and a basic o verview of applications and commercial instruments followed by a description of basic hardware and image for mation methods in SPECT and CT . Design trends in small animal SPECT de vices are then discussed, and this chapter concludes with a brief discussion of future directions in the f ield. F inally, in contrast to clinical SPECT imaging, it is impor tant to note that preclinical SPECT is f ar from an “old” or mature technolo gy; no vel instr umentation and techniques are developing at a rapid pace. Particularly exciting are single photon emission microscop y (SPEM) methods for achieving spatial resolutions of 250 µm and better at useful sensitivity. Many reviews of small animal SPECT applications and instr umentation have been pub lished in the last fe w years.1–6 An excellent treatment of small animal SPECT principles, instrumentation, and methods can be found in the book by Kupinski and Barrett,7 while basic principles of the ph ysics of radionuclide imaging are described in Physics in Nuc lear Medicine.8 More advanced topics on emission tomo graphy are discussed in the book b y Wernick and Aarsvold.9

HISTORY Animal imaging using SPECT has e xisted as long as the modality itself. The Humongotron, an earl y

Small Animal SPECT, SPECT/CT, and SPECT/MRI

SPECT instr ument for patients constr ucted using a standard clinical Anger camera mounted to an old cesium radiotherapy machine, w as used in the v alidation of cardiac SPECT for human subjects through do g heart imaging experiments.10 At nearly the same time, a similar study w as conducted b y Jaszczak and colleagues on the Searle Radio graphics prototype SPECT system.11 Spatial resolution for both de vices was probably close to 10 mm full width at half maximum (FWHM). Soon thereafter , a do g hear t inf arct model was used to v alidate a 72-pinhole time-modulated coded aper ture for human cardiac imaging. 12 Resolution of the device ranged from 3.8 mm FWHM at 4 cm from the aperture to 7.8 mm at 12 cm. The instrument, however, suf fered from the ar tifacts of limited-angle tomography (see Section “Image Formation and Reconstruction in SPECT and CT”). In a study of the temporomandibular joint with bone imaging agent 99m Tc-MDP in rhesus macaques, a single-slice prototype tomograph (SPRINT) for human brain imaging 13 was outf itted with a special imaging aper ture that reduced the transaxial f ield-of-view (FO V) to 10 cm while improving spatial resolution to 4.5 mm FWHM. In the quest for higher spatial resolution for imaging 123 I and 131I tumor x enografts in mice and rats with labeled monoclonal antibodies, an animal SPECT instrument w as de veloped based on the design of the SPRINT II. 14 The imaging aper ture allo wed 1 mm FWHM resolution for 99mTc labeled tracers and 2 mm FWHM for 131I. It w as e ven used briefl y for imaging positron emitting nuclides at high resolution for dosimetry studies in rats prior to the a vailability of modern small animal PET scanners. 15 The “modern” era of small animal SPECT star ted in the earl y 1990s with the introduction of clinical SPECT cameras modified to use pinhole aper tures at high magnifications. The adv antage of this approach, w hich is described in more detail in Section “Basic Principles,” was that the modest intrinsic resolution of the clinical SPECT camera could be de-emphasized and SPECT resolution of 1 to 2 mm FWHM and better w as easily attainable over a small FOV.16,17 The general approach has been extended to provide spatial resolutions toda y that approach 100 µm. Nevertheless, with a lack of w ell-developed small animal models of human disease at the time, interest in small animal SPECT faded. In the past decade, due in part to development of better small animal models (e g, gene knock-ins and knock-outs in mice) and new SPECT radiotracers, and in par t to the de velopment of better SPECT instr umentation, small animal SPECT has gained signif icant momentum and interest is rapidly increasing.

77

RADIONUCLIDES, RADIOTRACERS, AND APPLICATIONS Radiotracers and Applications Numerous radiotracers ha ve been used in small animal SPECT (primaril y in murine models) for applications ranging from central ner vous system (CNS) imaging 18–32 to hear t imaging, 33–40 oncology,41–45 stem cell tracking,36,40,46–49 and others. 50–56 As noted abo ve, one of the advantages of SPECT is that many radiolabeling “kits” are available for clinical SPECT imaging that cross over to use in animal imaging, eliminating the need for an on-site cyclotron and reducing or eliminating the need for radiochemists required in PET. Table 1 lists common SPECT radionuclides along with their half-lives, emitted photon ener gies, and sample radiotracers. By f ar the most common is the transition metal 99mTc, which is obtained in the for m of TcO4− ions through elution of a 99Mo generator.8 Its 140 k eV photon emission and absence of higher ener gy photons is nearl y ideal for SPECT image formation. Because of these favorable properties, it has been used to label a wide v ariety of radiotracers such as MIBI for m yocardial b lood flo w,33 MDP for bone, HMP AO (e xametazime) for cerebral blood flow, and Annexin V for apoptosis. 20 For reference, Figure 1 is a high-resolution 99mTc-MDP bone scan of a mouse, while Figure 2 shows gated tomographic short- and long-axis slices of myocardial blood flow in a mouse heart at end-systole and end-diastole using 99mTc-Sestamibi. Many times, the 6-hour half-life of 99mTc is too shor t relative to the time required to accumulate suf ficient tracer in tissue or to achie ve maximum tar get-to-background ratios. This is often the case with man y labeled peptides and monoclonal antibodies. Typically, 111In with its 2.8-da y half-life is used for labeling these tracers; however, 111In chelation is more challenging as w ell as less stable in vivo, and its higher energy photon emissions are less suitab le for high-resolution SPECT imaging. Despite these limitations, its use in SPECT is widespread. Several iodine isotopes are also widel y used in SPECT. A variety of 125I and 131I labeled compounds are available commercially (eg, see Web sites of P erkinElmer, GE Healthcare, or Biomedical Technologies). Their long half-li ves (59 and 8 da ys, respecti vely) are often adv antageous, but the emission ener gy of 131I is poorly suited to high-resolution SPECT, while that of 125I is not w ell suited for imaging animals much lar ger than mice (or rats for some applications). Ne vertheless, because of the number of a vailable compounds and the fact that very high spatial resolution is possible using 125I,

78

MOLECULAR IMAGING: PRINCIPLES AND PRACTICE

Table 1. COMMON SINGLE-PHOTON RADIONUCLIDES, CHARACTERISTICS, AND APPLICATIONS Isotope

Half-Life

Photon Energy (% Abundance)

Notes and Applications

99m

Tc

6.02 h

140 keV (89.1)

Most commonly used SPECT isotope. Generator produced and readily available in nuclear medicine clinics. Numerous radiotracer kits available including MIBI for myocardial perfusion and some tumors, MAG3 for renal scanning, HMPAO for brain blood flow, MDP for bone imaging. Autologous RBC labeling for blood-pool imaging. Labeled Bombesin for tumors and Annexin V for apoptosis. The 140 keV emission and absence of higher energy photons are nearly ideal for SPECT.

201

Tl

3.04 d

68–82 keV Hg X-rays, 167 keV (10.0)

Thallous chloride for myocardial perfusion—largely replaced by MIBI above. Sometimes useful in dual-isotope studies with 99mTc or 123I.

111

2.8 d

171 keV (90.7), 245 keV (94.1)

Monoclonal antibody, protein and peptide labeling, RBC labeling. General cell labeling for tracking.

123

13.3 h

125

59 d

159 keV (83.3), 528 keV (1.4) 27–32 keV Te X-rays, 35.5 keV (6.7) 364 keV (81.5), 637 keV (7.2), 723 keV (1.8)

All iodine isotopes have been used to label a wide variety of molecules such as peptides, monoclonal antibodies, and brain imaging agents such as iodoamphetamine and iodobenzamide. They have also been used as I− for assessing thyroid function (or for radioablation of thyroid tissue in the case of 131I). Clinically, 123I is most often used for diagnostic imaging because of its favorable photon energy, whereas 131I is used for internal radiotherapy applications due to its corresponding β emissions; its high photon energies are not particularly well suited to SPECT. Even the small—abundance of 637 and 723 keV photons significantly degrade imaging performance over 364 keV photons alone. A variety of 125I labeled compounds are commercially available. The low-energy photons of 125 I are readily absorbed by even a few centimeters of tissue but are easily imaged, and research SPECT instrumentation is capable of better than 100 µm resolution.

93 keV (35.7), 185 keV (19.7), 300 keV (16.0)

67

In I I

131

I

8.02 d

67

Ga

3.3 d

Ga-citrate—Hodgkin’s disease, lymphomas, etc. Some acute inflammatory lesions. Presence of 300 keV emission compromises ability to obtain high resolution.

SPECT = single photon emission computed tomography; RBC = red blood cell.

it is likely to become an increasingl y important radionuclide for small animal SPECT . At 159 k eV, the photon energy of 123I is better but relati vely fe wer compounds, such as 123I-IBZM and 123I-FP-CIT, are readily obtainable commercially.28 Of course, tracer synthesis and radioiodination can also be perfor med on site in an appropriatel y equipped radiochemistr y f acility, b ut these resources are likely unavailable in most small animal SPECT installations. One feature of SPECT is that multiple radiotracers can be imaged simultaneousl y and their distributions reconstructed indi vidually if the tracers are labeled with radionuclides emitting dif ferent ener gy photons. Figure 3 sho ws a reconstr ucted and rendered image using 201Tl (~80 keV, rendered as green) for myocardial perfusion, 99mTc-MDP (140 k eV, orange) for bone imaging, and 123I− (159 k eV, b lue) for th yroid uptak e. The SPECT images ha ve also been fused with a CT image of the mouse. Performance in accomplishing isotope separation depends upon the ener gy resolution of

the camera system. In this case, a cadmium zinc telluride (CZT) detector ha ving 4.5% FWHM ener gy resolution at 140 keV was used to achieve good separation between the 167 k eV emission of 201Tl and the 159 keV emission of 123I. Although the choice of radiolabel and tracer is important, tw o additional issues must be considered in small animal SPECT. The f irst is the mass of compound injected. In human imaging, a trace quantity of the compound is injected that does not e xhibit a phar macologic effect, w hereas a similar quantity injected into a rat or mouse may well have an effect.57 The second and related issue is radiation dose to the animal. Funk and colleagues estimated that w hole body radiation dose to mice for many SPECT studies is a signif icant fraction of the LD50/30 (50% mortality in 30 days) and high enough to result in increased gene e xpression.58 This is especiall y relevant if longitudinal SPECT imaging studies requiring multiple tracer injections are anticipated.

Small Animal SPECT, SPECT/CT, and SPECT/MRI

79

Figure 1. 99mTc-MDP mouse bone scan acquired using USPECT-II (Courtesy F. Beekman, MI-Labs).

Figure 3. Mouse injected with 99mTc MDP (orange) for bone metabolism, 201Tl (green) for myocardial perfusion, and 123I (blue) for thyroid imaging. The image is acquired with a single list-mode SPECT acquisition, and the data were energy-discriminated post acquisition to generate three separate SPECT images. The image also shows a coregistered CT scan as a rendered transparent skeletal surface to give anatomic reference to SPECT images (Courtesy K. Iwata, Gamma Medica-Ideas).

ED

ES

Figure 2. Gated 99mTc-sestamibi mouse heart at end-diastole (top panel) and end-systole (bottom panel). Leftmost images are sections across the short-axis of the left ventricle while two orthogonal sections along the long-axis are shown at the center and right (Courtesy F. Beekman, MI-Labs).

Commercial Small Animal SPECT Instruments A number of SPECT and SPECT/CT systems are a vailable commerciall y. By the time this book is pub lished, the offerings will have undoubtedly changed. Our objective is neither to make recommendations of specific systems nor to pro vide an e xhaustive sur vey and comparison of e xisting de vices. Rather , those contemplating the purchase of a SPECT system for small animal imaging should be a ware of se veral things. F oremost is whether the system—or even SPECT for that matter— is suitable for answering the relevant research questions. As opposed to PET , where a single instr ument may be useful for a wide range of tracers in both mice and rats, SPECT performance depends upon how well the system is matched to the imaging task. For example, is high resolution in a small FO V required? Is it necessar y to

follow rapidl y changing dynamic processes? The configuration of most SPECT de vices can be altered to suit the imaging task, but performance optimization will likely require a good w orking relationship with the manufacturer for e ven minor hardw are and softw are modifications. As an e xample of a commercial de vice, F igure 4 shows the FLEX Triumph scanner available from Gamma Medica-Ideas (http://www.gm-ideas.com). The instrument can combine SPECT, CT, and/or PET on a single gantr y. The SPECT system uses CZT as the detector (see Section “Performance Optimization and Design Trends”) and can use multiple image for mation or collimation schemes depending on the animal model and the imaging task. A high-resolution SPECT instr ument from MI-LABS (http://www.milabs.com), the U-SPECT II, is sho wn in Figure 5. It is based on a three-headed scintillation camera (described in Section “Basic Principles”) and multiple pinhole collimators (Section “P erformance Optimization and Design Trends”), which can be changed to suit applications for imaging mice and rats. Bioscan offers two systems, the NanoSPECT/CT with four SPECT detectors and the HiSPECT system (http://www.bioscan.com). The significance of the latter de vice is that it can pro vide small animal imaging capabilities b y adding multi-pinhole collimation to standard clinical Anger cameras. ISE (http://www .isesrl.com) of fers a combined SPECT/PET instr ument

80

MOLECULAR IMAGING: PRINCIPLES AND PRACTICE

y

p (s, θ)

Object s

SPECT Camera

θ

x f (x, y)

Figure 4. Gamma Medica-Ideas FLEX Triumph preclinical imaging system. Instrument can incorporate up to three modalities: SPECT, PET, and CT (Courtesy K. Iwata). Figure 6. Rotating camera SPECT image acquisition for a single tomographic slice. SPECT camera and collimator rotate around object and collect a set of line-integral projections p(s, θ) of the distribution of radiotracer f(x, y) in the object.

instrument (http://www.neurophysics.com). As noted , our list is not exhaustive and new small animal SPECT companies and products appear almost monthly.

BASIC PRINCIPLES Image Formation and Reconstruction in SPECT and CT

Figure 5. MI-Labs U-SPECT-II preclinical SPECT system. Interchangeable imaging apertures can be chosen for imaging both mice and rats at high spatial resolution (Courtesy F. Beekman, MI-Labs).

based on YAP scintillation cameras (Section “Performance Optimization and Design Trends”). In addition to its small animal PET systems, Siemens has introduced the In veon SPECT system (http://www .medical.siemens.com). General Electric has in the recent past of fered the eXplore SpeCZT camera, w hich uses a “slit-slat” collimation approach similar to SPRINT II (http://www .gehealthcare.com). Finally, Neurophysics offers the MollyQ SPECT instrument that uses a novel image formation method similar to confocal microscopy resulting in a relatively low-cost

Figure 6 illustrates the principles of SPECT data acquisition. A γ-ray imaging camera and image for mation collimator (discussed in detail belo w) collect a set of projections or line-inte grals of acti vity through the object for each parallel “slice” of a 3D volume. Conceptually, the simplest data acquisition scheme is shown where the detector rotates around the object at least 180° and collects a set of parallel-ray line-integral projections at each view-angle. Tomographic infor mation for the complete v olume is generally obtained b y using a tw o-dimensional gamma camera. Ne glecting the issue of γ-ray attenuation b y the object (discussed fur ther belo w), the set of line-inte gral projections for each slice can be represented as follo ws: R

R

p( s, θ) = ∫ ∫ f ( x, y )δ ( x cos θ + y sin θ − s )dxdy , −R −R

where f (x, y) is the radiotracer distribution in a 2D slice of the object, R is the radius of the maximum FO V, and p(s, θ ) is the projection data inde xed by the vie w-angle of the camera θ and the displacement s from the central

Small Animal SPECT, SPECT/CT, and SPECT/MRI

ray. Reco vering or reconstr ucting the distribution of radiotracer requires collection of projections over the 2D interval [−R, R] × [0, π]. Data acquisition for X-ray CT is similar and Figure 7 shows a fan-beam CT setup. An X-ray detector and X-ray source corotate around the object with the detector measuring the transmission of X-ra ys along each line integral. Using Beer’s Law, the underlying linear attenuation µ(x, y), which we wish to estimate, is related to the projections for each slice by the following: R

R

p(ϕ, θ) = I 0 (ϕ ) ∫ ∫ exp[ − µ ( x, y )δ ( x cos(θ + ϕ ) −R −R

+ y sin(θ + ϕ ) − X s sin ϕ )]dxdy, where I0 is the X-ra y source flux on the detector in the absence of the object and X s is the distance from the X-ray focal-spot to the rotation axis (isocenter). The measured projections for this f an-beam CT e xample are indexed by view-angle θ and offset-angle ϕ on the X-ray detector. As for SPECT , a complete set of CT data requires collecting line-integrals over the interval [−R, R] × [0, π] (where −R ≤ X s sin ϕ ≤ R). SPECT image reconstr uction up until recentl y was accomplished using the filtered bac kprojection reconstruction algorithm. An intuiti ve approach to reconstruction for the parallel-ray geometry is to simply take the projection at each vie w-angle and “smear” it

back across the image at the appropriate angle, accumulating the result at each point in the image estimate.This straight backprojection operation results in an image in which g ross object features are e vident but details are severely b lurred. The Fourier slice or Central Section Theorem shows that the tomo graphic measurement process overemphasizes the low spatial frequency information and that an unb lurred estimate of the object can be recovered by f iltering each projection prior to backprojection with a f ilter having magnitude that increases linearly with spatial frequenc y—ie, a ramp filter in Fourier frequency space.59 The operation can be also be done as a con volution in the spatial domain, and the resulting f iltered backprojection reconstr uction is as follows: π ∞

fˆ ( x, y ) = ∫ ∫ h( x cos θ + y sin θ − s ) p( s, θ)dsdθ , 0 −∞

where h(s) is the spatial domain representation of the ramp filter. The operations of f iltering and backprojection can be re versed with an adaptation of the f ilter. The projection data are first backprojected and are then filtered with the appropriate 2D f ilter representation. Reconstruction of CT data is similar except that (1) measured projection (transmission) data are transformed prior to reconstr uction to give the linear attenuation along each line-integral by the following: p(ϕ, θ) = log

y

p (ϕ, θ)

Obje ct

θ X-ra y De te ctor

x

µ(x, y )

ϕ X-ra y S ource Xs

Figure 7. Typical fan-beam X-ray CT data acquisition in which an X-ray source and a charge-integrating detector corotate around the object and collect a set of transmission measurements along lines through the object. With the transformation noted in the text, transmission measurements are converted to integrals of linear attenuation along each line and can be reconstructed with same method used for emission tomography.

81

I 0 (ϕ ) p(ϕ, θ)

,

and (2) v arious for ms of f an-beam rather than parallelray reconstructions are used. Further information on various f iltered backprojection reconstr uction methods can be found in the referenced literature. 59–61 Although f iltered backprojection methods are still widely used in CT reconstr uction, the y ha ve lar gely been replaced in SPECT and PET applications b y the statistically moti vated methods described in Section “Accounting for De gradations: Advanced SPECT Reconstruction,” which allow for straightforward compensation of de gradations occur ring in emission tomography. Sampling

For the simplest, slice-b y-slice parallel-ra y geometr y shown in F igure 6, the sampling increment Δs between parallel ra ys along each projection and the inter -slice increment is chosen as follows:

82

MOLECULAR IMAGING: PRINCIPLES AND PRACTICE

Δs


2 πR Δx

, Δθ
25 k eV) used in small animal SPECT ; instead, image formation relies on absorbing collimation in w hich most photons emitted in the direction of the camera are b locked to estimate the directions of those that are detected. More details on ph ysical mechanisms for image formation from photons and their suitability for SPECT can be found in Furenlid and colleagues. 71 The two most popular collimation schemes for small animal SPECT are e xamined belo w; a k ey feature the y share is use of a photon absorbing material. The ideal absorber w ould b lock photon transmission with a v ery thin la yer of material. This can, in f act, be closel y approached at lo wer photon ener gies. Table 2 sho ws several absorbing materials used in small animal SPECT collimators. Attenuation length is the mean free path of photons in the material and the distance at w hich 63% of

84

MOLECULAR IMAGING: PRINCIPLES AND PRACTICE

Table 2. ATTENUATION LENGTH AND PHOTOELECTRIC FRACTION FOR COMMON SHIELDING AND COLLIMATION MATERIALS Density (g/cm3)

Material (Z)

Attenuation Length (cm)/Photoelectric Fraction (%) 30 keV

140 keV

171 keV

245 keV

364 keV

Molybdenum (42)

10.28

0.004

97.6

0.20

68.4

0.30

57.7

0.55

36.8

0.86

18.8

Lead (82)

11.34

0.003

95.2

0.04

91.0

0.06

87.9

0.14

79.4

0.32

65.0

Tungsten (74)

19.25

0.002

94.3

0.03

89.4

0.04

85.5

0.10

75.0

0.23

58.1

Gold (79)

19.30

0.002

94.9

0.02

90.5

0.04

87.1

0.09

77.9

0.20

62.6

Uranium (92)

18.90

0.001

95.9

0.02

92.5

0.03

90.0

0.06

83.4

0.16

77.2

photons have been absorbed or deflected from their initial trajectory. At 30 keV, attenuation lengths range from 0.02 to 0.04 mm and collimators allo wing high spatial resolution (< 100 µm FWHM) can be f abricated. At 140 k eV, attenuation length has already increased b y an order of magnitude. More material is needed for ef fective absorption, and collimator feature size must generall y be larger making high resolution more dif ficult to achie ve. This becomes particularly evident at 364 keV. Since attenuation occurs w hen a photon has been either completely absorbed or deflected from its original path, attenuation length is not the w hole stor y. Photons with energies less than 1.022 MeV undergo three significant interactions in matter: photoelectric absor ption in which the photon is completel y absorbed, and Compton or coherent scattering in w hich the direction of the incident photon is merel y changed. 72 Obviously, photoelectric absorption is the most desirable interaction from the collimation vie wpoint. Scattered photons ma y be deflected a way from the camera, ma y be absorbed elsewhere in the collimator , or ma y escape and be detected by the camera, reducing the effectiveness of the collimation. High proton number (Z) materials and lo w energies f avor photoelectric absor ption. As sho wn in Table 2, the fraction of photoelectric interactions decreases rapidly with increasing energy even for high-Z collimator materials. Because of the influence of absorbing material on collimator perfor mance, in par ticular, on the ability to achieve high spatial resolution, man y small animal SPECT instruments have explored use of tungsten, gold, and even depleted uranium in contrast to the lead alloys commonly used in clinical SPECT collimation. 73–75 Parallel-Hole Collimator

Channel collimation, and specif ically the parallel-hole collimator, is presently the most common image formation

aperture in clinical nuclear medicine.These collimators are often used in animal SPECT when good performance over a lar ge FOV is desirab le and high spatial resolution is a secondary consideration. As the name suggests, the collimator consists of a large number of open parallel channels having w alls constr ucted of a suitab le γ-ray absorber . Photons traveling in nearly the same direction as the orientation of the channel will be detected , w hereas those outside the acceptance angle will be absorbed by the channel w alls. F igure 11 sho ws the constr uction of a typical collimator from cor rugated lead sheets as w ell as def initions of parameters used in the e xpressions below. The combination of detector and collimator resolution on the PSF is given by the following expression with the f irst term inside the square-root being the ef fect due to the collimator itself. Note that the collimator resolution depends on the ratio of the hole diameter d to channel length l: d2

RT ≈

l2

( l + z ) 2 + RD2 ,

where additionally z is the distance from the collimator face to a parallel plane containing the source and RD the detector resolution (PSF). The counting ef ficiency (the number of photons detected/those emitted from a point) is as follo ws:

⎛ d 2 ⎞ , η ≈ K ( d / l ) ⎜ 2 ⎟ ⎝ ( d + t ) ⎠ 2

2

where t is the thickness of the septa and K is a normalizing factor depending on collimator hole shape (~0.26 for hexagonal holes in a he x array).8 For collimators used in clinical SPECT at 140 k eV, typical ef ficiency is 2 to 3 photons detected for e very 10,000 emitted (2–3 × 10−4). Significance of collimation ef ficiency will become more apparent in the discussion of noise in SPECT.

Small Animal SPECT, SPECT/CT, and SPECT/MRI

85

De te ctor l

t

d

Figure 11. Typical parallel-hole collimator. Left: parallel-hole collimator principle and definitions. Photons within an acceptance angle determined by collimator parameters will pass through the channels to be detected by the camera, whereas those outside the acceptance angle will be absorbed by the channel. Right: photograph of a section of a parallel-hole collimator showing the channels. Given the parameters written on the collimator face (in inches), spatial resolution is ~8 mm FWHM at 10 cm.

The important features to note regarding parallel-hole collimation are that (1) resolution becomes w orse with increasing distance from the collimator , (2) collimator efficiency is constant ir respective of distance, (3) ef ficiency is related to the square of the PSF width (ie, doubling spatial resolution decreases ef ficiency b y four times), and (4) perfor mance of parallel-hole collimation (and absorbing collimation in general) decreases rapidly with increasing ener gy due to the rapid decrease in absorber performance shown in Table 2. For example, at 1.8% septal penetration, w hich is near the upper end of acceptable, a collimator for 140 k eV can achie ve 5 mm resolution at 10 cm with an ef ficiency of 6.7 × 10−5. To achieve the same resolution at 364 k eV, the ef ficiency drops to 1.0 × 10−5 due to the thick er septa required to maintain the same penetration. A good introduction to parallel-hole collimation is given by Gunter.76 Designing specialized collimators for specific tasks is not out of question for investigators using small animal SPECT. A useful Web-based calculator for collimator design is a vailable on the Nuclear Fields Web site (http://www.nuclearfields.com).

Composite resolution for the pinhole and camera (again, as measured through the PSF width) is gi ven by the following:

Pinhole Aperture

There are several important things to note. F irst, the efficiency depends on the distance of the source point to the pinhole—it drops in inverse square law fashion as the distance increases. And when viewed at an angle θ, the apparent opening of a pinhole aper ture g(θ ) decreases with cor responding loss of ef ficiency. Ev en in the best case, where there is no additional vignetting due to the pinhole geometr y (e g, the k eel thickness in F igure 12), efficiency f alls as cos 3θ. Second, spatial resolution also becomes worse with increasing distance. Attempting to improve resolution b y using a smaller pinhole results in

Although channel collimators are most common in clinical SPECT, the simplest and the most often used collimator in small animal SPECT at present is a “pinhole” aperture in a γ-ray absorbing material. 77 Figure 12 shows a photo graph of a lead pinhole collimator along with definitions of parameters used in the e xpressions for resolution and ef ficiency. Analogous to a camera lens, the image of the object on the detector is in verted b y the pinhole.

2

2

⎛ z + z A ⎞ 2 ⎛ z ⎞ 2 RT ≈ ⎜ deff + ⎜ ⎟ RD , ⎟ ⎝ z A ⎠ ⎝ z A ⎠ in which zA is the distance from the pinhole to the detector, z is the distance from the pinhole to a parallel plane containing the point in the object, and deff is the effective pinhole diameter.78,79 As noted abo ve, absorbers are not perfect so the actual size is some what lar ger than the physical opening and also depends on the absorbing material and actual pinhole shape. A great deal of w ork has been done on pinhole designs for v arious applications and more detail can be found in the referenced work.80–86 The ef ficiency of a simple pinhole aper ture is as follows: η≈

deff2 16 z

2

g (θ).

86

MOLECULAR IMAGING: PRINCIPLES AND PRACTICE

De te ctor ZA d e ff

P inhole

Z θ

S ource Figure 12. Pinhole Collimator. Left: pinhole aperture and definitions used in text. Photons emitted from the source either pass through the open aperture or absorbed. Optimum pinhole geometry depends on γ-ray energy. To obtain small effective diameters, pinholes often have a “keel” instead of the absorber thickness decreasing to a knife-edge. Right: example of a simple pinhole aperture in a lead absorber.

loss of ef ficiency proportional to the square of the resolution. Finally, some good ne ws for small animal imaging: effects of the intrinsic resolution RD of the radiation imaging detector on the image resolution can be minimized b y using a lar ge magnif ication ( zA/z). Ne vertheless, in Section “P erformance Optimization and Design Trends,” it is sho wn that signif icant perfor mance increases can be obtained if a high spatial resolution detector is used with a pinhole in conjunction with demagnification (zA < z).

Image Degradations Two signif icant image distor tions were noted above. As a by-product of nonzero ef ficiency, collimation methods introduce an uncertainty in the arrival direction of detected γ-rays resulting in f inite spatial resolution or b lurring of reconstructed images. Fur ther uncer tainty introduced b y the camera adds in quadrature to the collimator resolution. In addition to this b lurring, there are also de gradations introduced by the randomness of radioactive decay and by interactions of the emitted x- or γ-ray photons with the object. This is of course tr ue in PET as well as SPECT. Photon Attenuation

In the case of X-ra y CT, photon attenuation b y the object is important because it encodes the desired signal. Attenuation by the object also occurs in emission tomo graphy, but instead of providing a useful signal, it becomes a distortion that if not cor rected introduces systematic quantitative inaccuracies in reconstr ucted images. For example, reconstructions from a uniformly emitting cylindric object would appear hollo wed out or cupped instead of ha ving uniform intensity.

Although photon attenuation is a signif icant issue in clinical SPECT imaging, it cannot necessarily be ignored in small animal SPECT especiall y if high quantitati ve accuracy is desired. F or e xample, at 140 k eV, the attenuation length of photons in soft tissue is appro ximately 6.5 cm— 63% of the photons will be either scattered or absorbed from this depth. At 1 and 2 cm, the fractions are 14 and 26%, respectively. At 30 k eV, attenuation length decreases to 2.6 cm. The cor responding fractions absorbed or scattered at 1 and 2 cm are 32 and 54%. The effect of attenuation can be compensated in the reconstr uction but requires knowledge of the attenuation distribution.Typically, this can be provided by transforming the map of linear attenuation measured by X-ray CT to the appropriate energy.2,87 Instruments capable of both SPECT and CT simplify matters, but correction can also be accomplished b y appropriately fusing information acquired on separate instr uments. Compton Scatter

As noted, attenuation occurs an y time a photon is either deflected from its initial trajectory or absorbed. As it turns out, attenuation in tissue at 140 k eV and abo ve is almost entirely due to Compton scattering. A Compton interaction is an inelastic process in which the scattered photon always has lower energy. The energy E of the scattered photon as a function of incident ener gy E0 in k eV and scattering angle θ is given by the following: E0

E= 1+

E0 511

.

(1 − cos θ)

A signif icant fraction of scattered photons escape the object and interact in the camera. If these scattered photons

Small Animal SPECT, SPECT/CT, and SPECT/MRI

are used indiscriminatel y in the reconstr uction, additional blurring and loss of contrast—especiall y in small cold regions—will result. However, if the detector is capab le of measuring the ener gy of each detected photon, it can discriminate against those that ha ve Compton scattered. F or example, a typical SPECT study at 140 keV with an Anger camera having 10% FWHM energy resolution might be set up to record onl y those e vents in a ±10% energy window centered at 140 keV. How effective this rejection is in practice depends on both the energy resolution of the detector and the emission energies of the radionuclides. The higher the emission energy, the larger the energy range over which the scattered photon spectrum is spread. For example, photons that have scattered at 90° would lose 1.7, 30, and 151 k eV for incident energies of 30, 140, and 364 k eV, respectively. Thus, although it is relati vely easy to use ener gy to ef fectively discriminate against scatter at 140 k eV and abo ve, it is nearly impossib le for 125I because of its closel y spaced emission energies (see Table 1) and the small ener gy loss with each scatter. Fortunately, at all but the lowest energies of interest, attenuation and Compton scatter in small animals such as mice and rats—w hile not negligible—are small.

Effects of the abo ve distor tions, f inite spatial resolution, attenuation, and Compton scatter can, in principle, be reduced—even eliminated in the limit—b y using an image reconstr uction technique that appropriatel y accounts for these de gradations. Ne vertheless, the “monkey wrench” limiting the ef ficacy of these recovery methods is counting noise in the projection measurements. Radioactive deca y is a random process. Assuming that the half-life is long with respect to the obser vation time, the probability of the emission of a given number of events in a time inter val of duration T is gi ven b y the Poisson distribution: P ( N = n | λ, T ) =

e − λT ( λT ) n n!

,

where λ is the rate of emission. Randoml y selecting emitted γ-rays via a collimator , for e xample, results in another P oisson process but with reduced rate gi ven by η × λ. The mean and v ariance in the number of obser ved events (measured through the collimator) are given by the following:

2 and σ ( N ) = ηλT ,

N = ηλT

respectively, where the angle brack ets denote the e xpectation operation. Note that both the mean number of e vents recorded and the uncer tainty in the number of e vents increase over time. Moreo ver, the v ariance in the number of events is equal to the mean. Rather than the ra w number of recorded e vents, however, in SPECT w e are more typicall y interested in the underlying concentration of radiotracer independent of the measurement time and instr ument efficiency. An appropriate estimator for the rate of photon detection in each projection measurement is as follows: ˆλ ( N ) = N . ηT Calculating the mean and v ariance of the estimated rate, we find that λˆ ( N ) =

N ηT

= λ,

ie, this simple estimator of rate is unbiased and that σ λˆ ( N ) = 2

Counting Noise

87

)

(

σ2 (N ) 2

ηT

2

=

λ ηT

.

The uncer tainty in emission rate estimate decreases with increasing observation time or collimation efficiency. To reduce the standard deviation by half, the product of the observation inter val and collimation ef ficiency must be four times g reater. More details on counting noise can be found in Knoll 72 and Barrett.88

Accounting for Degradations: Advanced SPECT Reconstruction If we assume that the three-dimensional distribution of radiotracer in the object can be represented adequatel y by a f inite number of basis functions—cubic v oxels (volume pictur e elements ) ha ving scale coef ficients λ1,...,λb, for e xample—then the lik elihood of obser ving the measurements—the number of counts yd in each detector channel—given that each is P oisson distributed and statistically independent is gi ven by the product of the individual probabilities, viz: P ([ y1 ,..., yD ] | [ λ1 ,..., λ B ]) T∑ a λ ) ( =∏ D

d =1

b

db

b

yd

(

exp −T ∑ b adb λ b yd !

),

88

MOLECULAR IMAGING: PRINCIPLES AND PRACTICE

where the coef ficients adb quantify the probability that a photon emitted from the bth v oxel in the representation of the object is detected in the dth detector channel. Note that the set of coefficients {adb}, often referred to as a system response matrix A, can model man y of the nonideal characteristics of the imaging process including photon attenuation, Compton scatter, and finite spatial resolution, while the for m of the lik elihood function itself pro vides appropriate handling of the kno wn statistical characteristics of the measurements. 89,90 Conceptually, the optimum image is obtained b y choosing non-ne gative v oxel v alues λ1,...,λb that maximize the likelihood of observing the projection measurements y1,...,yD. The solution is usually obtained iteratively using variations of the Expectation-Maximization (EM) algorithm. The EM algorithm for emission tomo graphy introduced by Shepp and Vardi91 and Lange and Carson92 is as follows: λˆ (bk +1) =

λˆ (bk )

adb yd , ∑ (k ) ˆ T ∑ adb d = 1 ∑ adb λ b d

D

b

with superscripts indicating the iteration number for each voxel. Since the abo ve algorithm is slo w to con verge, numerous adaptations have been developed. The ordered subsets expectation-maximization (OSEM) algorithm exhibits faster convergence and is often used in small animal PET and SPECT applications. 93 Despite capabilities for accuratel y modeling the SPECT measurement process, maximizing the ra w likelihood seldom produces a desirab le reconstr uction. Reconstruction generall y entails a trade-of f betw een accuracy, e g, resolution or bias, and precision or v ariance in the estimated voxel values. Maximum likelihood estimation stri ves for perfect resolution or zero bias. Reconstruction error, however, is a function of both variance and bias (and often higher order moments of the error distribution) and is generall y not minimum when bias is zero. P oisson counting noise in the projection measurements introduces v ery lar ge er rors in reconstructed images as one attempts to reco ver resolution or force the bias toward zero. There is an inherent trade-off between bias and v ariance or similarl y between resolution and noise. To limit the v ariance, the solution needs to be regularized, which can be accomplished b y stopping EM or OSEM iterations well short of convergence, through the use of penalty functions,94 the similar Bayesian methods,95 or b y post-smoothing the ra w maximum lik elihood

reconstruction by a desired PSF. Regularization typically leads to an image that has better f idelity with respect to the underlying tracer distribution. Ne vertheless, the method used depends on the purpose for which the images will be used. As noted in the ne xt section, measures of image quality and perfor mance e valuation are f ar from solved problems. Success in reco vering spatial resolution as well as in reducing the ef fects of Compton scatter and attenuation also depends on ho w accurately the imaging system is modeled. Pinhole SPECT , in par ticular, has received extensive attention in this regard.78,96–108

PERFORMANCE OPTIMIZATION AND DESIGN TRENDS Performance Optimization In contrast to PET, SPECT offers many possibilities for optimizing perfor mance through selection of dif ferent FOV sizes, collimators, and detectors. Ne vertheless, performance optimization is a comple x issue requiring the definition of a measure of perfor mance relative to a specific task. For example, is the task one of detecting the presence or absence of a lesion? Or rather, is the task one of quantifying the amount of uptak e in a specif ic region? What is the anticipated distrib ution of radiotracer and how will it dif fer among the animals used in the study? For the case of lesion detection, detectability may be the appropriate measure or perhaps recei ver operating characteristic cur ves that quantify inherent trade-off between false-positives and missed lesions. On the other hand, for quantification tasks some measure of the error is likely more appropriate. As noted in Section “Accounting for De gradations: Advanced SPECT Reconstruction,” er ror can generall y be thought of as being composed of systematic er rors or bias, which is due to f inite spatial resolution, for e xample, and variance, which is introduced both by counting noise and by variations among animals. Mean-squared er ror is the sum of the variance and the squared bias, but other measures such as maximum er ror or mean absolute er ror may be more appropriate depending on the task. To further complicate matters, it is often difficult to categorize the imaging task so con veniently as pure detection or pure quantif ication. After all, reconstr ucted images are typically used to answer a variety of questions. Optimization of medical imaging system performance remains an acti ve research area. A good overview of the rele vant issues can be found in an article b y F ryback and Thornbury109 and an e xcellent

Small Animal SPECT, SPECT/CT, and SPECT/MRI

introduction to perfor mance e valuation methods in Report 54 of the ICRU.110 Numerous methods have been proposed for SPECT performance optimization: refer to the series of ar ticles b y Bar rett for e xamples.111–114 Alternatives ha ve also been proposed for e valuating multipinhole SPECT systems (although the y are relevant to an y SPECT aper ture).115,116 A recent ar ticle by Clarkson and colleagues describes methods for quantifying perfor mance in multimodality imaging systems.117 But ignoring the f iner issues of image quality measurement and optimization, there are some obvious routes to improving SPECT performance.

Design Trends Based on the discussion in Section “Basic Principles,” the most clear-cut path to better small animal SPECT performance is to increase the counting efficiency of the instrument while maintaining spatial resolution. At first glance, the relationship betw een resolution and sensiti vity for both pinholes and parallel-hole collimators seems to work against it—doubling resolution decreases efficiency by a f actor of four. However, note that the object can be surrounded b y multiple cameras to increase ef ficiency with no loss in resolution. This approach has of course been taken in clinical SPECT where it is common to surround the patient with tw o or three Anger cameras. The same approach is used in small animal imaging. The basic idea is to sur round the object with detectors and to fully use the available detector area by projecting as many images of the FO V as possib le onto it without o verlap. Although fan- or cone-beam channel collimators can be used, multiple pinhole aper tures mak e this technique especially straightforward. If M pinholes view the object, efficiency for a point in the FOV equidistant from all pinholes increases by a factor of M while resolution remains

89

the same. Note that as the FO V becomes smaller , more nonoverlapping images can be projected onto the detector and higher ef ficiency can be achie ved. F or e xample, Beekman and colleagues 118 use a 75-pinhole aper ture to project nono verlapping images of a small FO V onto a three-camera instrument. Of course, adding pinholes has limits: at some point, projections of the object star t to overlap on the detector . Imaging systems that allow overlapping projections have traditionally been ter med coded apertur es. As for nonmultiplexing pinhole aper tures, raw sensitivity increases by a f actor of M; however, since there is no w ambiguity associated with the pinhole that passed a γ-ray, each detection car ries less infor mation than in an unmultiplexed system. The increased sensiti vity obtained with additional multiple xing often outw eighs the additional noise resulting from “unfolding” the projection o verlap, and this is especially true if either the overlap is small or the radiotracer is onl y accumulated in re gions of the object that are small relative to the size of the FOV (“hot spots”). Indeed, allowing at least slight overlap in pinhole projections is cur rently the most popular method of increasing ef ficiency w hile maintaining resolution in small animal SPECT. Evaluating effects of multiple xing on perfor mance using statistical methods has been the subject of a number of research projects. 113,115,116 Figure 13 sho ws tw o dif ferent multiple xing apertures: one in w hich relatively few pinholes are used and another having a lar ge number in a uniformly redundant array.119 Although a number of multistep approaches have been used over the years to recover data from multiplexing aper tures, reconstr uction is best done with straightforward adaptation of the maximum lik elihood method presented in Section “ Accounting for De gradations: Advanced SPECT Reconstr uction.” The matrix A merely needs to account for the f act that for each projection measurement photons are arriving through more than

Figure 13. Two examples of multiple pinhole multiplexing apertures. Left: relatively few pinholes in a tungsten absorber. Right: a no-twoholes-touching uniformly redundant array with a large number of open apertures (Courtesy R. Accorsi, Children’s Hospital of Pennsylvania).

90

MOLECULAR IMAGING: PRINCIPLES AND PRACTICE

a single aper ture opening. A large number of multiple xing aper ture small animal SPECT systems ha ve been developed.119–122 Although multiple pinhole multiple xing aper tures are used to a g reat e xtent in cur rent small animal SPECT devices, there is an emer ging alter native. Suppose that for a gi ven FOV size and spatial resolution, enough pinholes are placed in a hemispheric shell of absorber such that projections of the FO V on a concentric hemispheric detector just touch. Reducing the radius of the detector , thereb y decreasing zA, and increasing detector resolution to compensate reduces the size of each projection and allo ws more pinholes to be inser ted such that the projections again just touch, which increases ef ficiency without additional multiplexing. But to maintain the same image resolution, the effective size—and therefore the ef ficiency—of each pinhole aper ture must also be reduced in accordance with the for mula in Section “Image De gradations.” It turns out that as zA decreases, the number of allo wable pinholes increases faster than the loss of ef ficiency resulting from smaller pinholes. Each projection of the FOV is demagnified onto a high-resolution detector . If such high-resolution detectors are a vailable, an o verall significant increase in ef ficiency at the desired resolution results. 123 Limits are imposed b y the inability to make pinholes ha ving ever smaller ef fective diameters given the ph ysical limits of absorbers and b y the f act that the high-resolution detectors ma y require the ability to deter mine the γ-ray interaction location in three dimensions rather than just two, but the idea has dri ven much of the cur rent research direction in small animal SPECT instrumentation.

Toward High-Resolution Detectors

Motivated partially by the above idea (but perhaps just as often to make a more compact SPECT instrument), a number of higher resolution, smaller alter natives to the Anger camera are under acti ve development. These can be categorized as direct and indirect conversion detectors. Scintillation cameras are indirect converters; interactions are first converted to visible light photons and then to an electrical signal using a photodetector. γ-Rays interacting in a direct conversion detector such as CZT, on the other hand, generate an electrical signal directl y. As expected, the radiation absorption characteristics of each detector material are important and are summarized in Table 3. The most common detection material used in small animal SPECT is sodium iodide followed by CZT and cesium iodide. A common alter native to the Anger camera w as enabled b y the de velopment of the position-sensiti ve photomultiplier (PSPMT) and pix elated scintillator array. In contrast to the lar ge ar ray of PMTs used in an Anger camera, w hich as a unit has nominal photodetector resolution of 75 mm for a clinical instr ument, PSPMTs typically provide < 1 mm resolution in a significantly more compact package. To achieve 3 to 4 mm spatial resolution in an Anger camera, scintillation light must be spread broadly enough that it is measured by multiple 75 mm diameter PMTs. In comparison, no light-spreading is necessar y for the PSPMT ; the pix elated scintillator merel y channels the light photons to the face of the photodetector. An example using a single large PSPMT and NaI(Tl) array with 2 × 2 mm crystals is shown in F igure 14. Designs using single and multiple PSPMTs have been extensively used in commercial

Table 3. ATTENUATION LENGTH AND PHOTOELECTRIC FRACTION FOR COMMON γ-RAY DETECTOR MATERIALS VERSUS PHOTON ENERGY Material

Attenuation Length (cm)/Photoelectric Fraction (%)

Direct detectors

30 keV

140 keV

171 keV

245 keV

364 keV

Si

0.30

80.8

2.87

5.7

3.16

3.3

3.66

1.2

4.29

< 1.0

Ge

0.014

96.4

0.67

49.6

0.92

37.2

1.40

19.3

1.90

8.4

CZT

0.008

96.3

0.24

77.6

0.39

70.0

0.78

49.8

1.36

29.1

CdTe

0.008

96.4

0.24

77.8

0.38

69.3

0.76

50.2

1.34

29.5

NaI(Tl)

0.037

89.8

0.38

77.8

0.60

69.4

1.20

50.3

2.10

29.6

CsI(Tl)

0.024

90.5

0.26

80.9

0.41

73.5

0.87

55.7

1.60

34.7

YAP:Ce

0.014

96.6

0.64

50.1

0.88

37.8

1.33

19.9

1.80

8.8

LaBr3:Ce

0.013

95.5

0.35

73.0

0.54

63.2

1.02

43.0

1.67

23.6

Scintillators

Small Animal SPECT, SPECT/CT, and SPECT/MRI

91

Figure 14. Example of a compact scintillation camera (disassembled), The camera was constructed from a 5-inch diameter position-sensitive photomultiplier and a pixelated sodium iodide scintillation array.

and research small animal SPECT designs and typically result in compact systems ha ving good performance.124–128 Spatial resolution for these systems depends to a lar ge extent on cr ystal pitch in the scintillator ar ray and is cur rently 1.5 to 2 mm FWHM w hile energy resolution is ~12% FWHM. There ha ve been recent attempts to use CsI(Tl) ar rays with cr ystal sizes as lo w as 0.2 mm separated b y a center -to-center distance of 0.4 mm coupled to a high-resolution PSPMT.129 The in vestigators repor t intrinsic detector resolution as good as 0.6 mm FWHM. One of the advantages of direct detectors as opposed to scintillation cameras is that it is relati vely straightforward to achieve high spatial resolution (at least in 2D) in thick (ie, ef ficient) detectors. Moreo ver, ener gy resolution is generall y superior to indirect con version devices, which allo ws better scatter rejection and better radiotracer separation in multiple isotope studies. The most widely in vestigated direct detector for small animal SPECT is CZT. A CZT detector ha ving ~1.6 mm square pixels and 4.5% energy resolution at 140 keV is shown in Figure 15 and is used in the commercial FLEX Triumph SPECT-PET-CT device from Gamma Medica-Ideas (see Figure 4). Investigators at the Center for Gamma-Ray Imaging at the University of Arizona have developed high-resolution CZT detectors ha ving 380 µm pix els in a 2 mm thick detector with the ultimate goal of realizing SPECT devices having high resolution and efficiency using the demagnification approach.130 Examination of Table 3 reveals that the attenuation length for CZT at 140 keV is 2.4 mm. If the incident angle of γ-rays is lar ge, as it could w ell be in a compact pinhole geometry where zA is small, detectors will

Figure 15. CZT-based SPECT detector consisting of 25 25 5 mm thick CZT modules (upper right) backed by high density low power ASIC readout electronics (upper left). These CZT modules are tiled 5 by 5 to form a 12.5 × 12.5 cm field-of-view 80 × 80 pixel compact gamma camera (bottom). Camera has ~1.6 mm spatial resolution and 4.5% FWHM energy resolution at 140 keV—a significant improvement over the 10–12% typical for scintillation cameras (Courtesy K. Iwata, Gamma Medica-Ideas).

require depth-of-interaction or 3D resolution, w hich is under development. For imaging the lo w-energy emissions from 125I, Accorsi and colleagues have developed a 1 mm thick CdTe detector having 55 µm pixels in a 256 × 256 array read out using the Medipix2 chip for use with multiplexing apertures similar to that sho wn in F igure 13. The detector has an active area of only 14 mm square so that many devices must be tiled together to construct a large detector.131 Although silicon is rarel y considered a good x- or γ-ray detector (its attenuation length at 140 keV is nearly 3 cm), it has several advantages that make it attractive for use at ~30 k eV. It has well-understood properties and is the most w ell-developed semiconductor material. Detectors have been a vailable for char ged par ticles and low-energy photons for decades. It is relati vely straightforward to construct detectors having spatial resolution of nominally 50 µm using doub le-sided or thogonal strip readout. Peterson is e xploring use of a stack of doublesided strip detectors for lo w-energy imaging. 132 Each detector is 6 cm × 6 cm × 1 mm thick with a strip pitch of 59 µm. The attenuation length of 30 k eV photons in silicon is 3 mm; therefore, a stack of detectors must be used to obtain reasonab le ef ficiency. One adv antage of

92

MOLECULAR IMAGING: PRINCIPLES AND PRACTICE

using a stack, however, is that it also pro vides depth resolution. Intrinsic silicon contains residual impurities that prevent fabrication of detectors of more than 1 to 2 mm thick; however, it is possible to compensate the effects of impurities b y drifting lithium ions into the detector thereby allo wing f abrication of thick er de vices. Use of 6 mm thick lithium-drifted silicon (SiLi) detectors having 1 × 1 mm elements has been proposed for lo w-energy SPECT imaging.133 Scintillation cameras using solid-state photodetectors instead of bulkier PMTs are also under acti ve investigation for high-resolution imaging. CsI(Tl) scintillators are typically used because of the e xcellent spectral match of the scintillation light with the sensiti vity of silicon photodetectors. As examples, a design that couples a 5-mm thick CsI(Tl) scintillator to a position-sensiti ve avalanche photodiode (APD) has been proposed and estimated resolution is 0.5 mm at 140 k eV.134 Fiorini and colleagues135,136 have coupled a continuous CsI(Tl) crystal 3-mm thick to a miniature ar ray of silicon drift photodetectors and have achieved 0.16 mm resolution with 14% energy resolution at 122 k eV although cooling is required, which can complicate the imaging system. Detection of scintillation light requires lo w-noise photodetectors such as PMTs, drift detectors, APDs, or the new silicon photomultipliers (SiPM), but it has been challenging to construct imaging arrays or position-sensitive photodetectors e xhibiting spatial resolution less than ~0.5 mm. On the other hand , standard CCD and CMOS camera chips are capable of high resolution but their noise precludes use directl y as a photodetector for scintillation light. Ho wever, a ne w de vice, the electron-multipl ying CCD (EMCCD), has become a vailable in the past fe w

years and has been incor porated into high-resolution detectors by several investigators. Like the APD or SiPM, the EMCCD has an inter nal gain mechanism that signif icantly increases the signal-to-noise ratio to the point that scintillation light is detectab le. The drawback is that the devices are small and therefore not w ell matched to FOV sizes used in small animal SPECT . Options for matching the small EMCCD to a lar ger FOV have explored lenses for coupling the scintillation light, 137 fiber optic tapers, 138 and an electronic demagnif ication tube.139 Research Instruments

It is impractical within the scope of this chapter to cover all devices used in small animal SPECT or e ven any in detail. Se veral commercial de vices w ere mentioned above, and w e highlight three research instr uments below. Refer to the current literature for more details on these and other instr uments while noting that dif ferent imaging applications often benef it from v ery dif ferent SPECT designs. 68,99,118,121,122,125,127,130,133,134,139–154 The FastSPECT II de vice under de velopment at the University of Arizona consists of a circular ar rangement of modular scintillation cameras (F igure 16). Each camera is similar to an Anger camera in that it comprises a continuous NaI(Tl) scintillator coupled to 9 PMTs. P osition resolution is some what better than clinical Anger cameras, but a key feature is that good position resolution can be attained close to the edges of each detector . The conventional Anger camera has a lar ge dead area around the edges, resulting in loss of ef ficiency w hen pack ed into ar rays. In contrast, the modular cameras can be close-packed with little efficiency loss.

Figure 16. The FastSPECT II system developed at the University of Arizona. Left: complete unit undergoing calibration. Right: partially assembled instrument showing the arrangement of modular scintillation cameras (Courtesy L. Furenlid, U. Arizona).

Small Animal SPECT, SPECT/CT, and SPECT/MRI

The main feature of F astSPECT II (and with an increasing number of SPECT instr uments) is that the object is viewed from many directions simultaneously by a large number of pinhole apertures (similar to the U-SPECT II, see F igure 5). The simultaneous vie ws satisfy the 3D sampling requirements discussed abo ve (at least for moderate spatial resolution) so that rapidl y changing dynamic processes can be easily followed as in PET. More conventional small animal SPECT instr uments that require translation or rotation of the aper ture do not collect complete data at an y time instant. Although following slowly varying dynamic processes is possible, imaging rapidly changing distributions can be prob lematic with more conventional devices. Meng and colleagues have developed a SPEM prototype with inte grated CT for imaging 125I in mice and in ex vivo specimens.155 The instrument, shown in Figure 17, has two SPEM detectors each consisting of a multipinhole apertures and thin CsI(Tl) scintillator coupled to an EMCCD camera through an electronic demagnif ication tube. To acquire a complete set of SPECT data, the specimen is placed on a rotating stage betw een the detectors. An X-ray source and amorphous silicon X-ray detector are used to acquire simultaneous cone-beam CT images. Spatial resolution for 125I is less than 200 µm, and efficiency in full system can be as high as 1× 10−3. The prototype has been used for tracking radiolabeled T-cells injected into the brain of a mouse. Finally, both SPECT and PET can require an hour or more of imaging time, and in order to minimize motion, animals must typicall y be anesthetized or otherwise restrained. Ho wever, both can alter the functional

93

processes measured with SPECT and this can be especially important in CNS imaging. A team of in vestigators has developed a SPECT instrument that can be used for imaging awake, unrestrained mice. 156 Designed for brain imaging, the de vice uses an IR camera tracking system with retroreflecting spheres attached to the mouse’ s head (Figure 18). P osition and orientation are track ed throughout the scan and used to reconstr uct the acquired data into a reference frame that can then be fused with a CT scan acquired using the same instrument. To limit gross motion, the instrument also incor porates a “bur row” in w hich the mouse feels comfortable.

SPECT/CT AND SPECT/MR SPECT/CT The previous discussions have focused on design of the SPECT portion of a SPECT/CT or SPECT/MRI instrument because at least for SPECT/CT, this is the par t of the instrument that will vary most widely among applications. Adding CT capability to a SPECT instr ument is straightforw ard and is typicall y accomplished b y adding a fixed-anode X-ray tube and appropriate X-ray detector. Operating characteristics for the X-ray source are nominally from 50 to 90 kV and 1 to 5 mA depending on the size of the animal and application. Typically, a 2D CMOS or amor phous silicon flat-panel X -ray detector is used to collect a set of cone-beam projections that are usually reconstructed using the Feldkamp reconstruction63 or v ariations that more appropriatel y handle the incomplete data resulting from the circular orbit. The small cone-angle associated with most CT imaging geometries for small animals signif icantly reduces the sampling ar tifacts noted abo ve resulting from use of a circular source trajector y. Numerous commercial and research instr uments ha ve incor porated CT with SPECT.145,146,157–159

SPECT/MR

Figure 17. A prototype single photon emission microscopy device with integrated X-ray CT. Two electron-multiplying CCD cameras, electronic demagnification tubes, scintillators, and multipinhole apertures (left and right) view a mouse-sized field-of-view. Complete sampling is obtained by placing the mouse on a rotating stage. X-ray tube for CT is located at the bottom of the figure, whereas an amorphous silicon X-ray detector is located at the top.

Performing simultaneous MRI and SPECT is considerably more challenging due to mutual interference betw een the MRI and SPECT hardw are. F irst, electric char ges (eg, electrons) mo ving in a magnetic f ield are subject to a Lorentz force that alters their trajector y making devices with such con ventional PMTs dif ficult—if not impossible—to use in the high static MRI f ield. Ev en direct detectors such as CZT e xperience depth-dependent positioning errors if the electric drift-f ield of the detector is not aligned with the magnetic f ield direction. Another

94

MOLECULAR IMAGING: PRINCIPLES AND PRACTICE

Figure 18. SPECT/CT instrument for awake, unrestrained mice. Left: SPECT/CT instrument for brain imaging in conscious mice. In addition to SPECT and CT, instrument has infrared-tracking cameras to continuously record position and orientation of the mouse’s head. Right: awake mouse is shown in burrow with retroreflecting targets for IR light (Courtesy D. Weisenberger, Thomas Jefferson National Accelerator Facility).

problem is interference of the g radient switching and RF sequences with the sensiti ve SPECT electronics. Much of this issue can lik ely be resolv ed with appropriate electromagnetic shielding. F inally, the SPECT hardw are—especially the EMI shielding—can interfere with MRI acquisition. This can be minimized b y appropriate design that places RF coils within the bore of the SPECT de vice rather than vice v ersa. As for PET , de velopment of SPECT/MR instruments is an active research area. A CZT SPECT detector inser t for an instr ument under de velopment in a collaboration between investigators at John Hopkins Uni versity and Gamma Medica-Ideas is sho wn in Figure 19. Other research SPECT/MR systems ha ve been developed,160 but as for PET/MR, the field is in its infancy and numerous advances will be made in the coming years.

FUTURE DIRECTIONS Radiation Sensors for Better Integration with Other Imaging Modalities Although the potential of the combination of SPECT and other modalities has been well demonstrated, development of multimodality instr umentation has reall y onl y just begun. Cur rent emphasis has been more on assemb ling multimodal de vices that demonstrate proof-of-concept rather than on achie ving the optimum perfor mance. Taking combined SPECT/MR systems as an e xample, the ideal detectors should be immune to the magnetic f ield induced distor tion and ha ve minimal interference to MR data acquisition. The detectors should also ha ve a high spatial resolution and at least moderate ener gy resolution allowing better inte gration of SPECT infor mation with MR. Given the limited space available inside MR systems, compact γ-ray sensors w ould allow for a more f avorable imaging geometr y for SPECT . As noted abo ve, much research ef fort is no w focused on de velopment of compact, high-resolution SPECT detectors.

Figure 19. MR-compatible SPECT prototype based on rings of eight CZT modules arranged in an octagonal configuration around a sleeve-collimator/RF coil. This system can acquire MR and SPECT images simultaneously with SPECT acquired without any movement (as in FastSPECT II). In the photo above, three rings are populated with 24 modules working as independent cameras. When four rings are populated (32 modules total), the field-of-view will cover an entire mouse. The outer diameter of the system is 12.0 cm to fit standard preclinical MR systems (Courtesy K. Iwata, Gamma Medica-Ideas).

Adaptive SPECT Imaging In addition to adv anced detectors, future SPECT systems will also benef it from adapti ve data acquisition schemes that maximize the ef ficiency for collecting infor mation relevant to a desired imaging task. Ideall y, an adapti ve SPECT system could alter its measurement geometr y in real time based on the imaging infor mation that has been collected and on the given task specified by the user.161 This would help both to of fset the intrinsicall y lo w detection efficiency of SPECT and to mak e aperture selection more robust with respect to the underl ying object. Adaptive imaging has been widel y proposed or demonstrated in

Small Animal SPECT, SPECT/CT, and SPECT/MRI

astronomic,162–164 ultrasound,165,166 and MRI imaging applications.167,168 Recent ef forts ha ve focused on the development of a mathematic framework161 and a prototype system for demonstrating the concept in SPECT .169 To realize the potential of this approach, future in vestigations are needed in the de velopment of (1) meaningful methods for task-based image quality assessment that could be used to guide the self-adaptive process, (2) rapid computation for both reconstruction and evaluation of performance indices, and (3) adaptive detection hardware and associated control mechanisms.

CONCLUSIONS In contrast to clinical SPECT , small animal SPECT is a rapidly expanding field buoyed by the number of radiolabeled compounds a vailable and rapid adv ances in instr umentation. The relati ve ease with w hich SPECT can be incorporated into the biomedical research laborator y is a significant adv antage. Although the f ield of combined SPECT/MR instr uments is still y oung, SPECT/CT is a well-developed technolo gy with se veral instr uments available commercially. In the coming years, expect to see continued expansion of small animal SPECT applications with new and e xisting radiotracers and with commercial availability of advanced concepts such as SPEM, adaptive SPECT apertures, and combined SPECT/MR systems.

REFERENCES 1. Del Guerra A, Belcari N. State-of-the-art of PET, SPECT and CT for small animal imaging. Nucl Instr um Methods Ph ys Res A 2007;583:119–24. 2. Hasegawa BH, Iw ata K, Wong KH, et al. Dual-modality Imaging of function and physiology. Acad Radiol 2002;9:1305–21. 3. King MA, Pretorius PH, F arncombe T, Beekman FJ. Introduction to the ph ysics of molecular imaging with radioacti ve tracers in small animals. J Cell Biochem 2002;221–30. 4. Madsen MT . Recent adv ances in SPECT imaging. J Nucl Med 2007;48:661–73. 5. Meikle SR, K ench P, Banati RB , Kassiou M. Small animal SPECT and its place in the matrix of molecular imaging technolo gies. Phys Med Biol 2005;50:R45–61. 6. Peremans K, Cornelissen B, Van den Bossche B, et al. A review of small animal imaging planar and pinhole spect Gamma camera imaging. Vet Radiol Ultrasound 2005;46:162–70. 7. Kupinski MA, Bar rett HH, editors. Small-animal SPECT imaging. New York: Springer; 2005. 8. Cherry SR, Sorenson J , Phelps M. Ph ysics in nuclear medicine. 3rd ed. Philadelphia (PA): Saunders; 2003. 9. Wernick MN, Aarsvold JN. Emission tomography: the fundamentals of PET and SPECT. New York: Academic Press; 2004. 10. Keyes JW, Leonard PF, Brody SL, et al. Myocardial infart quantification in the dog by single photon emission computed tomo graphy. Circulation 1978;58:227–32. 11. Jaszczak RJ, Murphy PH, Huard D , Burdine JA. Radionuclide emission tomography of the head with 99m-Tc and a scintillation camera. J Nucl Med 1977;18:373–80.

95

12. Rogers WL, Koral KF, Mayans R, et al. Coded-aper ture imaging of the heart. J Nucl Med 1980;21:371–8. 13. Rogers WL, Clinthor ne NH, Stamos J , et al. SPRINT : a stationar y detector single photon ring tomo graph for brain imaging. IEEE Trans Med Imaging 1982;1:63–8. 14. Rogers WL, Clinthor ne NH, Shao L, et al. SPRINT II: a second generation single photon ring tomo graph. IEEE Trans Med Imaging 1988;4:291–7. 15. Wrobel M. Slit Aperture SPECT and Simplified in vitro Methods for the Dosimetry of Positron Emitting Radiotracers. School of Public Health [Ph.D disser tation], Ann Arbor (MI): Uni versity of Michigan; 1988. 16. Jaszczak RJ, Li J, Wang H, et al. Pinhole collimation for ultra-highresolution, small f ield-of-view SPECT . Ph ys Med Biol 1994; 39:425–37. 17. Weber D A, Iv anovic M, F ranceschi D , et al. Pinhole SPECT : an approach to in vi vo high resolution SPECT imaging in small laboratory animals. J Nucl Med 1994;35:342–8. 18. Acton PD, Choi SR, Plossl K, K ung HF. Quantification of dopamine transporters in the mouse brain using ultra-high resolution singlephoton emission tomography. Eur J Nucl Med Mol Imaging 2002; 29:691–8. 19. Andringa G, Dr ukarch B, Bol, JGJM, et al. Pinhole SPECT imaging of dopamine transpor ters cor relates with dopamine transpor ter immunohistochemical anal ysis in the MPTP mouse model of Parkinson’s disease. Neuroimage 2005;26:1150–8. 20. Blankenberg FG, Kalin yak J , Liu LP , et al. Tc-99-HYNIC-annexin VSPECT imaging of acute stroke and its response to neuroprotective therapy with anti-F as ligand antibody . Eur J Nucl Med Mol Imaging 2006;33:566–74. 21. de Win MML, de Jeu RAM, de Br uin K, et al. Validity of in vi vo [I-123]beta-CIT SPECT in detecting MDMA-induced neuroto xicity in rats. Eur Neuropsychophar macol 2004;14:185–9. 22. Garcia-Alloza M, Bacskai BJ . Techniques for brain imaging in vi vo. Neuromolecular Med 2004;6:65–78. 23. Hwang LC, Chang CJ , Liu HH, et al. Imaging the a vailability of serotonin transpor ter in rat brain with I-123-AD AM and small-animal SPECT. Nucl Med Commun 2007;28:615–21. 24. Nikolaus S, Antke C, Beu M, et al. In-vi vo quantif ication of dosedependent dopamine transpor ter b lockade in the rat striatum with small animal SPECT. Nucl Med Commun 2007;28:207–13. 25. Nikolaus S, Larisch R, Beu M, et al. In vestigating the dopaminer gic synapse in vivo. II. Molecular Imaging studies in small laboratory animals. Rev Neurosci 2007;18:473–504. 26. Nikolaus S, Larisch R, Wirrwar A, et al. [I-123]Iodobenzamide binding to the rat dopamine D-2 receptor in competition with haloperidol and endogenous dopamine—an in vi vo imaging study with a dedicated small animal SPECT . Eur J Nucl Med Mol Imaging 2005;32:1305–10. 27. Nikolaus S, Wirrwar A, Antke C, et al. Quantitation of dopamine transporter blockade by methylphenidate: f irst in vi vo investigation using [I-123]FP-CIT and a dedicated small animal SPECT . Eur J Nucl Med Mol Imaging 2005;32:308–13. 28. Pissarek MB, Oros-Peusquens AM, Schramm NU . Challenge b y the murine brain: multi-pinhole SPECT of I-123-labelled pharmaceuticals. J Neurosci Methods 2008;168:282–92. 29. Scherfler C, Donnemiller E, Schock e M, et al. Ev aluation of striatal dopamine transporter function in rats by in vivo beta-[I-123]CIT pinhole SPECT. Neuroimage 2002;17:128–41. 30. Scherfler C, Scholz SW , Donnemiller E, et al. Ev aluation of [I-123]IBZM pinhole SPECT for the detection of striatal dopamine D-2 receptor availability in rats. Neuroimage 2005;24:822–31. 31. Seo Y, Gao D W, Hase gawa BH, et al. Rodent brain imaging with SPECT/CT. Med Phys 2007;34:1217–20. 32. Wall JS, Kennel SJ, Paulus M, et al. Radioimaging of light chain amyloid with a fibril-reactive monoclonal antibody. J Nucl Med 2006; 47:2016–24.

