DEPARTMENT OF RADIOLOGY SCIENTIFIC REPORT

2007

a a a a a a a a

DEPARTMENT OF RADIOLOGY

Cover image: “The tally of the year’s progress” X-ray of a late 70’s calculator

Erasmus MC

Universitair Medisch Centrum Rotterdam

CONTENTS

PREFACE

4

6

HIGHLIGHTS Relative CitationRate Continues to Rise Global Press Attention for Erasmus MC Radiology Publications Honors & Awards Conference Honors & Special Lectures Memberships & Chairs

8 9 9 10 11 11

RESEARCH GRANTS

14

RESEARCH STAFF Research Committee Senior Scientists Junior Scientists and Fellows PhD Students Master’s Students Guest Researchers Scientific Support Staff

22 23 23 24 24 26 27 27

INFRASTRUCTURE Research Committee Unit Research, Education & Training Medical Technology Information Technology (IT) Cluster 7 Contract Research Imaging Facilities

30 31 31 33 33 34 35 36

DEVELOPMENT OF ACQUISITION AND PROCESSING TECHNIQUES FOR DIAGNOSTIC IMAGING HIGH RESOLUTION & MOLECULAR IMAGING Cell Tracking by MRI In Vivo Application of Labeled Cells Monitoring and Predicting Response to Cancer Therapy Detection and Assessment of Inflammatory Disease High Resolution MRI Collaborations BIOMEDICAL IMAGING GROUP ROTTERDAM Cardiovascular Image Analysis Cellular and Molecular Image Analysis Neuro Image Analysis Oncological Image Analysis Image Guidance in Interventions Collaborations IMPLEMENTATION OF DIAGNOSTIC AND INTERVENTIONAL IMAGING TECHNOLOGY NON-INVASIVE CARDIAC IMAGING Non-Invasive Cardiac Imaging: Magnetic Resonance and Computed Tomography Magnetic Resonance Imaging of Acquired Heart Disease. Imaging of Congenital Heart Defects Collaborations NEUROLOGICAL & NEUROVASCULAR IMAGING Cerebrovascular Disease & Atherosclerotic Plaque Epidemiological Studies Functional MRI Neuroimaging in the Pediatric Patient Collaborations INTERVENTIONAL RADIOLOGY Applications of Transjugular Intrahepatic Portosystemic Shunt (TIPS) Improving Patency of Endovascular Interventions

44 46 46 48 50 51 53 64 68 68 73 76 81 84 88 90 92 92 98 101 104 108 108 113 116 119 124 128 128 130

5

contents



6

Vertebroplasty Collaborations BODY IMAGING Detection and Monitoring of Pediatric Lung Diseases Gastrointestinal and Abdominal Diseases Minimally Invasive Autopsy Breast Imaging Musculoskeletal Imaging Collaborations

132 134 136 136 139 144 146 148 152

ASSESSMENT OF DIAGNOSTIC AND INTERVENTIONAL IMAGING TECHNOLOGY Outcome Assessment of Minimally Invasive Image-Guided Therapies Methodology for the Assessment of Diagnostic Imaging Technology Assessment of Diagnostic Imaging Strategies for Cardiovascular Disease Assessment of Diagnostic Imaging for Trauma Prevention and Early Diagnosis Collaborations

156 158 160 162 163 164 166

DISSERTATIONS Arlette Odink: Epidemiologic Studies on Arterial Calcification: The Rotterdam Study Marc Kock: Diagnostic Imaging of Peripheral Arterial Disease with Multi-Detector Row Computed Tomography Angiography Jochem van den Berg: Functional Cardiovascular Assessment in Congenital Heart Disease Simone Boks: MR Imaging in Patients with Knee Injury: An Observational Study in General Practice Marco van Strijen: Diagnosing Pulmonary Embolism: Establishing and Consolidating the Role of Spiral CT Liang Wang: Magnetic Resonance Imaging of the Prostate Indra van den Bos: State-of-the-Art Magnetic Resonance Imaging in the Work-Up of Primary Hepatocellular Tumors Empar Rollano-Hijarrubia: Imaging of Small High-Density Structures in Computed Tomography Jan-Jaap Visser: Ruptured Abdominal Aortic Aneurysms: Endovascular Repair versus Open Surgery, a Decision-Analytic Approach

168 169

ABSTRACTS OF SELECTED TOP PUBLICATIONS

208

SCIENTIFIC OUTPUT Books Dissertations Chapters Peer-Reviewed Journal Articles Miscellaneous Publications

218 219 220 221 224 238

CONTRIBUTIONS TO SELECTED CONFERENCES European Congress of Radiology Joint Meeting of the International Society for Magnetic Resonance in Medicine and the European Society for Magnetic Resonance in Medicine and Biology Radiological Society of North America

240 241

COLOPHON

250

NOTES

251

244 246

172 176 181 184 191 195 199 205

7

PREFACE

Preface

The year 2007 marks my 10-year anniversary as Chairman of the Erasmus MC Department of Radiology. It has been 10 years of building: Building infrastructure, building clinical organization, building attitudes, building collaborations, and building a core of dedicated research personnel. I am extremely proud to look back at the past 10 years and realize how far our Department as a whole and especially our research program has come. Although good research remains hard work, we are now able to enjoy many of the fruits of this period of intensive building. Our output is good, with high impact; we are consistently acquiring grants to sustain our research; and we have enthusiastic, hard-work-

MRI, a movement to amend the “EU Physical Agents Directive 2004/40/EC (EMF)” to allow clinical MRI to continue. We helped found the European Institute for Biomedical Imaging Research (EIBIR), and we currently lead several of the EIBIR networks. Of course, we intend to continue and to broaden our involvement in these types of activities. 2007 saw Rotterdam playing a leading role in placing biomedical imaging in the agenda of the European Strategy Forum on Research Infrastructures (ESFRI), and we hope that the Dutch Roadmap will include a major imaging node. Reality is that this will be a huge boost to radiology research, both in the Netherlands and in Europe.

Gabriel Krestin: “I am proud of the developments within our research enterprise. The inspiration, motivation, and hard work of our researchers are bearing fruit in the form of international success.”

“Everything begins at the beginning”

ing, and respected researchers whose work is honored by prizes at international conferences and in international competitions. Furthermore, we have established formal, intensive, functional collaborations with the Erasmus MC Departments of Cardiology, Epidemiology, and Medical Informatics, as well as with the Delft University of Technology, which add expertise complementary to our own. Our relations with industry — and especially the close relationships with Siemens Medical Solutions, GE Healthcare, and Philips Medical Systems — are beneficial. During the past 10 years, Erasmus Radiology has played an increasingly important role in public policy, both on the national and on the European level. We have actively supported, and continue to support, the successful Alliance for

I would like to thank the researchers and research support personnel of our Department for their dedication and input over the years, and I hope that they are as pleased with and proud of our total accomplishments as I am. I would also like to thank our collaborators and industrial partners, who have formed an integral part of our research enterprise and supported us throughout my tenure at Erasmus MC. I am curious what the next decade will hold for us and whether we will be able to make our visions become reality, but I am sure that our staff and partners will be crucial to our success. Gabriel P. Krestin Professor and Chairman Department of Radiology Erasmus MC

Conventional X-ray of an X-ray tube. 8

9

highlights

highlights

Relative citation rate continues to rise

relative citation rate (indexed)

On the basis of information from the Institute for Scientific Information (ISI), the Center for Science and Technology Studies (CWTS) in Leiden, the Netherlands, calculates the average number of citations per publication (citation rate) for each Dutch scientific department. They also calculate the citation rate for all qualifying publications in the field (Journal Subject Category as defined by the ISI). Using these figures, they derive an indexed score, showing the impact of the department’s publications relative to that of an average global publication in the field. The analysis is

2

twice average

1

average

0

19

“In the spotlight”

97

-20

00 19

98

-20

01 19

99

-20

02 20

00

-20

03 20

01

-20

04 20

02

-20

05 20

03

-20

06

Citation rate relative to scientific field. Bars give indexed scores relating the citation rate of publications from the Erasmus MC Department of Radiology to that of all publications in the ISI subject category “Radiology, Nuclear Medicine, & Medical Imaging”. Data was analyzed in overlapping 5-year periods. A score above 1 indicates that the department’s publications are cited more frequently than an average publication in the field. More information at www.cwts.nl.

performed in overlapping 5-year blocks to average out annual fluctuations, and self-citations, Editorials and Abstracts are excluded from the analysis. The Erasmus MC Department of Radiology continues to improve both in numbers of publications and in score (see figure). For the most recent analysis period (2003-2006), we earned an index of 2.21, meaning that — on average — our publications are cited 2.21 times more often than an average global publication in our field. Combined with the publications numbers, this means not only that we are publishing more articles each year, but that the global impact of our articles is concomitantly increasing.

Global press attention for Erasmus MC Radiology publications Several of our research lines have received press attention this year and in years past, but none has received so much attention and caused so much discussion as the New England Journal of Medicine article in which Meike Vernooij and colleagues Arfan Ikram, Herve Tanghe, Arnaud Vincent, Albert Hofman, Gabriel Krestin, Wiro Niessen, Monique Breteler and Aad van der Lugt demonstrate that brain abnormalities are quite common in non-symptomatic adult volunteers.

A CT image of a spotlight post-processed with a virtual rendering technique 10

11

highlights

Honors & Awards • Timo Baks received the annual Philips Prize honoring “the best radiological PhD dissertation published in the Netherlands” for his 2006 dissertation entitled MRI and MSCT for the Assessment of Myocardial Function and Viability. This is the second consecutive year that the prize has gone to an Erasmus MC candidate. • Annick Devos was awarded the 2007 Armed Forces Institute of Pathology (AFIP) prize from the Radiological Society of the Netherlands, honoring her high score on the biennial RSN-AFIP quiz. • Gabriel Krestin and Monique Bernsen initiated and developed ENCITE – the European Network for Cell Imaging and Tracking Expertise, which was allocated €12,000,000 by the European Union. • Wiro Niessen was interviewed by the Radio 5 program Hoe?Zo! (How?This Way!) on 21 June 2007 for a segment entitled “Computer Diagnosis”, part of a series of interviews of members of the Young Academy of the Royal Netherlands Academy of Arts and Sciences.

Examples of the global press attention for the article: Meike Vernooij, Arfan Ikram, Herve Tanghe, Arnaud Vincent, Albert Hofman, Gabriel Krestin, Wiro Niessen, Monique Breteler and Aad van der Lugt. “Incidental findings on brain MRI in the general population” New England Journal of Medicine 357:1821-8 (2007)

12

• Arlette Odink was granted the 2500ste PhD from the Erasmus University for her work on arterial calcification in epidemiological research. The occasion warranted a great deal of attention in the national media.

• Marion Smits was awarded the Lourens Penning prize for the “best Dutch or English publication in Neuroradiology by a Dutch or Belgian author” for her articles on the CHIP study. The prize was awarded during the 2007 “Radiologendagen”, at which time Marion also presented a Special Lecture on the CHIP study.

• Ihor Smal received an Honorable Mention for the Conference Contribution entitled “Rao-Blackwellized marginal particle filtering for multiple object tracking in molecular bioimaging”, which was presented in Kerkrade, the Netherlands at the 20th International Conference on Information Processing in Medical Imaging.

• Harm Tiddens was awarded a WB MacDonald Visiting Professorship at the Princess Margaret Hospital for Children in Perth, Australia.

• Meike Vernooij received the Fellowship Award for her lecture “Cerebral blood flow, white matter lesion volume and cognitive function” at the 3rd Congress of the International Society for Vascular, Behavioural, and Cognitive Disorders (Vas-Cog).

Conference Honors & Special Lectures • Ylian Liem won a 1st place Lee B. Lusted Prize for the best student presentation at the 2007 Annual Meeting of the Society for Medical Decision Making. • Danielle Robbers-Visser was honored with the Young Investigator Award (Congenital) at the Society of Cardiovascular Magnetic Resonance Imaging (Rome, Italy; 1-4 February 2007) for her abstract “Cardiac MRI combined with low dose dobutamine stress reveals an abnormal stress response in children and young adults after Fontan operation at young age (J Cardiovasc MR 9:106).” Peter Pattynama and Wim Helbing were among the co-authors.

• Piotr Wielopolski, Gabriel Krestin, Monique Bernsen, and co-authors received the ICRS-Genzyme Award for Excellence in Cartilage Research 2007 for the Conference Contribution entitled “Tracking bone marrow stromal cells with iron labeling,” which was presented in Warsaw, Poland at the 7th World Congress of the International Cartilage Repair Society. • Piotr Wielopolski, Sandra van Tiel, Gavin Houston, Gabriel Krestin, Monique Bernsen, and co-authors were awarded an RSNA Certificate of Merit for their poster entitled “Molecular Imaging: Translational research on clinical MRI scanning platforms.”

13

highlights

Membership & Chairs • Monique Bernsen was appointed Technology Specialist for the Imaging Spearpoint of the multicenter SmartMix project “TeRM – Translational Regenerative Medicine”. • Johanna Bosch was elected to the Scientific Committee of the European Society of Medical Decision Making. She was also selected to participate in the Erasmus MC Female Career Development Program.

• Erik Meijering was selected as a Senior Member of the Institute for Electrical and Electronics Engineers (IEEE). Less than 10 percent of the IEEE members worldwide have achieved this recognition. • Peter Pattynama was elected President of the Union of European Medical Specialists (UEMS) Section of Radiology.

• Myriam Hunink and Wiro Niessen served on the Program Committee of the European Congress of Radiology. • Myriam Hunink additionally served on the Health Services Policy and Research Subcommittee of the Scientific Program Committee of the Radiological Society of North America (RSNA), on the Guideline Committee Minimal Head Injury of the Dutch Institute for Healthcare Improvement (CBO), and on the Committee on Radiation Protection of the Health Council of the Netherlands. She also led EuroAIM – the European network for the Assessment of Imaging in Medicine, was appointed Vice-Chairperson of the Erasmus MC Committee for Academic Promotions, and served as co-Chairperson of the Steering Committee of the Erasmus MC Consultation Center for Patient-Oriented Research.

14

15

Research grants

research grants

Lejla Alić: Netherlands Organization for Scientific Research (NWO) – Mosaiek Fellowship 2005-2009: “Quantification of tumor vessel morphology: a tool to monitor treatment” Monique Bernsen and Ingrid Renes (Neonatology): Erasmus MC Translational Research Seed Grant 2005-2007: “Improved detection accuracy of early IBD lesions with MRI through local contrast enhancement using lipid-based nanocarriers”

Johanna Bosch and Marc van Sambeek (Vascular Surgery): Dutch Society for Body and Life 2007-2011: “Differences between men and women in presentation of vascular disease and differences in treatment outcome” Monique Breteler (Epidemiology), Aad van der Lugt and Peter Koudstaal (Neurology): Erasmus MC Research Grant 2006-2010: “Clinical relevance of cerebral microbleeds”

Linda Everse: “Our research funding continues to increase each year, and we are increasing our participation in European consortia. With these developments, the presence of a unit dedicated to supporting researchers in the administrative and legal burdens associated with external funding becomes more a necessity than a luxury.” Johanna Bosch, Marc van Sambeek (Vascular Surgery), Lukas van Dijk, and Myriam Hunink: Erasmus MC Health Care Efficiency Research 2005-2007: “Cost-effectiveness of endoprosthetic management of acute abdominal aortic aneurysms.”

“Money, money, money ...”

Johanna Bosch and Marc van Sambeek (Vascular Surgery): Dutch Society for Body and Life 2005-2008: “Endovascular therapy or conventional open surgery for the repair of acute abdominal aortic aneurysm”

Marleen de Bruijne: Netherlands Organization for Scientific Research (NWO) – VENI Grant 2007-2010: “Robust multi-object segmentation” Kees van Dijke, Timo ten Hagen (Surgical Oncology) and Jifke Veenland: Erasmus MC Translational Research Grant 2003-2007: “Intravital microscopy and tumor drug uptake studies to assess the predictive value of a clinically applicable indicator in correlation to dynamic contrast-enhanced MRI”

A CT image of coins post-processed with a virtual rendering technique 16

17

research grants

Kees van Dijke, Timo ten Hagen (Surgical Oncology) and Jifke Veenland: Erasmus MC Translational Research Seed Grant 2005-2007: “Early treatment response evaluation using MRI to test the suitability of therapy” Pim de Feyter and Myriam Hunink: ZonMW Health Technology Assessment Grant 2004-2007: “MSCT coronary angiography in patients with stable and unstable angina: A multicenter study” Pim de Feyter, Jeroen Bax (LUMC), & Maarten-Jan Cramer (UMC): Netherlands Heart Foundation Large Grant Program 20072012: “Modification of risk and management with MSCT coronary imaging in high risk asymptomatic individuals” Alejandro Frangi (Universitat Pompeu Fabra), Johan van der Lei (Medical Informatics), Miriam Sturkenboom (Medical Informatics), Wiro Niessen, Aad van der Lugt, and consortium partners: European Union Integrated Project Grant 2006-2009: “@ neurist: Integrated Biomedical Informatics for the Management of Cerebral Aneurysms” Robert-Jan van Geuns: Pie Medical Imaging Research Grant 2006-2008: “Quantitative analysis of magnetic resonance images for left ventricular function, myocardial perfusion and viability evaluation”

18

Wim Helbing & Dennis Dooijes (Clinical Genetics): Sophia Children’s Hospital Foundation / Wilhelmina Children’s Hospital Foundation 2005-2008: “Hemodynamic and genetic factors in longterm outcome of Fontan circulation”

Johannes Jeekel (Surgery), Johanna Bosch and Myriam Hunink: ZonMW Health Care Efficiency Research Grant 2005-2008: “Inguinal hernia: Operation or observation?”

Gabriel Krestin (ESR), Linda Everse, and consortium partners: European Union Network of Excellence Grant 2007-2009: “EIBIR: European Institute for Biomedical Imaging Research”

Wim Helbing & Barbara JM Mulder (AMC): Netherlands Heart Foundation 2006-2010: “Early diagnosis of right ventricular dysfunction in patients operated for tetralogy of Fallot: A multicenter study with serial follow-up”

Marion de Jong (Nuclear Medicine), Jifke Veenland, Monique Bernsen, Wiro Niessen, and Gabriel Krestin: Netherlands Cancer Society Grant 2008-2012: “Functional imaging of tumor uptake and therapy response: Combining CT, SPECT and MRI to monitor Peptide Receptor Radionuclide Therapy”

Gabriel Krestin, Monique Bernsen, Wiro Niessen, Erik Meijering, and consortium partners: European Union Large Scale Integrating Project 2008-2012: “ENCITE – the European Network for Cell Imaging and Tracking Expertise”

Myriam Hunink: ZonMW Health Technology Assessment Grant 2004-2007: “Setting research priorities, optimizing study design, and guiding the choice of relevant outcome measures in the evaluation of diagnostic (imaging) tests” Myriam Hunink: Dutch Health Care Insurance Board Technology Assessment Grant 2005-2007: “Cost-effectiveness of decision rules for the use of CT for minimal head injury” Myriam Hunink: ZonMW Prevention Grant 2007-2009: “Cost-effectiveness of CT screening to identify individuals at risk for cardiovascular events: a computer simulation study” Myriam Hunink: Erasmus MC Health Care Efficiency Research Grant 20082011: “Cost-effectiveness of diagnostic imaging strategies for the diagnosis of coronary artery disease in patients with new onset chest pain suggestive of stable angina pectoris”

Marion de Jong (Nuclear Medicine), Jifke Veenland, and Monique Bernsen: Erasmus MC Research Grant 2008-2012: “Multimodal molecular tumor imaging: Combining SPECT, CT and MRI” Willi Kalender (University of Erlangen), Gabriel Krestin, Winnifred van Lankeren, Inge-Marie Obdeijn, Marcel van Straten, and consortium partners: European Union Collaborative Project 2008-2011: “Dedicated CT of the Female Breast: Feasibility, optimisation and comparison to competing imaging modalities” Gabriel Krestin: GE Healthcare Contract Research Agreement 2004-2009: “Use of 3 Tesla MR in molecular, cardiovascular, neurological, and abdominal imaging” Gabriel Krestin: Siemens Medical Solutions Grant: “Coronary Angiography using State-of-the-Art CT Imaging”

Boudewijn Lelieveldt (LUMC/DUT), Erik Meijering, and Monique Bernsen: Medical Delta – Health Science & Technology Research Grant 2008-2012: “Molecular image processing from cell to organism: An integrative approach” Aad van der Lugt: Netherlands Organization for Scientific Research (NWO) – Clinical Fellowship 2004-2008: “Detection of vulnerable plaque in the carotid artery with multislice CT” Aad van der Lugt: Schering AG Clinical Research Grant 2006-2007: “Improved characterization of atherosclerotic plaque in the carotid artery with contrast-enhanced (MS-325), high resolution 3T MRI” Aad van der Lugt and Diederik Dippel (Neurology): Netherlands Heart Foundation 2008-2012: “Serial CT Angiography of atherosclerotic carotid plaque: determinants and prognosis of change in volume, composition and morphology”

19

research grants

Folkert J Meijboom (Cardiology) & Wim Helbing: ZonMW Health Care Efficiency Research Grant 2007-2009: “Comparison of 3D echocardiography and MRI for assessment of right ventricular function in congenital heart disease”

Wiro Niessen: Netherlands Organization for Scientific Research (NWO) – VICI Grant 2006-2010: “3D multimodal vascular image analysis for improved diagnosis and therapy”

Peter Pattynama: Erasmus MC Health Care Efficiency Research Seed Grant 2006-2008: “TIPS with covered stentgrafts versus endoscopic treatment of acute bleeding of esophageal varices”

Erik Meijering: Netherlands Organization for Scientific Research (NWO) – VIDI Grant 2005-2010: “Model-driven spatiotemporal tracking for quantitative analysis of subcellular dynamics”

Wiro Niessen and consortium partners: Dutch Ministry of Economics ‘Peaks in the Delta’ Program 2008-2012: “The heart in three dimensions”

Peter Pattynama, Johanna Bosch, and Myriam Hunink: ZonMW Health Care Efficiency Research Grant 2006-2010: “Transjugular Intrahepatic Porto-systemic Shunt (TIPS) with Gore-tex covered stent-graft versus endoscopic treatment for acute bleeding of esophageal varices”

Nico Mollet: Nuts-OHRA Grant 2006-2009: “Early detection of coronary artery disease by non-invasive, multislice CT coronary imaging in asymptomatic, high-risk individuals” Nico Mollet: Erasmus University Rotterdam (EUR) Fellowship 2006-2010: “Early detection of coronary artery disease by non-invasive, multislice CT coronary imaging in asymptomatic, high-risk individuals” Wiro Niessen: Netherlands Organization for Scientific Research (NWO) Exact Sciences Open Competition 2005- 2009: “MOSAIC: Multiscale Modelling of Object Shape and Appearance for Analyzing 3D Image Content” Wiro Niessen, Monique Breteler (Epidemiology), Aad van der Lugt, and Henri Vrooman: Erasmus MC Seed Grant 2006-2008: “Automated analysis of diffusion tensor MR brain data”

20

Wiro Niessen, Miriam Sturkenboom (Medical Informatics), Aad van der Lugt, Diederik Dippel (Neurology), and Henk Bijvoet (Neurosurgery): Erasmus MC Seed Grant 2008-2009: “4D imaging of brain aneurysms: A potential hemodynamic biomarker for aneurysmal growth and rupture” Louis Nguyen (Harvard Medical School) and Myriam Hunink: National Heart, Lung, and Blood Institute K23 Award 20062010: “Percutaneous treatment of claudication: The role of diabetes and inflammation” Eugenio Parati (National Neurological Institute “Carlo Besta”), Wiro Niessen, Aad van der Lugt, and consortium partners: European Union Specific Targeted Research Grant 20062008: “Neuroweb – Integration and sharing of information and knowledge in neurology and neurosciences”

Hans Reiber (LUMC), Rik Stokking, Aad van der Lugt, and consortium partners: Dutch Ministry of Economic Affairs – SENTER Grant 20052009: “Automatic Diagnostic Vascular Analysis of CTA Examinations (ADVAnCE)” Pina Sanelli (Weill Medical College) and Myriam Hunink: Neuroradiology Education and Research Foundation Award 2007-2008: “CT perfusion in aneurysmal subarachnoid hemorrhage patients” Han van Schie (Ortho) Harrie Weinans (Ortho), Wiro Niessen, and Abida Ginai: Erasmus MC Seed Grant 2006-2008: “Imaging of tendinopathy of the Achilles tendon” Marion Smits: ISMRM Educational Stipend 2007: “A presentation at the ISMRM 2007”

Marion Smits: Netherlands Organization for Scientific Research Van Walree Travel Award 2007: “Three presentations at the ECR 2007” Harm Tiddens: Roche Netherlands Research Grant 2005-2008: “Optimalisation of DNase treament in CF” Harm Tiddens: Australian Cystic Fibrosis Foundation Seed Grant 2006-2010: “Prediction model for survival in CF with Severe Advanced Lung Disease (SALD) using computed tomography” Harm Tiddens: Cystic Fibrosis Foundation Clinical Research Award: Award 2006-2011: “Visiting Professorship University of Washington” Harm Tiddens: Dutch Cystic Fibrosis Foundation Research Grant 20072009: “Lung abnormalities and therapy of CF mouse” Harm Tiddens: Italian Cystic Fibrosis Foundation Seed Grant 2007-2010: “Prediction model for survival in CF with Severe Advanced Lung Disease (SALD) using computed tomography” Meike Vernooij: ISMRM Educational Stipend 2007: “Three presentations at the ISMRM 2007”

21

research grants

Harrie Weinans (Orthopedics), Gerjo van Osch (Orthopedics), Jan Verhaar (Orthopedics), Koen Bos (Orthopedics), Holger Jahr (Orthopedics), Monique Bernsen, and consortium partners: Dutch Ministries of Economics and of Education, Culture & Science – Smart Mix Grant: “TeRM – Translational Regenerative Medicine”

Jacqueline Witteman (Epidemiology), Albert Hofman (Epidemiology), Erik Duckers (Cardiology), Caroline Cheng (Cardiology), Patrick Serruys (Cardiology), André Uitterlinden (Internal Medicine), Aad van der Lugt, Gabriel Krestin, Dirk Duncker (Cardiology): Erasmus MC Research Grant 2008-2012: “A genetic approach to enhance pathophysiologic understanding of plaque formation and plaque vulnerability”

Jolanda Wentzel (Cardiology), Cornelis Slager (Biomedical Engineering), and Aad van der Lugt: Dutch Interuniversity Cardiological Institute / Royal Dutch Academy of Sciences (ICIN/KNAW) Project Grant 20052009: “High sheer stress contributes to plaque rupture by spatially restricted endothelial anti-inflammatory signaling” Jacqueline Witteman (Epidemiology), Aad van der Lugt and Gabriel Krestin: Netherlands Heart Foundation 2004-2007: “Predictive value of coronary, carotid and aortic arch calcification for risk of coronary heart disease and stroke” Jacqueline Witteman (Epidemiology), Aad van der Lugt and Gabriel Krestin: Netherlands Heart Foundation 2006-2010: “Predictive value of vulnerable plaques in the carotid arteries for risk of coronary heart disease and stroke”