96

MOLECULAR IMAGING: PRINCIPLES AND PRACTICE

33. Acton PD, Thomas D, Zhou R. Quantitati ve imaging of m yocardial infarct in rats with high resolution pinhole SPECT . Int J Cardiovasc Imaging 2006;22:429–34. 34. Constantinesco A, Choquet P, Monassier L, et al. Assessment of left ventricular perfusion, v olumes, and motion in mice using pinhole gated SPECT. J Nucl Med 2005;46:1005–11. 35. Liu ZL, Bar rett HH, Ste venson GD, et al. High-resolution imaging with Tc-99m-glucarate for assessing myocardial injury in rat heart models exposed to dif ferent durations of ischemia with reperfusion. J Nucl Med 2004;45:1251–9. 36. Terrovitis J, Kwok KF, Lautamaki R, et al. Feasibility of in vivo cardiac stem cell tracking, by SPECT and PET imaging, using the sodiumiodide symporter as reporter gene. Circulation 2007;116:374. 37. Thomas D, Bal H, Arkles J, et al. Noninvasive assessment of myocardial viability in a small animal model: comparison of MRI, SPECT, and PET. Magn Reson Med 2008;59:252–9. 38. Vanhove C, Lahoutte T, Defrise M, et al. Reproducibility of left v entricular volume and ejection fraction measurements in rat using pinhole gated SPECT. Eur J Nucl Med Mol Imaging 2005;32:211–20. 39. Wu MC, Hase gawa BH, Dae MW. Performance evaluation of a pinhole SPECT system for m yocardial perfusion imaging of mice. Med Phys 2002;29:2830–9. 40. Zhou R, Thomas DH, Qiao H, et al. In vi vo detection of stem cells grafted in infarcted rat myocardium. J Nucl Med 2005;46:816–22. 41. Carlson SK, Classic KL, Hadac EM, et al. In vi vo quantitation of intratumoral radioisotope uptake using micro-single photon emission computed tomo graphy/computed tomography. Mol Imaging Biol 2006;8:324–32. 42. Mathe D, Foldes I. Visualization and quantif ication of mouse thyroid gland 99mTc and 131I-uptake by a small animal SPECT/CT system. Eur J Nucl Med Mol Imaging 2007;34:S398–S398. 43. Mueller C, Forrer F, Schibli R, et al. SPECT study of folate receptor-positive malignant and nor mal tissues in mice using a no vel Tc-99mradiofolate. J Nucl Med 2008;49:310–7. 44. Muller C, Hohn A, Schubiger PA, Schibli R. Pre-clinical evaluation of novel organometallic Tc-99m-folate and Tc-99m-pteroate radiotracers for folate receptor -positive tumour tar geting. Eur J Nucl Med Mol Imaging 2006;33:1007–16. 45. Schipper ML, Riese CGU, Seitz S, et al. Efficacy of Tc-99m pertechnetate and I-131 radioisotope therapy in sodium/iodide symporter (NIS)-expressing neuroendocrine tumors in vivo. Eur J Nucl Med Mol Imaging 2007;34:638–50. 46. Hildebrandt IJ , Gambhir SS. Molecular imaging applications for immunology. Clin Immunol 2004;111:210–24. 47. Kanwar B, Gao D-w, Pan M-h, et al. Pinhole SPECT imaging and quantification of intestinal Cd4+ T cells in the de xtran sulf ate sodium model of murine colitis. Gastroenterology 2007;132:A169–70. 48. Pittet MJ, Grimm J, Berger CR, et al. In vivo imaging of T cell delivery to tumors after adopti ve transfer therapy. Proc Natl Acad Sci U S A 2007;104:12457–61. 49. Scopinaro F, Paschali E, Di Santo G, et al. Bombesin receptors and transplanted stem cells in rat brain: high-resolution scan with Tc99m BN1.1. Nucl Instrum Methods Phys Res A 2006;569:525–8. 50. Cisbani E, Cusanno F , Garibaldi F , et al. Molecular imaging with radionuclides, a po werful technique for studying biolo gical processes in vi vo. Nucl Instr um Methods Ph ys Res A 2007; 571:169–72. 51. de Graaf W, Vetelainen RL, de Br uin K, et al. Tc-99m-GSA scintigraphy with SPECT for assessment of hepatic function and functional volume during li ver regeneration in a rat model of par tial hepatectomy. J Nucl Med 2008;49:122–8. 52. Gambhir SS, Herschman HR, Cher ry SR, et al. Imaging transgene expression with radionuclide imaging technolo gies. Neoplasia 2000;2:118–38. 53. Gotthardt M, Lal yko G, v an Eerd-Vismale J, et al. A new technique for in vi vo imaging of specif ic GLP-1 binding sites: f irst results in small rodents. Regul Pept 2006;137:162–7.

54. Ostendorf B, Scherer A, Wirrwar A, et al. High-resolution multipinhole single-photon-emission computed tomography in experimental and human arthritis. Arthritis Rheum 2006;54:1096–104. 55. Paulmurugan R, Ra y P , De A, et al. Imaging protein-protein interactions in li ving subjects. Trends Analyt Chem 2005; 24:446–58. 56. Pham TQ, Berghofer P, Liu X, et al. Preparation and biolo gic evaluation of a no vel radioiodinated benzylpiperazine, I-123-MEL037, for malignant melanoma. J Nucl Med 2007;48:1348–56. 57. Kung M-P, K ung HF. Mass ef fect of injected dose in small rodent imaging by SPECT and PET. Nucl Med Biol 2005;32:673–8. 58. Funk T, Mingshan S, Hase gawa BH. Radiation dose estimate in small animal SPECT and PET. Med Phys 2004;31:2680–6. 59. Kinahan PE, Defrise M, Clackdo yle R. Analytic image reconstr uction methods. In: Wernick MN, Aarsvold JN, editors. Emission tomography: the fundamentals of PET and SPECT . Ne w York: Academic Press; 2004. p. 421–42. 60. Horn BKP . Density reconstr uction using arbitrar y ra y-sampling schemes. Proc IEEE 1978;66:551–62. 61. Horn BKP . F an-beam reconstr uction methods. Proc IEEE 1979; 67:1616–23. 62. Tuy HK. An inversion formula for cone-beam reconstr uction. SIAM J Appl Math 1983;43:546–52. 63. Feldkamp LA, Da vis LC, Kress JW. Practical cone-beam algorithm. J Opt Soc Am A 1984;1:612–9. 64. Wang G, Yu HY, De Man, B . An outlook on X-ra y CT research and development. Med Phys 2008;35:1051–64. 65. Ayan AS, Metzler SD . Axial resolution of helical-orbit pinhole SPECT with synchronized and unsynchronized motion. IEEE Trans Nucl Sci 2008;55:524–30. 66. Bal G, Zeng GL, Acton PD, et al. Study of different pinhole configurations for small animal tumor imaging. 2004 IEEE Nuclear Science Symposium Conference Record 2004;3133–7. 67. Barrett HH, Gif ford HC. Cone-beam tomo graphy with discrete data sets. Phys Med Biol 1994;39:451–76. 68. Beekman FJ , Vastenhouw B . Design and simulation of a highresolution stationary SPECT system for small animals. Phys Med Biol 2004;49:4579–92. 69. Damiani C, Del Guer ra A, Moretti E, et al. Sampling considerations for high resolution small animal SPECT . IEEE Trans Nucl Sci 2005;52:555–9. 70. Metzler SD , Greer KL, Jaszczak RJ . Helical pinhole SPECT for small-animal imaging: a method for addressing sampling completeness. IEEE Trans Nucl Sci 2003;50:1575–83. 71. Furenlid L, Chen Y-C, Hyunki K. SPECT imager design and data acquisition systems. In: Kupinski MA, Barrett HH, editors. Small animal SPECT imaging. New York: Springer; 2005. p. 115–38. 72. Knoll GF. Radiation detection and measurement. 3rd ed. Ne w York: Wiley; 2000. 73. Kastis GA, Wu MC, Balzer SJ , et al. Tomographic small-animal imaging using a high-resolution semiconductor camera. IEEE Trans Nucl Sci 2002;49:172–5. 74. Tenney CR, Tornai MP, Smith MF, et al. Uranium pinhole collimators for 511-k eV photon SPECT imaging of small v olumes. IEEE Trans Nucl Sci 2001;48:1483–9. 75. Vastenhouw B , Beekman F . Submillimeter total-body murine imaging with U-SPECT-I. J Nucl Med 2007;48:487–93. 76. Gunter DL. Collimator design for nuclear medicine. In: Wernick MN, Aarsvold JN , editors. Emission tomo graphy: the fundamentals of PET and SPECT . New York: Academic Press; 2004. p. 153–68. 77. Beekman F, van der Have F. The pinhole: gateway to ultra-high-resolution three-dimensional radionuclide imaging. Eur J Nucl Med Mol Imaging 2007;34:151–61. 78. Accorsi R, Metzler SD . Analytic deter mination of the resolutionequivalent effective diameter of a pinhole collimator . IEEE Trans Med Imaging 2004;23:750–63.

Small Animal SPECT, SPECT/CT, and SPECT/MRI

79. Accorsi R, Metzler SD. Resolution-effective diameters for asymmetric-knife-edge pinhole collimators. IEEE Trans Med Imaging 2005;24:1637–46. 80. Shokouhi S, Fritz MA, Metzler SD, et al. Design of a multi-pinhole collimator in a dual-headed , stationar y, small-animal SPECT . 2006 IEEE Nuclear Science Symposium Conference Record 2007;2399–402. 81. Smith MF, Meikle SR, Weisenberger AG, Maje wski S. Design of multipinhole collimators for small animal SPECT . 2003 IEEE Nuclear Science Symposium. Conference Record 2004;2291–5. 82. Tae Yong S, Byung-Tae K, Jin Ho J, et al. Optimization of pinhole collimator for small animal SPECT using Monte Carlo simulation. IEEE Trans Nucl Sci 2003;50:327–32. 83. van der Ha ve F, Beekman FJ . Penetration, scatter and sensiti vity in channel micro-pinholes for SPECT : a Monte Carlo in vestigation. IEEE Trans Nucl Sci 2006;53:2635–45. 84. van der Ha ve F , Beekman FJ . Photon penetration and scatter in micro-pinhole imaging: a Monte Carlo in vestigation. Ph ys Med Biol 2004;49:1369–86. 85. Vanhove C, Defrise M, Lahoutte T, Bossuyt A. Three-pinhole collimator to improve axial spatial resolution and sensitivity in pinhole SPECT. Eur J Nucl Med Mol Imaging 2008;35:407–15. 86. Williams MB, Stolin AV, Kundu BK. Investigation of efficiency and spatial resolution using pinholes with small pinhole angle. IEEE Trans Nucl Sci 2003;50:1562–8. 87. Hwang AB, Hasegawa BH. Attenuation correction for small animal SPECT imaging using X-ra y CT data. Med Ph ys 2005; 32:2799–804. 88. Barrett HH. Detectors for small-animal SPECT II. In: Kupinski MA, Barrett HH, editors. Small-animal SPECT imaging. Ne w York: Springer; 2005. p. 49–86. 89. King MA, Glick SJ , Pretorius PH, et al. Attenuation, scatter , and spatial resolution compensation in SPECT . In: Wernick MN , Aarsvold JN, editors. Emission tomo graphy: the fundamentals of PET and SPECT. New York: Academic Press; 2004. p. 473–98. 90. Lalush DS, Wernick MN. Iterative image reconstruction. In: Wernick MN, Aarsvold JN, editors. Emission tomo graphy: the fundamentals of PET and SPECT . Ne w York: Academic Press; 2004. p. 443–72. 91. Shepp LA, Vardi Y. Maximum likelihood reconstruction in positron emission tomography. IEEE Trans Med Imaging 1982;1:113–22. 92. Lange K, Carson RE. EM reconstr uction algorithms for emission and transmission tomo gpraphy. J Comput Assist Tomogr 1984; 8:306–16. 93. Hudson HM, Larkin RS. Accelerated image reconstr uction using ordered subsets of projection data. IEEE Trans Med Imaging 1994;13:601–9. 94. Fessler JA, Hero AO. Penalized maximum-likelihood image reconstruction using space-alter nating generalized EM algorithms. IEEE Trans Image Process 1995;4:1417–29. 95. Qi J , Leah y RM, Cher ry SR, et al. High resolution 3D Ba yesian image reconstr uction using the microPET small animal scanner . Phys Med Biol 1998;43:1001–13. 96. Andreyev A, Defrise M, Vanhove C. Pinhole SPECT reconstr uction using blobs and resolution reco very. IEEE Trans Nucl Sci 2006; 53:2719–28. 97. Bal G, Acton PD. Analytical derivation of the point spread function for pinhole collimators. Phys Med Biol 2006;51:4923–50. 98. DiFilippo FP. Geometric characterization of multi-axis multi-pinhole SPECT. Med Phys 2008;35:181–94. 99. DiFilippo FP, Rif fe MJ, Heston WD, et al. Detached multipinhole small animal SPECT de vice with real-time calibration. IEEE Trans Nucl Sci 2006;53:2605–12. 100. Funk T, Kirch DL, Sun MS, et al. Simulation and validation of point spread functions in pinhole SPECT imaging. IEEE Trans Nucl Sci 2006;53:2729–35.

97

101. Gross KA, Kupinski MA, Hesterman JY. A fast model of a multiplepinhole SPECT imaging system. Proc SPIE 2005;5749:118–27. 102. Hsu CH, Huang PC. A geometric system model of f inite aperture in small animal pinhole SPECT imaging. Comput Med Imaging Graph 2006;30:181–5. 103. Hwang A, Taylor C, Seo Y, et al. Improving the quantitative accuracy of a dedicated small animal SPECT/CT scanner . Med Ph ys 2005;32:2133–4. 104. Israel-Jost V, Choquet P, Salmon S, et al. Pinhole SPECT imaging: compact projection/backprojection operator for efficient algebraic reconstruction. IEEE Trans Med Imaging 2006;25:158–67. 105. Metzler SD, Jaszczak RJ . Simultaneous multi-head calibration for pinhole SPECT. IEEE Trans Nucl Sci 2006;53:113–20. 106. Metzler SD, Jaszczak RJ, Greer KL, Bowsher JE. Angular-dependent axial-shift cor rection for pinhole SPECT . IEEE Trans Nucl Sci 2007;54:124–9. 107. Smith MF, Jaszczak RJ. An analytic model of pinhole aper ture penetration for 3D pinhole SPECT reconstr uction. Ph ys Med Biol 1998;43:761–75. 108. Zeniya T, Watabe H, Aoi T, et al. A new reconstruction strategy for image impro vement in pinhole SPECT . Eur J Nucl Med Mol Imaging 2004;31:1166–72. 109. Fryback DG, Thornbury JR. The ef ficacy of diagnostic imaging. Med Decis Making 1991;11:88–94. 110. ICRU. Medical imaging—the assessment of image quality (repor t 54). Bethesda (MD): Inter national Commission on Radiation Units and Measurements; 1996. 111. Barrett HH. Objective assessment of image quality—effects of quantum noise and object variability. J Opt Soc Am A 1990;7:1266–78. 112. Barrett HH, Abbey CK, Clarkson, E. Objecti ve assessment of image quality . III. R OC metrics, ideal obser vers, and lik elihood-generating functions. J Opt Soc Am A 1998;15:1520–35. 113. Barrett HH, Denny JL, Wagner RF, Myers KJ. Objective assessment of image quality. 2. F isher information, Fourier crosstalk and f igures of merit for task-performance. J Opt Soc Am A 1995;12:834–52. 114. Barrett HH, Yao J, Rolland JP, Myers KJ. Model observers for assessment of image quality. Proc Natl Acad Sci U S A 1993;90:9758–65. 115. Meng LJ, Clinthor ne NH. A modif ied unifor m Cramer-Rao bound for multiple pinhole aper ture design. IEEE Trans Med Imaging 2004;23:896–902. 116. Vunckx K, Beque D, Nuyts J, Defrise M. Single and multipinhole collimator design e valuation method for small animal SPECT . 2005 IEEE Nuclear Science Symposium Conference Record 2006; 2223–7. 117. Clarkson E, K upinski MA, Bar rett HH, Furenlid L. A task-based approach to adapti ve and multimodality imaging. Proc IEEE 2008;96:500–11. 118. Beekman FJ, van der Have F, Vastenhouw B, et al. U-SPECT-I: a novel system for submillimeter -resolution tomography with radiolabeled molecules in mice. J Nucl Med 2005;46:1194–200. 119. Accorsi R, Gasparini F , Lanza RC. A coded aper ture for highresolution nuclear medicine planar imaging with a con ventional Anger camera: e xperimental results. IEEE Trans Nucl Sci 2001; 48:2411–7. 120. Garibaldi F, Accorsi R, Cinti MN , et al. Small animal imaging b y single photon emission using pinhole and coded aper ture collimation. IEEE Trans Nucl Sci 2005;52:573–9. 121. Meikle SR, Fulham MJ, Fulton RR, et al. A prototype coded aperture detector for small animal SPECT . IEEE Trans Nucl Sci 2002; 49:2167–71. 122. Schramm NU, Ebel G, Engeland U , et al. High-resolution SPECT using multipinhole collimation. IEEE Trans Nucl Sci 2003; 50:315–20. 123. Barrett HH, Hunter WCJ. Detectors for small-animal SPECT I : overview of technolo gies. In: K upinski MA, Bar rett HH, editors. Small-animal SPECT imaging. New York: Springer; 2005. p.9–48.

98

MOLECULAR IMAGING: PRINCIPLES AND PRACTICE

124. Bradley EL, Cella J, Majewski S, et al. A compact gamma camera for biological imaging. IEEE Trans Nucl Sci 2006;53:59–65. 125. Del Guerra A, Bartoli A, Belcari N, et al. P erformance evaluation of the fully engineered YAP-(S)PET scanner for small animal imaging. IEEE Trans Nucl Sci 2006;53:1078–83. 126. Loudos GK, Nikita KS, Giokaris ND , et al. A 3D high-resolution gamma camera for radiophar maceutical studies with small animals. Appl Radiat Isot 2003;58:501–8. 127. McElroy DP, MacDonald LR, Beekman FJ, et al. Performance evaluation of A-SPECT: a high resolution desktop pinhole SPECT system for imaging small animals. IEEE Trans Nucl Sci 2002; 49:2139–47. 128. Schramm N , Wirrwar A, Sonnenber g F, Halling H. Compact high resolution detector for small animal SPECT. IEEE Trans Nucl Sci 2000;47:1163–7. 129. Cinti MN, Scafe R, Pellegrini R, et al. CsI(Tl) micro-pix el scintillation ar ray for ultra-high resolution gamma-ra y imaging. IEEE Trans Nucl Sci 2007;54:469–74. 130. Kim H, Furenlid LR, Cra wford MJ , et al. SemiSPECT : a smallanimal single-photon emission computed tomo graphy (SPECT) imager based on eight cadmium zinc telluride (CZT) detector arrays. Med Phys 2006;33:465–74. 131. Accorsi R, Celentano L, Laccetti P, et al. High-resolution I-125 small animal imaging with a coded aper ture and a h ybrid pixel detector. IEEE Trans Nucl Sci 2008;55:481–90. 132. Peterson TE, Wilson DW, Bar rett HH. Application of silicon strip detectors to small-animal imaging. Nucl Instr um Methods Ph ys Res A 2003;505:608–11. 133. Choong WS, Moses WW, Luke PN, Tindall CS. Design for a highresolution small-animal SPECT system using pix ellated Si(Li) detectors for in vivo 125 I imaging. IEEE Trans Nucl Sci 2005; 52:174–80. 134. Funk T, Shah KS, Despres P . A multipinhole small animal SPECT system with submillimeter spatial resolution. Med Ph ys 2006; 33:1259–68. 135. Fiorini C, Gola A, Zanchi M, et al. Silicon drift photodetectors for scintillation readout in medical imaging. Nucl Instr um Methods Phys Res A 2007;571:126–9. 136. Fiorini C, Perotti F. Small prototype of anger camera with submillimeter position resolution. Rev Sci Instrum 2005;76:(044303)1–8. 137. Soesbe TC, Lewis MA, Richer E, et al. De velopment and evaluation of an EMCCD based gamma camera for pre-clinical SPECT Imaging. IEEE Trans Nucl Sci 2007;54:1516–24. 138. Teo BK, Shestakova I, Sun M, et al. Evaluation of a EMCCD detector for emission-transmission computed tomo graphy. IEEE Trans Nucl Sci 2006;53:2495–9. 139. Meng LJ, Clinthor ne NH, Skinner S, et al. Design and feasibility study of a single photon emission microscope system for small animal I-125 imaging. IEEE Trans Nucl Sci 2006;53:1168–78. 140. Accorsi R, Autiero M, Celentano L, et al. MediSPECT : Single photon emission computed tomography system for small field of view small animal imaging based on a CdTe hybrid pixel detector. Nucl Instrum Methods Phys Res A 2007;571:44–7. 141. Aoi T, Zeniya T, Watabe H, et al. System design and development of a pinhole SPECT system for quantitative functional imaging of small animals. Ann Nucl Med 2006;20:245–51. 142. Cardi CA, Kar p JS, Zixiong C, et al. Pinhole PET (pPET): a multipinhole collimator inser t for small animal SPECT imaging on PET cameras. 2005 IEEE Nuclear Science Symposium Conference Record 2006;1973–6. 143. Furenlid LR, Barrett HH, Wilson DW, et al. FastSPECT II: a secondgeneration high-resolution dynamic SPECT imager . IEEE Trans Nucl Sci 2004;51:631–5. 144. Hesterman JY, Kupinski MA, Bar rett HH, et al. The multi-module, multi-resolution system (M 3 R): a no vel small-animal SPECT system. Med Phys 2007;34:987–93.

145. Hong KJ, Choi Y, Lee SC, et al. A compact SPECT/CT system for small animal imaging. IEEE Trans Nucl Sci 2006;53:2601–4. 146. Kastis GA, Furenlid LR, Wilson D W, et al. Compact CT/SPECT small-animal imaging system. IEEE Trans Nucl Sci 2004;51:63–7. 147. Kubo N, Zhao S, Fujiki Y, et al. Ev aluating performance of a pix el array semiconductor SPECT system for small animal imaging. Ann Nucl Med 2005;19:633–9. 148. Kundu BK, Stolin AV, Pole J, et al. Tri-modality small animal imaging system. IEEE Trans Nucl Sci 2006;53:66–70. 149. Lackas C, Schramm NU, Hoppin JW, et al. T-SPECT: A novel imaging technique for small animal research. IEEE Trans Nucl Sci 2005;52:181–187. 150. McDonald BS, Shokouhi S, Barrett HH, Peterson TE. Multi-energy, single-isotope imaging using stack ed detectors. Nucl Instr um Methods Phys Res A 2007;579:196–9. 151. Peterson TE, Shok ouhi S, Wilson DW, Furenlid LR. Multi-pinhole SPECT imaging with silicon strip detectors. 2005 IEEE Nuclear Science Symposium Conference Record 2006;2752–6. 152. Walrand S, Jamar F, de Jong M, Pauwels S. Evaluation of novel wholebody high-resolution rodent SPECT (Lino view) based on direct acquisition of linogram projections. J Nucl Med 2005;46:1872–80. 153. Zeng GL, Gagnon D . CdZnTe strip detector SPELT imaging with a slit collimator. Phys Med Biol 2004;49:2257–71. 154. Zeniya T, Watabe H, Aoi T, et al. Use of a compact pixellated gamma camera for small animal pinhole SPECT imaging. Ann Nucl Med 2006;20:409–16. 155. Meng LJ, Fu G, Roy EJ, et al. An ultrahigh resolution SPECT system for I-125 mouse brain imaging studies. IEEE Trans Nucl Sci 2007. [Submitted]. 156. Weisenberger AG, Gleason SS, Goddard J , et al. A restraint-free small animal SPECT imaging system with motion tracking. IEEE Trans Nucl Sci 2005;52:638–44. 157. Zingerman Y, Golan H, Gersten A, Moalem A. A compact CT/SPECT system for small-object imaging. Nucl Instrum Methods Phys Res A 2008;584:135–48. 158. Di Domenico G, Cesca N , Zavattini G, et al. CT with a CMOS flat panel detector inte grated on the YAP-(S)PET scanner for in vi vo small animal imaging. Nucl Instr um Methods Ph ys Res A 2007; 571:110–3. 159. Weisenberger AG, Wojcik R, Bradle y EL, et al. SPECT -CT system for small animal imaging. IEEE Trans Nucl Sci 2003;50:74–9. 160. Goetz C, Breton E, Choquet P , et al. SPECT lo w-field MRI system for small-animal imaging. J Nucl Med 2008;49:88–93. 161. Barrett HH, Furenlid L, Freed M, et al. Adaptive SPECT. IEEE Trans Med Imaging 2008;27:775–88. 162. Gonsalves RA. Phase retrie val and di versity in adaptive optics. Opt Eng 1982;21:829–32. 163. Paxman RG, Schulz TJ, Fienup JR. Joint estimation of object and aberrations using phase diversity. J Opt Soc Am A 1992;7:1072–85. 164. Rousset G. Wavefront sensing. In: Roddier F, editor. Adpative optics in astronomy. Cambridge, UK: Cambridge University Press; 1999. 165. Gauss RC, Trahey GE, Soo MS. Adaptive imaging in the breast. Proceedings of the IEEE Inter national Ultrasonics Symposium 1999;1563–9. 166. Ng GC, Freiburger PD, Walker WF, Trahey GE. A speckle target adaptive imaging technique in the presence of distributed aber rations. IEEE Trans Ultrason Ferroelectr Freq Control 1997;44:140–51. 167. Cao Y, Levin DN. Using prior knowledge of human anatomy to constrain MR image acquisition and reconstruction: half k-space and full k-space techniques. Magn Reson Imaging 1997;15:669–77. 168. Zientara GP. F ast imaging techniques for inter ventional MRI. In: Young I, Jolesz F A, editors. Inter ventional MR. London, UK: Martin Dunitz; 1995. p. 25–52. 169. Freed M, Kupinski MA, Furenlid LR, et al. A prototype instr ument for single pinhole small animal adapti ve SPECT imaging. Med Phys 2008;35:1912–25.

7 INSTRUMENTATION AND METHODS TO COMBINE SMALL ANIMAL PET wITH OTHER IMAGING MODALITIES CRAIG S. LEVIN, PHD

INTRODUCTION Earlier chapters of this book discussed positron emission tomography (PET) as a clinical tool. But in recent years, PET has also emer ged as a po werful pre-clinical research tool and is cur rently used in small laborator y animal research to visualize and track certain molecular processes associated with diseases, such as cancer, heart disease, and neurological disorders, in li ving small animal models of disease. in vi vo small animal imaging PET assays enable very sensitive studies of the cellular and molecular bases of disease in its natural state and may be used to guide the discovery and development of new treatments. Ho wever, to be ab le to visualize and quantify the often subtle preferential accumulation of a PET molecular probe in v ery small str uctures within small animals, such as rodents, requires special very high-resolution PET systems. As a result, during the past decade there has been substantial research and energy devoted to the de velopment of pre-clinical PET systems for rodent research. This work has resulted in the de velopment of numerous research prototypes and several commerciall y a vailable high-resolution PET systems that aim to enhance PETs ability to detect, visualize, and quantify low concentrations of probe interacting with its tar get, to more accuratel y study the subtle signatures associated with cellular and molecular processes of interest. However, in modern biology there is often a desire to measure more than one feature or parameter associated with a disease state to gain a more deep understanding of its nature. PET cannot readily achieve this “multiplexing”

as there is onl y one possible wavelength of photon emissions resulting from the positron annihilation process, and the PET radionuclide often has a relati vely long half-life. Thus, the desire to measure dif ferent disease characteristics, such as gene expression, ligand-receptor interactions, or enzymatic acti vity, using PET almost al ways requires multiple studies to be perfor med on distinct da ys. However, there are other in vi vo imaging modalities, such as X-ray computed tomo graphy (CT), single photon emission computed tomography (SPECT), magnetic resonance imaging (MRI), and optical imaging, that, through a multimodality imaging approach, can provide a range of complementary anatomic, ph ysiological, and/or cellular/molecular infor mation and contrast mechanisms to PET for more powerful characterization of disease states. Thus, there has recentl y been much acti vity to inte grate PET instr umentation with other imaging modalities for in-series or e ven simultaneous measurements of multiple complementary parameters of disease. This chapter begins with a basic re view of the principles of PET and cur rent commerciall y a vailable highresolution, small animal PET instr umentation. It then describes some of the research ef forts under w ay to combine small animal PET instrumentation with other imaging modalities, such as CT, SPECT, MRI, and optical imaging. A more in depth review of PET and other imaging modalities is left to the references and other chapters in this book. If these novel efforts to combine PET with other imaging modalities succeed without signif icant performance compromises of the subsystems in volved, this multimodality approach will likely continue to increase PET’s role in the study of disease and development of novel treatments.

99

100

MOLECULAR IMAGING: PRINCIPLES AND PRACTICE

PRE-CLINICAL PET There are se veral desirab le features of PET as a small animal (ie, pre-clinical) imaging modality.1,2 First, as with other biological imaging techniques, such as optical imaging, PET can be used to study the cellular and molecular processes associated with disease in li ve animals. Lik e optical imaging methods that e xploit the phenomena of bioluminescence and fluorescence, PET can detect a v ery low concentration of the probe molecule. Ho wever, unlike optical methods, PET can probe these subtle molecular signals deep (man y centimeters) within tissue, with high spatial resolution and contrast resolution, and thus provide quantitatively accurate spatial and temporal probe biodistribution data. The molecular probes used in PET are typically v ery small, lo w mass, biolo gically rele vant molecules, and thus can readil y reach their molecular target(s) without per turbing the natural states of cells and tissues. Finally, since PET is already a clinical standard of care, successful pre-clinical molecular imaging assays that are proven to be safe can be translated to the clinic. PET is currently used in small animal research to noninvasively study the molecular basis of disease and to guide the development of novel probes and molecular-based treatments.1–30 Many ne w molecular probes labeled with positron-emitting radionuclides and associated PET imaging assays are under de velopment to tar get, detect, visualize, and quantify v arious e xtracellular and intracellular molecules and processes associated with diseases, such as cancer, heart disease, and neurolo gical disorders. 1–30 Thus, there is a continuing need to impro ve PET’ s molecular sensitivity, that is, the capability to detect and quantify the subtle signatures associated with molecular tar gets and processes.31,32

REVIEW OF SMALL ANIMAL PET IMAGING SYSTEM TECHNOLOGY Small Animal PET System Design Issues To visualize and accuratel y quantify the biodistribution of a PET molecular probe in small structures within small animal disease models requires special high-resolution PET systems. During the past tw o decades, there has been substantial research in the de velopment of preclinical PET systems for rodent research.33 This work has resulted in the de velopment of numerous small animal PET research prototypes (e g,34–45) and commerciall y available systems 46–52 (Figure 1). Some of these g roups have made attempts to combine their PET technolo gy with other modalities.

With the e xception of one commerciall y a vailable gas-based multiwire propor tional counter (MWPC) system design, 44,52 nearly all small animal PET system designs (e g,34–43,45–51) use v ariations of the same basic position-sensitive annihilation photon detector design concept, which is essentiall y a miniature v ersion of that used in clinical PET system designs. The system is built from detector modules arranged into a ring. Each module comprises ar rays of long and nar row scintillation cr ystal rods with their small ends coupled to a position-sensiti ve photodetector (Figure 2). Each cr ystal is co vered with a very thin reflector, except for the end f ace coupled to the photodetector, so that each is opticall y isolated from its neighbors. An incoming photon that interacts in one of the crystal rods creates a small flash of light. The light pulse reflects off the crystal faces and exits the end of the cr ystal rod, which forms the basic light signal to be collected by the photodetector . In small animal PET systems, the crystals may be directl y coupled to the photodetector(s), coupled through a light diffuser, or coupled through f iber optics. Coupling scintillation cr ystal to f ibers or f iber bundles before the photodetector is not prefer red because fiber coupling al ways introduces v arying le vels of light signal loss and energy/time dispersion. For higher resolution, narrower crystal rods (eg, < 2 mm) are used (Figure 2), but this tends to trap the scintillation light, w hich introduces light collection variations that depend on the photon interaction point as well as further dispersion in the magnitude and transit time of the light signal reaching the photodetector. These f actors af fect scintillation detector signal-to-noise ratio (SNR), which leads to degradation of multiple perfor mance parameters. To some what relie ve this prob lem, in most pre-clinical PET system designs the crystals are made shor t (eg, < 10 mm), with substantial compromise to photon sensitivity, but with higher and less-varying light collection ef ficiency into the photodetector. The photodetector(s) collect and convert the available light signal into electronic pulses, which are amplified and processed to estimate the incoming interaction location, total energy deposited, and ar rival time of each incoming annihilation photon. The photodetectors currently used in small animal PET system designs are almost al ways photomultiplier tubes (PMTs), in particular position-sensitive PMTs (PSPMTs), w hich contain an ar ray of char ge collecting anodes within a single e vacuated v acuum tube. The signals induced on the anodes can be used to localize the scintillation light flash to within 2 mm or less. A variation of the basic scintillation detector described is to use semiconductor photodetectors, such as avalanche photodiodes (APDs), instead of PMTs to read out the scintillation

Instrumentation and Methods to Combine Small Animal PET with Other Ima ging Modalities

Source size Non-colinearity

J

Subtracted Hammersmith

3 FWHM Resolution (mm)

101

Exact HR (CTI)

SHR-2000 (Hamamatsu)

RatPET (UCLA)

BaF2/TMAE (VUB)

Tomitani

SHR-7700 A-PET (Philips) (Hamamatsu) MADPET HRRT ATLAS (NIH) Donner 600 (Munich) (CTI)(Berkeley) microPET (UCLA) ClearPET APD-BGO YAPPET (Ferrara) (Sherbrooke) Ge eXplore microPET II Focus Light Sharing (b, 2.1 mm) MADPET ll LabPET HIDAC

2

1

Electronic Coding (b, 1.1 mm)

LabPET ll

Individual Coupling (b, 0 mm) Crystal Resolution (d/2)

0 0

1

2 3 Crystal Size (mm)

4

5

Figure 1. Intrinsic spatial resolution (quantified by the full-width-at-half-maximum [FWHM] of a point source response along a given direction) at the center achieved with several existing small animal positron emission tomography (PET) system designs versus scintillation crystal array element size used (source size and annihilation photon acolinearity effects de-convolved). The symbol b parametrizes different degrees of light or charge multiplexing inherent to a PET detector technology that contribute toward spatial resolution degradation. Due to scintillation light sharing, charge sharing in the photodetector or electronic readout, inter-crystal scatter, and positron range, intrinsic spatial resolution often does not reach the fundamental limit set by 1/2 the crystal element width d (represented by the dashed line). Nearly all of these designs have significant resolution degradation as the point source moves away from the system center. Adapted from Bloomfield PM et al.34 Courtesy of Roger Lecomte, University of Sherbrooke.

light. A re view of APDs and other ne w detectors and methods proposed is provided in the study by Levin.32

PET System Performance Issues Performance parameters dictate a PET system’s ability to visualize and quantify a molecular signal in the presence of background.31,32 To understand the effects of PET system design compromises that have been made to achieve current PET-based multimodality system designs as w ell as the ef fects of operating another nearb y modality on PET system perfor mance, here w e re view PET system performance basics. There are se veral important parameters of PET system performance, such as photon sensitivity, spatial resolution, energy r esolution, coincidence time r esolution, and count r ate perf ormance. The ener gy and temporal resolutions as w ell as count rate perfor mance w ork together to def ine the instr ument contrast r esolution, which is the ability to dif ferentiate a subtle probe concentration from background or two slightly different concentration levels of probe in adjacent targets. The photon sensitivity, spatial resolution, and contrast resolution work together to define the molecular signal sensitivity of a PET instrument.31,32

2x2x10 mm3

1x1x10 mm3 Photodetector

Figure 2. Nearly all small animal positron emission tomography (PET) system designs use long and narrow scintillation crystal rod elements (left) to assemble an array that is coupled end-on to a position-sensitive photodetector (right). The crystals rods are typically optically isolated with a thin inter-crystal reflector. The PET system comprises rings of these fundamental module subunits.

Photon Sensitivity

In radionuclide imaging, the photon e vents and resulting signals are collected and processed one at a time rather than running in photon integration mode, which is used in X-ray imaging. The system photon sensitivity is the fraction of all coincident 511 k eV photon pairs emitted from the imaging subject that are recorded b y the system and is also refer red to as the coincidence photon detection efficiency. This parameter determines the statistical quality of image data for a given acquisition time. Photon sensitivity impacts image quality because it influences the noise le vel of images reconstr ucted at a desired spatial

102

MOLECULAR IMAGING: PRINCIPLES AND PRACTICE

resolution. Photon sensiti vity in PET is impro ved by (1) increasing the probability that emitted photons will traverse detector material, which is known as the geometric efficiency and (2) b y increasing the lik elihood that photons tra versing detector material will be stopped and counted, ter med the intrinsic detection ef ficiency. The geometric efficiency is enhanced b y bringing the detectors as close as possible to the body, and covering the subject with as much detector area as possib le; these factors decrease the chance that photons will escape without traversing detector material. Ho wever, bringing the detectors closer to the subject can lead to position-dependent parallax positioning er rors (hence loss of spatial resolution uniformity) due to photon penetration into the detector elements 32 (Figure 3). The intrinsic detection efficiency is improved by tightly packing the detector elements to gether with little or no spaces, using denser , higher atomic number ( Z) and thick er (longer) detector elements to improve the 511 keV stopping power. Typical small animal PET detector system photon sensiti vities range from < 1% (one coincidence photon pair collected for every 100 emitted) to several percent. Spatial Resolution

The spatial resolution describes a system’s ability in order to distinguish tw o closel y spaced molecular probe concentrations and is important to detect and visualize subtle

molecular signals emitted from miniscule structures. PET spatial resolution is limited by the fact that one is trying to precisely deter mine the location of a positron-emitting radionuclide attached to the probe molecule indirectl y using the line drawn between the two annihilation photon hits in the detectors. Because this line results from tw o electronically deter mined interaction points, this process is called electronic collimation . The spatial resolution is typically measured b y imaging a point-lik e positron radioactive source and measuring its obser ved spread in the reconstructed images. The fundamental spatial resolution limit is dictated b y (1) the positron r ange effect, which is due to v ariations in direction and path length of all the possib le positron trajectories (F igure 4) created from a given point positron source; the e xtent of this resolution degrading effect depends upon the range of energies of the emitted positrons and the medium traversed by the positrons before the y annihilate; (2) the photon acollinearity effect, which is caused by the fact that since the positron and electron are not al ways at rest when they combine, the tw o annihilation photons are not al ways emitted 180° apart, and hence the line def ined by the two detector elements that w ere hit will not al ways pass through the point of the positron-electron annihilation; this acollinearity ef fect on spatial resolution is w orse for larger system diameters; (3) the size of the photon detector element (a.k.a. detector resolution or pixel size), which determines how precisely a system can localize the photon

4.5

Dd

Dd 5 2 mm

ui R

Dx r ui

Radial blurring (mm FWHM)

4

Dd 5 5 mm 3.5

Dd 5 10 mm

3 2.5 2 1.5 1 0.5 0 0

0.5

1

1.5

2

2.5

3

3.5

4

Radial coordinate (cm) Figure 3. A, Depiction of the radial resolution blurring Δx due to photon interaction depth variation in the detector crystal elements for coincident photons emitted from a point source located at radial coordinate r, entering obliquely into two isolated crystals. The resolution blurring is insignificant when the source is at the center (radial coordinate r = 0) and the photons enter normal (θi = 0) to the crystal elements, but away from the center can be shown to roughly vary linearly with r and Δd, the photon interaction depth resolution, and inversely with the system radius R.32 B, The estimated radial resolution degradation Δx (full-width-at-half-maximum [FWHM]) plotted as a function of radial position r of a point source for an 8 cm diameter positron emission tomography (PET) system for different photon interaction depth resolutions Δd.32 Very few PET systems incorporate photon interaction depth measuring capabilities, and thus Δd just equals the length of the crystal elements used, which for small animal PET is typically on the order of 10 mm.

Instrumentation and Methods to Combine Small Animal PET with Other Ima ging Modalities

hits. The size of the detector element used in PET has been gradually decreasing throughout recent y ears to impro ve spatial resolution. At present, typical clinical systems use 4 to 6 mm width detector elements and most small animal systems now use 1.5 to 2 mm width detector elements. Figure 5 shows the combined spatial resolution limit from these three effects as a function of detector pixel size for various system diameters ranging from small animal to clinical PET systems for an 18F point source.53 We see that in principle spatial resolution ma y be impro ved signif icantly by reducing the 511 k eV detector pix el size. This element size dominates spatial resolution for small diameter (< 20 cm) animal PET systems because the acolinearity effect on spatial resolution is minor for small detector ring diameters. Ho wever, de veloping 511 keV photon detector arrays with miniscule detector elements is challenging and typicall y results in perfor mance compromises in other impor tant system parameters. F or example, using a point 18F positron source, a 20 cm detector system diameter for small animal PET , and 1 mm scintillation crystal pixels, Figure 5 indicates that it is possible in principle to achie ve submillimeter full-width-athalf-maximum (FWHM) spatial resolution at the center of the system, pro vided there are enough counts in the acquired data (adequate 511 k eV photon sensiti vity) to reconstruct images at that desired spatial resolution without requiring signif icant smoothing.

103

However, it is very difficult to collect a high fraction of the a vailable light out of nar row (eg, < 2 mm width) and long (> 2 cm) scintillation cr ystals,54 and thus individual cr ystals may not be resolv ed in a detector flood histogram. Fur thermore, this light collection ef ficiency varies as a function of interaction location within the crystal, causing ener gy and ar rival time dispersion of the resulting detector signal, and so ener gy and time resolutions suf fer as a result. 54 Typically, to achie ve acceptable light collection with 1 mm cr ystal pix els, their length is limited to ≤ 10 mm (e g, see 35–43,54), but this significantly compromises the probability of absorbing 511 k eV photons, and hence limits the o verall photon sensitivity performance. Variations in photon interaction depths in cr ystals that can cause energy and arrival time dispersion can also further de grade spatial resolution if the PET detector design does not incor porate a means to measure photon interaction depth (see Figure 3). Energy and Coincidence Time Resolution

Energy resolution is the precision to which one can measure the incoming photon ener gy. Because photons that scatter lose energy, good ener gy resolution means one ma y use a narrow energy window to reduce scatter photon contamination in image data without signif icantly compromising

10 18F

18

POSITRONS

F (Emax 5 635 keV)

5

0

25

4 mm

10,000 EVENTS 2 10 2 10 2500

25 0 5 END POINT COORDINATE (mm) 100,000 EVENTS

EVENTS

2000

100 TRACKS

10

18

F

1500 1000

0.102 mm FWHM

500

1.03 mm FWTM

0

4 mm

22

21 0 1 2 END POINT COORDINATE (mm)

Figure 4. (Left) 100 18F positron tracks in water emanating from a point source. (Top right) Scatter plot of endpoints for 10,000 positron tracks in water. Annihilation photons are created at or very near the endpoints. (Bottom right) Histogram of x-coordinates of 18F positron endpoints shows a cusp-like function with only ~100 micron full-width-at-half-maximum (FWHM) (full-width-at-tenth-maximum [FWTM] is ~1 mm). Adapted from Levin CS and Hoffman EJ.53

104

MOLECULAR IMAGING: PRINCIPLES AND PRACTICE

System Diameter

SPATIAL RESOLUTION (mm)

6

FWHM 0 cm 10 cm 20 cm 80 cm

FWTM 0 cm 10 cm 20 cm 80 cm

18F

5 4 3 2 1 0 0

1 2 3 4 DETECTOR ELEMENT WIDTH (mm)

5

Figure 5. Spatial resolution limit (full-width-at-half-maximum [FWHM] and full-width-at-tenth-maximum [FWTM]) determined from three fundamental effects of positron range, photon acolinearity, and crystal element size for an 18F point source in water versus the latter for different system diameters. In principle, using an 18F point source placed at the system center, submillimeter spatial resolution is possible. Adapted from Levin CS and Hoffman EJ.53

photon sensitivity. A nar row energy window also helps to reduce the rate of random (unpaired) photon contamination because man y of these photons also under go scatter. The coincidence time r esolution determines how well one can decide w hether tw o coincident photons tr uly are detected simultaneously. Analogous to benef its of good energy resolution, good coincidence time resolution means one may use a nar row time windo w to reduce random e vents without compromising photon sensiti vity. Ener gy and coincidence time resolution are impro ved by using scintillation cr ystals that generate brighter and faster light pulses, low noise photodetectors, and by collecting a higher and constant fraction of the a vailable scintillation light into the photodetector to create larger, nonvarying, more robust electronic pulses. A typical value for PET ener gy resolution is 25% FWHM at 511 keV and 3 ns FWHM for coincidence time resolution. A figure of merit for SNR in PET images is provided by the noise-equivalent count rate (NECR), defined by T 2/(T + S + kR), where S and R are the photon scatter and random coincidence rates, respecti vely, and T is the tr ue, non-scattered, non-random photon coincidence rate (a.k.a. the signal). The factor k in front of R equals two when a separate measurement is made to estimate R, otherwise it is unity. Count Rate Performance

Each detector signal recorded in a PET system has a finite processing time. If too many photons hit the detectors in a

given time, the front-end photon detectors or subsequent acquisition electronics in the PET system can saturate due to piling up of more than one electronic detector pulse within the required electronic signal processing duration. Typically, the de gree of pile up is limited b y the photon detector signal processing time, w hich depends upon the decay time of the scintillation cr ystal, the ef fective integration time of the electronics, and the photon e vent rate seen by the detector. For example, suppose a 10 mCi (370 MBq) point source is placed at the center of a PET system with 10 detector modules providing 5% coincidence photon detection ef ficiency. Then, the a verage photon e vent rate per detector module is roughly 3.7 × 108 (radionuclide decays per second) × 2 (photons per event) × 0.05 (photon sensitivity) ÷ 10 (photon detector modules) = 3.7 × 106 counts per second. If each system detector module requires 1 µs of processing time per event, there could be significant pile up of e vents. For a gi ven system photon sensitivity, for the best count rate performance, the system should use scintillation cr ystals with fast decay time, fast processing electronics, and limited activity within the sensitive field-of-view (FoV).

Present Commercially Available Small Animal PET System Designs Present commercially available small animal PET system designs are technolo gy transfers from academic research system developments33 (see Figure 1), and nearl y all are variations of the same scintillation detector theme described. As will be seen, it has been dif ficult to obtain exceptional perfor mance in all parameters simultaneously; the majority of cur rently a vailable small animal PET system designs use trade-of fs betw een parameters, such as photon sensiti vity, spatial resolution, scintillation light collection ef ficiency, energy resolution, and coincidence time resolution. F or example, no system design to date can provide the following “wish list” of small animal PET system perfor mance values: ≥ 15% photon sensitivity, ≤ 1 mm FWHM spatial resolution that is unifor m throughout the sensitive FoV, > 90% scintillation light collection ef ficiency, ≤ 13% FWHM ener gy resolution at 511 keV, and ≤ 3 ns FWHM coincidence time resolution. To date, attempts to achie ve one of these perfor mance parameters, for e xample, 1 mm spatial resolution, has resulted in the signif icant compromise of one or more of the others. All of the current commercially available scintillation detector-based system designs ha ve nonunifor m spatial resolution that rapidly degrades with distance from the center, and only a few percent photon sensitivity using a narrow energy window (high energy threshold).

Instrumentation and Methods to Combine Small Animal PET with Other Ima ging Modalities

A technology development at University of California, Los Angeles in the late 1990s 35 was transfer red to what has become the Siemens microPET product line.46 The detector module design uses an optical f iber interface between a lutetium-oxyorthosilicate (LSO) scintillation crystal array and a multianode PSPMT to enab le a high inter -module packing fraction. The cur rent generation system, the In veon, uses b lock modules comprising 20 × 20 arrays of 1.5 × 1.5 × 10 mm 3 LSO crystals coupled through a tapered light guide into PSPMTs; these modules are ar ranged in a 16.1 cm diameter ring, with a 12 cm diameter bore and 10 cm transaxial and 12.7 cm axial FoV. The spatial resolution and photon sensiti vity are 1.4 mm FWHM and 10% (100 keV threshold), respectively, at the system center. The detector scheme developed at University of Pennsylvania37 went into the Philips Mosaic small animal PET system, 47 which uses just under 17,000 discrete 2 × 2 × 10 mm 3 gadolinium orthosilicate (GSO) cr ystals coupled through a single lar ge annular light dif fuser to a bank of standard PMTs. The system has a 21 cm diameter bore and a 12.8 cm transaxial and 11.6 cm axial FoV. Note that unlik e the other scintillation cr ystal designs, the Mosaic is not par titioned into indi vidual block modules. The reconstr ucted spatial resolution ranges betw een 2.7 mm FWHM at the center and 3.2 mm FWHM at a radial offset of 45 mm from the center with an absolute photon sensitivity of 0.65% (410 keV threshold).55 Work at National Institutes of Health, Depar tment of Nuclear Medicine 36 was translated to a Spanish compan y called Suinsa and subsequentl y became w hat is no w known as the General Electric (GE) eXplore Vista PET system.48 This detector concept uses a tw o-layer scintillation crystal array or phoswich to allow coarse estimation of photon depth of interaction (DOI). Each la yer is made from a dif ferent scintillation cr ystal type (lutetiumyttrium-oxyorthosilicate [LYSO] and GSO) with a different scintillation light deca y constant. Pulse shape discrimination methods are used to determine which array layer was hit by an incoming photon. The system has two rings of 6084 LYSO-GSO crystals arranged into modules, each containing 13 × 13 ar rays of 1.5 × 1.5 × 15 mm 3 LYSO-GSO pairs. The detector diameter is 11.8 cm, with a 6 cm transaxial and 4.6 cm axial FoV. The reconstructed spatial resolution is 1.6 mm FWHM at the center with a 4% absolute photon sensitivity (250 keV threshold).56 A research concept conceived at University of Texas38 has been incor porated into a small animal PET system built by Gamma Medica-Ideas49 (and currently distributed by GE). The system uses bismuth ger minate (BGO) scintillation cr ystals with a light sharing technolo gy that

105

allows them to be read out b y standard PMTs. The edge crystals of each module are trapezoidal to promote high inter-detector module cr ystal packing fraction (reduce gaps between modules). The system specif ications are an axial FoV of 11.8 cm, a useful transaxial F oV of 10 cm, and a center point source photon sensiti vity and spatial resolution of 10% (100 k eV threshold) and 1.5 mm, respectively. A University of Sherbrooke development39 was transferred to Advanced Molecular Imaging, Inc. under the name “LabPET” and more recentl y to Gamma MedicaIdeas.49 The basic detector design uses an array of two different scintillation cr ystal elements (L YSO and LGSO) with different decay times coupled to a singleAPD device. Using pulse shape discrimination enab les the identif ication of w hich of the tw o crystals is hit per module b y an incoming photon, using only one electronic readout channel. The resulting system has a 15.6 cm detector ring diameter with an 11 cm diameter aper ture, 10 cm useful transaxial FoV, and either a 3.75, 7.5 or 11.25 cm axial FoV with 1536, 3072 or 4608 APDs, respecti vely. The specified center point source photon sensiti vity is 2, 4, and 6%, respecti vely, for the three a vailable axial F oVs. The spatial resolution is 1.5 mm FWHM at the center . The ClearPET LYSO/lutetium-yttrium aluminum perovskite (LuYAP) phoswich scanner51 is a technology transfer from the ideas de veloped within the ClearPET g roup41 of the Cr ystal Clear Collaboration (CCC), a scintillation crystal research or ganization based at Or ganisation Européenne pour la Recherche Nucléaire (CERN), an international particle physics laboratory located in Geneva, Switzerland. The system has tw o adjustab le detector diameters, 13.5 and 28.5 cm, with an open gantry space of 12.5 and 22.0 cm, respecti vely. The phos wich detectors comprise two layers of 2 × 2 × 10 mm 3 crystals of LYSO and LuYAP coupled to PSPMTs. Due to signif icant gaps between the detectors, the system rotates around the subject to enable full angular sampling. At the system center , the specified spatial resolution is 1.5 mm and the absolute photon sensitivity is 3.8%. The Y AP-PET system50 is an of f-shoot from University of F errara and Pisa. 40 The system comprises four rotating heads spaced 15 cm apar t, each with an active area of 4 × 4 cm 2, containing a 20 × 20 ar ray of 2 × 2 × 30 mm3 optically isolated YAP crystals coupled to PSPMTs, forming a 4 cm useful transaxial and axial FoV. The reconstructed spatial resolution and absolute photon sensitivity are 1.8 mm FWHM and 1.7% (50 keV threshold) for a centered point source, respecti vely. The Oxford Positron Systems high-density avalanche chamber (HID AC) PET system 52 is a culmination of

106

MOLECULAR IMAGING: PRINCIPLES AND PRACTICE

years of high-resolution gas MWPC imaging system developments at CERN 44 that were modified and ref ined for small animal imaging. 57 In this position-sensitive gas ionization chamber , the annihilation photons are converted b y lead cathode plates into electrons, w hich are subsequently detected and localized b y collecting their ionization generated as the y drift and a valanche in the gas. The lead cathode plates are for med b y la yers of laminated lead containing interlea ved insulated sheets, mechanically drilled with ~200,000 holes, each 0.4 mm diameter, each of w hich acts as an independent detector element. The hole pitch, ~0.5 mm, sets the limit to the intrinsic resolution. The quad-HID AC consists of four large rotating heads, each with eight HIDAC detector layers with an active area of 17 × 28 cm2, with head separation of 17 cm. These layers provide information about the photon interaction depth. The transaxial and axial F oV are 17 and 28 cm, respecti vely. The reconstructed spatial resolution is 1.0 mm in all three-dimensions (3D). Due to this capability to precisely localize the interaction coordinates of incoming photons in 3D, the spatial resolution is uniform throughout the F oV (not just at the center [see Figure 3], w hich is another distinguishing feature of the HIDAC technology), and the absolute photon sensiti vity is 1.8% for an effective 200 keV threshold. However, due to signif icant dispersion of the signal in the lead la yers and detection process, the ener gy and coincidence time resolutions are dif ficult to assess. Two v ariants for this technology are commerciall y a vailable comprising 16 and 32-modules.58

MULTIMODALITY PET IMAGING Introduction Multimodal in vivo imaging enables the measurement of multiple, complementar y image contrast mechanisms to study anatomic, ph ysiologic, cellular , and/or molecular pathways of disease in li ving subjects. The remainder of this chapter focuses on a discussion of instr umentation and algorithms for combining PET with other imaging modalities to f acilitate biolo gical studies in small animals. Multimodality molecular imaging with PET has become an essential tool to study the molecular pathways of disease in li ving subjects, to aid in the disco very and testing of novel therapeutic approaches in animal models of human disease, and for the development of new molecular contrast agents. Multimodality imaging requires easy and con venient access to multiple imaging systems and robust registration

of images generated b y the indi vidual modalities. To address these concerns, industry and academic institutions are developing v arious approaches specif ically designed for imaging small animals that allows one to combine the power of PET with other imaging modalities, such as X-ray CT, MRI, and optical imaging. The focus of this chapter is on tr uly inte grated or “h ybrid” multimodal systems that have been developed for small animal imaging. This approach is more complex but has the advantage of a higher lik elihood of successful image re gistration because there is less chance of subject movement or physiological changes when multimodal studies are performed within the same hybrid system and there is a signif icantly shorter time span betw een multimodal studies, or the y may even be run simultaneously. The hybrid approach also provides a high le vel of con venience for biolo gical researchers because the multimodal studies are a vailable immediately without having to move the subject betw een systems in dif ferent rooms or buildings or requiring the scheduling of multiple imaging studies on dif ferent machines. Although the commerciall y a vailable multimodality hybrid clinical systems are either PET/CT or SPECT/CT scanners, the g reater system fle xibility enab led b y small animal imaging research has resulted in the de velopment of several high-resolution dual- and tri-modality systems, such as PET/CT,59–61 PET/SPECT/CT,62–64 PET/MRI,65–85 and PET/optical. 86–89 Such multimodality systems f acilitate a range of in vi vo strategies to obtain rich, cor relative information about the molecular basis of disease and enhance inter pretation and quantif ication capabilities of data from the individual modalities involved.

Software-Based Fusion of Multimodality Images Over the years, many techniques have been developed for multimodal clinical image re gistration.90,91 However, although there ha ve been a fe w incidental studies, 92–98 in general image registration algorithms used in human studies have not been well characterized for small animal imaging. Softw are fusion of data from tw o separate imaging modalities is possible with the help of anatomical or f iducial mark ers (e g, small, dense str uctures or radionuclide sources placed at the subject contours) that allo w spatial registration of the two image volumes, but such efforts are most successful for studies of or gans and tissues that do not move with time, such as the brain. 99 As is the case for clinical imaging re gistration, some techniques for small animal imaging have used external fiducial markers to aid

Instrumentation and Methods to Combine Small Animal PET with Other Ima ging Modalities

the re gistration process, w hereas other researchers ha ve attempted fully automated algorithms that do not in volve user interaction. 92–98 State-of-the-art image re gistration techniques enable automatic image re gistration through a rigid body transfor mation, for e xample, using mutual information-based criteria, ignoring or gan defor mation due to ef fects, such as cardiac and respirator y motion or spatial linearity differences between modalities. There has also been progress in the development of nonrigid registration algorithms that can compensate for defor mation perceived by different imaging modalities or re gister images from dif ferent subjects. 100 However, despite some progress, man y image re gistration prob lems par ticularly for small animal imaging remain unsolv ed, and this is likely to continue to be an acti ve f ield of research in the future.

Hardware-Based Approaches to Combine Small Animal PET with Other Modalities The reco gnized limitations and incon veniences of software-based image fusion led to the development of hardware approaches to f acilitate more accurate re gistration of image data from multiple modalities. Some g roups have simply developed special fixtures that can be rigidly and reproducibly mounted on the imaging beds of distinct small animal scanners (eg, PET and CT). For example, in the study b y Cho w and colleagues, 101 using such a movable f ixture and a 3D g rid phantom with 1288 lines, a spatial transfor mation matrix for re gistration w as derived using a 15-parameter perspective model, yielding an average registration error between PET and CT mouse bone scans of less than 0.335 mm. The reproducibility and robustness of the system also enab led the use of CT images for accurate attenuation cor rection of the PET data to increase quantitative accuracy.102 The other approach for multimodality imaging is to develop a system that combines more than one type of imaging technolo gy into one inte grated unit. Such a hybrid system allo ws simultaneous and/or sequential acquisitions with the dif ferent modalities, ideall y without compromising the performance of either system, and provides additional cor relative infor mation that cannot be obtained as easily using the two modalities separately. Combined modality instr umentation that is capab le of truly simultaneous acquisition of the distinct modalities allows multiple, time-cor related biolo gical measurements of the same disease state. Such a h ybrid system may also enable simultaneous imaging of multiple molecular tar gets if dif ferent molecular contrast agents/mechanisms can be used. F or imaging methods

107

that are time-consuming, such as PET and MRI, w hich can last ~30 to 45 minutes each, simultaneous imaging also provides the benef it of reduced imaging time compared to performing the two studies sequentially. The standard design of PET/CT systems is an example of an approach that enab les onl y sequential acquisitions with the two modalities. PET is used to measure the modified cellular or molecular characteristics of a diseased state, and CT is used in series to pro vide high-resolution visualization of the cor responding anatomy w here the diseased tissue (eg, cancer) resides, without compromising the PET measurement. Adding CT to PET has the additional benefit of enhancing PETs accuracy and throughput by facilitating a rapid, low-noise, accurate estimate of photon attenuation coefficients,103,104 a feature that in the pre-clinical arena is perhaps more impor tant for rats than for mice due to their larger size and hence increased photon attenuation. 105 In addition to pre-clinical PET/CT designs, 59–61 several investigators are investigating dual-modality systems that combine PET with other v arious imaging technologies, such as PET and SPECT,62–64,106 PET and MRI,65–85 and PET and optical imaging.86–89 Moreover, tri-modality pre-clinical systems inte grated in a single gantr y for PET/SPECT/CT have also been achieved.62–64 As one lear ns about multimodal system designs under investigation, impor tant questions should be k ept in mind: Does the integration of multiple systems provide new or more accurate infor mation than w as otherwise available with the distinct stand-alone systems? Does integration compromise either system’ s perfor mance compared to when they stand alone? That is, were any of the indi vidual system perfor mance specif ications compromised to f acilitate integration? Should inte gration of multiple modalities occur e ven at the cost of signif icant performance degradation/limitation of one or more of the modalities involved? After integration, is there an y measurable mutual interference betw een modalities that further degrades their performance during operation? In the previous section, we learned that an impor tant factor that strongly affects PET system performance is the photon sensiti vity. Thus, an y multimodal PET system design with a small axial FoV coverage, thin (short) detector cr ystals, or gaps betw een detector modules/elements will compromise the number of 511 k eV photons recorded, and therefore, image quality and quantitati ve accuracy. We also lear ned that the SNR and ener gy/time dispersion of the scintillation light signal measured by the PET system detectors are impor tant f actors af fecting nearly all perfor mance parameters to dif ferent de grees. Thus, any design that compromises this light signal will also compromise the PET system perfor mance.

108

MOLECULAR IMAGING: PRINCIPLES AND PRACTICE

PET/CT

Despite the popularity and widespread interest in clinical dual-modality PET/CT imaging, only a few small animal PET/CT prototypes ha ve been de veloped at academic institutions,39,59–61 although more are a vailable commercially.46–49 Excellent micro-CT images of live animals are being obtained using cone-beam X-ra y CT imaging and reconstruction.107 In X-ra y CT, the spatial resolution is limited by a convolution of the X-ray beam focal spot size produced b y the tube aper ture and the X-ra y detector intrinsic spatial resolution. Image SNR at a gi ven reconstructed resolution is deter mined by the photon statistics of the detected X-ray flux. However, it has also been recognized for high-resolution small animal CT that dose issues are critical. 108 The Uni versity of Califor nia at Da vis g roup de veloped a prototype microCT/microPET dual-modality small animal imaging system for combining anatomic and molecular imaging of the mouse. 59,60 The microPET detectors used 9 × 9 arrays of 3 × 3 × 20 mm3 LSO scintillator cr ystal elements coupled through a f iber-optic taper to a PSPMT . These 60 × 60 mm 2 flat-panel PET detectors were placed on opposite sides and rotated about the animal with the annihilation photons from the positron emission detected in electronic coincidence. The X-ray CT system used a < 75 micron microfocus X-ra y tube and a 6.6 × 5.5 cm 2 amorphous selenium detector array of 1024 × 832 pixels, each 66 microns, coupled to a complementary metal-o xide semiconductor (CMOS) flat-panel readout ar ray.109 The same g roup later de veloped a more adv anced microCT/microPET II system. The microPET II is a fullring tomograph comprising three rings of 30detector blocks made from 14 × 14 arrays of 1 × 1 × 12 mm3 LSO crystals coupled through a tapered fiber-optic bundle to PSPMTs. In total, the system has nearl y 18,000 cr ystal elements and 52 million response lines for med between all cr ystal elements. The bore diameter is 16.0 cm with axial and transaxial F oVs of 4.9 and 8.5 cm, respecti vely. The microCT system comprises a 50 kVp, 1.5 mA f ixed tungsten anode X-ray tube, with 70 µm focal spot size, and 5 × 5 cm2 position-sensitive photodetector comprising a 48 µm pix el CMOS array and a fast gadolinium oxysulfide (GOS) intensifying screen.60 The system is mounted on a flexible C-arm gantry design with adjustab le detector positioning and is integrated in series (on the back) with the microPET II scanner (Figure 6). 110,111 Although the inte gration of PET and CT is in series, and onl y sequential acquisitions are performed, there is no de gradation of either modality’s performance compared to their stand-alone operation.

The pre viously described LabPET scanner de veloped by the group from Sherbrooke61 and commercialized by Gamma Medica-Ideas, Inc. comprises detector modules made from tw o rows of four cr ystals, one of LYSO and the other of LGSO, each with dimensions of 2 × 2 × 10 mm 3. The system has been enhanced b y adding adv anced X-ra y CT capability for acquiring anatomical images using the PET detectors. That is, the X-rays from a micro-focus source are collected using the same detector cr ystals and electronics to enab le simultaneous PET/CT scanning (~1 mm reconstr ucted spatial resolution for both). The claimed advantages of truly simultaneous PET and CT acquisitions are (1) simplified and more accurate re gistration because the subject does not ha ve relative motion betw een the tw o modalities, (2) reduced footprint compared to a PET/CT with separate detection subsystems, (3) the FoV is more accessib le for inter ventions during imaging, and (4) the system w ould cost less because an expensive, separate high-resolution CT detector system and gantry are not required. Because PET already uses

microCT microPET II

Figure 6. The UC Davis high-resolution small animal positron emission tomography (PET)/computed tomography (CT) system uses the series approach to multimodality system integration with the PET in front and the CT in back. Reproduced with permission from Liang H et al.60

Instrumentation and Methods to Combine Small Animal PET with Other Ima ging Modalities

photon-counting electronics, the same operating mode is used for the CT acquisition. This can be achieved by sampling the analog signal waveform using high-speed analog-to-digital con verters and digital processing in field pro grammable gate ar rays. The possibility to count and ener gy discriminate indi vidual X-ra y photons in CT mode39,61 provides foreseen benefits over the standard integration mode used in most X-ra y CT system designs, such as (1) a lo wer relati ve contribution from electronic noise that can be discriminated via a low-energy signal threshold and (2) a more appropriate energy w eighting independent of the X-ra y photon energy transmitted through tissue compared to the proportional w eighting of inte grating detectors that are typically used in X-ray CT. These factors could imply a lower dose required to obtain a gi ven image quality . The parallel architecture and f ast digital processing electronics allo w high count rates (25% deadtime or count loss at an e vent detection rate of 1.5 × 106 cps) for both PET and CT modes, whereas the modularity of the system design allo ws one to e xtend the number of channels up to 10 4 or more.

PET/SPECT/CT

There have been efforts to combine three modalities (PET, SPECT, and CT) and record quasi-simultaneous, complementary information gathered from each. One such example is the FLEX Triumph™ system commerciall y available from Gamma Medica-Ideas, Inc. (Nor thridge, CA).63 Proof-of-principle examples of its tri-modal imaging capabilities are shown in Figure 7 for normal mice. It has been claimed that the LabPET APD-based detector module proposed in 64 is another good technolo gy for the design of compact tri-modality (PET/SPECT/CT) imaging systems. The YAP-(S)PET scanner50 can perform both PET and SPECT studies on small animals. 106 Operating the scanner in SPECT mode is possib le using the same detector conf iguration b y mounting a high-resolution parallel-hole collimator in front of each detector panel and acquiring a standard circular orbit acquisition of photons. Although these multimodality systems require separate acquisitions of the PET , SPECT, and/or CT, there are no mutual interference ef fects. At this point in time, it is unclear w hat impor tant biomedical applications will require both PET and SPECT measurements, but it will be likely involve the need for a measurement of two complementary molecular signatures of disease in a living subject that is onl y a vailable with a separate PET and SPECT tracer and image acquisition.

109

PET/MRI

PET data pro vide high molecular sensiti vity. Magnetic resonance (MR) data provide a rich variety of anatomical and ph ysiological contrast mechanisms, through a wide range of a vailable pulse sequences, to study disease processes without using ionizing radiation. Thus, there has been considerab le interest to combine the tw o technologies65–85 and the topic has been w ell reviewed.33,74,75,77 The majority of MR pulse sequences involve relatively long acquisitions compared to CT. Thus, developing a combined system with a PET system in series with an MR system, with sequential acquisitions similar to most PET/CT configurations, would be impractical due to long acquisition times.As a result, unlike most system designs that combine PET and CT , nearl y all designs to inte grate PET with MR under in vestigation enable truly simultaneous (temporally and spatially registered) acquisitions of both PET and MR data. In all designs to date, the PET system inser t resides outside of the radio frequenc y (RF) coils and inside the g radient coils as depicted in Figure 8. Aside from the wide v ariety of multiparameter , dynamic contrast mechanisms a vailable with MRI, combining PET with MR has a couple of signif icant benefits over PET/CT. MRI does not use ionizing radiation. Dose reduction is especiall y impor tant for patients that are undergoing multiple repeated studies for follo w-up over time. Also, in soft tissue, MR demonstrates better visualization of contrast differences (a.k.a. contrast resolution). There are se veral impor tant challenges that must be overcome in designing and operating a combined PET/MRI imaging system. As we saw in the previous sections, nearl y all small animal PET detector designs proposed to date use PMTs w hose perfor mance can be seriously affected in the presence of extremely strong magnetic f ields produced b y moder n MRI scanners. Fur thermore, an MRI scanner relies on rapidl y switching gradient magnetic fields and RF signals to produce the MR image. The presence of the magnetic f ield gradients and RF signals cer tainly could disr upt the perfor mance of a PMT based PET detector if the y w ere located within or e ven adjacent to the magnet of the MRI system. Similarl y, the operation of the MRI system relies on a very homogenous, uniform, and stab le magnetic f ield to produce the MR image. The introduction of radiation detectors, electronics, and other bulk materials can per turb the main magnetic field, RF pulses, and g radient f ields in a w ay that introduces artifacts in the MR image. Early studies to design MR-compatib le PET units were made b y the Uni versity of Minnesota. 112–114 In the

110

MOLECULAR IMAGING: PRINCIPLES AND PRACTICE

CT/PET/SPECT

Figure 7. (Left) Tri-modality computed tomography (CT)/positron emission tomography (PET)/single photon emission computed tomography (SPECT) image of a normal mouse acquired with a single, integrated CT-PET/SPECT system—the FLEX Pre-clinical Imaging System by Gamma Medica-Ideas49 that contains separate CT, PET, and SPECT subsystems. The normal mouse was injected initially with 18F sodium fluoride and imaged with both the CT and PET subsystems. Subsequently, it was injected with 99mTc-methylene diphosphonate (MDP) through a catheter (without moving the animal from the bed) and imaged immediately after injection. The mouse skeletal structure is displayed with surface rendering of the CT image (cream color). The PET image (shown in green) displays bone uptake and clearance through the bladder. The SPECT imaging (shown in orange) was completed within a short time period compared to the bone uptake and clearance, thus the image shows the subject’s vasculature, heart, kidneys, and liver (which has a high blood volume fraction). This image indicates that MDP is still in the blood stream and also shows an early phase of clearance through the kidneys. (Middle and Right) PET/CT images of normal mice acquired with a single, integrated CT-PET-SPECT system—the FLEX Triumph Pre-clinical Imaging System by Gamma Medica-Ideas equipped with CT and LabPET subsystems. The LabPET system uses a fundamentally different PET detector design than the standard FLEX PET subsystem (different scintillation crystals and avalanche photodiode [APD] photodetectors rather than photomultiplier tubes [PMTs]). The middle image is a PET-CT overlay of a mouse injected with fluorodeoxyglucose (FDG), showing myocardial metabolic activity. The right image is a mouse injected with 18F sodium fluoride displayed in overlaid volume rendering. The integrated design of the FLEX Triumph scanner facilitates co-registration of the two image volumes using a pre-calibrated spatial transformation that enables the PET image dynamic range to match that of the CT images. Courtesy of Koji Iwata, Gamma Medica-Ideas.49

following years University of Califor nia, Los Angeles (UCLA)65–67 developed the f irst system conf igured as a 5.6 cm diameter, single scintillation cr ystal ring inser t for a 1.5 T MR system. The single ring comprised 72 2 × 2 × 10 mm3 LSO crystals each coupled side ways to 4 m long optical f ibers that were read out b y a single PSPMT coupled to readout electronics, both located outside the system bore. By k eeping the radiation-sensiti ve elements of the detector within the MR system, w hile operating the photodetector and electronics a way from the magnetic f ield, the combined system could perfor m simultaneous PET/MR imaging without measurab le mutual interaction effects.66 A main drawback for this design is that coupling to fibers results in loss of a signif icant fraction of the light signal, affecting energy and time resolution perfor mance and de grading the ability to identify w hich cr ystal

absorbed a photon. In par ticular, the side ways manner in which the scintillation cr ystals were coupled to the f ibers in67 as well as light attenuation within the f ibers yielded low scintillation light collection ef ficiency and e vent energy/time dispersion, resulting in an energy resolution of 45% FWHM at 511 keV and coincidence time resolution of 26 ns FWHM. Another main drawback is that the single crystal ring and data acquisition design resulted in extremely poor photon sensiti vity (~10 −4). Ne vertheless, the indi vidual cr ystals w ere resolv ed suf ficiently b y the PSPMT to form a single image slice with 2.1 mm FWHM spatial resolution near the system center and produced the first simultaneously acquired PET and MR images. 67 Collaborators at Guys and St. Thomas Hospital, London placed the UCLA-built system inside of a 9.4 T nuclear magnetic resonance (NMR) spectrometer to

Instrumentation and Methods to Combine Small Animal PET with Other Ima ging Modalities

PET Insert Gradient Coils

RF Transmit and Receive Coil

Main Super Conducting Magnet

Figure 8. In nearly all integrated positron emission tomography (PET)/magnetic resonance (MR) system designs, the PET system insert is placed outside the radio frequency (RF) transmit/receive coils and inside the gradient coils. Courtesy of Peter Olcott, Stanford University.

study metabolism in an isolated , perfused rat heart model. 32P-NMR spectra from phosphor ylated glucose were acquired simultaneousl y with PET images of 18 F-fluorodeoxyglucose (FDG) uptak e in the myocardium68,69 and the tw o data w ere compared for consistency gi ven the kno wn biochemistr y of glucose metabolism. The group planned to extend this MR-compatible f iber-coupled PET scanner concept to de velop another single ring of 480 LSO crystals arranged in three layers (160 cr ystals per layer) for photon DOI measurement capability, with a ring diameter of 11.2 cm, and a 5 cm diameter useful FoV that is large enough to accommodate an animal within a stereotactic frame.70 The chosen fiber length of 3.25 m resulted in roughl y 70% light loss70 (other such long-f iber-coupled designs had up to ~90% light loss85). However, it was argued that the number of scintillation photons w as suf ficiently abo ve the noise le vel of the PSPMT to deter mine the cr ystal of interaction within the detector block, although it is clear that energy and time resolutions, and therefore contrast resolution and quantif ication capabilities w ould be diminished significantly with the long f iber coupling. Other subsequent approaches based on PMT -based PET detectors used either the same f iber-coupling-based principles79,80 or relied upon more comple x magnet designs, including a split magnet 83 or a f ield-cycled MRI85 and ha ve been re viewed else where.74,75,77 In the latter case, the PET and MR acquisitions are interlea ved between the MR f ield cycling, and thus the two data sets are not acquired simultaneously. New PET system designs have arisen in recent years due to the availability of new semiconductor photodetector technolo gies and other adv ances that for PET/MR

111

allowed one to a void using PMTs and reduce the length of or eliminate the need for optical f iber coupling. Se veral research g roups are cur rently investigating methods to integrate a PET system into an MRI scanner by designing detectors made from relatively nonmagnetic materials that can be placed within the magnetic field of an MRI or MR spectroscopy system. Here w e highlight a fe w PET systems conf igured with suitab le semiconductor photodetectors that are insensitive to magnetic f ields and can consequently be operated within an MR system for combined PET/MR imaging. Some g roups ha ve tested APDs within a high magnetic f ield and ha ve produced PET and MR images that appear to be free of se vere ar tifacts and distortion.71–78 Other g roups are studying Silicon PMTs (SiPMs),115–119 which are a relati vely new type of semiconductor photodetector (Figure 9) that show promise in the design of combined PET/MR scanners; in addition to being insensitive to magnetic f ields, these de vices have much lar ger signals than APDs, approaching that of PMTs (hence the name), and lower operating bias, which relaxes readout electronics requirements, and thus facilitates operation inside an MRI system. 116–119 The UCLA design w as improved upon at UC Da vis (in collaboration with Califor nia Institute of Technology [Caltech]) by using shor ter, bent optical f ibers and position-sensitive APDs (PSAPDs) (Figure 10) to replace the long optical f ibers and PSPMTs, respecti vely. The PET system inser t consists of one 60 mm diameter ring of detector modules, each comprising an 8 × 8 array of 1.43 × 1.43 × 6 mm 3 LSO cr ystals (1.5 mm pitch) coupled through a 10 cm long f iber-optic b undle that bends 90 degrees from the cr ystal rod ends do wn the axis and away from the sensiti ve region of the MR system to the PSAPDs. The PSAPDs are read out b y low noise charge sensitive preamplifiers that drive long coaxial cables to a

Figure 9. Compact silicon photomultiplier (SPM) array (left) configured into 4 × 4 pixels each with 3 × 3 mm2 area and comprising 3640 Geiger-mode avalanche photodiode cells (right) with a 35 micron active area and 42 micron pitch. This is a promising new photodetector technology for PET scintillation detector arrays operating inside a strong magnetic field. Courtesy of Joe O’Keeffe, SensL USA (Mountain View, CA).

112

MOLECULAR IMAGING: PRINCIPLES AND PRACTICE

data acquisition system residing outside the MR system. The useful axial and transaxial F oV are 12 and 35 mm, respectively. The entire inser t is encased in copper for electromagnetic shielding with an outer diameter of 11.8 cm, f itting tightly within the gradient coil set. The UC Davis/Caltech group have performed a number of tests of the PET system inser ted in the Bruker 7-T Biospec animal MRI system operating with spin-echo and g radient-echo pulse sequences that are commonl y used for small animal MRI studies. Since the scintillation crystals and f ibers are nonparamagnetic and nonconductive, and all front-end electronic components are located beyond the sensiti ve region of the MR system, this system design has v ery little effect on the MR perfor mance and vice versa.76 PET perfor mance compromises made in the UC Davis/Caltech PET system inser t design are the shor t (6 mm length) cr ystals and small (12 mm) axial F oV, which considerably compromise PET photon sensiti vity, and the relati vely poor scintillation light detection SNR and energy/time dispersion resulting from the light loss from f iber bending and attenuation do wn the f iber. The measured single cr ystal-fiber ener gy and coincidence time resolution before system constr uction w as > 24% FWHM at 511 keV and > 5 ns FWHM, respectively. Due to a position-dependent delay of crystal-fiber location on 8 mm A

C

Scintillation Light flash Corner contacts

(X,Y)

A, B, C, D

B

D

X

(A B) (C D) A B C D

Y

(A C) (B D) A B C D

Figure 10. Position-sensitive avalanche photodiode (PSAPD) is another promising photodetector technology that can operate inside a strong magnetic field. Light from a small crystal impinging on the PSAPD (as depicted in blue) creates charge that is collected by four corner contacts (A, B, C, D) on the back side of the device as well as a common signal contact on the top of the device. From these signals one may accurately estimate, for each incoming 511 keV photon event, the x-y coordinate of the center of the light flash (using the positioning logic shown) as well as the energy and arrival time of the event. The PSAPD is used in the small animal positron emission tomography (PET) insert described in reference76.

the PSAPD , after system constr uction the coincidence timing window used w as 40 ns, leading to an increased random coincidence rate. With the cur rent detector configuration and ring packing geometr y, it would be a challenge to e xtend the axial F oV fur ther. But with the general scintillator -fiber coupled PET detector scheme used, increasing the axial F oV w ould lik ely still not substantially alter MRI performance. The UC Davis/Caltech group has also be gun studying applications of combined PET/MR studies using dual-PET and MR molecular probes to tar get and image microphages involved in stenotic plaques in rat models of arterial injury.77 On the PET side, the probe molecule 64 Cu-labeled ligand designed to bind is based on a specifically to certain receptors on microphages. On the MR side, they have used polymer-based agents containing gadolinium w hich interact directl y with w ater protons to shor ten T1 relaxation times, resulting in an increase in signal intensity compared to nonlabeled neighboring tissues. This approach is capab le of high dynamic range of signals because the y produce positi ve contrast. They ha ve also produced nanopar ticle-based agents containing iron o xides that shor ten T2 relaxation times through a magnetic f ield ef fect, resulting in decreased signal intensity . This contrast mechanism is generally more sensiti ve, producing measurab le ef fects at lower MR probe concentrations. 77 The University of Tuebingen developed an MR compatible PET detector that does not use f iber-optic coupling.71,72 They later developed an improved system design (Figure 11) that uses 10 detector modules each comprising a 12 × 12 LSO scintillator ar ray (1.6 × 1.6 × 4.5 mm 3) coupled through a light diffuser to a 3 × 3 APD array, and custom charge sensitive pre-amplif ier electronics, located within a thin copper shielded housing.73–75 This design has an adv antage in ter ms of scintillation detector SNR because the crystals are coupled to the photodetector without using optical f ibers. The multiring PET scanner inser t has an axial and transaxial F oV of 19 and 40 mm, respectively, and generates 23 tomo graphic slices with a slice thickness of ~0.8 mm. The signals emanating from the 10 detector modules are read out and processed with dedicated PET electronics (Siemens Pre-clinical Solutions, Knoxville, TN). The housing was designed in such a w ay to limit potential interference with basic components of a 7 T ClinScan animal MRI scanner (Br uker BioSpin MRI, Ettlingen, Ger many) including the magnetic f ield, f ield gradients, and RF recei ver/transmitter electronics. The cylindrical PET shell is inser ted to f it inside the g radient coil set and outside the 35 mm diameter RF coil of the MRI scanner, which has a 30 cm bore diameter.

Instrumentation and Methods to Combine Small Animal PET with Other Ima ging Modalities

B

A PET Insert Gradient Set

RF-Coil

C 7 Tesla Magnet ClinScan

PET Detector Module

LSO crystal block Amplifier and electronics and APD array

Figure 11. (A) Depiction of the University of Tuebingen positron emission tomography (PET) insert for the 7T Bruker BioSpin magnetic resonance imaging (MRI) system, as well as pictures of (B) the insert by itself and (C) individual copper-shielded scintillation detector array module used to build the 19 mm axial and 40 mm transaxial field-of-view (FoV) insert. Reproduced with permission from Judenhofer MS et al.73

The sensiti ve F oV of the Tuebingen combined PET/MRI conf iguration is 19 mm in the axial direction (limited by the PET inser t) and 35 mm in the transaxial direction (limited b y the RF coil). This FoV is adequate to image an entire mouse brain/hear t, or e ven lar ge tumors on rodents. It has been sho wn that there is onl y minor mutual interference betw een PET and MRI w hen operated simultaneously even when using more demanding MR sequences lik e echo planar imaging (EPI) for functional magnetic resonance imaging (fMRI).72,73 Moreover, the g roup showed that NMR spectroscop y is 73 The feasible in parallel with PET data acquisition. group also in vestigated cancer imaging applications of simultaneous PET/MR imaging. In one study using a mouse model of colon carcinoma, 73 the g roup used a PET cell proliferation mark er, 18F-fluoro-thymidine, T1-weighted MR, and dynamic image acquisition to help differentiate acti ve malignant cells from necrosis and inflammation. The authors argued that this differentiation was improved by the temporally correlated PET and MR data acquired dynamically. The main perfor mance compromises of the Tuebingen PET system inser t, as with the UC Da vis system, is the relatively low photon sensitivity due to shor t crystals and small (< 2 cm) axial F oV. Also, with the cur rent detector design it is unclear how to expand the axial FoV while a voiding fur ther de gradation of the MR data because the number of electronic components, amount of shielding, and co-axial signal transmission cabling inside the sensitive MR region would increase.

113

A Brookhaven National Laboratory (BNL) group has developed an MR-compatible PET insert prototype based on the technology used for the most recent version of the Rat Conscious Animal PET (RatCAP), a complete 3D tomograph designed to image the brain of an awake rat.42 This system conf igures the PET system into a highl y compact ar rangement of LSO/APD ar rays with the help of custom-made, highl y inte grated electronics b uilt at BNL.81,82 The PET system comprises a 4 cm diameter detector ring containing 12 b lock detectors, each conf igured as a 4 × 8 array of 2.3 × 2.3 × 5 mm 3 LSO crystals read out with a matching array of APDs. As with the Tuebingen system, the PET detector modules are positioned just outside the RF pickup coil of the MRI scanner and inside the gradient coils. The plan is to configure the system to f it within a radial distance of ~3 to 4 cm outside the RF coil and arranged to have a fairly small number of cables e xiting from the bore of the MRI scanner . The team designed a special RF pickup coil comprising tw o orthogonal Helmholtz coils that f it inside the RatCAP and enab le compensation for residual magnetic f ield effects or eddy cur rents produced by the presence of the PET insert. As with the other designs, the cr ystals in the BNL design are shor t and the axial F oV is nar row, leading to poor photon sensitivity. There is clearl y still a need for inno vative, highperformance small animal PET system designs that can be inserted into MR systems. The ideal small animal PET insert for an MR system would have uncompromised specifications, similar to the ideal stand-alone PET system: a large axial FoV (eg, > 10 cm) and thick (e g, > 2 cm long) crystals to be ab le to image an entire rat with high photon sensitivity (eg, ≥ 15%), and superb spatial resolution (eg, ≤ 1 mm FWHM, unifor m throughout the sensitive F oV), energy resolution (e g, ≤ 13% FWHM at 511 k eV), and coincidence time resolution (e g, ≤ 3 ns FWHM). These performance specif ications are not y et possib le with current designs. It is clear that there is a need to develop an MR-compatible PET detector design that combines the superior scintillation detector SNR a vailable with direct scintillation crystal-photodetector coupling with the minimal ef fect on MR perfor mance ensured with f iber-optic signal transmission. Such a design is under development.120 PET/Optical

For completeness, w e briefly discuss attempts to inte grate PET with optical imaging systems that use light-emitting molecular probes, a subject of focus of a later chapter. Sensitive, cooled char ge-coupled de vice (CCD) cameras which detect emitted light from fluorescent/bioluminescent

114

MOLECULAR IMAGING: PRINCIPLES AND PRACTICE

probes within a living organism have shown their potential in tracking pro gression of disease in murine models. 2,121 The goal of simultaneousl y recording radionuclide and optical signals is being pursued b y several g roups86–89,121 and offers the possibility of measuring multiple molecular based processes concurrently on the same living subject. In the combined PET-optical (OPET) system under constr uction at UCLA, 87–89 the scintillation cr ystal ar ray plays the dual role of coupling the optical signal from bioluminescence emitted from the animal to the photodetector as w ell as channeling optical scintillations created from the annihilation photon interactions to the photodetectors. The PET component comprises a he xagonal conf iguration of six detector blocks with an inner system radius of 15.6 mm. Each detector consists of a 2D ar ray of 8 × 8 GSO scintillation crystals each with a 2 × 2 mm 2 cross-sectional area and varying lengths from ~8 mm in the ar ray center to ~10 mm near the array edge that create a concave shape for each array to couple directly to the rodent to collect the bioluminescence light as well as the PET 511 keV annihilation photons emitted from the animal. Unlik e all other scintillation detectors used in PET, the ends of the crystals of the OPET detectors that are in contact with the animal are open and not covered with reflector material to let the bioluminescence light into the detectors. Special electronics were built to integrate the optical signal emitted from the animal over a desired frame duration, but also r un at the same time in single pulse processing mode for the scintillation light produced b y the 511 k eV annihilation photons. The crystal ar rays are read out by multichannel PMTs. 87–89 Extensive detector studies and Monte Carlo simulations w ere perfor med to study the feasibility of the concept and assess the ef fect of v arious geometric parameters on the perfor mance characteristics of the system. 87,88 There are a fe w performance compromises of note for the design of the PET component of the OPET system. The first is signif icant detector SNR de gradation because at least one half of the scintillation signal is lost due to the absence of the top reflector on the cr ystal ar rays. Other compromises are the relati vely short (< 10 mm long) GSO crystals, lo w ef fective Z (59) and density (6.71 g/cc) of GSO, and narrow axial FoV. Regarding the bioluminescence imaging capabilities, perfor mance is compromised due to the use of the scintillation crystals as bioluminescence light guides into the PMTs and their relatively low spatial resolution compared to the superior SNR and high spatial resolution available with a cooled, lens-coupled CCD imager used in standard in vi vo bioluminescence imaging. The OPET system is reviewed in more depth in Chapter 9, “Fiber Optic Fluorescence Imaging” of this book.

CONCLUDING REMARKS To date, there are numerous small-animal PET system designs either commercially available or under investigation and this remains an acti ve field of research. To date, most small-animal PET system designs tend to compromise other performance parameters in favor of achieving high (eg, < 2 mm) spatial resolution at the system center. Furthermore, there are se veral hybrid designs that combine tw o or more small-animal imaging modalities, including PET/CT , PET/SPECT , PET/MRI, and PET/optical. To date, h ybrid designs to inte grate PET hardware with that of other modalities (e g, MRI) with truly simultaneous operation in mind lead to fur ther performance trade-offs due to compromises, such as f ibercoupling, short crystals, and narrow axial FoV inherent in the design as w ell as perfor mance de gradations due to inter-modality interference during operation. Discovering compelling applications that moti vate truly simultaneous, temporall y cor related imaging with PET and other modalities, as opposed to separate acquisitions and softw are fusion of image data, is also an acti ve field of research. F or e xample, the ideal applications for simultaneous PET/MR imaging are y et to be deter mined but will lik ely in volve dynamic acquisitions and e xploit temporal correlations between molecular information provided with PET and the v ariety of multiparameter , dynamic contrast mechanisms available with MRI. Assuming such compelling applications for simultaneous acquisitions arise, it is clear that fur ther ef fort is required to develop integrated system designs that do not compromise performance of the imaging modalities in either the basic PET design specif ications or during multimodal data acquisition. Thus, there is still room for inno vative instrumentation designs.

ACKNOWLEDGMENT The author w ould lik e to ackno wledge Dr . Virginia Spanoudaki, currently at Stanford University, for her help in reviewing this chapter.

REFERENCES 1. Gambhir SS. Molecular imaging of cancer with positron emission tomography. Nat Rev Cancer 2002;2:683–93. 2. Massoud TF, Gambhir SS. Molecular imaging in living subjects: seeing fundamental biolo gical processes in a ne w light. Genes De v 2003;17:545–80. 3. Haubner R, Weber WA, Beer AJ, et al. Noninvasive visualization of the acti vated αvβ3 integrin in cancer patients b y positron emission tomo graphy and [F-18]galacto-RGD . PLoS Med 2005;2:244–52.

Instrumentation and Methods to Combine Small Animal PET with Other Ima ging Modalities

4. Vavere AL, Welch MJ. Preparation, biodistribution, and small animal PET of Ti-45-transferrin. J Nucl Med 2005;46:683–90. 5. McQuade P, Martin KE, Castle TC, et al. Investigation into 64Cu-labeled bis(selenosemicarbazone) and bis(thiosemicarbazone) complexes as hypoxia imaging agents. Nucl Med Biol 2005;32:147–56. 6. Robinson MK, Doss M, Shaller C, et al. Quantitati ve immunopositron emission tomography imaging oil HER2-positive tumor xenografts with an iodine-124 labeled anti-HER2 diabody . Cancer Res 2005;65:1471–8. 7. Lyons SK. Advances in imaging mouse tumour models in vi vo. J Pathol 2005;205:194–205. 8. Waldherr C, Mellinghoff IK, Tran C, et al. Monitoring antiproliferative responses to kinase inhibitor therap y in mice with 3 ʹ′-deoxy3ʹ′-18F-fluorothymidine PET. J Nucl Med 2005;46:114–20. 9. Thakur ML, Aruva MR, Gariep y J, et al. PET imaging of onco gene overexpression using 64Cu-vasoactive intestinal peptide (VIP) analog: comparison with 99mTc-VIP analo g. J Nucl Med 2004; 45:1381–9. 10. Aliaga A, Rousseau J A, Ouellette R, et al. Breast cancer models to study the expression of estrogen receptors with small animal PET imaging. Nucl Med Biol 2004;31:761–70. 11. Barthel H, Wilson H, Collingridge DR, et al. In vivo evaluation of [F18]fluoroetanidazole as a ne w mark er for imaging tumour hypoxia with positron emission tomo graphy. Br J Cancer 2004; 90:2232–42. 12. Sundaresan G, Yazaki PJ, Shively JE, et al. 124I-labeled engineered anti-CEA minibodies and diabodies allo w high-contrast, antigen-specific small animal PET imaging of xenografts in athymic mice. J Nucl Med 2003;44:1962–9. 13. Solomon B , McAr thur GA, Cullinane C, et al. Applications of positron emission tomo graphy in the de velopment of molecular targeted cancer therapeutics. BioDrugs 2003;17:339–54. 14. Yang HH, Berger F, Tran C, et al. MicroPET imaging of prostate cancer in LNCAP-SR39TK-GFP mouse x enografts. Prostate 2003;55:39–47. 15. Ray P, Pimenta H, P aulmurugan R, et al. Nonin vasive quantitati ve imaging of protein-protein interactions in living subjects. Proc Natl Acad Sci U S A 2002;99:3105–10. 16. Luker GD, Sharma V, Pica CM, et al. Noninvasive imaging of proteinprotein interactions in li ving animals. Proc Natl Acad Sci U S A 2002;99:6961–6. 17. Adonai N, Nguyen KN, Walsh J, et al. Ex vivo cell labeling with 64Cupyruvaldehyde-bis (N4-methylthiosemicarbazone) for imaging cell trafficking in mice with positron-emission tomo graphy. Proc Natl Acad Sci U S A 2002;99:3030–5. 18. Wang J , Maurer L. P ositron emission tomo graphy: applications in drug discovery and drug development. Curr Top Med Chem 2005; 5:1053–75. 19. Frangioni JV, Hajjar RJ. In vivo tracking of stem cells for clinical trials in cardiovascular disease. Circulation 2004;110;3378–83. 20. Cao F, Lin S, Xie X, et al. In vivo visualization of embryonic stem cell survival, proliferation, and mig ration after cardiac deli very. Circulation 2006;113:1005–14. 21. Blasberg RG, Tjuvajev JG. Molecular -genetic imaging: cur rent and future perspectives. J Clin Invest 2003;111:1620–9. 22. Pither R. PET and the role of in vi vo molecular imaging in personalized medicine. Expert Rev Mol Diagn 2003;3:703–13. 23. Shah K, Jacobs A, Breakefield XO, Weissleder R. Molecular imaging of gene therapy for cancer. Gene Ther 2004;11:1175–87. 24. Luker GD. Special conference of the American Association for Cancer Research on molecular imaging in cancer: linking biolo gy, function, and clinical applications in vi vo. Cancer Res 2002; 62:2195–8. 25. Wu JC, Inubushi M, Sundaresan G, et al. Positron emission tomography imaging of cardiac repor ter gene e xpression in li ving rats. Circulation 2002;106:180–3.

115

26. Jacobs A, Voges J, Reszka R, et al. P ositron-emission tomography of vector-mediated gene e xpression in gene therap y for gliomas. Lancet 2001;358:727–9. 27. Inubushi M, Wu JC, Gambhir SS, et al. P ositron-emission tomography reporter gene e xpression imaging in rat m yocardium. Circulation 2003;107:326–32. 28. Hnatowich DJ. Observations on the role of nuclear medicine in molecular imaging. J Cell Biochem 2002;87:18–24. 29. Chakrabarti A, Zhang K, Aruva MR, et al. Radioh ybridization PET imaging of KRAS G12D mRNA expression in human pancreas cancer x enografts with [64Cu]DO3A-peptide nucleic acid-peptide nanoparticles. Cancer Biol Ther 2007;6:1–9. 30. Ingvar M, Eriksson L, Ro gers GA, et al. Rapid feasibility studies of tracers for positron emission tomography: high resolution PET in small animals with kinetic anal ysis. J Cereb Blood Flo w Metab 1991;11:926–31. 31. Levin CS. Primer on molecular imaging technolo gy. Eur J Nucl Med Mol Imaging 2005;32:S325–45. 32. Levin CS. New imaging technologies to enhance the molecular sensitivity of positron emission tomography. Proc IEEE 2008; 96:439–67. 33. Levin CS. Cur rent trends in pre-clinical PET system design. PET Clinics 2007;2:125–60. 34. Bloomfield PM, Rajeswaran S, Spinks TJ, et al. The design and physical characteristics of a small animal positron emission tomo graphy. Phys Med Biol 1995;40:1105–26. 35. Cherry SR, Shao Y, Silverman RW, et al. MicroPET: a high resolution PET scanner for imaging small animals. IEEE Trans Nucl Sci 1997;44(Pt 2):1161–6. 36. Seidel J, Vaquero JJ, Green MV. Resolution uniformity and sensitivity of the NIH Atlas small animal PET scanner: comparison to simulated LSO scanners without depth of interaction capability. IEEE Trans Nucl Sci 2003;50(Pt 2):1347–50. 37. Surti S, Kar p JS, P erkins AE, et al. Design e valuation of A-PET: a high sensiti vity animal PET camera. IEEE Trans Nucl Sci 2003;50(Pt 2):1357–63. 38. Uribe J, Wong W-H, Li H, et al. Signal characteristics of indi vidual crystals in a high resolution BGO detector design using PMT quadrant sharing. IEEE Trans Nucl Sci 2003;50:355–61. 40. Motta A, Damiani C, Del Guerra A, et al. Use of a fast EM algorithm for 3D image reconstr uction with the YAP-PET tomo graphy. Comput Med Imaging Graph 2002;26:293–302. 41. Ziemons K, Auffray E, Barbier R, et al. The ClearPET (TM) project: development of a 2nd generation high-perfor mance small animal PET scanner. Nucl Instrum Methods Phys Res A 2005;537:307–11. 42. Vaska P, Woody CL, Schl yer DJ, et al. RatCAP: miniaturized headmounted PET for conscious rodent brain imaging. IEEE Trans Nucl Sci 2004;51:2718–22. 43. Wu HY, Pa D, O’Sullivan JA, Tai YC. A feasibility study of a prototype PET inser t device to convert a general-pur pose animal PET scanner to higher resolution. J Nucl Med 2008;49:79–87. 44. Jeavons A, Kull K, Lindberg B, et al. A proportional chamber positron camera for medical imaging. Nucl Instr um Methods 1980; 176:89–97. 45. Spanoudaki VC, McElroy DP, Torres-Espallardo I, Zie gler SI. Ef fect of temperature on the performance of proportional APD-based modules for gamma ra y detection in positron emission tomo graphy. IEEE Trans Nucl Sci 2008;55:469–80. 46. Knoxville (TN): Siemens Pre-clinical Solutions. 47. Andover (MA): Philips Medical Systems. 48. Madrid, Spain: GE-Suinsa. 49. Northridge (CA): Gamma Medica-Ideas. 50. Pisa, Italy: Ingegneria dei Sistemi Elettronici. 51. Straubenhardt, Germany: Raytest GmbH. 52. Oxford, UK: Oxford Positron Systems. 53. Levin CS, Hoffman EJ. Calculation of positron range and its effect on the fundamental limit of positron emission tomo graphy system

116

54.

55.

56.

57.

58.

59.

60. 61.

62.

63.

64.

65.

66.

67. 68.

69. 70.

71.

72.

73.

74. 75.

76.

MOLECULAR IMAGING: PRINCIPLES AND PRACTICE

spatial resolution [cor rigendum: Ph ys Med Biol 2000;45:559]. Phys Med Biol 1999;44:781–99. Levin CS. Design of a high resolution and high sensitivity scintillation crystal array for PET with nearl y complete light collection. IEEE Trans Nucl Sci 2002;45:2236–43. Surti S, Kar p JS, Perkins AE, et al. Imaging perfor mance of A-PET: a small animal PET camera. IEEE Trans Med Imaging 2005;24:844–52. Wang Y, Seidel J, Tsui BM, et al. P erformance evaluation of the GE healthcare e xplore VISTA dual-ring small animal PET scanner . J Nucl Med 2006;47:1891–900. Jeavons AP, Chandler RA, Dettmar CAR. A 3D HIDAC-PET camera with sub-millimetre resolution for imaging small animals. IEEE Trans Nucl Sci 1999;46:468–73. Schafers KP, Reader AJ, Kriens M, et al. P erformance evaluation of the 32-module quadHID AC small animal PET scanner . J Nucl Med 2005;46:996–1004. Goertzen AL, Meadors AK, Silverman RW, Cherry SR. Simultaneous molecular and anatomical imaging of the mouse in vi vo. Ph ys Med Biol 2002;47:4315–28. Liang H, Yang Y, Yang K, et al. A microPET/CT system for in vi vo small animal imaging. Phys Med Biol 2007;52:3881–94. Bérard P, Riendeau J, Pepin C, et al. Investigation of the LabPETTM detector and electronics for photon-counting CT imaging. Nucl Instrum Methods Phys Res A 2007;571:114–7. Parnham KB , Cho wdhury S, Li J , et al. Second-generation, trimodality pre-clinical imaging system. IEEE Nucl Sci Symp Conf Rec 2006;3:1802–5. Patt B , P arnham K, Li J , et al. FLEX: tri-modality small animal tomography combining PET, SPECT and CT in a single modular gantry [abstract]. J Nucl Med 2005;46:207. Saoudi A, Lecomte R. A novel APD-based detector module for multimodality PET/SPECT/CT scanners. IEEE Trans Nucl Sci 1999; 46:479–84. Farahani K, Slates R, Shao Y, et al. Contemporaneous positron emission tomography and MR imaging at 1.5 T. J Magn Reson Imaging 1999;9:497–500. Slates R, Farahani K, Shao Y, et al. A study of artefacts in simultaneous PET and MR imaging using a prototype MR comptatible PET scanner. Phys Med Biol 1999;44:2015–27. Shao Y, Cherry SR, Farahani K, Meadors K. Simultaneous PET and MR imaging. Phys Med Biol 1997;42:1965–70. Garlick PB, Marsden PK, Cave AC, et al. PET and NMR dual acquisition (P ANDA): applications to isolated perfused rat hear ts. NMR Biomed 1997;10:138–42. Marsden PK, Strul D, Keevil SF, et al. Simultaneous PET and NMR. Br J Radiol 2002;75:S53–9. Mackewn JE, Str ul D, Hallett WA, et al. Design and de velopment of an MR-compatible PET scanner for imaging small animals. IEEE Trans Nucl Sci 2005;52:1376–80. Pichler BJ, Judenhofer MS, Catana C, et al. P erformance test of an LSO-APD detector in a 7-T MRI scanner for simultaneous PET/MRI. J Nucl Med 2006;47:639–47. Judenhofer MS, Catana C, Sw ann BK, et al. Simultaneous PET/MR images, acquired with a compact MRI compatible PET detector in a 7 Tesla magnet. Radiology 2007;244:807–14. Judenhofer MS, Wehrl HF , Ne wport DF , et al. Simultaneous PET/MRI: a ne w perspecti ve for functional and mor phological imaging. Nat Med 2008;14:459–65. Pichler BJ , Judenhofer MS, Wehrl HF . PET/MRI h ybrid imging: devices and initial results. Eur Radiol 2008;18:1077–86. Pichler BJ , Wehrl HF, K olb A, Judenhofer MS. P ositron emission tomography/magnetic resonance imaging: the ne xt generation of multimodality imaging? Semin Nucl Med 2008;38:199–208. Catana C, Wu Y, Judenhofer MS, et al. Simultaneous acquisition of multislice PET and MR images: initial results with a MR-compatib le PET scanner. J Nucl Med 2006;47:1968–76.

77. Cherry SR, Louie AY, Jacobs RE. The integration of positron emission tomo graphy with magnetic resonance imaging. Proc IEEE 2008;96:416–38. 78. Catana C, Procissi D, Wu Y, et al. Simultaneous in vivo positron emission tomo graphy and magnetic resonance imaging. Proc Natl Acad Sci U S A 2008;105:3705–10. 79. Raylman RR, Majewski S, Velan SS, et al. Simultaneous acquisition of magnetic resonance spectroscop y (MRS) data and positron emission tomography (PET) images with a prototype MR-compatib le, small animal PET imager. J Magn Reson 2007; 186:305–10. 80. Raylman RR, Majewski S, Lemieux SK, et al. Simultaneous MRI and PET imaging of a rat brain. Ph ys Med Biol 2006;51:6371–9. 81. Woody C, Schlyer D, Vaska P, et al. Preliminary studies of a simultaneous PET/MRI scanner based on the RatCAP small animal tomograph. Nucl Instrum Methods Phys Res A 2007;571:102–5. 82. Schlyer D, Rooney W, Woody C, et al. De velopment of a simultaneous PET/MRI scanner . IEEE Nucl Sci Symp Conf Rec 2004; 6:3419–21. 83. Handler WB, Gilbert KM, Peng H, Chronik B A. Simulation of scattering and attenuation of 511 k eV photons in a combined PET/field-cycled MRI system. Phys Med Biol 2006;51:2479–91. 84. Lucas AJ, Hawkes RC, Ansorge RE, et al. De velopment of a combined microPET -MR system. Technol Cancer Res Treat 2006;5:337–41. 85. Yamamoto S, Takamatsu S, Murayama H, Minato K. A block detector for a multislice, depth-of-interaction MR-compatib le PET. IEEE Trans Nucl Sci 2005;52:33–7. 86. Peter J, Semmler W. A modular design triple-modality SPECT -CTODT small animal imager [abstract]. Eur J Nuc Med Mol Imaging 2007;34:S158. 87. Rannou FR, K ohli V, Prout DL, Chatziioannou AF. In vestigation of OPET perfor mance using GA TE, a Geant4-based simulation software. IEEE Trans Nucl Sci 2004;51:2713–7. 88. Alexandrakis G, Rannou FR, Chatziioannou AF. Tomographic bioluminescence imaging b y use of a combined optical-PET (OPET) system: a computer simulation feasibility study . Phys Med Biol 2005;50:4225–41. 89. Vu NT, Silverman RW, Chatziioannou AF. Preliminar y perfor mance of optical PET (OPET) detectors for the detection of visib le light photons. Nucl Instrum Methods Phys Res A 2006;569:563–6. 90. Maintz JB, Viergever MA. A sur vey of medical image re gistration. Med Image Anal 1998;2:1–36. 91. Pluim JP, Maintz JB, Viergever MA. Mutual-information-based registration of medical images: a sur vey. IEEE Trans Med Imaging 2003;22:986–1004. 92. Cross DJ, Minoshima S, Nishimura S, et al. Three-dimensional stereotactic surface projection anal ysis of macaque brain PET : development and initial applications. J Nucl Med 2000;41:1879–87. 93. Shimada Y, Uemura K, Ardekani BA, et al. Application of PET-MRI registration techniques to cat brain imaging. J Neurosci Methods 2000;101:1–7. 94. Vaquero JJ, Desco M, P ascau J, et al. PET, CT, and MR image re gistration of the rat brain and skull. IEEE Trans Nucl Sci 2001; 48:1440–5. 95. Humm JL, Ballon D , Hu YC, et al. A stereotactic method for the three-dimensional registration of multi-modality biolo gic images in animals: NMR, PET, histology, and autoradiography. Med Phys 2003;30:2303–14. 96. Jan M-L, Chuang K-S, Chen G-W , et al. A three-dimensional re gistration method for automated fusion of micro PET -CT-SPECT whole-body images. IEEE Trans Med Imaging 2005;24:886–93. 97. Fei B, Wang H, Muzic RF Jr , et al. Defor mable and rigid re gistration of MRI and microPET images for photodynamic therapy of cancer in mice. Med Phys 2006;33:753–60. 98. Deroose CM, De A, Loening AM, et al. Multimodality imaging of tumor x enografts and metastases in mice with combined small animal PET , small animal CT , and bioluminescence imaging. J Nucl Med 2007;48:295–303.

Instrumentation and Methods to Combine Small Animal PET with Other Ima ging Modalities

99. Hill DLG, Batchelor PG, Holden M, Ha wkes D. Medical image re gistration. Phys Med Biol 2001;46:R1–45. 100. Slomka PJ. Software approach to mer ging molecular with anatomic information. J Nucl Med 2004;45:S36–45. 101. Chow PL, Stout DB, Komisopoulou E, Chatziioannou AF. A method of image re gistration for small animal, multi-modality imaging. Phys Med Biol 2006;51:379–90. 102. Chow PL, Rannou FR, Chatziioannou AF. Attenuation correction for small animal PET tomographs. Phys Med Biol 2005;50:1837–50. 103. Kinahan PE, Townsend DW, Beyer T, Sashin D. Attenuation correction for a combined 3D PET/CT scanner. Med Phys 1998;25:2046–53. 104. Townsend DW, Beyer T. A combined PET/CT scanner: the path to true image fusion. British Jour nal of Radiology 2002;75:S24–30. 105. Yang Y, Cherry SR. Observations regarding scatter fraction and NEC measurements for small animal PET. IEEE Trans Nucl Sci 2006; 53:127–32. 106. Del Guerra A, Bartoli A, Belcari N, et al. Performance evaluation of the fully engineered YAP-(S)PET scanner for small animal imaging. IEEE Trans Nucl Sci 2006;53:1078–83. 107. Kiessling F, Greschus S, Lichy MP, et al. Volumetric computed tomography (VCT): a ne w technolo gy for nonin vasive, high-resolution monitoring of tumor angiogenesis. Nat Med 2004; 10:1133–8. 108. Boone JM, Velazquez O, Cherry SR. Small animal X-ra y dose from micro-CT. Mol Imaging 2004;3:149–58. 109. Andre MP, Spivey BA, Martin PJ, et al. Inte grated CMOS-selenium X-ray detector for digital mammo graphy. Proc SPIE 1998; 3336:204–9. 110. Tai Y-C, Chatziioannou A, Yang Y, et al. MicroPET II: design, development and initial perfor mance of an improved microPET scanner for small animal imaging. Phys Med Biol 2003;48: 1519–37. 111. Yang Y, Tai YC, Siegel S, et al. Optimization and perfor mance evaluation of the microPET II scanner for in vi vo small animal imaging. Phys Med Biol 2004;49:2527–45.

117

112. Hammer BE, Christensen NL, Heil BG. Use of a magnetic f ield to increase the spatial resolution of positron emission tomo graphy. Med Phys 1994;21:1917–20. 113. Christensen NL, Hammer BE, Heil BG, Fetterly K. Positron emission tomography within a magnetic f ield using photomultiplier tubes and light-guides. Phys Med Biol 1995;40:691–7. 114. Raylman RR, Hammer BE, Christensen NL. Combined MRI-PET scanner: a Monte-Carlo e valuation of the impro vements in PET resolution due to the ef fects of a static homo geneous magnetic field. IEEE Trans Nucl Sci 1996;43:2406–12. 115. Dolgoshein B, Balagura V, Buzhan P, et al. Status repor t on silicon photomultiplier de velopment and its applications. Nucl Instr um Methods Phys Res A 2006;563:368–76. 116. Moehrs S, Del Guer ra A, Herber t DJ, Mandelkern MA. A detector head design for small animal PET with silicon photomultipliers (SiPM). Phys Med Biol 2006;51:1113–27. 117. Spanoudaki VC, Mann AB, Otte AN, et al. Use of single photon counting detector ar rays in combined PET/MR: characterization of L YSO-SiPM detector modules and comparison with a LSO-APD detector. J Instr 2007;2:P12002(1–16). 118. Herbert DJ, Saveliev V, Belcari N , et al. F irst results of scintillator readout with silicon photomultiplier . IEEE Trans Nucl Sci 2006; 53:389–94. 119. Otte AN, Barral J, Dolgoshein B, et al. A test of silicon photomultipliers as readout for PET . Nucl Instr um Methods Ph ys Res A 2005;545:705–15. 120. P.D. Olcott, H. Peng, and C.S. Levin. Novel Electro-Optical Coupling Technique for Magnetic Resonance–Compatib le P ositron Emission T omography Detectors. Molecular Ima ging, Vol 8, No 2 (March–April, 2009): pp. 74–86 121. Park JM, Gambhir SS. Multimodality radionuclide, fluorescence, and bioluminescence small animal imaging. Proc IEEE 2005; 93:771–83.

8 FUNCTIONAL IMAGING USING BIOLUMINESCENT MARKERS CHRISTOPHER H. CONTAG, PHD

OVERVIEW Biomedical tools based on optical methods are widely used for sensing and imaging in research and in the clinic; recent advances in optics ha ve led to a ne w generation of optical tools for biomedical imaging that enab le imaging of molecular and cellular processes in li ving subjects. The clinical utility of optical molecular imaging is conf ined to a narrow set of applications due to the limited penetration of light through mammalian tissues and the small number of approved contrast agents. Ho wever, optical imaging is ideally suited for the study of small laborator y animals. 1,2 As such, optical imaging has contributed considerab ly to the study of mammalian biolo gy and will continue to uncover biological mechanisms and accelerate drug development. In pre-clinical studies, the in vestigator has the opportunity to genetically manipulate transplanted cells, or those of the host, and therefore reporter genes can be incorporated into animal models with both specif icity and genetic control. This has led to an e xpansive set of tools based on optical reporter genes that can advance the study of animal models of human biolo gy and disease. Among the optical imaging tools used in small animals, in vi vo bioluminescence imaging (BLI) has had, and will continue to ha ve, a signif icant role in the f ield of molecular imaging.3–5 BLI is based on the use of optical reporter proteins called luciferases, and the genes that encode these light emitting enzymes are being de veloped as repor ter genes for use in mammalian subjects. 6–9 In addition, the proteins encoded by these genes can be conjugated directly to other targeting and sensing entities and used as repor ter proteins, or fusion proteins comprised of a luciferase and a targeting entity can be created geneticall y.10–13 The versatility of BLI presents a number of unique oppor tunities, and given the tremendous signal to noise ratios that can be 118

achieved, can pro vide sensiti ve in vi vo bioassa ys that reveal the location and magnitude of a wide range of biological processes. This is e videnced b y the ability to build informative animal models by incorporating reporter genes into target cell DNA or by creating unique repor ter conjugates through linking luciferases to tar geting molecules, such as antibodies for use as opticall y-based detectors.10,11,14 Creation of genetic constr ucts encoding fusion proteins comprised of coding sequences for the targeting antibody and that of a luciferase can be used as molecular probes in vivo.10,11 Understanding the properties of the v arious luciferases (T able 1), the parameters that control transmission of light through mammalian tissues, and the required cof actors for light emission are essential for selecting the appropriate repor ter for a gi ven application and for creating the most infor mative bioassay. A number of repor ter genes ha ve been de veloped for BLI and the proteins the y encode can be designed with specificity for a given biological process leading to markers that respond with optical signals to changes in the labeled process. These signals can be detected using lo wlight imaging systems that are external to the bodies of the study subjects. The repor ter genes used in BLI encode light-emitting enzymes, luciferases, and those that ha ve been used for BLI include genes from ter restrial and marine or ganisms that encode enzymes with emission peaks from 490 to 620 nm (seeTable 1). As with all modalities that operate in the visib le and near -infrared (NIR) region of the spectrum, the principles of tissue optics apply to the in vi vo detection of luciferase acti vity. Ho wever, there are some unique considerations including expression levels in target cells and the background emission from live animals that are unique to the in vi vo detection of these bioluminescent repor ters. The relati ve opacity of tissue permits only limited transmission of visib le light through

Functional Imaging Using Bioluminescent Markers

119

Table 1. CHARACTERISTICS OF LUCIFERASES THAT HAVE BEEN USED IN VIVO a

Luciferase

Species of Origin

Peak Emission [nm]

Substrate Used

Protein Size [kD]

Subcellular Localization

Energy Source

Cofactors

Citations

562/612

Luciferin

61

Cytoplasmic

ATP

O2b, Mg2+

Zhao et al.49

Pyrophorus plagiophthalamus (click beetle)

615

Luciferin

61

Cytoplasmic

ATP

O2

Wood et al.41,104,209 Zhao et al.49

CBLucGr68

Pyrophorus plagiophthalamus

543

Luciferin

61

Cytoplasmic

ATP

O2

Wood et al.41,104,209 Zhao et al.49

RLuc

Renilla reniformis (sea pansy)

480

Coelenterazine

36

Cytoplasmic Substrate

O2

Zhao et al.49

Rluc-mod

Renilla reniformis

547

Coelenterazine

36

Cytoplasmic Substrate

O2

Loening et al.48,110,111

Gluc

Gaussia princeps (copepod)

480

Coelenterazine

19.9

Extracellular Substrate

O2, Na+

Tannous et al.45 Wiles et al.210 Verhaegent and Christopoulos211

Aequorin (ALuc)

Aequorea victoria (jellyfish)

469

Coelenterazine

Cytoplasmic Substrate

O2, Ca2+

Shimomura212

Luxc

Photorhabdus luminescens (bacterium)

490

Decanal

Intracellular

O2

Fisher et al.213

Fluc

Photinus pyralis (firefly beetle)

CBLucred

77

FMN

nm = nanometers; kD = kilodaltons (kd). a Emission peak (λmax) measured at 37°C. b All luciferases characterized to date are oxygenases, however, some require cofactors, such as magnesium, calcium or sodium ions. c The lux genes are encoded on a five gene operon. The heterodimeric luciferase is encoded by two genes, Lux A and Lux B, which produce a 37 and 40 kd protein, respectively.

the tissues of mammals, and although largely attenuated by both absorption and scattering, light in the visible and NIR region of the spectr um has been successfull y used for imaging of biological process in living subjects.15,16 Bioluminescent signals are typically collected noninvasively using macro-optics pro viding a w hole-body image such that detectable signals from any tissue of the animal can be localized , however microscopic detection is also possib le. However, the relati vely weak signals of luciferase reactions and the inability to easil y counter stain tissues and retain enzymatic acti vity mak e microscopic detection of bioluminescent signals in tissue sections less than ideal. To o vercome this limitation bioluminescent repor ters are often combined. Bioluminescent repor ters are often combined with fluorescent reporter genes to create dual or multifunctional repor ters that can be detected with multiple methods for increased data per study and/or for v alidation of one repor ter with the other. To provide multiple methods of detection and for validation. Fluorescence is particularly well suited for

microscopic detection w here the collecting lens is a microscope that is placed on e xposed tissues, and fluorescence microscopes with bulk optics have been used to detect fluorescence in li ving animals; this method is called intravital microscopy.17,18 In addition, less invasive miniaturized microscopes ha ve been used in both mice and man to detect fluorescent signals, and these endomicroscopic tools are increasing in v ersatility and capability.19,20 Thus, bioluminescence and fluorescence are complementary modalities pro viding optimal macroscopic and microscopic detection. Macroscopic imaging with bioluminescence can provide information relative to when and w here to look with microscopic detection or with procedures that require anal yses after biopsy or necropsy. As such image guidance serves to refine animal models and optimize the data obtained from pre-clinical studies. The versatility of BLI has led to its use in di verse f ields, and a number of inno vations have increased the range of assa ys that can be perfor med in vivo. Recent technological advances include dual enzyme

120

MOLECULAR IMAGING: PRINCIPLES AND PRACTICE

assays for the study of tw o processes simultaneousl y,21 bioluminescence resonance ener gy transfer (BRET) for wavelength shifts and tagging cells,22 modified substrates for altered enzyme acti vity,23 and split luciferases for analyses of protein-protein interactions in li ve animals.24–27 As we are increasing the types of data that can be obtained in vi vo, BLI will continue to increase our understanding of mammalian biolo gy, ref ine preclinical studies, and accelerate the development of new therapeutic approaches. BLI has already become an essential tool for conducting predicti ve and infor mative studies of small animal models of human biolo gy and disease and responses to therap y, and with increased numbers of applications in industr y and academics its impact will continue to expand.

IMAGING RODENT MODELS OF HUMAN BIOLOGY AND DISEASE BLI is primaril y a preclinical imaging modality used mostly in laboratory rodents, and so in thinking about BLI one must f irst ask the question, “Wh y image a mouse?” There are two basic answers to this question. The first and perhaps the most ob vious answer is to use animal models for the de velopment of clinicall y relevant tools w here the tools themselves are f irst evaluated in animal models with the intent of translating the method or instr ument to the clinic. The second ans wer to the question, “Wh y image a mouse?” is the generation of ne w knowledge about mammalian systems and pathophysiology that can be translated to the clinic. This new infor mation can be used to ref ine clinical studies or to pro vide new insights that open ne w areas of clinical investigation. For the purpose of developing and testing of imaging instr umentation and imaging probes for clinical use, BLI has limited translational capability, with onl y v ery fe w potential niche applications in the clinic. Therefore, the v ast majority of preclinical BLI imaging studies are not aimed at building tools based on bioluminescence that translate to the clinic, but rather these studies are designed to v alidate potential therapeutic targets, test ne w compounds that tar get the basis of disease, and de velop deli very tools that can car ry estab lished or experimental compounds to the target site. The use of BLI to evaluate compounds and methods that can be translated to the clinic comprises the lar gest number of the studies using luciferases to report biological functions. In a g rowing number of studies, BLI is used to re veal basic features of mammalian biolo gy and biolo gical responses to insult. How imaging with bioluminescence can be used to test therapeutic approaches and to create opportunities for new clinical studies are the focus of this chapter .

Prior to the de velopment of imaging tools for small animals, animal studies lar gely involved serial sacrif ice to obtain temporal data. This meant that predeter mined time points needed to be selected and assa ys were performed on tissues from predeter mined locations. This effectively biases these studies b y linking the study design to a priori kno wledge and preconcei ved notions. As such this type of study has, in some respects, an inherent circularity—the e xperiment is being perfor med to learn about a bi ological process, b ut one needs to suf ficiently understand the process to appropriatel y select the times and tissues to study . Animal e xperimentation has used this approach for decades and although it has provided essential information, we now have better tools and can approach biolo gical questions with less bias. By applying molecular imaging tools to the study of animal models, we can remo ve some of the biases because the entire body can be imaged and the temporal resolution is within the rele vant range of seconds, minutes, or hours. Because imaging studies can remove many of these limitations and can serve to guide the investigator to specific times and tissues, e xperiments can be designed that reveal ne w features of disease processes and pro vide additional infor mation about a biolo gical process. 28,29 Because preclinical molecular imaging tools enab le realtime data acquisition using dynamic measures of biolo gical function, the animal e xperiments are more informative and the data sets are more complete. Moreover, given that imaging obviates serial sacrif ice, studies are not limited to assa ys on tissue samples obtained at necropsy where each time point is represented b y a different group of animals. Therefore, imaging reduces error in a given study, decreases the number of animals needed, and enab les the possibility of identifying outliers and individual variation. Prior to the widespread use of preclinical imaging tools, outliers w ere not studied and this important source infor mation w as not accessib le. The outliers may be due to e xperimental error, or may reveal some unique biological mechanism, and imaging enables making this distinction. By understanding e xperimental error and remo ving variability inherent in studies w here each time point is represented by a different group of animals, the statistics of animal studies are considerab ly improved. Real-time access to cellular and molecular infor mation in intact tissues and organs of animal models enables the researcher to obser ve mechanisms of action or cascading events within the animal that w ould not otherwise be detected using conventional methods. Imaging modalities that are already used clinicall y are being used in preclinical studies; however, BLI and other optical imaging tools w ere de veloped specif ically for the study of

Functional Imaging Using Bioluminescent Markers

small-animal models. 30–34 The use of visib le light of fers the advantages of rapid and high-throughput measures of biological function using relati vely low cost instr umentation yet with tremendous sensitivity. Because of their ease of use, optical imaging methods will continue to ha ve an impact on dr ug studies w here it is necessar y to e valuate disease process in models with intact biological pathways that interact with the potential therapies. The complexity of the re gulatory networks in disease processes needs to be studied in the context of intact organs and living tissues and cannot be readil y modeled in culture. Because the range of estab lished animal models that ha ve been w ell characterized is signif icant and optical imaging strategies can be superimposed on e xisting animal models, the use of imaging in preclinical studies is expanding rapidly, and new models that incor porate optical repor ter genes for imaging are being developed.

121

A

Light

B

ATP

D-Luciferin

ADP 1 Pi

Oxyluciferin Mg21 , O2 Luciferase

C

BASIC CONCEPTS AND METHODS Generating Signals and Sources of Noise BLI has been widel y adopted as a molecular imaging modality for preclinical studies o ver a wide range of disciplines largely because of its v ersatility, sensitivity, and accessibility. There are a number of ar ticles and re views describing the use of this approach to study mammalian biology, disease mechanisms, and therapeutic response.3,4,35–37 The light sources used in BLI are lightemitting enzymes, all of w hich are o xygenases and thus require oxygen in addition to a chemical substrate, generally referred to as luciferin (F igure 1), to produce optical signals. Because neither the enzymes nor most of the substrates produce signals alone, or in sera, backg round signals are minimal and a s witch occurs w hen the enzyme and substrate pairs come to gether to produce the signal, also ser ving to k eep the noise at a minimum. The luciferases from marine organisms use coelenterazine as a substrate and this high-ener gy molecule pro vides energy to the reaction such that additional cellular ener gy sources, such as adenosine triphoshate (A TP), are not needed. However, the autocatal ysis of coelenterazine can be a source of autoluminescence; ie, noise. The signals from these tw o enzymes are bright but not deepl y penetrating in mammalian tissues (see section “Enzyme Reporters”), so at the superf icial sites, the signal-to-noise ratios are still quite good. BLI is an e xtremely sensiti ve imaging modality in laboratory rodents due to the extraordinary signal-to-noise ratios that can be achieved,38 given the absence of noise in the reporter system and that naturally occurring sources of

HO

N

S

S

N

COOH

280.33 g/mol D-Luciferin: D-(-)-2-(6’-hydroxy-2’benzothiazolyl)thiazoline-4-carboxylic acid Figure 1. Firefly luciferase and its substrate. Bioluminescence reactions are the source of signal in bioluminescence imaging and the most commonly used luciferase is that from the North American firefly (A). This reaction (B) requires substrates and cofactors that link it to the cellular metabolism. The substrate for this reaction is D-luciferin (C) and the pharmacology of this molecule permits in vivo use with excellent signal-to-noise ratios.

light in mammalian tissues are infrequent39 or nonexistent. Signal-to-noise ratios approaching 10,000 can be obtained in BLI, and despite the signals from bioluminescent sources in the body being relati vely weak and the loss of signal due to absorbance b y mammalian tissues e xtreme, the absence of noise mak es these signals detectable, even using detectors that are external to the animal’s body. Unlike fluorescence, bioluminescence does not require an e xcitation light source and therefore e xcitation of naturall y occur ring fluors is absent in BLI. In contrast, autofluorescence is a signif icant source of noise in in vi vo fluorescence imaging, 38 thus e ven though fluorescent signals can be e xtremely intense when e xcited with a bright e xcitation source, the autofluorescence also scales with excitation intensity, and reduces signal to noise ratios. Autofluorescence is especially prob lematic at shor ter w avelengths of light (300–600 nm) in fluorescence imaging. At

122

MOLECULAR IMAGING: PRINCIPLES AND PRACTICE

longer wavelengths of light (> 600 nm) absorbance, autofluorescence decreases dramaticall y, and the scatter is also diminished but not as signif icantly. Because of the optical proper ties of mammalian tissue, optical imaging probes with longer w avelengths of emission are less attenuated than those with shor ter emission, and in all areas of optical imaging contrast agents and probes that function at the longer w avelengths are being sought.

Enzyme Reporters Luciferases used in BLI have been derived from a variety of different marine and terrestrial organisms (see Table 1) and have been e xtensively modif ied with adaptations to mak e them suitable as repor ter genes in mammalian cells. 6,9,40,41 These modifications include codon optimization for mammalian expression and elimination of sequences that tar get the luciferase to subcellular compar tments (peroxisomes42; or to the extracellular space).43–45 Increasing the wavelength of emission of the luciferases for use in BLI has onl y been moderately successful with a 615 nm peak for the longest emitting luciferases (both the click beetle [CBLuc] and firefly [FLuc] at 37°C). 9,46–48 All luciferases have broad emission spectra such that 60% of the emission for the longest emitting luciferases (FLuc and the red click beetle luciferase [CBLuc red]) is above 600 nm. 49 Transmission of light through mammalian tissue is most ef ficient at these longer wavelengths due to diminished absorbance of light by hemoglobin when wavelengths above 600 nm are used. This affects the signals that can be detected externally from internal bioluminescent sources (Figure 2). However, these broad emission peaks can present prob lems in scenarios where spectral resolution of multiple reporters would benefit the e xperiment; in such an e xperiment, the spectra are likely to overlap. Spectrally resolved imaging of luciferase emission enables localizing emitters of two different wavelengths in the body and this has been used in a limited number of studies. 50 Alternatively, because the chemistries are unique for some of the bioluminescent reporters, the use of luciferases with distinct substrates enables sequential imaging for localization of tw o markers.51 In this approach, the substrate that is cleared more rapidl y is used f irst followed by the second substrate. The broad emission spectra, ho wever, offers a solution to deter mining depth information by enabling a ratio of short to long wavelengths within a given emission spectra as a means of revealing depth of the source in the tissue. The use of light-emitting enzymes, that is, luciferases, and external imaging of the light transmitted through mammalian tissue was first demonstrated using bacteria labeled

through the e xpression of a bacterial repor ter gene, 32 and the versatility of this method has led to rapid adaptation to revealing tumor growth and response to therapy,52 stem cell engraftment and proliferation, 53 gene e xpression,33 and protein-protein interactions (See Chapter 47, “Molecular Imaging of Protein–Protein Interactions, ” b y Massoud et al.). 54–58 The enzyme from the Nor th American f irefly, FLuc, and its deri vatives have been used most commonl y. The emission of this luciferase is at 612 nm w hen assayed at 37°C making it among the longest emitting luciferases at mammalian body temperatures.49 Of the enzymes that have been examined for altered emission with changes in temperature, this w as the onl y one with a temperature shift where the λmax (emission peak) is at 560 nm at 22°C and 612 nm at 37°C—other luciferases appear to be temperature stab le.49 However, the enzymatic acti vity of man y luciferases is greater at 37°C than at room temperature. 49 BLI typicall y requires the use of geneticall y encoded markers of biolo gy, and this can be an adv antage because specif icity can be designed into the genetic constructs and the v ersatility of the repor ters is extreme given that an y number of genes can be tagged using a handful of a vailable bioluminescent repor ter genes. However, this is par t of the limitation to translation for BLI. Transfer of genetically engineered cells to humans can be justified for therapeutic genes, but use of bioluminescent repor ter genes is more dif ficult to justify given their limited use for deep tissue imaging. There are, however, bioluminescent fusions to antibodies that ma y ha ve clinical utility gi ven that the fusion protein is transfer red and not an engineered cell. Relative to other reporter gene strategies for imaging, BLI is a sensiti ve, rapid , relati vely ine xpensive approach for the study of laborator y animals and has high-throughput. Like other whole-body optical imaging approaches, it is a lo w-resolution imaging modality with limited areas of clinical translation. Luciferase reactions require oxygen and an ener gy source and therefore the signals are often tied to the metabolic activity of the tagged cell; this can be used to assess cellular function or tissue physiology,59 but it can also lead to v ariable data. Understanding these links to ph ysiology are essential for inter pretation of image data. Enzymes that use adenosine triphosphate (ATP) or other cellular cofactors as energy sources must be intracellular to produce signals—this includes the CBLuc and FLuc enzymes. Those enzymes that use coelenterazine as both the substrate and the ener gy source, Gaussia luciferase (GLuc) and Renilla luciferase (RLuc), are acti ve extracellularly and can therefore be used to tag antibodies and ligands for imaging.

Functional Imaging Using Bioluminescent Markers

A

B B

32

Open

500

520

24

540

2 1

560

580

600

1

620

8

C C

Lung

4

Liver

In Vitro

100

Skin

660

720

740

Lung

680

700

0

Open

Liver 5

100

4

80

3

60

3

2

40

2

1

20

1

0 400 450 500 550 600 650 700 750

0

0 400 450 500 550 600 650 700 750

60 40

CBGr68

4

5

100

4

80

4

60

3

60

3

40

2

40

2

20

1

20

1

0

0 400 450 500 550 600 650 700 750

0

0 400 450 500 550 600 650 700 750

100 80

Flu c 1

CBRe d

5

Molar Extinction Coefficient (cm2 1/M, 3 10000)

Normalized Light Intensity (A.U.)

640

5

h RLu c

80

20 0

Photons/sec (3 103)

28

S kin

123

Wavelength (nm) Figure 2. Transmission of bioluminescent light through murine tissues—the dependence of emission spectra and tissue depth on signal detection. Bioluminescent light emission was collected from four different sources in the skin, lung, or liver in the three animal models studied (A). The four luciferases were RLuc, FLuc, CBLuc,red and CBLuc.green68 A representative animal is shown for each model with the pseudocolor (Blue-Red) representing the bioluminescence data that is superimposed over a grayscale reference image showing the location and orientation of the animal. Spectral scanning was performed by collecting sequential bioluminescence images using an IVISTM200 loaded with 20 nm-band pass filters (filter wavelength indicated below each image); B, The open-filter images acquired before and after this spectral sequence are shown. These were used to calibrate the signal degradation during the scanning process. The entire sequence was acquired in less than five min to minimize signal deterioration. Spectra collected from skin, lung, and liver are compared with those from labeled cells in vitro (35°C); C, Photon fluxes are normalized to the values at 680 nm for beetle luciferases and 640 nm for Renilla luciferase (indicated by a green arrow). Hemoglobin (Hb) absorption curves (average values of oxy-Hb and deoxy-Hb) are plotted as background (shaded in gray). Reproduced with permission from Zhao H et al.49

The ability to study live cells in the context of living organs and live tissues in the body of fers new opportunities for examining pathologies—for example, because the cells are alive and biologically active, the molecular markers can be amplif ied through the e xpression of reporter genes 32,33,37,60–70; alter natively, amplif ication can be accomplished b y tar geting a probe to an 71–74 enzymatic acti vity that is intrinsic to the tissue. However, there are significant challenges to performing assays in vi vo, relative to standard histopatholo gy, that include the difficulty of getting a vital dye or molecular probe across biolo gical bar riers to the tar get tissue in concentrations that can produce a detectable signal, and

the inability to w ash of f unbound reagents for the purpose of reducing the noise in the image. These challenges are being met through interdisciplinar y approaches for code velopment of ne w chemical compounds, with phar macokinetic proper ties that are w ell suited for imaging, in conjunction with instruments and algorithms with impro ved sensiti vity and signif icant noise reduction. Because luciferases require a substrate to generate their optical signals, the phar macology of these substrates in mammals is relevant to interpretation of imaging data. The detection of optical signals in BLI is the focus of this section. Since enzyme acti vity is used as the readout for BLI it is essential that the

124

MOLECULAR IMAGING: PRINCIPLES AND PRACTICE

substrate be in e xcess, otherwise, the assa y will be measuring substrate le vels. At doses of 150 millig rams per kilogram body weight (mg/kg) of luciferin the substrate is generally in excess.

Detection Detection in BLI is typicall y planar and uses imaging systems that are sensiti ve to the w eak signals that escape the scattering and absorbing en vironment of the mammalian body. These systems are typically based on chargecoupled device (CCD) detectors fitted with optimal lenses and filters that operate in the visible to NIR regions of the 33 spectrum and collect the maximum amount of light. Because the detectors are sensiti ve and the signals are weak, any ambient light or noise generated b y the instr ument will reduce the image quality and sensiti vity. Therefore, the detectors are placed in light-tight chambers to exclude ambient light, and the materials in these chambers are selected for those with v ery little long-li ved fluorescence. Although most plastics ha ve a characteristic longlived fluorescence, appropriate materials for inside the chamber are limited. There are a variety of low-light imaging CCD architectures that include those that increase signal (amplified) or reduce noise (cooled); the uses of these cameras for BLI has been pre viously re viewed.10 The materials used for signal amplification can impose a spectral bias to ward the shor ter w avelengths, and although such cameras w ere used pre viously, most detectors currently being used for BLI preserve the spectral range of the CCD (roughl y 250–1100 nm). Back-thinned cooled CCDs are most widely used in the field, however, electromechanical cooling (EMC) detectors are gaining popularity due to f ast data collection and the ability to perfor m video rate imaging e ven if the signals are w eak. This is particularly impor tant for imaging using the calciumsensitive aquorin luciferase, ALuc, as an indicator of calcium fluxes in vivo. Devices used for BLI tend to be less e xpensive than those used in other imaging modalities, however, anatomic resolution is relati vely poor in w hole-body images and usually the reference images consist of a photograph of the subjects. When necessary a high-magnification lens can be directed at sites in the body w here labeled cells have been localized b y w hole-body imaging to produce highresolution images that complement the lo wer resolution images taken noninvasively of the w hole animal. The use of a high-magnif ication lens in li ving animals, intra vital microscopy, has been pub lished for a number of cell trafficking studies and tremendous insights ha ve been gained in this manner.75–80 The extreme attenuation of light, in the

visible and NIR regions of the spectrum, by human tissues will limit the translation of optical imaging modalities to specialized niches in the clinic, but since radioacti vity is not required and the instr uments are relatively lower cost, optical modalities are more accessib le and a vailable tools for the study of small-animal models.

Cellular and Molecular Biology as a Basis for in Vivo Optical Imaging Luciferase reporter genes ha ve been used in cell culture for several decades as assays for gene expression and for assessing the le vels of ATP, calcium, and other biolo gically relevant molecules. These assays were then used in thin and transparent organisms, and when initially used in larger animals had been assayed after necropsy or biopsy, in tissue homo genates. These studies ha ve pro vided a substantial amount of infor mation that can no w be used as w e transition from cells and tissue homo genates to imaging assa ys in li ving animals. The de velopment of luciferases for use in mammalian cell culture has led to well-developed luciferase repor ter genes that are expressed well in mammalian cells.44,45,49,81,82 In addition, large-scale sequencing ef forts and tremendous adv ances in high-density screens and gene e xpression studies have identified many key genetic elements in ph ysiology and disease that can be used to drive or control the expression of reporter genes as indicators of gene function. 83 These advances form the foundation on w hich in vivo BLI was built and continues to suppor t advancements in the f ield. Biological responses that comprise ph ysiology or pathophysiology typically consist of coordinated expression of multiple interrelated genes, proteins, and biological pathways, and the level of expression of a given gene and its protein product ma y be re gulated in this coordinated response depending on the interaction of multiple cell types. Because there are both resident and recr uited cells at the site of the biological response, differentiating between gene activation in the resident population versus an influx of cells already e xpressing a given gene is relevant to data inter pretation. Spatiotemporal patter ns of gene expression in dynamic environments are best e valuated through imaging, at least initiall y, as tissue str ucture and circulation can influence expression levels. The dynamic changes in e xpression can be monitored with imaging and then the infor mation can be used to guide the more labor -intensive assa ys on reco vered cells and tissues. Reconstr ucting the pathw ays and netw orks comprised of multiple single elements will require technological advances and a systems approach to biology be yond the capabilities of toda y’s imaging tools.

Functional Imaging Using Bioluminescent Markers

Such studies will require ef fective links betw een multiplexed ex vivo and in vi vo assays. Imaging approaches that can be used to monitor biochemical e vents in living cells, tissues, and in w hole organisms, are emer ging as an inte gral par t of systems biolo gy. Opticall y-based imaging methods will be at the forefront of these de velopments gi ven their ease of use, amenability to multiplexing, and their availability for biologists.