22

23

research staff

research staff 2007

Research Committee Linda A Everse, PhD – Research Coordinator Pim J de Feyter, MD, PhD – Professor of Non-Invasive Cardiac Imaging Myriam GM Hunink, MD, PhD – Professor of Clinical Epidemiology and Radiology Gabriel P Krestin, MD, PhD – Professor of Radiology, Chairman of the Department Aad van der Lugt, MD, PhD – Associate Professor of Neuroradiology Wiro J Niessen, PhD – Professor of Medical Image Processing Peter MT Pattynama, MD, PhD – Professor of Interventional Radiology

Senior Scientists

“on call 24/7”

Monique R Bernsen, PhD – Research Associate Cell Biology & Molecular Imaging Johanna L Bosch, PhD – Assistant Professor of Clinical Epidemiology and Radiology Marleen de Bruijne, PhD – Assistant Professor of Medical Image Processing in Radiology Filippo Cademartiri, MD, PhD – Research Associate University Hospital Parma, Italy Jos N van der Geest, PhD – Research Associate Neurosciences Abida Z Ginai, MD, PhD, FRCR – Associate Professor of Radiology

Robert Jan M van Geuns, MD, PhD – Research Associate Cardiology Wim A Helbing, MD, PhD – Professor of Pediatric Cardiology Maarten H Lequin, MD, PhD – Radiologist HW (Erik) Meijering, PhD – Assistant Professor of Medical Image Processing Harm AWM Tiddens, MD, PhD – Pediatric Pulmonologist Jifke F Veenland, PhD – Assistant Professor of Medical Image Processing Albert M Vossepoel, PhD – Professor of Medical Image Processing Henri A Vrooman, PhD – Assistant Professor of Medical Image Processing Theo van Walsum, PhD – Assistant Professor of Medical Image Processing Piotr A Wielopolski, PhD – Research Associate MRI

A CT image of a beeper post-processed with a virtual rendering technique 24

25

research staff

Junior Scientists and Fellows Timo Baks, MD, PhD – Research Associate Cardiology Lukas C van Dijk, MD, PhD – Radiologist Ineke J Hartmann, MD, PhD – Radiologist Majanka H Heijenbrok-Kal, PhD – Post-doc John J Hermans, MD – Radiologist Ashis Jalote Parmar, MDes – Research Fellow TU Delft Maka Kekelidze, MD – Research Fellow Winnifred van Lankeren, MD, PhD – Radiologist Edwin van der Linden, MD – Radiologist Rashindra Manniesing, PhD – Post-doc Adriaan Moelker, MD, PhD – Resident in Radiology Amber D Moelker, PhD – Post-doc Nico RA Mollet, MD, PhD – Resident in Radiology Galied SR Muradin, MD, PhD – Radiologist Koen Nieman, MD, PhD – Cardiologist Jeroen J Nikken, MD, PhD – Radiologist A Inge-Marie Obdeijn, MD – Radiologist Rody Ouwendijk, MD, PhD – Resident in Radiology Niels van Pelt, MD – Research Fellow Cardiology

26

Nomdo Renken, MD – Research Fellow Reinier de Graaf Hospital Marcel van Straten, PhD – CT Physicist Hervé LJ Tanghe, MD – Radiologist François EJA Willemssen, MD – Radiologist

PhD Students Lejla Alić, MSc (Medical Informatics) – PhD expected 2009 Nóra Baka, MSc (Medical Informatics) – PhD expected 2011 Ineke van den Berg, MSc (Epidemiology) – PhD expected 2010 W Jochem BW van den Berg, MD (Pediatric Cardiology) – completed PhD in 2007 Renske de Boer (Epidemiology) – PhD expected 2010 Simone Boks, MD (Clara Hospital) – completed PhD in 2007 Indra C van den Bos, MD – completed PhD in 2007 Quirijn van den Bouwhuijsen, MD (Epidemiology) – PhD expected 2011 Gerben M van Buul, MSc (Orthopedics) – PhD expected 2012 Chris A Cocosco, MEng (Medical Informatics) – PhD expected 2009 Jeroen Dudink, MD (Neonatology) – PhD expected 2009

Soender (Roy) Dwarkasing, MD – PhD expected 2008 Oleh Dzyubachyk (Medical Informatics) – PhD expected 2010 Suzette E Elias-Smale, MD – PhD expected 2010 Bas Ferket (Epidemiology) – PhD expected 2010 Azadeh Firouzian, MEng (Medical Informatics) – PhD expected 2011 H Zwenneke Flach, MD – PhD expected 2009 Remy WF Geenen, MD – PhD expected 2009 Alina G van der Giessen, MEng (Thorax Center) – PhD expected 2009 Harald C Groen (Biomedical Engineering) – PhD expected 2010 Bas Groot Koerkamp, MD (Epidemiology) – PhD expected 2008 Nathalie Grootenboer (Epidemiology) – PhD expected 2011 Joke Hendriks, MD (Vascular Surgery) – PhD expected 2009 Philip J Homburg, MD – PhD expected 2010 Alfonso Isola (Medical Informatics) – PhD expected 2010 Yusef Karamermer, MD (Cardiology) – PhD expected 2012 Sharon WM Kirschbaum, MD (Cardiology) – PhD expected 2010 Marc CJM Kock, MD – completed PhD in 2007 Ylian S Liem, MD (Epidemiology) – completed PhD in 2008 Fedde van der Lijn (Medical Informatics) – PhD expected 2009 Martine Loeve, MD (Pediatrics) – PhD expected 2010

Saskia Luijnenburg, MD (Pediatric Cardiology) – PhD expected 2011 W Bob Meijboom, MD (Cardiology) – PhD expected 2008 Coert T Metz (Medical Informatics) – PhD expected 2010 Carlos AG van Mieghem, MD (Cardiology) – PhD expected 2009 Cecilé de Monyé, MD – PhD expected 2009 Lisan AE Neefjes, MD (Cardiology) – PhD expected 2011 Arlette E Odink, MD (Epidemiology) – completed PhD in 2007 Edwin HG Oei, MD – PhD expected 2009 Mohamed Ouhlous, MD – PhD expected 2008 Marielle MF Poels, (Epidemiology) – PhD expected 2012 Francesca Pugliese, MD – PhD expected 2008

27

research staff

Daniëlle Robbers-Visser, MD (Pediatric Cardiology) – PhD expected 2009 M Empar Rollano-Hijarrubia, MSc (Medical Informatics) – completed PhD in 2007 Sietske Rozie, MD – PhD expected 2010 Michiel Schaap (Medical Informatics) – PhD expected 2010 Caroline KL Schraa-Tam, MSc, MEng – PhD expected 2009 Ihor Smal, MSc (Medical Informatics) – PhD expected 2009 Marion Smits, MD – completed PhD in 2008 Marco JL van Strijen, MD (UMC Leiden) – completed PhD in 2007 Tirza Springeling, MD (Cardiology) – PhD expected 2011 Sandra Spronk, MSc (Ikazia Hospital) – completed PhD in 2008 Renate MC Swarte, MD (Neonatology) – PhD expected 2009 Maarten GJ Thomeer, MD – PhD expected 2009 Meike W Vernooij, MD (Epidemiology) – PhD expected 2008 Jan Jaap Visser (Epidemiology, Surgery) – completed PhD in 2007 Marion van Vliet, MD – PhD expected 2009 Sonia Volpi, MD (Pediatrics) – PhD expected 2012 Danijela Vukadinovic (Medical Informatics) – PhD expected 2010 Liang Wang, MD (Memorial Sloan-Kettering Cancer Center) – completed PhD in 2007 Thomas T de Weert, MD – PhD expected 2009

28

Annick C Weustink, MD – PhD expected 2009 Alexandra Wils, MD – PhD expected 2010

Aletta Tholen (Epidemiology) – MSc expected 2009 Evert FS van Velsen (Medical Informatics) – completed MSc in 2006 Arthur R Wijsmuller (Surgery) – MSc expected 2009

Master’s Students

Guest Researchers

Liliane Caldeira (Medical Informatics) – completed MSc in 2007 Farzin Fakhry (Epidemiology) – MSc expected 2009 Tessa Genders (Epidemiology) – MSc expected 2008 Marius de Groot (Medical Informatics) – MSc expected 2009 B Els van der Ham (Epidemiology) – MSc expected 2008 Marcus de Jong (Epidemiology) – MSc expected 2009 Guido C Kaandorp (Epidemiology) – MSc expected 2008 Gert Jan R ten Kate (Cardiology) – completed MSc in 2007 Bob van Kempen (Epidemiology) – MSc expected 2009 Balinder Paul (Medical Informatics) – MSc expected 2008 Fabian van der Sluis (Surgery) – MSc expected 2009 Tim Smith (Epidemiology) – MSc expected 2009 Loek Tan (Epidemiology) – MSc expected 2009

Filippo Alberghina, MD – University of Palermo, Italy Riccarda Failo, MD – University of Verona, Italy Deepa Gopalan – Northern General Hospital Sheffield, UK Katarzyna J Gruszczynska – University of Silesia, Katowice, Poland Patrizia Malagutti, MD – University of Trieste, Italy Alessandro A Palumbo, MD – University Hospital Parma, Italy

Marco Rengo, MD – University of Rome “La Sapienza”, Italy Alexia Rossi, MD – University of Trieste, Italy

Scientific Support Staff Caroline H van Bavel-Hamburg – Trial Coordinator Ronald Booij – Specialized CT Technologist Annemarie M Brugmans-den Toom – Specialized CT Technologist Charlotte Bruidegom-van Heerden – Research Technologist MRI Ommoord Sylvia Bruininks – Research Technologist fMRI Lyda Buist-Ezerman – Research Technologist MRI Ommoord Hilda GB Buitenhuis-Brand – Research Technologist MRI Ommoord Marcel L Dijkshoorn – Research Technologist CT Gabriela N Doeswijk – Research Technician Pauli M van Eldik – Research Technologist MRI Ommoord Suzanne JPM van Engelen – Research Technician AW (Ton) Everaers – Graphic Designer AC (Karin) Haak – Research Technologist MRI Ommoord K (Reinhard) Hameeteman, MEng – Programmer Gavin C Houston, PhD – Applied Scientist in the service of General Electric Berend P Koudstaal – Chief CT Technologist

29

research staff

Wibeke J van Leeuwen, BSc – Trial Coordinator, Experimental Animal Coordinator J (Hanneke) Muharam-Kamminga, BSc – Chief MR Technologist Erik-Jan Schoonen, MSc – Unit Manager OO&O Frans HD Sebus – Administrative Assistant Research Office Corine J Stout-Blom – Radiological Technologist Sandra T van Tiel – Research Technician JM Hannie Versteeg – Radiological Assistant Els C van der Wiel – Research Coordinator Pediatric Lung Imaging CMM (Karin) ten Wolde – Administrative Assistant Research Office

30

31

infrastructure

research infrastructure 2007

Research Committee The Research Committee is core to the structured development of our research programs. It develops departmental research policy, makes management and administrative decisions that transcend the bounds of individual research lines, and monitors progress of the graduate students. Members of the Research Committee in 2007 were Linda Everse, Pim de Feyter, Myriam Hunink, Aad van der Lugt, Gabriel Krestin, Wiro Niessen, and Peter Pattynama. Erik-Jan Schoonen is a valued advisor to the Committee.

Schoonen (Unit Manager), who run the subunit on the basis of Integral Management. As applied to our research, Integral Management means that the Linda (responsible for content) and Erik-Jan (responsible for organization) work in tandem to achieve targets set by the Chairman and the Research Committee (executive powers), as well as to help our researchers achieve their own targets. In this way the balance between content and organizational matters can and is continuously optimized. With the administrative support

Unit Research, Education & Training The Erasmus MC Department of Radiology feels strongly that researchers should be able to dedicate as much of their time as possible to actual research. In order to support our researchers and educators with organizational, bureaucratic, and administrative tasks, we have a Unit Research, Education & Training (Unit OO&O — Onderzoek, Onderwijs & Opleiding).

“Instruments for daily practise”

The Unit is quite unique in the world, and we believe that this organized approach to research support has contributed to the success of our research. The primary goal of the research subunit is to facilitate research within the Department of Radiology in any way feasible. The research subunit is headed by Linda Everse (Research Coordinator) and Erik-Jan

Research Committee: Front: Peter Pattynama; middle: Pim de Feyter, Gabriel Krestin, Myriam Hunink; back: Wiro Niessen, Erik-Jan Schoonen, Linda Everse, Aad van der Lugt.

CT 3D reconstruction of some disposibles used in the radiology department 32

33

infrastructure

of Frans Sebus, Linda and Erik-Jan locate funding opportunities, track deadlines, give critical feedback on grant proposals and manuscripts, administer grants, negotiate contracts, liaise with the legal department, provide notes for lab meetings, arrange and publicize symposia, coordinate the use of various research support services, and manage the entire research process, including the preparation of the yearly account of research activities for the Erasmus MC Department of Research Policy and this Annual Report.

Wibeke is also the Experimental Animal Coordinator for Radiology. As such, she prepares submissions for the local Board of Animal Ethics and liaises with the Animal Welfare Officer. Her combined expertise as biotechnician and radiological technologist adds to the strength of both our own experiments and those of other departments. Linda has accreditation and training permitting her to design and implement experiments with animals. As such, the Research Office can provide expert assistance in study design, imaging technique, and interpretation of images in animal experimentation.

The Unit also employs three specialized Trial Nurse/Data Managers to coordinate and expedite clinical trials. Wibeke van Leeuwen, Caroline van Bavel, and Hannie Versteeg prepare documents for The Unit OO&O also includes the Institutional Review Board, specialists in graphics to proinclude patients, liaise with the vide professional support in Unit Research, Education & Training. Front: Erik-Jan clinic to arrange logistics, and aslayout, design, and printing Schoonen, Linda Everse, Karin ten Wolde, Ton Everaers; back: Marcel Dijkshoorn, Joke Loeve, Wibeke van Leeuwen, Frans semble, enter, and track data. They of publications — including Sebus. Not shown: Caroline van Bavel, Sandra van Tiel. also perform quality control to asboth our PhD dissertations sure performance levels. Wibeke and this Annual Report — and Caroline, as well as Linda, are and large-format posters for trained in GCP, to ensure optimal standardization, data scientific conferences. Graphic designer Ton Everaers also management, and compliance with patient ethics. produces high-quality slides (analogue or digital) for pre-

34

Erik-Jan Schoonen: “In building my own personal infrastructure, an event in 2007 made me realize that miracles really do happen: my first child, daughter Myrthe, was born.” sentations and journal-quality figures for publication, by photographing specialized apparatus for illustration, by processing radiological images, or by providing line art to specification. Ton is assisted by Karin ten Wolde. Medical Technology The Department of Radiology is technically supported by a Clinical Physicist and a group of Medical Technologists. Clinical Physicist Niels Matheijssen decided to leave us in 2007, and we were joined by CT Physicist Marcel van Straten on 1 Jan 2008. Marcel will pursue his qualification as Clinical Physicist while setting up a research line in CT Development. The Medical Technologists include Adri Bot, Ronald van Haaren, Aad Seesink, Robbert Stam en Paul Visser. They maintain the radiological equipment, solve equipment failures, and liaise with industrial technicians as needed to fulfill these tasks. The group is also responsible for the quality control of the facilities of the Department of Radiology. Their work allows researchers to acquire validated and reliable data for their research.

Information Technology (IT) The clinical department operates in an almost entirely filmless and paperless environment. A core group of IT specialists monitor and maintain the PACS and adjacent IT systems: Bert van Heerebeek, Rick Prinsen, Jeroen van IJperen, Ludwig Mayer, and Mart Rentmeester. Tony Merien left the IT

Medical Technology. Aad Seesink, Paul Visser, Ronald van Haren, Marcel van Straten, Robbert Stam, Adrie Bot.

35

infrastructure

Contract Research

cluster employees are also responsible for the Planning and Control cycle of the entire Department of Radiology. The cluster saw a great deal of personnel changes in 2007. Ine Zijlstra, cluster manager, moved to a medical insurance company, where we hope she will be very happy. Her duties have been taken over by Yvonne Koppelman. Adri van der Vorm, head of the HRM group, retired; Yolanda Kievits, personnel advisor, chose to dedicate herself to her children; and Eveline van Kan, our Policy Advisor, moved on to new challenges. They leave large shoes to fill.

Information Technology (IT). Jeroen van IJperen, Ludwig Mayer, Rick Prinsen, Mart Rentmeester, Bert van Heerebeek.

group in 2007 to return to his roots as one of our radiological technologists. The IT group also supports Teleradiology services and speech recognition software. Importantly, this group also provides functional databases to the Research Office to log and track data from (pre-)clinical trials as well as maintaining the research computers. Cluster 7 The Erasmus MC is organized into management clusters, each encompassing 3-5 departments. Within a cluster, the departments share staff for Financial Administration, Human Resource Management (HRM), and Policy Advice. The

36

We divide our research into that of scientific importance to our department, which you will find presented in greater detail elsewhere in this book, and that which we support for the benefit of other departments or industrial relations, but which required very limited scientific input from our researchers. The last type of projects are referred to as “Service Projects,” are performed on a contract basis, and are arranged by the Unit OO&O Trial Coordinators in conjunction with the chief radiological technologists. In 2007, we worked on over 35 Service Projects, mostly to support the research of other Erasmus MC departments.

Despite the many changes, we can always depend on the cluster. Lyda Kramp continues to administer finances for both the Radiology clinic and our research. She provides the infrastructure for grant administration and financial accountability. The HRM group has been taken over by Loes Wijsman, and includes 3 personnel advisors, Annelies Schellevis, Annalisa Saraceno, and Annette Moreland. They run the entire personnel system, including preparing employment contracts, maintaining the personnel files, and giving advice on career planning. Since the Department of Radiology encourages intensive research collaboration, many of our researchers have joint appointments, and the HRM group assures that these have been properly arranged. Cluster 7. Top to bottom: Loes Wijsman, Yvonne Koppelman, Lyda Kramp, Annalisa Saraceno, Annette Moerland, Michel Gouweleeuw, Thea Beyer

37

infrastructure

Imaging Facilities Magnetic Resonance Imaging Brand

Equipment

Year of acquisition

Location

Brand

Equipment

Year of acquisition

Location

GE Medical

3.0T Signa HD

2003

Central Hospital

Siemens

Somatom Definition

2006

Central Hospital

1.5T Signa HDx

2006

DDHK

Sensation s64 Straton

2004

Central Hospital

1.5T Signa [Excite 2]

2005

Off-site

Sensation 16 Straton

2003

Central Hospital

1.5T Signa HD

2003

Sophia

Sensation 16

2002

DDHK

1.5T Signa HD (CVi)

2000

Central Hospital

Emotion 6

2004

Sophia

Siemens

Magnetom Sonata MRE

1999

DDHK

Injector Stellant

2006

Sophia

Esaote

Artoscan

1998

Central Hospital

2004

Central Hospital

Philips

ACS NT

1996

Central Hospital

2003

Central Hospital

Medrad

Injector Solaris

2003

Sophia

2003

Central Hospital

Injector Spectris

2003

Central Hospital

1998

DDHK

2000

Central Hospital

2004

Sophia

2000

DDHK

2003

Central Hospital

1998

Central Hospital

2003

Central Hospital

1998

DDHK

2002

Central Hospital

2006

DDHK

2003

Sophia

2000

Central Hospital

1998

Central Hospital

GE Medical

Philips

38

Computer Tomographic Imaging

Workstation Advantage Windows (ver 4.2)

Workstation EasyVision

Medrad

Siemens

Workstation Leonardo

39

infrastructure

Angiography / Interventional Radiology

Ultrasonic Imaging

Brand

Equipment

Year of acquisition

Location

Brand

Equipment

Year of acquisition

Location

Siemens

Angiostar Plus OR

2000

Central Hospital

Biomedic

SSD 620

2007

Central Hospital

Multistar TOP

1997

DDHK

Aloka Alpha 10

2006

DDHK

V3000

1995

Central Hospital

Aloka Alpha 5

2006

DDHK

C2000

1994

Sophia

C2000

1992

Central Hospital

Tyco

Cooltip RF system

2003

Central Hospital

Medicor

Rita RF systeem

1999

Central Hospital

Medrad

Provis angio injector

2006

Injector Mark V Provis

2000

Injector Mark V Plus

1996

Central Hospital

Injector Mark V Plus

1996

Sophia

Workstation Virtuoso

2000

Central Hospital

Philips

Siemens

2005

Central Hospital

SSD 400

2003

Central Hospital

HDI 5000

2005

Central Hospital

2000

Central Hospital

Central Hospital

2000

Central Hospital

Central Hospital

2000

Sophia

5000 Sono CT

2003

Sophia

HDI 3000

1995

Central Hospital

Sonoline Elegra Advanced

2000

Central Hospital

Sonoline Siena

2000

Central Hospital

Flomap 5500

1999

Central Hospital

Philips

Siemens Endosonics

40

41

infrastructure

Fluoroscopic Imaging

Conventional X-ray Imaging

Brand

Equipment

Year of acquisition

Location

Brand

Equipment

Year of acquisition

Location

Siemens

Axiom Iconos MD

2005

DDHK

Siemens

Vertix 3D

2004

Central Hospital

Polystar TOP

1998

Central Hospital

Vertix 3D

2004

Central Hospital

Uroskope D2

1983

Sophia

Axiom Aristos FX

2004

Central Hospital

Urodiagnost MRF

1999

Central Hospital

Axiom Aristos TX

2002

DDHK

Diagnost 76+

1983

Sophia

Axiom Aristos VX

2001

Central Hospital

Multix Top

2002

DDHK

Multix Top

1998

Central Hospital

Multix H + Vertris pro

1992

Central Hospital

Multix UH

1991

Central Hospital

Diagnost TH

2003

Sophia

Diagnost HDH

1993

Sophia

Ortophos XG3DS

2005

Central Hospital

Ortophos 3 DS

2004

DDHK

Ortophos 3 DS

2003

Sophia

C-arm BV Libra

2004

Sophia

C-arm Pulsera

2003

Central Hospital

C-arm BV Libra

2002

Central Hospital

C-arm BV Libra

2002

DDHK

C-arm BV 29

1998

Central Hospital

C-arm BV300

1998

Sophia

C-arm BV 29

1996

Central Hospital

Philips

Mammography Equipment

Year of acquisition

Location

Lorad

Lorad M-IV

2004

DDHK

Lorad

42

Philips

Brand

Selena Digitale mammograaf

2004

DDHK

Multicare

1998

DDHK

Workstation

2004

DDHK

Demedis Dental

Philips

43

infrastructure

Siemens

44

Mobilet Plus HP

2004

Central Hospital

2004

Sophia

2004

Sophia

2001

Central Hospital

2001

Central Hospital

2001

DDHK

2001

Sophia

45

development

development of Acquisition and Processing Techniques for Diagnostic Imaging

PROGRAM LEADERS: Wiro J Niessen, PhD Johan van der Lei, MD, PhD Gabriel P Krestin, MD, PhD The general goal of this research program is to investigate and validate new diagnostic and therapeutic imaging techniques for Computed Tomography (CT), Magnetic Resonance Imaging (MRI), and Interventional Radiology. At the moment, our acquisition research focused primarily on MRI, but we hope that with the arrival of our new CT Physicist, Marcel van Straten, more developmental research in the area of CT will start. This research in embedded in the NIHES research school (EMC NIHES-03-30-03). Within the MR development, Molecular Imaging plays a growing role. As with all our research, we find integration of specialists from relevant fields important to our research progress. In our Molecular Imaging line, we have a small but growing group of cell biologists working in tandem with MR specialists to produce solid, cutting-edge results. In previous years, Molecular Imaging had its own chapter

in our Annual Report, reflecting its independent management. With the introduction this year of sub-programs, Molecular Imaging has taken its rightful place under the Development program, both in this report and in the management structure. On the image processing side, the main research focuses are the improvement of existing and development of novel computerized techniques for the registration and segmentation of high-dimensional diagnostic images. Specific goals are (a) to increase the outcome of image segmentation techniques, (b) to develop new ways of visualizing high-dimensional data sets by combining existing techniques (surface and volume rendering) with newly developed methods (virtual reality), and (c) to improve user-interfaces and develop automated procedures to assist the end-user with the post-processing, segmentation and visualization of radiological image data. The Biomedical Imaging Group Rotterdam also participates in the ASCI Research School based in Delft.

“Combining modalities” A CT image of a MRI coil post-processed with a virtual rendering technique 46

47

high resolution & molecular imaging

High Resolution & Molecular Imaging coordinator: Gabriel Krestin, MD, PhD

A

B

Within the sub-program “High Resolution & Molecular Imaging,” research is focused on detailed visualization of pathological and biological processes in vivo. Within this program we strive to facilitate the translation from basic research to clinical application. To achieve this goal we have Figure 1. Cell labeling with paramagnetic probes. A. Light microscopy image of a culture dish containing stem cells close interactions between labeled with iron-oxide nanoparticles (stained blue). B. MR image of a culture dish containing stem cells labeled with basic scientists and clinicians iron-oxide nanoparticles (hypo-intense dots). within the Department of Radiology and with external collaborators. Through these collaborations we are not Cell Tracking by MRI only integrating different disciplines, but also various imaging modalities. Within Radiology, clinical 1.5 and 3.0 T Principal Investigator: Monique Bernsen, PhD MRI scanners and custom built receiver coils, suitable for micro imaging in whole-body MR systems, and (multiThe ability to monitor the fate of transplanted cells is of modality) contrast agents are central to our translational crucial importance for the development and validation of research projects. In addition, The Applied Molecular Im(stem-) cell-based therapies. For clinical translation of in aging at Erasmus MC (AMIE) program is currently realizing vivo cell tracking, MRI is currently considered the modalan “Animal Imaging Facility” with state-of-the-art imaging ity of choice due to its excellent resolution capabilities and equipment including a dedicated high-field animal MR high penetration depth and the lack of ionizing radiation. scanner. The combination of preclinical research in laboraIn order to visualize and track cells by MRI, the cells of intertory animals and top-of-the-line human imaging facilities est need to be labeled with paramagnetic probes. While will advance translational research in molecular medicine. the feasibility of such approaches has been extensively proven, several challenges regarding the most optimal la-

48

beling strategies, image acquisition parameters and image analysis and processing techniques are still faced. Through a variety of collaborative efforts we are developing and validating tools to overcome these challenges. Optimal labeling strategies The suitability of a variety of paramagnetic contrast agents, based on iron-oxide nano-particles and Gd-chelates, as intra-cellular probes is being studied regarding their incorporation efficiency, intra-cellular retention and their effects on cell survival, cell proliferation and cell function (Figure 1). Optimal labeling of cells with paramagnetic probes, does not only involve maximal loading of the cells with probe. For cell tracking as an adjunct to cell-based therapy, it is also important that the probe is retained for at least several weeks within the cell, and that the label and/or labeling procedure does not cause toxic effects. A large variety of direct labeling strategies have been used and described for

use in experimental settings. Using a variety of biochemical, cell biological and functional assays, we try to establish optimal labeling strategies for various cell types, such that translation to the clinic is facilitated. Optimal image acquisition For cell imaging and molecular imaging in general, high resolution, sensitive imaging is required. For optimized signal to noise and contrast to noise, we are using customized coils, modified protocols and pulse sequences (see also “High Resolution MRI”). Optimal analysis and processing A major advantage that Molecular Imaging offers to biomedical research is the ability to visualize dynamic biological or pathological processes in vivo. To monitor the fate of transplanted cells in vivo, an integrated assessment

Monique Bernsen: “MRI is a versatile tool for translational research applications.” the incorporation of paramagnetic probes into cells. Direct labeling strategies are based on the exposure of the cells to exogenous probe leading to incorporation of the probe into the cell via endocytotic pathways. Most of these strategies have been aimed at maximal or sufficient incorporation of probe for in vitro or in vivo visualization and for

of cell fate, function and response to treatment will be required. Analysis of such multi-parametric data in time series, poses an enormous challenge in terms of image informatics. While much progress has been made in terms of registration techniques and kinetic modeling, existing post-processing techniques do not accommodate the

49

high resolution & molecular imaging

complexities posed by 4D data. In collaboration with the Dept. of Medical Informatics, we are working on novel algorithms for sensitive detection and registration, matching, and anatomical motion correction. (See also Biomedical Imaging Group Rotterdam).