Reporter Genes Optimal repor ter genes for in vi vo studies ha ve se veral well-defined characteristics. They should encode w ellcharacterized gene products, proteins that can result in generation or accumulation of deepl y penetrating light emission with the potential for a high signal-to-noise ratio. The signal-to-noise ratio is dependent on le vels and location of repor ter gene e xpression and also on the optical properties of mammalian tissues.15 Once a reporter gene is expressed, signal detection is dependent on the absorbing and the scattering proper ties of the tissue, w hich ha ve a considerable influence on shor ter w avelengths of light.38,84,85 The versatility of BLI is due to the ability to engineer many different cell types and tag man y different genes with a small set of repor ter genes, but the variety of optical reporters is increasing and will lik ely contribute to increased numbers of application areas and uses of BLI in biomedical research and to the ability to multiple x these assays. The hardware for detection as described abo ve has been fairly well established with the promise of only incremental advances. However, we are only in the initial stages of probe development and advances in molecular contrast agents ha ve tremendous potential for di versifying and improving reporters used in BLI. By analogy to light microscopy where dyes are used to stain cells in tissues, adv ances in BLI will be in the area of new optical contrast that can mark cells and molecular processes in the body . Unlike histopathology, the inability to wash off unbound dye in vivo is a signif icant limitation. To address this issue, cells are typicall y labeled outside the body b y transfer of the repor ter gene prior to injection into an animal model and imaging either macroscopically or microscopically.18,86,87 Reporter genes encode reporter proteins that function as “labels” that can be traced noninvasively over time. The proteins interact with “repor ter probes,” applied substrates (luciferins) to generate a signal that can be initiall y localized from outside the body, and then these data can serve as a guide for selecting the appropriate tissues and times w here sampling will be most meaningful. Alternatively, genes can be transfer red to cells in vi vo, as in gene therap y

125

approaches, using viral and non viral gene transfer techniques. In vi vo gene transfer of luciferase genes serves as a model for de veloping new delivery tools and viral vectors for use in nucleic acid-based therapies, 88–91 and efficient gene transfer techniques are being used to deliver tar get genes for e valuation of RN A inhibition strategies for genetic control. 91–97 Because these reporter genes were initially developed for use in cell-based assa ys or for detection in e xcised tissues, after imaging their activity in live subjects, measurements of the reporter activity can also be made on e xcised tissue using the same optical signal that generated the in vivo image. As such the in vi vo and ex vivo measures are linked to each other and to cor relative cell culture assa ys (see section “Cor relative Cell Culture Assays”). In this manner, BLI becomes an iterati ve study of biolo gy using integrated assays that e xtend from biochemical assa ys to live animal models. Effective use of optical imaging therefore results in studies of animal models that are more informative and predicti ve than studies that onl y use conventional assays on excised tissues. For measurements of tumor burden or for assessing cell trafficking patterns, it is necessary to use genetic constructs consisting of a strong constitutively expressed promoter (such as those from the β actin, and ubiquitin c genes, or from vir uses, such as cytomegalovirus [CMV] and other sources) dri ving expression of a reporter gene, and these are integrated into the genome. Inte gration into the genome can be accomplished with viral v ectors, mammalian transposons or random integration with subsequent selection of cells with integrated repor ter genes. Inte gration circumv ents problems of loss of signal due to dilution from cell di vision as well as the confounding detection of signals that ha ve dissociated from the viab le tar get cell population. Despite these adv antages, these ectopicall y e xpressed repor ter genes can be silenced b y chromosomal modif ications. Additional de velopment in the area of optical repor ter genes will be motivated by the need for g reater sensitivity and specificity. Advances in understanding gene e xpression patterns as integrated sets using arrays of 30,000 elements can be used to indicate w hich promoters from w hich genes are best used to e xpress a repor ter designed mark a selected biological process. In this manner , the tar geting is at the level of transcription w here the cells are labeled b y targeting specif ic genetic re gulatory elements, that is, promoters, and linking them to sequences that encode reporter genes. Extensi ve use of this approach in cell biology has produced a wide ar ray of genetic elements from w hich to dra w from for molecular studies in animals. Furthermore, the transition from cell-based assa ys

126

MOLECULAR IMAGING: PRINCIPLES AND PRACTICE

to in vi vo assa ys has been enab led b y the adv ances in imaging instr umentation specif ically designed for small animals that are readil y a vailable. Repor ter genes ha ve also been developed that are re gulated posttranscriptionally and can be constitutively expressed without expressing an optical signal until another specif ic molecular event occurs (see section “Cor relative Cell Culture Assays”). In this case, the activation of the reporter is typically posttranslational.3,54–58,98,99

Correlative Cell Culture Assays Validation of in vi vo measurements of optical repor ters can be performed on excised tissues and in culture using the same repor ters that are used in vi vo. These sets of assays have shared roots in molecular and cell biolo gy. Correlative cell culture assays can be used as a means of developing an imaging strategy prior to in vivo use of the imaging approach and then can be used as a follo w-up assay for v alidation. The instr uments that detect optical signals from inside the animal can typicall y accommodate cell culture plates and thus even the detectors are the same. Imaging of cell culture cor relates can also be used to design probes and test their v alidity. The v alidation studies are conducted using biochemical measures on tissue lysates using well-established measures of enzyme activity, messenger RN A, and protein le vels.88,100 The data obtained in vivo has tremendous utility for selecting the times and tissues for anal yses that can be selected based on the image data or cell numbers, and this image guidance can be used to impro ve the data set b y conf ining the study to the rele vant times and tissues. The new advances in imaging are based on adv ances in cellular and molecular biolo gy, and the tools of these f ields are essential components of the f ield of molecular imaging and of BLI.

BIOLUMINESCENT REPORTERS AND THEIR SUBSTRATES FOR IN VIVO USE Luciferase enzymes ha ve been found in a wide range of organisms from several different genera (see Table 1), but there are essentially three basic biochemistries among the enzymes characterized to date. 6 Each of the biochemistries uses dif ferent substrates and conditions, and each chemistr y has been used in BLI. Genes encoding members of the three classes of luciferases ha ve been cloned and their chemistries characterized to a point where they can be routinel y used in the cor relative cell culture assays and in BLI. These include the luciferases from beetles (coleoptera), jell yfish and sea pansies

(cnidaria), and bacteria ( Vibrio spp. and Photorhabdus luminescens). All luciferases are o xygenases and require energy and a specific luciferin, for light production, often in the presence of cof actors.

Luciferase Enzymes The luciferases from f ireflies and related insects are single polypeptides related to the CoA ligase f amily of proteins101,102 and use a benzothiazole luciferin substrate to generate light in the presence of ATP, magnesium, and oxygen. The gene encoding FLuc is the most commonly used bioluminescent repor ter and has been codon optimized for expression in mammalian cells. In addition, mammalian transcription f actor binding sites have been removed with the intent of pre venting inadvertent control of e xpression, and man y of the cr yptic splice donor and acceptor sites ha ve been remo ved to maintain e xpression. These modif ications ha ve been largely conducted b y the v endors of these genes (Promega Cor p, Madison, WI) and ha ve produced reporter genes with high levels of expression in the target cells and tissues and that represent e xpression patterns of targeted genes. 103 Other beetle luciferases ha ve also been codon optimized, including CBLuc. 41,47,104 These proteins are broad spectrum emitters with a signif icant red component of the spectr um. Luciferases from the sea pansy ( Renilla reniformis; RLuc), 105–107 the jell yfish ( Aequorea aequorea),58,108,109 and marine copepod ( Gaussia princeps; GLuc) 45 have also been cloned , characterized, and used as in vi vo repor ter genes. All the enzymes from marine organisms that have been described today use coelenterazine as their substrate. These three coelenterazineusing enzymes dif fer in aspects that gi ve them unique characteristics for interrogating biology in vivo. Selection of the appropriate reporter gene for a given study requires some understanding of their characteristics, the bioa vailability of their substrates, and their unique biochemistries. The greatest diversity of bioluminescent proteins is in the enzymes derived from marine organisms and it is unfortunate that all of these are b lue emitters and use the less desirable in vivo substrate, coelenterazine. Since shor ter w avelengths of light are lar gely absorbed b y hemo globin, the b lue emission from the luciferase deri ved from marine or ganisms is absorbed in vivo and the sensiti vity of detection is dramatically reduced. Attempts have been made to shift the emission spectr um of RLuc to ward the red with mutations and substrate optimization, 48,110,111 but the shifts have been modest. The g reatest adv antage of these

Functional Imaging Using Bioluminescent Markers

mutations is lik ely its g reater stability in vi vo. The emission of nati ve RLuc peaks at a w avelength w here there is a signif icant dip in hemo globin absor ption (at 490 nm), and the red-shifted mutants emit light at a peak in the hemo globin absor ption (betw een 510 and 600 nm). The biodistribution of coelenterazine in mammals is not optimal for imaging as it has a relati vely short circulation time in vivo and some associated autoluminescence.49,112 The relatively short circulation time, however, mak es it a prefer red compound for the f irst reaction in dual enzyme assays, as there is little residual activity detectab le after 10 minutes. Codon-optimized versions of the Renilla enzyme ha ve ser ved as a light source for self-illuminating quantum dots (Qdots) (below).13 The luciferases from Renilla and Gaussia (RLuc and GLuc) are e xtremely bright and do not require cof actors from the host cell; this opens up the possibility of monitoring e xtracellular events and using fusion proteins (e g, antibody-luciferase fusions) as imaging agents in animals and possib ly humans. 10,11,14

127

within a fe w seconds to a minute after injection. D-luciferin can be injected IP or IV .112 Substrate derivatives and conjugates of fer ne w oppor tunities for imaging,23,119,120 and a v ariety of ne w deri vatives ha ve been described. It has also been reported that D-luciferin and coelenterazine are substrates for multidr ug resistance (MDR) proteins that pump xenobiotics out of cells.121–123 The two compounds are subject to different mechanisms with coelenterazine being a substrate for P gl ycoprotein121,122 and D-luciferin123 for alternate pathways. These observations have been used to de velop assays to assess the le vels of xenobiotic pumps on cells and to de vise methods of circumventing multidrug resistance in cancer.121 The differences in emission spectra and biochemistries of the luciferases offer the potential for multiple xing assays in cell culture and in vi vo. At present tw o, and possib ly more, biolo gical processes can be studied simultaneously.86,120 Combining bioluminescence with other enzyme activities offers another layer for in vivo bioassay development.

Substrates and Their Bioavailability In BLI, to generate a signal the substrate must be present in the same subcellular compar tment or tissue site and there are a number of cellular and tissue bar riers that the substrate must cross. These substrates are much lik e drugs that need to reach their tar get, and biodistribution of the substrate controls signal intensity in BLI. Pharmacokinetics of D-luciferin and coelenterazine, relative to tissues and or gans, is an impor tant consideration in BLI. Luciferin appears to access cells in most if not all tissues and crosses the blood-brain and placental bar riers.88,89,113 D-luciferin is cleared slo wing with the peak le vels in most tissues at 15 to 30 minutes after intraperitoneal (IP) injection; this is w hen mice should be imaged. After intravenous (IV) injections, the D-luciferin seems to peak at 1 min and be cleared within 5 min, but does appear to be more e venly distributed and higher concentrations appear to be present in the central nervous system.114–117 The optimal time from administration to data acquisition in BLI depends upon both the route of administration and the rate of clearance of the substrate in the tar get tissue. In contrast to D-luciferin, coelenterazine is subject to rapid clearance rates. D-luciferin is relati vely stab le in the body and has a relatively long circulation time. 33 In contrast, the substrate for RLuc, coelenterazine, is rapidl y cleared from the body and binds to serum proteins.112 Therefore, imaging protocols using coelenterazine generall y require IV injection and data acquisition must be complete

Multifunctional Reporter Genes Combinations of reporter genes that encode gene fusions that produce multifunctional single proteins, 121,122 or that incorporate genetic elements that enable multiple proteins to be made from a single RN A transcript (pol ycistronic message 53,103,123), ha ve been used to create multimodality imaging strategies. These are used for the purpose of pro viding additional data to a study in the form of v alidation or to link macroscopic and microscopic imaging modalities for a more complete anal ysis. Such approaches can be used to provide different types of information from a single repor ter. The ease of linking genes through genetic tools has led to the development of a number of these reporters that can be detected by two or more modalities. 100,121,124 Multireporter gene fusions were f irst described using luciferase and g reen fluorescent protein (GFP) for use in cells and flies,125 and similar fusions ha ve been used as a means of connecting BLI measurements in mice, to e x vi vo assa ys such as flo w cytometry, and fluorescence microscop y using GFP and related fluorescent proteins. 100 Triple gene fusions ha ve been used to link optical imaging and e x vi vo assa ys to positron emission tomography (PET) imaging.121,122,124 These multifunction repor ter genes can considerab ly increase the fle xibility of a repor ter for in vi vo assa ys and perhaps even more importantly offer validation measures that are not possib le using a single function reporter.100,126,127

128

MOLECULAR IMAGING: PRINCIPLES AND PRACTICE

Gene Transfer A significant limitation to the use of repor ter genes is the transfer of these genes to cells in intact or gans of li ving mice with stab le inte gration and reliab le e xpression. Transformation of tumor cells with repor ter genes is generally straightforward and selection of transfor med cells with drug selection markers enables creation of stable cell lines. However, different cell types can require dif ferent methods and selection can tak e time and create cell v ariants with characteristics that differ from the parental line. In comparison, labeling primar y cells isolated from their host is more challenging, and transfer of genes to cells in intact or gans of li ving mice has remained a signif icant barrier despite adv ances in both non viral- and viralmediated gene transfer methods. One means of addressing this problem is to create transgenic mice that e xpress the reporter genes, single or dual function, with e xpression targeted to a specif ic cell type 128 or tissue. 129 These animals can be studied directl y or ser ve as cell and tissue donors for transplantation or cell trafficking studies.53 Transgenic donor mice that e xpress repor ter genes from strong constitutive promoters have been used for the study of trafficking and expansion of hematopoietic stem cells and immune cells 53,123 and the study of organ transplantation. By crossing these transgenic mice with transgenic lines of mice that ha ve been engineered to spontaneously develop malignancies under the control of doxycycline, the crossed lines can generate transplantable tumors. The onco genic potential of these transplanted cells can be controlled and the process of tumor g rowth, response to therap y, and relapse can be studied. 130 Studies of these transplanted tumors have revealed a persistent low-level signal after inacti vation of the onco gene that initiated the initial tumor growth.130 At the nadir when the disease burden w as minimal but detectab le, the labeled cells had a nor mal appearance by all measures, but after reactivation of the onco gene, these cells pro vided a source of relapse. The insight gained from this study was that in states of remission, the cells that persist ma y have stem cell characteristic giving rise to tissues with nor mal appearance in the absence of the onco gene or cancer when the onco gene is reacti vated. These insights w ere revealed b y the ability to monitor the entire disease course and use the images to guide tissue selection. In BLI, the repor ter genes are often inte grated into the mouse genome, and the signals are not lost during the course of the disease; this enab les long-ter m study of cells through each of the stages of malignanc y. The use of repor ter genes in BLI has made this modality ideall y suited for the de velopment of nucleic

acid-based therapies. By placing the reporter gene into the gene transfer vector or other gene delivery tool, the transfer of genetic material to the cells in tissues can be assessed.88,91,131–133 Conversely, with the repor ter gene in the cells as a transgene or temporaril y e xpressed gene, inhibitory nucleic acids, such as small inhibitor y RNAs (siRNA), can be deli vered and the reduction in signal measured as a reporter of effective delivery.92–94,134 In both of these scenarios, the inte grity of e xpression of the reporter gene is crucial to interpretation of the data. In all reporter gene imaging approaches, there is the potential of the promoter element to be silenced or otherwise modulated. This may be due to methylation of CpG dinucleotides in the promoter of to other epigenetic events.135,136 Promoters deri ved from vir uses are more likely to be silenced (e g, the immediate earl y promoter from CMV and that from simian vir us 40), than those derived from eukar yotic cells. Thus, in situations w here strong constituti ve e xpression is desired , the promoters from ubiquitin C and β actin are used. 53,123

BRET Bioluminescence is a biological source of light that can be manipulated and directed through genetic engineering and the gene products used for a v ariety of pur poses. One of these applications is BRET . BRET uses a bioluminescent luciferase that is geneticall y fused to a protein or peptide that can interact with a fluorescent moiety that can accept the photons from the luciferase and con vert them to a longer, lower energy signal.22 In the case of protein-protein interactions, the tw o fusion proteins bring the luciferase and the fluorescent protein close enough for resonance energy transfer to occur, leading to a signal s witch. BRET was first demonstrated in cell culture 22 and has since been used to assess protein-protein interactions in vi vo.137–140

Self-Illuminating Qdots One subset of BRET is the use of luciferases to e xcite Qdots. Qdots are particles of different sizes that fluoresce at dif ferent w avelengths using a common e xcitation wavelength, and this common w avelength matches the emission spectr um of RLuc. Thus, combining bioluminescence with Qdots has the adv antage of illuminating the dot at its surf ace to create a self-illuminating particle.13 Since the excitation light for Qdots is generally of short wavelength, by conjugating the excitation source to the fluor, the loss of e xcitation light due to absor ption by mammalian tissues is ob viated, and the distance that

Functional Imaging Using Bioluminescent Markers

the excitation light travels is signif icantly less than w hat would be needed for e xternal e xcitation.13 Such conjugates use BRET (also kno wn as chemiluminescence resonance ener gy transfer and luminescence resonance energy transfer) 22,141–143 to produce signals. In the study by So and colleagues, carboxylate-presenting Qdots were conjugated to a mutant for m of the luciferase from R reniformis. The long w avelength of emission from the conjugate and the colocalization of the e xcitation source with the fluorescence resulted in impro ved sensitivity in small-animal imaging and a high signal-to-noise ratio in comparison to Qdots requiring an e xternal e xcitation source.

DETERMINANTS OF DATA QUALITY Evaluation of BLI data requires an understanding of the determinants of data quality and signif icance. The sensitivity of detection for BLI is deter mined b y the brightness of the labeled cell (ie, photon flux from the source). Detection is then dependent on the depth of the source within the tissue and the optical proper ties of the tissues between the source and the detector—for example, liver is absorbing because it is highl y vascularized and therefore contains considerab le le vels of hemoglobin and bone scattering due to its str ucture. The extent of absorbance of the signal b y mammalian tissues is determined by the wavelength of the emission and the tissue composition. Therefore, grayscale reference images tak en under lo w-light illumination to reveal the position of the animal that are link ed to the optical data of v arious tissues can be used to impro ve data quality . Re gardless of the source intensity and depth, the ability to detect signals is dependent on the quantum ef ficiency and noise of the detector and the efficiency of the optics used to direct the signal to the detector. Because the source of the signal is a biochemical reaction that requires substrate, the intensity will also depend on the substrate availability at sites of reporter gene expression.

APPLICATIONS OF BLI: BLI IN THE DEVELOPMENT OF NEW THERAPEUTIC STRATEGIES A significant understanding of the molecular basis of disease has been generated o ver decades of reductionist science, and this has led to the identification of a number of molecular targets for therapy. As such the paradigm of directing therapies to specif ic molecular tar gets through

129

rational design, rather than targeting physiological states, such as accelerated cell di vision, is being applied to a wide range of diseases. The validation of molecular targets for such therapeutic advances requires testing in animal models, and BLI is being used to accelerate and refine this testing. An e xample of this type of study involves mouse models of cancer that are the result of a disregulated myc oncogene, and the effects of specifically inhibiting this molecule in v arious cancers. 130 In these models, cells with disre gulated myc expression are labeled and the effect of drugs that decrease myc expression are studied in li ving animals using luciferase as an indication of tumor burden. Traditionally, cancer therapies ha ve been tested using cell lines that represent frank malignancy and often end-stage disease. The sensitivity of BLI enab les the study of earl y disease and minimal residual disease, and therefore the effects of myc overexpression as an initiating event or as a cause of relapse, as well as decreasing myc as a therapeutic objective could be investigated.130 Although most patients that present with cancer in the oncolo gy clinic respond to established chemotherapies and radiotherapies, and ma y even demonstrate e xtended periods of remission, most patients will relapse. Thus, w e should be directing the development of ne w therapies to states of minimal disease and aim at controlling relapse as an impor tant therapeutic objecti ve. The sensiti vity of BLI enab les the study of minimal disease states and is pla ying a role in changing treatment paradigms. Imaging is opening windows into disease states that were previously inaccessible (Figures 3 and 4) and with advances in imaging w e will be ab le to study ne w disease mechanisms and develop therapies specif ically for these steps in disease pro gression. This includes targeting the small number of cancer cells that exist in the initial stages of disease or during states of minimal residual disease. 86,144–147 Elimination of these cells is being approached with ne w small molecule inhibitors,1,91,92,95,148 variations of e xisting chemotherapeutic agents, 146 and ne w combinations of therapies.86,146,149 In each of these approaches, BLI can assist in the de velopment and testing of these ne w tools and can be used as a general indicator of the e xtent of disease, that is, tumor burden in cancer, gene expression in genetic disease, or bacterial load in infection (see Figure 4), or of a specific biolo gical process, such as tumor hypoxia.59,150 Bioluminescent mark ers of h ypoxia have been developed by a number of groups and can be based either on transcriptional re gulation through tar geting hypoxia-inducible factors (HIF) 151 or on protein de gradation by directing a bioluminescent repor ter protein to

130

MOLECULAR IMAGING: PRINCIPLES AND PRACTICE

500

1000

450

800

400

600

350 400 300 200

7d

9d

16d

Figure 3. Sensitive imaging modalities provide access to new information about minimal residual disease in cancer models through in vivo bioluminescence imaging. Mice with labeled lymphoma cells were followed over time beginning prior to therapy (7 d) and after whole-body radiation and a bone marrow transplant (9 d and 16 d posttumor inoculation). 2 days after therapy a diminution in signal is observed, but the persistent minimal residual disease is apparent at 16 d—this is 11 d after treatment. The ability to monitor minimal residual disease enables the development of new therapies that can target the small number of remaining cells and thus offer the potential of preventing relapse. Adapted from Edinger M et al.100,126

genes to re veal biolo gical processes in vi vo, that is, controlling repor ter protein e xpression at the le vel of transcription4,33,153 or constitutively expressing a destabilized reporter protein that is stabilized by the targeted process.154,155 Therapies directed at gene e xpression include the development of nucleic acid-based therapies for either gain of function (gene therap y88,90,131,156–158) or loss of function (antisense RNA or RN Ai strategies91,93,95,134,159,160). These strategies ha ve been adv anced signif icantly b y appl ying preclinical imaging tools to the study of the tar get and the therapy. Through imaging, w e ha ve a better sense of promoter strength and silencing, 161,162 and duration of expression or effect on the target cells and tissue. 93–97 This will lead to improved therapies and refined clinical studies with more insightful study designs. The use of optical reporter genes in these studies can be a precursor to using reporter genes that confer detectable signatures on the molecular therapy in the clinic. These reporters would be those that can be assessed using nuclear medicine approaches. Thus, the information gleaned through imaging of laboratory animals will help direct the use of imaging in clinical studies.

APPLICATIONS OF BLI: ENABLING ADVANCES IN CLINICAL IMAGING MODALITIES

Figure 4. Imaging infection. Bacterial luciferases can be used to label bacteria and assess location of infection. In this example, Pseudomonas aeruginosa was labeled and used to infect the animals. Despite keeping the intranasal route of administration constant, both unilateral and bilateral infections were noted.

the proteosome for de gradation with an o xygen-dependent protein destabilization domain. 152 In the latter approach, the repor ter protein is stabilized in h ypoxic conditions and the signal is thus increased as the tumor vasculature is compromised and the pO 2 decreases. Development of such assa ys depends on understanding the oxygen sensitivity of each repor ter system 59 and on the ability to engineer a biological system using a sensitive reporter that can reveal subtle changes. These studies demonstrate tw o basic strate gies for using repor ter

By linking preclinical modalities that are based on clinical imaging, such as PET, single photon emission computed tomo graphy (SPECT), and magnetic resonance imaging (MRI), with imaging tools that are best suited for animal models, lik e BLI and fluorescence imaging, we can better develop strategies for imaging humans as one modality guides the other and can be used to v alidate new imaging tools. A g reat example of this is the use of gene fusions comprised of those encoding bioluminescent reporters and those that encode proteins that concentrate radiolabeled compounds, such as thymidine kinase and its substrates.121,122,124 These gene fusions enable monitoring of gene transfer for impro ved genebased therapies and for the de velopment of ne w cellbased therapies. Linking imaging modalities through mixed function repor ter genes creates po werful multimodality approaches for preclinical testing and better enables translation through validation studies.

Improved Delivery Tools The advances in small molecule therapies ha ve not been matched by a comparab le development of deli very tools

Functional Imaging Using Bioluminescent Markers

1d

4d Dorsal

Lateral

Ventral

131

7d

Lateral

Figure 5. Visualizing bone marrow transplantation. Transgenic donor mice that express firefly luciferase from strong constitutive promoters (eg, hybrid promoters from β-actin and cytomegalovirus) serve as sources of labeled cells for long-term study of transplantation and tissue regeneration. Single lateral views are shown for days 1 and 7, and two lateral, a dorsal, and a ventral view are shown for day 4. The multiple views demonstrate how the images of the same animal at the same time point differ with different perspectives. Adapted from Cao YA et al.123

d12

d17

d21

d31

Figure 6. Temporal analyses of hematopoietic stem cell (HSC) engraftment and expansion after transfer of a single labeled HSC. Transfer of a single HSC via intravenous injection results in hematopoietic reconstitution over a 30 day time course. Foci are detected at 6 to 12 d posttransfer and these early foci consist of approximately 2000 cells. These foci serve as a source of hematopoietic reconstitution and eventually signal is detected from the entire animal as the labeled cells repopulate the blood. Adapted from Cao YA et al.53

for directing these dr ugs to their therapeutic tar gets. The emerging nanotechnologies that are designed to address this problem as well as the more established technologies of liposomes and antibodies are advancing at an unprecedented rate due to the use of preclinical imaging tools that are based on optical repor ter genes. 159,163–165 The development of molecular transpor ters as methods of improved dr ug deli very ha ve adv anced through imaging.166 In addition, an increased understanding of immune cell and stem cell mig ration has been re vealed through imaging of laborator y animals, and this has created ne w opportunities for using cells to target specific pathologies and deliver therapies to the target tissues.5,53,86,123,124,167,168 To effect their function, immune cells must migrate to the

target tissue. It follo ws that these cells w ould be useful, therefore, to carry therapeutic constructs to sites of tissue destruction in autoimmune diseases or to deli ver oncolytic agents to malignant cells, and BLI has been used to e valuate and adv ance each of these strategies.86,132,133,144,169,170 The de velopment of such comple x combination biolo gical therapies benef its considerab ly from imaging, as the immune cell, the therapeutic, and/or the target can be labeled and imaged. 86,144 The complexity of de veloping cell-based therapies pales in comparison to the sophistication required for re generative medicine and stem cell therap y, and BLI is contributing to advances in this emerging area of research. BLI has become an essential par t of studies in regenerative medicine where bioluminescent reporter genes can be used to re veal the location and the number of transplanted stem cells, 53,63,123,168,171–177 the duration of g raft survival,123,168 and the maturation of a stem cell. 158,168,178 To date, this has been accomplished using constituti vely expressed reporter genes in stem cells (F igures 5 and 6); however, refined reporter genes with developmentally regulated promoters are be ginning to be used to assess the extent of stem cell dif ferentiation, and not just to assess cellular location and cell numbers. The ability to monitor engraftment and assess graft survival presents new opportunities for discovery that have not previously been possible. F or e xample, for an allo graft to sur vive immune suppression it is necessary to prevent rejection, and transfer of immunomodulator y genes or sets of genes into luciferase-labeled g rafts may prolong their sur vival. Use of BLI would enable assessing the best genes, and combinations of genes, for this pur pose. These approaches are beginning to emerge in the literature and comprise a functional genomics approach based on imaging where a large

132

MOLECULAR IMAGING: PRINCIPLES AND PRACTICE

number of genes can be screened for function in vi vo by conferring a g rowth advantage on the transplanted cells. Use of this approach will be realized in identif ication of genes that prolong engraftment of replacement tissues and those that accelerate cancer g rowth to be de veloped as new targets for therapy. The development of transgenic mice with strong constitutive promoters dri ving e xpression of the repor ter gene have been useful for transplantation studies53,123 and have also been found to be useful for de veloping ne w tools for small-molecule deli very.179 Since these animals were developed to express luciferase in every cell of their body, and luciferase acti vity depends on substrate a vailability, by controlling the delivery of a given substrate as a model small molecule, new tools can be developed that direct therapeutic compounds to target tissues. Controlled delivery has been demonstrated in skin 166 and in tissue window chambers. 179 Use of substrate deri vatives23 with different log p v alues should per mit the de velopment of improved delivery tools for a v ariety of small molecules with different pharmacological properties. Even though many of the optical imaging modalities are constrained in the clinic b y poor tissue penetration, they e xcel in the preclinical studies, and therefore are being widely used to direct clinical studies and adv ance many scientif ic f ields that depend on laborator y rodents as models. Since optical repor ters can be built into animals and linked to target genes and cells, the oppor tunities for de veloping ne w rob ust and infor mative models are significant. In addition, many of the optical tools can be superimposed on e xisting models without modif ication and this strengthens the study of w ell-developed approaches that are already pro viding new data, advancing our understanding of new targets, and accelerating the development of new therapies.

Using BLI to Improve Our Understanding of Mammalian Biology The second reason to image laborator y rodents and develop preclinical imaging tools is to improve our understanding of mammalian biolo gy, and here the end product is not a ne w dr ug, delivery tool, or imaging reagent, but new knowledge, and this is what is translated to the clinic. Transgenic reporter mice that are engineered to emit a bioluminescent signal in response to v arious chemical or physical stressors, or infectious insults can be used to guide 50,180–183 These the anal yses of the tissue responses. reporters can be designed b y tar geting transcriptional activity and using specif ic genetic re gulatory elements, promoters, to direct e xpression of the repor ter gene, or

through other tools of molecular biolo gy that are used in cell culture assa ys. These transgenic repor ter animals can be de veloped to respond to stimuli as di verse as malignancy, infection, chemicals—to xicology,184 physical stress—thermal, and physiologic changes. An alternative is to use strong constitutive promoters to drive the expression of reporter genes in transgenic animals, and this has led to the de velopment of labeled donor mice, as discussed above, that can be used as a source of labeled tissues and cells for transplantation into unlabeled recipients and the graft studied over time.53,123 These animals are particularly useful for cell trafficking studies and the study of the mammalian immune responses to infection and malignanc y. Immune cell function is based on sensing foreign proteins or cellular stress and recr uiting other immune cells to sites of insult b y either mig rating to a l ymphoid tissue, lik e dendritic cells, 185 and stimulating other immune cells to respond to the insult or secreting cytokines and chemokines that recruit effector cells to the target tissues. 100,186–189 As such, cell mig ration is a k ey feature of the mammalian immune response, and preclinical imaging is pro viding insights into these processes and increasing our knowledge of mechanisms and kinetics of immune cell function.5,167,190,191 As cell migration is also the foundation of metastatic disease and many of the cell signaling mechanisms used by immune cells are used by cancer cells (eg, the chemokine stem cell derived factor [SDF] and its cognate receptor CCR479), optical imaging is pro viding insights into mechanisms of disease progression in malignanc y and re vealing new targets for therapeutic intervention.192–195 In a study by Lin and colleagues,79 intravital microscop y w as used to re veal cell migration patter ns, and this study implicated SDF as a key f actor for metastatic disease and directed mig ration of cancer cells to specif ic regions of the bone. Coupling this microscopic technique with a macroscopic modality, such as BLI, creates a powerful combination because BLI is nonin vasive, and without injur y to the animal, can direct the in vestigator to specif ic times and tissues for placement of the bulk optics of the microscope objecti ve that is used for high-resolution imaging of cell mig ration patterns. Intravital microscopy is often a ter minal procedure that the animals do not reco ver from, and therefore, using a noninvasive, although low resolution, modality as a guide can lead to more insightful studies and can reduce wasted time and lost data. Advanced studies of oncogenesis and metastasis are enabled by imaging (see Figure 3), and although these processes ma y be thought of as disregulated mammalian de velopment, the use of imaging to study developmental biology will also lend insights into these disease processes.

Functional Imaging Using Bioluminescent Markers

The basis of development is, in par t, control of stem cells and their dif ferentiation and self-rene wal. Understanding self-renewal of stem cells or the alternative pathway of dif ferentiation and tissue re generation is the foundation of the f ields of stem cell biolo gy and tissue regeneration. Because these processes can onl y be studied in the context of the living body, imaging is inextricably link ed to the development of these areas of research, and BLI is refining models of regeneration and repair.53,63,168,171,172,174,176 It is apparent in studies that look at stem cell biolo gy that imaging pro vides a guide that directs tissue sampling to the appropriate sites and times such that the biochemical anal yses of specif ic molecules can be perfor med in a directed manner to pro vide more information. This is especiall y tr ue for assa ys that comprise genomics, proteomics, ph ysiomics, and gl ycomics because the number of multiple xed assa ys that can be performed can be se verely limited b y time and cost. Directed studies are thus imperati ve if we are to use our resources ef ficiently. The combination of imaging as a functional readout and high-throughput multiple xed assays has adv anced functional genomics w hereby w e can assess more than e xpression levels and can be gin to ascribe function to changes in gene expression. Although imaging is no w largely used to direct e x vivo assays of gene expression or protein levels in the areas of genomics and proteomics, it is becoming ob vious that w e can use imaging to de velop multiple xed in vi vo assa ys of gene expression and protein function in preclinical models. There are a limited number of unique bioluminescent proteins and chemistries that have been used in vivo, and therefore these assays will use one or two bioluminescent reporters45,49,51,120 but can incorporate multiple genes that alter the function of the target cell. Development of these techniques is the future of functional genomics as the y reveal the role that changes in expression have in specific biological processes and can link these changes to outcome—this cannot readil y be accomplished with standalone multiplexed assays. The benef its of preclinical imaging using bioluminescent, and other reporters, are obvious as is the role that these tools ha ve in testing compounds and deli very schemes for therapy. The use of these tools to study mammalian biolo gy and the transfer of this kno wledge will have a signif icant impact on ho w we approach the study of human biolo gy and in the management of disease. Imaging enab les the in vi vo study of cell biolo gy, and when integrated with thorough studies in culture and e x vivo, can re veal the nuances and subtleties of disease mechanisms and of therapeutic responses. In this w ay, visible animal models of human biolo gy and disease

133

comprise one of the most impor tant contributions of molecular imaging to human health as the y ser ve to accelerate and refine the analyses of mammalian biology and offer a rapid readout for the development of new therapies. The use of imaging in preclinical studies is apparent, but in applying these tools it is essential to ask which modality will most effectively and economically answer a given biological question with the greatest sensitivity and specificity. There is a wide range of choices and knowing what each modality can provide is crucial for developing an effective study design. As this chapter focuses on BLI, the ne xt section will attempt to ans wer the question, “Why perform imaging with bioluminescent markers?”

SUMMARY AND FUTURE OF BIOLUMINESCENCE REPORTERS IN LIVING ANIMALS The de velopment of ne w luciferase proteins and their respective substrates for use as molecular imaging tools will of fer po werful approaches for studying molecular changes in tissues and cells under physiological conditions where the conte xtual influences of intact or gan systems can be evaluated. BLI has already played a significant role in the study of animal models and , because it can be used to ref ine and accelerate these studies, has become a cornerstone technology in preclinical studies and the development of novel therapies. BLI and other imaging strate gies contribute to the de velopment of sophisticated animal models of human biolo gy and disease. The detection of internal sources of light that comprise BLI has largely been planar, and this too is evolving. The vast majority of studies using BLI represent the data as planar projections using pseudo-colored images to represent signal intensity, and these images are localized over g rayscale reference images of the subjects. Recent developments in three-dimensional (3D) reconstr uction are leading to instruments that generate 3D data sets from either multiple images from se veral vie ws, or from temporal196 or spectral data. 197,198 Reconstructing 3D images from bioluminescence data sets is not trivial199 and work in this area continues b y a number of g roups.200–202 All the advances in instr umentation for BLI are based on the same basic design that includes an imaging chamber that e xcludes ambient light and a sensiti ve CCD as a detector.15 Despite the simple design, there ha ve been a number of instr umentation advances for detecting bioluminescent signals in the body . A majority of these advances have been for the purpose of obtaining data that would permit the reconstr uction of 3D data sets. 197,203–205 These designs include a ring of detectors for obtaining

134

MOLECULAR IMAGING: PRINCIPLES AND PRACTICE

multiple vie ws, a stage for mo ving the animal coupled with a rotating mir ror for obtaining multiple vie ws, and improved spectral imaging using f ilters.197,206–208 Many of these advances are relatively new and are beginning to be applied to biological questions. We are at the v ery beginning of w hat promises to be a re volution in biolo gical investigation and in vi vo analyses of patholo gic changes. We ha ve seen the impact on studies in laborator y animals and its impact on clinical care is onl y beginning. As we approach the era of personalized medicine, the influence of the emerging tools of molecular imaging on biomedicine will lik ely be signif icant. in vi vo measures of cellular and molecular changes using imaging has re volutionized the study of laborator y animals and as these methods become inte grated into all f ields of biomedical research this will continue to grow. Perhaps the g reatest contributions to medicine of fered b y the field of molecular imaging will be increased understanding of the molecular basis of disease in animal models and the subsequent de velopment of ne w therapies that target these biological processes.

REFERENCES 1. Shah K, Weissleder R. Molecular optical imaging: applications leading to the de velopment of present da y therapeutics. NeuroRx 2005;2:215–25. 2. Gross S, Piwnica-Worms D. Spying on cancer: molecular imaging in vivo with genetically encoded reporters. Cancer Cell 2005;7:5–15. 3. Contag CH, Ross BD . It’s not just about anatom y: in vi vo bioluminescence imaging as an e yepiece into biolo gy. J Magn Reson Imaging 2002;16:378–87. 4. Contag CH, Bachmann MH. Advances in in vi vo bioluminescence imaging of gene e xpression. Annu Re v Biomed Eng 2002; 4:235–60. 5. Negrin RS, Contag CH. In vi vo imaging using bioluminescence: a tool for probing g raft-versus-host disease. Nat Re v Immunol 2006;6:484–90. 6. Hastings JW. Chemistries and colors of bioluminescent reactions: a review. Gene 1996;173:5–11. 7. Hastings JW , Morin JG. Photons for repor ting molecular e vents: green fluorescent protein and four luciferase systems. Methods Biochem Anal 2006;47:15–38. 8. Wilson T, Hastings JW. Bioluminescence. Annu Rev Cell De v Biol 1998;14:197–230. 9. Wood KV. Luc genes: introduction of colour into bioluminescence assays. J Biolumin Chemilumin 1990;5:107–14. 10. Venisnik KM, Olafsen T, Gambhir SS, Wu AM. Fusion of Gaussia luciferase to an engineered anti-carcinoembryonic antigen (CEA) antibody for in vi vo optical imaging. Mol Imaging Biol 2007;9:267–77. 11. Venisnik KM, Olafsen T, Loening AM, et al. Bifunctional antibodyRenilla luciferase fusion protein for in vi vo optical detection of tumors. Protein Eng Des Sel 2006;19:453–60. 12. So MK, Loening AM, Gambhir SS, Rao J. Creating self-illuminating quantum dot conjugates. Nat Protoc 2006;1:1160–4. 13. So MK, Xu C, Loening AM, et al. Self-illuminating quantum dot conjugates for in vivo imaging. Nat Biotechnol 2006;24:339–43.

14. Olafsen T, Cheung CW , Yazaki PJ, et al. Co valent disulf ide-linked anti-CEA diabody allo ws site-specif ic conjugation and radiolabeling for tumor tar geting applications. Protein Eng Des Sel 2004;17:21–7. 15. Rice BW, Cable MD, Nelson MB . In vi vo imaging of light-emitting probes. J Biomed Opt 2001;6:432–40. 16. Tromberg BJ, Shah N, Lanning R, et al. Non-invasive in vivo characterization of breast tumors using photon mig ration spectroscopy. Neoplasia 2000;2:26–40. 17. Jain RK, Munn LL, Fukumura D. Dissecting tumour pathophysiology using intravital microscopy. Nat Rev Cancer 2002;2:266–76. 18. Halin C, Rodrigo Mora J , Sumen C, v on Andrian UH. In vi vo imaging of l ymphocyte traf ficking. Annu Re v Cell De v Biol 2005; 21:581–603. 19. Hsiung PL, Hardy J, Friedland S, et al. Detection of colonic dysplasia in vi vo using a tar geted heptapeptide and confocal microendoscopy. Nat Med 2008;14:454–8. 20. Wang TD, Friedland S, Sahbaie P, et al. Functional imaging of colonic mucosa with a f ibered confocal microscope for real-time in vi vo pathology. Clin Gastroenterol Hepatol 2007;5:1300–5. 21. Wehrman TS, von Degenfeld G, Krutzik PO, et al. Luminescent imaging of β-galactosidase activity in living subjects using sequential reporter-enzyme luminescence. Nat Methods 2006;3:295–301. 22. Xu Y, Piston DW, Johnson CH. A bioluminescence resonance energy transfer (BRET) system: application to interacting circadian clock proteins. Proc Natl Acad Sci U S A 1999;96:151–6. 23. Shinde R, Perkins J, Contag CH. Luciferin derivatives for enhanced in vitro and in vi vo bioluminescence assa ys. Biochemistr y 2006; 45:11103–12. 24. Kanno A, Ozawa T, Umezawa Y. Intein-mediated reporter gene assay for detecting protein-protein interactions in li ving mammalian cells. Anal Chem 2006;78:556–60. 25. Ozawa T. Designing split repor ter proteins for anal ytical tools. Anal Chim Acta 2006;556:58–68. 26. Chan CT, Paulmurugan R, Gheysens OS, et al. Molecular imaging of the efficacy of heat shock protein 90 inhibitors in li ving subjects. Cancer Res 2008;68:216–26. 27. Villalobos V, Naik S, Piwnica-Worms D. Detection of protein-protein interactions in li ve cells and animals with split f irefly luciferase protein fragment complementation. Methods Mol Biol 2008;439:339–52. 28. Hardy J, Francis KP, DeBoer M, et al. Extracellular replication of Listeria monoc ytogenes in the murine gall b ladder. Science 2004; 303:851–3. 29. Hardy J, Margolis JJ, Contag CH. Induced biliar y excretion of Listeria monocytogenes. Infect Immun 2006;74:1819–27. 30. Germano G, Chen BC, Huang SC, et al. Use of the abdominal aor ta for arterial input function deter mination in hepatic and renal PET studies. J Nucl Med 1992;33:613–20. 31. Choi Y, Hawkins RA, Huang SC, et al. Parametric images of myocardial metabolic rate of glucose generated from dynamic cardiac PET and 2-[18F]fluoro-2-deo xy-d-glucose studies. J Nucl Med 1991;32:733–8. 32. Contag CH, Contag PR, Mullins JI, et al. Photonic detection of bacterial pathogens in living hosts. Mol Microbiol 1995;18:593–603. 33. Contag CH, Spilman SD, Contag PR, et al. Visualizing gene expression in li ving mammals using a bioluminescent repor ter. Photochem Photobiol 1997;66:523–31. 34. Tjuvajev JG, Stockhammer G, Desai R, et al. Imaging the e xpression of transfected genes in vivo. Cancer Res 1995;55:6126–32. 35. Sadikot RT, Blackwell TS. Bioluminescence imaging. Proc Am Thorac Soc 2005;2:537–40, 511–32. 36. Contag CH. In vivo pathology: seeing with molecular specif icity and cellular resolution in the li ving body . Annu Re v P athol 2007; 2:277–305. 37. Contag PR, Olomu IN , Stevenson DK, Contag CH. Bioluminescent indicators in living mammals. Nat Med 1998;4:245–7.

Functional Imaging Using Bioluminescent Markers

38. Troy T, Jekic-McMullen D , Sambucetti L, Rice B . Quantitati ve comparison of the sensitivity of detection of fluorescent and bioluminescent reporters in animal models. Mol Imaging 2004;3:9–23. 39. Steele RW. Clinical applications of chemiluminescence of g ranulocytes. Rev Infect Dis 1991;13:918–25. 40. Shapiro E, Lu C, Bane yx F. A set of multicolored Photinus p yralis luciferase mutants for in vi vo bioluminescence applications. Protein Eng Des Sel 2005;18:581–7. 41. Wood KV, Lam YA, Seliger HH, McElroy WD. Complementary DNA coding click beetle luciferases can elicit bioluminescence of different colors. Science 1989;244:700–2. 42. de Wet JR, Wood KV, DeLuca M, et al. Firefly luciferase gene: structure and expression in mammalian cells. Mol Cell Biol 1987;7:725–37. 43. Michelini E, Cevenini L, Mezzanotte L, et al. Combining intracellular and secreted bioluminescent repor ter proteins for multicolor cell-based assays. Photochem Photobiol Sci 2008;7:212–7. 44. Shao N, Bock R. A codon-optimized luciferase from Gaussia princeps facilitates the in vivo monitoring of gene expression in the model alga Chlamydomonas reinhardtii. Curr Genet 2008;53:381–8. 45. Tannous BA, Kim DE, Fernandez JL, et al. Codon-optimized Gaussia luciferase cDNA for mammalian gene expression in culture and in vivo. Mol Ther 2005;11:435–43. 46. Doyle TC, Burns SM, Contag CH. In vi vo bioluminescence imaging for integrated studies of infection. Cell Microbiol 2004;6:303–17. 47. Stolz U, Velez S, Wood KV, et al. Darwinian natural selection for orange bioluminescent color in a Jamaican click beetle. Proc Natl Acad Sci U S A 2003;100:14955–9. 48. Loening AM, Wu AM, Gambhir SS. Red-shifted Renilla renifor mis luciferase v ariants for imaging in li ving subjects. Nat Methods 2007;4:641–3. 49. Zhao H, Doyle TC, Coquoz O, et al. Emission spectra of bioluminescent reporters and interaction with mammalian tissue determine the sensitivity of detection in vivo. J Biomed Opt 2005;10:41210. 50. Kadurugamuwa JL, Modi K, Coquoz O, et al. Reduction of astrogliosis by early treatment of pneumococcal meningitis measured b y simultaneous imaging, in vivo, of the pathogen and host response. Infect Immun 2005;73:7836–43. 51. Bhaumik S, Le wis XZ, Gambhir SS. Optical imaging of Renilla luciferase, synthetic Renilla luciferase, and firefly luciferase reporter gene expression in living mice. J Biomed Opt 2004;9:578–86. 52. Sweeney TJ, Mailander V, Tucker AA, et al. Visualizing the kinetics of tumor-cell clearance in living animals. Proc Natl Acad Sci U S A 1999;96:12044–9. 53. Cao YA, Wagers AJ, Beilhack A, et al. Shifting foci of hematopoiesis during reconstitution from single stem cells. Proc NatlAcad Sci U S A 2004;101:221–6. 54. Paulmurugan R, Umeza wa Y, Gambhir SS. Nonin vasive imaging of protein-protein interactions in li ving subjects b y using repor ter protein complementation and reconstitution strate gies. Proc Natl Acad Sci U S A 2002;99:15608–13. 55. Ray P, Pimenta H, P aulmurugan R, et al. Nonin vasive quantitati ve imaging of protein-protein interactions in li ving subjects. Proc Natl Acad Sci U S A 2002;99:3105–10. 56. Paulmurugan R, Gambhir SS. Monitoring protein-protein interactions using split synthetic renilla luciferase protein-fragment-assisted complementation. Anal Chem 2003;75:1584–9. 57. Luker KE, Piwnica-Worms D. Optimizing luciferase protein fragment complementation for bioluminescent imaging of protein-protein interactions in li ve cells and animals. Methods Enzymol 2004; 385:349–60. 58. Wang Y, Wang G, O’Kane DJ , Szalay AA. A study of protein-protein interactions in li ving cells using luminescence resonance energy transfer (LRET) from Renilla luciferase to Aequorea GFP. Mol Gen Genet 2001;264:578–87. 59. Cecic I, Chan D A, Sutphin PD, et al. Oxygen sensiti vity of repor ter genes: implications for preclinical imaging of tumor hypoxia. Mol Imaging 2007;6:219–28.

135

60. Gambhir SS, Herschman HR, Cher ry SR, et al. Imaging transgene expression with radionuclide imaging technolo gies. Neoplasia 2000;2:118–38. 61. Chalfie M. Green fluorescent protein. Photochem Photobiol 1995; 62:651–6. 62. Chalfie M, Tu Y, Euskirchen G, et al. Green fluorescent protein as a marker for gene expression. Science 1994;263:802–5. 63. Tolar J, Osbor n M, Bell S, et al. Real-time in vi vo imaging of stem cells follo wing transgenesis b y transposition. Mol Ther 2005; 12:42–8. 64. Zhang M, Alicot EM, Chiu I, et al. Identif ication of the tar get selfantigens in reperfusion injury. J Exp Med 2006;203:141–52. 65. Siegel RM, Chan FK, Zacharias D A, et al. Measurement of molecular interactions in li ving cells b y fluorescence resonance ener gy transfer betw een v ariants of the g reen fluorescent protein. Sci STKE 2000;2000:PL1. 66. Tsien R Y. The g reen fluorescent protein. Annu Re v Biochem 1998;67:509–44. 67. Tsien RY, Miyawaki A. Seeing the machiner y of li ve cells. Science 1998;280:1954–5. 68. Cubitt AB, Heim R, Adams SR, et al. Understanding, impro ving and using g reen fluorescent proteins. Trends Biochem Sci 1995; 20:448–55. 69. Louie AY, Huber MM, Ahrens ET, et al. In vivo visualization of gene expression using magnetic resonance imaging. Nat Biotechnol 2000;18:321–5. 70. Moore A, Basilion JP, Chiocca EA, Weissleder R. Measuring transferrin receptor gene e xpression by NMR imaging. Biochim Biophys Acta 1998;1402:239–49. 71. Funovics M, Weissleder R, Tung CH. Protease sensors for bioimaging. Anal Bioanal Chem 2003;377:956–63. 72. Jaffer FA, Tung CH, Gerszten RE, Weissleder R. In vi vo imaging of thrombin acti vity in e xperimental thrombi with thrombin-sensitive near-infrared molecular probe. Arterioscler Thromb Vasc Biol 2002;22:1929–35. 73. Chen J, Tung CH, Mahmood U , et al. In vi vo imaging of proteol ytic activity in atherosclerosis. Circulation 2002;105:2766–71. 74. Weissleder R, Tung CH, Mahmood U, Bogdanov A Jr. In vivo imaging of tumors with protease-acti vated near -infrared fluorescent probes. Nat Biotechnol 1999;17:375–8. 75. von Andrian UH, Berger EM, Ramezani L, et al. In vi vo behavior of neutrophils from tw o patients with distinct inherited leuk ocyte adhesion deficiency syndromes. J Clin Invest 1993;91:2893–7. 76. Mazo IB , Gutier rez-Ramos JC, F renette PS, et al. Hematopoietic progenitor cell rolling in bone mar row micro vessels: parallel contributions b y endothelial selectins and v ascular cell adhesion molecule 1. J Exp Med 1998;188:465–74. 77. Dewhirst MW, Shan S, Cao Y, et al. Intravital fluorescence facilitates measurement of multiple ph ysiologic functions and gene e xpression in tumors of live animals. Dis Markers 2002;18:293–311. 78. Sumen C, Mempel TR, Mazo IB , v on Andrian UH. Intra vital microscopy: visualizing immunity in conte xt. Immunity 2004; 21:315–29. 79. Lin DA, Wei X, Wu JW, et al. In vi vo imaging of specialized bone marrow endothelial microdomains for tumour eng raftment. Nature 2005;435:969–73. 80. Germain RN, Castellino F, Chieppa M, et al. An extended vision for dynamic high-resolution intra vital immune imaging. Semin Immunol 2005;17:431–41. 81. Gooch VD, Mehra A, Lar rondo LF , et al. Full y codon-optimized luciferase uncovers novel temperature characteristics of the Neurospora clock. Eukaryot Cell 2008;7:28–37. 82. Patterson SS, Dionisi HM, Gupta RK, Sayler GS. Codon optimization of bacterial luciferase (lux) for e xpression in mammalian cells. J Ind Microbiol Biotechnol 2005;32:115–23. 83. Nielsen TO, West RB, Linn SC, et al. Molecular characterisation of soft tissue tumours: a gene expression study. Lancet 2002;359:1301–7.

136

MOLECULAR IMAGING: PRINCIPLES AND PRACTICE

84. Benaron D A, Cheong WF, Ste venson DK. Tissue optics. Science 1997;276:2002–3. 85. Jobsis FF. Noninvasive, infrared monitoring of cerebral and myocardial o xygen suf ficiency and circulator y parameters. Science 1977;198:1264–7. 86. Thorne SH, Negrin RS, Contag CH. Synergistic antitumor effects of immune cell-viral biotherapy. Science 2006;311:1780–4. 87. Sosnovik D, Weissleder R. Magnetic resonance and fluorescence based molecular imaging technologies. Prog Drug Res 2005;62:83–115. 88. Lipshutz GS, Gr uber CA, Cao Y, et al. In utero deli very of adenoassociated viral v ectors: intraperitoneal gene transfer produces long-term expression. Mol Ther 2001;3:284–92. 89. Lipshutz GS, Titre D, Brindle M, et al. Comparison of gene e xpression after intraperitoneal delivery of AAV2 or AAV5 in utero. Mol Ther 2003;8:90–8. 90. Lee CT, Lee YJ, Kwon SY, et al. In vi vo imaging of adeno virus transduction and enhanced therapeutic ef ficacy of combination therapy with conditionall y replicating adeno virus and adenovirus-p27. Cancer Res 2006;66:372–7. 91. McCaffrey A, Kay MA, Contag CH. Advancing molecular therapies through in vi vo bioluminescent imaging. Mol Imaging 2003; 2:75–86. 92. McCaffrey AP, Meuse L, Karimi M, et al. A potent and specific morpholino antisense inhibitor of hepatitis C translation in mice. Hepatology 2003;38:503–8. 93. Hickerson RP, Smith FJ, Reeves RE, et al. Single-nucleotide-specific siRNA targeting in a dominant-negative skin model. J Invest Dermatol 2008;128:594–605. 94. Ilves H, Kaspar RL, Wang Q, et al. Inhibition of hepatitis C IRESmediated gene expression by small hairpin RNAs in human hepatocytes and mice. Ann N Y Acad Sci 2006;1082:52–5. 95. Smith FJ, Hickerson RP, Sayers JM, et al. Development of therapeutic siRNAs for pachyonychia congenita. J In vest Dermatol 2008; 128:50–8. 96. Wang Q, Contag CH, Ilv es H, et al. Small hair pin RNAs efficiently inhibit hepatitis C IRES-mediated gene e xpression in human tissue culture cells and a mouse model. Mol Ther 2005;12:562–8. 97. Wang Q, Ilv es H, Chu P , et al. Deli very and inhibition of repor ter genes by small interfering RNAs in a mouse skin model. J In vest Dermatol 2007;127:2577–84. 98. Zhang GJ, Kaelin WG Jr. Bioluminescent imaging of ubiquitin ligase activity: measuring Cdk2 activity in vivo through changes in p27 turnover. Methods Enzymol 2005;399:530–49. 99. Laxman B, Hall DE, Bhojani MS, et al. Noninvasive real-time imaging of apoptosis. Proc Natl Acad Sci U S A 2002;99:16551–5. 100. Edinger M, Cao YA, Verneris MR, et al. Re vealing l ymphoma growth and the ef ficacy of immune cell therapies using in vivo bioluminescence imaging. Blood 2003;101:640–8. 101. Franks NP, Jenkins A, Conti E, et al. Str uctural basis for the inhibition of f irefly luciferase b y a general anesthetic. Bioph ys J 1998;75:2205–11. 102. Conti E, F ranks NP, Brick P. Cr ystal str ucture of f irefly luciferase throws light on a superf amily of aden ylate-forming enzymes. Structure 1996;4:287–98. 103. O’Connell-Rodwell CE, Shriver D, Simanovskii DM, et al. A genetic reporter of ther mal stress def ines ph ysiologic zones o ver a defined temperature range. FASEB J 2004;18:264–71. 104. Wood KV, Lam YA, McElroy WD, Seliger HH. Bioluminescent click beetles revisited. J Biolumin Chemilumin 1989;4:31–9. 105. Karkhanis YD, Cormier MJ. Isolation and properties of Renilla reniformis luciferase, a lo w molecular w eight ener gy con version enzyme. Biochemistry 1971;10:317–26. 106. Matthews JC, Hori K, Cor mier MJ. Purif ication and proper ties of Renilla reniformis luciferase. Biochemistry 1977;16:85–91. 107. Srikantha T, Klapach A, Lorenz WW, et al. The sea pansy Renilla reniformis luciferase serves as a sensitive bioluminescent reporter

108.

109.

110.

111.

112.

113.

114.

115.

116.

117. 118.

119.

120.

121. 122.

123.

124.

125.

126.

127.

128.

for differential gene e xpression in Candida albicans. J Bacteriol 1996;178:121–9. Shimomura O , Johnson FH. Chemical nature of bioluminescence systems in coelenterates. Proc Natl Acad Sci U S A 1975;72: 1546–9. Greer LF III, Szalay AA. Imaging of light emission from the expression of luciferases in li ving cells and or ganisms: a review. Luminescence 2002;17:43–74. Loening AM, F enn TD, Gambhir SS. Cr ystal str uctures of the luciferase and g reen fluorescent protein from Renilla renifor mis. J Mol Biol 2007;374:1017–28. Loening AM, F enn TD, Wu AM, Gambhir SS. Consensus guided mutagenesis of Renilla luciferase yields enhanced stability and light output. Protein Eng Des Sel 2006;19:391–400. Zhao H, Do yle TC, Wong RJ, et al. Characterization of coelenterazine analogs for measurements of Renilla luciferase acti vity in live cells and living animals. Mol Imaging 2004;3:43–54. Luo J, Ho P, Steinman L, Wyss-Coray T. Bioluminescence in vi vo imaging of autoimmune encephalom yelitis predicts disease. J Neuroinflammation 2008;5:6. Lin AH, Luo J , Mondshein LH, et al. Global anal ysis of Smad2/3dependent TGF-β signaling in li ving mice re veals prominent tissue-specific responses to injury. J Immunol 2005;175:547–54. Ando T, Hatsushika K, Wako M, et al. Orally administered TGF-β is biologically active in the intestinal mucosa and enhances oral tolerance. J Allergy Clin Immunol 2007;120:916–23. Luo J, Lin AH, Masliah E, Wyss-Coray T. Bioluminescence imaging of Smad signaling in living mice shows correlation with excitotoxic neurodegeneration. Proc Natl Acad Sci U S A 2006;103:18326–31. Tesseur I, Zou K, Berber E, et al. Highly sensitive and specific bioassay for measuring bioactive TGF-β. BMC Cell Biol 2006;7:15. Levi J, De A, Cheng Z, Gambhir SS. Bisdeo xycoelenterazine derivatives for impro vement of bioluminescence resonance ener gy transfer assays. J Am Chem Soc 2007;129:11900–1. Jones LR, Goun EA, Shinde R, et al. Releasable luciferin-transporter conjugates: tools for the real-time anal ysis of cellular uptak e and release. J Am Chem Soc 2006;128:6526–7. Bhaumik S, Gambhir SS. Optical imaging of Renilla luciferase reporter gene expression in living mice. Proc Natl Acad Sci U S A 2002;99:377–82. Ray P, De A, Min JJ, et al. Imaging tri-fusion multimodality reporter gene expression in living subjects. Cancer Res 2004;64:1323–30. Ray P , Tsien R, Gambhir SS. Constr uction and v alidation of improved triple fusion reporter gene vectors for molecular imaging of living subjects. Cancer Res 2007;67:3085–93. Cao YA, Bachmann MH, Beilhack A, et al. Molecular imaging using labeled donor tissues re veals patter ns of eng raftment, rejection, and sur vival in transplantation. Transplantation 2005;80:134–9. Ponomarev V, Doubrovin M, Serganova I, et al. A novel triple-modality reporter gene for whole-body fluorescent, bioluminescent, and nuclear noninvasive imaging. Eur J Nucl Med Mol Imaging 2004; 31:740–51. Day RN , Kawecki M, Ber ry D. Dual-function repor ter protein for analysis of gene e xpression in li ving cells. Biotechniques 1998;25:848–50, 852–4, 856. Edinger M, Hof fmann P, Contag CH, Ne grin RS. Ev aluation of effector cell f ate and function b y in vi vo bioluminescence imaging. Methods 2003;31:172–9. Mandl S, Mari C, Edinger M, et al. In vi vo dynamics of tumor cell death: multi-modality imaging identif ies key imaging times for assessing response to chemotherap y using an or thotopic mouse model of lymphoma. 2003. [Submitted] Yang Y, Contag CH, F elsher D, et al. The E47 transcription f actor negatively regulates CD5 e xpression during th ymocyte development. Proc Natl Acad Sci U S A 2004;101:3898–902.

Functional Imaging Using Bioluminescent Markers

129. Smith SJ , Zhang H, Cler mont AO, et al. In vi vo monitoring of pancreatic β-cells in a transgenic mouse model. Mol Imaging 2006;5:65–75. 130. Shachaf CM, K opelman AM, Arvanitis C, et al. MYC inacti vation uncovers pluripotent dif ferentiation and tumour dor mancy in hepatocellular cancer. Nature 2004;431:1112–7. 131. Rehemtulla A, Hall DE, Ste gman LD, et al. Molecular imaging of gene expression and efficacy following adenoviral-mediated brain tumor gene therapy. Mol Imaging 2002;1:43–55. 132. Slavin AJ, Tarner IH, Nakajima A, et al. Adoptive cellular gene therapy of autoimmune disease. Autoimmun Rev 2002;1:213–9. 133. Tarner IH, Nakajima A, Seroogy CM, et al. Retro viral gene therapy of collagen-induced ar thritis b y local deli very of IL-4. Clin Immunol 2002;105:304–14. 134. Eefting D, Grimbergen JM, de Vries MR, et al. Prolonged in vi vo gene silencing b y electroporation-mediated plasmid deli very of small interfering RNA. Hum Gene Ther 2007;18:861–9. 135. Chen ZY, Riu E, He CY , et al. Silencing of episomal transgene expression in li ver by plasmid bacterial backbone DN A is independent of CpG methylation. Mol Ther 2008;16:548–56. 136. Chen ZY, He CY, Meuse L, Ka y MA. Silencing of episomal transgene expression by plasmid bacterial DNA elements in vivo. Gene Ther 2004;11:856–64. 137. De A, Gambhir SS. Noninvasive imaging of protein-protein interactions from li ve cells and li ving subjects using bioluminescence resonance energy transfer. FASEB J 2005;19:2017–9. 138. Hoshino H, Nakajima Y, Ohmiya Y. Luciferase-YFP fusion tag with enhanced emission for single-cell luminescence imaging. Nat Methods 2007;4:637–9. 139. Massoud TF, Paulmurugan R, De A, et al. Reporter gene imaging of protein-protein interactions in living subjects. Curr Opin Biotechnol 2007;18:31–7. 140. Carriba P, Navarro G, Ciruela F, et al. Detection of heteromerization of more than tw o proteins by sequential BRET-FRET. Nat Methods 2008;5:727–33. 141. Arai R, Nakagawa H, Kitayama A, et al. Detection of protein-protein interaction b y bioluminescence resonance ener gy transfer from firefly luciferase to red fluorescent protein. J Biosci Bioeng 2002; 94:362–4. 142. Issad T, Boute N , P ernet K. The acti vity of the insulin receptor assessed by bioluminescence resonance ener gy transfer. Ann N Y Acad Sci 2002;973:120–3. 143. Boute N, Jockers R, Issad T. The use of resonance energy transfer in high-throughput screening: BRET v ersus FRET. Trends Pharmacol Sci 2002;23:351–4. 144. Thorne SH, Contag CH. Integrating the biological characteristics of oncolytic viruses and immune cells can optimize therapeutic benefits of cell-based delivery. Gene Ther 2008;15:753–8. 145. Dickson PV, Hamner B, Ng CY, et al. In vivo bioluminescence imaging for early detection and monitoring of disease progression in a murine model of neuroblastoma. J Pediatr Surg 2007;42:1172–9. 146. Nathan CO , Amirghahari N , Rong X, et al. Mammalian tar get of rapamycin inhibitors as possible adjuvant therapy for microscopic residual disease in head and neck squamous cell cancer . Cancer Res 2007;67:2160–8. 147. Wetterwald A, van der Pluijm G, Que I, et al. Optical imaging of cancer metastasis to bone mar row: a mouse model of minimal residual disease. Am J Pathol 2002;160:1143–53. 148. Flanagan JM, Truksa J, Peng H, et al. In vi vo imaging of hepcidin promoter stimulation by iron and inflammation. Blood Cells Mol Dis 2007;38:253–7. 149. Kaneda Y. Impro vements in gene therap y technolo gies. Mol Urol 2001;5:85–9. 150. Schmid T, Young MR. Lights on for lo w o xygen: a nonin vasive mouse model useful for sensing o xygen def iciency. Mol Inter v 2006;6:20–2.

137

151. Harada H, Kizaka-Kondoh S, Hiraoka M. Optical imaging of tumor hypoxia and evaluation of efficacy of a hypoxia-targeting drug in living animals. Mol Imaging 2005;4:182–93. 152. Safran M, Kim WY, O’Connell F, et al. Mouse model for nonin vasive imaging of HIF prolyl hydroxylase activity: assessment of an oral agent that stimulates er ythropoietin production. Proc Natl Acad Sci U S A 2006;103:105–10. 153. Contag CH, Ste venson DK. In vi vo patter ns of heme o xygenase-1 transcription [discussion]. J Perinatol 2001;21 Suppl 1:S119–27. 154. Stankunas K, Ba yle JH, Ha vranek JJ, et al. Rescue of de gradationprone mutants of the FK506-rapam ycin binding (FRB) protein with chemical ligands. Chembiochem 2007;8:1162–9. 155. Banaszynski LA, Selme yer M, Thorne SH, et al. In vi vo control of protein stability. Nat Med 2008. [In press] 156. Ciana P, Raviscioni M, Mussi P, et al. In vi vo imaging of transcriptionally active estrogen receptors. Nat Med 2003;9:82–6. 157. Luker GD, Bardill JP, Prior JL, et al. Nonin vasive bioluminescence imaging of her pes simplex vir us type 1 infection and therap y in living mice. J Virol 2002;76:12149–61. 158. Wu JC, Inubushi M, Sundaresan G, et al. Optical imaging of cardiac reporter gene e xpression in li ving rats. Circulation 2002; 105:1631–4. 159. Bartlett DW, Su H, Hildebrandt IJ, et al. Impact of tumor-specific targeting on the biodistribution and ef ficacy of siRNA nanoparticles measured by multimodality in vivo imaging. Proc Natl Acad Sci U S A 2007;104:15549–54. 160. Zhang GJ, Safran M, Wei W, et al. Bioluminescent imaging of Cdk2 inhibition in vivo. Nat Med 2004;10:643–8. 161. Moldt B, Yant SR, Andersen PR, et al. Cis-acting gene re gulatory activities in the terminal regions of sleeping beauty DNA transposon-based vectors. Hum Gene Ther 2007;18:1193–204. 162. Wilber A, F randsen JL, Wangensteen KJ , et al. Dynamic gene expression after systemic delivery of plasmid DNA as determined by in vi vo bioluminescence imaging. Hum Gene Ther 2005;16:1325–32. 163. Kim SI, Shin D , Choi TH, et al. Systemic and specif ic deli very of small interfering RNAs to the liver mediated by apolipoprotein A-I. Mol Ther 2007;15:1145–52. 164. Suzuki R, Takizawa T, Ne gishi Y, et al. Tumor specif ic ultrasound enhanced gene transfer in vi vo with no vel liposomal bubb les. J Control Release 2008;125:137–44. 165. Iyer M, Berenji M, Templeton NS, Gambhir SS. Nonin vasive imaging of cationic lipid-mediated delivery of optical and PET reporter genes in living mice. Mol Ther 2002;6:555–62. 166. Wender PA, Goun EA, Jones LR, et al. Real-time anal ysis of uptake and bioactivatable cleavage of luciferin-transpor ter conjugates in transgenic repor ter mice. Proc Natl Acad Sci U S A 2007; 104:10340–5. 167. Prins RM, Shu CJ, Radu CG, et al. Anti-tumor activity and trafficking of self, tumor -specific T cells against tumors located in the brain. Cancer Immunol Immunother 2008;57:1279–89. 168. Li Z, Wu JC, Sheikh AY, et al. Differentiation, survival, and function of embr yonic stem cell deri ved endothelial cells for ischemic heart disease. Circulation 2007;116:I46–54. 169. Costa GL, Sandora MR, Nakajima A, et al. Adoptive immunotherapy of experimental autoimmune encephalomyelitis via T cell delivery of the IL-12 p40 subunit. J Immunol 2001;167:2379–87. 170. Nakajima A, Seroo gy CM, Sandora MR, et al. Antigen-specific T cell-mediated gene therapy in collagen-induced arthritis. J Clin Invest 2001;107:1293–301. 171. Sheikh AY, Lin SA, Cao F, et al. Molecular imaging of bone marrow mononuclear cell homing and eng raftment in ischemic myocardium. Stem Cells 2007;25:2677–84. 172. Degano IR, Vilalta M, Bago JR, et al. Bioluminescence imaging of calvarial bone repair using bone mar row and adipose tissuederived mesenchymal stem cells. Biomaterials 2008;29:427–37.

138

MOLECULAR IMAGING: PRINCIPLES AND PRACTICE

173. BitMansour A, Bur ns SM, Traver D , et al. My eloid pro genitors protect against invasive aspergillosis and Pseudomonas aeruginosa infection following hematopoietic stem cell transplantation. Blood 2002;100:4660–7. 174. Chan KM, Raikw ar SP, Za vazava N . Strate gies for dif ferentiating embryonic stem cells (ESC) into insulin-producing cells and development of non-in vasive imaging techniques using bioluminescence. Immunol Res 2007;39:261–70. 175. Heckl S. Future contrast agents for molecular imaging in strok e. Curr Med Chem 2007;14:1713–28. 176. Okada S, Ishii K, Yamane J, et al. In vivo imaging of engrafted neural stem cells: its application in e valuating the optimal timing of transplantation for spinal cord injury. FASEB J 2005;19:1839–41. 177. Olivo C, Alblas J, Verweij V, et al. In vi vo bioluminescence imaging study to monitor ectopic bone for mation b y luciferase gene marked mesenchymal stem cells. J Or thop Res 2008;26:901–9. 178. Tanaka M, Swijnenburg RJ, Gunawan F, et al. In vivo visualization of cardiac allo graft rejection and traf ficking passenger leuk ocytes using bioluminescence imaging. Circulation 2005;112:I105–10. 179. Kim JB, Leucht P, Mor rell NT, et al. Visualizing in vi vo liposomal drug delivery in real-time. J Dr ug Target 2007;15:632–9. 180. Roberts ES, Malstrom SE, Dreher KL. In situ pulmonar y localization of air pollution par ticle-induced o xidative stress. J Toxicol Environ Health A 2007;70:1929–35. 181. Dohlen G, Odland HH, Carlsen H, et al. Antioxidant activity in the newborn brain: a luciferase mouse model. Neonatolo gy 2008; 93:125–31. 182. Su H, van Dam GM, Buis CI, et al. Spatiotemporal expression of heme oxygenase-1 detected b y in vi vo bioluminescence after hepatic ischemia in HO-1/Luc mice. Liver Transpl 2006;12:1634–9. 183. Wilmink GJ, Opalenik SR, Beckham JT, et al. Assessing laser-tissue damage with bioluminescent imaging. J Biomed Opt 2006; 11:041114. 184. Weir LR, Schenck E, Meakin J , et al. Biophotonic imaging in HO-1.luc transgenic mice: real-time demonstration of gender specific chlorofor m induced renal to xicity. Mutat Res 2005; 574:67–75. 185. Schimmelpfennig CH, Schulz S, Arber C, et al. Ex vi vo expanded dendritic cells home to T-cell zones of l ymphoid organs and survive in vi vo after allo geneic bone mar row transplantation. Am J Pathol 2005;167:1321–31. 186. Chan JK, Hamilton CA, Cheung MK, et al. Enhanced killing of primary o varian cancer b y retar geting autolo gous c ytokineinduced killer cells with bispecif ic antibodies: a preclinical study. Clin Cancer Res 2006;12:1859–67. 187. Kim D, Hung CF , Wu TC. Monitoring the traf ficking of adopti vely transferred antigen- specific CD8-positive T cells in vivo, using noninvasive luminescence imaging. Hum Gene Ther 2007;18:575–88. 188. Scheffold C, K ornacker M, Schef fold YC, et al. Visualization of effective tumor targeting by CD8+ natural killer T cells redirected with bispecif ic antibody F(ab’)(2)HER2xCD3. Cancer Res 2002;62:5785–91. 189. Zhang C, Lou J , Li N , et al. Donor CD8+ T cells mediate g raftversus-leukemia acti vity without clinical signs of g raft-versushost disease in recipients conditioned with anti-CD3 monoclonal antibody. J Immunol 2007;178:838–50. 190. Beilhack A, Schulz S, Baker J, et al. In vivo analyses of early events in acute g raft-versus-host disease re veal sequential inf iltration of T-cell subsets. Blood 2005;106:1113–22. 191. Lee MH, Lee WH, Van Y, et al. Image-guided analyses reveal that nonCD4 splenocytes contribute to CD4+ T cell-mediated inflammation leading to islet destr uction by altering their local function and not systemic trafficking patterns. Mol Imaging 2007;6:369–83.

192. Garcia T, Jackson A, Bachelier R, et al. A con venient clinicall y relevant model of human breast cancer bone metastasis. Clin Exp Metastasis 2008;25:33–42. 193. Cowey S, Szafran AA, Kappes J , et al. Breast cancer metastasis to bone: evaluation of bioluminescent imaging and microSPECT/CT for detecting bone metastasis in immunodef icient mice. Clin Exp Metastasis 2007;24:389–401. 194. Hensley H, P ollack A. In vi vo visualization of metastatic prostate cancer. Cancer Biol Ther 2003;2:661–2. 195. Rosol TJ, Tannehill-Gregg SH, LeRo y BE, et al. Animal models of bone metastasis. Cancer 2003;97:748–57. 196. Bloch S, Lesage F , McIntosh L, et al. Whole-body fluorescence lifetime imaging of a tumor -targeted near -infrared molecular probe in mice. J Biomed Opt 2005;10:054003. 197. Chaudhari AJ, Darvas F, Bading JR, et al. Hyperspectral and multispectral bioluminescence optical tomography for small animal imaging. Phys Med Biol 2005;50:5421–41. 198. Ntziachristos V, Ripoll J , Wang LV, Weissleder R. Looking and listening to light: the e volution of w hole-body photonic imaging. Nat Biotechnol 2005;23:313–20. 199. Burcin Unlu M, Gulsen G. Ef fects of the time dependence of a bioluminescent source on the tomo graphic reconstr uction. Appl Opt 2008;47:799–806. 200. Li S, Zhang Q, Jiang H. Two-dimensional bioluminescence tomo graphy: numerical simulations and phantom experiments. Appl Opt 2006;45:3390–4. 201. Dehghani H, Da vis SC, Jiang S, et al. Spectrall y resolved bioluminescence optical tomography. Opt Lett 2006;31:365–7. 202. Alexandrakis G, Rannou FR, Chatziioannou AF. Tomographic bioluminescence imaging b y use of a combined optical-PET (OPET) system: a computer simulation feasibility study . Phys Med Biol 2005;50:4225–41. 203. Dogdas B, Stout D, Chatziioannou AF, Leahy RM. Digimouse: a 3D whole body mouse atlas from CT and cryosection data. Phys Med Biol 2007;52:577–87. 204. Soloviev VY. Tomographic bioluminescence imaging with v arying boundary conditions. Appl Opt 2007;46:2778–84. 205. Ahn S, Chaudhari AJ, Dar vas F, et al. F ast iterati ve image reconstruction methods for full y 3D multispectral bioluminescence tomography. Phys Med Biol 2008;53:3921–42. 206. Kumar D, Cong WX, Wang G. Monte Carlo method for bioluminescence tomography. Indian J Exp Biol 2007;45:58–63. 207. Wang G, Cong W, Shen H, et al. Ov erview of bioluminescence tomography—a ne w molecular imaging modality . F ront Biosci 2008;13:1281–93. 208. Tonary AM, Pezacki JP. Simultaneous quantitati ve measurement of luciferase repor ter acti vity and cell number in tw o- and threedimensional cultures of hepatitis C virus replicons. Anal Biochem 2006;350:239–48. 209. Wood KV , Lam YA, McElro y WD. Introduction to beetle luciferases and their applications. J Biolumin Chemilumin 1989; 4:289–301. 210. Wiles S, Ferguson K, Stef anidou M, et al. Alternative luciferase for monitoring bacterial cells under adverse conditions. Appl Environ Microbiol 2005;71:3427–32. 211. Verhaegent M, Christopoulos TK. Recombinant Gaussia luciferase. Overexpression, purif ication, and analytical application of a bioluminescent repor ter for DN A h ybridization. Anal Chem 2002; 74:4378–85. 212. Shimomura O. A short story of aequorin. Biol Bull 1995;189:1–5. 213. Fisher AJ, Thompson TB, Thoden JB , et al. The 1.5-A resolution crystal str ucture of bacterial luciferase in lo w salt conditions. J Biol Chem 1996;271:21956–68.

9 OPTICAL MULTIMODALITY TECHNOLOGIES ARION F. CHATZIIOANNOU, PHD

In the pre vious chapters, v arious molecular imaging technologies and their combinations w ere introduced. As is commonplace with most technolo gical inno vations, new capabilities and features are often combined and inte grated with prior ar t, resulting in imaging systems that are f ar more capab le and v ersatile. We have already seen that optical bioluminescence imaging (BLI) has pro ven to be a tr uly e xceptional tool at the disposal of biological scientists (see Chapter 8, “Functional Imaging Using Bioluminescent Mark ers”).1,2 It combines a number of impor tant advantages, the most significant of which are (1) high sensitivity, (2) ease of operation, (3) low overall cost, and (4) an implementation that is an e xtension of assa ys that biolo gical researchers are already f amiliar with from the benchtop aspects of their w ork. When compared with radionuclide technolo gies, another k ey adv antage of the optical technologies is that the enzymatic reactions that produce visible light can continue as long as there is locall y a vailable ener gy, substrate, and enzyme, thereby producing thousands of photons per a vailable enzyme. In contrast, radionuclides produce a single emission from each parent isotope, based on a random decay process, meaning that there is no physical way of turning of f or quenching these emissions. Due to the unique combination of these adv antages, the penetration of in vi vo BLI into the instr umentation toolbox of molecular imaging scientists has been the f astest and highest compared with vir tually e very other in vi vo molecular imaging modality.3 On the flip side of these advantages is the limitation that visib le light propagation in li ving tissues is seriousl y hindered b y v ery strong scattering and strong absor ption, especially for wavelengths below 600 nm. 4 This fundamental ph ysical ef fect depends on the inherent optical proper ties of tissues, cells, and intracellular organelles and limits the application of BLI

to the superf icial tissues of lar ger animals, or the w hole body of v ery small mammals (mice). Fur thermore, this effect causes an une ven sensiti vity prof ile with strong preference for tar gets closer to the surf ace and tissues with low b lood content due to the absor ption of hemoglobin,5 and it precludes implementation in the general whole body sense in lar ger animals and cer tainly in humans.6 That does not mean that optical imaging is not applicable to humans, w here it can ha ve signif icant uses in endoscopic, intraoperati ve, and other similar conte xt applications. Another consequence of optical photon absorption and scattering is that the same physical effects also seriously affect image quantitation as well as spatial resolution, even within the small body of a mouse. 7,8 The adv antages listed abo ve are e xtremely compelling, and they have naturally stirred signif icant interest by biological researchers for in vivo BLI technologies. This de facto success has spa wned a ne w set of scientif ic endeavors, in which many research groups worldwide have embarked upon ef forts to provide (1) improved dedicated imaging instrumentation, (2) improved molecular imaging probes, and (3) methodolo gies for image reconstr uction algorithms, which will provide improved quantitation and spatial resolution (see Chapter 11, “Fluorescence Tomography” and Chapter 14, “Dif fuse Optical Tomography and Spectroscopy”). The goals of all of the abo ve efforts are two-fold: to first improve the qualitati ve and quantitati ve accuracies of routine optical imaging e xperiments and measurements, and second and perhaps most impor tant, to facilitate the translation of these measurements and the knowledge the y impar t to practical and useful clinical diagnostic applications in humans. 6 Achievement of these goals, as in essentiall y all aspects of molecular imaging, is a multidisciplinar y task that includes image reconstr uction algorithms, 8,9,10 photon propagation physics,11,12 advanced and sensitive 139

140

MOLECULAR IMAGING: PRINCIPLES AND PRACTICE

instrumentation technologies13, and the de velopment of novel molecular imaging probes. 14,15 One only needs to consider the significant leap of technologies from the in vitro realm of the cell culture flask and the petri dish, to the in vivo realm of small-animal imaging experiments, to the realm of human applications, to realize that multiple cr ucial v alidation e xperiments for both the technologies and the biolo gical conte xt are necessar y at each step. Fur thermore, in practice, radicall y dif ferent technologies have been optimized over the years and are used for each of these physical domains, ranging all the way from advanced microscopes, to clinical whole-body tomographic imaging systems (Figure 1). As stepping stones bridging this gap and f acilitating this translational validation, a series of technolo gies have been de veloped that span multiple molecular imaging modalities. These technolo gies include multimodality molecular imaging probes that are described in Chapter 29 “Multimodality Agents,” as w ell as multimodality imaging instr uments that include combinations of positron emission tomography (PET) and single photon emission computed tomo graphy (SPECT) systems with optical bioluminescence,16–19 combinations of optical and acoustic imaging detailed in Chapter 16 “Molecular Photoacoustic Tomography,” combinations of anatomic via X-rays and magnetic resonance imaging (MRI) with bioluminescence and fluorescence, 20–22 and e ven other optical imaging modalities such as optical transmission tomography.23–25 In this chapter, we will outline and describe the rationale and the goals behind the technical instr umentation development in multimodality PET/SPECT and X-ra y computed tomography (CT) with bioluminescence optical imaging and ho w the y f it together within the lar ger puzzle of molecular imaging. We will also discuss some of the technical hurdles that need to be overcome and how

A A

BB

Figure 1. A, Standard microscope, optimized for light collection from small specimens; B, Clinical imaging tomograph, optimized for whole-body surveys.

together with the work that describes the development of multimodality imaging probes in Chapter 29 “ Multimodality Agents,” these combinations should help us advance molecular imaging and bring us closer to the goals of the f ield.

MULTIMODALITY OPTICAL IMAGING GOALS One of the first issues that is considered within the context of the molecular imaging f ield is that the ultimate goal of an y of these de velopments is their application in human disease diagnosis and treatment. Some of the molecular imaging technologies have fundamental limitations when it comes to applications in humans as discussed abo ve. On one hand , optical imaging suf fers from poor tissue penetration. On the other hand , multiple high-sensiti vity molecular imaging modalities, whose response is independent of tar get depth, ha ve been established (PET/SPECT) and have been proven in the clinical environment.26,27 These modalities are based on imaging of radionuclide emissions, w hose fundamental infor mation car riers are man y orders of magnitude more energetic than those of optical photons in the visib le range (~10 5 eV vs 1–2 eV). Besides that significant dif ference, in se veral senses, in vi vo bioluminescence/fluorescence and PET/SPECT imaging are actually quite similar. They both are based on the measurement of photon signal emanating from a source inside the body and the y both depend on highl y sensitive photon detectors that are positioned outside the body. Furthermore, the generated signal at the source is, in principle, propor tional to the biolo gic process the y are examining. That is, these modalities are inherentl y dependent on direct signal measurements and not on signal contrast as in traditional MRI, 28 X-ray CT, and ultrasound applications. It is, therefore, these similarities that mak e it natural to consider that ne w optical molecular imaging probes and methodologies should be translated at their f inal stage into clinicall y a vailable radionuclide technologies. From the biolo gy side consequentl y, the o verall concept is to de velop and estab lish biological constr ucts that can be used in a similar f ashion to produce bioluminescence or fluorescence signal w hen used in small-animal preclinical models, or yield PET/SPECT signals, w hen used in the clinic with patients. Fur thermore, these dual probes can provide additional information when combined with intravital or local sur gical techniques, w here excised tissue can be visualized opticall y under a microscope. 29,30

Optical Multimodality Technologies

Because the optimization process of the de velopment of these molecular imaging methodolo gies is v ery time consuming and technicall y demanding, 31 the development of multimodality biological constructs is critical to f acilitate the ultimate translation of molecular imaging w ork based on bioluminescence and fluorescence on a biolo gist’s bench, to the clinic. Research from se veral g roups32–34 has yielded biological constr ucts acting as multimodality repor ter systems that can bind (or con vert and trap) either a PET/SPECT probe or produce emission of either bioluminescence or fluorescence photons from the same cells. Since the choice of these signals can be based on the experimental conte xt, the same biolo gical e xperiment can in essence be used and translated betw een in vi vo small-animal preclinical measurements and patients. This remarkab le technolo gical achie vement is an enabling technology that opens doors for translation of many molecular biolo gy e xperimental results based on optical bioluminescence and fluorescence into the clinical domain. Even under those circumstances, the tw o signals emanate as the result of dif ferent protein products, 32 and there is an underlying assumption that the two signals will be if not equal, at least proportional with each other.35 Because the measurements resulting from these biolo gic e xperiments should be directl y related to the le vel of produced photon signals, an impor tant step is the validation process for the underlying photon signal flux. This process naturall y entails the use of multimodality molecular imaging instruments that can simultaneously detect and quantify the signal from both optical and radionuclide emissions. These instruments should provide biologists with a tool to directl y quantify the signal under multiple dif ferent e xperimental conditions. A number of research g roups have developed or are in the process of de veloping such technologies. They are combinations of PET and SPECT systems with optical bioluminescence and fluorescence. With these imaging systems, one can simultaneously detect optical and radionuclide signals and examine different aspects of the complete imaging process that can be schematically described here in the following steps for imaging of gene e xpression: (1) delivery of the repor ter and therapeutic genes via the vector, (2) delivery of the imaging probe, (3) binding of the probe to its tar get, (4) clearance of the nonspecifically bound probe from its tar get, (5) signal generation at the source, and (6) signal propagation from the source through the tissues.

141

OPTICAL SIGNAL QUANTITATION A key technical parameter in this process is the quantif ication of optical signal at its source, deep within the body, which is something v ery different than the quantif ication of the signal emanating from the surf ace of a preclinical model. Due to the significant photon attenuation and scattering in tissues, absolute quantif ication and position information of optical signal at its source is a v ery difficult problem indeed and is still in its earl y stages. 36–38 In the de velopment process of these multimodality instr uments, as w ell as for the single modality imaging instr uments, it has been shown that the quantification of photon signal is highly dependent on (1) the exact optical properties of the tissues, (2) their spatial distribution, and (3) the wavelength of light at the source. Although (2) and (3) above can in principle be considered constants for each measurement in each animal, the e xact scattering and absorption coefficients of tissues could be affected by the physiologic state of the model at the time of the study , including b lood flow, fur ther complicating the measurements. Ev en so, as a result of the ef forts to pro vide answers to items (1) and (2) above, researchers have been led to the creation of multimodality optical imaging instruments that include strictly anatomic imaging instr uments such as X-ra y CT , MRI, and also systems that include optical transmission measurements. 9,21,25,39 The problem of robust reconstr uction of optical signal at the source, e ven with all this a priori infor mation is still exceedingly difficult and has y et to reliab ly replace semiquantitative estimates. It should be noted here that this is a significant research area in dif fuse optical tomo graphy (DOT) and that the concept of signal quantif ication at the source has multiple le vels of ans wers, ranging from the qualitative decision about a signal being increased or decreased “up or down,” to log order quantif ication, to the absolute quantif ication of the number of enzymatic reactions taking place in each second at the tar get site. Ev en though many researchers are stri ving for the latter def inition in tight conf idence intervals in an attempt to provide a truly quantitative explanation of the underlying biology,40,41 the f act is that biolo gical research has made signif icant progress by using simpler quantif ication schemes that lend themselves to easier data interpretation and higher throughput methodologies. A prime example of this approach is traditional BLI, which tends to serve at an earlier stage of the discovery process and feeds into more quantitati ve, but lower throughput and higher cost, imaging modalities.31 We will summarize here some of the technical aspects of these technically challenging and fascinating efforts.

142

MOLECULAR IMAGING: PRINCIPLES AND PRACTICE

PRINCIPLES OF OPTICAL AND RADIONUCLIDE IMAGING For the reader to understand the rationale behind the technologies used in these combinations, we will lay out here and summarize the principles of optical and radionuclide imaging. Although both optical and radionuclide imaging use electromagnetic radiation in the for m of photons for signal transport, the energy between the two is vastly different. This is a mixed blessing in the sense that although optical photons are easil y absorbed in v ery small distances in low density, low atomic number materials such as silicon (Si), the y are at the same time hea vily attenuated and scattered b y tissues. Therefore, it is onl y necessary to place a small amount of Si as in a typical charge-coupled de vice (CCD) camera pix el to e xactly determine the location of photon interaction. Advanced optical f ilters w hen used in the photon path can also selectively block or diffract different wavelengths, adding color information in the image, at some cost in sensiti vity. When standard optics are also placed in front of the CCD camera, the y ef fectively collimate the signal and thereby provide the direction of travel in a scattering free media such as air (Figure 2A). This information provides the point of origin of these photons from the last scattering position just under neath the tissue surf ace, in the form of an image of the surf ace distribution of photons. In contrast, photons emitted by radionuclides have a very low probability of interaction in most low atomic number Z materials typicall y in volved in tissues. To ef fectively stop these photons, specialized high-density and high atomic number scintillator materials are used such as LaBr, NaI, BGO, and LSO.42 Alternatively, a few systems are using solid-state semiconductor detector materials such as CdTe and TlBr.43 The scintillators act as energy modulators. They con vert the high-ener gy photons to

multiple low-energy photons in the visib le range, essentially the same photons that are used in bioluminescence and fluorescence. These photons can in tur n be detected with similar technolo gies as optical photons since the y have the same ener gy range. There is a k ey dif ference though, that a single high-energy photon from a radionuclide emission is now converted into a g roup of multiple photons numbering in the fe w hundreds to a fe w thousands, emitted typicall y within a fe w tens to a fe w hundreds of nanoseconds. 44 The number of these photons is directly related to the ener gy of the originating radionuclide emission. Therefore, PET and SPECT systems need to accumulate and measure the total amount of optical photons in this scintillation pulse within a small time window, to estimate the ener gy and ar rival time of the radionuclide emission and use this infor mation to estimate the coincidence time and ener gy for PET , and energy discrimination for SPECT . The ener gy infor mation for both PET and SPECT can be used to help in the rejection of scattered high-ener gy photon e vents in both the patient and in the detectors. Because these high-ener gy γ-ray photons are so penetrating, no conventional lens systems similar to those available for optical photons e xist, although attempts to create such systems are ongoing. 45 The principle of operation of these γ-ray optical lenses is based on small angle scattering without ener gy loss, and the best cur rent systems typically work at lower energies (< 40 keV) and with small scattering angles under typicall y low efficiencies.46 Therefore, for SPECT imaging, the direction of photon travel needs to be distinguished b y selective photon rejection accomplished through a hea vy absorber , called a collimator (Figure 2B). This collimator is typicall y made out of lead, or other high atomic number and density material, including e xotic options such as depleted uranium. 47 This device works effectively as a lens in optical systems

Scintillation detector

CCD

Collimator

PET/Optical detector

A A

B B

C

Figure 2. A, Traditional lens-based imaging system; B, Traditional single photon radionuclide detector module; and C, Combined positron emission tomography (PET)/optical detector.18

Optical Multimodality Technologies

and also typically rejects a very large fraction of the overall photons (10 −4 for a parallel hole collimator). In contrast, one of the k ey technical dif ferences betw een PET and SPECT is that this mechanical collimation is not necessary for PET. The direction of photon travel is determined electronically by nanosecond coincidence detection.48 This key difference pro vides a signif icant intrinsic adv antage to PET in ter ms of o verall system sensiti vity per unit of injected radionuclide activity.

SIMULTANEOUS OPTICAL AND RADIONUCLIDE IMAGING With these principles of image for mation in mind , the following methodologies of combining the tw o modalities become ob vious: (1) use tw o completel y separate imaging instr uments, inte grated perhaps in a single gantry and (2) use the same detector technolo gy and same physical detector, with the addition of an ener gy modulator at the front end to enab le detection of both photon signals. Although the f irst approach has merits in its simplicity and ease of implementation, it generates two separate data sets that require spatial co-registration and can possib ly have a higher cost. This methodology has been the choice of a number of research g roups and will be discussed belo w. The second approach is more technically challenging and less straightforw ard to implement, but it is also potentially lower in overall cost and should pro vide inherentl y spatiall y co-re gistered images. It has been the method of choice of a smaller set of research groups and will also be discussed belo w.

COMBINING BIOLUMINESCENCE WITH RADIONUCLIDES THROUGH SEPARATE INSTRUMENTS The first interest in the combination of radionuclide technologies with optical imaging was reported by Huber and colleagues.49 That initial system combined a single photon radionuclide detector with a parallel hole collimator for the detection and imaging of Tc-99m, with a highsensitivity CCD camera and an optical lens for the detection of bioluminescence. The radionuclide detector w as based on a cesium iodide (CsI) scintillator , coupled to a photodiode ar ray with specialized high-perfor mance electronics. The single nonrotating head radionuclide imaging system allowed for a significant amount of clear space that w ould accommodate the optical imaging system, and the tw o systems could operate simultaneousl y

143

and independentl y. A similar approach w as tak en b y another g roup,16 who combined a planar single photon detector with a fluorescence imaging system. The initial single photon radionuclide detector was based on Si technology that had rather lo w detection ef ficiency resulting in a lo w sensiti vity system. The detector w as later replaced by CdTe that works very well for single photon emitters up to about 140 k eV. The optical e xcitation source of this instr ument w as based on raster scanning illumination of the subject with a g reen pulsed laser through a scanning mir ror, and detection through a standard CCD camera and lens as reported in the first in vivo results.50 The fluorophore they used was intramuscularly injected hematoporphyrin that accumulated in the tumor xenografts. These early, projection-only systems were followed b y a commercial v endor system that inte grated bioluminescence, fluorescence, and X-ra y projection imaging in a single device, which later acquired the capability to image noncollimated radionuclide emissions with low spatial resolution. 51 Several other groups developed inte grations of tomo graphic systems; one of the earliest ones developed a tomographic SPECT and multiprojection optical detection system, 52 which used a pinhole collimator placed in the front of an ar ray of NaI scintillators, read out by large area flat panel photomultiplier tubes. The optical imaging system w as based on a lens coupled, cooled CCD camera, and a set of mir rors placed in the front of the pinhole collimator , providing the same view as the radionuclide image. A laser source was a vailable to pro vide raster scanning of the subject with excitation light for fluorescence imaging. In a later development, the same group developed the concept of a more advanced system that focused on PET and provided the additional capability of bioluminescence and fluorescence imaging 53 (Figure 3A). For the optical data detection, instead of a single lens and CCD , the g roup proposed the use of multiple lar ge area Complementar y Metal Oxide Semiconductor (CMOS) detectors and instead of con ventional lenses for visib le light collimation, it used a specially designed micro-lens array and an optical septum positioned in front of each of the Si-based detectors. The low atomic number of the CMOS detectors, together with the lo w mass of the micro-lens ar ray, allowed positioning of the optical detectors inside the PET gantry, with minimal attenuation of the PET photons based on simulations. Laser excitation was made possible with thin optical f ibers positioned betw een the scintillator detectors. The complete system concept w as demonstrated b y a combination of simulations and measurements on a clinical PET tomo graph.

144

MOLECULAR IMAGING: PRINCIPLES AND PRACTICE

AA

B B

Figure 3. A, Combined positron emission tomography (PET) and optical bioluminescence imaging (BLI) and fluorescence imaging system, with separate detectors53; B, Optical-PET (OPET), combined PET and bioluminescence imaging system using the same detectors.18

COMBINING BIOLUMINESCENCE WITH RADIONUCLIDES THROUGH ONE INSTRUMENT The more challenging approach of using the similarity in the nature of optical and radionuclide signals w as taken by another g roup18 that combined a PET and a BLI instrument, using the same detectors. The scintillators were used for the dual role of ener gy modulation and light guide, providing improved light collection from the subject (F igure 2C) w hile photomultiplier tubes w ere used as photodetectors. It should be noted here that the overall light collection of lens-based optical imaging systems is typically very low, significantly smaller than 1%, even for large aperture optics.54 In effect, optical imaging systems e xchange fle xibility and ease of operation for absolute detection sensitivity. Direct coupling of the light detectors to the source with optical f ibers and optical fiber bundles would allow a signif icant increase in their light collection efficiency, equivalent to one or two orders of magnitude. 55 In this approach for the combination of optical and PET detection, the optical light collection efficiency w as also considerab ly increased in a similar fashion as in f iber-optic coupled systems, at the e xpense of f ield-of-view flexibility and system throughput capability. Despite the improved geometry for light collection efficiency b y that approach, the use of photomultiplier tubes with their inherently low quantum efficiency counteracted somewhat this increase in sensitivity. With a single detector , the system could detect both optical and radionuclide emissions, and specialized electronics enabled separation betw een the visib le light photons emanating from bioluminescence, and the pulses of light emanating from the interaction of radionuclide emissions

in the scintillators.56 At a later development, other groups followed this approach and created similar detectors with additional features such as multiple la yers that pro vide depth of interaction capability for the PET signal, and dichroic mir rors that impro ved the collection ef ficiency of the scintillation light. 57

OPTICAL SOURCE RECONSTRUCTION As se veral g roups combined their ef forts to de velop these combined instr uments, the other aspect of the technology that in volves mathematical modeling of photon tracks, interactions, and signal reconstr uction demonstrated that to resolve the issue of 3D optical photon quantification, in the general sense and in heterogeneous li ving mouse tissues, a priori infor mation on optical properties of tissues in situ is required , together with anatomic infor mation of the spatial distrib ution of these tissues. 23,58–60 When in addition to that infor mation, extra knowledge from the w avelength of the emitted and detected photons is used , successful reconstructions for the location and magnitude of a source inside a mouse w as achieved.7,23 This result was demonstrated for both bioluminescence and fluorescence and uses the information from differential photon absorption and scattering as a function of w avelength. On the basis of these results, a number of g roups decided to fuse anatomic imaging modalities lik e X-ray CT and MRI, with their v olumetric high-spatial resolution, to gether with or gan se gmentation softw are and atlas and lookup tab le based results of tissue optical properties toward image reconstruction.20–22 Although these ef forts ha ve produced good results in simulation e xperiments, despite the signif icant

Optical Multimodality Technologies

mathematical challenges the y had to o vercome in this ill-posed prob lem,12 the f act is that e ven the best estimates of optical properties from the literature4,7 are based on measurements from e xcised or frozen tissue samples or from dif ferent species alto gether. Fur thermore, biologic motions and pock ets of air in the lungs, the intestines, and clear fluid in the stomach that mo ve during a study create prob lems in the modif ied dif fusion approximation or other models used for most solutions. Consequently, these atlas-based results are suboptimal w hen used in vi vo and methods that measure the optical proper ties in vi vo, in situ are being de veloped. Those include illumination with a series of optical f iber sources, measurements from multiple vie ws, and other simpler methods with a local f iber next to the e xit measurement point, w hich pro vide a good estimate of the local proper ties at the time of the e xperiment.24 These produce a better estimate of the source intensity with a more acceptable error tolerance.

FUTURE DIRECTIONS A signif icant wealth of kno wledge has resulted from the flurry of the scientific efforts described here. Technologies from optical multimodality probes to optical multimodality instr uments ha ve been de veloped and as this f ield evolves, a few new technological requirements are slo wly establishing themselves. One important and clear aspect is that a convergence of fields seems to be taking place, with the fields of fluorescence tomography (Chapter 11, “Fluorescence Tomography”), DOT (Chapter 14, “Diffuse Optical Tomography and Spectroscop y”), and multimodality BLI/fluorescence imaging coming closer to gether, to produce more quantitati ve and more accurate reconstr ucted results. That is certainly not a surprising or unexpected outcome, since all these modalities rel y heavily on accurate modeling of visib le light propagation of dif ferent w avelengths in tissues. Therefore, w hen the goal is to obtain quantitative infor mation about a ne w molecular imaging probe, instruments that can perform both optical transmission and optical emission measurements in a similar f ashion as a PET/SPECT and X-ra y CT system are a natural choice. This advancement will cer tainly benef it all three optical imaging modalities, bioluminescence, fluorescence, and DO T. The inclusion of additional imaging modalities such as PET and SPECT with their X-ray CT counter part will be a natural continuation that will enable the development and optimization of both optical and radionuclide probes. It will also allo w the ans wer of biological questions in a multiparameter space, on both the biodistribution of a probe and its tar get acti vity, and possibly its interactions in situ in vivo.

145

CONCLUSIONS In this very exciting f ield, many bright minds are producing outstanding scientif ic results and are solving dif ficult problems. It still remains to be seen though w hich, if any, of the methodolo gies will pro ve to be clearl y superior to others in terms of ease of use, accuracy of results, and cost. As we all strive for the quest of the ultimate multimodality instrument, we need to be reminded of the reason why this quest be gan, that is the enrichment of the infor mation obtained from preclinical molecular imaging studies and the translation of this infor mation into the clinical realm. Optical bioluminescence is attracti ve due to its simplicity , low cost, high throughput, lo w backg round, and ease of use. A methodology that would enable and simplify optical signal transmission measurements, in a similar f ashion as transmission scans did for PET and SPECT ,61 would greatly improve the accuracy of optical image reconstr uctions and the quantif ication of the signal. The integration of such a de vice with other imaging modalities such as PET and SPECT will be very useful indeed.

REFERENCES 1. Christopher H, Contag BDR. It’s not just about anatomy: in vivo bioluminescence imaging as an eyepiece into biology. J Magn Reson Imaging 2002;16:378–87. 2. Negrin RS, Contag CH. In vi vo imaging using bioluminescence: a tool for probing g raft-versus-host disease. Nat Re v Immunol 2006;6:484–90. 3. Cherry SR. Multimodality in vi vo imaging systems: twice the po wer or double the trouble? Annu Rev Biomed Eng 2006;8:35–62. 4. Cheong WF, Prahl SA, Welch AJ. A review of the optical-proper ties of biological tissues. IEEE J Quantum Electron 1990;26:2166–85. 5. Srinivasan S, Pogue BW, Jiang S, et al. Inter preting hemoglobin and water concentration, o xygen saturation, and scattering measured in vivo b y near-infrared breast tomo graphy. Proc Natl Acad Sci 2003;100:12349–54. 6. Culver J, Akers W, Achilefu S. Multimodality molecular imaging with combined optical and SPECT/PET modalities. J Nucl Med 2008;49:169–72. 7. Alexandrakis G, Rannou FR, Chatziioannou AF. Ef fect of optical property estimation accurac y on tomo graphic bioluminescence imaging: simulation of a combined optical-PET (OPET) system. Phys Med Biol 2006;51:2045–53. 8. Rice BW, Cable MD, Nelson MB . In vi vo imaging of light-emitting probes. J Biomed Opt 2001;6:432–40. 9. Dehghani H, Da vis SC, Jiang S, et al. Spectrall y resolv ed bioluminescence optical tomography. Opt Lett 2006;31:365–7. 10. Hebert T, Leah y R. A generalized EM algorithm for 3-D Ba yesian reconstruction from Poisson data using Gibbs priors. IEEE Trans Med Imaging 1989;8:194–202. 11. Arridge SR, Schw eiger M, Hiraoka M, Delp y DT. A f inite element approach for modeling photon transpor t in tissue. Med Ph ys 1993;20(2 Pt 1):299–309. 12. Boas DA, Brooks DH, Miller EL, et al. Imaging the body with diffuse optical tomography. Signal Processing Mag IEEE 2001;18:57–75. 13. Contag CH, Spilman SD, Contag PR, et al. Visualizing gene expression in li ving mammals using a bioluminescent repor ter. Photochem Photobiol 1997;66:523–31.

146

MOLECULAR IMAGING: PRINCIPLES AND PRACTICE

14. Herschman HR. Molecular imaging: looking at prob lems, seeing solutions. Science 2003;302:605–8. 15. Weissleder R, Ntziachristos V. Scaling down imaging: molecular mapping of cancer in mice. Nat Med 2003;9:123–8. 16. Celentano L, Laccetti P, Liuzzi R, et al. Preliminar y tests of a prototype system for optical and radionuclide imaging in small animals. Nucl Sci IEEE Trans 2003;50(5 Part 2):1693–701. 17. Peter J, Schulz RB, Semmler W. PET-MOT—a novel concept for simultaneous positron and optical tomography in small animals. In: 2005 IEEE, Nuclear Science Symposium Conference Record, 2005. 18. Prout DL, Silv erman R W, Chatziioannou A. Detector concept for OPET—A combined PET and optical Imaging system. IEEE Trans Nucl Sci 2004;51:752–6. 19. Tsyganov EN, Antich PP, Kulkarni PV, et al. Micro-SPECT combined with 3D optical imaging. In: 2004 IEEE, Nuclear Science Symposium Conference Record, 2004. 20. Allard M, Côté D, Davidson L, et al. Combined magnetic resonance and bioluminescence imaging of li ve mice. J Biomed Opt 2007; 12:1083–3668. 21. Joshi A, Rasmussen JC, Kw on S, et al. Multi-modality CT -PET-NIR fluorescence tomography. In: Biomedical Imaging: From Nano to Macro, 2008. ISBI 2008. 5th IEEE Inter national Symposium on 2008. 22. Yujie L, Jie T, Wenxiang C, Ge W. Experimental study on bioluminescence tomo graphy with multimodality fusion. J Biomed Imaging 2007;2007:9–9. 23. Chaudhari AJ, Dar vas F, Bading JR, et al. Hyperspectral and multispectral bioluminescence optical tomo graphy for small animal imaging. Phys Med Biol 2005;50:5421–41. 24. Comsa DC, Farrell TJ, Patterson MS. Quantitative fluorescence imaging of point-lik e sources in small animals. Ph ys Med Biol 2008; 53:5797–814. 25. Gulsen G, Xiong B, Birgul O, Nalcioglu O. Design and implementation of a multifrequency near-infrared diffuse optical tomography system. J Biomed Opt 2006;11:014020. 26. Phelps ME. PET: the merging of biology and imaging into molecular imaging. J Nucl Med 2000;41:661–81. 27. Sharma V, Luker GD, Piwnica-Worms D. Molecular imaging of gene expression and protein function in vi vo with PET and SPECT . J Magn Reson Imaging 2002;16:336–51. 28. Gleich B, Weizenecker J. Tomographic imaging using the nonlinear response of magnetic particles. Nature 2005;435:1214–7. 29. Blasberg RG. In vi vo molecular -genetic imaging: multi-modality nuclear and optical combinations. Nucl Med Biol 2003;30:879–88. 30. Cai W, Chen K, Li ZB , et al. Dual-function probe for PET and near-infrared fluorescence imaging of tumor v asculature. J Nucl Med 2007;48:1862–70. 31. Massoud TF, Gambhir SS. Molecular imaging in living subjects: seeing fundamental biolo gical processes in a ne w light. Genes De v 2003;17:545–80. 32. Ray P, De A, Min JJ, et al. Imaging tri-fusion multimodality repor ter gene expression in living subjects. AACR 2004;302:605–8. 33. Ponomarev V, Doubrovin M, Serganova I, et al. A novel triple-modality reporter gene for whole-body fluorescent, bioluminescent, and nuclear nonin vasive imaging. Eur J Nucl Med Mol Imaging 2004;31:740–51. 34. Kesarwala AH, Prior JL, Sun J, et al. Second-generation triple repor ter for bioluminescence, micro-positron emission tomography, and fluorescence imaging. Mol Imaging 2006;5:465–74. 35. Gambhir SS, Herschman HR, Cher ry SR, et al. Imaging transgene expression with radionuclide imaging technolo gies. Neoplasia 2000;2:118–38. 36. Hielscher AH. Optical tomo graphic imaging of small animals. Cur r Opin Biotechnol 2005;16:79–88. 37. Kuo C, Coquoz O , Troy TL, et al. Three-dimensional reconstruction of in vivo bioluminescent sources based on multispectral imaging. J Biomed Opt 2007;12:024007.

38. Ntziachristos V, Ripoll J, Wang LV, Weissleder R. Looking and listening to light: the e volution of w hole-body photonic imaging. Nat Biotechnol 2005;23:313–20. 39. Zavattini G, Vecchi S, Mitchell G, et al. A hyperspectral fluorescence system for 3D in vi vo optical imaging. Ph ys Med Biol 2006;51:2029–43. 40. Phelps ME. PET : molecular imaging and its biolo gical applications. New York: Springer; 2004. 41. Kim SJ, Doudet DJ, Studenov AR, et al. Quantitati ve micro positron emission tomography (PET) imaging for the in vivo determination of pancreatic islet graft survival. Nat Med 2006;12:1423–8. 42. van Eijk CWE. Inor ganic scintillators in medical imaging detectors. Nucl Inst Methods Phys Res A 2003;509:17–25. 43. Madsen MT . Recent adv ances in SPECT imaging. J Nucl Med 2007;48:661–73. 44. Knoll GF. Radiation detection and measurement. 3rd ed. Ne w York: Wiley and Sons; 2000. 45. Snigirev A, Kohn V, Snigireva I, Lengeler B . A compound refractive lens for focusing high-energy X-rays. Nature 1996;384:49–51. 46. MacDonald CA, Gibson WM. Applications and advances in polycapillary optics. X-Ray Spectrum 2003;32:258–68. 47. Meikle SR, K ench P, Kassiou M, Banati RB . Small animal SPECT and its place in the matrix of molecular imaging technolo gies. Phys Med Biol 2005;50:R45–61. 48. Phelps ME. Application of annihilation coincidence detection to transaxial reconstruction tomography. J Nucl Med 1975;16:210–24. 49. Huber JS, Sudar D, Moses WW. Conceptual design of a dual modality optical and radionuclide imaging camera. In: High Resolution Imaging In Small Animals. Rockville, MD: 2001. 50. Autiero M, Celentano L, Cozzolino R, et al. Experimental study on in vivo optical and radionuclide imaging in small animals. Nucl Sci IEEE Trans 2005;52:205–9. 51. Feke GD, Lee vy WM, Or ton S, et al. Har nessing multimodality to enhance quantif ication and reproducibility of in vi vo molecular imaging data. Nat Met 2008;5. 52. Peter J, Ruehle H, Stamm V, et al. De velopment and initial results of a dual-modality SPECT/optical small animal imager . IEEE Nuclear Science Symposium Conference Record; 2005. 53. Peter J, Unholtz D, Schulz RB, et al. De velopment and initial results of a tomo graphic dual-modality positron/optical small animal imager. Nucl Sci IEEE Trans 2007;54:1553–60. 54. Liu H, Karellas A, Harris LJ, D’Orsi CJ. Methods to calculate the lens efficiency in optically coupled CCD X-ray imaging systems. Med Phys 1994;21:1193–95. 55. Liu H, Karellas A, Har ris L, D’Orsi C. Optical proper ties of f iber tapers and their impact on the performance of a fiber optically coupled CCD X-ray imaging system. SPIE; 1993. 56. Prout DL, Silv erman RW, Chatziioannou A. Readout of the optical PET (OPET) detector. Nucl Sci IEEE Trans 2005;52:28–32. 57. Takahashi K, Inadama N , Murayama H, et al. Preliminar y study of a DOI-PET detector with optical imaging capability. IEEE Nuclear Science Symposium Conference Record; 2007. 58. Alexandrakis G, Rannou FR, Chatziioannou AF. Tomographic bioluminescence imaging b y use of a combined optical-PET (OPET) system: a computer simulation feasibility study . Phys Med Biol 2005;50:4225–41. 59. Guven M, Yazici B, Intes X, Chance B. Diffuse optical tomography with a priori anatomical infor mation. Ph ys Med Biol 2005; 50:2837–58. 60. Li A, Boverman G, Zhang Y, et al. Optimal linear in verse solution with multiple priors in dif fuse optical tomo graphy. Appl Opt 2005;44:1948–56. 61. Kinahan PE, Townsend DW, Beyer T, Sashin D. Attenuation correction for a combined 3D PET/CT scanner. Med Phys 1998;25:2046–53.

10 FIBER OPTIC FLUORESCENCE IMAGING RABI UPADHYAY, BS AND UMAR MAHMOOD, MD, PHD

The clinical endoscope has ser ved as an imaging tool since the se venteenth centur y for e xaminations of the canals and cavities of the human body. Even when compared to cross-sectional imaging modalities toda y, modern fiber-optic catheters intrinsically provide high spatial resolution images of anatomic aber rations ranging from subtle mucosal patterns to gross luminal narrowings. The images are a vailable in real time during acquisition, and the technology requires minimal device and maintenance costs. Hence, f iber-optic catheters pro vide a lo w bar rier of entry for clinical translation of fluorescence molecular imaging and much potential for rapid testing and de velopment of new probes and disease models. This chapter will begin with a brief overview of current endoscopy implementations and disease applications to describe the standard of care and the unmet clinical needs that moti vate better imaging methods. In our discussion of preclinical f iber-optic technolo gies, w e will begin with spectroscopic methods for molecular sensing and se gue to techniques that create tr ue spatial images. The imaging will be g rouped b y methods that use endogenous fluorophores, basic exogenous fluorophores, and sophisticated tar geted and acti vatable molecular probes. We will conclude with a brief discussion of the necessary instrumentation and current design constraints for fiber-optic fluorescence imaging.

CLINICAL STATE OF THE ART Direct visualization of colorectal cancer and precancerous adenomas by minimally invasive colonoscopy has been used over decades, is a mainstay of early detection toda y through screening pro grams,1 and per mits immediate treatments such as pol ypectomy w here applicable. Other widespread endoscopic applications include inspection of the upper gastrointestinal tract for Barrett’s esophagus, laparoscopy of the peritoneal cavity,

bronchoscopy of the respirator y tract, c ytoscopy of the urinary tract, and mediastinoscopy of the thorax. In addition, inter ventional cardiolo gists and radiolo gists use percutaneous coronar y angioscop y with f iber-optic catheters to deter mine plaque mor phology, guide stent placement, and repair aneur ysms. The high spatial resolution of f iber optics has enabled the earl y detection of v ery small anatomic changes and the differentiation of adjacent neoplastic and nonneoplastic lesions on or gan surf aces or within lumens. In complement, the high temporal resolution has enabled immediate inter vention such as the biopsy of neoplastic lesions leaving behind surrounding healthy tissue in situ. It is impor tant to note these primar y advantages as the introduction of molecular imaging into the clinic should onl y enhance traditional endoscop y and should not diminish cur rent functions. The resolution of most clinical endoscopes is a function of the number of individual imaging fibers that can be fit within the fiber bundle of the catheter. Typical numbers range from less than ten thousand f ibers within a 0.8 mm angioscope to several hundred thousand fibers within a 3 cm colonoscope. This number is fur ther limited w hen a w orking channel for biopsy forceps or fluid deli very consumes par t of the cross-sectional area a vailable for imaging and illumination f ibers. Recentl y, microchip cameras able to f it onto the tips of catheters ha ve largely supplanted traditional f iber-optic-based endoscopes thereby fur ther impro ving image resolution and clarity . These cameras ha ve also been engineered to f it entirely within a capsule. 2 Such systems are ingested b y the patient, image snapshots of the entire gastrointestinal tract during their passage, and are par ticularly helpful in the detection of small bowel pathology. However, during colonic e valuation with capsule endoscopy, the stochastic time and location of each image acquisition has been sho wn to miss a number of 147

148

MOLECULAR IMAGING: PRINCIPLES AND PRACTICE

prominent lesions. 3 In f act, e ven f iber-optic-based adenoma detection has ultimately been shown to be largely operator and time dependent,4 given the subtlety of detecting earl y lesions across a lar ge organ surf ace. In studies using back-to-back colonoscopies, more than 20% of colonic pol yps w ere missed. 5 Such anatomic imaging is e ven more lik ely to miss flat lesions, w hich may ha ve higher rates of dysplasia compared with polyps.6 Direct visualization of adenomas and colorectal cancer b y endoscopic methods remains the clinical standard, but engineering impro vements of the white light (WL) imaging paradigm, by themselves, are unlikely to resolve the missed lesion rate. This problem is further exacerbated, for instance, in ulcerati ve colitis in which dysplasia can develop in macroscopically normal appearing mucosa. Cur rent colonoscopic sur veillance in patients with ulcerative colitis relies on random biopsies throughout the colon, which is relatively insensitive and cumbersome. 7 In the case of intraperitoneal spread of cancer , intraoperati ve detection of small metastatic foci may be limited by the similar luminosity of tumors compared with adjacent nor mal tissue. These clinical needs ma y be addressed in par t b y using no vel approaches that combine ne w f iber-optic devices and fluorescence detection. As we co ver each molecular imaging technology, we can inquire into its usefulness compared with standard endoscopy. Is it practical to cover large organ surface areas with the technique or is the sampling as stochastic as random biopsies? Does the acquisition proceed in real time allowing for simultaneous inter vention? How deep into the mucosal surf ace can be imaged? What is the sensitivity and specificity for the disease application?

SCATTERING SPECTROSCOPY Some of the methods f irst developed to glean molecular information with fiber optics include spectroscopic measurements of photon scattering in epithelial tissue. Because the f iber optics that detect photon scattering events ha ve matured, the y ha ve been slo wly introduced into the working channels of endoscopes for human studies in the detection of v arious epithelial pathologies. Two types of photon scattering b y a molecule e xist: elastic and inelastic (or Raman) scattering. 8 Both Raman scattering and fluorescence alter the optical w avelength but through dif ferent mechanisms. In both cases, the excited molecule relaxes to an energy level of the ground state and emits a photon. In Raman scattering, a Stok es transition occurs when an interacting photon is less energetic after interaction and an Anti-Stokes transition

occurs w hen the interacting photon is more ener getic after interaction with the molecule. Inelastic scattering more typically shows specific chemical composition and molecular str ucture of tissue, w hereas elastic scattering more typically shows the size distribution of the scatterers. Both techniques are amenab le to in vi vo measurements as the photon flux and excitation wavelengths used are nondestructive to the tissue. 9 In practice, elastic light scattering spectroscop y (LSS) has been used to deter mine the size distribution of cell nuclei.10 The diameter of nondysplastic cell nuclei is typically 5 to 10 µm, w hereas dysplastic nuclei are often larger, up to 20 µm across. Epithelial cell nuclei can be modeled as transparent spheroids w hose refractive index is higher than that of the sur rounding c ytoplasm. The backscattered light characteristically varies depending on nuclear size and refracti ve inde x. F or a collection of nuclei of different sizes, the light-scattering signal is a superposition of these v ariations, enab ling the nuclear size distribution and refractive index to be determined from the spectrum of light backscattered from the nuclei. Once the nuclear size distribution and refractive index are known, quantitati ve measures of nuclear enlar gement, crowding, and hyperchromasia can be obtained. These are the same criteria used by pathologists to diagnose tissue biopsies for mucosal dysplasia. The potential of LSS has been tested in multiple patient studies 11,12 to diagnose dysplasia and carcinoma in situ in different human organs with different types of epithelium: columnar epithelia of the colon and esophagus, transitional epithelium of the urinar y bladder, and squamous epithelium of the oral ca vity. The technique delivers a weak pulse of WL through a f iber-optic bundle threaded through a standard endoscope. After pulsing a 1-mm2 tissue surface on the order of milliseconds at wavelengths of 350 to 650 nm, the fiber bundle collects the diffusely reflected light. The spectra consist of a lar ge background from submucosal tissue, on which is superimposed a small (2 to 3%) component secondary to scattering by cell nuclei in the mucosal la yer. Although the spectral analysis can be cumbersome, pre vious datasets can help create decision algorithms for immediate histologic classification based on nuclear enlargement and density. Beyond a certain threshold nuclear diameter and population, a sampled tissue location may be designated as dysplasia or carcinoma. Hence, LSS has shown a strong potential to detect epithelial pre-cancerous lesions in an objective manner. In the same vein as LSS, Raman (inelastic scattering) spectroscopy also relies on pre-def ined spectroscopic models of v arious patholo gies. It interrogates the vibrations of molecular bonds and pro vides a direct

Fiber Optic Fluorescence Imaging

method to quantify the chemical composition of biological tissue. The modeling approach is based on the assumption that the Raman spectr um of a mixture is a combination of the spectra of its components and that signal intensity and chemical concentration are linearl y related. The resulting fit coefficients yield the contribution of each basis spectrum to the macroscopic tissue spectr um thereby elucidating the chemical mor phological mak eup of the lesion. Spectroscopic models 13 usually f it macroscopic tissue spectra with a linear combination of basis spectra from Raman microscopy of components such as epithelial cell cytoplasm, the cell nucleus, f at, β-carotene, collagen, calcium hydroxyapatite, calcium oxalate dihydrate, cholesterol-like lipid deposits, and water. Tissue composition extracted through modeling is used as the basis of a diagnostic algorithm capable of differentiating between a normal and diseased state. Although initial Raman studies e xamining inter nal body tissues required long collection times in the range of 5 to 30 s, recent advances in catheter design have resulted in fle xible Raman f iber-optic catheters capab le of collecting spectra with large signal-to-noise in 1 second or less. This has, for e xample, enab led the technolo gy’s application to in vivo spectral pathology of human atherosclerosis and vulnerable plaque.14 The optical fiber may be advanced through a catheter during carotid endarterectomy and femoral bypass surgeries to obtain Raman spectra of endothelial tissue. 15 Representative f it coefficients of the major components from a pre viously de veloped morphological model described b y Buschman and colleagues16,17 (including collagen, elastin, cholesterol cr ystals, necrotic core, calcifications, adventitial fat, smooth muscle cells, and β-carotene cr ystals) can therefore be generated in situ. The model has previously shown accurate tissue characterization (as confirmed by histology of the sur gical biopsies) and achie ved a sensiti vity and specificity of 79 and 85%, respecti vely. Similar devices have been used in diagnosing breast cancers b y dif ferentiating benign and malignant lesions based on chemical composition.18,19 In this application, the fit coefficients for fat and collagen are the k ey parameters in the diagnostic algorithm, which classifies tissue samples according to their specif ic patholo gical diagnoses. In patient studies, the spectroscopic technique attained 94% sensitivity and 96% specif icity for distinguishing cancer from normal and benign tissues. The technique has particularly shown its ef fectiveness for i n vivo margin assessment during partial mastectomy breast surgery.19 Both spectroscop y techniques, LSS and Raman, clearly of fer an enhanced ability o ver traditional endoscopy to glean molecular infor mation from tissue

149

and to detect very subtle pathologies in their early stages. Although not always, the creation of quasi real-time data processing algorithms can enab le in situ measurements through the working channel of an endoscope. Ho wever, these methods are still inherentl y point measurements and do not generate tr ue spatial images. The problem of grossly undersampling the v ast surface area of epithelial tissues remains unsolved, and the questions remain as to what advantage spectroscopy provides over the histology of surgical biopsies? What is needed in the clinical realm, therefore, is a spatial imaging technology that approaches the sensiti vity/specificity of spectroscop y and that can quickly survey entire organ surfaces. Fluorescence generated from either endo genous or e xogenous fluorophores in epithelial tissue of fers exactly these benef its and will be discussed in the remainder of this chapter .

IMAGING ENDOGENOUS AUTOFLUORESCENCE As described above, the phenomena of fluorescence is a (Stokes) shift in emitted w avelength gi ven a par ticular incident wavelength that excites an electron in a molecule to a stationary state. A variety of endogenous substances in biological tissue show this property at different excitation wavelengths collectively giving tissue, which is commonly known as autofluorescence. Although almost all tissue components e xhibit autofluorescence at some le vel, the signal obser ved during fluorescence endoscop y is predominantl y generated b y molecules in the mucosa such as collagen, elastin, tr yptophan, nicotinamide adenine dinucleotide, fla vin adenine dinucleotide, and por phrins.20 Such molecules ma y differentially accumulate in areas of dysplasia, leading to autofluorescent characteristics that ma y help distinguish between nor mal and neoplastic tissue. By placing standard band-pass f ilter w heels within the e xcitation and emission light paths, imaging can easil y switch between normal WL endoscopy and autofluorescence endoscop y, which highlights the signal from these endo genous fluorophores.21–23 The wavelength ranges for autofluorescence typicall y f all into the lo wer re gion of the visib le light spectrum emphasizing blue and green fluorophores. However, this is completel y dependent on the fluorophores of interest, and sometimes, the range may reach higher than 1000 nm. There have been numerous studies of standard autofluorescence endoscopy coupled with WL endoscopy, but the main applications in humans ha ve been for the detection of Bar rett’s esophagus 24 and colon neoplasia. 25 In addition, the paradigm has been extended to diseases in

150

MOLECULAR IMAGING: PRINCIPLES AND PRACTICE

the breast, lung, and other or gans. The majority of these studies ha ve sho wn that autofluorescence-guided endoscopy improves the diagnostic yield for neoplasia in comparison with the conventional approach using WL only and four-quadrant biopsies 20 (Figure 1A, B). Ho wever, it has also been e xtensively sho wn that autofluorescence alone is not suitab le for replacing the standard four -quadrant biopsy protocol 24 and that autofluorescence detection is associated with a f alse positive rate as high as 51%. As the techniques for autofluorescence imaging have been ref ined, a specif ic type of autofluorescence imaging dubbed narrow band imaging (NBI) has shown promise in the detection of Bar rett’s esophagus. 26 NBI enhances the visualization of superficial mucosal structures b y nar rowing the band-pass ranges of the g reen and b lue components of the e xcitation light. This causes the relative intensity of the b lue emission spectrum to increase and impro ves the visualization of mucosal blood vessels (since the blue light excitation is highly absorbed by hemoglobin).27 NBI emphasizes features such as capillar y and cr ypt patter ns, and this technique has potential for diagnosing gastrointestinal diseases at an early stage. The results of NBI in human studies are similar to that of other autofluorescence techniques. Although NBI has some what enhanced conventional endoscopic detection of disease, the added benef it is incremental at best, and fur ther trials are required to deter mine the true advantage compared with conventional endoscopy. One of the factors hindering the extraction of quantitative biochemical information from measured tissue autofluorescence is the presence of potentiall y signif icant distortions introduced b y tissue scattering and absor ption. Although a number of methods have been proposed for the recovery of the intrinsic (undistor ted) tissue fluorescence, the y are not easil y implemented in a clinical setting or they have limited applicability in the 400- to 500-nm spectrum because of high hemoglobin absorption levels. One potential solution has been to combine information in simultaneously measured tissue autofluorescence and diffuse reflectance.28 Such a technique can e xtract intrinsic (undistor ted) tissue autofluorescence and isolate and quantify the spectral contributions of N AD(P)H and collagen. This can pro ve useful because the relati ve contribution of these tw o fluorophores to the intrinsic tissue autofluorescence ma y be modif ied during the de velopment of pre-cancerous lesions in tissues such as Bar ratt’s esophagus and the uterine cer vix.29 Thus, the y may ser ve in some cases as in vi vo biomarkers of pre-malignant change, without the need for tissue remo val.

A

B B

C C

Figure 1. Clinical example of autofluorescence endoscopy. A 64-year-old man with atrophic gastritis underwent endoscopic treatment for early gastric cancer on the posterior wall of the upper gastric body seen as a whitish 1-cm elevated lesion in the white light channel (A). Autofluorescence imaging in the 490 to 625 nm light band (B) reveals the tumor as purple and additionally detects a flat tumor extension on the distal side that is not clear in the WL image. Chromoendoscopy with 0.04% indigo carmine solution (C) confirms the extent of the tumor. Reproduced with permission from Uedo N et al.23

Fiber Optic Fluorescence Imaging

FLUORESCENCE LIFETIME TECHNIQUES Many of the fluorescence endoscopy techniques, including the a utofluorescence a pproaches d escribed a bove, a re steady state methods that dif ferentiate tissue based upon differential emission spectral prof iles. The autofluorescence techniques discussed above often have a reasonable sensitivity for the detection of earl y cancers but a lo w specificity and a high f alse positive rate. 30 An alter native approach is fluorescence lifetime imaging (FLIM). 31 The method is based on the measurement of the temporal decay in fluorescence intensity following excitation. Because the fluorescence lifetime is derived from relative intensity values, FLIM can provide useful information concerning fluorochrome localization in spite of differences in scattering and variation in fluorophore concentration. Wide-field microscop y and FLIM ha ve already been used in studies of tissue constituents, 32 cell cultures,33 and the human skin.34 FLIM can be perfor med in the frequenc y domain for w hich a high-frequency modulated laser beam e xcites the sample and the fluorescence lifetime is deter mined from the demodulation and phase shift of the fluorescence signal. FLIM can also be performed in the time domain for which the fluorescence decay is directl y measured after pulsed laser excitation. FLIM is not only sensitive to the type of fluorophores but also may depend on its environment. This functionality has been e xploited35 to quantify ph ysiological parameters including pH, [Ca 2+], and pO 2. Differences in the fluorescence lifetimes betw een normal and neoplastic tissue ha ve been sho wn in the colon,36 breast,37 and brain. 38 However, the instr umentation required for these studies has been generall y costly, bulky, and very sensitive to small changes in tissue f ixation and optical calibration. The difficulties mostly arise from the very sophisticated lasers and detectors required to e xcite and sense photons with nanosecond temporal resolution, w hich ha ve not been practical options for a clinical instrument. Until now, there have been only a few reports of high-speed wide-f ield FLIM, and these ha ve usuall y been restricted to a reduced number of pix els.39 The imaging paradigm has just recently been applied to fiberoptic catheters such that FLIM de vices may be threaded through standard endoscopes and potentiall y image tissue in vivo with high frame rates and reasonab le spatial resolutions.40 The design relies on a standard gated optical image intensif ier with a rapidl y s witchable dela y generator, and much pre vious w ork has been done on implementing an anal ytic rapid lifetime deter mination algorithm. Although the research has so f ar onl y been applied to the autofluorescence of healthy tissues (human

151

stomach and kidne y), FLIM endoscop y has much potential in the future to generate fluorescence lifetime “maps” through a catheter at a video rate so that the technology can become more clinically relevant.

IMAGING UNTARGETED EXOGENOUS FLUOROPHORES Administration of exogenous fluorescent dyes with a high quantum yield can achie ve a much stronger e xtrinsic fluorescence contrast in epithelial tissue compared with most autofluorescence techniques. Such compounds ma y be delivered intravenously before examination or intravitally through the working channel of the endoscope during an imaging session. Chromoendoscopy is the practice of spra ying a fluorescent dye onto the mucosa using a spra y catheter passed through a standard endoscope (see F igure 1C). Although various dyes have been tried , methylene blue is the most common agent, and it is primaril y used in the detection of intraepithelial neoplasia and colon cancer. It is an inexpensive, absor ptive stain that, in contrast to other substances such as indigo carmine, is taken up by the intestinal epithelium after its local application, resulting in a relatively stable staining pattern and the visualization of the opening of the glandular pits during chromoendoscopy. It has been shown in large human studies6 that chromoendoscopy impro ves earl y diagnosis of adenomas and earl y colorectal cancers. It allo ws prediction of the nature of mucosal lesions in the colorectum b y using the so-called pit patter n classif ication41 for mucosal staining patter ns to dif ferentiate betw een neoplastic and nonneoplastic changes. A randomized , controlled trial42 has shown that chromoendoscopy permits more accurate diagnosis of the extent and severity of the inflammator y activity in ulcerati ve colitis compared with con ventional colonoscopy. Limited success has also been repor ted43 using the technique for the detection of dysplasia in Bar rett’s esophagus. Despite all these positive preliminary results, the technique has yet to be rigorously proven as an advantage during endoscopy. Because meth ylene b lue is inherentl y a reducing agent, some studies ha ve sho wn that it can induce oxidative damage of DNA when photosensitized by WL during chromoendoscop y and therefore can accelerate carcinogenesis.44 Such risks need to be carefully balanced against the possib le benef its of improved early disease detection. One e xogenous fluorophore that has already been heavily tested for adv erse reactions 45 is indoc yanine green (ICG). It has been used for more than 30 years as

152

MOLECULAR IMAGING: PRINCIPLES AND PRACTICE

a water-soluble dye with a peak absor ption at ~800 nm that rapidly binds to blood proteins (primarily albumin) after intravenous injection. Its pre vious clinical applications ha ve been for deter mining cardiac output, hepatic function, and ophthalmic angiography.46 Given its low reactivity, intravenous ICG has been e xploited for fluorescent endoscop y of v ascular str uctures. Its near infrared absorption and emission has enabled ICG to w ell delineate esophageal v arices, where structural changes in the vascular wall are caused by portal hypertension, and this method has recentl y been extended to the detection of vascular lesions in the digestive tract.47 In an observational study of 30 patients with gastric tumors,48 fluorescence was positive in 8 of 10 cancers with submucosal in vasion and in 1 of 20 adenomas or intramucosal cancers. It is believed that the retention of ICG is cor related with the size of the submucosal v ascular bed, and near -infrared (NIR) endoscop y of ICG can enhance the visibility of deeper v essels within a gastric tumor. ICG has also been used for the detection of metastases to a sentinel l ymph node (the f irst lymph node that receives drainage from a cancer). Intraoperati vely injecting dy e at the site of a melanoma to identify sentinel nodes is a well-characterized technique,49 and it has been recently been paired with NIR laparoscopic imaging of ICG in gastric cancer50 and in lung cancer patients. 51 The findings so f ar suppor t the ef ficiency of sentinel node navigation using ICG for detecting clinicall y nodenegative cancers. Several other agents function similarl y to ICG as fluorescent markers of vasculature during perfusion. Fluorescein (and its deri vative fluorescein isothiocyanate, [FITC]) has long been used in the realms of microscopy and angiographic ophthalmology. However, it is important to note that its peak e xcitation (494 nm) and emission (521 nm) wavelengths are well below that of ICG, and therefore signal from fluorescein is significantly absorbed in vivo by surrounding tissue. In addition, both fluorescein and ICG are relati vely small molecules that quickl y leak through the v asculature into interstitial space. Given the parameters of optical w avelength and molecular size, agents de veloped to image and quantitate tumor v asculature ha ve been slo wly optimized.52 One simple approach has been to attach organic dyes to long circulating de xtranated nanoparticles.53 Other methods use high molecular w eight (around 250 kDa) pe gylated g raft copol ymers with indocyanine-type fluorophores optimized for nonquenching54 (Figure 2A, B). Because these imaging

agents are nonimmuno genic (due to pe gylation and isotonia), longitudinal endoscopic imaging studies are also possible. A

B

C

D

E

F

G

H

II

J

Figure 2. Preclinical mouse endoscopies depicting white light (WL) (left column) and near-infrared (NIR) (right column) imaging channels. In vivo images from peritoneum vasculature (A,B) illustrate the ability to quantify vascular leak of a fluorescent blood pool agent by comparing intravascular with adjacent extravascular signal. Colonoscopy shows the detection of orthotopic tumor implantations (C,D) with a modified NIR fluorescence agent containing a cyclic RPMC peptide motif that was derived from a library screen. Significantly lower signal is observed in a control experiment (E,F) with a scrambled peptide sequence. Quantitative real-time colonoscopy shows the detection of two colon tumors barely discernable in the WL channel with a protease-activatable NIR probe (G,H). Fluorescent signal from the lesions is seen to remain constant (I,J) as the tip of the catheter is advanced closer to the tissue. Adapted from Sheth RA et al.84; Kelly K et al.61; and Upadhyay R et al.80

Fiber Optic Fluorescence Imaging

IMAGING TARGETED/ACTIVATABLE MOLECULAR PROBES As f iber-optic fluorescence de vices are maturing, a diverse array of sophisticated optical probes is being synthesized to complement the technology. These probes are exogenously delivered and generate v ery large signal-tonoise ratios by either targeting or interacting with biological processes on a molecular le vel. Most are completing preclinical validation in mouse models, and a fe w have entered human imaging studies along with con ventional endoscopes modified to detect fluorescence. One such probe originated as an agent for photodynamic therapy. Specifically, the photosensitizer pre-cursor 5-aminolevulinic acid (5-ALA) is intracellularly converted to protoporphyrin IX (PpIX) a few hours after intravenous administration. Con ventional photodynamic therap y proceeds with laser excitation of the photosensitizer, generates a singlet state o xygen molecule, destr uctively reacts with any nearby biomolecules, and it may result in apoptosis or necrosis.55 However, it has been disco vered that at lo w doses, PpIX can also ser ve as an imaging agent because it fluoresces red under b lue illumination. 56 Because 5-ALA elicits synthesis and preferential accumulation of PpIX in epithelial neoplastic tissue, it has strong potential as a targeted/activatable imaging agent in addition to its cur rent clinical role as a therapeutic. Low doses of 5-ALA are already appro ved for use in patients, which has enab led human studies to assess the performance of fluorescent endoscopy of PpIX, for example, for detecting intraepithelial neoplasia and earl y cancers in Bar rett’s esophagus. 57,58 Briefly, the fluorescence detection was seen to achieve a similar performance compared with four-quadrant random biopsy (the current gold standard), but it resulted in signif icantly fewer biopsies. This has motivated the development of next generation photodynamic agents that minimize extraneous phototoxicity and improve delivery and efficacy.59,60 Several other tar geted probes in the preclinical pipeline that are specif ically designed for endoscopic (and not necessarily therapeutic) applications ha ve been demonstrated in mouse models and imaged with prototype endoscopic systems. The probe designs in volve attaching a fluorochrome to a targeting moiety such as a peptide sequence. One probe w as engineered to be specific to a colon cancer b y deri ving an af finity ligand from a phage librar y screen. The result w as a c yclic peptide motif that could be fluorescentl y labeled 61 and imaged with fluorescence endoscop y. Such librar yderived imaging agents can be easil y v alidated in vi vo and compared with control molecules to sho w

153

cancer-specific tar geting (F igure 2C– F). This endoscopic imaging paradigm can be e xtended to an y other targeted optical probe developed for intraoperative applications such as fluorescentl y labeled antibodies, 62 small molecules,63 and nanoparticles.64 Rather than targeting surface receptors and proteins, targeted probes can also be incor porated into biolo gical vectors. This has been demonstrated, for instance, with a herpes simple x viral v ector with cancer -selective infection and replication including a transgene for green fluorescent protein. 65 Fluorescence-aided minimally invasive endoscopy showed microscopic tumor deposits unreco gnized b y con ventional laparoscop y/thoracoscopy. The imaging vector model was also in vitro confirmed in 110 types of cancer cells from 16 different primar y or gans, and the vector was shown to infect tumors and metastases in both immunocompetent and immunodef icient mice. Despite such preclinical results, similar fluorescent proteins encoded in biolo gical vectors are unlik ely to translate clinicall y due to concer ns re garding long-ter m immunogenicity. Finally, there has been e xtensive work in the de velopment of a ne w class of optical imaging agents that change their fluorescent proper ties after tar get interaction.66 These smar t probes are initiall y opticall y silent, secondary to fluorochrome–fluorochrome quenching (ie, when the 2 fluorochromes are in close pro ximity on a backbone molecule, the y absorb the other lights) but become brightly fluorescent in areas of disease. One specific target is the increased protease expression present in neoplastic tissue, w hich mediates enzymatic clea vage of fluorochromes from a delivery backbone, resulting in signal amplif ication of up to several hundredfold. This imaging paradigm has pro ven particularly successful with endoscopic applications because high signal-tonoise imaging of protease o verexpression allo ws rapid screening of a lar ge surf ace area (F igure 2G, H). The probe has been e xploited in colonoscopies, 67–70 thoracoscopies,71 and laparoscopies, 72 and it is no w poised to enter initial clinical trials. In addition, the same model for an activatable probe has been translated to other types of enzyme o verexpression for application in cardio vascular73 and autoimmune diseases, 74 where f iber-optic fluorescence imaging is also feasible.