Participating researchers Dept. of Radiology: Sandra van Tiel, Amber Moelker, Gabriel Krestin, Erik Meijering, Wiro Niessen, Piotr Wielopolski

Dutch Ministries of Economics and of Education, Culture & Science – Smart Mix Grant: “TeRM – Translational Regenerative Medicine”

study clinically relevant principles of cardiac stem cell therapy in small laboratory animals on a clinical scanner. Using the standard clinical imaging gradients, pulse sequences and ECGs, in combination with sensitive small radius surface coils, we are able to image all relevant pulse sequences in rats as well as mice, and detect relatively large number of labeled cells in the myocardium (Figure 2). While visualization of labeled cells in tissue from body parts not affected by uncontrolled motion such as breathing and heartbeat is quite feasible with relevant sensitivity and resolution, comparable sensitivity is severely limited in cardiac tissue, due to motion artifacts caused by the beating heart. We are currently looking at options for increased detection sensitivity, such that homing and/or survival of various cell types can be related to their t therapeutic potential.

In Vivo Application of Labeled Cells

Participating researchers Dept. of Radiology: Sandra van Tiel, Amber Moelker, Robert-Jan van Geuns, Gavin Houston, Gabriel Krestin, Piotr Wielopolski.

Funding: European Union Large Scale Integrating Project 2008-2012: “ENCITE – the European Network for Cell Imaging and Tracking Expertise” Medical Delta – Health Science & Technology Research Grant 2008-2012: “Molecular image processing from cell to organism: An integrative approach”

A

Principal Investigator: Monique Bernsen, PhD

B Figure 2. Cell tracking for monitoring of cellular cardiomyoplasty. A. MR images of a rat with a myocardial infarction showing the akinetic area of the myocardium (left panel) and the contrast enhanced, non-viable tissue (right panel). B. Detection of stem cells labeled with a Gd-based probe following injection into the myocardium.

50

Our experience and knowledge regarding cell labeling and cell imaging is used in translational research regarding stem cell therapy in collaboration with other departments in the Erasmus MC. Two main in vivo applications for stem cell therapy are being studied. Stem cell therapy for myocardial infarction In collaboration with the Department of Cardiology, we are implementing cell imaging using high resolution MRI to

I B C

Stem cell therapy for musculoskeletal disorders

Figure 3. Detection of transplanted stem cells in cartilage lesion. MR image showing the knee cap of a horse with iron-oxide labeled stem cells implanted in a cartilage lesion.

In collaboration with the Department of Orthopedics, we are studying the use of cell imaging by high resolution MRI for cell-based therapy in musculoskeletal disorders. We have assessed the feasibility of labeling mesenchymal stem cells and chondrocytes with iron-oxide nanoparticles. We have already established that both cell types can be labeled with iron-oxide particles and are able to retain the iron oxide particles for several weeks in vitro and in vivo. In addition, we have also demonstrated that MSC can be

efficiently labeled with iron-oxide particles without compromising their multi-lineage potential. Further research efforts will be focused on developing imaging protocols for the detection of labeled cells in affected joints and assessment of therapeutic potential by in vivo imaging (Figure 3). Participating researchers Dept. of Radiology: Sandra van Tiel, Gavin Houston, Gabriel Krestin, Piotr Wielopolski.

51

high resolution & molecular imaging

lection are important issues. It is becoming increasingly clear that treatment efficacy is not only dependent on the tumor cell itself, but also on tumor micro-environment characteristics. Within this research line imaging tools for the assessment of tumor micro-environment characA B teristics and early response parameters are being deFigure 4. Imaging of cancer treatment efficacy. A. MR images of the vasculature of a tumor in a rat tumor model. B. Corresponding images of combined SPECT/CT and SPECT/MR images showing presence of bound radiolabeled veloped and validated. Curpeptide in relation to tumor vasculature.(left panel: SPECT/CT, right panel SPECT/MRI). rently our efforts are focused on peptide receptor targeted Funding: European Union Large Scale Integrating Project radionuclide therapy and biochemotherapy. 2008-2012: “ENCITE – the European Network for Cell Within the Department of Nuclear Medicine, radiolabeled Imaging and Tracking Expertise” receptor binding peptides have been developed as an important class of radiopharmaceuticals for tumor diagnosis Dutch Ministries of Economics and of Education, Culture and therapy. Further improvements of peptide receptor & Science – Smart Mix Grant: “TeRM – Translational radionuclide therapy (PRRT) depend on the optimization Regenerative Medicine” of radiation doses to tumor versus normal organs. Preclinical high-resolution SPECT and autoradiography studies have indicated that peptide distribution in the tumor is Monitoring and Predicting often heterogeneous. By combining microSPECT/CT with Response to Cancer Therapy MRI, we aim at elucidating the influence of tumor (microenvironmental) characteristics on tumor uptake and distriPrincipal Investigator: Monique Bernsen, PhD, and Jifke bution of radiolabeled peptides and on treatment effects Veenland, PhD after peptide receptor radionuclide therapy (Figure 4). Within the department of Experimental Surgical Oncology, In assessing and improving the efficacy of novel anti-cancer local and targeted biochemotherapy are being developed treatment strategies response monitoring and patient seand validated in cancer treatment strategies. In order to

52

assess the efficacy of such treatment strategies, imaging methods play a crucial role. Within this collaborative effort multi-modality imaging strategies, combining optical imaging with MRI, are also being studied. Participating researchers Dept. of Radiology: Sandra van Tiel, Marion van Vliet, Lejla Alić, Jifke Veenland, Gavin Houston, Wiro Niessen, Gabriel Krestin, Piotr Wielopolski Funding: Erasmus MC Translational Research Grant 20032007: “Intravital microscopy and tumor drug uptake studies to assess the predictive value of a clinically applicable indicator in correlation to dynamic contrastenhanced MRI” Erasmus MC Translational Research Seed Grant 2005-2007: “Early treatment response evaluation using MRI to test the suitability of therapy” Erasmus MC Research Grant 2008-2012: Multimodal molecular tumor imaging: Combining SPECT, CT and MRI Netherlands Cancer Society Grant 2008-2012: Functional imaging of tumor uptake and therapy response: Combining CT, SPECT and MRI to monitor Peptide Receptor Radionuclide Therapy

Figure 5. Local contrast enhancement in a rat model of inflammatory bowel disease. Single slice coronal MR image of a rat with inflamed intestine after administration of Gd-liposomes.

Detection and Assessment of Inflammatory Disease Principal Investigator: Monique Bernsen, PhD Inflammatory processes can cause major tissue damage and are a considerable health risks. While the existence of an inflammatory reaction can be determined from blood analyses and the occurrence of fever or a general feeling of malaise, it is not always possible to determine the exact location and extent of the inflammatory process. By using (targeted) contrast agents that will accumulate at the site of inflammation, the presence, location and extent of an inflammatory process could be visualized by in vivo imaging.

53

high resolution & molecular imaging

A

C

B

D

E

Figure 6. High resolution MRI on a 3.0T clinical scanner. Small surface coils can provide adequate signal-to-noise ratio (SNR) necessary for high resolution imaging applications for translational research on clinical scanner platforms. Above, a thin section of human brain showing exquisite gray and white matter delineation and a small microbleed collected with a 2 cm loop coil (A) and a closer, higher resolution image at 40x40x100 µm3 acquired using a 1 cm loop coil (B). MRI could track the development of the muscle fibers in the heart at high resolution and stem cells labeled with iron-oxide particles. Below, a proton density (C), a T2-weighted (D) and a T1-weighted (E) scans from a rat heart using a 2 cm loop coil with a 40x40x600 µm3 resolution.

Local contrast enhancement for increased detection sensitivity of early IBD lesions with MRI The current golden standard for diagnosis and followup of inflammatory bowel disease is endoscopy with biopsy. These procedures, however, can cause considerable discomfort to patients. MRI is considered a valuable alter-

54

native diagnostic modality without being invasive. Regrettably, MRI performs poorly in detecting early stage lesions in inflammatory bowel disease. Local contrast enhancement at inflammatory sites is thought to improve accuracy of disease assessment by MRI and could be obtained by incorporating MRI contrast agents (Gadolinium) in long circulating liposomes. These liposomes preferentially accumulate at inflammatory sites. In a rat model of inflammatory bowel disease we are developing protocols by which we can assess whether local contrast enhancement using Gadolinium liposomes indeed improves MRI detection accuracy of early IBD lesions and thus would provide a feasible clinical application (Figure 5). These studies are performed in collaboration with the Erasmus MC Departments of Gastroenterology, Neonatology, & Experimental Surgical Oncology; the Delft University of Technology; and the Eindhoven University of Technology. Participating researchers Dept. of Radiology: Sandra van Tiel, Remy Geenen, Gabriel Krestin, Piotr Wielopolski, Gavin Houston Funding: Erasmus MC Translational Research Seed Grant 2005-2007: “Improved detection accuracy of early IBD lesions with MRI through local contrast enhancement using lipid-based nanocarriers.”

A

B

C

Figure 7. Large FOV coverage in a mouse using multiple receivers. Multiple receivers can be used to augment the effective field-of-view (FOV) and yet obtain optimal SNR as compared to a single-channel volume coil. A sagittal slice acquired in a mouse is demonstrated using a 4-coil reception interface..

High Resolution MRI Principal Investigator: Piotr A. Wielopolski, PhD Magnetic resonance imaging (MRI) is a non-invasive technique that can provide a multitude of resolution and contrast possibilities and accommodate research interests from basic science to advanced clinical evaluations. This research line explores these possibilities for translational research and molecular imaging techniques, setting up optimized tools for researchers and clinicians in order to obtain both morphologic and quantitative data at all resolution ranges on the 1.5T and 3.0T clinical imaging platforms. These tools cover distinct areas, such as multi-channel signal reception and dedicated receiver coils (in collaboration with Herman Flick, Flick Engineering Solutions BV, the Netherlands), MRI pulse sequence development and support, and measurement protocol improvements (in collaboration with Gavin Houston, General Electric Healthcare, Applied Science Lab­

Figure 8. High resolution scans in different regions-of-interest using multiple receivers. Scanning of several animals or regions-of-interest (ROI) can reduce setup and imaging time with optimal SNR with independent coils positioned in the imaging volume. Two BN175 sarcoma tumors implanted in the hind limbs of two rodents are shown. (A) 3D T2*-weighted scan (16 minutes) delineating blood by-products and scar tissue. A MR angiogram was performed on the smaller tumor after the injection of an intravascular contrast agent. (B) 3D surface rendering of this tumor. (C) maximum intensity projection (MIP) of the tumor’s vascular bed (10 minute acquisition - 3D T1-weighted protocol). Scans performed on the 1.5T clinical MRI system.

oratory Europe). This research line creates a collaboration platform so that a multi-disciplinary approach can be instantiated when new methodologies are available to facilitate communication channels between different imaging modalities.

55

high resolution & molecular imaging

MR micro- and molecular imaging: Small coil arrays and parallel imaging on whole-body unmodified clinical scanners This project investigates the micro-imaging potential of clinical whole-body MRI scanners to provide a flexible platform for translational research and molecular imaging. Specially developed 4 and 8 phased array channel connection interfaces (manifolds) and surface coils make our 1.5T and Figure 9. Vascular architecture in tumors using susceptibility contrast at 1.5T. Long TE - T2*-weighted 3D velocity (flow) compensated scans (TE = 24 ms) can display the vascular architecture of a tumor. Arrows point at regions of lower venous related signal attenuation, locations that may indicate more intense tumor growth. A 500 µm minimum intensity projection (minIP) slice is represented in 3 different planes (original axial acquisition and reconstructed sagittal and coronal planes, respectively). The appearance of hypo intense areas in the periphery of the tumor may be more related to blood by-products. Higher magnetic field strengths, e.g. 3.0T and greater, enhance the effect of magnetic susceptibility at shorter TEs.

Figure 10. High resolution vascular imaging using intravascular contrast agents in a rat skin fold window model. This figure represents a volumerendered depiction of a growing sarcoma (arrows) in a magnetically compatible rat skin fold window model. Using the 3.0T clinical scanner and a custom designed 2 cm ID coil, high SNR can be achieved with voxels on the order of 90x90x100 µm3 with scan times below 20 minutes. The thickness of the window containing the tumor is approximately 1.5 mm. Gd-DTPA-Albumin or GdDTPA liposomes (Gd-Lip) have been used as intravascular contrast agents to provide exquisite detail of the vascular signal surrounding a developing tumor.

3.0T whole-body MRI clinical imaging scanners well suited for this purpose. The typical clinical gradient hardware (a nominal strength of 40 mT/m - 260 ms rise time to gradient maximum) in combination with tailored MRI pulse sequences has been adapted to perform imaging at inplane micron resolution (20-100 µm2 pixels) with very thin sections (Figure 6). The phased array interface makes it possible to extend the field-of-view to suit particular applications ranging from continuous larger area scanning (e.g. 2x8 cm2, Figure 7) to multi-local imaging in different samples with good signal- and contrast-to-noise ratios (SNR/ CNR) (Figure 8) to streamline studies in multiple animals. Participating researchers Dept. of Radiology: Monique Bernsen, Gavin Houston

56

Piotr Wielopolski: “Despite its exquisite contrast and resolution possibilities, translational molecular imaging using magnetic resonance imaging as a tool will require the help of other imaging modalities and extensive basic research to fully realize its potential for diagnosis and therapeutic monitoring.” A

B

Figure 11. High resolution vascular imaging using intravascular contrast agents. A collaboration with the Dept. of Nuclear Medicine has been initiated to understand the relation between the tumor vascular distribution network using high resolution MRI in conjunction with intravascular contrast agents and the signals provided by diagnostic radionuclides with SPECT/ CT to determine the range of action and predict therapeutic efficiency. Here, a high resolution T2-weighted scan (A) provides a signature of the different characteristics of a growing tumor while a fat suppressed, T1-weighted high resolution MR angiogram is formed to realize its blood supply (B). The arrows point at signals provided by the intravascular contrast agent present in the vessels. A dedicated 2 cm ID coil at 3.0T can acquire this data with imaging times below 20 minutes with a 100x100x600 µm3 (A) and 80x80x100 µm3 (B).

High resolution MRI using susceptibility contrast for the detection of vascular structures This research line is intended to provide additional information to help characterize tumor development and the effects of potential therapies. This can be done by using a more novel imaging strategy; exploiting the magnetic susceptibility differences between tissues to provide enhanced contrast, e.g. making it possible for improved visualization for particular structures containing iron or lower oxygen levels. This contrast is generated exclusive using T2*-weighted gradient echo (GRE) sequences (using longer echo times, TE, than usually employed). T2*-weighted scans have been extensively used in the clinics in recent years to produce venograms, to localize tumors and detect vascular malformations in the brain. A high resolution imaging example (Figure 9) of a BN175 sarcoma implanted in the hind limb of rat demonstrates darker signals that corre-

57

high resolution & molecular imaging

spond to venous structures. The T2*-weighted contrast provides also an effective mean of tracking blood by-products, such as hemosiderin in micro-hemorrhages and other chronic conditions (in the brain and endometriosis). This contrast has been also used to track ironloaded cells, such as macrophages, after the injection of ultra-small super-paramagnetic iron particles (USPIOs) in regions with active atherosclerosis. The susceptibility contrast increases dramatically with higher magnetic field strengths so applications are best suited at 3.0T or higher. Participating researchers Dept. of Radiology: Marion van Vliet, Monique Bernsen, Gavin Houston Angiogenesis: High resolution vascular imaging Angiogenesis relates to the formation of new vessels in both benign and malignant processes. It has been documented that the formation of tumor metastasis can be linked effectively to the intensity of this angiogenic activity; a greater number of microvessels have been shown to shed more malignant cells into the blood stream. Therefore, high resolution imaging of the tumor vascular bed is of great interest to us. T1-weigthed 3D imaging at resolutions ranging between

58

Figure 12. Monitoring tumor growth of iron-labeled tumor cells on a clinical 3.0T MRI scanner. The development of ironlabeled tumor cells injected under the skin in the limb of a rat are have been tracked over 4 weeks. For tracking these cells a high resolution T2*-weighted 3D scan is performed in conjunction with a 2 cm surface coil with scan times under 20 min at voxel resolutions of 90x90x100 µm3. Volume rendered images generated to visualize the 3D distribution of the iron-labeled cells, represented in shades of blue. Figure 13. Tracking of Gd-DTPA-liposome labeled tumor cells on a clinical 3.0T MRI system. Tracking Gd-DTPA-liposome (Gd-Lip) labeled cells can have major advantages. The many confounding dark features found in T2*-weighted images can be mistaken easily by labeled cells (iron outside dead cells, macrophages, scar tissue or fat chemical shift). Gd-Lip labeled cells can stand bright in a darker background. For tracking, a high resolution T1-weighted 3D scan is used with a 2 cm ID surface coil and a scan time under 13 min for a voxel resolution of 100x100x200 µm3 at 3.0T. It is hoped that more uniform intensity gradient will be detected with this setting, leading to a better understanding of cell division and tumor growth pattern. The tumor depicted at 6 days post-injection of approximately a million tumor cells (only 1% cells labeled) shows ample contrast for tracking the Gd-DTPA signature (arrows).

60-100 µm2 pixels in-plane with 100 µm slices had been possible on our clinical MRI platforms displaying data that is comparable to intravital confocal microscopy. Specially designed MRI blood pool contrast agents, available in house, make it possible to image the vascular tree at high resolution with high SNR with specialized surface coils. Imaging is performed in models such as the skin fold tumor window (Figure 10), isolated limb (ILP) perfusion and other models (Figure 11). We are assessing the accuracy of the blood pool contrast agents with large molecular weight that we have available to relate tumor aggressiveness with capillary leakage. These developments open the potential to monitoring tumor response more adequately and translate directly the results obtained from therapeutic trials in animal models to clinical reality. This line of research promotes an intensive collaborative channel with other research departments (Nuclear Medicine, Surgical Oncology and Pathology).

Participating researchers Dept. of Radiology: Monique Bernsen, Gavin Houston High resolution imaging for monitoring proliferation and trafficking of magnetically labeled stem cells

Magnetic labeling of cells and tracking with MRI has received in recent years great interest especially the advent of molecular imaging and stem cell based therapies. We have continued strongly on this research line by making our clinical 1.5T and 3.0T MRI systems more suitable to provide high resolution imaging for tracking positively and negatively labeled stem cells for direct translational imaging in the clinics. The idea to use the clinical platforms for this type of research is well justified; research performed on animal MRI systems operating at higher magnetic field strengths may be difficult to translate down to the more commonly used clinical MRI field A B C strengths. At higher field strengths, human imaging is also more difficult and the technology is not yet mature or widespread. Likewise, the inherent changes in the T1 and T2 (T2*) properties in tissues and contrast enhancing agents can be potential obstacles. We believe Figure 14. High resolution MRI of rat joint and implanted scaffold of SPIO-labeled stem cells in equine cartithat our 3.0T clinical MRI scanner lage. Stem cell therapy is being sought to repair damaged cartilage in several models. (A) is a collage of slices can provide the optimal setup refrom the fat-suppressed proton-density weighted 3D volume collected in the joint on a rat acquired under 35 minutes with a voxel size of 60x80x100 µm3 using a surface coil of 2 cm ID at 1.5T. To view the signature garding operational field strength of iron-labeled cells in cartilage, a scaffold without/with iron-labeled chondrocytes has been implanted in a with enough imaging gradient sample of equine cartilage previously damaged (B,C). The green arrow in (B) points at a susceptibility artifact, performance for high resolution similar to the signature of iron-labeled cells, representing remnant air. The white arrow in (C) is a pocket of un-labeled cells in the implanted scaffold.. T2*-weighted contrast for tracking

59

high resolution & molecular imaging

mouse

rat

man

Figure 15. Cardiac imaging potential on MRI clinical platforms – from mouse to man. Functional and anatomical cardiac imaging can be performed effectively in clinical scanners with minor software modifications and with the inclusion of specialized surface coils for high SNR and acquisition throughput. Scanning has been performed effectively down to 100 µm2 in plane resolution with thin slices (0.7-1.6 mm) in mice with all relevant clinical sequences such as black blood fast spin echo imaging, functional gradient echo cine without/with tagging and delayed myocardial infarct scans as applied to humans.

iron-loaded cells or T1-weighted contrast with short TE for vascular imaging or positively enhanced labeled cells (e.g. using Gd-DTPA bearing liposomes, Gd-Lip, with high R1 relaxivity). Tracking iron labeled cells can be visualized using the negative enhancement provided by the larger R2 relaxivity present and using exclusively 3D gradient echo scans (Figure 12). Single SPIO labeled cells have been imaged with a 1 cm ID loop in vitro with a maximum resolution of 20x20x100 µm3 in less than 30 minutes. For tracking Gd-Lip loaded cells, T1-weighted 3D scanning has produced images of developing tumors using a positive contrast agent (Figure 13). With Gd-Lip loaded cells we hope we will be able to enhance our interpretation on cell tracking and differentiation capabilities of therapeutic stem cells. The possibility of working on our wide-bore clinical MRI scanners at standard field strengths makes our setup competitive to specialized animal systems as it can prove effec-

60

tive for imaging many samples in parallel. Special setups for animal handling have been devised for different experiments that, in conjunction with our specialized multichannel reception hardware, coil sets (1-3 cm ID loops), large area coverage and faster scanning using parallel imaging. In our view, optimal coil setups are relatively inexpensive in comparison to animal systems at higher magnetic fields (e.g. 4.6-11T). Whole body human MRI scanners at higher magnetic field strengths, exemplified by the recent installed base of 7.0 T research scanners worldwide, can prove extremely difficult to work with in these type of cell tracking applications given the problems related to B1 field inhomogeneities and radiofrequency power deposition (tissue heating, high specific absorption ratios). Participating researchers Dept. of Radiology: Monique Bernsen, Amber Moelker, Gavin Houston High resolution imaging for cartilage and monitoring of cartilage repair therapies Tissue engineering and regenerative medicine offer potential for effective treatment of a vast range of diseases and disabilities in the near future. Many of these treatments involve the use of autologous cells being culture expanded and re-implanted into patients. Cartilage degeneration in the knee is a common illness. MRI can provide high spatial resolution with exquisite contrast to detect cartilage detects in routine clinical practice. With the idea that stem cell therapy can improve flexibility in the therapeutic procedure, we have experimented with SPIO labeled stem cells to follow cartilage repair and evaluate its status in sev-

eral animal models. Follow up is done using high resolution scanning on our clinical 1.5 T and 3.0T scanners with a set of high SNR reception hardware custom made for these applications. Figure 14 demonstrates a high resolution 3D proton-density weighted scan of a knee of a rodent imaged at 1.5T and an implanted scaffold of iron-labeled chondrocytes in equine cartilage. Participating researchers Dept. of Radiology: Monique Bernsen, Sandra van Tiel, Gavin Houston Rodent cardiac imaging in a clinical platform MRI is definitively the method of choice for providing superb soft tissue contrast, high resolution and parametric imaging, making it possible to investigate subtle structural abnormalities and provide a deeper insight into anatomy and function. In general, work with small animal models has been performed on dedicated small animal MRI systems at higher field strengths (4-11 Tesla). With a rising interest in genetically modified animal models for medical translational research work and the widely available base of clinical scanners in research institutions, it is interesting to explore in depth the imaging possibilities of unmodified clinical scanners for this type of work. Experimental models for cardiovascular disease imaging research have been on the rise. Along the way, MRI has proven as the tool of excellence to measure myocardial mass, end-systolic and end-diastolic volumes, ejection fraction (EF) and aortic blood flow. However, cardiac work on clinical MRI systems for small laboratory animals can be difficult because specialized hardware is deemed necessary to make it possible for high in-plane combined with good temporal resolution

A

B

Figure 16. MRI histology in rat and porcine models of myocardial infarction. MRI can provide a non-invasive mean for tracking iron-labeled stem cells in the heart. Nonetheless, confounding issues can occur depending on the choice of MR pulse parameters when looking for the dark signature of iron labeled cells around a myocardial infarction. (A) depicts a collage of high resolution 2 mm thick minimum intensity projections (mIP) reconstructed from a T2*-weighted 3D scan in a rat heart with a myocardial infarction. Although the infarct could not be clearly visualized, the darker patterns seen in this histological preparation correspond to venous structures and remnant blood in the cardiac chambers. In (B) a T2*-weighted 3D collected in the myocardial infarct region in a pig demonstrates, after 1 week post infarction, a dark signature which corresponds to hemosiderin deposits (no iron-labeled cells were injected). For both examples, a 2 cm ID loop coil used for signal reception at 3.0T.

and good signal-to-noise ratios (SNR) for good image quality. In this research line we explore the cardiac microimaging possibilities of our clinical scanners for stem cell tracking in rodents. In-plane resolution has been possible between 80-150 µm2, with 0.7-3.0 mm sections with adequate temporal resolution for scanning mice with heart rates up to 350 beats per minute (Figure 15). The cardiac MRI pulse sequences and gating equipment present in the clinical scanner is used in conjunction with the standard cardiac electrodes on the

61

high resolution & molecular imaging

paws of the animals or the pulse gating interface. All clinical relevant pulse sequences (black blood, cine, tagging and delayed contrast enhanced scans) have been possible with good image quality at all resolution levels. With this performance, our clinical scanners can help us understand the therapeutic possibilities at lower field strengths, the ones relevant for day to day work flow, with suitable animal model that can demonstrate the process of cardiac repair. Participating researchers Dept. of Radiology: Robert Jan van Geuns, Monique Bernsen, Sandra van Tiel, Amber Moelker MRI of myocardial infarcts: Therapy and tracking of iron and Gd-DTPA labeled stem cells

A

Figure 17. Diffusion tensor imaging (DTI) in normal and infarcted pig hearts. Myocardial fibers can be tracked similarly as demonstrated in the brain using DTI. A region of interest is placed in the myocardium and tailored software is used to track the fibers. The tracked pattern is reminiscent to the spiral pattern seen in histology (A). This DTI scan was acquired on a fresh, in-situ heart using 25 diffusion tensor directions with b=600. In a 5 week pig infarct model the fiber bundles stop at the region of the infarct (arrow) (B). This was acquired in a heart in formaldehyde using 6 diffusion tensor directions. Both examples were collected at 1.5T.