PROBE DESIGN CONSTRAINTS AND QUANTIFICATION Given the previous discussion of untargeted, targeted, and activatable e xogenous molecular probes, w e can be gin

154

MOLECULAR IMAGING: PRINCIPLES AND PRACTICE

to outline general constraints and considerations in the design of optical molecular imaging agents for fiber-optic detection. Optimizing the strength of the fluorescence and the ability to quantify the signal detected by the catheter are par ticularly impor tant to in vi vo applications. Of primar y impor tance to the signal strength is the optical wavelength of the tar get.75 Most fluorescent proteins, FITC conjugations, rhodamine conjugations, and methylene blue emit in the visible-light range where surrounding tissue markedly absorbs the signal. The excitation light is similarl y attenuated. The attenuation is conversely decreased relative to shorter wavelengths for NIR probes such as conjugations to c yanine dyes, ICG, and other synthetic NIR fluorochromes. Furthermore, the autofluorescence of endogenous fluorophores in the surrounding tissue tends to be the strongest in the visiblelight range. This illustrates another adv antage in the signal-to-noise ratio of NIR probes, and it helps e xplain the limited success of pre vious autofluorescence endoscopy studies in humans. 24 Complementary to signal strength, the schedule and endurance of the peak signal are especiall y impor tant to serial f iber-optic imaging of the same tissue location. Enzyme activatable probes generally reach their peak fluorescence 24 to 48 hours after injection, and the signal may endure several more days. Most blood pool imaging agents are designed to remain in the vasculature a specific amount of time, primarily depending on the size of the molecule. 52 The use of multiple molecular imaging agents necessitates optically distinct wavelengths that are each discer nable by the fiber-optic instrumentation. For enzyme activatable probes, the absolute value of fluorescence obtained is a function of the intensity of the incident light from the endoscope and the depth and size of the lesion producing the fluorescence. In addition, the amount of acti vated product produced b y the protease reflects not onl y the enzyme acti vity but also the delivery of the imaging probe substrate. This complicates the quantif ication of the enzyme of interest because both play a role in the f inal signal intensity. To address this issue, dual fluorochrome probes ha ve been synthesized that separate the tw o processes. 76 The design is based upon standard nanopar ticles with clea vable peptide spacers attached through a C-ter minal c ysteine and a fluorochrome attached to the N ter minus. In addition to the activatable/quenched fluorochrome, a second fluorochrome is attached directl y to the macromolecule car rier and is resistant to proteol ytic acti vation. Endoscopic i maging o f t his p article a t t wo d istinct

wavelengths of fers both pieces of infor mation. Signal strength of the acti vatable fluorochrome represents probe delivery and activation, whereas signal strength of the reporter fluorochrome only correlates to delivery. Hence, the ratio of the tw o signals can be used to nor malize for differences in the size and depth of a target lesion and differences in probe delivery. Although the above probe optimizations are sufficient to quantify signal with f ixed geometry intraoperative fluorescence imaging systems,77 catheter-based systems face an e xtra complication. F iber-optic catheters do not ha ve the advantage of a static-controlled imaging en vironment in which distances from the illumination source to the target tissue and from the target tissue to the charge-coupled device (CCD) are fixed, and the illumination intensity across the tar get remains unifor m over time. These distances become dynamic, essentiall y uncontrollab le v ariables with catheter-based systems, and the fluorescence emission of a par ticular location is no longer constant. Photon fluence decreases as the square of the distance between the tar get tissue and the catheter tip increases, resulting in a marked change in NIR photon counts as one approaches or retreats from the disease being investigated. Moreover, sharp angles of incidence between the catheter and the NIR signal source cause objects closer to the catheter to appear brighter than more distant objects within the same video frame. The fluorescence microscop y community has resolved similar concerns regarding the quantitative ability of their instr uments by imaging a unifor mly fluorescent reference sample to estab lish a baseline sample image and then dividing all subsequent images by the reference sample image. 78,79 A comparab le algorithm can also be applied to real-time catheter -based systems. 80 Rather than dividing by a constant reference image, however, each NIR frame can undergo pixel-wise division by a simultaneously acquired WL image. This allows for a dynamic frame-b y-frame nor malization that accounts for v ariations in signal intensity betw een and within individual frames due to changes in catheter position (Figure 2G–J and Figure 3).

INSTRUMENTATION DESIGN CONSTRAINTS The addition of fluorescent molecular imaging to current clinical endoscope instr umentation poses some necessary design constraints. One fundamental prerequisite is the need for a dichroic mir ror (beam splitter) within the

Fiber Optic Fluorescence Imaging

155

B

A Concentration 1 µM 5 µM 10 µM 20 µM

Distance to Surface

Raw NIR Counts/msec

30 25 20 15 10 5

Constant ROI

0

C

1

2

3 4 Distance (mm)

5

D

0.3

40 % Change

Corrected NIR Signal

60

0.2

0.1

20

Corrected Data

0 20 40 Raw NIR Data

60 0

1

2

3 4 Distance (mm)

5

80

1

2

3 4 Distance (mm)

5

Figure 3. Distance dependence of raw near-infrared (NIR) pixel values versus white light (WL) division corrected NIR values. A, Schematic of the experimental design shows the various concentrations of fluorescent dye used in the phantoms. ROI = region of interest. B, Distance dependence curves for raw NIR counts per ms for the concentrations show a large percentage change over the distance of a few mm. Error bars denote the standard of error. C, Distance dependence curves for the corrected NIR pixels for the same concentrations show an approximately flat relationship between corrected signal intensity and distance. The difference in corrected NIR signal intensity between all four concentrations is significant (p < .001), as determined with analysis of covariance. D, Percentage change from the mean values for raw and corrected NIR. The reduced absolute slope of the corrected data (compared with the slope of the raw data) shows the reduced dependence of the signal intensity on distance. Reproduced with permission from Upadhyay R et al.80

endoscopic light path to enab le simultaneous WL and NIR imaging 72 (Figure 4A). This allows the operator to visualize NIR molecular signals and co-re gister them with a natomic l andmarks ( as s een i n c onventional endoscopy). Currently, dichroic mir rors can onl y be placed after fiber-optic catheters in the optical train, and they are not compatib le with cer tain commercial endoscopes containing CCD chips on the tip of the catheter . This constraint produces a slight loss in spatial resolution (depending on the substituted f iber bundle), but it also allows for a modular design of two or more cameras that independently acquire WL and NIR images. Dif ferent excitation light bands (such as a x enon lamp and a NIR laser) are also usually combined with a dichroic before en tering t he f iber-optic i llumination b undle. Interference band-pass f ilters can be appropriatel y placed throughout the optical path to avoid cross-talk between imaging channels. 81

When multiple distinct w avelength probes or fluorochromes ha ve to be imaged , tw o instr umentation designs are possib le. One ar rangement69 places an initial dichroic to transmit WL to a CCD and reflect all NIR signal to a second dichroic and a full y reflective mir ror that separates the light into tw o independent images (lower and higher emission ranges) onto different portions of a single CCD chip. Although this method can image two molecular tar gets, it reduces the NIR spatial resolution, complicates image processing, and does not scale well be yond tw o NIR w avelengths. A more direct approach simply adds independent cameras for each NIR channel desired. Thus, e xposure times can be independently controlled , and digital images can be perfectl y aligned b y standard se gmentation82 and re gistration83 algorithms. Such re gistration is par ticularly impor tant to the image di vision algorithms (discussed abo ve) that can enable quantification of NIR signal.

156

MOLECULAR IMAGING: PRINCIPLES AND PRACTICE

A

Dichroic Mirror Emission Filter

NIR CCD

12 Mb/sec

Fiberoptic Catheter Input Relay Lenses Visible Light CCD

Real-time Processing

12 Mb/sec

Initialize both cameras’ hardware

B

Generate ROI via segmentation routine Align cameras via registration routine Generate histogram for ROI

Camera 1

Camera 2

Set camera exposure time

Set camera exposure time

Start camera

Start camera

Wait for a single image to be transfered to RAM

Wait for a single image to be transfered to RAM

Display 8-bit image

Display 8-bit image

Calculate new exposure time

Calculate new exposure time

Save 12-bit data

Save 12-bit data

Stop camera

Stop camera

Calculate pixel value corresponding to 95% of histogram area Compare pixel value to setpoint and adjust exposure time accordingly

Real-time Visualization

Data Storage: Raw Images and Parameters

Figure 4. Schematic of fluorescent endoscopy instrumentation optics and software pipeline. Emission light from a fiber-optic catheter is optically separated by a dichroic mirror into WL and NIR light beams (A). These are focused onto CCD cameras with additional NIR filtering through a band-pass filter. The 12-bit data streams reach the computer via 100-Mbit connections. Incoming data are managed completely in random access memory to achieve real-time latency (B). The top arrow diverges into two parallel threads executed simultaneously but independently to allow for different frame rates for the two cameras. Left and right arrows loop for each camera image. Expanded bubble shows histogram-based calculations. ROI = region of interest. Reproduced with permission from Upadhyay R et al.80

Fiber Optic Fluorescence Imaging

In k eeping wi th s everal o ther c linical m odalities (including computed tomo graphy, magnetic resonance, and ultrasound), the CCD cameras in f iber-optic instrumentation that image the various wavelengths require a lar ge dynamic range (generally 12 bit or higher). The detection of very weak NIR signals is generall y a limiting factor in such imaging. The WL image stream in con ventional endoscop y is usuall y ne ver count star ved, and the frames can be acquired at video rate (roughl y 30 frames per second). As NIR CCDs are engineered to be more sensitive, minimum e xposure times to acquire practical signal-to-noise ratios can be set e ven f aster, and the NIR imaging frequency can also slo wly approach a tr ue video rate. Regardless, independent control of each CCD enables individual e xposure times to be set based on the cur rent signal strength in each imaging channel. Autoexposure algorithms in fixed geometry modalities, such as microscopes, can be based on either the brightest pix el v alue or the mid-range of an entire histogram of pixel values from a previous image. The latter is especially desirab le in f iber-optic applications because strong illumination from a point source onto a wet epithelial tissue often produces direct (specular) reflection that artificially saturates a fe w pixels. For example, selecting the pixel value that occurs at 95% of the histo gram area avoids artifacts from specular reflection but is still indicative of the o verall image brightness. The autoe xposure algorithm can g radually bring this pix el value to a user defined set point such that there is sufficient overall signal without any substantial pixel saturation. However, if the cameras operate independently and shutter variable exposure times, the raw pixel values of different images are no longer comparab le betw een cameras or successi ve images. One solution is to normalize each image’s pixel values by its cor responding exposure time because there is a linear relationship between the tw o variables. This ratio of counts/second can be used as a constant, e xposure time-independent data value. The new value is also required by an image division algorithm in w hich pixels within the WL and NIR images need to ha ve the same units before an y mathematical operation can be perfor med. In the case of typical 12-bit image acquisition, the raw pixel values can range from 0 to 4096 (2 12). However, the new counts/second v alue has additional fle xibility in exposure time: data can take any value from 0 to 4096 di vided b y the range of possib le e xposure times. For e xample, if there are 2 20 possible e xposure times (typical for modern cameras), this results in 2 32 possible values, correlating to an effective 32-bit dynamic range. Thus, the counts/second v alue can also dramaticall y

157

increase the dynamic range and sensiti vity of the digital imaging. After initial images ha ve been acquired in each channel, future exposure times have been determined, and the di vision algorithm has been perfor med, the operator needs to be ab le to visualize the cur rent frames before acquisition can proceed. Because all the frames are acquired with a high bit depth and video displays are limited to an 8-bit dynamic range, all images must under go proper g radation (windo wing and le veling) before the y can be displayed. Typically, NIR frames are rescaled using a pseudo-coloring algorithm that colorizes the images based on each pixel’s percent change from a user -defined variable baseline. The WL image must under go a demosaicing algorithm to inter polate a complete color image from the 12-bit ra w data recei ved from the color -filtered image sensor. Ultimately, both routines can produce color WL (as seen in con ventional imaging) and pseudo-color NIR images for immediate displa y, and the ra w data can be saved with the original dynamic range. Given the demands of acquisition, autoe xposure, window leveling, and other image processing algorithms that function in real time to maintain a tr ue video rate, modern fiber-optic fluorescence imaging systems require significant computing power. Software frameworks have been developed84 that rely on multithreaded code that can distribute the computational load across multiple processors and computers to achieve real-time signal processing (Figure 4B). The softw are design remains impor tant because the ability to process more data must scale as additional NIR channels (and CCDs) are introduced to simultaneously visualize multiple optical probes.

REFERENCES 1. Wolff WI. Colonoscop y: histor y and de velopment. Am J Gastroenterol 1989;84:1017–25. 2. Lux G, Kn yrim K, Scheubel R, Classen M. Electronic endoscop y— fibres or chips? Z Gastroenterol 1986;24:337–43. 3. Schoofs N , De viere J , Van Gossum A. PillCam colon capsule endoscopy compared with colonoscopy for colorectal tumor diagnosis: a prospective pilot study. Endoscopy 2006;38:971–77. 4. Barclay RL, Vicari JJ, Doughty AS, et al. Colonoscopic withdra wal times and adenoma detection during screening colonoscop y. N Engl J Med 2006;355:2533–41. 5. van Rijn JC, Reitsma JB , Stoker J, et al. P olyp miss rate deter mined by tandem colonoscopy: a systematic review. Am J Gastroenterol 2006;101:343–50. 6. Rembacken BJ, Fujii T, Cair ns A, et al. Flat and depressed colonic neoplasms: a prospective study of 1000 colonoscopies in the UK. Lancet 2000;355:1211–14. 7. Bernstein CN, Weinstein WM, Levine DS, Shanahan F. Physicians’ perceptions of dysplasia and approaches to sur veillance colonoscopy in ulcerati ve colitis. Am J Gastroenterol 1995; 90:2106–14.

158

MOLECULAR IMAGING: PRINCIPLES AND PRACTICE

8. Wang L V, Wu H-I. Biomedical optics: principles and imaging. Hoboken (NJ): Wiley-Interscience; 2007. 9. Hanlon EB, Manoharan R, Koo TW, et al. Prospects for in vivo Raman spectroscopy. Phys Med Biol 2000;45:R1–59. 10. Wallace MB, Perelman LT, Backman V, et al. Endoscopic detection of dysplasia in patients with Bar rett’s esophagus using lightscattering spectroscopy. Gastroenterology 2000;119:677–82. 11. Backman V, Wallace MB , P erelman L T, et al. Detection of pre-invasive cancer cells. Nature 2000;406:35–6. 12. Georgakoudi I, Jacobson BC, Van Dam J , et al. Fluorescence, reflectance, and light-scattering spectroscop y for e valuating dysplasia in patients with Bar rett’s esophagus. Gastroenterolo gy 2001;120:1620–29. 13. Motz JT, Gandhi SJ, Scepanovic OR, et al. Real-time Raman system for in vivo disease diagnosis. J Biomed Opt 2005;10:031113. 14. Romer TJ, Brennan JF 3rd , Fitzmaurice M, et al. Histopatholo gy of human coronary atherosclerosis by quantifying its chemical composition with Raman spectroscopy. Circulation 1998;97:878–85. 15. Motz JT , F itzmaurice M, Miller A, et al. in vi vo Raman spectral pathology of human atherosclerosis and vulnerab le plaque. J Biomed Opt 2006;11:021003. 16. Buschman HP , Deinum G, Motz JT , et al. Raman microspectroscopy o f h uman c oronary a therosclerosis: b iochemical assessment of cellular and extracellular morphologic structures in situ. Cardiovasc Pathol 2001;10:69–82. 17. Buschman HP, Motz JT, Deinum G, et al. Diagnosis of human coronary atherosclerosis b y mor phology-based Raman spectroscop y. Cardiovasc Pathol 2001;10:59–68. 18. Haka AS, Shafer-Peltier KE, Fitzmaurice M, et al. Diagnosing breast cancer b y using Raman spectroscop y. Proc Natl Acad Sci USA 2005;102:12371–76. 19. Haka AS, Volynskaya Z, Gardecki J A, et al. in vi vo margin assessment during partial mastectomy breast surgery using raman spectroscopy. Cancer Res 2006;66:3317–22. 20. Borovicka J, Fischer J, Neuweiler J, et al. Autofluorescence endoscopy in sur veillance of Bar rett’s esophagus: a multicenter randomized trial on diagnostic efficacy. Endoscopy 2006;38:867–72. 21. Haringsma J, Tytgat GN. The value of fluorescence techniques in gastrointestinal endoscop y: better than the endoscopist’ s e ye? I: the European experience. Endoscopy 1998;30:416–18. 22. Uedo N, Iishi H, Tatsuta M, et al. A novel videoendoscopy system by using autofluorescence and reflectance imaging for diagnosis of esophagogastric cancers. Gastrointest Endosc 2005;62:521–28. 23. Uedo N, Iishi H, Ishihara R, et al. Novel autofluorescence videoendoscopy imaging system for diagnosis of cancers in the digestive tract. Digestive Endoscopy 2006;18:S131–6. 24. Curvers WL, Singh R, Song LM, et al. Endoscopic tri-modal imaging for detection of early neoplasia in Barrett’s oesophagus: a multicentre feasibility study using high-resolution endoscop y, autofluorescence imaging and nar row band imaging incor porated in one endoscopy system. Gut 2008;57:167–72. 25. Izuishi K, Tajiri H, Fujii T, et al. The histological basis of detection of adenoma and cancer in the colon b y autofluorescence endoscopic imaging. Endoscopy 1999;31:511–16. 26. Gono K, Obi T, Yamaguchi M, et al. Appearance of enhanced tissue features in narrow-band endoscopic imaging. J Biomed Opt 2004; 9:568–77. 27. Yoshida T, Inoue H, Usui S, et al. Nar row-band imaging system with magnifying endoscop y for superf icial esophageal lesions. Gastrointest Endosc 2004;59:288–95. 28. Georgakoudi I, Jacobson BC, Müller MG, et al. NAD(P)H and collagen as in vi vo quantitati ve fluorescent biomark ers of epithelial pre-cancerous changes. Cancer Res 2002;62:682–87. 29. Drezek R, Sokolov K, Utzinger U, et al. Understanding the contributions of N ADH and collagen to cer vical tissue fluorescence spectra: modeling, measurements, and implications. J Biomed Opt 2001;6:385–96.

30. Bard MP, Amelink A, Skurichina M, et al. Improving the specificity of fluorescence bronchoscopy for the analysis of neoplastic lesions of the bronchial tree b y combination with optical spectroscop y: preliminary communication. Lung cancer 2005;47:41–7. 31. Siegel J, Elson DS, Webb SE, et al. Studying biological tissue with fluorescence lifetime imaging: microscopy, endoscopy, and complex decay profiles. Appl Opt 2003;42:2995–3004. 32. Dowling K, Dayel MJ, Lever MJ, et al. Fluorescence lifetime imaging with picosecond resolution for biomedical applications. Opt Lett 1998;23:810–12. 33. Bastiaens PI, Squire A. Fluorescence lifetime imaging microscop y: spatial resolution of biochemical processes in the cell. Trends Cell Biol 1999;9:48–52. 34. Hartmann P, Mirtolouei R, Untersberger S, et al. Non-invasive imaging of tissue PO2 in malignant melanoma of the skin. Melanoma Res 2006;16:479–86. 35. Lakowicz JR. Principles of fluorescence spectroscop y. Ne w York: Springer; 2006. 36. Mycek MA, Schomacker KT, Nishioka NS. Colonic polyp differentiation using time-resolv ed autofluorescence spectroscop y. Gastrointest Endosc 1998;48:390–94. 37. Tadrous PJ, Siegel J, French PM, et al. Fluorescence lifetime imaging of unstained tissues: early results in human breast cancer. J Pathol 2003;199:309–17. 38. Marcu L, Jo JA, Butte PV, et al. Fluorescence lifetime spectroscopy of glioblastoma multiforme. Photochem Photobiol 2004;80:98–103. 39. Agronskaia AV, Tertoolen L, Gerritsen HC. Fast fluorescence lifetime imaging of calcium in living cells. J Biomed Opt 2004;9:1230–7. 40. Munro I, McGinty J, Galletly N, et al. Toward the clinical application of time-domain fluorescence lifetime imaging. J Biomed Opt 2005;10:051403. 41. Kudo S, Tamura S, Nakajima T, et al. Diagnosis of colorectal tumorous lesions b y magnifying endoscop y. Gastrointest Endosc 1996;44:8–14. 42. Kiesslich R, F ritsch J, Holtmann M, et al. Meth ylene blue-aided chromoendoscopy for the detection of intraepithelial neoplasia and colon cancer in ulcerative colitis. Gastroen terology 2003;124:880–8. 43. Sharma P, Weston AP, Topalovski M, et al. Magnification chromoendoscopy for the detection of intestinal metaplasia and dysplasia in Barrett’s oesophagus. Gut 2003;52:24–7. 44. Olliver JR, Wild CP, Sahay P, et al. Chromoendoscopy with methylene blue and associated DNA damage in Barrett’s oesophagus. Lancet 2003;362:373–4. 45. Hope-Ross M, Yannuzzi LA, Gragoudas ES, et al. Adverse reactions due to indocyanine green. Ophthalmology 1994;101:529–33. 46. Flower RW, Hochheimer BF. Indocyanine green dye fluorescence and infrared absor ption choroidal angio graphy perfor med simultaneously with fluorescein angio graphy. Johns Hopkins Med J 1976; 138:33–42. 47. Okamoto K, Mugur uma N , Kimura T, et al. A no vel diagnostic method for e valuation of v ascular lesions in the digesti ve tract using infrared fluorescence endoscopy. Endoscopy 2005;37:52–7. 48. Kimura T, Muguruma N, Ito S, et al. Infrared fluorescence endoscopy for the diagnosis of superf icial gastric tumors. Gastrointest Endosc 2007;66:37–43. 49. Morton DL, Wen DR, Wong JH, et al. Technical details of intraoperative l ymphatic mapping for earl y stage melanoma. Arch Sur g 1992;127:392–9. 50. Ishikawa K, Yasuda K, Shiromizu A, et al. Laparoscopic sentinel node navigation achieved by infrared ray electronic endoscopy system in patients with gastric cancer. Surg Endosc 2007;21:1131–34. 51. Ito N, Fukuta M, Tokushima T, et al. Sentinel node navigation surgery using indocyanine green in patients with lung cancer. Surg Today 2004;34:581–85. 52. Weissleder R, Bo gdanov A Jr , Tung CH, Weinmann HJ. Size optimization of synthetic g raft copolymers for in vi vo angiogenesis imaging. Bioconjug Chem 2001;12:213–9.

Fiber Optic Fluorescence Imaging

53. Montet X, Ntziachristos V, Grimm J , Weissleder R. Tomographic fluorescence mapping of tumor tar gets. Cancer Res 2005; 65:6330–6. 54. Montet X, Figueiredo JL, Alencar H, et al. Tomographic fluorescence imaging of tumor v ascular v olume in mice. Radiolo gy 2007; 242:751–8. 55. Peng Q, Ber g K, Moan J , et al. 5-Aminole vulinic acid-based photodynamic therap y: principles and e xperimental research. Photochem Photobiol 1997;65:235–51. 56. Stepp H, Sroka R, Baumgar tner R. Fluorescence endoscop y of gastrointestinal diseases: basic principles, techniques, and clinical experience. Endoscopy 1998;30:379–86. 57. Stepinac T, Felley C, Jornod P, et al. Endoscopic fluorescence detection of intraepithelial neoplasia in Bar rett’s esophagus after oral administration of aminolevulinic acid. Endoscopy 2003;35:663–68. 58. Endlicher E, Knuechel R, Hauser T, et al. Endoscopic fluorescence detection of low and high g rade dysplasia in Bar rett’s oesophagus using systemic or local 5-aminolae vulinic acid sensitisation. Gut 2001;48:314–19. 59. McCarthy JR, P erez JM, Br uckner C, Weissleder, R. P olymeric nanoparticle preparation that eradicates tumors. Nano Lett 2005;5:2552–6. 60. McCarthy JR, Weissleder R. Model systems for fluorescence and singlet o xygen quenching b y metallopor phyrins. ChemMedChem 2007;2:360–5. 61. Kelly K, Alencar H, Funo vics M, et al. Detection of in vasive colon cancer using a novel, targeted, library-derived fluorescent peptide. Cancer Res 2004;64:6247–51. 62. Gee MS, Upadhyay R, Bergquist H, et al. Multiparameter noninvasive assessment of treatment susceptibility , drug target inhibition and tumor response guides cancer treatment. Int J Cancer 2007; 121:2492–500. 63. Weissleder R, K elly K, Sun EY , et al. Cell-specif ic tar geting of nanoparticles by multivalent attachment of small molecules. Nat Biotechnol 2005;23:1418–23. 64. Sun EY, Josephson L, Weissleder R. “Clickable” nanoparticles for targeted imaging. Mol Imaging 2006;5:122–8. 65. Adusumilli PS, Stiles BM, Chan MK, et al. Real-time diagnostic imaging of tumors and metastases b y use of a replication-competent herpes vector to facilitate minimally invasive oncological surgery. Faseb J 2006;20:726–28. 66. Weissleder R, Tung CH, Mahmood U, Bogdanov A Jr, in vivo imaging of tumors with protease-acti vated near -infrared fluorescent probes. Nat Biotechnol 1999;17:375–8. 67. Alencar H, Funovics MA, Figueiredo J, et al. Colonic adenocarcinomas: near -infrared microcatheter imaging of smar t probes for early detection—study in mice. Radiology 2007;244:232–8.

159

68. Alencar H, King R, Funo vics M, et al. A no vel mouse model for segmental orthotopic colon cancer. Int J Cancer 2005;117:335–9. 69. Funovics MA, Alencar H, Montet X, et al. Simultaneous fluorescence imaging of protease e xpression and v ascularity during murine colonoscopy for colonic lesion characterization. Gastrointest Endosc 2006;64:589–97. 70. Funovics MA, Alencar H, Su HS, et al. Miniaturized multichannel near infrared endoscope for mouse imaging. Mol Imaging 2003; 2:350–7. 71. Figueiredo JL, Alencar H, Weissleder R, Mahmood U. Near infrared thoracoscopy of tumoral protease acti vity for improved detection of peripheral lung cancer. Int J Cancer 2006;118:2672–7. 72. Funovics MA, Weissleder R, Mahmood U . Catheter -based in vi vo imaging of enzyme activity and gene expression: feasibility study in mice. Radiology 2004;231:659–66. 73. Chen J, Tung CH, Allport JR, et al. Near-infrared fluorescent imaging of matrix metalloproteinase acti vity after m yocardial inf arction. Circulation 2005;111:1800–05. 74. Izmailova ES, P az N, Alencar H, et al. Use of molecular imaging to quantify response to IKK-2 inhibitor treatment in murine arthritis. Arthritis Rheum 2007;56:117–28. 75. Weissleder R, Ntziachristos V. Shedding light onto li ve molecular targets. Nat Med 2003;9:123–28. 76. Kircher MF, Weissleder R, Josephson L. A dual fluorochrome probe for imaging proteases. Bioconjug Chem 2004;15:242–48. 77. De Grand AM, F rangioni JV . An operational near -infrared fluorescence imaging system prototype for large animal surgery. Technol Cancer Res Treat 2003;2:553–62. 78. Model MA, Burkhardt JK. A standard for calibration and shading correction of a fluorescence microscope. Cytometry 2001;44:309–16. 79. Zwier JM, Van Rooij GJ , Hofstraat JW, Brak enhoff GJ. Image calibration in fluorescence microscopy. J Microsc 2004;216:15–24. 80. Upadhyay R, Sheth RA, Weissleder R, Mahmood U . Quantitati ve real-time catheter-based fluorescence molecular imaging in mice. Radiology 2007;245:523–31. 81. Mahmood U, Tung CH, Tang Y, Weissleder R. F easibility of in vi vo multichannel optical imaging of gene e xpression: e xperimental study in mice. Radiology 2002;224:446–51. 82. Yoo TS, Ackerman MJ, Lorensen WE, et al. Engineering and algorithm design for an image processing Api: a technical repor t on ITK— the Insight Toolkit. Stud Health Technol Inform 2002;85:586–92. 83. Woods RP, Grafton ST, Holmes CJ, et al. Automated image registration: I. General methods and intrasubject, intramodality v alidation. J Comput Assist Tomogr 1998;22:139–52. 84. Sheth RA, Upadhyay R, Weissleder R, Mahmood U. Real-time multichannel imaging frame work for endoscop y, catheters, and f ixed geometry intraoperative systems. Mol Imaging 2007;6:147–55.

11 FLUORESCENCE TOMOGRAPHY VASILIS NTZIACHRISTOS, MSC, PHD

From cell labeling to immunohistochemistr y and micro-array studies, fluorescence has been one of the most common sources of contrast in the biomedical laboratory. The development of fluorescence probes and of fluorescence proteins as reporter molecules for subcellular function has further solidified the versatility and wide application of fluorescence imaging in a v ariety of applications, for e xample, the study of protein function and interactions, gene transcription, and visualizing molecular pathways or cellular trafficking.1,2 In parallel, the associated photon sensing and imaging technolo gy has progressed over the y ears, to allo w visualization at many dif ferent scales, spanning from the nanometer scale to clinical imaging. 3,4 Fluorescence microscopy has been one of the common methods of fluorescence imaging of cellular mono-la yer assays or thin tissue sections, such as histolo gical slides. The de velopment of adv anced microscop y methods, for example, confocal and multiphoton microscopy, has further allowed imaging of thick er tissues. 5,6 These methods can effectively m inimize t he e ffect o f s cattering, o ffering images of high resolution compared with con ventional microscopy images. By sequentiall y imaging at dif ferent depths, three-dimensional images can be also generated. Depending on the tissue’s optical properties and the particular microscopic implementations, depths of up to 500 to 800 microns can be reached , especially when using multiphoton microscopy. These methods are reviewed in Chapter 13, “Intravital Microscopy.” To image deeper in tissue, it becomes essential to use tomographic techniques operating in meso and macroscopic scale. Tomography generally refers to the ability to obtain cross-sectional images from intact tissues and animal or human bodies, thus of fering a three-dimensional picture of fluorescence bio-distribution. In contrast to microscopic three-dimensional tissue-sectioning imaging, tomography and reconstruction imply the formulation of a 160

mathematical inverse problem, whose solution yields the tomographic images,7 in analogy to methods used in x-ray computed tomography (CT), single photon emission computed tomography (SPECT) or positron emission tomography (PET). Major technological approaches have been developed to achieve fluorescence tomography beyond the limits of intra vital microscopy, including the use of theoretical models of photon propagation in tissue.As a general rule, microscopy retains higher resolution and sensiti vity compared with tomographic methods developed for imaging deeper in tissue. Con versely, tomo graphy of fers the ability to offer noninvasive imaging of larger tissues allowing for in vivo imaging of insects, f ish, small animals, and even humans. The follo wing parag raphs detail the principles of in vivo fluorescence tomo graphy be yond the intra vital microscopy limit and showcase imaging applications.

TECHNOLOGY FOR FLUORESCENCE TOMOGRAPHY A common denominator of each tomography method is the ability to transmit a for m of ener gy to the sample of interest and detect changes in this transmitted energy due to its interaction with the sample from multiple depths or at dif ferent projections. The measurement of these changes are then digitized and used to gether with an inversion mathematic model to for m images. In the case of fluorescence, the transmitted energy is light that can be absorbed by the fluorochrome. Different substances can absorb light of different energy and emit fluorescence light at a lo wer energy. For imaging within only a few hundred microns to a few millimeters, light in the visib le can be used to e xcite common fluorescent proteins, such as the g reen fluorescent protein and its variants, and other spectrally shifted fluorescent proteins,

Fluorescence Tomography

such as the y ellow fluorescent protein, the red fluorescent protein, etc. In addition, it can be used to excite common or ganic fluorochromes, such as fluorescein, Texas Red, etc, or quantum dots. However, for penetration beyond a fe w millimeters, it becomes essential to use light in the near -infrared (NIR), that is, light of wavelength higher than 650 nm. This is because the light attenuation in tissue is signif icantly lo wer in the NIR, compared to the visible, due to the characteristic absorption spectr um of o xy- and deo xy-hemoglobin, w hich absorbs light mainly in the visible and not in the NIR. 8,9 Therefore, NIR light can penetrate much deeper in tissue compared with visib le light, up to se veral centimeters. One of the major challenges for tomo graphic imaging of tissues therefore is not the tissue absor ption but the tissue scattering. To achieve multiprojection illumination, the de velopment of appropriate imaging setups is required , which can direct a beam of light to the tissue surface. Figure 1 depicts a typical mode of imaging, that is, transillumination tomography w here the e xcitation light is projected from one side of the object and light that has propagated inside tissue is collected on the other side. This is a preferred mode of operation as it attains the most relaxed dynamic range requirements for the optoelectronic technology used while achieving vir tually symmetric v olume illumination coverage.

161

One of the impor tant aspects for fluorescence tomography is the use of accurate theoretical models that describe photon propagation in tissues. These models typically use some appro ximate solution to the transpor t equation.7 Tomographic de velopment generall y needs first to v alidate a suggested theoretical model against experimental measurements before using it in subsequent applications. The most common models used in optical tomographic imaging use solutions to the diffusion equation.10–13 Typically, these solutions w ork well for highl y scattering media of propagation dimension lar ger than 1 cm. F or smaller dimensions or moderatel y scattering media, more advanced models have been proposed. 14 The appearance of a typical optical tomo graphy problem takes the following form: N

φsc (r, rs) = ∑ W(r, rn, rs)O(rn), n=1

(1)

where W(r, rn, rs) represents a “w eight” that associates the effect of the optical proper ty O(rn) at position rn to a measurement at r owing to a source at rs. For a number of measurements, M, a system of linear equations is then obtained, resulting in a matrix equation: y = Wx,

(2)

where W is the weight matrix, x represents the distribution O(rn) of optical proper ty in each of the N voxels

Figure 1. Typical optical tomography arrangement in transillumination mode. Excitation light illuminates the object on one side, whereas light is collected on the opposite side of the object.

162

MOLECULAR IMAGING: PRINCIPLES AND PRACTICE

assumed, and y is the cor responding measurement vector. Inversion of this system of equations yields the unknown image x. Depending on the e xact for mulation of the problem, x can represent a two-dimensional or three-dimensional image. Ho wever, because photon propagation is fundamentally a volumetric phenomenon, x is typically a three-dimensional image. Computation times in optical tomography scale with the number of source-detector pairs used in the measurements. The use of CCD cameras readily offer millions of such pairs that need f irst to be processed (g rouped together) to reduce the size of v ector y. Even then inversion times may reach several minutes to hours, depending on the exact tomographic implementation, although typical commercial systems ha ve reconstruction times in the minute range.

MODES OF ILLUMINATION OPERATION There are three major modes of illumination that are typically considered in fluorescence tomo graphy15: continuous w ave (CW), time domain (TD), and frequenc y domain (FD) illumination. These categories relate to the technology used in generating light and result in different photon propagation characteristics. Constant wave methods use light of constant intensity and typically use CCD cameras for light detection. This approach requires low-cost implementations and offers high detection sensitivity and signal-to-noise ratio. They are well suited for fluorescent tomography although they cannot distinguish betw een absor ption and scattering, and are not prefer red for measurements of endo genous contrast (ie, absor ption or scattering). Compared with more adv anced illumination settings, CW methods have the limitation that they cannot minimize the effect of tissue scattering, that is, the resolution achieved is limited by the tissue optical proper ties. In addition, the y are not suitable for measuring fluorescence lifetime. Time domain methods use nar row photon pulses (typically < 10 ps). The detection in this case is achie ved with time-gated or time-resolv ed techniques, w hich record the ar rival of photons as a function of time, with time resolution on the order of picoseconds to hundreds of picoseconds. Time domain systems have the capability to independentl y disentangle the scattering and absor ption coefficients and to image fluorescence lifetime. Furthermore, b y using time-gating techniques, highl y scattered photons can be rejected , yielding images of improved resolution compared with CW systems. Compared with the CW method , TD systems yield lo wer

signal-to-noise ratio due to the low-duty cycles used due to pulsing, require more adv anced optical designs for optimal operation, and cost more. Finally, the frequenc y domain method uses light of modulated intensity and correspondingly uses demodulation methods to measure the amplitude and phase of the photon wave that is established into the object image. Modulation intensities span the 100 to 1,000 MHz range, and corresponding systems can operate at a single or multiple frequencies. Similar to the TD method, FD systems can measure lifetime, and by using high modulation frequencies, they can impro ve the image resolution over CW systems. A further advantage of the FD method is the ability to reject ambient, nonmodulated light. Practical implementation of this method is more challenging compared with the CW method due to the need of incor porating high frequenc y response optoelectronics and corresponding demodulation techniques. Again, these methods are more costly than CW.

TOMOGRAPHIC APPLICATIONS Fluorescence tomo graphy has been applied at dif ferent resolution scales, from the penetration limits of multiphoton microscopy (~500 microns) to human imaging. A particular class of fluorescence tomo graphy of tissues w as developed specif ically for molecular inter rogations of tissue in vi vo, ter med fluorescence molecular tomo graphy (FMT). Fluorescence molecular tomo graphy combines measurements at both emission and e xcitation wavelengths to quantify and to reconstruct fluorochromes of high molecular specif icity three-dimensionall y.15 The technique is used in conjunction with the systemic administration of fluorescence probes tar geting specif ic enzymes or other proteins. An associated technique, fluorescence protein tomo graphy (FPT) 16 is a tomo graphic method adapted to imaging fluorescence proteins by using appropriate spectral decomposition methods for reducing tissue autofluorescence ef fects w hen strong autofluorescence contributions are present. Figure 2 sho ws the application of FPT , operating in CW mode using an Ar+ laser to image mor phogenetic movements occur ring within the Drosophila pupae, in a study by Vinegoni and colleagues.17 In this case, the morphogenesis of the GFP-e xpressing wing discs and thorax was followed during the first few hours after preparation. The depicted time-lapse fluorescence imaging sequence shows the mor phogenesis of the wing imaginal discs and is in good ag reement with the cor responding histolo gic sections. In this case, a homemade system, similar to that

Fluorescence Tomography

A

B

C

D

E

F

163

Figure 2. Time-lapse imaging of Drosophila wing imaginal discs. The images are acquired from a single live specimen, at six different time points. Three different projections at 0, 50, and 180° with respect to the pupa’s dorsal view are shown. The reconstructions in the first column correspond to the sections indicated by the red lines in the third, fourth, and fifth columns. Comparison with DAPI-stained histological sections is shown in the second column. The histological images were acquired from morphologically matched areas of different staged pupae.

Figure 3. Imaging of lung inflammation. A, D, Photographs of a challenged and a control animal respectively. B, E, Fluorescence images (in color) for the experimental and control animals superimposed on the photographs of (A) and (D). C, F, H&E staining of lung sections obtained from the experimental and control animal 24 h after challenge. Inflammatory response of the challenged animal is seen as diffuse alveolar wall injury with noted thickening. Adatpted from Haller J et al.20 See also Chapter 67.

of Figure 1, was used. Laser light was beam expanded and directed t hrough a l ow n umerical a perture o bjective (Olympus, PlanN 10 × /0.25) to one side of the live pupa. Light propagation through the pupa body was captured by a CCD camera using a Leica Z16 APO apochromatically corrected zoom lens, using appropriate nar row band-pass interference filters (ex: 488 ± 5 nm, em: 513 ± 5 nm; the narrow width reducing autofluorescence ef fects). To improve image quality and reduce photon dif fusion, a polarization anal yzer w as placed in front of the microscope, oriented in parallel to the incident polarization light so as to reject highl y scattered photons, that is, photons that lose their polarization state. Theoretical modeling in this case w as based on the Fermi appro ximation inte grated over the ph ysical area seen by each pixel on the CCD camera used. To account for refractive index mismatch between inner volume of the pupae and the sur rounding air, the Fermi-based forward model w as fur ther cor rected b y calculating the correct angle of propagation in the medium using the Snell’s law of optical refraction. In version in this case was based on a back-projection algorithm. Fluorescence tomography of larger animals has been also shown, and imaging systems are now commercially available. Se veral other studies elucidate on specif ic

disease applications of this technolo gy.18,19,15 One specific example is shown in Figure 3, which depicts the results from imaging of inflamed mouse lungs, after administration of a cathepsin-acti vatable fluorescence probe (Prosense-680). In this case, imaging was also based on transillumination scanning using a limited angle projection scanner . The major dif ference of this approach, compared to the system shown on Figure 1, is that a focused laser beam is scanned on the animal surface when the animal is stationary, instead of the animal being rotated. Photographs of an animal challenged with intranasal instillation of O VA and a control animal treated with saline instillation only are shown in Figures 3A and 3D , respectively; see also Chapter 67 for more in-depth information on pulmonar y FMT imaging. Correspondingly, F igures 3B and 3E sho w fluorescence images obtained through the e xperimental and control animals, depicting a mark ed fluorescence increase from the experimental animal, congruent with the lung. Histologic analysis conf irmed parameters of an acute inflammatory response in the challenged lung (F igure 3C), featuring re gions of alv eolar w all thick ening and collapse with accompan ying edema and inflammator y cell penetration (obser ved at higher magnif ications; not shown), relative to that of healthy lung tissue (Figure 3F).

164

MOLECULAR IMAGING: PRINCIPLES AND PRACTICE

TOMOGRAPHIC CHARACTERISTICS Description of the de velopment of li ving organisms and of disease has lar gely relied on histolo gic sections or on other in vitro laboratory tests, allowing for measurements obtained from euthanized or e xcised tissues. Histolo gic analysis is challenging when monitoring quickly evolving phenomena or responses over time, for example, in monitoring disease e volution or dr ug ef fects. Although the GFP was imaged in the abo ve e xample, the technolo gy can be more generally applied to imaging red-shifted proteins or NIR fluorescence probes, in lar ger penetration scales. For example, it w as shown that the tomo graphic imaging of indoc yanine g reen accumulation is possib le through the human breast or small animals in vivo. Therefore, fluorescence tomo graphy becomes a cr ucial technology for studying the bio-distribution of new classes of fluorescent molecular probes or fluorescent p roteins i n l iving s ystems n oninvasively. Generally, a linear relation of reconstructed fluorochrome concentration and targeted molecule exist in biologically relevant concentrations, w hich allow for tr ue quantif ication, assuming that the nonlinearity associated with photon propagation in tissues is accounted for b y the appropriate theoretical model used.

REFERENCES 1. Giepmans BNG, Adams SR, Ellisman MH, Tsien RY. Review—the fluorescent toolbo x for assessing protein location and function. Science 2006;312:217–24. 2. Weissleder R, Ntziachristos V. Shedding light onto live molecular targets. Nat Med 2003;9:123–8. 3. Hell SW . Toward fluorescence nanoscop y. Nat Biotechnol 2003; 21:1347–55.

4. Ntziachristos V, Yodh AG, Schnall M, Chance B. Concurrent MRI and diffuse optical tomo graphy of breast after indoc yanine g reen enhancement. Proc Natl Acad Sci USA 2000;97:2767–72. 5. Jain RK. Nor malization of tumor v asculature: an emerging concept in antiangiogenic therapy. Science 2005;307:58–62. 6. Zipfel WR, Williams RM, Webb WW. Nonlinear magic: multiphoton microscopy in the biosciences. Nat Biotechnol 2003;21:1368–76. 7. Arridge SR. Optical tomography in medical imaging. Inverse Probl 1999;15:R41–R93. 8. Jobsis FF. Noninvasive, infrared monitoring of cerebral and m yocardial oxygen sufficiency and circulatory parameters. Science 1977; 198:1264–7. 9. Chance B. Optical method. Annu Rev Biophys Biophys Chem 1991; 20:1–28. 10. Patterson MS, Chance B , Wilson BC. Time resolv ed reflectance and transmittance for the nonin vasive measurement of tissue opticalproperties. Appl Opt 1989;28:2331–36. 11. Arridge SR, Hebden JC. Optical imaging in medicine II: modelling and reconstruction. Phys Med Biol 1997;42:841–53. 12. Yodh AG, Chance B. Spectroscopy and imaging with diffusing light. Phys Today 1995;48:34–40. 13. Chang JH, Graber HL, Barbour RL. Imaging of fluorescence in highly scattering media. IEEE Trans Biomed Eng 1997;44:810–22. 14. Klose AD, Netz U , Beuthan J , Hielscher AH. Optical tomo graphy using the time-independent equation of radiati ve transfer—part 1: forward model. J Quant Spectrosc RadiatTransf 2002;72:691–713. 15. Ntziachristos V, Ripoll J, Wang LHV, Weissleder R. Looking and listening to light: the e volution of w hole-body photonic imaging. Nat Biotechnol 2005;23:313–20. 16. Zacharakis G, Kambara H, Shih H, et al. Volumetric tomography of fluorescent proteins through small animals in-vi vo. Proc Natl Acad Sci USA 2005;102:18252–57. 17. Vinegoni C, Pitsouli C, Razansk y D , et al. In vi vo imaging of Drosophila melano gaster pupae with mesoscopic fluorescence tomography. Nat Methods 2007. [10.1038/ nmeth1149]. 18. Patwardhan SV, Bloch SR, Achilefu S, Culver JP. Time-dependent whole-body fluorescence tomo graphy of probe bio-distributions in mice. Opt Express 2005;13:2564–77. 19. Ntziachristos V, Tung C, Bremer C, Weissleder R. Fluorescence-mediated tomography resolves protease activity in vivo. Nat Med 2002; 8:757–60. 20. Haller J, De Kleine R, Hyde D, Ntziachristos V. Fluorescence tomography of inflammatory responses in the lung. J Appl Physiol 2007. [In press].

12 ENDOMICROSCOPY SEOK H. (ANDY) YUN, PHD AND CHARLES P. LIN, PHD

INTRODUCTION The aim of endoscopic microscop y is to enab le highresolution imaging of internal organs or tissue compartments at the cellular le vel through an optical probe that is introduced into the body in a minimall y in vasive manner.1 Endoscopic microscopy bridges an impor tant gap separating, on the one hand, traditional whole-body imaging modalities, such as magnetic resonance imaging and positron emission tomo graphy that lack the sensitivity and resolution (both spatial and temporal) to visualize single-cell dynamics in the body , and on the other hand, high-resolution optical imaging techniques, such as confocal and multiphoton microscopy, that lack the ability to image deep into tissue. By replacing the objective lens and other bulk y components of the standard microscope with a small-diameter probe, the ability of optical microscop y can be e xtended be yond easily accessib le surf aces, such as the skin or the cornea. F or small-animal molecular imaging research, cellular processes in inter nal or gans, including the brain, the gastrointestinal (GI) tract, the spleen, and the lymph nodes, can be visualized with much less tissue manipulation compared with traditional intra vital microscopy approaches.

TYPES OF ENDOMICROSCOPES Endomicroscopes Based on Fiber-Optic Bundles The most common type of endoscope uses a fle xible fiber-optic bundle to rela y images from the distal end of the bundle to the pro ximal end, where the image can be viewed b y e ye or captured b y a char ge-coupled de vice (CCD) or similar electronic cameras. The f ibers are arranged in a coherent manner such that the position of

each fiber within the bundle is maintained throughout the length of the bundle. In this w ay, an image is transmitted from one end of the bundle to the other end without getting scramb led. Depending on the imaging modality , a two-dimensional (2D) image is either transmitted in its entirety through all f ibers of the bundle at once, as in wide-field microscop y, or transmitted pix el-by-pixel sequentially through one f iber at a time, as in laser scanning microscop y. The dif ferent imaging modalities are discussed in section “Imaging Modalities.” The number of picture elements (pix els) within the image is determined by the number of fibers contained in the bundle. F or e xample, the Sumitomo IGN-08/30, a popular f iber-optic bundle, is 800 µm in diameter and contains 30,000 f ibers. Smaller bundles are a vailable with lower number of fibers (eg, IGN-028/06 is a 280 µm diameter bundle with 6,000 individual fibers). The ability to pack so man y f ibers into a small-diameter probe is remarkable, yet the numbers are lo w in comparison with the standard video resolution, such as the VGA for mat, which has 640 × 480 or a total of ~300,000 pix els. An issue related to the relati ve low pixel count is the low pixel density. To minimize cross talk between adjacent fibers, the interf iber spacing (center -to-center distance between neighboring f ibers) is al ways lar ger than the diameter of the light-transmitting cores. F or e xample, individual f ibers in the IGN-08/30 bundle mentioned above have a 2.4 µm core diameter, but the inter-fiber distance is about 4 µm. The dead space betw een adjacent cores reduces the light-coupling efficiency and also gives rise to a pixelated or honeycomb appearance in the image. The pix elation ar tifacts can be remo ved to some e xtent using image processing algorithms, such as lo w-pass f iltering or interpolation,2 but the underlying limit in resolution is not impro ved by such processing methods. If d is the inter-fiber spacing projected onto the sample (taking into account the demagnif ication f actor of the objected 165

166

MOLECULAR IMAGING: PRINCIPLES AND PRACTICE

lens), then according to sampling theor y, the highest spatial frequency that can be resolv ed is 1/2d. Therefore, to achieve a lateral resolution of 1 to 2 µm, the inter -fiber spacing will have to be demagnif ied by a factor of 4 to 8. While that is possib le, such resolution can onl y be achieved b y a signif icant reduction in light throughput 2 because of the relati vely high di vergence of the f iber output as dictated by the f iber numerical aperture (NA). Objective Lenses for Endomicroscopy

The output end of the f iber-optic bundle can be used in direct contact with the tissue sample (in w hich case the lateral resolution is determined by the inter-fiber spacing) or used with an objecti ve lens that (1) projects a (demagnified) image of the f iber bundle output surf ace onto the sample, (2) collects the remitted light from the sample, and (3) reimages it onto the f iber bundle. The physical size of the objecti ve lens, rather than the diameter of the fiber bundle, is often the limiting f actor in miniaturizing the probe. The NA of the objective lens, in a conventional microscope, deter mines the optical resolution and the light-gathering ef ficiency of the imaging system. In a fiber bundle endomicroscope, the resolution is usually not limited by the NA of the objective lens but by the (demagnified) inter-fiber distance at the sample plane. F or optimal coupling of the remitted light into the f iber-optic bundle, the lens should be designed for the specif ic fiberoptic bundle being used, such that the NA on the backside (fiber) of the lens matches the N A of the f iber. For fluorescence molecular imaging, chromatic aberration is of particular concern, especially if more than one fluorophore is to be used for multicolor imaging. Rouse and colleagues3 have designed a multielement lens with an NA of 0.46 and a maximum diameter of 3 mm. The lens has an achromatic range of 480 to 660 nm and is therefore suited for fluorescence molecular imaging applications. Karlson and colleagues4 and Chidley and colleagues5 used injection-molded plastic lenses with aspheric surfaces that can signif icantly reduce the manuf acturing cost. These lens assemblies are 7 mm in diameter and ha ve an NA of 1.0. However, they are designed for reflectance confocal microscopy at 1064 nm and ma y have not been tested for fluorescence imaging over a range of shorter wavelengths. Graded-index (GRIN) lens represents an attracti ve alternative to the con ventional spherical lenses. GRIN lens is made of a glass rod doped with metal ions, such as silver and thallium, to ha ve the inde x of refraction decreasing with radius. The “g raded” parabolic inde x profile causes the light ra ys to bend continuousl y within the lens and sta y focused on a spot, allo wing the real

image to be for med. The ion-exchange doping process 6,7 replaces the need for tightly controlled surface curvatures of miniature lenses. Commercial GRIN lenses are a vailable with NAs from 0.05 to 0.6, diameters from 0.35 to a few millimeters, and v arious pitches with relati vely low cost. An impor tant consideration is that the c ylindrical shape and planar end surf aces of a GRIN lens considerably facilitate miniature assembly. For their simplicity and a vailability, GRIN lenses ha ve been widel y used in v arious types of miniature endoscopic probes, despite some drawbacks, such as aber ration. Endomicroscopes with f iber-optic bundles coupled to GRIN objective lenses ha ve been described b y Knittel and colleagues8 and by Göbel and colleagues. 9

Endomicroscopes Based on Single Optical Fibers A second type of f iber-optic microscope uses a single optical f iber instead of a bundle of f ibers. This kind of microscope is inherently a scanning microscope because a single f iber cannot transmit an image. Consequentl y, the distal (output) end of the f iber has to be scanned in tw o dimensions to acquire an image, either b y mechanically vibrating10–12/rotating13 the f iber tip, the objecti ve lens together with the f iber tip, 14 or by using miniature beam scanners built into the endoscope.15 To achieve diffractionlimited focusing, single-mode f ibers are used w hose small-core diameters (3 to 7 µm) supports the propagation of a single mode of electromagnetic field within the fiber. One major advantage of the single fiber microscope is that images are free of the pixelation artifacts, unlike its fiberoptic bundle counter part. Ho wever, the need for distal scanning adds complexity and bulk to the probe. A tubular probe with a diameter as small as 2.4 mm has been realized that contains tw o pairs of piezoelectric actuators capable of resonantl y driving a cantilevered optical f iber in a 2D (spiral) scan patter n.12 Future advances in microfabrication processes, for e xample, using microelectromechanical systems technology,15,16 will undoubtedly lead to fur ther component miniaturization. F igure 1 sho ws a prototype side-looking optical probe assemb led on a micro-fabricated active silicon optical bench being de veloped for two-photon endomicroscopy.25 Spectrally Encoded Endomicroscopy

As an alternative to 2D beam scanning, a technique called spectrally encoded endoscopy (SEE)17 requires only onedimensional (1D) scanning to produce a 2D image. This

Endomicroscopy

is made possib le by a broadband light source and a dispersive element at the distal end of a single-mode f iber. Each w avelength component from the broadband light source is spatiall y separated b y the dispersi ve element and is brought to a dif ferent focus along a line at the sample b y the objecti ve lens (F igure 2). Backscattered light from different points along this line is brought back through the same dispersi ve element, w hich undoes the spatial separation, and delivered by the same single-mode optical fiber to a spectrometer. The spectrometer analyzes the amount of backscattering light at each location, encoded by its own wavelength. In this means, a 1D (line) image is formed, and physical scanning is only needed in the second dimension (in the direction per pendicular to the line) to obtain a 2D image. 18 Three-dimensional (3D) imaging is also possib le using optical interferometr y.19 A probe diameter of only 350 µm has been used to obtain the v olumetric images with about 400,000 resolv able points at 30 frames/second. 20 For fluorescence molecular imaging, the retur n photons do not ha ve the same w avelength, so an additional de gree of encoding is necessar y. In addition, the retur n (Stok es-shifted) photons do not retrace the same optical path through the dispersi ve

A

167

element, and consequentl y, it is not easil y coupled into the same optical fiber. Considerable reengineering is still needed to implement this interesting technology for fluorescence molecular imaging.

Rigid Endoscopes For small-animal imaging research, which does not need a flexible light guide o ver extended distances, the rigid endomicroscope is an attracti ve choice, allo wing minimally invasive imaging of inter nal or gans by inser ting the rigid probe tip through a small opening. Two types of rigid probes are a vailable. The GRIN endomicroscope22 is made by fusing together a short GRIN objective lens with a high N A (0.4 to 0.6) for tight beam focusing and a rod lens with lo w N A (0.1 to 0.2) for beam rela y.23 By replacing the f iber-optic b undle (see section “Endomicroscopes Based on Single Optical Fibers”) with the rod lens for beam rela y, the GRIN endomicroscope gains in light throughput and image quality (no pixelation artifact) but sacrifices mechanical flexibility. Typical length and diameter of the GRIN endomicroscopes are in the range of 10 to 25 mm and

B

Figure 1. Illustration (A) and image (B) of a prototype side-looking optical probe assembled on a microfabricated-active silicon optical bench. A double-clad photonic crystal fiber enters the device from the right. A microprism and a graded-index (GRIN) lens are visible in the center of the device. Arrows point to microfabricated-contoured thermomechanical actuators that move the prism and the GRIN lens.25

Figure 2. Spectrally encoded endoscope. Light from a broadband source that emerges from the single-mode fiber is spread along the wavelength axis by a dispersive element built into the probe tip so that different positions along this axis are encoded by a different wavelength (color). Tissue is scanned in the perpendicular direction to obtain a 2D image. Adapted from Yelin et al.20

168

MOLECULAR IMAGING: PRINCIPLES AND PRACTICE

0.3 to 1 mm, respectively. A second type of rigid probe, commonly called the stick lens, is a set of narrow diameter objective lenses developed by Olympus specifically for live-animal imaging pur poses.24 They are a vailable in 1.3 mm and 3.5 mm diameters, with N A of 0.5 and 0.7, and are cor rected for near-infrared wavelengths up to 1000 nm.

IMAGING MODALITIES Wide-Field Endomicroscopy In wide-f ield (epifluorescence) endomicroscop y, the entire field of view (FOV) is illuminated, and the image is acquired by a 2D array detector, such as a CCD. The widefield method does not pro vide optical sectioning since light from different tissue depths, both in and out of focus, contributes to image for mation and results in poor contrast. Whether a str ucture can be visualized depends on how bright the tar get is labeled and w hether it is strong enough to be visualized abo ve the backg round. Because no scanning is required, wide-field imaging has the advantages of simpler and less e xpensive instr umentation and faster full-frame acquisition. Both GRIN endoscope probes22,26 and f iber-optic bundles 27 have been used for wide-field endomicroscopy.

Endoscopic Confocal Microscopy In confocal microscop y, a point in the sample is illuminated b y a focused laser spot and remitted light from this point is detected through a confocal pinhole. The point is then scanned in two dimensions to build up an image. Conceptuall y, a straightforw ard way to convert the f iber bundle endomicroscope into a confocal system is to illuminate the indi vidual f ibers in the bundle one at a time and detect the remitted light coming back through the same f iber while rejecting light coming back through adjacent f ibers that car ry multipl y scattered photons and signals from out-of-focus objects. Although it is often stated that the illumination f iber acts as its own confocal pinhole in the detection path, a physical pinhole is in fact necessary to block light coming back through sur rounding f ibers that car ry unwanted photons. To obtain a 2D image, dif ferent fibers in the bundle are illuminated sequentiall y and each f iber detected confocall y. In the simplest implementation, a focused laser is scanned continuousl y over the entrance surf ace of the f iber bundle in a 2D (x–y) raster patter n. This raster scan patter n is rela yed to the distal end of the bundle and imaged onto the sample b y

the miniature objecti ve lens. Because the scan is continuous, the laser beam spends a considerab le portion of the scan time focused on the cladding instead of on the core of the f ibers, reducing the coupling ef ficiency, and increasing the backg round (nonconfocal) fluorescence le vel. Replacing the f iber bundle with a GRIN rod lens alle viates these prob lems and also removes the pixelation artifact, but it limits the imaging targets to str uctures that can be reached with a shor t rigid probe. Using the f iber bundle or the GRIN lens to relay the image has the adv antage that scanning is accomplished outside the endoscope, making it unnecessary to integrate a scan head into the probe. To increase the frame rate, an entire linear ar ray of fibers can be illuminated at once and detected by an array detector through a confocal slit aper ture.28 This arrangement is similar to a slit-scanning confocal microscope, which needs scanning only in the second dimension perpendicular to the long dim ension of the slit. With this approach, images can be obtained at video rate and above. The gain in speed and simplicity with the slit aperture design is balanced b y a compromise in its ability to reject of out of focus light. The rejection has a 1/z dependence as opposed to 1/z2 for the pinhole confocal system, where z is the distance away from the focal plane. 28 As described in section “Endomicroscopes Based on Single Optical Fibers,” a fiber-optic microscope based on a single-mode optical fiber requires a scanning mechanism at the output end of the fiber. The return light is descanned by the same scanner and coupled into the same optical f iber, which acts as a confocal pinhole. This type of endomicroscope is therefore, intrinsically, a scanning confocal microscope. In regular confocal microscopy, when detecting very weak fluorescence signals, it is customar y to increase the detection pinhole size, which increases the sensitivity while sacrificing the axial resolution. In f iber-optic confocal microscopy, light-collection ef ficiency can be impro ved using a doub le-clad f iber that has a center core, an inner cladding, and an outer cladding. The illumination light is carried in a single-spatial mode do wn the center core, whereas the return light is coupled into the multimode inner cladding, w hose diameter is much lar ger than the center core, thereby increasing the light-gathering efficiency.29

Endoscopic Two-Photon Microscopy Two-photon excitation fluorescence (TPF) microscop y has emerged as a powerful intravital imaging technique for biomedical research, par ticularly in small-animal models. Additionally, it has the potential to become a clinical tool for intraoperative tissue characterization and molecular

Endomicroscopy

diagnosis.30 Despite its use of near -infrared excitation that allows deeper tissue penetration and less photodamage, imaging depth in most tissues remain limited , typically to less than a fe w hundred microns from the surf ace. TPF endomicroscopy has been demonstrated with GRIN lenses as focusing optics in conjunction with either GRIN rela y lenses or with f iber-optic delivery.9,11,22 A special consideration for fiber-optic TPF endomicroscop y is the pulsebroadening ef fect w hen ultrashor t laser pulses are propagated through optical f ibers. This is undesirab le because the efficiency of two-photon excitation is reduced as the pulse duration increases. One cause of pulse broadening is g roup velocity dispersion, w hen slightly different “colors” within the laser pulse tra vel at dif ferent velocity through a medium, such as glass (because the inde x of refraction varies with wavelength). As a result, the different colors get out of step and the pulse broadens. Different colors exist in an ultrashor t laser pulse because of the uncertainty relation: the shor ter the pulse the broader its spectrum. F or e xample, a typical Titanium (T i):sapphire laser pulse used for tw o-photon microscop y has a pulse width of about 100 fsec and a spectral width of appro ximately 10 nm, centered at 800 nm (ie, a color spread o ver the 795 to 805 nm range). Group velocity dispersion can be compensated for by introducing a dif ferent delay length to each color (a technique called prechir ping) so that all the colors get back in step again. Another cause of pulse broadening is nonlinear self-phase modulation that occurs at high intensities and is more difficult to compensate. New hollow core or large mode area photonic cr ystal f ibers can deliver femtosecond laser pulses without self-phase modulation. Ho wever, these f ibers are not a vailable as bundles; therefore, it is necessar y to use single f iber with postfiber scanning for two-photon imaging.

Optical Coherence Endomicroscopy Optical coherence tomography (OCT) uses a low coherence (broadband) light source and interferometric techniques to perform depth-resolv ed imaging with resolutions ranging from 2 to 15 µm.31 With large penetration depths of about 2 mm in tissue, cross-sectional imaging capabilities, and relative ease of incor poration into f iber-optic catheters and endoscopes,32 OCT endomicroscop y is a promising candidate for comprehensi ve screening or intraoperati ve tissue characterizations. 33,34 Recent human studies ha ve demonstrated that OCT is capab le of identifying dysplasia in Barret’s esophagus and colonic adenomas and can identify all the histopatholo gic features of vulnerab le plaque.35–37 In the past fe w y ears, the imaging speed of OCT has been impro ved signif icantly using frequenc y

169

domain, instead of time domain, ranging. 33 A minimall y invasive procedure, such as catheterization, is used to deliver the OCT probe to the tar get or gan or system. Through rotation and longitudinal pullback of the inter nal portion of the probe, the OCT system records the full 3D microstructure of the tissue. Subsequent image processing can then be used to produce a for m of vir tual histolo gy, where arbitrar y cross-sectional vie ws of the tissue can be viewed to screen for disease. Gi ven the indi vidual v oxel dimensions of 15 µm × 15 µm × 10 µm, the systems obtain data at a rate of 60 mm 3/s permitting visualization of large tissue volumes with microscopic resolution. The coherent detection scheme used in OCT cannot directly detect fluorescence. Ho wever, several molecular contrast methods ha ve already been suggested for conventional OCT , such as Coherent anti-Stok es Raman shift, molecular probes, and second harmonic generation. Furthermore, functional techniques are on the horizon, including Doppler -OCT for quantitati ve assessment of blood flow38 and polarization-sensitive OCT for quantification of collagen content,39 providing further molecularand ph ysiologic-contrast mechanisms that ma y be used for the study of disease and primar y diagnosis. Recent developments in v arious contrast agents for OCT are reviewed in Boppart and colleagues. 40

IMPLEMENTATION AND APPLICATIONS System Description After taking into consideration the v arious design options discussed in the pre vious two sections, w e have chosen to implement a combined confocal and multiphoton endomicroscopy system based on a GRIN lens probe for in vi vo imaging of small animals. The system uses continuous wave lasers at 491 nm, 532 nm, and 635 nm for single-photon fluorescence e xcitation and a mode-lock ed tunab le Ti:Sapphire laser (780 to 920 nm) for multiphoton e xcitation, to provide broad spectral coverage to a wide range of molecular probes. A schematic of the system is sho wn in Figure 3. Flip mir rors and dichroic splitters are used to select and direct a laser beam to a raster beam scanner comprising a silv er-coated polygon scanner ( x-axis) and a galvanometer (y-axis). We designed the microscope to ha ve a FOV of 250 µm, when a 40 × 0.6 NA objective lens is used. The proximal end of the optical probe is placed appro ximately at the focal plane of the objecti ve lens. A confocal pinhole and a photomultiplier tube (PMT) are used for confocal fluorescence and reflectance imaging, w hereas another PMT-2 is used for tw o-photon fluorescence and second harmonic generation imaging. An 8-bit 3-ch frame

170

MOLECULAR IMAGING: PRINCIPLES AND PRACTICE

A A

B B

C C

D D

Figure 3. Rigid graded-index (GRIN) microendoscope probes. A, A doublet probe with a high numerical aperture (NA) GRIN lens (NA 0.45–0.6, pitch 0.16–0.25) coupled to a low NA relay lens (NA 0.1–0.2, 3/4 pitch). B, A triplet with two high NA GRIN lenses (NA 0.45–0.6, pitch 0.16–0.25) sandwiching a low NA relay lens (NA 0.1–0.2, 3/4 pitch). C, Photograph of three probes with diameters of 1, 0.5, and 0.35 mm, respectively. Each probe is a compound triplet GRIN lens, as depicted in (B), consisting of a coupling lens, an objective lens, and a longer relay lens in between. Minor ticks on the scale bar represent 1 mm. D, Video-rate confocal and multiphoton imaging system. (A–C) adapted with permission from Jung et al.23 and (D) adapted from Kim et al.51

grabber digitizes the output of each PMT at 10 MS/s, acquiring 500 × 500 pixels/frame. The system acquires and displays images in real time at a frame rate of 30 Hz and can save them to hard disk simultaneousl y. There is a real-time display at a video rate ( > 15 frames/second) that considerably facilitates image-guided navigation of the probe within the animal.

Optical Probe and Resolution Figure 3 shows the principle of GRIN endoscopic probes. The scanning strategy shown in Figure 3A involves a doublet probe with a 3/4-pitch relay lens. The collimated laser beam (red ar rows) is focused onto the back f ace of the endoscope probe. The laser focus is scanned transv ersely (dashed arrow line). The scanning focus is rela yed to the image plane within the tissue sample (dashed ar row line). The two-photon-excited fluorescence (green arrows) generated at the sample focal spot returning through the endoscope probe is detected with a PMT via a dichroic mir ror. There are several variants of this strategy. The triplet probe shown in F igure 3B is comprised of an objecti ve lens, a relay lens, and a 1/4-pitch GRIN coupling lens with high NA matching that of a microscope objecti ve that couples the excitation laser light into the probe. We designed and f abricated se veral probes using 1-mm-diameter commercial GRIN lenses (NSG and Grintech) in the triplet str ucture. It comprises tw o high NA (0.45 to 0.6) GRIN lenses (pitch 0.16 to 0.25) and a

half-pitch low NA (0.1 to 0.2) relay lens in between. The probes w e used in the e xperiments described here are 15 mm long and have 0.45-imaging NA, a FOV diameter of 250 to 300 µm, and a working distance of 0 to 300 µm with water immersion. In two-photon excitation imaging at 800 nm, the measured resolution was 1.1 ± 0.08 µm in x or y and 13.4 ± 0.3 µm in z, defined as full width at half maximum. In confocal imaging, we measured the resolution as a function of the pinhole diameter . Pinhole sizes greater than the Airy disk are commonl y used for deeptissue imaging because lar ger pinhole sizes impro ve the photon collection ef ficiency at the e xpense of reduced axial resolution. Consistent with the theory, we measured the transverse resolution to be relatively insensitive to the pinhole size and found a strong dependence for both axial resolution and signal strength (Figure 4). We chose to use a pinhole with one Airy disk size (50 µm). This provided a resolution of 1.5 ± 0.08 µm in x or y and 12.4 ± 0.3 µm in z, compared with that of tw o-photon imaging. These values are approximately two times larger than theoretical diffraction limits of 0.45-NA optics, which is attributed to spatial aberration of the GRIN probe.

Method to Mitigate Chromatic Dispersion for Multicolor Confocal Imaging Multicolor imaging in confocal modality is challenging because the GRIN probe has signif icant chromatic aberration. We measured that the focal depth w as dif ferent

Endomicroscopy

A A

171

B B

Figure 4. Measure resolution (A) and signal strength (B) as a function of confocal pinhole size. Dotted lines in (A) denote the theoretical diffraction limits of a lens with numerical aperture 0.45. Adapted from Kim et al.51

between 491-nm and 635-nm e xcitation wavelengths by 35 to 90 µm, depending on specif ic GRIN lenses used. Such a large variation of focus, cor responding to several cell la yers, w ould be a signif icant prob lem fr ustrating multicolor-excitation cellular imaging. To mitigate this problem, we implemented a simple technique; w henever the excitation wavelength was changed, we adjusted the vertical position of the objecti ve lens b y a precalibrated distance (35 to 90 µm) accordingly so that the focal plane of the probe w as kept unchanged. Fur thermore, we optimized the pinhole position and the spectral width of a filter to minimize image b lur due to the w avelength differences between e xcitation and emission and within the emission band.

Images of Test Samples To test the imaging system, we acquired confocal images of a pine embryo at excitation wavelengths of 491 and 635 nm, respectively, using the chromatic aber ration compensation technique described earlier . A mer ged image (F igure 5) demonstrates co-re gistration of the red and g reen fluorescence images within a fe w microns o ver nearly the entire FOV. Image blurring near the FOV boundary is due to f ield curvature. Each image w as averaged over 30 frames (total acquisition time 1 sec) at the same sample position.

In Vivo Imaging of Intact Skin of Mice To demonstrate in vi vo tissue imaging, we conducted noninvasive confocal and multiphoton endoscop y in the ear skin of an anesthetized mouse. Using the GRIN probe, w e were able to visualize indi vidual dendritic cells e xpressing green fluorescent proteins (GFP) in epider mis and der mis (Figure 6A), blood plasma labeled with Rhodamine-B dyes conjugated with dextran after tail-vein injection (2,000,000

Figure 5. Two-color confocal image of a test sample (pine embryo) taken with the graded-index endoscope. Scale bar = 25 µm. Adapted from Kim et al.51

MW, 200 µg/200 µL) (Figure 6B), and collagen f ibers via endogenous SHG in the der mis (F igure 6C). Video-rate monitoring not onl y considerably facilitates image na vigation but also allo ws monitoring of f ast dynamic processes, such as cell traf ficking in the b lood stream. F igures 6D–F show the mo vie frames acquired b y confocal microendoscopy from a GFP+ Tie-2 mouse in which all the vasculature endothelial cells are GFP +, after tail-v ein injection of human ovarian cancer cells labeled with a membrane dy e, DiD (Invitrogen). Flowing cells could be obser ved clearly, and the cell count and v elocity could be measured. The maximum penetration depth was 50 to 100 µm from the surface of the skin, limited b y finite signal-to-noise ratio and the out of focus backg round.

In Vivo Imaging of the Mouse GI Tract A number of mouse models of inflammator y bowel disease and colon cancer are a vailable to study the mucosal immune system, colitis, and development of cancer in the gut. However, due to the small size of the GI or gans in mice, standard microscopy would require extensive surgical opening and , therefore, has not been widel y used. In the past fe w years, f iber-optic scanning endomicroscop y systems ha ve become commerciall y a vailable and ha ve been used for cellular and molecular imaging of the GI organs in mice and humans. 41,42 In addition, miniature wide-field endoscopes ha ve been demonstrated for fluorescence mouse colonoscop y.43,44 We ha ve recentl y

172

MOLECULAR IMAGING: PRINCIPLES AND PRACTICE

A

B

C

D

E

F

Figure 6. Images from the intact skin of an anesthetized mouse, taken with the graded-index endomicroscope. A, A confocal fluorescence image showing major histocompatibility complex (MHC) class-II+ Langerhans cells expressing green fluorescent proteins (GFP) in the epidermis. The MHC class II GFP mouse was kindly provided by Dr Marianne Boes at Harvard Medical School. Excitation 491 nm, emission 520/35 nm. B, A two-photon fluorescence image of blood vessels labeled with rhodamine dextran. Excitation 800 nm, emission 590/80 nm. C, Endogenous collagen second harmonic signal excited at 800 nm and detected at 400 nm. D–F, A sequence of frames showing ovarian cancer cells (red) in blood circulation, superimposed on a green fluorescence image of blood vessels in a GFP+ Tie-2 mouse. Scale bar = 50 µm. Adapted from Kim et al.51

studied the possibility of using GRIN microendoscope probes to obtain high-resolution fluorescence images. In one experiment, we used the 1-mm-diameter GRIN probe to image the small and lar ge intestines of a major histocompatibility comple x class-II GFP mouse via a minimally in vasive laparotom y procedure (F igure 7A). The dendritic-like cells in the lamina propria of the colon are clearly resolved (Figure 7B). The penetration depth of the confocal GRIN probe was sufficient to visualize the GFP+ epithelial dendritic cells through the entire colonic w all (Figure 7C). For noninvasive colonoscopy, we fabricated a side-looking GRIN probe with a 90º prism reflector attached to the distal end (F igure 7D). The colonoscope was introduced via rectal inser tion. The side-looking probe design a voids the need for insuf flation b y an air pump. Figure 7E shows the same GFP structure as seen in Figure 7F, now imaged from the inside of the colon.

In vivo Imaging of Mouse Model of Heart Transplant Coronary ar tery disease represents the major threat to patients w ho ha ve under gone hear t transplantation. Endomicroscopy can visualize the immune responses and disease progression in the mouse models of chronic or gan rejection,45 but cellular imaging of a transplanted hear t has not been possible due to heartbeat motion. To overcome this problem, we have developed a miniature tissue-holding suction tube and used it with the GRIN endoscope probe for

heart imaging. We modified a 15-gauge hypodermic needle (inner diameter 1.37 mm, outer diameter 1.83 mm) and connected it to a mini v acuum pump with a flexible plastic tube. A 1-mm-diameter GRIN probe w as inserted through the inner channel of the needle tube. The blunted distal tip of the needle is made to contact the tissue so that it can hold it by gentle suction. We found an optimum range of vacuum pressure in the neighborhood of 100 mm Hg or 13 kP a, which is sufficient to freeze the tissue mo vement but does not cause apparent adverse effects, such as tissue damage or ischemia. We have used this de vice to obtain images from the inter nal or gans of mice, including small and lar ge intestines, spleen, kidney, and transplanted heart. Figure 8 shows a sequence of images obtained from a coronar y ar tery in a syngeneic hear t transplantation mouse model w here the hear t is not rejected. The hear t was placed in the abdominal ca vity in an actin GFP + recipient mouse.46 The images show the recipient’s GFP + cells flowing in the coronar y ar tery, some of w hich are arrested in the endothelial w all and extravasate (circles). During the course of imaging for about 30 min, the tissue surface was held by suction; no adverse phenomena were detected and the blood flow appeared normal.

CHALLENGES AND OUTLOOK In summary, endoscopic microscopy is a promising technology for molecular imaging research because it provides unique ways to look into tissue at a resolution pre viously

Endomicroscopy

C

B

A

D

173

EE

Figure 7. A, In vivo imaging of the colon of a MHC class II green fluorescent proteins (GFP) mouse using the confocal graded-index (GRIN) endomicroscope. The probe was inserted through a small incision in the skin. B, Dendritic-like GFP+ cells in the lamina propria. C, GFP+ epithelial dendritic cells in the colonic wall. D, A GRIN endoscope with a side-looking prism. E, The same GFP+ epithelial dendritic cells imaged from the inside of the colon using the side-looking probe. Scale bar = 50 µm.

A

B B

Figure 8. A, Anatomy of transplanted heart in the abdominal cavity. B, A sequence of images showing the recipient’s green fluorescent proteins+ cells flowing in the coronary artery, some of which are arrested in the endothelial wall and extravasate (circles). Scale bar = 50 µm.

unattainable with traditional imaging modalities. Se veral commercial endomicroscopes are cur rently available providing a range of options from fiber-optic bundles (Mauna Kea) to single f iber deli very (OptiScan) and rigid stick lens probes (Ol ympus). As in other fluorescence-based techniques, rapid pro gress in the de velopment of fluorescent molecular sensors and repor ters will enab le more selective imaging of molecular targets and assessing cellular functions. In addition, ne w contrast mechanisms, such as Coherent anti-Stok es Raman scattering (CARS) microscopy, are also being studied for chemical imaging

of molecular species based on their vibrational signatures.47 Indeed, CARS microscopy through a 1.3 mm rigid probe (Olympus stick lens) has recentl y been repor ted.48 Fiber-optic transmission of picosecond laser pulses for CARS microscop y49 is potentiall y less demanding, because of lower peak power and narrower spectral width, compared with transmission of femtosecond laser pulses used for two-photon microscopy. Ongoing challenges include fur ther component optimization and miniaturization, using the latest microfabricated and nanof abrication technolo gy. In addition,

174

MOLECULAR IMAGING: PRINCIPLES AND PRACTICE

system-level inte gration will become more critical for practical users. Issues, such as FO V, na vigation, and motion of the li ve subject will confront da y-to-day practitioners and will ha ve to be addressed , likely in an application-specific manner.

FOV In general, FOV of a microscope decreases with increasing magnification and the N A of the objecti ve lens. The FOV also decreases as the size of the objecti ve lens is reduced while maintaining the NA and magnification. For example, the 0.5-N A microprobe ha ving a diameter of 1.3 mm, developed by Olympus, gives a FOV of 200 µm, whereas a standard objective lens with the same NA would allow at least three times larger FOV. As a result, sampling volume tends to be limited in small-diameter high-resolution endomicroscopy. A small FO V can cause sampling er rors in quantitative analysis or clinical diagnosis. In addition, it is not possib le to monitor spatiall y distant e vents at the same time, when separated more than the FOV. Therefore, FOV is an impor tant design parameter in de veloping or using endomicroscop y. In some applications, such as intraluminal imaging, rapid scanning of the probe o ver a wide area larger than the otherwise limited FO V may be a viable solution.18,33 The resulting large data set places a premium demand on the computation po wer to acquire, process, analyze, and display volumetric information.

Navigation The small FOV of a endomicroscope also renders it difficult to kno w e xactly w hat inter nal str ucture is being imaged. Knowledge of the anatomy alone may not be sufficient to na vigate the probe to the right tar get location, and a navigation strategy may need to be de veloped that combines imaging modalities capab le of visualizing different size scales, so that major anatomic landmarks can be used for initial cursory positioning of the probe, before zooming in to the precise tar get location. In the most demanding experiments, one may want to image the same cell or group of cells over time. Such experiments require repeated positioning accuracy to better than 100 µm. Fast scanning (ie, acquiring images at video rate or close to video rate) is v ery useful for rapidl y surveying the local tissue landscape 50 and orienting the probe. 51 Alternatively, the probe can be implanted or affixed to tissue, for example, with a head-mounted gear that allows long-term imaging of neurons in the brain, even in awake animals.11

Motion Physiological motion of the live subject, such as breathing and heartbeat, can be a serious prob lem in high-resolution endomicroscopy. A tissue mo vement of more than the instrumental resolution during image acquisition results in undesirable ar tifacts, such as b lur and distor tion. Image processing methods or time-gated data acquisition in synchronous with a periodic mo vement ma y be useful in reducing motion ar tifacts. Application of immobilizing drugs, gentle pressing of the tissue surf ace with the end of endoscopic probe or suck device, such as the one described in section “in vi vo Imaging of Mouse Model of Hear t Transplant” are all possible solutions to stabilize the tissue. Perhaps the most elegant solution, one that does not introduce any perturbation to the tissue, is to acquire images at very high frame rates such that during each frame the motion is ef fectively frozen. Image co-re gistration algorithms can then be used to eliminate motion-induced image shift from frame-to-frame. 50 Extension of this method to 3D image stabilization is in principle feasib le but remains technically challenging.

REFERENCES 1. Flusberg BA, Cocker ED, Piyawattanametha W, et al. F iber-optic fluorescence imaging. Nat Methods 2005;2:941–50. 2. Delaney PM, Harris MR. In: Pawley JB, editor. Handbook of biological confocal microscop y. Ne w York: Plenum Press; 1995. p. 515–23. 3. Rouse AR, Kano A, Udovich JA, et al. Design and demonstration of a miniature catheter for a confocal microendoscope. Appl Opt 2004;43:5763–71. 4. Carlson K, Chidley M, Sung KB, et al. In vivo fiber-optic confocal reflectance microscope with an injection-molded plastic miniature objective lens. Appl Opt 2005;44:1792–7. 5. Chidley MD, Carlson KD, Richards-Kortum RR, Descour MR. Design, assembly, and optical bench testing of a high-numerical-aper ture miniature injection-molded objecti ve for f iber-optic confocal reflectance microscopy. Appl Opt 2006;45:2545–54. 6. Ohmi S, Sakai H, Asahara Y, et al. Gradient-index rod lens made by a double ion-exchange process. Appl Opt 1988;27:496–99. 7. Messerschmidt B, Possner T, Goering R. Colorless g radient-index cylindrical lenses with high numerical aper tures produced b y silver-ion exchange. Appl Opt 1995;34:7825–30. 8. Knittel J, Schnieder L, Buess G, et al. Endoscope-compatib le confocal microscope using a g radient index-len system. Opt Commun 2001;188:267–73. 9. Göbel W, Kerr JND, Nimmerjahn A, Helmchen F. Miniaturized tw ophoton microscope based on a flexible coherent fiber bundle and a gradient-index lens objective. Opt Lett 2004;29:2521–23. 10. Helmchen F, Fee MS, Tank DW, Denk W. A miniature head-mounted two-photon microscope high-resolution brain imaging in freel y moving animals. Neuron 2001;31:903–12. 11. Flusberg BA, Jung JC, Cocker ED, et al. In vivo brain imaging using a por table 3.9 g ram tw o-photon fluorescence microendoscope. Opt Lett 2005;30:2272–74. 12. Myaing MT, MacDonald DJ, Li XD. Fiber-optic scanning two-photon fluorescence endoscope. Opt Lett 2006;31:1076–79.

Endomicroscopy

13. Bird D , Gu M. Two-photon fluorescence endoscop y with a micro-optic scanning head. Opt Lett 2003;28:1552–54. 14. Ota T, Fukuyama H, Ishihara Y, et al. In situ fluorescence imaging of organs through compact scanning head for confocal laser microscopy. J Biomed Opt 2005;10:024010. 15. Piyawattanametha W, Barretto RPJ, Ko TH, et al. Fast-scanning twophoton fluorescence imaging based on a microelectromechanical systems tw o- dimensional scanning mir ror. Opt Lett 2006; 31:2018–20. 16. Dickensheets DL, Kino GS. Micromachined scanning confocal optical microscope. Opt Lett 1996;21:764–66. 17. Tearney GJ , Webb RH, Bouma BE. Spectrall y encoded confocal microscopy. Opt Lett 1998;23:1152–54. 18. Yelin D, Boudoux C, Bounia BE, Tearney GJ. Lar ge area confocal microscopy. Opt Lett 2007;32:1102–04. 19. Yelin D , White WM, Motz JT , et al. Spectral-domain spectrall yencoded endoscopy. Opt Express 2007;15:2432–44. 20. Yelin D , Rizvi I, White WM, et al. Three-dimensional miniature endoscopy. Nature 2006;443:765–765. 21. Motz JT, Yelin D, Vakoc BJ, et al. Spectral- and frequenc y-encoded fluorescence imaging. Opt Lett 2005;30:2760–2. 22. Jung JC, Schnitzer MJ . Multiphoton endoscop y. Opt Lett 2003;28:902–4. 23. Jung JC, Mehta AD, Aksay E, et al. In vi vo mammalian brain imaging using one- and tw o-photon fluorescence microendoscop y. J Neurophysiol 2004;92:3121–33. 24. Alencar H, Mahmood U , Ka wano Y, et al. No vel multiw avelength microscopic scanner for mouse imaging. Neoplasia 2005;7:977–83. 25. Chen SC. [PhD disser tation]. Depar tment of Mechanical Engineering, Massachusetts Institute of Technology; 2007. 26. Monfared A, Ble vins NH, Cheung ELM, et al. In vi vo imaging of mammalian cochlear b lood flo w using fluorescence microendoscopy. Otol Neurotol 2006;27:144–152. 27. Hirano M, Yamashita Y, Miyaka wa A. In vi vo visualization of hippocampal cells and dynamics of Ca2+ concentration during anoxia: feasibility of a f iber-optic plate microscope system for in vivo experiments. Brain Res 1996;732:61–68. 28. Sabharwal YS, Rouse AR, Donaldson LT, et al. Slit-scanning confocal microendoscope for high-resolution in vivo imaging. Appl Opt 1999;38:7133–44. 29. Yelin D , Bouma BE, Yun SH, Tearney GJ . Doub le-clad f iber for endoscopy. Opt Lett 2004;29:2408–10. 30. Zipfel WR, Williams RM, Christie R, et al. Live tissue intrinsic emission microscop y using multiphoton-e xcited nati ve fluorescence and second harmonic generation. Proc Natl Acad Sci U S A 2003; 100:7075–80. 31. Fujimoto JG. Optical coherence tomo graphy for ultrahigh resolution in vivo imaging. Nat Biotechnol 2003;21:1361–67. 32. Tearney GJ, Brezinski ME, Bouma BE, et al. In vivo endoscopic optical biopsy with optical coherence tomo graphy. Science 1997; 276:2037–39. 33. Yun SH, Tearney GJ, Vakoc BJ, et al. Comprehensive volumetric optical microscopy in vivo. Nat Med 2006;12:1429–33.

175

34. Yaqoob Z, Wu J, McDowell EJ, et al. Methods and application areas of endoscopic optical coherence tomography. J Biomed Opt 2006; 11:063001. 35. Poneros JM, Brand S, Bouma BE, et al. Diagnosis of special ized intestinal metaplasia by optical coherence tomography. Gastroenterology 2001;120:7–12. 36. Westphal V, Rollins AM, Willis J, et al. Cor relation of endoscopic optical coherence tomo graphy with histolo gy in the lo wer-GI tract. Gastrointest Endosc 2005;61:537–46. 37. Jang IK, Tearney GJ, MacNeill BM, et al. In vi vo characterization of coronary atherosclerotic plaque using optical coherence tomography. Circulation 2005;111:1551–55. 38. Vakoc BJ , Shishk o M, Yun SH, et al. Comprehensi ve esophageal microscopy b y using optical frequenc y-domain imaging (with video). Gastrointest Endosc 2007;65:898–905. 39. Nadkarni SK, Pierce MC, P ark BH, et al. Measurement of collagen and smooth muscle cell content in atherosclerotic plaques using polarization-sensitive optical coherence tomo graphy. J Am Coll Cardiol 2007;49:1474–81. 40. Boppart SA, Oldenbur g AL, Xu CY, Marks DL. Optical probes and techniques for molecular contrast enhancement in coherence imaging. J Biomed Opt 2005;10:41208. 41. Goetz M, F ottner C, Schir rmacher E, et al. In-vi vo confocal real-time mini-microscopy in animal models of human inflammatory and neoplastic diseases. Endoscopy 2007;39:350–6. 42. Kiesslich R, Goetz M, Lammersdorf K, et al. Chromoscop y-guided endomicroscopy increases the diagnostic yield of intraepithelial neoplasia in ulcerative colitis. Gastroenterology 2007;132:874–82. 43. Becker C, Fantini MC, Neurath MF. High-resolution colonoscopy in live mice. Nat Protoc 2006;1:2900–04. 44. Alencar H, Funo vics MA, F igueiredo J, et al. Colonic adenocarcinomas: near-infrared microcatheter imaging of smart probes for early detection—Study in mice. Radiology 2007;244:232–8. 45. Benichou G, Akiyama Y, Roughan J, et al. Mechanisms of alloreco gnition. In: Wilkes D, Burlingham W, editors. Immunobiolo gy of organ transplantation. Ne w York: Kluv er Acad./Plenum Press; 2004. p. 107–37. 46. Hasegawa T, Visovatti SH, Hyman MC, et al. Heterotopic vascularized murine cardiac transplantation to study g raft arteriopathy. Nat Protoc 2007;2:471–80. 47. Evans CL, Potma EO, Pouris’haag M, et al. Chemical imaging of tissue in vivo with video-rate coherent anti-Stokes Raman scattering microsocpy. Proc Natl Acad Sci U S A 2005;102:16807–12. 48. Wang H, Huff TB, Fu Y, et al. Increasing the imaging depth of coherent anti-Stok es Raman scattering microscop y with a miniature microscope objective. Opt Lett 2007;32:2212–4. 49. Legare F , Ev ans CL, Ganikhano v F , Xie XS. Towards CARS endoscopy. Opt Express 2006;14:4427–32. 50. Veilleux I, Spencer J A, Biss DP , et al. In vi vo cell tracking with video rate multimodality laser scanning microscop y. IEEE J Select Top Quantum Electron 2008;14:10–18. 51. Kim P, Puoris’haag M, Cote D , et al. In vi vo confocal and multiphoton microendoscopy. J Biomed Opt 2008;13:010501.

13 INTRAVITAL MICROSCOPY THORSTEN R. MEMPEL, MD, PHD

The introduction of optical imaging, that is, the use of light (the visib le par t of the electromagnetic spectr um) as a source of image contrast, into the practice of molecular imaging has created an intersection with the field of intravital microscopy (IVM). Both disciplines are often refer red to as “in vivo imaging” but have developed from quite distinct conceptual origins. While the idea of molecular imaging has been de vised b y radiolo gists as an approach to improve the diagnosis of disease in human patients, IVM began as a means to satisfy the curiosity of natural scientists about the inner w orkings of v arious nonhuman subjects, such as bats and fro gs. Since the introduction of fluorescence to their w ork, intra vital microscopists ha ve been gathering e xperience with this tool and are now at a stage where improved imaging technolo gy will allow the dynamic visualization of molecular processes at subcellular resolution in vi vo. Molecular imaging and IVM will thus likely benef it from each other , and cross-fertilization should be culti vated by regular peeks into the neighbor’ s yard. In this chapter, some historical and technical aspects of IVM and its application to biomedical research will be reviewed in order to pro vide a resource and star ting point for further exploration for interested researchers rooted in molecular imaging.

WHY IVM? IVM is the obser vation of biolo gical processes within the physiological conte xt of a li ving specimen using a microscope. It provides dynamic infor mation about th e mechanics of multicellular organisms over a wide spatial and temporal range (less than a micrometer to more than a millimeter and m icroseconds to months). Its unique power lies in providing the ability to study the integrated function of cellular and subcellular systems subjected to comple x information inputs the y encounter in vivo. It compares to “seeing an animal in its natural 176

environment rather than in a zoo” (Mark Davis, personal communication). In multicellular organisms, biological processes are largely controlled at the single-cell level. Each individual cell is equipped with sensors for e xternal stimuli, a subcellular network of information pathways to transmit and integrate these inputs, and v arious w ays to respond b y altered behavior (eg, migration, cell division) or by communicating change back to the en vironment (expression of secreted or cell surf ace molecules). When multicellular systems are reconstituted in vitro, the external stimuli can be fairly well controlled (defined stimuli are varied while unknown and ar tifactual stimuli are k ept constant) to study specif ic responses at the single cell or population level. i n vivo, however, cells are subjected to a comple x and rapidl y changing spectr um of signals that v ary between dif ferent p hysiological and patho-ph ysiological states of an or ganism and betw een its dif ferent anatomic microenvironments. Except for some physical stimuli (eg, pressure, light, or temperature), these signals are received from other cells either through direct ph ysical contact, soluble factors (eg, cytokines, hormones) or via extracellular matrix components. These for ms of cell-cell communication therefore constitute another infor mation network at the supracellular le vel. The sum of these signals, which in their entirety are either not kno wn or cannot be simulated in vitro, to gether with a cell’ s genetic make-up, determines its function. Due to the dynamically changing nature of tissue environments and their spatial hetero geneity, the situation of two cells of the same lineage and at the same developmental stage, within the same tissue but separated b y a fe w micrometers, may be dramatically different, and their subsequent fates may be entirely distinct. Such hetero geneity is overlooked in e x vivo population measurements, w here similarity and dissimilarity of cells are deter mined based on a fe w selected def ining characteristics and w here the

Intravital Microscopy

spatial infor mation about otherwise tr uly similar cells is lost. In order to obtain an inte grated view of how cellular functions are controlled , one must def ine the rele vant forms of cellular infor mation exchange and then measure the biolo gical response of indi vidual cells i n vivo. The only w ay to achie ve such discriminator y potential is through single-cell measurements i n vivo, and cur rently, the only available methodology for this purpose is IVM. As a consequence of its unique capabilities, IVM has had major impact in multiple areas of biomedical research, first in microcirculator y, inflammation, and angio genesis research, later in cancer biolo gy, immunology, microbiology, de velopmental biolo gy, and neuroscience. F or the purpose of illustrating the principles of IVM, the focus of this chapter will be on w ork perfor med to in vestigate phenomena of inflammation and of the immune system, but other f ields will be included to highlight additional technological or conceptual aspects of IVM.

A SHORT HISTORY OF IVM IVM technolo gy is b y no means a recent de velopment. Some of the scientists using the earl y microscopes of the 17th century were intravital microscopists, among them Marcello Malpighi and Antonie v an Leeuw enhoek, the discoverer of bacteria. Inspired b y the privilege to be the first ones to redisco ver the w orld at a dif ferent spatial scale, they studied and recorded a g reat variety of specimen, ranging from unicellular or ganisms to insects to vertebrates. Their attention must ha ve been par ticularly drawn to the dynamic phenomena of the blood circulation. Malpighi studied small ar teries and v eins in the lungs of live frogs, but it was only in dried lung preparations where he discovered the existence of small channels connecting the ar teries and the v eins.1 It was Leeuwenhoek who, during his numerous intravital observations of blood flo w in the micro vessels of tadpoles, f ish, crabs, rooster combs, rabbit ears, and bat wings, was the first to demonstrate blood flow in these smallest vessels in living animals and thus conf irmed William Har vey’s theor y of the blood circulation, although he did not yet identify capillaries as distinct str uctural units. 2 He e ven constructed various specialized specimen containment apparatus, such as an aquatic chamber for eels (Figure 1), with which he demonstrated the microcirculation of their tailfin to the Russian czar Peter the Great. Leeuwenhoek used onl y a v ery simple microscope, consisting of a single high-magnif ication lens incor porated into a metal plate and scre ws to adjust the relati ve position of the specimen. Ho wever, the technolo gical

177

future w as with the compound microscope, as used b y Robert Hooke.3 Yet, it was only following improvements in glass manufacturing and the development of the achromatic lens in the 1830s that the microscopes produced were of a quality that allo wed English ph ysician and physiologist Marshall Hall to identify capillaries as distinct structural units4 and Hall as well as German pathologist and zoologist Rudolph Wagner to describe the cellular components of the b lood and their dif ferential behavior in the microcirculation of various amphibians and fish5 (see Figure 1). The British neurophysiologist Augustus Waller (the name-gi ver of Wallerian degeneration of peripheral ner ves), after unsuccessful attempts to image the human prepuce, also resorted to cold-blooded animals when first describing the extravasation of blood cells into tissues through inter-endothelial gaps in the frog tongue in vivo and estab lished the identity of the pus cor puscles with the leuk ocytes of the b lood.6 Julius Cohnheim later rediscovered Waller’s w ork but had also independentl y come to the same conclusion, w hich was in stark contrast to the beliefs of his academic teacher Rudolph Virchow, who thought that all extravascular cells were derived from tissue-resident precursors.7,8 Cohnheim thus laid a cornerstone to our cur rent understanding of inflammation through the use of IVM. A subsequent technological breakthrough was the superfusion of membranous tissues and or gan surfaces with temperature-controlled ph ysiological saline, which g reatly impro ved IVM in w arm-blooded animals, such as do gs, cats, rabbits, and guinea pigs, and indicates that ef forts to maintain obser ved tissues in a near-physiological state w ere already reco gnized as critical.9 IVM setups, although still quite bulk y, had already remarkab le similarity to the v arious contraptions used for the same pur pose today (see F igure 1). Using this approach, Bizzozero disco vered thrombocytes as the third cor puscular b lood component and carefully described their contrib ution to thrombosis in the mesentery of the rabbit and guinea pig. 10 One w ay b y w hich IVM studies in w arm-blooded animals w ere usuall y limited w as that the preparations were not stab le for more than a fe w hours, precluding studies of phenomena that occur over the range of days or weeks. This shor tcoming w as g reatly alle viated b y the development of chronic animal windo w chambers through Sandison under the guidance of Eliot and Eleanor Clark.11 The wound-healing processes occurring in these chambers allo wed researchers to monitor angiogenesis in endother ms.12 Glen Algire later adapted the chamber model to the dorsal skinfold of the mouse 13 and together with Harold Chalk ey star ted to implant tumors

178

MOLECULAR IMAGING: PRINCIPLES AND PRACTICE

into these chambers to study the angiogenic response of the host, thus initiating the f ield of angio genesis in cancer biology.14 It was probably ophthalmologists who f irst used fluorescence for their mesoscopic obser vations of the human eye, for example, after peroral administration of fluorescein sodium to patients, 15 but soon after the de velopment of the fluorescence microscope at the be ginning of the 20th century, intravital microscopists also took advantage of this new modality (see Figure 1).16 Ellinger and Hirt were aware of the pH dependence in the fluorescence of fluorescein during their studies on glomer ular filtration in the amphibian kidne y, making their studies also the f irst to mak e functional fluorescence-based measurements i n vivo.16 Beyond this, the introduction of fluorescence transfor med IVM in at least two more ways. First, the chemical properties of many of the subsequently used fluorescent intravital dyes allowed specific tagging of anatomic compartments or various tissue components or e ven subcellular compar tments, an approach later extended through their conjugation to other molecules with functional proper ties (such as dextrans or antibodies). Second , the possibility to obtain high quality optical images through epi-illumination extended the range of specimen suitab le for IVM obser vation from translucent tissues to solid or gans. As a consequence, researchers could advance from studying the basic principles of the microcirculation in tissues selected for their practical adaptability to IVM to studying the par ticularities of vascular systems in organs selected for their biological significance. Several other technolo gical de velopments, w hose impact on research was not specific to the practice of IVM, nevertheless generated significant advances in the field. The ability to record and document in an objective manner what an indi vidual researcher sa w under the microscope, using microphotography or microcinemaphotography, provided the oppor tunity for independent inter pretation of observations and their v alidation b y other researchers. While the f irst preser ved microcinemato graphic recordings of a di viding sea urchin embr yo date back to the first decade of the 20th century,17 motion picture recordings of intravital microscopic obser vation did not become popular until the 1940s.18 The recording of dynamic phenomena at re gularly spaced time inter vals also uncoupled the timing of the vie wing of indi vidual frames from the timing of the recording. This permitted the visualization of dynamic processes that are otherwise either too f ast or too slow for our visual perception through the use of slo w motion or time-lapse displa y and literally opened up the temporal dimension as the microscope had opened up the spatial dimension 300 years earlier.

As most other disciplines, IVM research also benefited from the increasing focus on inbred rodents as e xperimental subjects and the oppor tunity to study the molecular mechanisms of phenomena obser ved i n vivo by e xperimental manipulation through the use of b locking monoclonal antibodies or mutant mouse strains. An additional benef it of impro ved recording technolo gy has been the possibility of more in-depth, of f-line analysis of IVM recordings, especiall y due to the no w widespread use of digital storage mediums, w hich has in some areas allowed for some de gree of automation of the otherwise cumbersome and occasionall y bias-prone manual anal ysis of IVM data. As a consequence of improved means of data analysis and quantitation, IVM research, for instance in immunology, has recentl y again found g reater acceptance by researchers outside the circles of specialists. The last re volution to the practice of IVM has been the introduction of nonlinear optical imaging modalities to biological research, in particular multiphoton microscopy (MPM), in 1990 by the group of Watt Webb.19 As discussed belo w, multiphoton intra vital microscop y (MP-IVM) for the f irst time allowed researchers to mak e three-dimensionally resolv ed obser vations of dynamic processes deep within intact tissues in living organisms at microscopic resolution, w hile at the same time minimizing the unw anted effects of light-illumination. Neuroscientists were the first by a wide margin to make use of this powerful new technology for their studies,20,21 followed by developmental biologists,22 cancer biologists,23 physiologists,24 and immunologists.25 Currently, we are still in the ascending slope of the e xponential g rowth phase in se veral of these f ields with re gard to technolo gical development of IVM and its application to biolo gical questions.

SOURCES OF IMAGE CONTRAST In principle, an y for m of light-matter interaction can be used as a source of optical image contrast in IVM. I n practice, utilization of light absor ption, dif fraction, and reflection, as w ell as of fluorescence, has traditionall y dominated, but several new, especially nonlinear, modalities, have started to enter the f ield.

Bright-field Illumination and Related Conventional Techniques The most straightforw ard approach to obtain images of tissues by IVM is by trans- or epi-illumination through a conventional broad-spectrum light source, such as a halogen lamp. Optical contrast is hereb y generated either

Intravital Microscopy

A

B

D

C

E

179

180

MOLECULAR IMAGING: PRINCIPLES AND PRACTICE

through the differential absorption or through diffraction of light by various tissue components. Image for mation through absor ption requires the object of interest to contain signif icant concentrations of light-absorbing molecules and to be sur rounded by nonabsorbing tissue elements. Absorption leads to a decrease in amplitude of transmitted light, which becomes visible at the image plane through diminished signal intensity that can be detected by eye or recorded b y camera. Unstained biolo gical tissues are usually poorly absorptive, but one example of a strongly absorbing molecule is hemoglobin, which is contained in red b lood cells, thereb y enab ling the prominent visualization of blood vessels by IVM (Figure 2). Sources of light dif fraction, on the other hand , are abundant in tissues since diffraction is caused by virtually all inhomo geneities in refracti ve inde x, so-called phase gradients, provided by objects that are at a similar spatial scale as the w avelength of light used , such as cell membranes or g ranules. One approach to visualize dif fraction is to e xclude parts of the dif fracted light from the detection pathw ay in an asymmetric f ashion so that ne gative and positive interference of the diffracted light can lead to detectable net changes in light amplitude at the image plane. Differential interference contrast (DIC), oblique illumination, and Hof fman modulation contrast are techniques to achie ve this. Ho wever, even despite the added benefit of their usuall y e xcellent axial resolution, phase gradient techniques have not gained widespread popularity among intra vital microscopists. This ma y in par t be due to the difficulty of integrating the required equipment into IVM setups. DIC lenses, for e xample, are usuall y bulky and have a short working distance, whereas lenses used for IVM are ideally slender and have a long working distance, which facilitates access to surgically exposed tissues. Notable examples are, for instance, the work of Alan Groom who used oblique transillumination of the edges of solid organs, such as the li ver and spleen, with f iberoptic light sources, to obtain high-resolution mor phological images.26,27 More recently, reflected light ob lique transillumination has been achie ved through placement of a tilted mir ror under a translucent specimen and has

been used f or time-lapse i ntravital v isualization o f neutrophil transendothelial and e xtravascular mig ration (see Figure 2).28 A special for m of bright-f ield transillumination can be achieved by epi-illumination of scattering tissue with polarized light. Some of the incident light is scattered multiple times in the tissue and e xits the tissue in a direction opposite to that of the illuminating light. Those rays that are scattered more than 10 times are depolarized and are thus able to pass to a second polarizing filter in the detection light path with orientation or thogonal to the polarization of the illumination light. On their w ay out of the tissue, these depolarized ra ys are attenuated b y structures that absorb at a chosen illumination wavelength, for instance ~550 nm for the visualization of hemo globin, thus g enerating a “ quasi-transillumination” o r b ackillumination image of red b lood cells in b lood vessels on the wide-f ield detector. This type of contrast for mation has been proven to be an effective method to visualize the microcirculation in tissues w here transillumination and use of fluorescence are not possible29 and has even been used to build devices that allow for the visualization of the microcirculation in humans in clinical settings.30 A related approach to obtain nonfluorescent image contrast is confocal reflectance microscopy, which relies on the confocal detection of light reflected b y various tissue components, such as melanin, upon epi-illumination 31 and w hich for instance yields useful image contrast at depths of up to 350 µm in human skin in vivo when using infrared light.32

Fluorescence Despite their utility in generating a rich source of morphological infor mation, the visualization techniques listed above have one critical shor tcoming. The identity of morphological objects cannot be determined with certainty but must al ways be assumed based on inferences from existing knowledge of their shape, size, optical properties, or behavior. Also, molecular events cannot be directly studied due to the limits of optical resolution. This is why fluorescence, and the ability to fluorescentl y tag biolo gical

Figure 1. History of intravital microscopy (IVM). A, Timeline of some important advances made in IVM-based research (grey panels) or technical advances relevant to IVM investigation (white panels). B, Original drawing of one of the first IVM setups, the eel chamber used around 1690 by Antonie Leeuwenhoek (left panel) and a redrawn illustration of the eel microcirculation as observed by van Leeuwenhoek (right panel). C, Original illustrations to IVM studies of the microcirculation of the frog interdigital web published in 1839 by Rudolph Wagner. Top: an individual web at 3 × resolution. Bottom: a postcapillary venule of the interdigital web at 350 × resolution (adapted from Wagner R5). D, An IVM setup from an 1878 publication by Thoma, designed to irrigate the membranous tissues of warm-blooded animals with temperature-controlled saline buffer (Adapted from Thoma R9). E, Intravital micrographs of the kidney from two different eras. Left, from the first publication using fluorescence in IVM in the toad in 192916; right, from a recent publication in 2002 using MP-IVM in the mouse.24 In both cases, epithelial nuclei were stained with an intravenously-injected intravital dye, Trypaflavin on the left, Hoechst 33342 on the right.

Intravital Microscopy

181

A

Conventional bright-field transillumination

B

C

Figure 2. Sources of image contrast. A, Enhanced visualization of extravasated leukocytes through use of phase gradients to generate image contrast. Left, a bright-field transmitted light intravital micrograph of the mouse cremaster muscle, where darker image elements result mostly from light absorption, eg, by hemoglobin in blood vessels. Right, image of a similar preparation obtained by reflected light oblique transillumination, where extravasated leukocytes are more clearly discernable due to their diffractive properties (Adapted from Mempel TR et al 28). B, From left to right, after intravenous injection, FITC-dextran (green) highlights the lumen of blood microvessels in the murine skin, while Rhodamine 6G (red) labels some intravascular leukocytes that roll along the vessel wall, as well as some endothelial cells and adventitial cells of the tissue surrounding the blood vessels. In a mouse lymph node, T and B cell areas are delineated by T (red) and B cells (green) harvested from a donor mouse and injected intravenously after fluorescent labeling with organic fluorochromes. Blood vessels are visualized through injection of a mixture of green- and red-labeled dextrans (adapted from von Andrian UH and Mempel TR116). In mice, which express major histocompatibility class II molecules fused with EGFP, endogenous dendritic cells (green) in lymph nodes can be detected by MP-IVM interacting with adoptively transferred T cells (red). In reactive lymph nodes, macrophages (yellow-white) stand out by virtue of their autofluorescence. Images were obtained by confocal or MP-IVM. Second harmonic signals from collagen fibers are shown in blue. C, Fluorescent labeling with molecular specificity. A fluorescently tagged monoclonal antibody (MECA-79) specific for a carbohydrate epitope modification was injected intravenously to identify the vascular beds in lymph nodes in which endothelial cells express the enzymatic activity required to generate this modification (adapted from M’Rini C et al 49).

182

MOLECULAR IMAGING: PRINCIPLES AND PRACTICE

structures and molecules with a high de gree of specificity, has adv anced the possibilities of intravital obser vation tremendously. Given the exponentially growing availability of fluorescent probes for the identif ication of cellular and subcellular structures, it can be assumed that this modality will also for some time in the future be dominating our efforts to understand biology through in vivo imaging, be it IVM or molecular imaging by optical means. Fluorescence is generated b y molecules containing functional chemical g roups that ef ficiently absorb electromagnetic energy through interaction with photons and can, after some of this energy is lost through nonradiative processes, again release most of this ener gy by emitting secondary photons. Due to the nonradiati ve ener gy losses, the emitted photons are, at least with single-photon e xcitation techniques, such as con ventional widefield or confocal fluorescence microscopy, of a longer wavelength than the e xcitation light and can be used to obtain spatial information about the fluorophor. In multiphoton excitation techniques, where the combined energy of se veral lo wer ener gy photons is absorbed b y the fluorophor, the emission is of a shor ter wavelength than the excitation light. Besides location, fluorescence encodes infor mation in the form of intensity (number of emitted photons per unit time), wavelength (of excitation and emitted light), and the lifetime of fluorescence, which makes it suitable for multiplexed for ms of in vestigations. The variety of fluorescent probes applicab le to IVM is enor mous and consists of organic chemical compounds, fluorescent proteins (FPs), and semiconductor nanocrystals as will be discussed below.

Nonlinear Imaging Modalities The success of multiphoton e xcitation fluorescence microscopy has brought high-ener gy, puls ed infrared lasers to man y biol ogical laboratories, including those engaged in the practice of IVM. Ho wever, the high laser peak po wers achie ved with these instr uments (se veral hundred kilowatts) not only permit multiphoton excitation but also make several other nonlinear optical contrast techniques, such as second-har monic generation (SHG), third-harmonic generation (THG) and coherent antiStokes Raman scattering (CARS) microscopy possible. The principle of SHG is dif ficult to comprehend on an intuitive basis but can be lik ened to the phenomenon of har monic overtones produced b y vibrating strings in musical instruments (Irving Bigio, personal communication). It is the exact frequency doubling that occurs when high-energy light strik es matter with cer tain str uctural features such as re gularity, noncentrosymmetr y, and

nanometer scale. 33 These conditions are fulf illed for a variety of biological macromolecules. One of these, which is abundant in tissues and has strong SHG properties, is collagen. 34 Intravital microscopists of several disciplines ha ve star ted to mak e use of this endo genous source of contrast to obtain a str uctural conte xt for the observation of dynamic cellular events35,36 or to study the dynamic alterations of collagen str ucture itself, for instance during anticancer therapy.37 THG is frequency tripling of light through interaction with refractive index inhomogeneities as found at cellular membranes. It has not been used in IVM studies y et, but Debarre and colleagues 38 have used it together with SHG and multiphoton-excited autofluorescence to image fresh rat liver explants. Curiously, they found that the strongest signals were generated by lipid bodies contained in hepatocytes. Unfortunately, the laser powers required to induce THG are even higher than those used in multiphoton excitation or SHG, w hich ma y mak e it impractical for the continuous recordings in IVM due to the in this case undesirab le simultaneous tw o- and three-photon absorption of endogenous or exogenous fluorophores.38 CARS microscop y, f inally, allo ws one to appl y the spectroscopic specif icity of Raman imaging in order to visualize the chemical constituents of a tissue by a nonlinear optical process. 39,40 Recently, this modality has been applied to IVM of the mouse ear and has enab led the distinction of various lipid species that dominate different fatcontaining compar tments of this or gan.41 CARS imaging requires significant hardware upgrades of today’s standard multiphoton microscopes, but the f act that spectroscopic imaging of protein or DN A appears to be within reach 41 may mak e such an in vestment w orthwhile for intra vital microscopists with related biological interests. It should be mentioned that all three of the listed nonlinear nonfluorescence optical processes generate signals in the forward direction of the illumination light beam. What makes them applicable to imaging in epi-detection mode, as required for IVM studies in solid organs, is that the forw ard-generated signal is par tially re versed in direction by multiple scattering events in turbid biological tissues. Since signal for mation, in principle (as in MPM), occurs onl y in the objecti ve’s focal point, these backscattered photons can be used for image for mation.

FLUORESCENT PROBES The property of fluorescence is commonl y linked to the presence of multiple aromatic g roups in a molecule, allowing for the for mation of lar ge conjugated electron systems. The number of carbon-carbon doub le bonds is

Intravital Microscopy

inversely correlated with the energy required to excite the molecule and thus positively correlated with its excitation and emission w avelength. This proper ty is shared b y a large number of or ganic compounds and proteins (see Chapters 27, “Optical Imaging Agents” and Chapter 48, “Fluorescence Readouts of Biochemistr y in Li ve Cells and Organisms”). A distinct mechanism provides fluorescence to semiconductor cr ystals,42 which w ere recentl y added to the palette of probes used for IVM.

Intravital Stains Many membrane-permeable fluorescent dyes have chemical properties that cause their accumulation in specific subcellular compar tments. Among these are acridine orange, w hich accumulates in cell nuclei, and the mitochondrial stain rhodamine 6G (see F igure 2). Both of these dyes have been used to label blood leukocytes in situ by simple intra venous bolus injection of the reagent into live animals and subsequent visualization of leuk ocyte behavior in the microcirculation by fluorescence IVM.43,44 The convenience of use of b lue fluorescent dyes in MPM has also repopularized another highl y dif fusible nuclear stain, Hoechst 33342, as an intra vital dy e to label nuclei of tissue-resident cells. 24 Propidium iodide, a nuclear dy e that onl y permeates membranes of apoptotic or necrotic cells, has been injected intra venously together with Hoechst 33342 into mice to monitor the viability of tubular epithelial cells b y MP-IVM in the kidney after ischemia and reperfusion injur y.24 Recently, probes have emerged that even allow for the specif ic visualization of tissues af fected by certain disease processes. Brad Hyman’s group has for instance validated various reagents that after intra venous injection selectively target plaques in the brains of mice suffering from a condition mimicking Alzheimer’s disease in humans. 45 Miller and colleagues ha ve used the fluorescein-based chemical 5-(and-6)-carbo xyfluorescein diacetate, succinimidyl ester (CFDA, SE, commonly simply termed CFSE) to tag antigen-presenting dendritic cells (DCs) in mice in vivo b y subcutaneousl y injecting a mixture of a model antigen, CFSE, and aluminum h ydroxide to for m a long-lasting local depot. The normally nonfluorescent, cellpermeable CFSE is g radually released from this depot and after local uptake by DC together with the antigen rendered fluorescent and char ged and thus cell imper meable by the action of intracellular esterases, in addition to binding to cellular amine groups. Migratory DC that eventually reach the draining lymph node (LN) can thus be imaged by MPM of explanted LNs. 46 This elegant approach ma y have been

183

inspired b y t he pr actice o f i ntravital m icroscopists i n neurobiology to load neurons of the central nervous (CNS) system with fluorescent dy es in vi vo through direct intracellular injection via micropipettes. 21

Functionalized Fluorochromes A second modality of assigning specificity to a fluorescent label for IVM is to attach it to a particle or macromolecule with certain biological properties or to use it to label cell lines or ex vivo purified primary cell populations for subsequent injection into animals. Classical e xamples of macromolecules tagged with fluorochromes to track their behavior after application to a live specimen are dextrans of various molecular weights and albumin to highlight either b lood plasma or l ymph compartments after intra venous or interstitial injection (see F igure 2). 47,48 Intravenous injection of conjugated molecules or unconjugated dyes can also be used to quantitatively measure vascular permeability.47 An e xtension of this principle is to use the binding specificity of antibodies to tar get fluorochromes to specific antigenic epitopes in li ve animals. Ulrich v on Andrian’s group has visualized carbohydrate moieties displayed on the luminal surf ace of specialized microvessels in lymph nodes that mediate the tethering of lymphocytes to the endothelium, which initiates their recruitment from the bloodstream into the tissue (see Figure 2). By simultaneous visualization of the b lood-borne lymphocytes, they could directly correlate the molecular endothelial staining with the biolo gical beha vior of immune cells. 49 Sipkins and colleagues 50 similarly characterized the recr uitment sites of leukemic cells to the bone mar row. One caveat of this approach is that the Fc portion of antibodies also possesses considerab le binding af finity to its receptors. Hauser and colleagues 51 therefore generated F ab fragments of an anti-CD21 antibody through papain digestion to label follicular DCs in l ymph nodes draining the subcutaneous injection site of their reagent. A modality of molecular imaging is the use of fluorochromes conjugated to molecules that confer them with specificity for tissues in dif ferent disease states either by virtue of their specific retention or by their alteration and activation through enzymatic or other molecular activities. It seems lik ely that, on the one hand , intravital microscopists will star t to use no vel tools generated in this f ield. One the other hand , molecular imaging might benefit from contributions b y IVM to their preclinical validation. In analogy to molecules, cells are also suitab le vehicles for fluorescence, allowing for the visualization of their

184

MOLECULAR IMAGING: PRINCIPLES AND PRACTICE

behavior by IVM (see F igure 2). In f act, the bur geoning field of immunoimaging b y IVM often relies on adopti ve transfer of fluorescentl y labeled immune cells, as will be discussed later in this chapter. Cell labeling mostl y relies on intracellular retention proper ties of dy es, as described earlier for CFSE, which makes this dye a useful reagent for the purpose of long-term tracking of adoptively transferred cells52; today, a wide spectrum of dyes covering the entire visible spectrum are available. Lastly, packaging fluorescence in for m of par ticles, such as microspheres or beads a vailable in sizes from a few tens of nanometers to tens of micrometers, pro vides a mean to target them to cells with phagocytic properties, such as macrophages of the spleen. 53 The same par ticle characteristics ha ve also hampered the use of quantum dots, nanometer -size semiconductor cr ystals with v ery favorable fluorescent proper ties,42 as mark ers for specific targeting reagents, and their use in IVM has been limited.54–56

Fluorescent Proteins The green fluorescent protein (GFP) was discovered more than 30 years ago,57–59 cloned in 1992,60 and first used as a marker for gene expression in 1994.61 Since then, GFP and its multicolored variants have transformed all areas of biological research b y ser ving as intracellular fluorescent reporters for visualization-based specimen anal ysis after introduction of recombinant DN A. It has also changed IVM by allowing for the nonin vasive fluorescent tagging of cell lineages or subcellular compartments (see Figure 2) or for reporters of transcriptional activity in whole animals through transgenesis or gene-tar geting approaches and through ballistic (“gene gun”), viral, electroporation, 62 or other ways of exogenous gene-delivery in vivo. Early versions of FPs, such as GFP derived from Aequorea Victoria or DsRed from Discosoma sp, were limited in their utility due to moderate fluorescence intensity, slow maturation, or the tendency to multimerize in cells. The latter proper ty for instance prohibited the transgenic expression of DsRed in rodents due to toxicity during embr yogenesis w hile slo w maturation (and thus acquisition of fluorescence) lo wers its utility as a transcriptional repor ter. Extensive mutagenesis of the wildtype proteins, ho wever, has b y no w not onl y lar gely overcome these limitations but has also g reatly extended the palette of available colors.63,64 In addition to their use to nonin vasively label endogenous cells in li ving e xperimental specimen, FPs are also adv antageous for the long-ter m tracking of rapidly proliferating, adoptively transferred tumor cells 65

or immune cells,66 as well as bacteria67,68 or other cellular pathogens b y IVM. Fluorescence le vels are maintained throughout cell di vision, w hereas or ganic cell track er dyes w ould be diluted in each generation of daughter cells, quickly rendering them nondetectable by IVM. Besides ser ving as stab ly expressed mark ers to tag cells for identif ication, FPs with shor t maturation times will be useful as in vivo reporters of transcriptional activity. Lastl y, the a vailability of s everal FP pairs (e g, BFPCFP, CFP-YFP , or GFP- RFP) with susceptibility for Forster resonance ener gy transfer (FRET) upon spatial proximity opens up the oppor tunity to use their fusions with probes for various molecular activities as molecular sensors (see Miyawaki63,69 for reviews).

Autofluorescence Some biological tissues and cells stand out in their content of v arious organic molecules with autofluorescent properties, such as serotonin, fla vins, or retinol, w hich can therefore be a useful source of contrast in fluorescence microscop y.70 The redo x sensiti vity of the coenzyme nicotinamide adenine dinucleotide (NAD+/NADH) has allo wed neurobiolo gists to study metabolic activity in the brain of living mice at cellular resolution.71 Cellular N AD+ along with N ADP+ can also be used to visualize macrophages b y MPM in explanted tumor tissue72 or in lymph nodes in vivo (see Figure 2). Since the N AD+/NADP+ redox state is also variable in these cells, their inducible changes in autofluorescence73 might in the future also provide a means to monitor their cellular activation status by IVM.

IVM INSTRUMENTATION Bright-field Microscopy Conventional IVM in its simplest for m uses brightfield transillumination of the specimen through a condenser on either an in verted or , more commonl y, an upright microscope. A con ventional halo gen lamp serves as a source of light that is focused through a collecting lens into the front aper ture of the condenser . This requires special long working distance condensers but allo ws for even illumination of the microscopic field for optimum image contrast and resolution (so-called Köhler illumination). The transmitted light is collected through long w orking distance, usuall y water-immersion, objecti ve-lenses, and detected on a charge-coupled device (CCD) camera for recording on video tape or , more and more frequentl y, on digital

Intravital Microscopy

storage media. Although primary digital data recording makes subsequent transfor mation and displa y of the data much more con venient, the often impracticab ly large data sets resulting from several minutes of recording at video-rate (25 or 30 images per second) still today make analog technology worthwhile in some situations. Although bright-f ield IVM generates a w ealth of morphological infor mation, its utility is limited to translucent tissues, and quantitati ve data can onl y be obtained in a few situations where the object under study is unambiguousl y identif ied b y mor phological criteria. This is for instance the case for leuk ocytes that tether , roll, and adhere to the luminal side of blood v essel endothelia and become easil y discer nable from other blood components as dif fractive spheres. 74 In other cases, the visualization of suf ficient mor phological detail can occasionall y be achie ved also outside b lood vessels through contrast-enhancing optic techniques 26,28 or through additional use of color as a source of contrast.75,76

Wide-field Fluorescence Microscopy For the reasons detailed earlier , the range of applications of con ventional IVM can be signif icantly e xtended through the use of fluorescence. Due to the lo w quantum yield of most fluorochromes, this requires the use of more powerful light sources, such as mercur y or x enon lamps. Dichroic mir rors and optical f ilters nar row do wn the illumination light to a spectr um suitable for e xcitation of the fluorochromes of interest w hile excluding light of the wavelengths of the e xpected flu orescence emission. Unlike with most bright-f ield techniques, the specimen is illuminated through the same objecti ve lens that also collects the emitted signals. Emitted light is typicall y f iltered to exclude excitation light and to select for specif ic fluorescence emission and then detected b y a monochrome camera. Color cameras allow for the simultaneous recording of the emission of multiple fluorochromes, bu t they ha ve ne ver gained popularity in fluorescence IVM because the y are either less sensiti ve, slo wer, or simpl y more expensive than monochrome cameras. The high light intensities required in fluorescence IVM unfortunately also command consideration of phototoxic effects in every experimental investigation, especially if continuous obser vation is emplo yed.77 One approach to reduce the risk of inducing ar tifacts through phototoxicity is to use highl y sensitive, for instance, silicone-intensified target tube or, more recently, back-illuminated electron-multipl ying CCD (EM-CCD) cameras

185

in combination with schemes to limit light exposure, such as stroboscopic illumination. Video-triggered xenon arc stroboscopes can pro vide light flashes of one microsecond duration at rates of 30/s, w hich reduces light e xposure of the specimen by a factor of 3 × 104 compared with continuous illumination.74,78 The strength of wide-f ield e xcitation fluores cence microscopy compared with the techniques discussed below is its high speed at full frame, allowing for data acquisition at video-rate or faster. Its disadvantages are the impracticality of multichannel recordings and its poor axial resolution. Nonconfocal optical sectioning techniques such as deconvolution79 or str uctured illumination 80 to impro ve axial resolution have not yet found entr y into the practice of IVM.

Laser-Scanning Microscopy Molecular and cellular mo vement in li ving specimen occurs in all three dimensions (3Ds), and acquisition of two-dimensional image data is therefore limiting in the study of solid tissues. Various modalities of laser-scanning microscop y can o vercome this prob lem. Lasers provide near -monochromatic, coherent light that can be focused to excite fluorescence in a diffraction limited spot within the focal plane of the objecti ve lens. The emitted light is usually detected by photomultiplier tubes (PMTs), and the spatial infor mation of the signal is preserved through synchronization of image re gistration with point-by-point scanning of the sample. Images are digitally rendered from the individual pixel data acquired during the laser scan. During single-photon excitation, sample fluorescence is also generated within the cones of illumination light abo ve and belo w the focal plane. To achieve confocal detection, this fraction of the fluorescence emission, along with signals fla wed b y sampleinduced light scattering in the e xcitation or in the emission path, is excluded by use of a pinhole aperture in an intermediate image plane before it reaches the d etector. During multiphoton e xcitation, as discussed belo w, extraction of full spatial infor mation in 3D is already achieved through selecti ve excitation of fluorescence in the focal plane, w hich obviates the need for descanning of the emitted light or use of a confocal pinhole aperture. Due to the sequential re gistration of image information, laser -scanning de vices are intrinsicall y inferior to wide-field excitation technology in terms of acquisition speed. Using the most frequently used galvanometer-driven xy mir ror scan-head, recording of a single full-frame (for e xample, 512 × 512 pixels) at a pixel dwell time of a fe w microseconds requires close

186

MOLECULAR IMAGING: PRINCIPLES AND PRACTICE

to a second. Various ways exist to improve scan speed, such as the use of rotating polygonic81 or resonant scan mirrors,82,83 acousto-optic deflectors, 84 or lens ar rays to scan the specimen with multiple beams. 85,86 All of these allow confocal image acquisition at video-rate or faster. Confocal microscop y has been implemented in IVM,44,87–89 but its use has not gained widespread popularity. This may be related to the limitation of optical penetration depth (typicall y less than 100 µm), w hich is greatly improved in MPM. Confocal microscopy may, however, remain the method of choice in tissues w here high concentrations of pigments or other sources of autofluorescence make multiphoton excitation problematic, such as the skin.90–92

MPM The near-simultaneous interaction of two or more lowenergy photons with a fluorophor can induce its excited state, and subsequentl y fluorescence emission, similar to w hat is achie ved with one indi vidual high-ener gy photon. This phenomenon is put to use in MPM w here conditions conducive to multiphoton-fluorophor interactions are generated b y condensing the infrared light from femtosecond-pulsed lasers in the focal point of high-numerical aperture objectives.19,93 Since photon density f alls of f quadraticall y with distance from the focal point and the efficiency of multiphoton excitation depends on the square of the photon density (for twophoton e xcitation), fluorescence f alls of f as a quar tic function outside the focal point of the microscope objective. F or practical pur poses, all e xcited fluorescence therefore originates from the focal point and can be used for image for mation, even the fraction of scattered emission, w hich w ould be e xcluded in confocal microscopy and w hich increases with increasing depth in the sample. The latter fact accounts for the high signal-to-noise ratios achie ved in MPM, especiall y in greater tissue depths. In addition, limiting fluorescence excitation to the focal point prevents significant energy absorption outside the focal plane and thus limits photo-bleaching and phototo xic ef fects. Fur thermore, the low scattering coefficient of the infrared e xcitation light typicall y used allo ws for focusing of the laser beam deep in turbid tissues. As a result, imaging depths off up to 1 mm ha ve been achie ved in the mouse brain.94 Finally, the broader and frequently overlapping two-photon cross sections of most fluorochromes (compared with their single-photon absorption spectra) make it possib le to simultaneousl y visualize multiple

fluorescent labels. This includes b lue fluorophores whose single-photon excitation otherwise requires special ultra violet light-transmissi ve optics that are not conducive to simultaneous imaging of red and f ar-red light. All of these qualities to gether mak e MPM the currently most po werful technolo gy for threedimensionally resolv ed fluorescence IVM studies in solid tissues of li ving specimen, and rapidl y ongoing technological developments guarantee that the range of possibilities will increase in the future.

Signal Detection Minimalization of light exposure should be a goal of all IVM methodology, and optimization of light detection is one way to contribute to this goal. For both CCD cameras and PMTs, ne w technolo gy is a vailable to mak e light detection more sensiti ve. Back-illuminated EM-CCD will lik ely ha ve utility for wide-f ield fluorescence IVM techniques, w hile Gallium arsenide phosphide (GaAsP)-based PMTs, w hich are more than twice as quantum-efficient as conventional bialkali or multialkali photocathodes, show promise for confocal and MPM. 95 The use of the GaAsP detectors however is still limited by their low damage threshold and their small photosensitive area (nondescanned detection of scattered emission benefits from larger detection areas). As a general r ule, CCD cameras are used in IVM to record bright-field and fluorescence wide-field microscopy images, while PMTs serve to acquire data in laser-scanning microscopes. Ho wever, a fe w e xceptions appl y. F or instance, when the specimen is scanned b y multiple beamlets generated through lens ar rays (Bewersdorf and Br uist, 1998) or a series of 50:50 beam-di viders,96,97 the spatial image information from the multitude of focal points must be preser ved through wide-f ield detection b y CCD cameras. This approach has the adv antage of faster frame rates and of the higher quantum ef ficiency of CCDs compared with PMTs. Multifocal illumination is thus an alternative to resonant scanners or AODs to achie ve video-rate image acquisition using multiphoton e xcitation85,86,96 for IVM. 97 However, the requirement for wide-f ield detection eliminates the ability to collect scattered emission light, which is the key advantage of multiphoton excitation for deep tissue imaging. Since the benef it of deeper tissue penetration of infrared light and of restricting phototo xicity and photobleaching to the focal plane remain, multifocal MPM ma y nevertheless be useful for IVM applications requiring video-rate. The gain in speed may for instance prove beneficial in the anal ysis of fluorescence lifetimes. 98 Low data sampling efficiency makes this latter imaging modality,

Intravital Microscopy

which otherwise has high potential for utility in IVM investigations, dramatically slower than fluorescence intensity measurements. Multifocal multiphoton e xcitation combined with CCD-based detection may offset this shortcoming and make the fluorescence lifetime-based observation of the dynamic events typically encountered in vivo practical, while the PMT-based detection of fluorescent lifetimes currently remains a challenge.

IVM MODELS The list of animals that have been subjected to IVM investigation is long. Cold-b looded species, such as fro g, toad, and eel, are cur rently more of historic interest, and lar ge warm-blooded v ertebrates, such as cats, do gs, rabbits, rooster, or turtles, have also lost popularity in comparison to small rodent species, which today dominate the work of most intra vital microscopists (T able 1). 74 On the other hand, a few new species have been adopted for use in IVM. Although the ph ysiology of fr uit flies, zebra f ish, or the nematode caenorhabditis ele gans is more dif ferent from ours than that of rodents, their fast-paced ontogenesis, ease of genetic manipulation, and f avorable optical characteristics have also made them attracti ve subjects for the IVMbased studies of de velopmental biologists.20,99–102 For the purpose of this chapter , several general aspects of animal preparation for IVM will be reviewed and a fe w selected rodent IVM models will be discussed in order to illustrate individual technical challenges in their design. Every obser vation by IVM requires a violation of the subject’s physical integrity, which harbors the risk of confounding the results of the in vestigation and antagonizes the ideal of IVM, the obser vation of biological processes in a tr uly physiological environment. Even the most noninvasive approaches, such as IVM of the eye or the skin, demand either some means of specimen immobilization, such as anesthesia, or optical coupling of an objective lens to the tissue with immersion medium, w hich may change the ph ysiological tissue temperature, or simpl y illumination at light energies that are not nor mally encountered at this site. The approach to tackle this problem is to take any available measures to minimize in vasiveness and to k eep in mind the v arious possibilities of inducing ar tifacts, putting controls in place wherever possible.

Anesthesia Some IVM models allow, with limitations, imaging in the awake animal. Examples are the dorsal skinfold chamber in hamsters, which can be calmed by positioning them in

187

tubes that mimic their natural ca vernous habitats, thus immobilizing them sufficiently for video-rate imaging.103 Studies in neurobiolo gy occasionally demand conscious animals, and MP-IVM in the brain of nonanesthetized freely moving rat has been achie ved through f iber-optic coupling of the microscope to a chronic brain windo w preparation.104 Generally, ho wever, IVM studies are conducted under general anesthesia of the animal for the pur pose of immobilization and to per mit surgical exteriorization of the tissue of interest. Yet, all available anesthetic reagents have not onl y neurolo gical but also cardio vascular side effects,105 and the choice of anesthetic is dictated b y the selected animal species and the anticipated minimal interference of its side effect with the experimental observations. Especiall y, studies in w hich micro vascular dynamics play a role should ideally use some form of cardiovascular monitoring to guide the narcotic re gimen. Finally, anesthetic agents may also exert uncharacterized effects o n v arious o ther p hysiological f unctions, f or instance on inflammatory processes.106,107

Surgery Except for studies of the skin and e ye, and future endoscopic approaches, IVM requires sur gical preparation of tissues to mak e them accessib le to the microscope optics, either through acute preparations or through the installation of chronic animal windows. The principle of minimal invasiveness hereby dictates the use of the least traumatic and, if feasible, aseptic technique. Membranous tissues and surfaces of inner or gans should be constantl y kept moist with appropriate buf fers that are free of contamination, for example, with bacterial to xins. The rodent cremaster muscle preparation, for instance, requires a buffer with defined pH and ionic strength since otherwise muscle fasciculations antagonize imaging efforts. The exposed brain, on the other hand, should be ir rigated with ar tificial cerebrospinal fluid for best results. Dr ying of tissues rapidl y causes microvascular dysfunction and often times ir reversible deterioration of optical transparenc y. The main concer ns with chronic animal windo w preparations are infection and e xcessive wound healing reactions, w hich both will confound not only microvascular studies but also the investigation of tumor biology or of immunological phenomena. Despite utmost care, some de gree of local tissue trauma during specimen preparation is una voidable, and immediate effects, such as acute local or e ven systemic inflammation, need to be tak en into consideration w hen interpreting the results of IVM studies. The opened cremaster muscle model, for instance, requires a comparably

188

MOLECULAR IMAGING: PRINCIPLES AND PRACTICE

Table 1. SELECTED ORGANS/TISSUES USED FOR INTRAVITAL MICROSCOPY STUDIES IN MICE Organ/Tissue

Field of Study

Microscopy Technique

Reference

Microcirculation; leukocyte migration

BF, WF, MP

Mazo and colleagues,115 Cavanagh and colleagues155

Microcirculation

WF

Carvalho-Tavares and colleagues156

Brain parenchyma

Neuronal differentiation; neuronal activity; leukocyte migration

MP

Svoboda and colleagues,21 Davalos and colleagues157

Spinal cord meninges

Microcirculation

WF

Vajkoczy and colleagues158

Microcirculation; leukocyte migration

BF, OT, WF

Mempel and colleagues,28 Hickey and colleagues,76 Baez108

Microcirculation

BF, OP, WF

Algire13

Microcirculation; tumor/tissue growth

BF, OP, WF, MP

Algire and Chalkey14

Microcirculation; leukocyte migration

RE, WF

Becker and colleagues159

Microcirculation

WF, CF

Miyamoto and colleagues160

Kidney

Microcirculation, physiological function

BF, MP

Dunn and colleagues,24 Buhrle and colleagues161

Knee joint (fat body)

Microcirculation

WF

Veihelmann and colleagues162

Liver

Microcirculation; leukocyte migration; host-pathogen interaction

BF, OT, WF, CF

Geissmann and colleagues,113 Frevert and colleagues,152 McCuskey163

Lung

Microcirculation

WF

Tabuchi and colleagues112

Microcirculation; leukocyte migration

WF, MP

von Andrian78

Popliteal LN

Microcirculation; leukocyte migration

MP

Mempel and colleagues35

Mesenteric LN

Microcirculation

WF, CF

Grayson and colleagues89

Mesentery

Microcirculation; leukocyte migration; lymph flow

BF, WF

Bienvenu and colleagues,75 Dixon and colleagues,144 Atherton and Born164

Pancreas

Microcirculation

BF, OT

Covell,165 McCuskey and Chapman166

Peyer’s patch

Microcirculation

WF

Bjerknes and colleagues167

Skin

Microcirculation; leukocyte migration

WF, CF

Kissenpfennig and colleagues,90 Nishibu and colleagues,91 Eriksson and colleagues168

Skin transplants

Microcirculation

MP

Chakraverty and colleagues169

Small intestine

Microcirculation, host-pathogen interaction

WF

Chieppa and colleagues,67 Massberg and colleagues170

Large intestine

Microcirculation

WF

Soriano and colleagues171

Spleen

Microcirculation

BF, OT, WF, CF

Grayson and colleagues,88 McCuskey and colleagues172

Tail skin

Lymph flow

WF

Leu and colleagues142

Bone marrow CNS Brain meninges

Cremaster muscle

Dorsal skinfold chamber Striated muscle Tumor/tissue implants Eye Iris Retina

Lymph node Inguinal LN

BF = bright-field; CF = confocal fluorescence; MP = multiphoton fluorescence; OPS = orthogonal polarization spectroscopy; OT = oblique transillumination; RE = reflectance; WF = wide-field fluorescence.