The introduction of cells (myoblasts) that can restore damaged or dysfunctional myocardium is being sought as the only possible therapeutic measure that can bring the cardiac muscle to full functionality after an infarct. MRI can help interpret non-invasively the fate of the delivered cells and provide possible biodistribution and correlation between cell presence and outcome. We have successfully tracked cells labeled with super paramagnetic iron-oxide particles (SPIO) in the heart by looking at signal voids that can be easily recognized. SPIO-labeled cells have been used as a potential means to study cell engraftment after transplantation or intra-coronary delivery into myocardial infarcts. The first therapeutic procedure after a myocardial infarct occurs is to quickly open the blocked arterial territory to

62

B

lems is not an easy matter. We have started a research line for this matter using stem cells loaded with Gd-DTPA liposomes (Gd-Lip) to provide a bright signature in the acquired images. Nonetheless, as seen a week after the occurrence of hemorrhage after reperfusion therapy, bright signals from extracellular methemoglobin is present with a white signature on T1-weighted scans which is a confounding issue for cell tracking with positive contrast. Future work using MRI histology using the different contrast settings at 1.5T and 3.0T and specialized surface coils (as has been done previously by us in the porcine model of myocardial infarction) will be pursued in the infarcted rodent hearts to validate the appearance of infarct and possibilities for labeled cell tracking. Participating researchers Dept. of Radiology: Robert Jan van Geuns, Monique Bernsen, Sandra van Tiel, Amber Moelker Fiber tracking in healthy and infarcted porcine hearts

regain perfusion. As pursued in clinical practice by rapid mechanical or thrombolytic reperfusion, this procedure can cause microvascular obstruction (MVO) and hemorrhage which creates several problems for tracking labeled cells. We are studying the time course of myocardial infarction to determine if blood by-products generate signal voids that would make iron-labeled cell tracking possible in practice. We use myocardial infarcts models in mice, rat and the more human sized such as those of pigs. Hemorrhage can be identified using T2*-weighted scans. Although more challenging to detect hemorrhage in mice and rats, the case in pigs shows similarities to the human counterpart (Figure 16). Envisioning these prob-

This project is an initiative to find an alternative way of assessing cardiac function and regeneration after a myocardial infarction through analysis of the structure of the myocardium fibers. Visualization of the myocardial fibers has been recently possible in vitro and in-situ with the same technique that has been used for mapping the white matter tracks in the brain using diffusion tensor imaging (DTI). In the event of a myocardial infarction (Figure 17), the progression of fiber arrangement could be attempted in-vivo to study remodeling patterns. Eventually with the application of promising therapies, such as the incorporation of stem cells into the infarcted region, a better understanding

Figure 18. High resolution MRI of atherosclerotic plaque at 3.0T. Histological preparation using van Giessen’s and hematoxylin stains compare satisfactorily with high resolution MRI data collected in the axial plane for proton density weighted, T2-weighted, and T1-weighted contrasts. The flexibility of MRI regarding acquisition plane can be used to produce additionally high resolution images in the sagittal and coronal orientations. The microCT scan can depict nicely calcification (seen in MRI with a dark signal) but otherwise, poor contrast in the overall specimen. Diffusion weighted MRI is also under scrutiny in excised specimens as a potential contrast for detecting vulnerable plaque components.

on how cardiac function may improve by monitoring the creation of healthy myocardial tissue. Participating researchers Dept. of Radiology: Monique Bernsen, Gavin Houston, Annick Weustink

63

high resolution & molecular imaging

High resolution imaging of aneurysm formation and atherosclerotic plaque The vessel wall can be inspected at high resolution using magnetic resonance imaging with a multitude of contrasts for the detection and estimation of atherosclerotic plaque burden. Typically, in-vivo scanning involves the collection of proton density-, T2- and T1-weighted contrasts using a black-blood MRI pulse sequence that provides good contrast between vessel wall and lumen. Scan times and the spatial resolution of the images performed in patients cannot be compared to those produced with ex-vivo samples after surgical removal (endartectomy), such as shown in Figure 18, which provide exquisite detail that can compare adequately to the histological preparation. Nonetheless, this exercise provides as with a window to understand the potentials of MRI in the clinics for discovering new contrast possibilities to enhance plaque. Heavily diffusion weighted imaging is being sought as a new type of contrast to determine the vulnerability of a plaque but intravascular contrast agents are being explored at present as a more clinically feasible alternative to help classify vulnerable plaque components. Another possibility is to use T2*-weighted scanning at high resolution in patients without/with the use of ultra-small superparamagnetic iron oxide particles (USPIO). These particles accumulate in places with greater macrophage activity, places that correlate with vulnerable plaque locations. Participating researchers Dept. of Radiology: Aad van der Lugt

64

65

high resolution & molecular imaging

Collaborations: Collaborations with other researchers and research groups at Erasmus MC Rotterdam include:

Researchers primarily based elsewhere with whom a close Collaborations exists include:

Department

Collaborator

Country

City (State)

Institute

Department

Hepatogastroenterology

Ernst J Kuipers, MD, PhD

Austria

Vienna

EIBIR

Neonatology

Ingrid B Renes, PhD

Nuclear Medicine

Marion de Jong, PhD

Orthopedics

P Koen Bos, MD, PhD

Belgium

Mons-Bergen

University of Mons-Hainaut

Organic Chemistry

Robert N Muller, PhD

Eric J Farrell, PhD

Czech Republic

Prague

Institute for Clinical and Experimental Medicine

Diagnostic and Interventional Radiology

Milan Hájek, MEng, PhD, DSc

Gerjo JVM van Osch, PhD

France

Paris

Biospace Instruments, Inc

Eva Haas Monika Hierath

European Society of Radiology

Peter Baierl

Serge Maitrejean, PhD

Jan AN Verhaar, MD, PhD

Marie Meynadier, PhD

Harrie H Weinans, PhD

Institut Curie

Cancer and Immunity Unit

Sebastian Amigorena, PhD

Radiation Oncology

Jeroen Essers, PhD

University Paris Descartes

AMM (Lex) Eggermont, MD, PhD

University Paris Descartes Necker Laboratory of Imaging Research

Olivier Clément, MD, PhD

Surgical Oncology

Max-Planck-Institute for Neurological Research

In Vivo NMR Laboratory

Mathias Hoehn, PhD

Timo LM ten Hagen, PhD

Germany

Cologne

Gerben A Koning, PhD Thorax Center

Medres – Medical Research GmbH

Heleen MM van Beusekom, PhD Dirk J Duncker, MD, PhD Wim J van der Giessen, MD, PhD Nico de Jong, MEng, PhD JET (Annemiek) van Wamel, PhD

66

Collaborator

Israel

Stefan Wecker

Erlangen-Nuremberg

Friedrich-Alexander University

Dermatology

Gerold Schuler, MD

Freiburg

University of Freiburg

Diagnostic Radiology

Jürgen Hennig, PhD

Munich

GE Heathcare ASL Europe

Rehovot

Weizmann Institute of Science

Biological Regulation

Michal Neeman, PhD

Tel Aviv

Tel Aviv University

Slezak Super Center

Gil Navon, PhD

Timo Schirmer, PhD

67

high resolution & molecular imaging

Italy

Milan

University of Milan – Bicocca

Torin

Cage Chemicals, Inc

Biophysics, Biotechnology, Bioscience

Francesca Granucci, PhD

Nijmegen

Radboud University Medical Center

Nijmegen Center for Molecular Life Sciences

Camilla Cavallotti, PhD

Peter Friedl, MD, PhD

Ivan Menegotto, PhD University of Torin the Netherlands

‘s Hertogenbosch

GE Heathcare ASL Europe

Alkmaar

Medical Center Alkmaar

Bilthoven

Delft

Center for Molecular Imaging

I Jolanda M de Vries, PhD

Silvio Aime, PhD Gavin C Houston, PhD

Periodontology & Biomaterials

John A Jansen, DDS, PhD

Orthopedics

Wouter JA Dhert, MD, PhD, FBSE

CF (Kees) van Dijke, MD, PhD

Utrecht

University Medical Center

Progentix Orthobiology, Inc

Joost D de Bruijn, PhD

Winterswijk

Flick Engineering Solutions

Herman Flick, MEng

Signifix, Inc

Eliane Schutte, MSc

Spain

Pamplona

Foundation for Applied Medical Research

Ignacio J Melero, MD, PhD

United Kingdom

London

King´s College

Slough

GE Heathcare ASL Europe

Detroit (MI)

Wayne State University

University of Technology

Radiology

Mediamatics — Data Visualisation Group

Charl P Botha, PhD

Radiochemistry

HT (Bert) Wolterbeek, PhD Urszula D Woroniecka

Eindhoven

Carl G Figdor, PhD

Technical University

Biomedical Engineering

USA

Neuroimaging Research Group

Mike Modo, PhD Race R Yeung

Radiology

E Mark Haacke, PhD

Klaas Nicolay, PhD Gustav J. Strijkers, PhD

68

Soft Tissue Biomechanics & Engineering

Frank PT Baaijens, PhD

Enschede

University of Twente

Biomedical Technological Institute

Clemens A van Blitterswijk, PhD

Leiden

University Medical Center

Endocrinology Research Laboratory

Clemens WGM Löwik, PhD

Hematology

Willem E Fibbe, MD, PhD

Image Processing

Boudewijn PF Lelieveldt, PhD

69

biomedical imaging

Biomedical Imaging Group Rotterdam coordinator: Wiro Niessen, PhD The Biomedical Imaging Group Rotterdam (BIGR) is an initiative of and collaboration between the Departments of Medical Informatics and Radiology. Through innovative fundamental and applied research it aims at developing and validating advanced techniques for the processing and analysis of large, complex, and heterogeneous medical and biological image data sets. The research of BIGR is organized along five research themes: • Cardiovascular image analysis • Neuro image analysis • Cellular and molecular image analysis • Image analysis in oncology • Image guided interventions A strong research focus of the group is to develop, evaluate and validate quantitative imaging biomarkers, for staging disease and monitoring therapy. The validation and evaluation steps are carried out through close collaborations with clinical and research departments within Erasmus MC. The BIGR group closely collaborates with the Department of Imaging Science & Technology of Delft University of Technology, a.o. through joint professor appointments. In 2006 the research activity in all existing research lines has been expanding, and two new research lines were added. The research line: “Image analysis in oncology” was initiated, which will focus on multimodal image analysis to characterize both tumor anatomy and physiology, to

70

improve diagnosis and to evaluate therapy effects. The research line on image guided interventions was started to perform research towards improved 3D guidance in interventional radio­logy and cardiology. A more detailed overview of activities in the research lines can be found in the next sections.

Cardiovascular Image Analysis Principal Investigator: Wiro Niessen, PhD Vascular imaging has gone beyond the traditional depiction of vascular luminal morphology. State-of-the art imaging techniques have the potential to provide detailed information on the vessel wall, such as plaque composition, elastic wall properties, and even biochemical processes that take place in the plaque. In addition, dynamic and perfusion imaging can provide functional information, e.g. for determining the perfusion or motion of the heart, or to study tumor activity. Owing to the growing complexity and sheer size of cardiovascular data, in combination with the large increase in the number of studies in clinical practice and biomedical research, there is a strong and increasing interest in robust, automated processing tools to aid in the analysis of these data. This research line aims to develop and evaluate novel image processing techniques for visualization, quantification and integrated analysis of multimodal anatomical and functional cardiovascular imaging data.

Wiro Niessen: “In 2007 we celebrated our first PhD thesis defense, which is an important happening for a young research group. Equally important, we were able to strengthen all of our research lines, and we have now 12 PhD students associated with the Biomedical Imaging Group Rotterdam. In all research lines, we are very happy with the strong and fruitful collaborations we have: most of the exciting progress is achieved through multidisciplinary research, and in our case this is achieved by combining our knowledge in the field of advanced biomedical image analysis with biomedical and clinical research. Also, we are very happy to announce that in 2010 we will be hosting a major international conference, the IEEE Symposium on Biomedical Imaging, here in Rotterdam. It is both an honor and pleasure to organize this event.” Carotid plaque quantification in CTA data The primary cause of cardiovascular disease is atherosclerosis. A large percentage of strokes are caused by plaque build-up in carotid arteries, which may lead to lumen narrowing and/or plaque rupture. A stroke occurs when the artery becomes completely blocked or a piece of plaque breaks loose and the resulting thrombogenesis causes an obstruction. In order to better understand the pathogenesis of carotid artery plaque, and to develop diagnostic and therapy monitoring tools, the detection and quantification of plaque and plaque components is essential. Computed tomography angiography (CTA) is a noninvasive imaging modality which can be used to accurately

quantify the human vasculature. It is increasingly used to estimate the severity of stenosis in the evaluation of stroke patients. Owing to the increased evidence that carotid artery plaque composition plays an important role in the risk of stroke, there is large interest in quantifying plaque composition as well. This task requires automated analysis, as manual evaluation of quantitative plaque composition by a radiologist is labor intensive and subject to inter and intra observer variability. The most important step towards carotid artery plaque quantification is the difficult task of detecting the vessel boundary, i.e. the outer vessel wall, which encloses both the vessel lumen and plaque. Once defined, it facilitates the further quantification of plaque, a task commonly con-

71

biomedical imaging

Quantification of brain aneurysm morphology and dynamics

A

B

Figure 19. Example of imaging data for automatic analysis. Shown are a CTA scan of the coronaries (A) and MRA of the carotids (B). Automated quantitative analysis of coronary and carotid lumen and plaque is one of the main research efforts in the cardiovascular image analysis research theme.

The project is part of a large EU collaboration, which aims to build an IT infrastructure to integrate data from the genetic up to the population level, in order to improve the management of cerebral aneurysms. Within this project the focus is on the development and validation of image processing tools that enable the analysis of both aneurysm shape, the deformation of the aneurysm during the cardiac cycle, and the change of aneurysm shape over time. Analysis will be both performed both on computed tomography (CTA) data and 3D rotational angiography (3DRA) data as

sidered simpler since the different plaque components have more or less distinctive ranges of HU. In order to solve the automatic outer vessel wall segmentation problem we used a combined approach of a machine learning technique and à deformable model fitting. Participating researchers Dept. of Radiology: Danijela Vukadinovic, Rashindra Manniesing, Theo van Walsum, Sietske Rozie, Philip Homburg, Thomas de Weert, Aad van der Lugt, Harald Groen Funding: Dutch Ministry of Economic Affairs – SENTER Grant 2005-2009: “Automatic Diagnostic Vascular Analysis of CTA Examinations (ADVAnCE)” Figure 20. Carotid plaque quantification prototype.

72

Figure 21. Carotid plaque quantification prototype.

collected within the Erasmus MC and within the European project. Magnetic resonance angiography (MRA) may also be considered in a later stage of the project. The techniques that will be considered include pre-processing techniques (denoising, image smoothing) to improve image visualization and further analysis, aneurysm segmentation (and hence quantification), and motion analysis. Segmentation will be performed using deformable model and level set techniques. Based on aneurysm segmentation, aneurysm shape characteristics will be studied to investigate whether they correlate with hemodynamic factors and/or contain predictive information on rupture risk. Aneurysmal dynamics will be studied using novel dynamic image reconstruction techniques. Two approaches will be pursued. In the first approach, the method for aneurysm

segmentation, as used in the previous activities, is simply applied to all the frames within the cardiac cycle. Subsequently, it is investigated whether descriptors of aneurysmal wall motion, e.g. total motion between systole and diastole, are related to rupture risk. However, owing to the decreased signal to noise ratio in cine datasets, and the possibly very subtle motion that needs to be recovered, the uncertainty in the aneurysmal boundary detection could be too large. Therefore, an alternative approach, which is aimed at estimating the deformation between adjacent phases in the cardiac cycle, is also pursued. In this approach, first a segmentation of the static aneurysm, representing the average of the cine image data, is performed using the methods described in the previous activity. The image data and segmented surface are subsequently registered using non-rigid surface and volumetric image registration techniques. Participating researchers Dept. of Radiology: Azadeh Firouzian, Rashindra Manniesing, Zwenneke Flach, Aad van der Lugt Funding: European Union Integrated Project Grant 20062009: “@neurist: Integrated Biomedical Informatics for the Management of Cerebral Aneurysms” Coronary vessel analysis in CTA Coronary artery disease is one of the leading causes of death in the western world. Automatic methods are necessary for accurate and efficient diagnosis, and prediction, of coronary artery disease. This project involves the develop-

73

biomedical imaging

methods we are currently organizing the Coronary Artery Tracking Challenge during the upcoming MICCAI conference. The purpose of this challenge is to evaluate and compare coronary artery central lumen line extraction methods with standardized evaluation methods and datasets. Participating researchers Dept. of Radiology: Michiel Schaap, Coert Metz, Theo van Walsum, Alina van der Giessen, Annick Weustink, Nico Mollet, Pim de Feyter Funding: Netherlands Organization for Scientific Research (NWO) – VICI Grant 2006-2010: “3D multimodal vascular image analysis for improved diagnosis and therapy” CTA lumen segmentation Figure 22. Automated tracing of coronary artery trees.

ment and evaluation of automatic methods for the quantification of coronary artery disease. We are developing techniques for automatic extraction of coronary artery central lumen lines and methods for automatic segmentation of the coronary lumen. These techniques can be used for automated stenoses grading and improved visualization. Furthermore, the developed algorithms will be used in a system for characterizing coronary plaque. We have developed a robust probabilistic methodology for tracking elongated structures, such as vessels, and used this method for tracking vessels in CTA datasets. Furthermore, for a thorough comparison of coronary artery extraction

74

An often used and important imaging technique for the diagnosis, screening and treatment planning of the vessels in the neck and brain is computed tomography angiography (CTA). CTA can noninvasively visualize the vessels in 3D with high spatial resolution resulting in data sets of typically over 500 slices for one scan. The goal of this project is to segment these vessels automatically, robustly and accurately. The segmented vessels can then be used for improved visualization, for the detection of pathology such as aneurysmata, for hemodynamic modeling, for stenosis and plaque quantification. In recent work we have shown the feasibility of fully automated cerebral vasculature segmentation technique [1] but only applied to a relatively small number of data sets (N=18).Current research is aimed at increasing the robustness of the segmentation method.

Participating researchers Dept. of Radiology:Rashindra Manniesing, Cecile de Monye, Sietske Rozie, Philip Homburg, Thomas de Weert, Aad van der Lugt Funding: Dutch Ministry of Economic Affairs – SENTER Grant 2005-2009: “Automatic Diagnostic Vascular Analysis of CTA Examinations (ADVAnCE)” Distensibility measurements in ECG-gated carotid CTA data The distensibility of a vessel is measure of its compliance and thus of it’s elasticity: (delta V) / (V0*deltaP). Assuming a piecewise constant cross-sectional area, this leads to (Amax - Amin)/(Amin * deltaP) where Amax and Amin indicate the maximal and minimal cross-sectional Area of the vessel during the heart cycle and dP the maximal pressure change during that cycle. We want to automatically calculate the distensibility of the Carotid Artery in a ECG-gated CTA using a automatically segmented lumen and registration of the whole data set. All images in the series are registered to each other and the resulting deformation fields are used to calculate the distensibility. Contours have been drawn on planes orthogonal to the manually defined center line of the carotid, to generate a reliable measure of the distensibility at one cross-sectional plane. Using these contours, the RR-intervals of maximum and minimum cross-sectional area of the carotid are determined. The images of these time points are put into registration and using Gauss theorem the area change is computed by integrating the normal component of the

displacement field found by the registration over the manually drawn contour. The distensibility measure found this way is in close agreement with the manually found measurements. Participating researchers Dept. of Radiology: Theo van Walsum, Reinhard Hameeteman, Sietske Rozie, Aad van der Lugt Funding: Dutch Ministry of Economic Affairs – SENTER Grant 2005-2009: “Automatic Diagnostic Vascular Analysis of CTA Examinations (ADVAnCE)”

Cellular and Molecular Image Analysis Principal Investigator: Erik Meijering, PhD One of the main challenges of biomedical research in the post-genomic era is the unraveling of the molecular mechanisms of life. This is facilitated by recent advances in molecular probing and imaging technologies, which are having an enormous impact on the basic life sciences as well as human health care, by enabling a better understanding of disease mechanisms, the development of new biomarkers for early diagnosis, and enhanced preclinical validation of novel treatments in small-animal models as a first step towards clinical implementation. Studies into dynamic phenomena at the cellular and molecular levels generate vast amounts of spatiotemporal image data, containing much more relevant information than can be analyzed by

75

biomedical imaging

Figure 23. Probabilistic tracking techniques are developed and evaluated to follow structures over time in time lapse fluorescence microscopy.

human observers. Hence there is a rapidly growing need for automated quantitative analysis of time-lapse imaging data, not only to cope with the rising rate at which images are acquired, but also to reach a higher level of sensitivity, accuracy, objectivity, and reproducibility. The goal of this theme group is to develop advanced image processing and analysis methods to enable efficient, accurate, and reproducible quantification and characterization of cellular and molecular dynamic processes. This

is accomplished by 1) developing innovative, probabilistic, model-based algorithms for image segmentation, image registration, object detection, object tracking, and motion analysis, 2) making efficient and robust implementations of the developed methods in the form of software tools, 3) carrying out thorough evaluations of the methods, and comparing them with possible alternative methods, by means of computer simulations as well as by expert human judgments, and 4) assessing the practical value of the methods for molecular imaging studies by using them to answer biologically and clinically relevant research questions, in collaboration with other groups.

Cell motion registration and analysis

Figure 24. Microtubule tracking results.

Tracking in molecular bioimaging Motion analysis of nanoscale intracellular objects, commonly studied using fluorescence microscopy imaging, requires tracking of large and time-varying numbers of spots in noisy image sequences. Conventional approaches to tracking in molecular cell biology typically consist of two separate stages: object detection and temporal data association. Recent comparison studies have shown common techniques based on this principle to fail in practical cases of poor imaging conditions. In this project we develop new

Erik Meijering: “This year we made considerable progress in our cellular and molecular image analysis research. Powerful algorithms were developed for cell and particle tracking, which will help unravel the molecular mechanisms of life. In addition, we received two grants to further extend our research activities in this area.” 76

Funding: Netherlands Organization for Scientific Research (NWO) – VIDI Grant 2005-2010: “Model-driven spatiotemporal tracking for quantitative analysis of subcellular dynamics”

techniques for multiple-object tracking based on nonlinear Bayesian approaches. Since these better integrate available temporal information and application-specific prior knowledge, they can be shown to perform superiorly. A novel particle filtering based algorithm has been developed for quantitative analysis of subcellular dynamics. Compared to existing approaches in this field, the algorithm is a substantial improvement for detection and tracking of large numbers of spots in image data with low SNR. Results of experiments on synthetic image data suggest that the algorithm is potentially more accurate than manual tracking by expert human observers. Experiments on real fluorescence microscopy image sequences from microtubule dynamics studies showed comparable performance. Participating researchers Dept. of Radiology: Ihor Smal, Wiro Niessen

Analyzing the motion and deformation of large numbers of cells in image sequences is a recurrent task in many biological studies. Automated segmentation and tracking methods are increasingly needed to be able to analyze the large amounts of image data acquired for such studies and to improve sensitivity, objectivity, and reproducibility compared to human observers. Quantification of cell motion may help to gain insight in a variety of disease processes. In other studies it allows to correct for global motion and improve the tracking of subcellular structures. In this project we develop new model-evolution techniques for this purpose. A new level-set based cell tracking algorithm has been developed. In order to compare the performance of this new algorithm with other algorithms of the same class and with human observers an extensive validation study has been set up. Participating researchers Dept. of Radiology: Oleh Dzyubachyk Funding: Netherlands Organization for Scientific Research (NWO) – VIDI Grant 2005-2010: “Model-driven spatiotemporal tracking for quantitative analysis of subcellular dynamics”

77

biomedical imaging

Figure 25. Level-set based cell segmentation and tracking.

Neuro Image Analysis Principal Investigator: Henri Vrooman, PhD MR brain imaging is widely used in basic scientific research and in clinical practice, as it is a technique that non-invasively provides both anatomical and functional information. In order to study brain morphometry and function, and its relation to e.g. disease processes or patient characteristics, often large imaging databases are collected. Neu-

78

roimaging research in the Biomedical Imaging Group Rotterdam (BIGR) aims at developing and evaluating state-of-the-art classification tools for accurate and reproducible quantification of brain changes in healthy elderly subjects. Research is carried out within the scope of the Rotterdam Scan Study. The Rotterdam Study is a prospective cohort study ongoing since 1990 in the city of Rotterdam in the Netherlands. The study was initiated by the department of Epidemiology & Biostatistics and targets cardiovascular, neurological, ophthalmological and endocrine diseases. As of 2008 about 15,000 subjects aged 45 years or over comprise the Rotterdam Study cohort. In 1991, a random sample of 111 participants underwent axial T2-weighted magnetic resonance (MR) imaging to assess presence and severity of white matter lesions. In 1995, a random sample of 563 non-demented participants underwent brain MR imaging in the context of the Rotterdam Scan Study. The scanning protocol included series of axial proton-density, T2weighted and T1-weighted images, as well as a highresolution 3D-HASTE sequence. From August 2005 onwards, a dedicated 1.5 Tesla scanner is operational in the research centre of the Rotterdam Study, and brain imaging is performed in all study participants without contraindications every three to four years. The scanning protocol includes 4 high-resolution axial sequences (3D T1-weighted; 2D PD weighted; 2D FLAIR; and 3D T2* GRE), 2D phasecontrast imaging, and diffusion tensor imaging (DTI).

Fully automated brain tissue classification

compared for the atlas-based tissue segmentation method. The atlas-based brain tissue and WML segmentation method has been tested on a set of 209 subjects.

Quantitative analysis of MR brain imaging data is of interest in order to study the aging brain in epidemiological studies, to support diagnosis in clinical practice and to better Participating researchers Dept. of Radiology: Renske de Boer, understand how disease processes affect the brain. Manual Fedde van der Lijn, Albert Vossepoel, Wiro Niessen, Meike analysis of the brain imaging data is a tedious procedure, Vernooij, Aad van der Lugt prone to inter- and intraobserver variability. The Funding: Netherlands aim of this project is to Organization for Scientific develop fully automated Research (NWO) – VICI Grant brain tissue segmentation 2003-2008: “Solving dementia” methods. We compare the (Monique Breteler, Dept. of conventional k-nearest Epidemiology) neighbor classifier (kNN) that requires manually labeled data for training, Knowledge-based with a method that autosegmentation of brain mates this training phase structures using atlas registration. For every voxel in the input Several research questions studdata a decision is made ied in the Rotterdam Scan Study to classify the voxel as require segmentation specific CSF, GM or WM. Our tissue regions or structures in the brain. segmentation method These segmentations can be is extended with a white used in volumetry studies. For matter lesion (WML) seginstance, hippocampus size has Figure 26. Integrated visualization of MRI brain data and diffusion MRI-based tractography. mentation method. This been shown to be an early biomethod uses the brain tismarker for Alzheimer’s Disease. sue classification to determine the lesion threshold value Another application of brain structure segmentations is of the fluid-attenuated inversion recovery (FLAIR) scan. the definition of regions of interest. An example of this is Different types of atlas registration methods have been the segmentation of the ventricles to distinguish different

79

biomedical imaging

to solve this problem. This prior information can be combined with intensity information in a flexible energy framework. The main focus of this project is the development of different prior and intensity models for several structures. Non-rigid registration of a manually labeled atlas was used to delineate the brain lobes. These segmentations were used to measure regional grey and white matter segmentations. White matter lesions are commonly divided in two classes: periventricular lesions lining the ventricles and subcortical lesions. To investigate whether these lesion types show have different effects, an automated classification method was developed. The technique uses a combination of multiple atlas registrations and tissue classification to segment the ventricles. An automated hippocampus segmentation method was developed. The method combines information from multiple atlas registrations, statistical intensity models, and a regularizer in an energy framework that is globally minimized using graph cuts. It was used to revisit a manual

Figure 27. Example of brain tissue classification. T1-weighted, proton density and FLAIR images are shown in three orthogonal views. The right column shows the resulting tissue and WML classification.

types of white matter lesions. However, to delineate brain structures in large collections of images like the Rotterdam Scan Study, automated segmentation methods are almost indispensible.

80

Intensity models like the ones used for brain tissue classification are generally not sufficient for brain structure segmentation: many structures have overlapping intensity distributions and ill-defined borders. We use manually labeled atlases to introduce the additional information necessary

Figure 28. Visualization of white matter lesions. Axial view of a FLAIR sequence showing white matter lesions as hyperintensities. The lesions have been classified as periventricular (yellow) and subcortical (blue) using a ventricle segmentation obtained with an automated method.

volumetry study that revealed associations between hippocampal volume and cognitive performance. These associations were successfully replicated by the automated

Henri Vrooman: “We have developed and validated basic and more advanced image processing tools for the automated analysis of large numbers of MR brain imaging data. This has resulted in an increasing number of international journal papers on brain imaging in collaboration with the Department of Epidemiology. In my opinion, some of the published associations between input determinants and outcome measures could not have been found without the use of these advanced processing tools.” 81

biomedical imaging

Diffusion tensor imaging (DTI)

Figure 29. Automated hippocampus segmentation. Coronal view of automated hippocampus segmentation (blue) compared to the manual ground truth (red). The automated method is based statistical models for intensity and spatial distribution, and regularization.

The diffusivity of water and its directional dependence can be measured using Diffusion Tensor Imaging (DTI). Measures relating to the microstructural integrity of the white matter in the brain can be derived from the diffusion tensor, which is calculated for every voxel in the image. The aim of our research is to enable comparison of DTI scans in large groups of subjects. By building and adopting the required methodology to perform these analyses, we aim to allow research to be done into risk factors and pathophysiological processes underlying neurodegenerative diseases such as cognitive decline and dementia. A close

method, showing its potential for the analysis of large imaging studies.

A drawback of Tract Based Spatial Statistics is its inability to deal with microstructural changes that occur in the white matter tracts outside of the tract centers. For this reason, we are also working on methodology that can quantify the changes that occur in the entire white matter tracts. This exciting research is still in a preliminary stage.

Participating researchers Dept. of Radiology: Fedde van der Lijn, Balinder Paul, Albert Vossepoel, Wiro Niessen, Meike Vernooij, Aad van der Lugt Funding: Netherlands Organization for Scientific Research (NWO) Exact Sciences Open Competition 2005- 2009: “MOSAIC: Multiscale Modelling of Object Shape and Appearance for Analyzing 3D Image Content”

Participating researchers Dept. of Radiology: Marius de Groot, Renske de Boer, Albert Vossepoel, Wiro Niessen, Meike Vernooij, Aad van der Lugt Figure 30. Visualization of white matter lesions. Preliminary association graph from a Tract Based Spatial Statistics analysis, run on 832 subjects (left), and an individual masking image used during the mapping (right).

82

collaboration with the department of Epidemiology creates a synergy between clinical and methodological research. DTI scans were incorporated in the protocol for the Rotterdam Scan Study in 2005, since then, large numbers of subjects have been scanned. In 2007 we have adopted a state of the art method for the voxelwise analysis of microstructural integrity measurements. Tract Based Spatial Statistics, uses an initial alignment of the images, followed by a nonlinear mapping. The mapping stage warps tract centers to a common white matter Figure 31. Rigid body matching of SPECT/CT and SPECT/MRA. A pancreatic tumor was skeleton, this way correcting for residual misalignimplanted in the side of a rat. SPECT images were made following labeling with octment. Voxelwise analyses are performed by mulreotide. tiple linear regressions on the individual mapped skeletons. For this project, our main focus is on extending the statistical analyses that are performed. Oncological Image Analysis

Funding: Erasmus MC Seed Grant 2006-2008: “Automated analysis of diffusion tensor MR brain data”

Principal Investigator: Jifke Veenland, PhD In cancer imaging, advances in MR hardware and software have resulted in the ability to visualize biochemical processes superimposed on anatomic images. For example, the oxygenation status, the acidity, the Brownian motion of water molecules and the blood perfusion can be imaged. Currently, there is a strong interest in determining the value of these functional characteristics as non-invasive biomarkers to evaluate treatment response and outcome. In order to sensitively and reproducibly measure, changes in, these functional parameters, robust and automated processing tools are needed. This research line aims to develop and evaluate image processing techniques for visualization,

83

biomedical imaging

Quantification of multimodal tumor images

Evaluating the suitability of DCE-MRI derived permeability parameters to quantify the effect of cytotoxic agents on tumor tissue: Quantification of local contrast uptake

that treatment effects could only be assessed comparing the pre- and post treatment scans. The patients that responded to therapy had significantly different parametric maps after treatment.

Neuro-endocrine tumors show an overexpression of receptors for regulatory peptides, like somatostatin. Labeling a peptide analog with radionuclides enables targeted treatment of the tumor expressing this particular receptor. However, high-resolution Single Photon Emission Computed Tomography (SPECT) studies and autoradiography studies have demonstrated that tumor uptake of these peptides is often heterogeneous. It is assumed that, besides variations in the expression pattern of the receptors on tumors, tumor micro-environmental characteristics play a role. With MRI various functional characteristics of the tumor can be visualized and quantified in vivo e.g. The goal of this project is to quantify the functional characteristics of the tumor integrating the information present in the multi-modal images. This information is necessary to understand the factors that influence the local uptake and distribution of and response to radiopeptide therapy. With this knowledge PRRT treatment regimens can be optimized.

Evaluation of contrast uptake in tumors is usually by an approach based on region of interest analysis. This quantifies the mean contrast uptake in the tumor or the tumor rim, notwithstanding the fact that, histologically, tumors may be very heterogeneous. A local approach is essential when evaluating the response to therapy: necrotic parts and still vital tumor tissue need then be discriminated. Voxel based approaches are suitable to illustrate the heterogeneity of the tumor contrast uptake, however they require an exact registration and good signal to noise ratios. This project uses data of patients with STS who received ILPtreatment. Since 2001, these patients have been imaged with DCE-MRI before ILP and repeatedly thereafter. DCEMRI based pharmacokinetic parameters were computed per voxel and minimal neighborhood. Two radiologists visually scored the MRIs and the parametric maps. These visual scorings and different pharmacokinetic parameters derived from the DCE-MRIs were compared for the different clinical and histological outcome categories. We found

Participating researchers Dept. of Radiology: Lejla Alić, Wiro Niessen, Marion van Vliet

quantification and integrated analysis of anatomical and functional cancer imaging data.

Participating researchers Dept. of Radiology: Monique Bernsen, Piotr Wielopolski

84

A

B

Figure 32. BOLD images of a sarcoma in a rat limb. Imaging took place before (A) and two hours after (B) treatment with Isolated Limb Perfusion.

BOLD/DWI/MRA/DCE-MRI: Monitoring treatment effect Researchers in the Department of Surgical Oncology have developed a rat model with a soft tissue sarcoma that closely mimics the clinical situation, in terms of response patterns and histopathology, with respect to isolated limb perfusion. The suitability of the different sequences for monitoring short term treatment effects is tested by treating one group of rats with isolated limb perfusion, and another group with sham perfusion. Images of the sarcomas with BOLD, DWI, MRA and DCE-MRI sequences will be obtained before and shortly after treatment. Registration procedures will be optimized to register the different sequences. New methods have to be developed to quantify the local characteristics in the DWI and BOLD images. The MRA and DCE-MRI images will be quantified using the tools developed in the other studies. Finally, the computed tumor characteristics will be evaluated for their suitability to monitor treatment effects.

Funding: Netherlands Organization for Scientific Research (NWO) – Mosaiek Fellowship 2005-2009: “Quantification of tumor vessel morphology: a tool to monitor treatment”

A

B

Figure 33. Local contrast uptake in responding tumor. Parametric maps as computed from DCE MRI before (A) and after (B) treatment of a patient with a sarcoma who responded well to Isolated Limb Perfusion.

Jifke Veenland: “Our new collaboration with the Department of Nuclear Medicine enables us to evaluate tumors with real multi-modal imaging: MRI, CT and SPECT. The combined expertise in tumor models, nuclear agents, SPECT-imaging, MRI-imaging, and image processing will accelerate the quest for imaging biomarkers.” 85

biomedical imaging

Development of registration procedures for accurate assessment of DCE-MRI of tumors To assess the malignancy of tumors within the liver and breasts, Dynamic Contrast Enhanced Magnetic Resonance Imaging (DCE-MRI) is a powerful technique. DCE-MRI data sets typically consist of a base line scan and several scans after the admission of a contrast agent. The Time Intensity Curve (TIC) showing the contrast uptake within the tumor can than be used to characterize the tumor as being benign or malignant. To acquire reliable TICs it would be necessary for the patient to lie perfectly still. However there is always some motion due to cardiac activity and respiration or differences in breathhold positions. Moreover most patients will move a bit during the several minutes of scanning. To compensate for this motion the images within the DCE-MRI series need to be registered to each other, so a perfect spatial alignment is reached. Based on work of Mattes et al.(2003) Thévenaz et al. (2000) and Rueckert et al. (1999) a non-rigid registration algorithm was implemented. Because it is very unlikely that during the short scanning time the volume of the liver will change, we also developed a post processing step that removes all (small) volume changes in the displacement field, found by the registration algorithm. This post processing removes the (local) divergence in a vector field and thus and can be applied to any deformation field and is not limited to our implemented registration routines. Our results on the liver DCE-MRI sets show that the post processing step can reduce the average absolute value of the divergence by one to three orders in magnitude. Currently we investigate the effect of the volume preserving

86

post processing step on mamma-MRI. To assess the effect of registration on the TICs we looked at the variation of the TICs within a homogeneous tumor. Using Principal Components Analysis (PCA) we investigated the explained variance of the first 2 components. Preliminary results show that in the registered data sets these two components explain more variation (their energy is higher) as compared to the non registered data sets. Participating researchers Dept. of Radiology: Reinhard Hameeteman, Albert Vossepoel, Liliane Caldeira, Wiro Niessen, François Willemssen

Image Guidance in Intervention Principal Investigator: Theo van Walsum, PhD Image guided interventions are generally performed using interventional imaging modalities such as fluoroscopy and ultrasound. These modalities allow for instantaneous visualization of the interventional instruments in relation to the patient anatomy, but are limited in their visualization capabilities. X-ray imaging yields 2D projection images, which has bad soft-tissue contrast; for 3D information, biplane imaging systems are required. Ultrasound imaging generally images only one 2D slice. Diagnostic imaging modalities such as MRI and CT, on the other hand, are more versatile and allow for accurate 3D imaging of the anatomy. However, such imaging modalities are generally not suitable for or available in interventional settings.

Theo van Walsum: “The installation of a new state-of-the-art intervention room in 2008 will boost our research activities in the field of image guidance.” Bringing these 3D pre-interventional imaging modalities to the interventional suite has several advantages. It facilitates minimally invasive approaches, by offering additional visualization modalities during the interventions. Moreover, it may improve the accuracy, as it enables introduction of plans that are made on pre-interventional data into the interventional setting. The primary goal of this research line is to improve image guidance in interventions by introducing (the preoperative diagnostic) 3D images into the interventional suite. Whereas such navigation approaches are becoming state-of-the art in applications such as brain surgery and orthopedics, application of this type of technology in interventions in soft tissue anatomy has been hampered by issues such as tissue motion and deformation. Our focus is to bring pre-operative information on the target (e.g. location of a tumor or occlusion, or a treatment plan) to the intervention, also in cases where tissue motion and deformation may occur. We are developing and evaluating methods to keep the pre-operative data synchronized with the interventional situation by combining interventional imaging with additional functional data, position tracking information and motion/deformation models.

Figure 34. Image guidance in vertebroplasty. Left: image guidance using interventional imaging (fluoroscopy). Right: image guidance using preoperative imaging: 3D Rotational X-ray imaging.

The main research areas for image guidance in interventions therefore are: • • •

(3D) image processing, e.g. to enhance or segment the diagnostic images image registration: registration of the pre-operative images to the patient, e.g. by registration of the preoperative modality to the intra-operative modalities tracking of motion and deformation of the anatomy, to allow compensation of the pre-operative images to the intra-operative situation

87

biomedical imaging

• •

and, in relation this, modeling of (cyclic) deformation of tissue, such as a beating heart, to be able to account for these deformations in image guidance tracking of the instruments, to facilitate combined display of the instrument and the pre-operative data

In vertebroplasty, a needle in introduced in a vertebral body transpedicluarly. Accurate introduction is required, as perforation of the pedicle wall may lead to serious complications. Left image shows traditional image guidance: fluoroscopic imaging. If only one C-arm is used, the exact location of the needle with respect to the vertebral body is not known. Right image shows a screenshot of navigation using 3D Rotational X-Ray data. Here, the position of the needle is shown in direct relation to the spinal anatomy, which facilitates accurate needle placement. Improved image guidance in intravascular interventions

tion of the 3D patient anatomy. By developing methods to display the guide wire and catheter with respect to the 3D anatomy, navigation during complex intravascular procedures can potentially be improved considerably. In addition, this technique would enable the visualization of the vessel wall during the procedure, while using X-ray imaging the intra-operative visualization is limited to the vessel lumen. We aim to develop methods to relate the position of instruments in intra-vascular interventions to pre-operative 3D imaging data in real time, so as to complement intraoperative fluoroscopy with 3D information on the vascular lumen and wall, even in case of respiratory and cardiac motion. These techniques would enable improved navigation, and improved vessel wall visualization during interventions, compared to the state-of-the-art. Challenges faced in this project include: coronary artery tracking in 2D (X-ray) and 3D (CTA) images; co-registration of 3D pre-operative images to (multiple) 2D intra-operative images; and motion and deformation modeling from 3D+t or 2D+t images. A voxel classification method has been proposed to automatically find a point in the center of the aorta. Features

In the diagnosis trajectory leading to, or in the planning of, (cardio)vascular interventions, three-dimensional imaging data, e.g. CT(A)/MR(A) are frequently available. However, when maneuvering the guide wire or catheter, or when inspecting the anatomy at the position of these devices, the radiologist or cardiologist usually relies on intra-operative monoplane X-ray projection imaging. Therefore, in monitoring and carrying out a procedure, the clinician Figure 35. Automated detection of the aorta in CTA scans of the heart as initial step in coronary artery has to rely on a mental reconstructracking. Representative results for four datasets are shown.

88

used are: intensity, roundness, size and location of the aorta in the axial slices. Furthermore, a method has been proposed to track the coronary arteries with minimal user interaction (one-click) using a minimal cost path approach. Participating researchers Dept. of Radiology: Coert Metz, Michiel Schaap, Wiro Niessen, Annick Weustink, Nico Mollet, Pim de Feyter Funding: Netherlands Organization for Scientific Research (NWO) – VICI Grant 2006-2010: “3D multimodal vascular image analysis for improved diagnosis and therapy”

variations. We address a number of fundamental questions for the two different types of sparseness scenarios: • How can we optimally use the prior shape model for interpolation in areas where image information is absent? • Given a shape extraction from sparse data, can we estimate the accuracy with which the shape has been extracted? • Given a prior shape model, can we determine which information is required to achieve a surface extraction with a given precision?

3D statistical shape modeling for improved intraoperative guidance

Participating researchers Dept. of Radiology: Nora Baka, Wiro Niessen, Peter Pattynama

In intraoperative situations, especially when interventions involve moving or deforming objects, it is highly desirable to improve 3D anatomical guidance by relating pre-operative 3D imaging data or anatomical models to the sparse data acquired during the procedure. In this research we investigate the reconstruction of 3D shapes out of projectional/cross sectional sparse 2D data. We propose to impose shape constraints based on a-priori knowledge of the expected shape and/or intra-operative deformation. Over the last decade, statistical shape models have been developed, which have considerably improved the robustness and accuracy of image analysis, by incorporating a-priori knowledge about shape/motion. This a-priori knowledge is learned from a set of examples and is typically expressed as an average and a set of characteristic

Funding: Netherlands Organization for Scientific Research (NWO) Exact Sciences Open Competition 2005- 2009: “MOSAIC: Multiscale Modelling of Object Shape and Appearance for Analyzing 3D Image Content”

89

biomedical imaging

Collaborations:

Researchers primarily based elsewhere with whom a close Collaborations exists include:

Collaborations with other researchers and research groups at Erasmus MC Rotterdam include: Department

Collaborator

Cell Biology

Anna S Akhmanova, PhD

Country

City (State)

Institute

Department

Collaborator

Canada

Montreal

MNI, McGill

Brain Imaging Center

Louis Collins, PhD

Germany

Forchheim

Siemens Medical Systems

Hamburg

Philips Medical Systems

Alkmaar

Medical Center Alkmaar

Best

Philips Medical Systems

Thomas G Flohr, PhD

Niels J Galjart, PhD Epidemiology & Biostatistics

Monique MB Breteler, MD, PhD M Arfan Ikram, MD

Medical Informatics

the Netherlands

Roelof Risselada, MD

Radiology

CF (Kees) van Dijke, MD, PhD Wim Crooijmans, MBA

Miriam CJM Sturkenboom, PharmD, PhD

Hein PA Haas, PhD

Neurology

Diederik WJ Dippel, MD, PhD

Frenk M Sloff

Nuclear Medicine

Marion de Jong, PhD

Pathology

Adriaan B Houtsmuller, PhD

Radiation Oncology

Jeroen Essers, PhD

Reproduction and Development

WA (Gert) van Cappellen, PhD

Boudewijn PF Lelieveldt, PhD

Surgical Oncology

AMM (Lex) Eggermont, MD, PhD

JHC (Hans) Reiber, PhD

Thorax Center

Jan Timmer Leiden

Medis Medical University Medical Center

Timo LM ten Hagen, PhD

Rotterdam

Cardialysis Incorporated

Frank JH Gijsen, PhD

Utrecht

University Medical Center

Steve Ramcharitar, MRCP, DPhil Patrick WJC Serruys, MD, PhD AFW (Ton) van der Steen, PhD Jolanda J Wentzel, PhD

90

Chris A Cocosco

Bob Goedhart, PhD Image Processing

Gerrit-Anne van Es, PhD Image Sciences Institute

Max A Viergever, PhD

Neurology

Leonard H van den Berg, MD, PhD

Spain

Barcelona

University Pompeu Fabra

Computer Science

Alejandro f Frangi, PhD

United Kingdom

London

Imperial College

Computer Science

Daniel Rueckert, PhD

USA

Bothell (WA)

Philips Medical Systems

Chapel Hill (NC)

University of North Carolina

Princeton (NJ)

Siemens Corporate Research

Detroit (MI)

Wayne State University

Richard Kemkers Computer Science

Stephen M Pizer, PhD Chenyang Xu, PhD

Radiology

E Mark Haacke, PhD

91

implementation

implementation of Diagnostic and Interventional Imaging Technology

PROGRAM LEADERS: Gabriel P Krestin, MD, PhD Pim J de Feyter, MD, PhD Peter MT Pattynama, MD, PhD Aad van der Lugt, MD, PhD The research program focuses on the implementation of new techniques used for diagnostic and interventional imaging methods. As such, this program holds most of our clinical, applied research. Specific modalities of interest within the program are Magnetic Resonance (MR) technology, Multidetector Spiral Computed Tomography (MSCT), and interventional radiology. There is extensive collabora-

tion with clinical groups within Erasmus MC as well as with Erasmus MC basic science groups. This research program is embedded in the NIHES Research School (EMC NIHES03-30-01). New this year is that we have divided this program into the sub-programs Cardiac, Neuro, Interventional, and Body Imaging, each with its own coordinator. With the developments in the various research lines, the program had become too large to reasonably manage as a single entity. We hope that the decentralization of management and concomitant increased independence will prove fruitful for these researchers in the coming years.

“In search of novelty” A CT image of a magnifieng glass post-processed with a virtual rendering technique 92

93

non-invasive cardiac imaging

Non-Invasive Cardiac Imaging coordinator: Pim de Feyter, MD, PhD The sub-program “Non-Invasive Cardiac Imaging” is a structural and intensive collaboration between the Departments of Radiology, Cardiology, and Pediatric Cardiology. The sub-program is anchored not only within the program “Implementation of Diagnostic and Interventional Imaging Technology (EMC NIHES-03-30-01)” of the NIHES Research

sessment of myocardial viability for prediction of beneficial revascularization effects, and myocardial perfusion to assess the functional limitation of coronary stenoses, is gaining rapidly acceptance in the clinical cardiac arena and is becoming an invaluable tool for the diagnostic work-up of cardiac patients. Computed Tomography of the coronary arteries has rapidly evolved and is now considered as a reliable diagnostic tool to exclude the presence of coronary arteries disease

Pim de Feyter: “In our group, cardiologists and radiologists work in tandem to produce breakthroughs in cardiac imaging while training the next generation to be integrated and multifacetted cardiac imaging specialists..” School, but also within the program “Cardiovascular Imaging and Diagnostics (EMC COEUR-03-43-03)” of the COEUR Research School. In addition to the many graduate students who participate in this sub-program, the sub-program hosts numerous foreign research fellows each year.

Non-Invasive Cardiac Imaging: Magnetic Resonance and Computed Tomography Principal Investigator: Pim de Feyter, MD, PhD Non-invasive cardiac imaging with Magnetic Resonance (MR) for left and right ventricular function, infarct size, as-

94

while the detection of significant coronary lesions needs further improvement. Further research regarding cardiac MR an CT is necessary to 1) establish the diagnostic accuracy of both imaging techniques 2) to improve and adjust the relevant protocols for optimal use in clinical practice 3) to implement the techniques in the clinical setting of patient management and 4) to establish the clinical role of cardiac CT and MR in relation to other non-invasive diagnostic techniques (SPECT, PET, echocardiography). MR and CT imaging is a research line that can only optimally progress due to the collaboration between the Erasmus MC Departments of Radiology and Cardiology.

64 slice CT-coronary angiography in patients with stable and unstable coronary syndromes In a prospective multicenter, multivendor study, we have evaluated 360 patients with stable and unstable coronary syndromes in collaboration with University Medical Centers of Utrecht and Leiden, The Netherlands, using 3 different 64-slice scanners: Siemens Sensation 64, Toshiba Aquilion 64 and Philips Brilliance 64. There were no differences of the diagnostic performance of the different CT-scanners, and this large study confirmed that 64-slice scanners are reliable to exclude the presence of significant coronary artery disease (very high negative predictive value) but that the number of false positive outcomes is rather high due to overestimation of the severity of the stenosis. It is note worthy that in this study all patients and all coronary segments were evaluated, which underscores the fact that current CT-scanners are robust and provide high-quality coronary images. Conventional, invasive X-ray coronary angiography has been the standard of reference for the assessment of coronary artery disease. In patients with angina pectoris, coronary angiography is the key diagnostic procedure to indentify who would benefit from percutaneous coronary intervention (PCI) or coronary artery bypass surgery (CABG). However, coronary angiography is an invasive procedure, potentially harmful with a small risk of serious events (~0.1%) and discomfort to the patient. Furthermore the catheterization-procedure is rather expensive, and because of its invasive nature it involves admission to a hospital or day-care facility. The procedure is carried out by a team consisting of a cardiologist,

nurse and technician, which requires approximately 30 minutes. Non-invasive CT coronary angiography is a relatively simple procedure which is safe, reliable, and patient-friendly. The CT scan is carried out by a technician within 10 to 15 minutes in an outpatient setting. Participating researchers Dept. of Radiology: Bob Meijboom, Nico Mollet, Annick Weustinck, Majanka Heijenbrok-Kal, Myriam Hunink Funding: ZonMW Health Technology Assessment Grant 2004-2007: “MSCT coronary angiography in patients with stable and unstable angina: A multicenter study” 64-slice CT-coronary angiography: Clinical implementation The diagnostic performance of 64-slice CT-coronary angiography (CT-ca) and its continuing updates justifies implementation of newer protocols into the clinical area. We have shown in a fairly large clinical patient population that in patients with a low-intermediate pre-test risk of having significant coronary artery disease the CT-ca is extremely useful and reliable to rule out the presence of significant coronary artery disease. Furthermore CT-ca is less useful in patients with a high-pre-test likelihood of having significant coronary artery disease. In this clinical situation one might preferably refer these patients to invasive coronary angiography. We have shown that, somewhat unexpectedly, that the diagnostic performance of 64-slice CT-ca is almost similar in men and women, and that CT-ca is feasible

95

non-invasive cardiac imaging

Figure 36. Dual source CT coronary angiography. High quality imaging of the coronary tree without the use of pre-scan beta-blockers. HR during scanning 95 beats/minute.

and reliable in patients with unstable coronary syndromes (with low-pre-test risk of having significant coronary artery disease. In a preliminary study we have shown that 64slice CT-ca is useful to rule out the presence of significant coronary artery disease in patients who are referred for cardiac value surgery and who according to the guidelines require a pre-operative coronary angiography to establish

96

the presence or absence of coronary artery stenosis, which may have developed unnoticed by these patients The place of CT-ca with respect to other, well-established, non-invasive stress-imaging techniques including exercise ECG-testing, dobutamine stress echocardiography, SPECT or PET-scanning, needs to be assessed. A first study in which patients underwent exercise-ECG, CT and conven-

tional coronary angiography showed 1 LM the adjunctive value of CT as comLAD pared to the normal initial work-up 1 of patients suspected of having significant coronary artery disease. 2 2 In that study we clearly demonstrated 3 that CT-ca had additional value, su4 perior to exercise ECG stress testing, in the diagnostic work-up of patients 3 with stable coronary artery disease. RCA Patients after left main coronary artery stenting is another population LCx that may benefit from CT, because LM surveillance conventional coronary 4 angiography is recommended in these patients after 2 to 6 months due to unpredictable occurrence DSCT of in-stent restenosis, with its attenFigure 37. High-resolution imaging using dual source CT coronary angiography. dant risks. Seventy-four con-low-up conventional coronary artery stenting underwent 64-slice CT scanning. CT correctly indentified majority of stents that were implanted in larger coronary all patients with in-stent restenosis (10 of 70) but misclasvessel, that had >3 mm diameter. CT-ca was less reliable in sified 5 patients without in-stent restenosis. We concluded stents in smaller coronary arteries that 64-slice CT technology, in combination with optimal Post-coronary bypass patients presenting with recurrent heart rate control, allowed reliable non-invasive evaluation anginal complaints are notoriously difficult to evaluate of selected patients after left main coronary artery stenting. with non-invasive techniques such as exercise ECG testCT is safe to exclude left main in-stent restenosis and may ing, SPECT or dobutamine Echo. We have evaluated these therefore be an acceptable first-line alternative to convenpatients with 64 slice CT-ca and have shown that the artetional coronary angiography. rial and venous grafts can be reliable visualized and nearly We also evaluated diagnostic performance of 64-slice CTall stenoses were accurately detected. However the native ca in patients who had received stent implantation to relief coronary arteries were more difficult to evaluate due to anginal symptoms. In-stent restenosis was detected in the small size and severe calcifications.