Intravital Microscopy

invasive procedure and considerable tissue trauma, which correlates with the skills and experience of the investigator, is ine vitable.108 In f act, for the no vice in the use of this model, it is impor tant to note that the baseline of measured parameters of inflammation, such as leukocyteendothelium interactions or micro vascular per meability, will change with increasing sur gical prof iciency and decreasing duration of the procedure. To mitigate the effects of investigator variability on the results of studies on inflammation, researchers can standardize the de gree of inflammation by preinjecting the scrotum with inflammatory mediators, such as TNFα or IL-1β, and beginning the recording of micro vascular parameters at def ined time points after this stimulus. 109

Immobilization Under the microscope, tissue mo vements o ver a fe w micrometers, caused by respiratory excursion of the chest cavity, cardiac motion, or the pulse of a local ar terial blood vessel, translate into imaging ar tifacts that can in the best case mak e visual anal ysis of the recording cumbersome and in the w orst case pre vent useful image acquisition. Video-rate imaging is hereby generally much less profoundl y af fected than 3D laser -scanning techniques, in which collection of an individual frame in the form of a stack of optical sections usuall y tak es on the order of a few seconds to minutes. The lung and the hear t pose the g reatest challenges in terms of overcoming specimen motion. At least for videorate studies of the hear t b y IVM, electrocardio gramtriggered image acquisition has been successfull y used to circumvent the necessity of immobilization.110 Lung movements, on the other hand, might be too ir regular for a similar approach, but electrom yography of intercostals muscle or the diaphragm could pro vide a useful synchronization trigger pulse. So f ar, pulmonar y physiologists have relied on the immobilizing ef fect of inter mittent inter ruption of mechanical v entilation or of adhering the obser ved lung area to a chest windo w using the ne gative pressures physiologically present in the intrapleural space. 111,112 Other or gans and tissues are also af fected b y respiratory and cardiac mo vements to a de gree inversely proportional to their distance from the chest cavity. In the case of li ver, spleen, and kidne ys, these mo vements can be par tially alle viated b y using microscopes with an inverted geometry so that the weight of the animal resting on the organ contributes to immobilization.24,88,113 Organs more distal to the tr unk are more suitab le for high-resolution 3D microscopy but still require efforts to eliminate residual micrometer-scale movements for optimal results.

189

The popliteal lymph node situated in the back of the knee can for instance be shielded from respirator y motion through the application of mechanical f ixtures to bon y pivots of the spine and the hip, 35,74 while imaging of the brain or the skull bone mar row benef its from the use of stereotactic holders.114,115

Tissue Homeostasis Apart from the inflammator y reaction to the sur gical preparation, attention should also be paid to other aspects of tissue homeostasis, depending on the biolo gical question addressed and the organ under study. Control of baseline macrocirculator y and microcirculator y parameters such as blood flow velocity is a prerequisite for an y studies on vascular biology or leukocyte-endothelium interactions, but also e xtravascular events noticeably depend on intact tissue perfusion. The interstitial mig ration of lymphocytes in l ymph nodes ceases within seconds w hen arterial blood flow is interrupted (TRM, unpublished observations). Although imaging studies in explanted lymphoid organs superfused with oxygenated buffer have so f ar yielded similar results with re gard to l ymphocyte migration and interaction as intra vital studies, it is conceivable that at some le vel intact blood flow, lymph flow, and inner vation ha ve an impact on immunolo gical processes as evidenced by alterations in the microanatomy of lymph nodes deprived of afferent lymph flow (reviewed in von Andrian and Mempel 116). Most biolo gical processes are to v arying de grees temperature dependent. F or instance, Miller and colleagues117 have found that the migration speed of naive T cells in explanted lymph nodes is maximal in the range of 36ºC but steepl y decreases at belo w 32ºC or above 42ºC. Possible causes of subphysiological temperatures in vi vo can be an anesthesia-related decrease in core body temperature or peripheral hypoperfusion. More significantly, the use of water immersion turns the objective lens into a heat sink for the tissue under obser vation. Aside from k eeping the animal core temperature in the physiological range, local temperature should therefore ideally be controlled through the use of objective heaters, heat lamps, or local sources of heat in direct contact with the specimen, such as ther mal putty or heating coils.

Pitfalls Beyond f ailure of the sur gical preparation or the anesthetic regimen, there are a fe w less ob vious f actors that can negatively affect the experimental outcome or lead to the unnoticed introduction of ar tifacts.

190

MOLECULAR IMAGING: PRINCIPLES AND PRACTICE

Effects of the, in most cases unnatural, light exposure of the tissue under study are potentiated through the presence of light absorbers, such as natural fluorochromes in some tissues, genetically encoded probes, or exogenously introduced organic fluorochromes. Well-perfused tissues efficiently dissipate heat through convection, but fluorescence e xcitation also leads to increased generation of reactive oxygen species in biolo gical tissues, and ef fects of continuous light exposure on cellular behavior are well documented.77 Some fluorescent dy es also ha ve intrinsic to xicity, Rhodamine 6G, for instance, is an inhibitor of o xidative phosphorylation,118 in addition to being mutagenic at higher doses, 119 and even some FPs can be to xic to cells due to their agg regation tendencies. FPs can also be problematic in long-ter m adopti ve transfer studies of tumor cells or immune cells due their potential immuno genicity. Some inbred mouse strains seem to be more af fected by this than others. 120 As with any scientific experiments, care must also be applied to the approaches and methodolo gies used to extract data from image material and ho w to process, analyze, and interpret the results. Some considerations are reviewed by Mempel and colleagues. 74

APPLICATIONS OF IVM Intravital microscopic study has been used in numerous fields of biomedical research. The following is a selection of areas of study , w here IVM in vestigations ha ve contributed to our understanding of biological processes, and which will serve to illustrate the range of possibilities of this methodology.

Cellular Migration and Interaction One of the classical domains of IVM is the study of the adhesive interactions between cells in the bloodstream and the endothelium of microvascular beds that are conducive to cellular recr uitment to tissues. This process usuall y occurs as a sequence of molecular interactions, and the specificity of recr uitment of the appropriate cell type to the right tissue at the right time is f acilitated through the regulated expression of signaling and adhesion receptors and their ligands on both blood and endothelial cells.121,122 Sequential molecular interactions are reflected b y different behavioral patterns of blood cells, such as initial tethering to the vessel wall, rolling, firm adhesion, spreading, crawling, and endothelial diapedesis. These steps are most characteristic for cells of the hematopoietic system, which

migrate to lymphoid and peripheral tissues as pro genitors cells115 or as mature dif ferentiated cells of the immune system.74 Similar or abbreviated recruitment cascades are observed for platelets in studies that in vestigate their role in atherosclerosis 123 or thrombus for mation,124 as well as for tumor cells in settings that mimic aspects of metastasis formation.125 While the bioph ysical conditions within blood v essels, such as the shear rate, can be par tially recreated in in vitro flow chamber systems, the special differentiation state of any endothelium is rapidly lost in culture in absence of the b lood-borne and tissue-deri ved factors. In situ study b y IVM is thus essential and has made critical contributions to the clarif ication of the molecular specif icities. The pace of cellular recr uitment to tissues is such that imaging at video-rate (or be yond) is usually most appropriate to resolv e the details of the process. The earliest studies that demonstrated the significance of specific molecular events instead of merely mechanical factors for leuk ocyte e xtravasation in vestigated the behavior of endo genous neutrophil g ranulocytes of the experimental animal using bright-f ield IVM in rodent mesenteries subjected to inflammator y s timuli.126,127 Fluorescence microscop y later allo wed for the specif ic identification of adoptively transferred, purified leukocyte subpopulations that could be either manipulated ex vivo or obtained from genetically altered donor animals in the microcirculation of recipient animals. 128 Tissues that are most amenable to this type of in vestigation are characterized b y v essel geometries that allo w visualization of sufficiently long vessel segments in the focal plane of the microscope. The poor axial resolution of con ventional bright-field and wide-f ield fluorescence microscop y is in fact of benefit in this context because it extends the depth of the microscopic field and thus the number of leukocytes passing a vessel that can be recorded. Once cells have arrested and begin to migrate on the luminal endothelium, their speed of locomotion decreases by two to three orders of magnitude, while their movement directionality transitions from nearl y onedimensional trajectories to 3D migration patterns.92,129 Intravital observation of the slow movements of any kind of cell during intralumenal crawling or during interstitial migration (which typically occurs at instantaneous velocities in the range of 1–30 µm/min) following diapedesis is more practical with time-lapse recordings and g reatly facilitated b y the 3D resolving po wer of confocal 92 or multiphoton microscopes.25 Although the high mig ratory speeds of leuk ocytes were anticipated based on in vitro studies and earlier observations of neutrophils in inflamed rodent mesentery

Intravital Microscopy

and cremaster muscle, the relentless and seemingl y random motility of l ymphocytes in secondar y lymphoid organs surprised most immunologists, but helped explain how rare antigen specif ic T and B cells can f ind the specialized cells presenting their co gnate antigen in the v ast volumes of secondar y lymphoid organs for the initiation of an immune response. 25,35,46 During their e xtravascular e xcursions, leuk ocytes physically interact and e xchange signals not onl y with other immune cells, but also with man y other tissue residents. In the case of ef fector CD8 T cells, cellular encounters with antigen-presenting tar get cells, such as tumor cells or virall y infected cells of the parench yma, triggers c ytolytic function (F igure 3). Other interactions induce developmental or behavioral modifications, such as the acti vation, proliferation, and ef fector dif ferentiation of T cells upon encounter with antigen-presenting DCs35,46,130,131 or the migration of tumor cells upon interaction with tumor-associated macrophages.132 The description of the in vi vo dynamics and r ules of cellular engagement by IVM ha ve contributed a g reat deal to our understanding of these processes.

Cellular Signaling The dynamics of cellular migration and interaction in the aforementioned studies are the net effect of a multitude of intercellular and intracellular molecular processes, and the ability to monitor these in vi vo and to correlate them to behavioral patterns will improve our ability to inter rogate the functionality of cellular networks. Neurobiologists, owing to their ref ined methods of loading environmentally sensitive fluorescent dy es into

191

individual neurons through microinjection into intact CNS tissue, were the first to measure Ca2+-fluxes in vivo as a readout of membrane depolarization during neuronal activity, triggered by a physiological stimulus, such as w hisker mo vement in the rat. 21 It took until 2006 before immunolo gists w ere ab le to measure the Ca 2+ response in B cells 133 upon their encounter of co gnate antigen on DCs in l ymph nodes in vi vo. Organic Ca 2+indicator dy es are plagued b y poor cellular retention. Genetically encoded , FP-based repor ters will in the future enable studies of Ca 2+-fluxes and of specif ic signaling pathw ays using FRET or FLIM techniques b y IVM. Ca2+ levels have already been indirectly monitored in muscle cells through IVM by using the Ca2+-sensitive proteolytic activity calpain on a FP-based FRET-probe as an indirect readout. 62

Physiologic and Pathophysiologic Functions The use of fluorescence greatly enhances the ability to not only describe but also quantitatively measure physiological processes b y IVM. Classical e xamples are studies on glomerular filtration or tubular secretion of free or conjugated fluorescent dy es in the kidne y16,24,134 or of v ascular permeability using fluorescently tagged macromolecules. 47 The same principle w as also used e xtensively to monitor transvascular and interstitial molecular transport in tumor tissue in order to refine the delivery of anticancer drugs (for a review, see Jain and colleagues135). The tumor microenvironment has also been inter rogated for hetero geneities in oxygen pressure through a phosphorescence-based IVM method.136,137 Malignant transformation of epithelial cells is B

A

Figure 3. Multiphoton intravital microscopy (MP-IVM) studies of cellular effector activity in the immune system. A, To measure the cytotoxic activity of tumor-reactive CD8+ T cells (green) in a tumor-draining lymph node at the single-cell level by MP-IVM, B cells coated with a tumor-expressed antigenic peptide were injected into the bloodstream, from where they would rapidly migrate to the draining lymph node and serve as surrogate target cells for T cells. B cells were labeled ex vivo with cytoplasmatic (Celltracker Orange, red) and nuclear organic fluorescent dyes (Hoechst 33342, blue) and changes in their fluorescent properties (loss of red signal and gain in blue signal) could be used to monitor loss of B cell structural integrity during cytotoxic T cell-induced apoptosis, along with loss of cellular function reflected by cessation of cellular motility. Yellow dots indicate B cell path. B, Time-resolved measurement of changes in B cell motility and structural integrity (“red/blue ratio”) upon encounter with a cytotoxic T cell (grey-shaded area) allows for determination of a T cell’s cytotoxic capacity. Modified from Mempel TR et al.66

192

MOLECULAR IMAGING: PRINCIPLES AND PRACTICE

accompanied by changes in their metabolic prof ile that are reflected in the fraction of free versus protein-bound NAD+. Free NAD+ has a shorter fluorescence lifetime than its protein-bound form, and MP-IVM has been used to detect precancerous lesions in hamsters. 138,139 IVM studies on microvascular perfusion have assessed parameters such as blood vessel diameter, blood flow velocity, or functional capillary density in a large number of disease states af fecting microcirculatory function. 140 Also the regulation of ar terial contractility has been subjected to imaging studies cor related with electroph ysiologic recordings to monitor the role of gap junctions in endothelial cellsmooth muscle cell communication in vivo.141 The lymphatic system has not been explored to such great detail as the blood vascular system, and in absence (until recently) of reliable histological markers, most of our knowledge of the anatomy of initial lymphatics was obtained from in vivo microlymphangiographic studies, initially in humans48 and later in mice,142 where the surprising f inding was made that tumors lack a functional lymphatic vasculature.143 High-speed video microscopy (at 500 frames per second) has recently allowed for precise determination of peak flow velocities in contractile segments of microlymphatic vessels.144

Cellular Growth and Differentiation While developmental processes occur at a f ast pace in lower animals, such as zebra fish, and can thus be studied by conventional time-lapse recordings, dif ferentiation processes in adult mammals oftentimes occur at timescales that require observation over days to weeks. The investigation of vessel formation in warm-blooded animals was the initial incentive for the development of animal chamber techniques, and the oppor tunity to study angiogenesis in the mouse, especially in response to solid tumor growth, has been the strength of the dorsal skinfold chamber model 13,14,135,145 which allo ws for the longitudinal obser vation of the same micro vascular bed over the course of weeks. Neuronal plasticity is another slo w process, and Karel Svoboda’s g roup has made use of a thinned skull preparation to monitor the changes in dendritic spines in nerve cells of the optical cortex under conditions of visual deprivation over the course of a month. 146

Host-Pathogen Interface Identifying and understanding the strate gies of pathogens use to enter a host, evade its immune defense mechanisms, and sometimes establish more or less peaceful, long-ter m

coexistence requires approaches to track the patho gens in vivo in relevant disease models. Methods to fluorescentl y tag pathogens have started to allow intravital visualization of their interaction with the host at dif ferent stages of infection (for a recent re view, see Mansson and colleagues147 and Velazquez and colleagues 148). Samel and colleagues 149 studied the kinetics of translocation of GFP-transfected , li ve E. coli bacteria through the intestinal w all in a model of bo wel obstruction, while Chieppa and colleagues67 investigated the role of epithelial DCs in the uptake of attenuated salmonella from the gut lumen. Other studies relied on the intravenous or subcutaneous injection of attenuated or inactivated patho gens to in vestigate their interaction with vascular endothelium as a putati vely relevant step in the development of bacterial sepsis 150 or with macrophages lining the floor of the subcapsular sinus in skin-draining lymph nodes as a threshold event in viral infection.151 Uta Frevert has pro vided breath-taking intra vital footage of plasmodium sporozoites transmitted through the skin of rodents by the bite of infectious mosquitoes, then using Kupffer cells as gates to exit the liver sinusoids, and eventually entering hepatocytes.152 The additional detail in our kno wledge of the lifestyle of pathogens from studies lik e this will in the future accelerate the development of therapeutic strategies to interfere with their initial entr y or their persistence. On the other hand, since our immune system has likely e volved primaril y to fend of f patho gens or to establish equilibria with commensals at our epithelial surfaces, its study in the context of infections will teach much about its function.

OUTLOOK Many advances in imaging technolo gy that will increase the range of applications of IVM are already in place and need to be adapted to the special challenges of imaging in live animals. One cur rent limitation of IVM is still in man y cases imaging depth in turbid tissues. Impro vement is in sight through the use of longer w avelength e xcitation light (beyond the range of Titanium:sapphire lasers mostly used in MP-IVM today) and of red and infrared fluorochromes. Better beam conditioning through the introduction of ne gative g roup-velocity dispersion, adapti ve w ave front sensing to correct for wave front distortion through refractive inde x inhomo geneities in biolo gical tissues, 153 and higher peak po wers achie ved with re generative amplifiers94 are also de velopments with potential benef it, but theoretically predicted limits of optical penetration at about

Intravital Microscopy

1 mm ma y remain in place. The use of needle-shaped optical lenses, so-called g radient index-lenses, which can be introduced into solid or gans without e xcessive tissue trauma ma y tur n out to be v aluable additions to IVM setups.95 Their combination with f iber-optical technology (see Chapter 14, “Dif fuse Optical Tomography and Spectroscopy”) might turn out to be a powerful tool for the minimally invasive optical access to internal organs and could also f ind application in tissues diagnosis in humans as instruments to obtain “optical biopsies” from patients. The utility of fluorescence imaging is par tly due to the possibilities of multiple xed data acquisition, and the use of spectral detectors instead of a limited number of PMTs may enhance the breadth of infor mation that can be simultaneously extracted from an IVM specimen. Finally, w e are witnessing the de velopment of more and more ref ined fluorescent probes for the inter rogation of molecular events, along with tools to influence processes under observation in vivo, for example, through photo-activated release of biologically active molecules.154 Apart from all technolo gical possibilities, ho wever, the ultimate limit to scientif ic investigation will remain the imaginati veness of the in vestigator, and it will tak e time until researchers in their respecti ve f ields will be able to ask all the biologically relevant questions that can usefully be addressed given the recent advances in intravital imaging technology.

REFERENCES 1. Malpighi M. De Pulmonibus. Bologna, Italy: Observationes Anatomicae; 1661. 2. Leeuwenhoek A. Collected letters. Edited and annotated b y a c ommittee of Dutch scientists. Amsterdam, The Netherlands: Swets and Zeitlinger; 1939–1999. 3. Hooke R. Micrographia: or some physiological descriptions of minute bodies made b y magnifying glasses with obser vations and inquiries thereupon. London: Jo. Martyn and Ja. Allestry, printers to the Royal Society; 1665. 4. Hall M. Critical and e xperimental assa y on the circulation of the blood. London: Sherwood; 1831. 5. Wagner R. Erläuter ungstafeln zur Ph ysiologie und Entwicklungsgeschichte. Leipzig, Germany: Leopold Voss; 1839. 6. Waller A. Microscopic examination of some of the principal tissues of the animal frame of the li ving frog, toad etc. Microscopic observations on the perforation of the capillaries b y the cor puseles of the b lood, and on the origin of mucus and pus-globules. Lond Edinb Dubl Phil Mag J Sci 1846;29:271–87, 379–405. 7. Cohnheim J . Über Entzündung und Eiter ung. Virchows Arch 1867;40:1–79. 8. Cohnheim J. Lectures on general patholo gy: a handbook for practitioners and students. London: The New Sydenham Society; 1889. 9. Thoma R. Über entzündliche Stör ungen des Kapillarkreislaufs bei Warmblütern. Virchows Arch 1878;74:360–93. 10. Bizzozero J . Ueber einen neuen F ormbestandteil des Blutes und dessen Rolle bei der Thrombose und der Blutgerinnung. Virchows Arch 1882;90:261–332.

193

11. Sandison JC. A ne w method for the microscopic study of li ving growing tissues b y the introduction of a transparent chamber in the rabbit’s ear. Anat Rec 1924;28:281–7. 12. Clark ER, Clark EL. Microscopic observations on the growth of blood capillaries in the living mammal. Am J Anat 1939;64:251–301. 13. Algire GH. An adaptation of the transparent chamber-technique to the mouse. J Natl Cancer Inst 1943;4:1–11. 14. Algire GH, Chalkey HW. Vascular reactions of normal and malignant tissues in vivo. I. Vascular reactions of mice to wounds and to normal and neoplastic transplants. J Natl Cancer Inst 1945;6:73–85. 15. Thiel R. Untersuchungen zum Flüssigkeitswechsel am lebenden Menschenauge. Graefes Arch Clin Exp Ophthalmol 1924;113:329–46. 16. Ellinger P, Hirt A. Mikroskopische Untersuchungen an lebenden Or ganen. I. Mitteilung. Methodik: Intra vitalmikroskopie. Zeitschrift für die gesamte Anatomie 1929;90:791–802. 17. Ries J. Kinematographie der Befruchtung und Zellteilung. Archiv für Mikroskopische Anatomie 1909;74:1–34. 18. Fulton GP, Jackson RG, Lutz BR. Cinephotomicroscop y of nor mal blood circulation in the cheek pouch of the hamster . Science 1947;105:361–2. 19. Denk W, Strickler JH, Webb WW. Two-photon laser scanning fluorescence microscopy. Science 1990;248:73–6. 20. Potter SM, Wang CM, Gar rity PA, Fraser SE. Intra vital imaging of green fluorescent protein using tw o-photon laser -scanning microscopy. Gene 1996;173:25–31. 21. Svoboda K, Denk W, Kleinfeld D, Tank DW. In vivo dendritic calcium dynamics in neocor tical p yramidal neurons. Nature 1997; 385:161–5. 22. Squirrell JM, Wokosin DL, White JG, Bavister, BD. Long-term twophoton fluorescence imaging of mammalian embr yos without compromising viability. Nat Biotechnol 1999;17:763–7. 23. Brown E, Campbell RB, Tsuzuki Y, et al. Measurement of physiological parameters in tumors in vi vo using MPLSM. Proc Soc Photo Opt Instrum Eng 2001;4262:134–45. 24. Dunn KW, Sandoval RM, K elly KJ, et al. Functional studies of the kidney of living animals using multicolor two-photon microscopy. Am J Physiol Cell Physiol 2002;283:C905–16. 25. Miller MJ, Wei SH, Cahalan MD, Parker I. Autonomous T cell trafficking examined in vivo with intravital two-photon microscopy. Proc Natl Acad Sci USA 2003;100:2604–9. 26. MacDonald IC, Ragan DM, Schmidt EE, Groom AC. Kinetics of red blood cell passage through interendothelial slits into v enous sinuses in rat spleen, anal yzed by in vivo microscopy. Microvasc Res 1987;33:118–34. 27. Schmidt EE, MacDonald IC, Groom AC. Interactions of leuk ocytes with vessel walls and with other b lood cells, studied by high-resolution intra vital video microscop y of spleen. Micro vasc Res 1990;40:99–117. 28. Mempel TR, Moser C, Hutter J , et al. Visualization of leuk ocyte transendothelial and interstitial mig ration using reflected light oblique transillumination in intra vital video microscop y. J Vasc Res 2003;40:435–41. 29. Slaaf DW, Tangelder GJ, Reneman RS, et al. A versatile incident illuminator for intra vital microscop y. Int J Microcirc Clin Exp 1987;6:391–7. 30. Groner W, Winkelman JW, Harris AG, et al. Or thogonal polarization spectral imaging: a new method for study of the microcirculation. Nat Med 1999;5:1209–12. 31. Rajadhyaksha M, Grossman M, Estero witz D, et al. In vi vo confocal scanning laser microscop y of human skin: melanin pro vides strong contrast. J Invest Dermatol 1995;104:946–52. 32. Rajadhyaksha M, Gonzalez S, Za vislan JM, et al. In vi vo confocal scanning laser microscop y of human skin II: adv ances in instr umentation and comparison with histolo gy. J In vest Der matol 1999;113:293–303. 33. Ping-Chin C, Sun CK. In: Pawley JB, editor. Handbook of biological confocal microscopy. New York City: Springer; 2006. p. 703–21.

194

MOLECULAR IMAGING: PRINCIPLES AND PRACTICE

34. Zoumi A, Yeh A, Tromberg BJ. Imaging cells and extra cellular matrix in vi vo b y using second-har monic generation and tw o-photon excited fluorescence. Proc Natl Acad Sci USA 2002;99:11014–9. 35. Mempel TR, Henrickson SE, von Andrian UH. T-cell priming by dendritic cells in lymph nodes occurs in three distinct phases. Nature 2004;427:154–9. 36. Sahai E, Wyckoff J, Philippar U, et al. Simultaneous imaging of GFP, CFP and collagen in tumors in vi vo using multiphoton microscopy. BMC Biotechnol 2005;5:14. 37. Brown E, McKee T, diTomaso E, et al. Dynamic imaging of collagen and its modulation in tumors in vivo using second-harmonic generation. Nat Med 2003;9:796–800. 38. Debarre D, Supatto W, Pena AM, et al. Imaging lipid bodies in cells and tissues using third-har monic generation microscop y. Nat Methods 2006;3:47–53. 39. Maker PD, Terhune RW. Study of optical ef fects due to an induced polarization third order in the electric f ield strength. Ph ys Re v 1965;137:A801–18. 40. Cheng J-X, Xie XS. Coherent anti-stok es raman scattering microscopy: instr umentation, theor y, and applications. J Ph ys Chem B 2004;108:827–40. 41. Evans CL, Potma EO, Puoris'haag M, et al. Chemical imaging of tissue in vivo with video-rate coherent anti-stok es raman scattering microscopy. Proc Natl Acad Sci USA 2005;102:16807–12. 42. Bruchez M Jr, Moronne M, Gin P, et al. Semiconductor nanocr ystals as fluorescent biological labels. Science 1998;281:2013–6. 43. Goldberg M, Seraf in D, Klitzman B . Quantif ication of neutrophil adhesion to sk eletal muscle v enules following ischemia-reperfusion. J Reconstr Microsurg 1990;6:267–70. 44. Villringer A, Dirnagl U, Them A, et al. Imaging of leuk ocytes within the rat brain cortex in vivo. Microvasc Res 1991;42:305–15. 45. Bacskai BJ, Hickey GA, Skoch J, et al. Four-dimensional multiphoton imaging of brain entry, amyloid binding, and clearance of an amyloid-beta ligand in transgenic mice. Proc Natl Acad Sci USA 2003;100:12462–7. 46. Miller MJ, Safrina O, Parker I, Cahalan MD. Imaging the single cell dynamics of CD4+ T cell acti vation by dendritic cells in l ymph nodes. J Exp Med 2004;200:847–56. 47. Svensjö E. Bradykinin and prostaglandin E1, E2 and F2alphainduced macromolecular leakage in the hamster cheek pouch. Prostaglandins Med 1978;1:397–410. 48. Bollinger A, Jager K, Sgier F, Seglias J. Fluorescence microlymphography. Circulation 1981;64:1195–200. 49. M’Rini C, Cheng G, Schw eitzer C, et al. A no vel endothelial L-selectin ligand acti vity in l ymph node medulla that is re gulated b y a(1,3)-fucosyltransferase-IV . J Exp Med 2003; 198:1301–12. 50. Sipkins DA, Wei X, Wu JW, et al. In vivo imaging of specialized bone marrow endothelial microdomains for tumour eng raftment. Nature 2005;435:969–73. 51. Hauser AE, Junt T, Mempel TR, et al. Def inition of ger minal-center B cell mig ration in vi vo reveals predominant intrazonal circulation patterns. Immunity 2007;26:655–67. 52. Lyons AB, Parish CR. Determination of lymphocyte division by flow cytometry. J Immunol Methods 1994;171:131–7. 53. Hoffmeister KM, F elbinger TW, Falet H, et al. The clearance mechanism of chilled blood platelets. Cell 2003;112:87–97. 54. Larson DR, Zipfel WR, Williams RM, et al. Water-soluble quantum dots for multiphoton fluorescence imaging in vi vo. Science 2003;300:1434–6. 55. Stroh M, Zimmer JP, Duda DG, et al. Quantum dots spectrally distinguish multiple species within the tumor milieu in vi vo. Nat Med 2005;11:678–82. 56. Rieger S, K ulkarni RP, Darc y D, et al. Quantum dots are po werful multipurpose vital labeling agents in zebrafish embryos. Dev Dyn 2005;234:670–81.

57. Shimomura O , Johnson FH, Saiga Y. Extraction, purif ication and properties of aequorin, a bioluminescent protein from the luminous h ydromedusan, Aequorea. J Cell Comp Ph ysiol 1962; 59:223–39. 58. Morin JG, Hastings JW. Energy transfer in a bioluminescent system. J Cell Physiol 1971;77:313–8. 59. Morise H, Shimomura O, Johnson FH, Winant J. Intermolecular energy transfer in the bioluminescent system of Aequorea. Biochemistry 1974;13:2656–62. 60. Prasher DC, Eckenrode VK, Ward WW, et al. Primary structure of the Aequorea victoria g reen-fluorescent protein. Gene 1992; 111:229–33. 61. Chalfie M, Tu Y, Euskirchen G, et al. Green fluorescent protein as a marker for gene expression. Science 1994;263:802–5. 62. Stockholm D, Bar toli M, Sillon G, et al. Imaging calpain protease activity by multiphoton FRET in li ving mice. J Mol Biol 2005; 346:215–22. 63. Miyawaki A. Inno vations in the imaging of brain functions using fluorescent proteins. Neuron 2005;48:189–99. 64. Shaner NC, Steinbach PA, Tsien RY. A guide to choosing fluorescent proteins. Nat Methods 2005;2:905–9. 65. Farina KL, Wyckoff JB, Rivera J, et al. Cell motility of tumor cells visualized in living intact primary tumors using green fluorescent protein. Cancer Res 1998;58:2528–32. 66. Mempel TR, Pittet MJ , Khazaie K, et al. Re gulatory T cells reversibly suppress c ytotoxic T cell function independent of effector differentiation. Immunity 2006;25:129–41. 67. Chieppa M, Rescigno M, Huang AY, Germain RN. Dynamic imaging of dendritic cell extension into the small bowel lumen in response to epithelial cell TLR engagement. J Exp Med 2006;203:2841–52. 68. Mansson LE, Melican K, Boek el J, et al. Real-time studies of the progression of bacterial infections and immediate tissue responses in live animals. Cell Microbiol 2007;9:413–24. 69. Miyawaki A. Visualization of the spatial and temporal dynam ics of intracellular signaling. Dev Cell 2003;4:295–305. 70. Zipfel WR, Williams RM, Christie R, et al. Li ve tissue intrinsic emission microscop y using multiphoton-e xcited nati ve fluorescence and second harmonic generation. Proc Natl Acad Sci USA 2003; 100:7075–80. 71. Takano T, Tian GF, Peng W, et al. Cortical spreading depression causes and coincides with tissue hypoxia. Nat Neurosci 2007;10:754–62. 72. Mrass P, Takano H, Ng LG, et al. Random mig ration precedes stable target cell interactions of tumor -infiltrating T cells. J Exp Med 2006;203:2749–61. 73. Kable E P, K iemer AK. N on-invasive l ive-cell m easurement o f changes in macrophage N AD(P)H b y tw o-photon microscop y. Immunol Lett 2005;96:33–8. 74. Mempel TR, Scimone ML, Mora JR, von Andrian UH. In vivo imaging of leuk ocyte traf ficking in b lood v essels and tissues. Cur r Opin Immunol 2004;16:406–17. 75. Bienvenu K, Harris N, Granger DN. Modulation of leukocyte migration in mesenteric interstitium. Am J Physiol 1994;267:H1573–7. 76. Hickey MJ, Forster M, Mitchell D, et al. L-selectin facilitates emigration and e xtravascular locomotion of leuk ocytes during acute inflammatory responses in vivo. J Immunol 2000;165:7164–70. 77. Saetzler RK, Jallo J , Lehr HA, et al. Intra vital fluorescence microscopy: impact of light-induced phototoxicity on adhesion of fluorescently labeled leuk ocytes. J Histochem Cytochem 1997; 45:505–13. 78. von Andrian UH. Intravital microscopy of the peripheral lymph node microcirculation in mice. Microcirc 1996;3:287–300. 79. Agard DA, Sedat JW . Three-dimensional architecture of a pol ytene nucleus. Nature 1983;302:676–81. 80. Neil MAA, Juskaitis R, Wilson T. Method of obtaining optical sectioning b y using str uctured light in a con ventional microscope. Opt Lett 1997;22:1905–7.

Intravital Microscopy

81. White JG, Amos WB, F ordham M. An e valuation of confocal versus conventional imaging of biological structures by fluorescence light microscopy. J Cell Biol 1987;105:41–8. 82. Tsien RY, Bacskai BJ . In: P awley JB, editor. Handbook of confocal microscopy. New York: Plenum; 1995. p. 459–78. 83. Callamaras N, Parker I. Constr uction of a confocal micro scope for real-time x-y and x-z imaging. Cell Calcium 1999;26:271–9. 84. Lechleiter JD , Lin DT , Sienear t I. Multi-photon laser scanning microscopy using an acoustic optical deflector . Bioph ys J 2002; 83:2292–9. 85. Bewersdorf J, Pick R, Hell SW. Multifocal multiphoton microscopy. Opt Lett 1998;23:655–7. 86. Buist AH, M¸ller M, Squier J , Brak enhoff GJ . Real time tw ophoton absor ption microscop y using mutli point e xcitation. J Microsc 1998;192:217–26. 87. Villringer A, Haberl RL, Dirnagl U, et al. Confocal laser microscopy to study microcirculation on the rat brain surf ace in vi vo. Brain Res 1989;504:159–60. 88. Grayson MH, Chaplin DD, Karl IE, Hotchkiss RS. Confocal fluorescent intravital microscopy of the murine spleen. J Immunol Methods 2001;256:55–63. 89. Grayson MH, Hotchkiss RS, Karl IE, et al. Intra vital microscop y comparing T lymphocyte trafficking to the spleen and the mesenteric l ymph node. Am J Ph ysiol Hear t Circ Ph ysiol 2003; 284:H2213–26. 90. Kissenpfennig A, Henri S, Dubois B, et al. Dynamics and function of Langerhans cells in vi vo: dermal dendritic cells colonize lymph node areas distinct from slo wer mig rating Langerhans cells. Immunity 2005;22:643–54. 91. Nishibu A, Ward BR, Jester JV, et al. Behavioral responses of epidermal Langerhans cells in situ to local patholo gical stimuli. J Invest Dermatol 2006;126:787–96. 92. Auffray C, Fogg D, Garfa M, et al. Monitoring of b lood vessels and tissues b y a population of monoc ytes with patrolling beha vior. Science 2007;317:666–70. 93. Zipfel WR, Williams RM, Webb WW. Nonlinear magic: multiphoton microscopy in the biosciences. Nat Biotechnol 2003;21:1369–77. 94. Theer P, Hasan MT, Denk W. Two-photon imaging to a depth of 1000 microm in living brains by use of a Ti: Al2O3 regenerative amplifier. Opt Lett 2003;28:1022–4. 95. Levene MJ, Dombeck DA, Kasischke KA, et al. In vi vo multiphoton microscopy of deep brain tissue. J Neurophysiol 2004;91:1908–12. 96. Nielsen T, Fricke M, Hellweg D, Andresen P. High efficiency beam splitter for multifocal multiphoton microscop y. J Microsc 2001; 201:368–76. 97. Niesner R, Andresen V, Neumann J , et al. The power of single and multibeam two-photon microscopy for high-resolution and highspeed deep tissue and intra vital imaging. Bioph ys J 2007; 93:2519–29. 98. Gerritsen HC, Draaijer A, van den Heuv el DJ, Agronskaia AV. In: Pawley JB, editor. Handbook of biological confocal microscopy. New York City: Springer; 2006. p. 516–34. 99. Williams D W, Truman JW. Mechanisms of dendritic elabora tion of sensory neurons in Drosophila: insights from in vi vo time lapse. J Neurosci 2004;24:1541–50. 100. Herbomel P, Thisse B, Thisse C. Onto geny and beha viour of earl y macrophages in the zebraf ish embr yo. De velopment 1999; 126:3735–45. 101. Lawson ND, Weinstein BM. In vivo imaging of embr yonic vascular development using transgenic zebraf ish. De v Biol 2002; 248:307–18. 102. Orozco JT, Wedaman KP, Signor D , et al. Mo vement of motor and cargo along cilia. Nature 1999;398:674. 103. Endrich B, Asaishi K, Gotz A, Messmer K. Technical report––a new chamber technique for micro vascular studies in unanesthetized hamsters. Res Exp Med (Berl) 1980;177:125–34.

195

104. Helmchen F , F ee MS, Tank D W, Denk W. A miniature head-mounted two-photon microscope. High-resolution brain imaging in freely moving animals. Neuron 2001;31:903–12. 105. Janssen BJ, De Celle T, Debets JJ, et al. Effects of anesthetics on systemic hemodynamics in mice. Am J Ph ysiol Hear t Circ Ph ysiol 2004;287:H1618–24. 106. Brookes ZL, Reill y CS, La wton BK, Bro wn NJ. Intravenous anesthesia inhibits leuk ocyte-endothelial interactions and e xpression of CD11b after hemor rhage. Shock 2006;25:492–9. 107. Lee HT , Kim M, Kim N , et al. Isoflurane protects against renal ischemia and reperfusion injury and modulates leukocyte infiltration in mice. Am J Physiol Renal Physiol 2007;293:F713–22. 108. Baez S. An open cremaster muscle preparation for the study of blood vessels by in vivo microscopy. Microvasc Res 1973;5:384–94. 109. Ley K, Bullard DC, Arbonés ML, et al. Sequential contrib ution of L- and P-selectin to leuk ocyte rolling in vi vo. J Exp Med 1995; 181:669–75. 110. Chilian WM, La yne SM. Coronar y micro vascular responses to reductions in perfusion pressure. Evidence for persistent ar teriolar v asomotor tone during coronar y h ypoperfusion. Circ Res 1990;66:1227–38. 111. Kuebler WM, Parthasarathi K, Linder t J, Bhattachar ya J. Real-time lung microscopy. J Appl Physiol 2007;102:1255–64. 112. Tabuchi A, Mertens M, Kuppe H, et al. Intravital microscopy of the murine pulmonary microcirculation. J Appl Physiol 2007. 113. Geissmann F, Cameron TO, Sidobre S, et al. Intra vascular immune surveillance b y CXCR6+ NKT cells patrolling li ver sinusoids. PLoS Biol 2005;3:e113. 114. Melder RJ, Salehi HA, Jain RK. Interaction of activated natural killer cells with nor mal and tumor v essels in cranial windo ws in mice. Microvasc Res 1995;50:35–44. 115. Mazo IB , Gutier rez-Ramos J-C, F renette PS, et al. Hematopoietic progenitor cell rolling in bone marrow microvessels: parallel contributions by endothelial selectins and VCAM-1. J Exp Med 1998; 188:465–74. 116. von Andrian UH, Mempel TR. Homing and cellular traffic in lymph nodes. Nat Rev Immunol 2003;3:867–78. 117. Miller MJ, Wei SH, P arker I, Cahalan MD . Two-photon imaging of lymphocyte motility and antigen response in intact l ymph node. Science 2002;296:1869–73. 118. Gear ARL. Rhodamine 6G. J Biol Chem 1974;249:3628–37. 119. Umeda M. Experimental carcino genesis: rat sarcoma produced b y the injection of rhodamine 6G. Gan 1954;45:391–3. 120. Stripecke R, Carmen Villacres M, Skelton D, et al. Immune response to green fluorescent protein: implications for gene therap y. Gene Ther 1999;6:1305–12. 121. Springer TA. Traffic signals for lymphocyte recirculation and leukocyte emigration: the multistep paradigm. Cell 1994;76:301–14. 122. Butcher EC, Pick er LJ. Lymphocyte homing and homeostasis. Science 1996;272:60–6. 123. Massberg S, Gawaz M, Gruner S, et al. A crucial role of glycoprotein VI for platelet recruitment to the injured arterial wall in vivo. J Exp Med 2003;197:41–9. 124. Denis C, Methia N , Frenette PS, et al. A mouse model of se vere von Willebrand disease: defects in hemostasis and thrombosis. Proc Natl Acad Sci USA 1998;95:9524–9. 125. Ludwig RJ, Boehme B , Podda M, et al. Endothelial P-selectin as a target of heparin action in e xperimental melanoma lung metastasis. Cancer Res 2004;64:2743–50. 126. von Andrian UH, Chambers JD, McEvoy LM, et al. Twostep model of leukocyte-endothelial cell interaction in inflammation: distinct roles for LECAM-1 and the leukocyte β2 integrins in vivo. Proc Natl Acad Sci USA 1991;88:7538–42. 127. Ley K, Gaehtgens P. Endothelial, not hemodynamic, differences are responsible for preferential leuk ocyte rolling in rat mesenteric venules. Circ Res 1991;69:1034–41.

196

MOLECULAR IMAGING: PRINCIPLES AND PRACTICE

128. Stockton BM, Cheng G, Manjunath N , et al. Ne gative re gulation of T cell homing by CD43. Immunity 1998;8:373–81. 129. Phillipson M, Heit B, Colarusso P, et al. Intraluminal cra wling of neutrophils to emig ration sites: a molecularl y distinct process from adhesion in the recruitment cascade. J Exp Med 2006;203:2569–75. 130. Shakhar G, Lindquist RL, Sk okos D, et al. Stab le T cell-dendritic cell interactions precede the development of both tolerance and immunity in vivo. Nat Immunol 2005;6:707–17. 131. Garcia Z, Pradelli E, Celli S, et al. Competition for antigen determines the stability of T cell-dendritic cell interactions during clonal expansion. Proc Natl Acad Sci USA 2007;104:4553–8. 132. Wyckoff JB , Wang Y, Lin EY , et al. Direct visualization of macrophage-assisted tumor cell intravasation in mammary tumors. Cancer Res 2007;67:2649–56. 133. Qi H, Egen JG, Huang AY, Germain RN. Extrafollicular activation of lymph node B cells b y antigen-bearing dendritic cells. Science 2006;312:1672–6. 134. Steinhausen M, Muller P, Parekh N. Renal test dyes IV. Intravital fluorescence microscop y and microphotometr y of the tub ularly secreted dye sulfonefluorescein. Pflugers Arch 1976;364:83–9. 135. Jain RK, Munn LL, Fukumura D. Dissecting tumour pathophysiology using intravital microscopy. Nat Rev Cancer 2002;2:266–76. 136. Torres F ilho IP, Intaglietta M. Micro vessel PO2 measurements b y phosphorescence decay method. Am J Physiol 1993;265:H1434–8. 137. Torres Filho IP, Leunig M, Yuan F, et al. Nonin vasive measurement of microvascular and interstitial oxygen profiles in a human tumor in SCID mice. Proc Natl Acad Sci USA 1994;91:2081–5. 138. Skala MC, Riching KM, Bird DK, et al. In vivo multiphoton fluorescence lifetime imaging of protein-bound and free nicotinamide adenine dinucleotide in normal and precancerous epithelia. J Biomed Opt 2007;12:024014. 139. Skala MC, Riching KM, Gendron-Fitzpatrick A, et al. In vivo multiphoton microscopy of N ADH and FAD redox states, fluorescence lifetimes, and cellular mor phology in precancerous epithelia. Proc Natl Acad Sci USA 2007;104:19494–9. 140. Messmer K, Krombach F. Microcirculation research in experimental surgery. Chirurg 1998;69:333–8. 141. de Wit C, Hoepfl B, Wolfle SE. Endothelial mediators and communication through vascular gap junctions. Biol Chem 2006;387:3–9. 142. Leu AJ, Berk DA, Yuan F, Jain RK. Flo w velocity in the superf icial lymphatic netw ork of the mouse tail. Am J Physiol 1994; 267:H1507–13. 143. Leu AJ, Berk D A, Lymboussaki A, et al. Absence of functional lymphatics within a murine sarcoma: a molecular and functional evaluation. Cancer Res 2000;60:4324–7. 144. Dixon JB , Za wieja DC, Gashe v AA, Cote GL. Measuring microlymphatic flo w using f ast video microscop y. J Biomed Opt 2005;10:064016. 145. Brown EB, Campbell RB , Tsuzuki Y, et al. In vi vo measure ment of gene e xpression, angio genesis and ph ysiological function in tumors using multiphoton laser scanning microscop y. Nat Med 2001;7:864–8. 146. Grutzendler J, Kasthuri N, Gan WB. Long-term dendritic spine stability in the adult cor tex. Nature 2002;420:812–6. 147. Mansson LE, Melican K, Molitoris B A, Richter-Dahlfors A. Progression of bacterial infections studied in real time––no vel perspectives provided b y multiphoton microscop y. Cell Microbiol 2007;9:2334–43. 148. Velazquez P, Waite JC, Dustin ML. Dynamics of host defense: the view at the front lines. Nat Immunol 2007;8:1153–7. 149. Samel S, K eese M, Kleczka M, et al. Microscop y of bacterial translocation during small bo wel obstr uction and ischemia in vivo––a new animal model. BMC Surg 2002;2:6. 150. Laschke MW, Kerdudou S, Her rmann M, Menger MD . Intravital fluorescence microscopy: a novel tool for the study of the interaction of Staphylococcus aureus with the micro vascular endothelium in vivo. J Infect Dis 2005;191:435–43.

151. Junt T, Moseman EA, Iannacone M, et al. Subcapsular sinus macrophages in l ymph nodes clear l ymph-borne vir uses and present them to antiviral B cells. Nature 2007;450:110–4. 152. Frevert U, Engelmann S, Zougbede S, et al. Intravital observation of Plasmodium berghei sporozoite infection of the li ver. PLoS Biol 2005;3:e192. 153. Rueckel M, Mack-Bucher JA, Denk W. Adaptive wavefront correction in tw o-photon microscopy using coherence-gated w avefront sensing. Proc Natl Acad Sci USA 2006;103:17137–42. 154. Mulligan SJ, MacVicar BA. Calcium transients in astroc yte endfeet cause cerebrovascular constrictions. Nature 2004;431:195–9. 155. Cavanagh LL, Bonasio R, Mazo IB , et al. Activation of bone marrow-resident memory T cells by circulating, antigen-bearing dendritic cells. Nat Immunol 2005;6:1029–37. 156. Carvalho-Tavares J , Hick ey MJ , Hutchison J , et al. A role for platelets and endothelial selectins in tumor necrosis factor-alphainduced leuk ocyte recr uitment in the brain micro vasculature. Circ Res 2000;87:1141–8. 157. Davalos D , Gr utzendler J , Yang G, et al. ATP mediates rapid microglial response to local brain injur y in vivo. Nat Neurosci 2005;8:752–8. 158. Vajkoczy P, Laschinger M, Engelhardt B. Alpha4-integrin-VCAM-1 binding mediates G protein-independent capture of encephalitogenic T cell blasts to CNS white matter microvessels. J Clin Invest 2001;108:557–65. 159. Becker MD, Nobiling R, Planck SR, Rosenbaum JT . Digital videoimaging of leukocyte migration in the iris: intravital microscopy in a ph ysiological model during the onset of endotoxin-induced uveitis. J Immunol Methods 2000;240:23–37. 160. Miyamoto K, Ogura Y, Hamada M, et al. In vi vo neutralization of P-selectin inhibits leuk ocyte-endothelial interactions in retinal microcirculation during ocular inflammation. Micro vasc Res 1998;55:230–40. 161. Buhrle CP, Hackenthal E, Helmchen U, et al. The hydronephrotic kidney of the mouse as a tool for intra vital microscopy and in vitro electrophysiological studies of renin-containing cells. Lab Invest 1986;54:462–72. 162. Veihelmann A, Szczesny G, Nolte D , et al. A novel model for the study of synovial microcirculation in the mouse knee joint in vivo. Res Exp Med (Berl) 1998;198:43–54. 163. McCuskey RS. A dynamic and static study of hepatic ar terioles and hepatic sphincters. Am J Anat 1966;119:455–77. 164. Atherton A, Born GVR. Quantitati ve investigation of the adhesi veness of circulating polymorphonuclear leucocytes to blood vessel walls. J Physiol 1972;222:447–74. 165. Covell WP. A microscopic study of pancreatic secretion in the living animal. Anat Rec 1928;40:213–23. 166. McCuskey RS, Chapman TM. Microscopy of the li ving pancreas in situ. Am J Anat 1969;126:395–407. 167. Bjerknes M, Cheng H, Otta way CA. Dynamics of l ymphocyteendothelial interactions in vivo. Science 1986;231:402–5. 168. Eriksson E, Boykin JV, Pittman RN. Method for in vivo microscopy of the cutaneous microcirculation of the hairless mouse ear . Microvasc Res 1980;19:374–9. 169. Chakraverty R, Cote D, Buchli J, et al. An inflammatory checkpoint regulates recr uitment of g raft-versus-host reacti ve T cells to peripheral tissues. J Exp Med 2006;203:2021–31. 170. Massberg S, Enders G, Leiderer R, et al. Platelet-endothelial cell interactions during ischemia/reperfusion: the role of P-selectin. Blood 1998;92:507–15. 171. Soriano A, Salas A, Sans M, et al. VCAM-1, but not ICAM-1 or MAdCAM-1, immunob lockade ameliorates DSSinduced colitis in mice. Lab Invest 2000;80:1541–51. 172. McCuskey RS, Meinek e HA, Townsend SF. Studies of the hemopoietic microen vironment. I. Changes in the micro vascular system and stroma during er ythropoietic regeneration and suppression in the spleens of CF1 mice. Blood 1972;39:697–712.

14 DIFFUSE OPTICAL TOMOGRAPHY SPECTROSCOPY

AND

DAVID R. BUSCH, MS AND BRITTON CHANCE, PHD, SCD (CANTAB), MD (HON)

Many for ms of molecular imaging rel y on optical techniques, often using some form of absorption, scattering, fluorescence, phosphorescence, or bioluminescence; some of these techniques are described else where in this book. Absorption and fluorescent spectroscop y are ubiquitous bench-top tools; dif fuse optics e xtends these techniques deep into scattering media, allowing quantitative measurements of chromophores se veral centimeters into biological tissues. These diffuse measurements are distinct from surface or near surface measurements, trading spatial resolution for the ability to quantify chromophore concentrations centimeters into tissue. If one is only interested in surface or near surface information, other techniques are more appropriate. Optical measurements of cells in suspension or on the surface of tissue are easil y accomplished with a microscope or camera; optical coherence tomo graphy pro vides high-resolution imaging of features less than a millimeter into the tissue. However, w hen absorbing features or fluorophores are embedded in fe w millimeters of tissue, quantif ication of chromophore concentrations requires modeling of the scattered photon path lengths; in cer tain limits, these path lengths can be described through the dif fusion equation. This chapter addresses dif fuse optics: the re gime where scattering dominates and light ener gy ef fectively diffuses away from a source. Dif fuse optics in biomedicine seeks to measure the concentrations of intrinsic chromophores (primaril y hemo globin, o xyhemoglobin, water, and f at) and contrast agents, along with scattering parameters in deep tissue (~cm), using sources and detectors on the surf ace of the skin. F igure 1 shows low (most light scattering forward) and high (light directionality lost within a few mm) scattering regimes.

The sensiti vity of dif fuse optical measurements to hemoglobin provides a f ast and nonionizing technique to monitor muscle function, brain acti vation, various types of hematoma, sub-surf ace wound healing, angio genesis, and even cognition. Work with the sole absorption/fluorescence contrast agent (indocyanine green [ICG], described below) approved for humans suggests that there is signif icant potential for both scientif ically interesting and clinicall y useful w ork with optical molecular imaging agents currently under development, especially those targeting particular disease states. These targeted agents will increase the contrast between targeted and healthy tissue, improving the ability of diffuse optics to detect the tissue of interest. This chapter focuses on applications of dif fuse optics to clinical prob lems, especiall y that of breast cancer. Clinically, advances in detection, diagnosis, and therapy monitoring for breast cancer offer much potential benef it to both suf ferers and, due to the high incidence of breast cancer ,1 society. Breast cancer has attracted much attention from the dif fuse optics community because, in addition to the clinical need, the human breast is some what easier to probe with dif fuse light than other tissues. The deformability and low optical absor ption of the breast allo w researchers to use a variety of geometries and mak e measurements through thicker tissue than is possib le with other or gans. This clinical need and reduced e xperimental difficulty may allow breast cancer to be one of the f irst clinical applications of molecular imaging agents in dif fuse optics. Leff et al.,2 in a recent review of clinical applications of diffuse optics to breast cancer , suggested that ~85% of cancers repor ted in these w orks* are detectab le using intrinsic optical contrast; Chance’ s study of o ver

*

See cited papers for inclusion criteria. 197

198

MOLECULAR IMAGING: PRINCIPLES AND PRACTICE

Figure 1. An example of scattering: low scattering (left) and high scattering (right). The high scattering example approaches the diffusive regime with an almost spherical distribution of light energy. Images were made by filling vials with water and two concentrations of scattering agent, then illuminating with a HeNe laser from the left of the figure.

100 women3 suggests sensitivity and specif icity above 90%, again with intrinsic contrasts. As Ntziachristos discusses fluorescence tomo graphy else where in this book, w e will focus mostl y on absorption tomo graphy, mentioning a fe w results especially important to clinical work. Similarly, we will not e xamine endoscopic or interoperati ve applications of dif fuse optics nor photodynamic therap y (PDT). Also, Wang discusses the related field of photoacoustic tomography else where in this book. Table 1 lists the acronyms used in this chapter.

NEAR-INFRARED DIFFUSION REGIME IN TISSUE Light propagation in a scattering, absorbing media can be described b y the dif fusion appro ximation to the radiative transport equation (RTE) under certain limits.† In tissue, these requirements are often satisf ied b y †

light appro ximately betw een 650 and 950 nm, the near-infrared (NIR) re gion of the spectr um. Other situations may require the full RTE or more careful approximations.‡ Modeling light propagation in tissue with the diffusion equation was pioneered by Patterson et al.6 in muscle; a recent tutorial b y Jacques and P ogue7 describes anal ytical, per turbative, and numerical approaches to dif fuse light transpor t and pro vides a useful introduction to the f ield. Additional references cited in this study pro vide justif ication for use of the diffusion equation for describing light transpor t and detail the associated limitations. The dif fusion re gime is def ined as the ratio of the reduced scattering and absorption coefficients ( µ sʹ′ / µ a, see definition below). Jacques and Pogue7 suggest µ ʹ′s /µ a > 20 as a good limit for use of the diffusion equation to describe light transpor t. These limits appl y to measurements of many tissues with light in the NIR: for e xample, human breast tissue ( µ a ∼ 0.05cm −1 , µ ʹ′s ∼ 10 cm −1 ) has µ ʹ′s / µ a ∼ 200 at ~800 nm. Use of the dif fusion approximation also requires measurements made se veral photon mean free paths (each ∼ 1 / µ ʹ′s ) away from the source and effectively isotropic scattering. The NIR w avelengths used in dif fuse optics are safe for frequent and long-ter m measurements as the NIR photons are nonionizing, unlik e X-ra ys. Tissue damage mechanisms are considered to be e xclusively thermal b y both American National Standards Institute8 (ANSI) and the United States F ood and Dr ug Administration 9 (FDA). Skin and e ye maximum permissible exposures (MPE) both vary by wavelength but are generally well above the power levels required to obtain useful dif fuse optical signal o ver se veral centimeters of source-detector separation; man y instruments operate at about the same power as a common laser pointer. These low power requirements allow design of instr uments for continuous optical monitoring of tissue without ph ysiological damage. This is also tr ue, within limits, for magnetic resonance imaging (MRI) and ultrasound , but dif fuse optical devices are much less e xpensive than MRI, do not require exclusion of patients with metal implants, and are a vailable at the bedside. Ultrasound pro vides the same adv antages but pro vides str uctural tissue information, whereas diffuse optics provides hemodynamic functional information. Fluorescence tomo graphy, typicall y using e xogenous agents in the NIR, introduces both additional

See, for example, the P1 approximation in Case 4 applied to the problem of neutron transport. ‡ See Klose and Hielscher 5 for an overview of optical tomography in the context of the RTE.

Diffuse Optical Tomography and Spectroscopy

Table 1. ACRONYMS AND SYMBOLS ASL

Arterial spin labeling

ANSI

American National Standards Institute

BOLD

Blood oxygenation level dependent (fMRI signal)

c

Speed of light in the relevant medium

CCD

Charge coupled device

CSF

Cerebral spinal fluid

CW

D0

Continuous wave (e.g., FD with ω = 0) c Diffusion coefficient: 3µsʹ′ [ r ] Homogeneous diffusion coefficient

δ[x]

Kronecker delta function of [x]

Δx

Change in variable x

DCS

Diffuse correlation spectroscopy

DOS

Diffuse optical spectroscopy

DOS

Diffuse optical tomography

DP

Differential path length (see DPF)

DPF

Differential path length factor

DPDW

Diffuse photon density wave

DWS

Diffusing wave spectroscopy, synonym for DCS

FD

Frequency domain

FDA

Food and Drug Administration (United States)

FDG

Fluorodeoxyglucose, an F18 PET agent

FEM

Finite element method

fDOS

Functional DOS

fMRI

Functional MRI

fNIRS

Functional NIRS

g

Anisotropy, mean cosine of scattering angle

γ

Tikhonov regularization constant. Typically denoted λ

Γ

Tikhonov regularization matrix

Gd

Gadolinium, MRI contrast agent

Gd-DTPA

Gadolinium Diethylenetriamine Penta-acetic Acid; commonly used Gd chelate for MRI contrast

Hb

Hemoglobin

HbO2

Oxyhemoglobin

Hbt

Total [Hb] = Hb + HbO2

H2O

Water

ICG

Indocyanine green

IRF

Instrument response function, used in TD measurements

λ

In this work, wavelength in nm

k0

Complex wavenumber of DPDW

LD

Laser diode

MPE

Maximum permissible exposure

MRI

Magnetic resonance imaging

µa

Absorption coefficient

µs

Scattering coefficient

D[ r ]

(Continued)

199

200

MOLECULAR IMAGING: PRINCIPLES AND PRACTICE

Table 1. (Continued) µsʹ′

Reduced scattering coefficient

Nd

Number of Detectors

NIR

Near Infrared (~650–950 nm)

NIRS

Near Infrared spectroscopy; synonym for DOS

Ns

Number of Sources

ω

Angular frequency

OI

Optical index, composite parameter to increase breast cancer contrast developed at the University of Pennsylvania

PD

Photodiode

PDT

Photodynamic therapy

PET

Positron emission tomography

PMT

Photomultiplier tube

ψ

Photon fluence

rd

Detector Position

rs

Source Position

RTE

Radiative transport equation

Sat.

Oxygen saturation = HbO2/Hbt

SNL

Sentinel lymph node

SVD

Singular value decomposition

TCSPC

Time-correlated single photon counting, used for TD measurements

TD

Time domain

TIBS

Transillumination breast spectroscopy

TOI

Tissue optical index, composite parameter to increase breast cancer contrast developed at University of California at Irvine

TRS

Time-resolved spectroscopy

information (agent distribution) and a complication, as the positions of fluorescent emitters are not known. However, the contrast betw een health y and labeled tissues can be signif icantly impro ved if the agent concentration in tar geted tissue is suf ficiently elevated compared with health y tissue (the tar get to background ratio). Autofluorescence in the NIR is quite low, fur ther reducing backg round fluorescence. These fluorescent imaging techniques ha ve been hugel y useful in small animal imaging but are presentl y not in common use for human studies due to a paucity of FDA-approved fluorescent agents. The exception, ICG, has been used in humans for decades and is described in Section “Contrast Agents for Optical Tomography.” Ntziachristos focuses on fluorescence imaging elsewhere in the book; see Weissleder10 for a comparison of diffuse optical tomo graphy (DOT) with other fluorescence modalities. *

PHOTON DIFFUSION EQUATION Experiments in this dif fusion regime require understanding of the mathematical model used to describe light transport; this section pro vides an introduction to the photon dif fusion equation, brok en do wn b y the w ay in which the light source is modulated in time. If w e consider photons to be par ticles tra versing a scattering medium, w e can describe the probability of scattering as a photon travels some distance by considering an ef fective scattering cross section and a number density of the scatterers. * This probability can be converted into an a verage path length before scattering, the mean free path ( ls). In the dif fusion equation, w e are interested in the scattering coef ficient ( µ s = ls−1 ) : the average number of scattering events per distance. If w e assume that the angular dependence of the scattering can be described solely by the angle between

In tissue, there is a distrib ution of scatterer cross sections and densities; w e will simply look at the effective average quantities.

Diffuse Optical Tomography and Spectroscopy

the incident and scattered photon paths ( θ), we can use the a verage cosine of this angle to characterize the anisotropy of this medium. We def ine an anisotrop y g = and, from this, define the reduced scattering coef ficient µsʹ′= µs (1 − g ) . The reduced scattering coefficient can be thought of as the in verse distance between ef fectively isotropic scattering e vents: 100 individual e vents with high forw ard scattering ma y have occurred, but we can lump them all together into a single, effective, scattering e vent in w hich the photon loses all memor y of its initial direction. The reduced scattering coef ficient allo ws us to use the dif fusion equation, which assumes isotropic scattering, in tissue, even though the anisotrop y of tissue ranges from ~0.7 to 0.9911 (e.g., tissue is strongly forward scattering). The distance betw een these ef fectively isotropic e vents is sometimes denoted ls* = ( µ sʹ′ )−1 and is ~1 mm in breast tissue. Similarly, w e def ine an absor ption coef ficient µa = la −1 , where la is the mean free path before absor ption. F or a single absorbing species, this def inition is equivalent to µa = C ⋅ ∈, where ∈ is the e xtinction coefficient and C is the concentration of the species. † Scattering and absorbing cross sections v ary with wavelength and therefore both µ ʹ′s and µa are wavelength dependent. The idea of calculating the probability of absorption or scattering as a function of distance tra veled o ver many individual steps is ter med a “random w alk”; this is the concept behind Monte Carlo calculations in many fields. Indeed, the diffusion equation describes the large N limit of man y disparate processes, w hich can be described as a random w alk. Chandrasekhar 12 has a classic treatment; Feynman et al.13 provided an insightful explanation of the transition betw een random walks and the diffusion equation. Ishimaru14 described scattering of w aves in random media; F ishkin and Gratton 15 and Haskell et al.16 provided useful solutions to the diffusion equation in the conte xt of dif fuse optics. Note that the accurac y of solutions w hen applied to e xperiments depends significantly on how the boundary of the medium is modeled. Hask ell et al .16 compared se veral techniques, but ongoing w ork on this topic is signif icant. Ishimaru wrote the classic text on the problem; see also the review by Jacques and Pogue.7 Arridge et al.17 provided a useful explanation of the diffusion equation and a tabulation of anal ytic solutions †

201

in various geometries. Boas et al.18 assembled a valuable review of dif fuse optics applied to medical imaging. Gibson et al19 provided a mathematically focused review on applications of diffuse optics.

Photon Diffusion Equation in the Time Domain r The dif fusion equation for photon density at a point and time t due to an isotropic point source at the origin S [ r , t ] = S0 δ[t ]δ[ r ]‡ is shown as follows6,§:

⎛ ∂ ⎞ ( [ ] ) µ [ ] − ∇ ⋅ D r ∇ + c r a ⎜⎝ ∂t ⎟⎠ Ψ d [ r , t ] = cS0 δ[t ]δ[ r ] ,

(1)

where w e ha ve def ined a dif fusion constant ⎡ cm 2 ⎤ ⎡ cm ⎤ is the speed of light in c , c D[ r ] = ⎢ ⎥ ⎢ s ⎥ 3µ ʹ′s [ r ] ⎣ s ⎦ ⎣ ⎦ the medium, µa is the absor ption coef ficient, and ⎡ W ⎤ is the dif fuse photon fluence rate. Note ψ d ⎢ 2 ⎥ ⎣ cm ⎦ that there are several conventions for the definition of the diffusion coefficient. We will use D = c as it has the 3µsʹ′ standard units (length squared per time) for this coef ficient, and se veral authors 20,21 have suggested that this formulation for D is required if one is to preser ve terms of the same order in the deri vation of Equation 1 from the RTE. The time domain (TD) solution to Equation 1 in a inf inite homo geneous dif fusing medium ( µa [r ] = µa0 , D[ r ] = D0 ) is as follows: Ψ d [r , t ] =

cS0 −3/ 2 3/2

( 4 πD0t )

e

−

r2 − µ 0 ct 4 D0t a

,

(2)

where r = r . Equation 2 describes photon propagation away from a sub-nanosecond pulsed source in an inf inite medium; Figure 2 provides a schematic of this process in a semi-infinite medium. The pulse broadening in the TD of a brief pulse of light passing through a scattering medium

See Section “Physiologic Information from DOT and Spectroscopy” for an explanation relating this description of absor ption to the Beer-Lambert Law. ‡ δ [x] denotes the Kronecker delta function for x. § Patterson cites Chandrasekhar,12 who uses somewhat different notation.

202

MOLECULAR IMAGING: PRINCIPLES AND PRACTICE

(∇

2

+k

2 0

)

cS0 Ψ AC = − δ[ r ], D0

(6)

cµa − ιω and other v ariables are D as in Section “Photon Dif fusion Equation in the Time Domain.” The solution in an infinite homogeneous media is as follows: where k = k − ιk = 0 R I

S0 Ψ AC [ r ] = 4πD0

k |r|

el I e × . |r| Damping exponential − k R |r|

(8)

Spherical wave

Figure 2. A schematic of a time domain measurement in the remission geometry. A narrow input pulse (intensity is plotted versus time, black down arrow; red schematic) is introduced into the tissue. Photons scatter off the scattering centers (blue) and are absorbed by the absorbing centers (red). Some photons will travel short distances before escaping the tissue (magenta); others will be absorbed close to the input site (cyan). Most photons will travel considerable distance from the input site (green); a few of these will be detected (black) at a site several mean free paths away from the input (black arrow up). As photons will travel many different paths to arrive at the detector, the input pulse is broadened by the time it reaches the detector.

is perhaps more intuitively obvious than the phase shift one observes in the more commonl y used frequenc y domain (FD) measurement discussed below; the TD is also convenient for Monte Carlo simulations. Most impor tantly, TD measurements allow absolute quantif ication of scattering and absorption coefficients.

Photon Diffusion Equation in the FD Writing Equation 1 in the FD brings out the close analogy to near f ield optics. Consider a source modulated at frequency ω, placed at the origin. There will be a constant (DC) and oscillating (AC) part:‡ S [ r , ω ] = δ[ r ]S0 e −ιωt + δ[ r ]SDC .

(3)

Examining onl y the oscillating par t of the detected wave: ∇( D[ r ]∇Ψ AC ) + (ιω − cµ a [ r ]) Ψ AC = − cS0δ [ r ].

(4)

In a homo geneous medium, D[r] = D 0 and the oscillating part of the photon fluence rate satisfies: ∇ 2 ΨAC +

‡

ιω − cµa cS Ψ AC = − 0 δ[ r ], D0 D0

This notation is adapted from F ishkin.15

(5)