97

non-invasive cardiac imaging

Further studies and improvement of CT-technology is necessary for CT-ca to become a clinical useful diagnostic tool in the diagnostic work-up of patients with stable and unstable coronary syndromes.

Dual source CT-coronary angiography

Technical developments in CT-ca are rapidly evolving. A newly introduced Dual-Source CT system equipped with two X-ray tubes and two corresponding detectors, allows Participating researchers Dept. of Radiology: Nico Mollet, improved coronary imaging in patients with higher heart Bob Meijboom, Carlos van Mieghem, Francesca Pugliese, rates because of its improved temporal resolution of 83 Koen Nieman, Annick Weustink, Gert Jan ten Kate, Marco ms independent of patient’s heart rate. We reported the Rengo diagnostic performance of Dual-source CT-ca to detect or rule out significant coronary stenoses in the clinically relevant coronary tree in 100 patients with a wide spectrum of symptomatic coronary artery disease including atypical chest pain, stable and unstable angina pectoris, and non-ST-elevated myocardial infarcLIMA tion. We found that Dual-source CT has an excellent diagnostic accuracy to detect or rule out significant coronary stenosis without the necessity of pre-scan B-blockers. The improved temporal resolution of Dual-Source CT also allows more reliable evaluation of LIMA-LAD anastomosis the in-stent lumen in patients after percutaneous interventional procedures. We evaluated the performance of dual-source CT-ca to detect in-stent restenosis in patients with various stent-diameters. We found that dual-source CT-ca was highly accurate to detect in-stent restenosis in stents with a minimal diameter of at least 3 mm. However, in LAD run -off smaller stents it remains difficult to evaluate the presence of in-stent-restenosis. MIP Dual-source CT-ca is excellent in the evaluation of venous and arterial bypass grafts, but Figure 38. MSCT bypass grafts visualized using dual source CT coronary angiography.

98

Participating researchers Dept. of Radiology: Annick Weustink, Francesca Pugliese, Nico Mollet, Koen Nieman, Niels van Pelt Imaging of atherosclerotic plaque using cardiac CT

Figure 39. Plaque imaging using 64-slice CT coronary angiography. CT-ca shows a lesion in the left main stem bifurcation.

as may be expected the evaluation of native coronary arteries in post-CABG patients remains difficult. Dual-source CT-ca technology is rapidly improving and updated technology is constantly implemented in the dual-source CT configuration. These updates are focused to reduce the radiation exposure to the patient, while preserving the image quality. The introduction of new technology makes the CT-technique more and more complicated and allows a tailored-like approach to each individual patient, which requires thorough knowledge of the CT-scanner and the various protocols which are available to acquire high-quality images at reduced radiation-exposure. Currently we are investigating a step-and-shoot protocol, which is associated with a dramatic reduction in radiation exposure, but which may suffer from reduced robustness and image quality.

The first clinical manifestation of coronary atherosclerosis is sudden death or myocardial infarction which occurs in 40%-50% of individuals without preceding cardiac symptoms. Earlier detection of coronary atherosclerosis imaging would be highly desirable and may be helpful to institute preventive measures in high-risk populations. Contrast-enhanced CT-ca is able to detect the presence of non-obstructive coronary atherosclerosis. Crudely, CT-ca allows distinguishing between calcified and noncalcified plaques. Research is needed to better characterize coronary plaques into lipid/necrotic plaques (potentially vulnerable plaques), fibrous plaques, or calcified plaques. Studies will be performed to characterize and quantify coronary plaques using as the standard of reference invasive coronary imaging techniques: intravascular ultrasound, virtual histology, palpography or optical coherence tomography and when possible histology (atherectomy specimens). We are currently evaluating the coronary plaque burden which is expected to be present in high-risk cardiac asymptomatic population including individuals with familial hypercholesterolemia , diabetes mellitus or peripheral vascular disease. The CT-coronary plaque burden may harbor important predictive information. The coronary plaque burden is usually assessed by visually estimating the size of coronary plaques, but now we are developing and validat-

99

non-invasive cardiac imaging

ing quantitative algorithms to more objectively measure the coronary plaque burden and so to allow progression or regression of coronary atherosclerosis. Participating researchers Dept. of Radiology: Nico Mollet, Lisan Neefjes, Wiro Niessen Funding: Erasmus University Rotterdam (EUR) Fellowship 2006-2010: “Early detection of coronary artery disease by non-invasive, multislice CT coronary imaging in asymptomatic, high-risk individuals” Nuts-OHRA Grant 2006-2009: “Early detection of coronary artery disease by non-invasive, multislice CT coronary imaging in asymptomatic, high-risk individuals”

Magnetic Resonance Imaging of Acquired Heart Disease. Principal Investigator: Robert-Jan van Geuns, MD, PhD In the beginning Cardiac Magnetic Resonance (CMR) had it most important function in complex congenital anatomy followed by left and right ventricular function and vascular flow imaging. With the introduction of inversion recovery gradient echo imaging a very T1 sensitive technique has become available with a high contrast between diseased and normal myocardium. This technique is applicable to both acute and chronic myocardial infarctions but seems also of value in several other cardiac diseases as hypertrofic cardiomyopathy, sarcoidosis and amyloidosis. For all proj-

100

ects in CMR an intensive collaboration between both Radiology and Cardiology is a the key to our success. Myocardial viability in chronic myocardial infarction Delayed contrast enhanced MR is a recently developed technique that allows the differentiation of viable tissue from infarct tissue in dysfunctional myocardial regions depend on blood flow through highly diseased vessels. The amount of viable tissue is a strong predictor for recovery of function after revascularization. We have studied this technique in the setting of prediction of functional recovery after coronary angioplasty for acute myocardial infarction and recanalisation of chronic occluded coronary arteries. Currently, we are following patients treated with low-dose dobutamine for 3 years to determine the long-term relation between contractile reserve and transmural extent of infarction. Ongoing research compares this technique with other tools looking for reserve capabilities for improvement in contraction. Improvement of either technique by combination with MR tagging and quantitative analysis will be studied in the future. MR imaging of acute myocardial infarction and stem cell transplantation in myocardial infaction We now have an established protocol for myocardial infarct imaging which is used for several multicenter studies on new treatments on top of current state of the art primary angioplasty for acute myocardial infarct. We are now working on improved imaging protocols to detect

Robert Jan van Geuns: “Cardiac MRI is a important technique for evaluation of both left and right ventricular function. Assessment of infarct distribution by contrast enhancement is an important recent development with a resolution and contrast to noise ratio incomparable to other imaging techniques. This technique is becoming the major clinical indication for cardiac MRI in the adult population.” the no-reflow area and area at risk. In collaboration with Experimental Cardiology, more experimental studies for new treatment modalities are starting. Participating researchers Dept. of Radiology: Sharon Kirschbaum, Timo Baks, Adriaan Moelker, Pim de Feyter, Piotr Wieloposki MRI myocardial perfusion imaging Identification of myocardial perfusion abnormalities with MRI techniques has been a slow developing technique. Technical the technique has reached a status where clinical application is possible but the combination of long acquisitions protocols including rest and pharmacological stress series seems to limit this application to a selected group of patients only. Application of MRI myocardial perfusion imaging to only patients identified by Multislice CT-ca patients with atherosclerotic artery disease seems more attractive. Our center has a long experience with non-invasive coronary angiography and a good correlation between anatomical abnormalities identified on non-inva-

sive and invasive angiography has been established. However anatomical abnormalities do not always represent the functional limitations in coronary artery flow. To increase myocardial contraction during exercise a two to three fold increase of bloodflow through the coronary arteries is necessary which may be limited by atherosclerotic lesions in the vessels. Unfortunately the severity of obstruction on angiography is poorly related to the limitations in flow increase. Patients with suspected obstructive coronary artery disease on Multislice CT therefore need additional testing especially if multiple lesions are identified. In this research project first a validation process is started to correlated MRI myocardial perfusion techniques with established invasive techniques followed by a larger application in patients with suspected obstructive coronary artery disease on multislice CT to guide angioplasty procedures to the right target lesion. Participating researchers Dept. of Radiology: Sharon Kirschbaum, Annick Weustink, Niels van Pelt, Bob Meijboom, Adriaan Moelker, Pim de Feyter, Carlos van Mieghem

101

non-invasive cardiac imaging

Cardiomyopathies Cardiac MRI is an valuable tools to evaluate different types of genetic linked cardiomyopathies both for anatomy and function. Hypertrofic obstructive cardiomyopathy can now be treated by ablation of a septal artery with alcohol are coils. The therapeutic effect is measured and related to change in symptoms and function. Other Cardiomyopathies under investigation are: non-compaction cardiomyopathy, and Melas syndrome. Participating researchers Dept. of Radiology: Sharon Kirschbaum, Adriaan Moelker Left ventricular function In collaboration with Pie Medical Imaging (Maastricht, The Netherlands), we evaluated a new analysis software program for quantitative analysis of left ventricular function.

ventricular end-diastolic volume, end-systolic volume and ejection fraction. The results of this new concept has been published in several papers and is now are standard technique. In collaboration with the cardiology department this software was used to validate new concepts in 3D echocardiography for left ventricular function. Participating researchers Dept. of Radiology: Sharon Kirschbaum, Adriaan Moelker, Pim de Feyter, Alexia Rossi, Katarzyna Gruszcnska Aortic valve disease Aortic valve stenosis is a disease mainly of the elderly people. Approximately 20% of patients with clinical indications for aortic valve replacements are not operated due to comorbidity. Percutanous aortic valve replacement is a new development which is now available for such patients and in the future will be and alternative with a shorter hospital

Wim Helbing: “In patients with tetralogy of Fallot ... [d]obutamine stress imaging may unmask abnormalities in RV diastolic filling not appreciated with rest imaging alone.” (J vd Berg, et al, Radiology 2007).” This software uses the 3 dimensional information present in the combination of long-axis and short-axis cine loops for MR LV-function for contour detection. Additionally delineation of the mitral valve and apex is improved resulting in fast and very reproducible measurements of left

102

admission and faster recovery of the patient after therapy. For this procedure acquire preoperative measurements are important for adequate sizing and procedural planning. Measurement included are: aortic valve area, aortic annulus diameter, diameter of the aortic root, diameter of the

peripheral vessels. All these measurement can be obtained from a single MRI examination. In the development of this protocol we are optimizing the imaging protocol and validating the measurements. Another import group of patients with aortic valve disease are young people with a bicuspid instead of tricuspid aortic valve. In the Multicenter Procas study patients with stenotic bicuspid aortic valves are followed over time and randomized to a statin or placebo to investigate a possible effect on progression. At baseline a new strategy to measure aortic valve area on a single flow sensitive sequence has been validated. These patients will be followed by MRI and echocardiography in 2 years. Related to this project is the CAMP study were young females with known cardiac disease are followed during pregnancy clinically and with several additional tests like ECG, lab and imaging tools. MRI is used to measure hemodynamic changes and relate this with outcome. Participating researchers Dept. of Radiology: Yusuf Karamermer, Alexia Rossi, Adriaan Moelker, Sharon Kirschbaum

Imaging of Congenital Heart Defects Principal Investigator: Wim Helbing, MD, PhD This research line is intended to increase the body of knowledge that can contribute to the prevention of premature heart failure induced by (congenital) heart defects.

The research line is a structural collaboration with the Department of Pediatric Cardiology, with Prof. Wim Helbing and several of his graduate students holding joint appointments. The hypotheses to be tested in this research line have been divided into several projects. Evaluation of new imaging techniques Real-time 3D echocardiography is being compared to 2D cine-MRI for the assessment of Right Ventricular (RV) function in 120 patients with congenital heart defects. We hope to determine if real-time 3D echocardiography can replace MRI (the current gold standard) in the clinical measurement of RV function in this patient population. 3D echo is both cheaper and more patient-friendly. Funding: ZonMW Health Care Efficiency Research Grant 2007-2009: “Comparison of 3D echocardiography and MRI for assessment of right ventricular function in congenital heart disease” [Folkert J Meijboom, Cardiology] Early diagnosis of right ventricular dysfunction following operation for Tetralogy of Fallot. One of the most common forms of congenital heart defects is the so-called Tetralogy of Fallot. This defect forms a good model for long-term consequences of abnormal function of the RV. In a multicenter study coordinated by the Department of Pediatric Cardiology, patients with RV dysfunction following correction of Tetralogy of Fallot will undergo serial measurements with MRI, stress tests, and biomarkers. The goals are to describe the development

103

non-invasive cardiac imaging

over time of the RV size and function and to identify risk factors and markers for degradation of RV function. This information is expected to lead to novel insights into medications for RV dysfunction.

“Hemodynamic and genetic factors in long term outcome of Fontan circulation”

Participating researchers Dept. of Radiology: Saskia Luijnenburg, Piotr Wielopolski Funding: Netherlands Heart Foundation 2007-2010: “Early diagnosis of right ventricular dysfunction in patients operated for tetralogy of Fallot: A multicenter study with serial follow-up” Heart dysfunction and univentricular circulation: Role of hemodynamic and genetic factors. About 10% of patients with congential heart defects have a univentricular heart. Widespread application of the Fontan operation has tremendously increased the life expectancy of these patients, but long-term heart failure remains a significant problem. One goal of this project is to identify hemodynamic risk factors for heart dysfunction. A second goal is to find markers identifying degradation of heart function in this group. For these purposes, patients will undergo MRI-based measurement of heart function and stress tests, and their blood will be assessed for biomarkers. Participating researchers Dept. of Radiology: Danielle Robbers, Piotr Wielopolski Funding: Sophia Children’s Hospital Foundation / Wilhelmina Children’s Hospital Foundation 2005-2008:

104

105

non-invasive cardiac imaging

Collaborations: Collaborations with other researchers and research groups at Erasmus MC Rotterdam include: Department

Collaborator

Epidemiology & Biostatistics

Monique MB Breteler, MD, PhD

Ophthalmology

Huib J Simonsz, MD, PhD

Osama II Soliman, MD, PhD

Orthopedics

Harrie H Weinans, PhD

AFW (Ton) van der Steen, PhD

Thorax Center

Heleen MM van Beusekom, PhD

Jolanda J Wentzel, PhD

Eric Boersma, PhD

Sing-Chien Yap, MD

Thorax Center (continued)

Patrick WJC Serruys, MD, PhD

Ewout Jan van den Bos, MD, PhD Nico Bruining, PhD Folkert J ten Cate, MD, PhD Kadir Caliskan, MD Jaap W Deckers, MD, PhD Ron T van Domburg, MD, PhD HJ (Erik) Duckers, MD, PhD Dirk J Duncker, MD, PhD Marcel L Geleijnse, MD, PhD Wim J van der Giessen, MD, PhD Peter PT de Jaegere, MD, PhD JBoudewijn J Krenning, MD Daphne Merkus, PhD Michelle Michels, MD Amber D Moulker Steve Ramcharitar, MRCP, DPhil Jolien W Roos - Hesselink, MD, PhD

106

107

non-invasive cardiac imaging

Researchers primarily based elsewhere with whom a close Collaborations exists include: Country

City (State)

Institute

Germany

Forchheim

Siemens Medical Systems

Munich

GE Heathcare ASL Europe

‘s Hertogenbosch

GE Medical Systems [ASL Europe]

Amsterdam

Free University Medical Center

the Netherlands

Department

Collaborator USA

Gynecology Cardiology

Baltimore (MD)

GE Heathcare ASL

Sandeep N Gupta, PhD

Thomas G Flohr, PhD

Bethedsa (MD)

GE Heathcare ASL

Maggie Fung

TTimo Schirmer, PhD

Milwaukee (WI)

GE Heathcare ASL

R Scott Hinks, PhD

Gavin C Houston, PhD

Niskayuna (NY)

GE Global Research Center

Donna Collins

Marco JW Götte, MD AC (Bert) van Rossum, MD, PhD

Delft

University of Technology

Mediamatics — Data Visualisation Group

Charl P Botha, PhD

Dordrecht

Albert Schweitzer Hospital

Cardiology

Marcel J Kofflard, MD, PhD

Eindhoven

Technical University

Biomedical Engineering

Gustav J. Strijkers, PhD

Leiden

University Medical Center

Cardiology

Jeroen J Bax, MD, PhD

Maastricht

Pie Medical

Boudewijn JA Verstraelen Jean Paul MM Aben, BSc

Spain

108

University of Maastricht

Medical Informatics

Ed HBM Gronenschild, PhD

Nijmegen

St Antonius Hospital

Cardiology

Benno JWM Rensing, MD, PhD

Utrecht

University Medical Center

Cardiology

Maarten-Jan M. Cramer, MD, PhD

Winterswijk

Flick Engineering Solutions

Madrid

Hospital “Gregorio Marañón”

Cardiology

Esther Pérez-David, MD, PhD

Polytechnic University

Electronic Engineering

MJ Ledesma-Carbayo, PhD

Herman Flick, MEng

109

neurological & neurovascular imaging

Neurological & Neurovascular Imaging coordinator: AAD VAN DER LUGT, MD, PHD Neuroradiology serves as a strong partner to related clinical departments to solve clinical problems, to guide therapy, and monitor the effects of therapy by follow-up studies. The research is mainly focused on neurovascular disease, where we have built a strong collaboration with the Departments of Neurology and Vascular Surgery to validate and evaluate the role of CT and MRI in the imaging of atherosclerotic plaque. A second theme in the clinical domain is pediatric Neuroradiology, where research concentrates on determining the role of non-invasive imaging in diagnosing and monitoring pediatric diseases. Currently, brain imaging modalities find frequent use outside the clinical domain: functional MRI is increasingly used by basic researchers in the neurosciences. In addition to our own research in fMRI, we support other groups in the design and development of an fMRI program to test their own research hypotheses. An important area within this sub-program is epidemiological research. Imaging-derived parameters have finally been established as solid intermediate endpoints in epide-

miological studies. We participate in the Rotterdam Study, a large population-based cohort study of the elderly, and are currently considering participation in Generation R, a large cohort of children who are being followed from prenatal care throughout their childhood.

Cerebrovascular Disease & Atherosclerotic Plaque Principal Investigator: Aad van der Lugt, MD, PhD Ischemic cerebral infarcts are related to the presence of atherosclerotic disease in the carotid artery. Severity of the stenosis is a predictor of clinical symptoms and is used as parameter in the therapeutic decision as to which patients will benefit from carotid intervention. Next to stenosis severity, plaque morphology is thought to be a major determinant of symptoms. The concept of unstable or vulnerable plaque, which may rupture and lead to release of thromboembolic material, has been postulated for the coronary arteries and may also be applicable to the carotid arteries. This vulnerable plaque contains a large necrotic lipid core

Aad van der Lugt: “Continuous increases in image quality will lead to the detection of increasing numbers of incidental findings. It is not unlikely that each one of us has an abnormal anatomical structure somewhere in his or her body. Although the first enthusiastic response will be a call for extensive screening programs, we must remember that the efficacy and cost-effectiveness of screening need to be proven before launching such initiatives.” 110

covered by a thin or disrupted fibrous cap. Visualization of atherosclerotic plaque and assessment of vulnerability in humans with non-invasive imaging techniques greatly enhances the understanding of atherosclerotic disease and the cerebrovascular events. MRI of the atherosclerotic plaque High-resolution magnetic resonance imaging (MRI) has emerged as a potential modality for atherosclerotic plaque imaging. Our research has addressed optimal coil design and the analysis of newly designed pulse sequences (such as black-blood fast spin echo). We have validated MRI by comparing images with histology and assessed the accuracy of MRI in the detection of different plaque com¬ponents. With dedicated imaging hardware, we now study in great detail plaque and the processes around plaque for¬mation in-vivo and ex-vivo at 3.0T. The final goal is to find a non-invasive marker that makes it possible to track those atherosclerotic plaques that are vulnerable and most dan¬gerous. We will also evaluate the additional value of con¬trast agents in the delineation of atherosclerotic plaque and its components, and in the quantification of plaque and plaque components. Participating researchers Dept. of Radiology: Sietske Rozie, Quirijn van den Bouwhuijsen, Piotr Wielopolski Funding: Schering AG Clinical Research Grant 2006-2007: “Improved characterization of atherosclerotic plaque in the carotid artery with contrast-enhanced (MS-325), high resolution 3T MRI”

B

A

C

Figure 40. MRI of the carotid artery with atherosclerotic plaque. . 3D T1w images were acquired in the coronal plane (A). The high resolution allows the reconstructions of images in the axial plane (B,C).

CT of atherosclerotic plaque components and morphology Previous in-vitro and in-vivo validation studies have demonstrated that non-invasive multi-slice computed tomography (MSCT) with thin collimation (0.5 mm) was able to image different plaque components. The next step was the development of a software tool for quantification of atherosclerotic plaque volume and volumes of the different components. We demonstrated that interobserver variability of the different measurements was moderate. Nevertheless,

111

neurological & neurovascular imaging

we still expect that improvements in the measurement software will improve the observer variability. In our first prototype, full manual outlining of the outer contour and partial outlining of the lumen-calcification interface introduced measurement variability. A new approach exploits the contrast between vessel wall and peri-arterial fat to enable automated detection of parts of the outer contour of an artery and automated lumen segmentation. The analysis in the prototype was performed in axial two-dimensional images. Evaluating the artery both in axial slices and using longitudinal reformats will provide more information on the borders of the vessel wall. This would enable a better continuation of transversal contours in adjacent slices. The Biomedical Imaging Group Rotterdam will improve the post-processing software, which will subsequently be tested on clinical data. We have assessed the volume of the atherosclerotic plaque and the proportion of the different plaque components in 100 symptomatic carotid arteries. We analyzed the relationship of these volumes with severity of stenosis, risk factors for atherosclerosis and cerebrovascular pathology. Furthermore, we have compared the plaque composition in the symptomatic artery with that in the contralateral asymptomatic carotid artery. CTA is superior to DSA in the detection of plaque irregularities and ulcerations. We have assessed atherosclerotic plaque surface morphology in the carotid artery in a large consecutive cohort of more than 400 patients with cerebrovascular symptoms. Plaque surface morphology was related to severity of stenosis, cardiovascular risk factors, and cerebrovascular symptoms. The ultimate goal is to assess the relationship between ca-

112

Figure 41. Plaque ulceration. In a multiplanar reconstruction of CTA data of the carotid artery, contrast is visible beyond the plaque borders, indicating that the atherosclerotic plaque has ulcerated after rupture of the plaque surface.

rotid atherosclerotic plaque features (volume, components, rupture) and cerebrovascular events in a clinical follow-up study. Nine-hundred patients have now been recruited for this study. Participating researchers Dept. of Radiology: Thomas de Weert, Sietske Rozie, Philip Homburg Funding: NWO Clinical Fellowship – 2004-2008: “Detection of vulnerable plaque in the carotid artery with multislice CT”

Serial CT Angiography of atherosclerotic carotid plaque Not much is known about the natural course of atherosclerotic disease in the carotid arteries. In this project we will evaluate atherosclerotic carotid plaque features (plaque volume, plaque composition, plaque morphology, shear stress) with MDCTA. We hypothesized that a change in plaque features can be predicted based on risk factor profile, baseline plaque features and shear stress and that assessment of baseline plaque features and a change in plaque features will result in accurate identification of patients at high risk of stroke. Three-hundred patients who had a baseline MDCTA with atherosclerotic carotid disease will be invited for a follow-up MDCTA after 3 years. Baseline and follow-up datasets will be matched to enhance the detection of changes over time. Funding: Dutch Heart Foundation – 2008-2011: “Serial CT Angiography of atherosclerotic carotid plaque: determinants and prognosis of change in volume, composition and morphology” Dual Source CT angiography (ECG-DSCTA) of the carotid arteries Noninvasive measurement of mechanical properties of arteries is useful because it provides a functional parameter of a diseased artery, which allows the evaluation of the effect of atherosclerotic disease on the function of the artery. In addition, this parameter may play a role in risk prediction and the evaluation of the pharmacological intervention.

Distensibility of the carotid artery has been evaluated with ultrasound and MRI. Up till now CT has not been able to provide this functional information of the carotid bifurcation. Dual-source CT (DSCT) is a novel design principle for CT that combines two arrays of X-ray tube plus detectors that are arranged at a 90° angle. The temporal resolution of the system is increased with a factor 2 to 83 ms. With ECG gated dual Source CT angiography, data are collected during the heart cycle, but reconstructed per heart phase. With reconstruction of multiple phases, the arterial distensibility can be calculated. In addition, the reconstructed images are less influenced by motion of the carotid arterial wall, which may result in better image quality. The main objectives of this study are to assess the distensibility of carotid arteries and to relate the distensibility to the amount of atherosclerotic plaque and to the plaque type (calcified/ non-calcified). Participating researchers Dept. of Radiology: Sietske Rozie CT of intracranial calcifications Coronary artery calcification, visualized with electron beam or multidetector computed tomography (MDCT) and assessed with the Agatston score, has been the most important atherosclerotic plaque feature studied with imaging modalities. Coronary artery calcification reflects the total plaque burden, is associated with cardiovascular risk factors, and is an independent risk factor for future ischemic cardiac and cerebral events. Although atherosclerotic calcifications in the intracranial internal carotid arteries are very frequent, their association with cardiovascular risk factors

113

neurological & neurovascular imaging

and their predictive value for cerebrovascular events has not been studied extensively. We have developed a semiautomatic custom-made system to quantify calcifications. Calcifications were defined as pixels within the arterial wall with an attenuation above 500 Hounsfield Units. In this study we analyzed the association between cardiovascular risk factors (smoking, hypertension, diabetes, hypercholesterolemia, previous cardiac disease and previous cerebrovascular disease) and the presence and volume of calcifications, In addition, the relation between calcifications and cerebrovascular symptomology is assessed. Participating researchers Dept. of Radiology: Thomas de Weert, Philip Homburg Shear stress calculations based on CTA data Atherosclerotic plaques rupture most frequently at the entrance section of a stenosis, where shear stress is higher than normal. Stability of the cap of a vulnerable plaque hangs on the balance between cap enforcing matrix synthesis by smooth muscle cells and enzymatic matrix breakdown induced by macrophages. At the entrance of a stenosis, however, the concentration of macrophages exceeds that of smooth muscle cells. The direct mechanical effect of sheer stress probably does not cause tissue tearing, but the high sheer stress enforced biological activity of the endothelium at the stenosis entrance might explain the diminished smooth muscle content that likely contributes to cap thinning. In this study we investigate the co-localization of sheer stress (SS) and plaque composition, including inflammation

114

markers in human atherosclerotic carotid plaques, by noninvasive functional imaging and 3D histology. Multislice computed tomography data from the carotid bifurcation will be used to 3D-reconstruct the lumen of symptomatic patients undergoing surgery, which then serves to calculate local SS by computational fluid dynamics. Endarterectomy specimens will reveal plaque composition and protein concentrations. Reconstruction of the histological data into the CT data set allows studying the co-localization between SS and plaque components, including inflammation markers. This study may yield crucial data for the guidance of anti-inflammatory therapy in patients with vulnerable plaques. Participating researchers Dept. of Radiology: Harald Groen, Philip Homburg, Sietske Rozie Funding: Dutch Interuniversity Cardiological Institute (ICIN/KNAW) Project Grant 2005-2009: “High shear stress contributes to plaque rupture by spatially restricted endothelial anti-inflammatory signaling” CT, CTA, and CTP in acute stroke Thrombolysis with rtPA is an effective treatment for patients with ischemic stroke within 3 hours from onset. The 3-hour time window as a treatment indication is a crude way to select patients. Simple, yet reliable and valid methods to select patients who may benefit from thrombolysis are clearly needed, in order to target the treatment, and to extend the treatment indication beyond the restrictive 3-hour window. CT is a promising imaging modality

to select patients with a potential treatment response, as it measures cerebral perfusion parameters as well as patency of the internal carotid artery and major branches of the circle of Willis. CT has widespread availability, low sensitivity to motion artifacts, and short acquisition times. On the other hand, it is unknown whether the visualized penumbral area predicts a treatment effect with acceptable accuracy. Furthermore, we do not know whether perfusion CT is useful for patients with small subcortical infarcts, and which (combinations of ) perfusion parameters best predict outcome and treatment response. The purpose of this study is to assess the prognostic value of triple CT (CT, perfusion-CT, and CT-angiography) for acute ischemic stroke in relation to outcome and treatment effect of intravenous thombolysis.

rate differences in the two treatment arms are expected to be small, large numbers of patients have to be included. Asymptomatic, or ‘silent’ ischemic lesions, detected with diffusion weighted MRI (DWI) occur more frequently than clinical events after carotid intervention. These ‘silent’ lesions may be clinically important since they were found to be associated with cognitive dysfunction in the general population. These lesions may therefore become a valuable outcome measure in clinical trials of CAS versus CEA. The patients included in the ICSS trial (stent versus endarterectomy) are studied with DWI. This project is collaboration with the University Medical Center Utrecht. Participating researchers Dept. of Radiology: Zwenneke Flach, Lukas van Dijk

Participating researchers Dept. of Radiology: Cecilé de Monyé

Epidemiological Studies

Carotid artery intervention: Evaluation with diffusion weighted imaging

Principal Investigator for Radiology: Aad van der Lugt, MD, PhD

Carotid endarterectomy (CEA) combined with medical treatment is an established therapy for patients with symptomatic carotid artery stenosis. Carotid artery stenting (CAS) has recently been advocated as a less invasive alternative for CEA. Case series of endovascular treatment with angioplasty or stenting have shown morbidity and mortality rates (30-day stroke and death rate: 4.9-10%) comparable with those after CEA. Randomized multicenter studies are currently comparing CAS and CEA — these trials have stroke-free survival as primary outcome. Because stroke

The Rotterdam Study is a prospective, population-based study aimed at investigating determinants of chronic and disabling diseases in the elderly. Developed and coordinated by the Department of Epidemiology & Biostatistics, it started including volunteers in the Rotterdam neighborhood Ommoord in 1990. Base-line data, detailed exposure status information and follow-up are well secured in the research environment of the Rotterdam Study. In 2002 the Department of Radiology took an active part in the Rotterdam Study and has a strong collaboration with

115

neurological & neurovascular imaging

neuro-epidemiology (Prof. Monique MB Breteler) and cardiovascular epidemiology (Prof. Jacqueline W Witteman). From 2003-2005, around 2000 subjects have undergone multidetector CT of the heart, aorta and carotid arteries. Starting in 2005, imaging of the brain has been embedded in the core protocol with a 1.5 Tesla GE scanner installed in a central location in Ommoord. Starting in 2007, MRI of the atherosclerotic plaque in the carotid bifurcation has been added to the core protocol.

tery calcification is a strong predictor of future coronary heart disease, but studies were in small, selected populations. No data are available on the relation of calcification in the aorta and carotid arteries, as detected by multidetector CT, with risk of stroke. In this study we will examine the predictive value of coronary calcification for risk of coronary heart disease as well as that of aortic and carotid calcification for risk of stroke. 2000 participants have been scanned and data analysis has been done. Follow up Cardiovascular and results have to be awaited. cerebrovascular diseases Magnetic resonance imag(Atherosclerotic plaque) ing (MRI) has the potential to measure plaque strucAtherosclerosis has a preture in asymptomatic subdilection for the coronary jects. MRI of the coronary Figure 42.. MRI of the carotid artery . PDw-EPI of the carotid artery demonarteries, the aorta and the arteries remains difficult strates thickening of the vessel wall due to the atherosclerotic process. carotid arteries. Measuredue to their small size and ment of atherosclerosis at motion but the accuracy of these sites may improve risk prediction of coronary heart multi-sequence MRI for identifying features of vulnerable disease and stroke. The new generation of fast multidetecplaque in human carotid atherosclerosis has been demontor computed tomography (CT) scanners enables accurate strated. Plaque vulnerability has been found to be multifoassessment of calcification of the coronary and extra-corocal and therefore, although local factors play an important nary vessel beds. These calcifications are almost invariably role, plaque vulnerability is likely to be a generalized pheassociated with atherosclerotic plaques. nomenon. In the present proposal, we will examine the Several studies suggest that the presence of coronary arstructure of atherosclerotic plaques in the carotid arteries,

116

and their development to overt cardiovascular disease. The ultimate aim is to develop a novel risk assessment strategy that includes the assessment of vulnerable plaque. The research questions are: What is the predictive value of vulnerable carotid plaques detected by multi-sequence MRI for risk of (1) coronary heart disease and (2) stroke? The study is embedded in the Rotterdam Study. All subjects will undergo an ultrasound examination of the carotid arteries. Subjects with one or more carotid plaque of 3 mm or thicker on ultrasound will undergo an MRI scan to measure plaque structure. Based on an expected prevalence of carotid plaques of this thickness of 10%, 1000 subjects will be invited for an MRI scan. The risk of coronary heart disease and stroke of subjects with vulnerable plaques will be compared to that of subjects without carotid plaques and subjects with stable carotid plaques. Participating researchers Dept. of Radiology: Arlette Odink, Suzette Elias, Piotr Wielopolski, Quirijn van den Bouwhuisen Neurodegenerative diseases (Brain) Starting in September 2005, all participants in the Rotterdam Study have undergone MRI of the brain. The imaging protocol includes structural brain imaging for volumetric and shape analysis of various brain structures. This provides for assessing focal structural abnormalities - including brain infarcts and lacunes, white matter lesions and microbleeds. In addition, diffusion tensor imaging will yield quantitative information on the integrity of normal appearing white matter.

The main research questions are: How does vascular and degenerative brain pathology affect the development of dementia? What are the risk factors for cognitive decline and dementia? How can we predict an individual’s risk to develop dementia? A 30 minutes MRI protocol that strives to acquire high-resolution data has been defined. The new 1.5T MRI scanners with multichannel array technology speed up the protocols, provide more contrast capabilities, and deliver images with finer detail and better spatial coverage. The datasets have been selected to maximize the amount of information available within the imaging collection period and define features that can provide better automatic segmentation and volume quantification of WM lesions and brain tissues such as the cerebrospinal fluid (CSF), white matter and gray matter (see also the section on the Biomedical Imaging Group Rotterdam). Diffusion tensor imaging has also been incorporated to eventually identify connections between different parts of the brain and the potential modifications that may be seen in the fiber tracks as a result of WM lesions. To improve the visualization and detection of microbleeds we developed a custom-made higher-resolution accelerated 3D T2*GRE sequence. This sequence was tested in a sub-study with conventional 2D T2*GRE imaging in their ability to detect cerebral microbleeds. Over 3000 participants have already been scanned. The first analysis was focused on microbleeds, cerebral blood flow and DTI parameters (mean diffusivity and fractional anisotropy) and on association with cardiovascular risk factors. A segmentation tool for automated extraction of brain volume, grey and white matter volume and white matter

117

neurological & neurovascular imaging

lesion volume has been implemented. In the last year we have also focused on incidental findings. Due to the high resolution scan protocol, abnormalities were commonly found in the studied population. Participating researchers Dept. of Radiology: Piotr Wielopolski, Meike Vernooij, Marielle Poels, Wiro Niessen, Henri Vrooman Funding: Erasmus MC Research Grant 2006-2010: “Clinical relevance of cerebral microbleeds”

A

B

Functional MRI Principal Investigator: Aad van der Lugt, MD, PhD Functional magnetic resonance imaging (fMRI) allows visualization of the brain at work. It is at present the most commonly used functional neuroimaging technique due to its entirely noninvasive nature. Blood Oxygenation Level Dependent (BOLD) fMRI takes advantage of the tight link between local neuronal activity and blood flow (neurovascular coupling). When neuronal activity increases locally, local blood flow also increases, leading to an increase in oxygenated blood that is disproportionate to the increased need of oxygen for neuronal activity. As a result, local susceptibility effects, caused by the presence of paramagnetic deoxygenated hemoglobin, decrease, leading to a signal increase on T2* weighted images in those brain areas that are active. Since the start of the fMRI research program in the Department of Radiology in 2003, the collaboration with the Departments of Neurosciences, Neurology, Psychology, Rehabilitation Medicine and Psychiatry has led to a multitude of projects.

Figure 43. Incidental findings on brain MRI of non-symptomatic subjects.. Both panels show protondensity-weighted axial images with abnormalities indicated by arrows. Panel A shows an aneurysm of the anterior cerebral artery. Panel B shows a large meningioma in close proximity to the cavernous sinus and with encasement of the internal carotid artery

118

Eye movements: Optokinetic reflex and smooth pursuit (with the Department of Neurosciences)

Functional MRI of working memory and attention in patients with postconcussion syndrome (with the Department of Neurology)

Eye movements can be subdivided into several sub-types; i.e. vestibular reflexive, optokinetic reflexive, smooth pursuit, saccadic, and vergence eye movements. The combination and the interactions between these sub-types contribute to two basic functions of eye movements: stabilization of gaze during movements and shifting of gaze to new positions. We study brain activation patterns associated with the different types of eye movements and the interaction between the different types, with a focus on the cortex, the brainstem and the cerebellum. Five separate studies have been performed, two of which have now been accepted for publication.

Postconcussion syndrome comprises a large variety of complaints, such as headache, cognitive complaints, fatigue and depression. It affects 15-30% of patients after minor head injury. The etiology of this syndrome is unknown and diagnosis is difficult, since neurological and neuropsychological examinations are usually normal, as are conventional imaging studies with CT and MRI of the brain. Functional MRI is a promising new tool in both diagnosing this syndrome and understanding the underlying neural abnormalities. We postulate that postconcussion syndrome is caused by brain damage undetectable by conventional imaging techniques, and affects - amongst other things - cognitive functioning. Neural plasticity compensating for these abnormalities may explain normal neuropsychological tests, and can be detected with functional MRI as heightened neural activity. In this study 29 minor head injury patients and 12 healthy volunteers have undergone functional MRI to eval­uate differences in neural activity while performing tasks of working memory, continuous attention and selective attention. Publication of results is in process.

Participating researchers Dept. of Radiology: Caroline SchraaTam, Marion Smits

Figure 44. fMRI of reflexive saccade eye movement. Shown are activation clusters for reflexive saccades versus fixation. All areas were thresholded at p 30 mmHg Based on cardiovascular side-effects of growth hormone excess in diseases such as acromegaly, concerns have risen on the possible long-term cardiovas-

cular risks associated with high dosed growth hormone treatment. Turner syndrome is one of the currently wellestablished indications for growth hormone treatment during childhood. Data on cardiovascular condition after discontinuation of growth hormone therapy in Turner’s syndrome are lacking. Therefore the common purpose of chapters 6 and 7 was to assess the different aspects of functional cardiovascular condition in young adult women with Turner’s syndrome formerly treated with growth hormone. Participants represents a non-selective part of the Turner population from a previous Dutch growth hormone dose response study, based on voluntary participation in the described MRIevaluation study. Patient data was compared to findings in healthy age, gender and body surface area (m2) matched controls. MRI examinations were performed between May 2003 and June 2004. Chapter 1 provides an introduction to 1) the general problems seen during the follow-up of congenital heart diseases, 2) the specific problems seen with corrected TOF, and with Turner’s syndrome, and 3) diagnostic tools used throughout this thesis. At the end of the chapter (chapter 1.5) the specific study aims are presented. The specific study aims in tetralogy of Fallot were: - to assess clinical condition, including the assessment of biventricular volumes and function by cardiac MRI - to identify factors associated with RV dilation, RV dysfunction and decreased exercise tolerance - to assess RV diastolic function, in terms of RV filling pattern, at rest and during pharmacological stress - to evaluate the effect of restrictive RV physiology, indi-

cated by end-diastolic forward flow in the main pulmonary artery, on clinical condition - to assess the level of 1) neurohormonal activity, 2) biventricular contractile reserve, and 3) maximal exercise capacity, in relation to the amount of residual PR and RV volume - to assess pro-arrhythmogenic electrocardiographic changes in the surface ECG during maximal physical exercise - to explore possible relations between electrocardiographic changes during physical exercise with other determinants of clinical condition The specific study aims in Turner’s syndrome were: - to assess aortic dimensions and distensibility in young adults with Turner’s syndrome formerly treated with growth hormone - to assess the effect of growth hormone dose on aortic dimensions and distensibility - to assess biventricular size and function in young adults with Turner’s syndrome formerly treated with growth hormone - to evaluate the effect of growth hormone dose on biventricular size and function In chapter 2 clinical condition was assessed in 59 patients with corrected TOF, and risk factors associated with RV dilation, RV systolic dysfunction and decreased exercise tolerance were identified. Despite the common findings of RV dilation (85% of patients) and RV systo1ic dysfunction (49% of patients), chapter 2 demonstrated symptomatic status, exercise performance and rhythm status were rela-

179

dissertations

lively normal at mid- to long-term follow-up since correction, Severity of residual PR independently predicted the degree of RV dilation, but not of RV systolic dysfunction. Worse RV EF was independently predicted by abnormal wall motion of the RV outflow tract. Severity of both RV dilation and RV systolic dysfunction related to longer interval since repair, however the moderate associations found with interval since repair could not be used to predict the pace of decline in the individual patient. The most important predictor of worse exercise capacity was worse RVEF. Prolongation of QRSduration, which was expected to predict larger RV volumes, also independently predicted worse RVEF and worse exercise capacity, or in short functional decline. We suggest that further improvement of clinical condition may be obtained by improved preservation of RV outflow tract function (both in terms of preserved valvular function and in terms of preserved wall motion), and by a reduction of myocardial electrical inhomogeneity. In chapter 3 RV diastolic filling pattern at rest and during dobutannine stress were assessed in a non-selective subset of 36 patients. From previous studies the impact of restrictive RV physiology on clinical condition corrected TOF remains unclear. End-diastolic forward flow (EDFF) in the main pulmonary artery, a generally accepted sign of restrictive RV filling was shown to relate to worse exercise capacity, at mid- to long-term follow-up. Furthermore patients with EDFF had a higher mean PR-percentage and tended to have a larger RV compared to patients without EDFF. These data are in contrast with the previous concept that restrictive RV physiology ants protective against the detrimental effects of PR.

180

RVEF was comparable between patients with and without EDFF and could not be used to explain the observed difference in exercise capacity. Surprisingly, analyses of RV diastolic filling showed the RV filling pattern was most abnormal in patients without EDFF. We hypothesized stress imaging might provide additional information. Repeated measurements of RV diastolic inflow during low dose dobutamine stress showed no change in the RV filling pattern of patients without EDFF, while highly abnormal changes indicative of relaxation related dysfunction emerged in the RV filling pattern of patients with EDFF. We concluded dobutamine stress MRI may be used to ‘’unmask’’ abnormalities in RV diastolic filling not appreciated by rest imaging alone. Furthermore it was speculated that the highly abnormal RV filling pattern observed during stress in patients with EDFF contributed to our understanding of the observed difference in exercise limitation between subgroups. The most recent guidelines for LV heart failure management are partly based on the neurohormonal concept. Surprisingly few data are available on RV remodeling as a result of chronic PR in relation to neurohormonal activation and, or functional ventricular stress reserve. Chapter 4 describes studies in which 1) level of neurohormonal activation, 2) biventricular contractile reserve, and c) exercise performance were assessed in patients with TOF. In patients corrected at young age according to contemporary surgical strategies overall levels of neurohormonal markers were normal and biventricular contractile reserve was preserved, irrespective of the amount of residual PR or the subsequent degree of RV dilation.

Only a weak significant positive relation was found between the level of NT pro-BNP and PR-percentage. Risk levels of NT-proBNP were found in 22 (44%) patients. These patients showed smaller biventricular ESV with stress and smaller RV-CR compared to other patients. Furthermore patients with hsCRP levels above the median tended to have worse exercise capacity compared to the others. These results can be interpreted as a confirmation of the diagnostic potential of cardiac biomarkers as early signs of activated compensatory mechanisms in case of abnormal ventricular loading conditions. According to chapter 4, RV size and PR-percentage at best show poor correlations with impaired exercise capacity biventricular stress response or neurohormonal activation, in a population operated according to current surgical strategies. We therefore question the validity of PR or RV volume criteria for pulmonary valve replacement in this group and suggest patients should probably not undergo pulmonary valve replacement solely based on and volume or PR fraction criteria. Previous studies successfully identified indicators of electrical inhomogeneity in the resting ECG of patients with TOF to predict ventricular arrhythmia. In congenital heart disease exercise is considered a predisposing factor for the development of ventricular arrhythmia. Exercise testing may be a useful diagnostic tool to detect such arrhythmia. An extensive analysis of ECG changes with exercise was not yet performed in corrected TOF. Chapter 5 was used to assess pro-arrhythmogenic changes in the surface ECG of TOF patients and healthy controls during maximal physical exercise.

TOF patients showed significant lengthening of mean QTc duration during exercise, while in controls QTc remained unchanged. Furthermore both QTc and JTc dispersion increased in patients, while controls showed no change. At peak exercise mean JTc dispersion was significantly larger in patients compared to healthy controls and mean QTc dispersion tended to be larger. In addition a larger increase in QTc duration was found to correlate with a larger RV volume and with worse RV systolic function. Finally, a larger increase in JTc dispersion correlated with more severe PR. Although TOF patient showed no arrhythmia it may be speculated that the increase in inhomogeneity of repolarization indicates increased susceptibility to arrhythmia associated with physical exercise, especially in patients with severe residual PR. As such the data support exercise limiting advices given to TOF patients with severe residual PR. Using MRI aortic dimensions and distensibility at four predefined locations in young adult women with Turner’s syndrome formerly treated with growth hormone were studied in chapter 6. During childhood the patients studied randomly received 1, 1.5 or 2 times the regularly used growth hormone dose in Turner’s syndrome. In line with previous findings the ascending aortic diameter was enlarged in 30 - 50% of patients. In addition to previous data Turner patients were shown to have larger mean aortic diameters at all sites of measurement along the thoracic aorta. Compared to healthy controls aortic distensibility in Turner patients was smaller at 2 out of 4 levels of measurement.

181

dissertations

From this it was concluded that Turner patients, besides aortic dilation, showed signs of impaired wall distensibility. Our findings in Turner’s syndrome show similarity with results in Marfan’s syndrome and as such are in support of the hypothesis that connective tissue abnormalities may play a role in aortic disease in Turner’s syndrome Compared to healthy controls mean aortic diameters were larger at all thoracic levels of measurement in the subgroup of Turner patients that received the smallest dose of growth hormone. The other two subgroups only had a larger ascending aorta diameter compared to healthy controls. compared to controls aortic distensibility was smaller at 3 out of 4 levels of measurement in the subgroup that received the smallest dose of growth hormone, while no differences were found for the other two subgroups. As such results suggest severity of aortic abnormalities in young adulthood relate to growth hormone dose received during childhood, with a beneficial effect of a larger dose on aortic abnormalities. Chapter 7 of this thesis showed young adult women with Turner’s syndrome formerly treated with growth hormone do not have myocardial hypertrophy and show well preserved biventricular function, As such findings contradict the fear for pathologic myocardial remodeling with supraphysiologically dosed growth hormone treatment during childhood in Turner’s syndrome. Mean ventricular volumes especially that of the left ventricle, where found to be smaller in patients compared to controls, Participants to the present study are a non-selective subgroup of a larger postulation that participated in a Dutch growth hormone dose-response study. Together with the cardiac findings from this larger cohort, the cur-

182

rent data suggests disproportional cardiac growth in Turner’s syndrome during childhood, No differences where found between growth hormone subgroups for indices of cardiac size or function.

21 March 2007 Simone S Boks Graduate Advisors: Myriam GM Hunink, Sita MA Bierma-Zeinstra, and Dammis Vroegindeweij

Turner patients achieved comparable cardiac output to healthy controls, however at a higher mean resting heart rate. In Turner’s syndrome a higher mean resting heart rate has repeatedly been found. Left ventricular size may play a role in the underlying aetiology of a higher resting heart. From previous findings in Turner’s syndromes aetiology may be expected to be multi-factorial.

MR Imaging in Patients with Knee Injury: An Observational Study in General Practice

Considering the potential long-term (cardiovascular) mortality risk associated with a higher resting heart rate, we suggest future research in Turner’s syndrome to specifically address this topic. Finally in chapter 8 the main findings of this thesis are discussed. When applicable suggestions for future research are made.

Every year, many patients Seek medical treatment for sustained knee injury, The reported incidence in general practice ranges from 1.1 to 1.4%, since the introduction of MR imaging, several studies have reported on the appearance of normal and abnormal cruciate ligaments, collateral ligaments a no menisci. Compared to arthroscopy, which is considered the reference Standard, the sensitivity and specificity of MR imaging are high for dejecting abnormalities in these structures. Therefore, MR imaging is wideIy used to evaluate knee symptoms, and clinical decision-making is influenced by the results of these scans. It has been Shown that abnormalities in menisci, ligaments, and bone are found after knee trauma in approximately two thirds of patients. Such abnormalities (especially meniscal lesions) are also present in up to one third of asymptomatic knees, especially in those of older age. Therefore, in the trauma setting it is questionable which visualized lesions are the

resuit of the recent trauma and which lesions might have been pre-existent. There is little knowledge on the natural course and functional outcome of posttraumatic knee lesions. however, this information is necessary for making managements decisions and to inform patients about their prognosis. This study was conducted to provide additional information in this field of posttraumatic knee lesions, This thesis consists of four parts. Part I (Chapter 1) forms the introduction. Part I I (Chapters 2 and 3) presents two studies in which we systematically summarized the available literature addressing the natural course of posttraumatic knee lesions. Part III describes the design (Chapter 4) and the results (Chapters 5 to 8) of our MR imaging study on patients with sustained knee trauma presenting in primary care. Part IV (Chapters 9 and 10) contains the general discussion and summaries. In Chapter 2 we described the results of our systematic review of the literature on the natural course of posttraumatic ligamentous and meniscal knee lesions. Eleven

183

dissertations

studies were included: five concerning the posterior cruciate ligament (PCL), five concerning the anterior cruciate ligament (ACL), and one the menisci. Using a 14-item criteria list, only four studies received eight points or more and were considered ‘high quality’. The results showed that between 77% and 93% of the partial or complete PCL ruptures regained continuity at follow-up MRI Scanning. In case of partial or total ACL rupture, repair of continuity is also possible. There was a possible association between MRI continuity and clinical stability, Based on this review, no conclusions could be drawn about the natural course of meniscal or collateral ligament injury as seen on MRI.

posttraumatic side. Meniscal lesions and effusion were almost equally found in posttraumatic and contralateral knees. Multivariable analyses revealed that moderate to severe effusion was related to revert trauma and osteoarthritis, but not to history of trauma and age. The presence of any meniscal tear had a statistically sign if cant association with recent trauma, a history of old trauma and age, but not with osteoarthritis. History of trauma was more Strongly related to the group of radial, longitudinal and complex menisca1 tears than to horizontal tears. Recent trauma was not related to horizontal meniscal tears but was strongly related to others types of mensical tears.

In Chapter 3 we presented our review of the literature on the follow-up of MRldetected posttraumatic occult bone lesions. Eight of the 13 included studies scored eight points or more of the 15 available quality points and were considered ‘high quality’. The study results suggested a generally good clinical prognosis of bone bruises. Normalization of MR imageng appearance is possible and most often encountered after reticular lesions (i.e. serpiginous region of diminished T1W signal distant from subchondral bone plate, according to the Vellet classification). Cartilage loss at follow-up is often found in case of initial cartilage damage (impaction or osteochondral fracture).

In Chapter 6 we described the prevalence, severity, and determinants of bone bruise for the study population. In 81 of 136 posttraumatic knees (60%), 160 bone bruise lesions were found; 63% concerned reticular lesions, 55% geographic lesions, and 8% impaction, subcortical or osteochondral fractures. The lesions were almost equally divided over femur condyles and lateral and medial tibia plateau. Presence of bone bruise was related to medial meniscal lesions, anterior cruciate ligament, and medial collateral ligament tears. Bone bruises were more severe in the tibial or lateral location, in case of reticular lesion type, and in the presence of anterior cruciate ligament tears.

In Chapter 4 we described the design of our MRI study on patients presenting to their general practitioner with posttraumatic knee complaints. In Chapter 5 we found that, after sustained knee tracma, Iigamentous lesions were found almost exclusively at the

184

weeks. Healing was prolonged in case of a higher number of bone bruise lesions and osteoarthritis. Resolution of individual bone bruise lesions was prolonged in the presence of osteoarthritis and higher age. Reticular lesions were 1ess likely to be present after six months than other bone bruise types. Bone bruise lesion size and location, concomitant ligamentous or meniscal knee lesions, gender, obesity, workload, and Sports load had no prognostic value. In Chapter 8 we looked for a possible relationship between the presence of bone bruise in posttraumatic knees and pain. All patients underwent MR imaging of the posttraumatic knee at baseline and received a pain Score at baseline and approve lately three months, six months and 12 months after trauma. Patients with bone bruise underwent follow-up MR imaging. Firstly, we compared the pain scores at baseline of bone bruise patients to the baseline pain scores of non-bone bruise patients. Secondly, we analyzed the differences in the follow-up pain scores between participants with and without bone bruise at baseline. Thirdly, we investigated the relationship between pain and the presence of bone bruise at follow-up in patients with bone bruise on the baseline MRI scan. None of these analyses showed a statistically Significant relationship between the presence of bone bruise and pain. In Chapter 9 we have discussed the main findings of this thesis and placed them in a broader perspective.

In Chapter 7 we performed a Structured MR imaging followup Study of the patients with bone bruise in the posttraumatic knee. Follow-up was ceased when no bone bruise couId be discerned any longer, or after one year follow-up. The median healing time of bone bruise patients was 42.1

185

dissertations

22 March 2007 Marcel JL van Strijen Graduate Advisors: JL (Hans) Bloem (Leiden), Peter MT Pattynama, and Menno V Huisman (Leiden) Diagnosing Pulmonary Embolism: Establishing and Consolidating the Role of Spiral CT Chapter 1. The introduction gives the reasons for performing the Antelope study. The first Dutch consensus strategy was introduced in 1993. Although accurate and costeffective, the ‘Dutch consensus’ strategy was not widely applied. A survey in 1994 showed that not all hospitals had nuclear facilities available, and that Krypton ventilation scintigrapy was only limited available in other hospitals, Furthermore there was still reluctance to use pulmonary angiography in case of a non-high probability ventilation perfusion stratigraphy result. The diagnostic modalities, proposed in the consensus strategy were either not widely available or were considered invasive and therefore a potential risk for the patients clinically suspected for pulmonary embolism (PE). In 1996 the National Healthcare Council of

186

the Netherlands suggested to join several research proposals that were all in a way linked to the clinical problem of pulmonary embolism. This was the first step in the design of the Antelope study: a large multicenter trial for evaluating current strategies in diagnosing pulmonary embolism and the accuracy of the new technologies D-dimer, technegas stratigraphy and helical CT of the chest. Divided into two separate phases, the objective was to further clarify the role of recent developments in diagnostic technique, such as D-dimer testing, clinical decision rule and the use of spiral CT in patients clinically suspected of pulmonary embolism. In the first phase these diagnostic tests were prospectively compared to the current available consensus diagnostic strategy in The Netherlands in a specifically for this purpose designed algorithm. The results of this pad had to be used subsequently to design a new diagnostic algorithm that was to be more costeffective and clinically acceptable than the current poorly used strategy. The trial was named Antelope, an acronym that stands for Advances in New Tech-

nologies Evaluating Localization Of Pulmonary Embolism, The special focus of this thesis is the availability and use of spiral CT, the diagnostic accuracy compared to previously used tests and the additional value of this technique by means of transverse images and multiplane reconstructions of all anatomic structures of the thorax. Chapter 2. In 1999 a questionnaire was sent to all Dutch hospitals (n=128). The questionnaire contained separate sections with questions for the hospital management and the medical practitioners at the departments of radiology, nuclear medicine, internal medicine and pulmonology, Five hundred and eighty-four questionnaires were sent out to 128 hospitals with a total response rate of 68%. Fifty-four hospitals (43%) had no nuclear medicine facility, 13 (11%) had no pulmonary angiography facility, and 72 (59%) had no spiral CT scan. Seventy-two (57%) percent of the responding hospitals had a nuclear medicine facility, three of which used Technegas for ventilation studies. Twenty-two hospitals (18%) used the nuclear medicine facilities of a nearby hospital, Strategies with spiral CT were available in approximately 27% of the hospitals. Due to planned investments, on site availability of nuclear facilities could increase to approximately 55%. Strategies with Technegas were available in 2,4% of the hospitals, this number could increase to 25% if Technegas was proven accurate. The ‘Dutch consensus’ strategy was available in 68% of the hospitals. All other strategies were less feasible, because of lack of necessary diagnostic equipment.

Chapter 3. Apart from availability the questionnaire mentioned in chapter 2 also investigated the application of this consensus strategy. Therefore questions were posed to internists and pulmonologists regarding the number of patients with a clinical suspicion of pulmonary embolism, prevalence of pulmonary embolism and diagnostic strategy in the last patient. In total, 384 questionnaires were sent to internists and pulmonologists, with a response rate of 63% and 65%, respectively. The specialists reported to have followed the consensus strategy in 75% of the patients seen the month prior to the questionnaire. However, analysis of only the reposed last patient with the suspicion of pulmonary embolism revealed that the consensus strategy was followed in 55 of the 162 reported patients (34%). Both an overuse and an underuse of the different diagnostic facilities were documented. Furthermore almost a quarter of the patients were treated without an ascertained diagnosis, whereas 11% were not treated despite an improper exclusion of venous thrombo-embolism. Compared to a survey in 1994, the use of the ‘Dutch consensus strategy’ has not improved substantially. In 34% of the patients, the consensus strategy was strictly followed; however in only 67% of the patients a proper diagnosis was made. Chapter 4. Spiral CT has emerged as a potentially conclusive diagnostic test to exclude pulmonary embolism in patients with non-high probability perfusion ventilation stratigraphy (V/Q-scan) and is already widely used - sometimes as the sole primary diagnostic test in the diagnosis of suspected

187

dissertations

pulmonary embolism. lts true sensitivity and specificity has however not been evaluated previously in a large cohort of consecutive patients. Part of phase 1 of the Antelope study was a study that determined the sensitivity and specificity of helical CT, Patients with normal perfusion stratigraphy were excluded from further analysis, Single detector spiral CT scanning and ventilation stratigraphy were then performed in all patients to diagnose pulmonary embolism, while pulmonary angiography was performed as the gold standard. The only exception were those patients who had both a high-probability V/Q-scan and a CT scan positive for pulmonary embolism: these patients were considered to have pulmonary embolism and did not undergo additional pulmonary angiography. All imaging tests were read by independent expert panels. 517 patients were available for complete analysis. The prevalence of pulmonary embolism was 32%, Spiral CT correctly identified 88 of 128 patients with pulmonary embolism, and 92 of 109 patients without pulmonary embolism, for a sensitivity and specificity of 69% (95% C.I. 6375%) and 84% (95% C.I. 80-89%) respectively. The sensitivity of spiral CT was 86% (95% C.I. 80-92%) for segmental or Iarger puImonary embolism and 21% (95% C.I. 14-29%) in the group of patients with subsegmental pulmonary embolism. In this study the overall sensitivity of spiral CT for pulmonary embolism is too low to endorse its use as the sole test to exclude pulmonary embolism. This holds true even if one limits the discussion to patients with larger pulmonary embolism in segmental or larger pulmonary artery branches. We conclude that in patients with clinically suspected pulmonary embolism and an abnormal perfusion stratigraphy

188

single slice detector spiral CT is not sensitive enough to be used as the sole test to exclude pulmonary embolism. Chapter 5. In everyday practice spiral CT is used in a routine clinical setting, In most hospitals (ventilation) perfusion stratigraphy is still the first test of choice in patients suspected of pulmonary embolism, We prospectively studied the role of spiral CT in 279 patients suspected of pulmonary embolism and with an abnormal perfusion scan. All patients started their diagnostic algorithm with chest radiographs and perfusion stratigraphy. Depending on the results of perfusion stratigraphy patients proceeded to subsequent levels in the algorithm: stop if perfusion stratigraphy was normal; CT and pulmonary angiography if subsegmental perfusion defects were seen; ventilation stratigraphy followed by CT when segmental perfusion defects were seen; and pulmonary angiography in this last group when results of ventilation perfusion stratigraphy and CT were incongruent. The reference diagnosis was based on a normal perfusion stratigraphy, or a high probability perfusion/ventilation stratigraphy in combination with abnormal CT, or pulmonary angiography, If pulmonary embolism was present, the largest involved branch was noted on pulmonary angiography, or on spiral CT-scan in case of a high-probability ventilation-perfusion scan and a positive CT-scan. A distinction was made between embolism in a segmental branch or larger, or subsegmental embolism. In 27 patients spiral CT and/or pulmonary angiography was non-conclusive. Using spiral CT we correctly identified 117 of 135 patients with pulmonary embolism, and 106 out of 117 patients

without pulmonary embolism. Sensitivity and specificity is therefore 87% and 91%. Prevalence of pulmonary embolism is 53%. Positive and negative predictive values are respectively 91% and 86%. In the high probability group, sensitivity and specificity increase to 97% and 100% with a prevalence of 90%. In the non-high probability group sensitivity and specificity decrease to 61% and 89% with a prevalence of 25%. In a routine clinical setting current CT technology has limited value as a second diagnostic test because of low added value in patients with a high probability stratigraphy and low sensitivity in patients with non-high probability stratigraphy result. Chapter 6. Shortly after the introduction of spiral computed tomography (CT) for diagnosing pulmonary embolism (PE) in the early nineties, diagnostic accuracies of CT similar to that of pulmonary angiography were reported. In later series observed sensitivities and specificities varied considerably, a pad of which can be explained by observer agreement based on image quality. The aim of this study was to determine the relation of quality parameters of CT with observer agreement and sensitivity and specificity. The following three quality items were scored on standard questionnaires: 1. Quantitative contrast enhancement analysis 2. Qualitative contrast enhancement analysis and 3. Motion artifacts. To obtain a gold standard diagnosis the protocol required pulmonary angiography to be performed in all patients with a non-high probability V/Q-scintigram result and in patients with a high-probability V/Q scintigram result and

a normal CT scan result. Image parameters for CT image quality were scored in a subgroup of 129 CT scans. A diagnosis of PE was made in 128 of the 237 patients with a CT scan (54%) in whom the protocol was completed. The sensitivity and specificity of CT were respectively 69% and 84%. The overall image quality of the 129 CT scans was considered to be good in 109 (46%), moderate in 118 (5O%), and poor in 10 patients (4%), The sensitivity of CT with a good image quality was higher than in the group in which image quality was considered moderate or poor (p < 0.001), The specificity was not significantly different between the two groups. Sensitivity and specificity of CT in the detection of PE were higher when the two independent readers agreed than when a third observer was required (p = 0,045 and p < 0.001, respectively). The observer agreement on the diagnosis was better when image quality was good (87%) than when the image quality was considered moderate or poor (73%, p = 0.01). The observer agreement was influenced by contrast enhancement in the interlobar arteries (HU) (p=0,045), the presence of contrast medium in the aorta on the first slice (p=0,008), presence of contrast medium in the lower subsegmental pulmonary arteries (p=0,01), motion artifacts scored as being absent, few/moderate or severe (p=0,03), and motion artifacts that were considered to interfere with PE assessment (p=0,04). In a multiple regression analysis the following parameters were independently related to observer agreement: contrast enhancement in the interlobar arteries, presence of contrast medium in the lower subsegmental pulmonary arteries, and motion artifacts

189

dissertations

considered to interfere with PE assessment (p70%; hyperintense signal intensity on T2-weighted fat-suppressed sequences and early arterial enhancement. The surrounding liver parenchyma shows changed segmental distribution; steatosis and inhomogeneous patchy arterial enhancement. These results indicate that in patients with HCA, both lesions as well as surrounding liver parenchyma show similar soft tissue and vascular abnormalities. Chapter 4 illustrates three studies addressing the MR imaging findings of hepatocellular carcinoma (HCC), and a report of a rare hepatoid adenocarcinoma of the gallbladder as a mimicker of HCC. In chapter 4.1, the detailed imaging findings of developing hepatocellular carcinoma in cirrhosis at sequential MR imaging are described. In this study, stateof-the-art MR imaging displayed a spectrum of findings in the initial detection of developing HCC, including 1) localized fatty infiltration within a developing dysplastic nodule that gradually evolved into HCC in combination with a slowly increasing alpha-fetoprotein (AFP); 2) development of a focus of pathology-proven HCC with high signal intensity at T2-weighted imaging in a dysplastic nodule; 3) prominent neovasculature as the initial sign of developing, fast growing HCC, confirmed at histology after liver explantation. These findings provide insight in various pathways of step-wise carcinogenesis of developing hepatocellular carcinoma in cirrhosis, which may facilitate early detection, improve patient outcome, provide better identification of

198

patients for follow-up and further explain the genetic heterogeneity of hepatocellular carcinoma. In chapter 4.2, the relationship between the size of hepatocellular carcinoma and specific MR imaging findings is studied. Related to the improved MR imaging hardware, coils and software, accelerated imaging sequences have become available which allow detection and characterization of smaller (malignant) primary liver lesions. These developments have had their influence on the MR imaging findings, both because of different magnetization properties of the tissues related to changed pulse techniques, but also because lesions are detected when they are still small. In this study it is shown that smaller HCCs often show lower signal on fat-suppressed T2weighted sequences, more intense enhancement and less pronounced washout compared to larger tumors (typically >2 cm). In chapter 4.3, MR imaging findings are described in a young patient, who underwent Kasai portoenterostomy for biliary atresia as a neonate, and developed HCC as a rare complication. Kasai surgery has substantially increased the survival rates of this young patient group since its introduction in 1959, and consequently postpones subsequential liver transplantation. However, almost all patients develop features of end-stage liver disease and biliary cirrhosis. Since they are at risk for development of HCC, repeated sequential MR imaging exams of the native liver is necessary to monitor possible malignant transformation of liver nodules that may develop as a result of chronic cholestatic disease. Chapter 4.4 illustrates the MR imaging findings of a large gallbladder tumor in a patient with no known liver disease and strongly elevated AFP. Given the combination of tumor morphology and location in the gallbladder, signal intensity, contrast enhancement, high AFP and histol-

ogy, hepatoid adenocarcinoma (HAC) of the gallbladder was the most likely diagnosis. This is a rare form of adenocarcinoma with hepatoid differentiation of the tumor cells, which behaves and acts like HCC, but occurs in a multitude of organs outside the liver. Knowledge of these tumors is important, since HCC and HAC require a different therapeutic approach, especially in patients with non-cirrhotic liver disease. Chapter 5 presents two optimization studies which were performed to improve hepatobiliary imaging at higher magnetic field strength. The introduction of 3.0T MRI systems has provided a number of potential advantages for abdominal MR imaging, related to doubled signal-to-noise ratio (SNR) which allows faster imaging and improved resolution. However, imaging at 3.0T posed some challenges as well, including increased inhomogeneity of the magnetic field, changed magnetic tissue characteristics and increased specific absorption rate (SAR) levels, which potentially result in heating effects within the scanned subject. This is most pronounced for SAR-intensive sequences such as T2-weighted fast spinecho imaging. Since this sequence is essential for detection of focal liver lesions, the following two studies were designed to relieve some of the limitations for imaging at

3.0T whilst making use of the advantages of imaging at higher magnetic field strength. In chapter 5.1, variable-rate-selective excitation (VERSE) radiofrequency (RF) pulses were assessed and compared to standard RF pulses, for fat-suppressed T2-weighted fast spin-echo (FSE) liver MR imaging at 3.0T. The results of this study indicate that the use of VERSE RF pulses provided significantly increased slice coverage with lower SAR. Compared to 1.5T, the increased SNR on 3.0T enabled thin-slice volumetric imaging of the liver within a single breath-hold, with multiplanar reformatting possibilities as post-processing tool. In chapter 5.2, breath-hold diffusion-weighted

199

dissertations

black-blood echo-planar-imaging (BBEPI) was evaluated as potential alternative for SAR-intensive spin-echo sequences at 3.0T. The results of this study indicate that BBEPI can be used for ultrafast, low SAR, thin-slice morphologic imaging of the entire liver in a single breath-hold at 3.0T. Chapter 6 illustrates the current status of automated and color-coded post-processing techniques for analysis of dynamic multiphasic contrast-enhanced MR imaging of the liver. Post-processing of these images on dedicated workstations allows generation of time-intensity curves (TIC) as well as color-coded images, which provides useful information on (neo-) angiogenesis within a liver lesion, if necessary combined with information on enhancement patterns of surrounding liver parenchyma. Analysis of these images provides an easy to interpret schematic presentation of tumor behavior, providing additional characteristics for adequate differential diagnosis. Inclusion of TIC and color-coded images as part of the routine abdominal MR imaging work-up protocol may help to further improve the specificity of MR imaging findings and may facilitate the diagnostic work-up of disease for detection, staging, and monitoring of anti-tumor therapy. In chapter 7.1; and an overview article discussing the MR imaging findings of primary hepatocellular lesions in chapter 7.2.

200

24 October 2007 M Empar Rollano-Hijarrubia Graduate Advisors: Wiro J Niessen and Rik Stokking Imaging of Small High-Density Structures in Computed Tomography Computed tomography (CT) imaging of small high-density structures such as calsifications and scents is highly relevant in diagnosis, treatment planning, and follow-up of patients with cardiovascular disease, Calcification quantification is increasingly used as a potential marker for cardiovascular disease. However, owing to image blur, accurate calcification quantification is compromised. In addition, blur of high intensity structures hampers the visualization of surrounding soft tissue structures This for example affects the assessment of the degree of vessel stenches at locations close to calcified atherosclerotic plaques and or stunt struts. The research described in this thesis focuses on the improved of visualization and quantification of small high density structures, in particular calsifications. The contribution of this thesis is twofold. First, we study the capacity of modern Multi-row Detector CT (MDCT) scanners in visualizing and quantifying small high density

structures as a function of the imaging parameter settings. This study not only provides insight into the capacities and limitations of current scanners, but also provides information on how to optimize measurement accuracy of small objects and how to interpret image measurements. Second, we develop and evaluate a method to improve the visualization and quantification of atherosclerotic plaque containing calcifications. This method is based on the selective deconvolution of small highdensity structures in CT Angiography (CTA) image data. Below, we summarize in more detail the results obtained in the different chapters. In Chapter 2 different factors that affect the visualization and quantification of small high-density structures and their surrounding tissues are analyzed for a 64 MDCT scanner (Siemens AG, Medical Solutions, Erlangen, Germany). For this purpose, a custom made phantom consisting of Aluminum cylinders of known different diameters and lengths is scanned using different acquisition and reconstruction parameters. Then, the accuracy and uncertainty

201

dissertations

in measuring density, diameter, length, area and volume of each cylinder size is assessed. In addition, a bead phantom is imaged to estimate the Point Spread Function (PSF) for each imaging parameter setting, and simulations based on the PSF information are compared with the experimental results. Diameter, length, area and volume measurements are performed automatically using the 50% relative-treshold quantification method. We show that measurements on the simulated imaging data are in close agreement with those obtained in the experimental data. This indicates that, for objects with good Signal-to-Noise Ratio (SNR), errors in estimating object density and size mainly depend on the actual object size and the system PSF. Therefore, we hypothesize that the results obtained in this study for the 64 MDCT can be extended to other imaging systems and protocols just by accounting for the differences in the PSF. From the simulations and experiments we show that object density and size can accurately be determined (i.e, with a relative error smaller than ~5% for objects larger than two times the Full Width at Half Maximum (FWHM) of the PSF in each direction of space (i.e., for object sizes larger than ~2.0 mm, with current MDCT scanners). Below this size, the Hounsfield Units (HUs) of the inner object voxels decrease, and size measurements are overestimated due to the Partial Volume Effect (PVE) both in the axial and transaxial directions. These findings are in agreement with the work presented by Prevrhal et al. To characterize atherosclerotic plaque and to assess the lumen of small arteries, increasing the spatial resolution by applying sharper kernels and thinner slice thickness seems a logical option. However, there is a trade-off owing to the

202

associated increase in noise and artifacts. These two competing factors, which are partially patient dependent, make it very complicated to establish the optimal imaging parameter settings. In our experiments a good balance between spatial accuracy and SNR is achieved with a medium-sharp consultation kernel: i.e., B46. For this kernel, ringing artifacts are still negligible, and for the range of sizes larger than ~1 mm in each direction of space, the measurement accuracy rapidly increases with the object size (i.e., objects of 1.0 mm are overestimated by 11%, objects of 1.2 mm are overestimated by 5%, and objects larger than 2.5 mm are estimated with an error smaller than 4% In the range of sizes smaller than ~1 mm geometric measurements are not accurate: an average bias of 0.85 mm in each direction of space leads to large measurement overestimation in this range. For clinical applications aimed at volume quantification of high-contrast objects (such as calcifications), sharper kernels, e.g. B60, are recommended. These kernels increase accuracy and reduce threshold dependency because object spread is reduced. In Chapter 3 we investigate the differences in image noise and measurement accuracy for two consecutive generations of CT systems that are currently used in clinical practice: i.e., the 16 and 64 MDCT scanners. The accuracy in measuring object attenuation and object size is assessed in the range of sizes near and under the spatial resolution of the system. Quantitative analyses are based on two different thresholding methods: fixed-threshold based quantification and the 50% relative-threshold based quantification. Quantitative analyses with the 50% relative threshold show that, in the range of object sizes smaller than 2.0-2.5 mm,

measurement accuracy in the CT longitudinal direction is slightly, but not significantly, better for the 64 than for the 16 MDCT. Measurements in the transverse plane are very similar for the two scanners. For object sizes larger than 2.02.5 mm, both longitudinal and transverse geometric measurements are the same for the 16 and 64 MDCT scanners. However, when applying fixed-threshold quantification, measurement accuracy in the range of objects larger than ~1.5 mm is significantly better for the 64 MDCT. Measurement accuracy obtained when using the 50% relative threshold mainly depends on the object size and the image PSF. Therefore, for the MDCT scanners being evaluated (whose PSFs FWHM are ≤1 mm in each direction of space), measurement accuracy with the 50% relativethreshold is very high in the range of object sizes larger than ~2.0-2.5 mm. Above this interval measurements do net depend on the scanner and imaging protocol. However, with fixed thresholds geometric measurements are scanner and protocol dependent even for the range of sizes larger than ~2.0-2.5 mm. When using a fixed threshold, measurement accuracy depends not only on the object size and image PSF, but also on the attenuation values of the material surrounding the object, Another limitation of fixed thresholds is the missing of smaller/lower-density objects, even when their SNR is sufficiently high to be discriminated from their surrounding tissues. The results as presented in Chapter 2 and 3 provide insight into the accuracy limits of state-of-the-art MDCT scanners in imaging small high-density objects. In Chapter 2 it is shown that the measured PSF is an effective means to predict small object measurement accuracy for a given scanning system and imaging protocol. Thus, this enables the

interpretation of image measurements for structures in the range of sizes near or under the size of the scanner’s PSF. Chapter 3 shows that both geometric measurement errors and inter-scanner/inter-protocol measurement variability can be reduced by applying the 50% relative-threshold quantification, instead of conventional fixed- threshold based quantification methods. Moreover, the data of Chapter 3 can be used to understand and interpret differences in results obtained from clinical studies that have been conducted using different scanners or imaging protocols. The knowledge generated in Chapter 3 can assist to develop quantification algorithms aimed to reduce inter-scanner and inter-protocol measurement variability. Chapter 4 and 5 investigate the use of selective digital image deconvolution to improve the imaging of small highdensity structures, in particular calcifications Digital image deconvolution is frequently used in applications where the systems spatial resolution is insufficient to properly resolve small objects of interest from the background. Several authors have recently applied deconvolution to improve clinical diagnosis in CT. Most of the work has been aimed at improving visualization of small bony structures in human anatomy. Some studies suggested the application of this technique to enhance CTA vessel visualization. However, deconvolution to improve the visualization and quantification of atherosclerotic plaque calcifications has, to the best of our knowledge, not been performed and validated before. A number of deconvolution algorithms have been developed with the goal of reducing image blur while increasing the SNR. Most of them are based on an iterative minimization of a measure reflecting the difference between

203

dissertations

the original image and the estimate of the “true” object obtained in each iteration. Whereas these methods are powerful in improving the SNR, this comes at the expense of very long computational processing times. Computational requirements of 3D image deconvolution with a regularized Wiener filter are relatively small compared with iterative methods. This makes the Wiener Filter more advantageous for clinical practice. In Chapter 4 and 5 a novel method is proposed to locally deconvolve small high-density structures of the image, while avoiding noise amplification and edge-ringing artifacts on the remaining low-density tissues. This method, which we refer to as Histogram-based Selective Deblurring (HiSD), performs a strong 3D deconvolution with the Wiener Filter, and subsequently combines the image information of the original CT and its deconvolved data to generate a restored image. To combine the original and deconvolved image, three regions of interest are defined: a region where the original image intensity is used (background), a region where the deconvolved image is used (small high intensity structures) and a transition region between those two regions. In the original image this transition region corresponded to structures which were affected by the blur of the small high intensity structures. The definition of these regions is achieved by a combination of different thresholds. In Chapter 4, the HiSD method is applied and validated for CT images of phantom and in vitro carotid atherosclerotic plaques. In Chapter 5, HiSD is extended, and its performance is assessed on in vivo CTA images of carotid atherosclerotic plaques. The extension of the algorithm is twofold. First, the determination of the threshold values used to define the regions of interest is fully automated.

204

Second, the estimation of the HUs in the transition region is improved by applying a weighted interpolation of the original and deconvolved image. This results in a more realistic image pat-tern around the calcification. Moreover, small calcifications are not missed because both original and deconvolved calcification peaks contribute positively to the interpolated value. In order to apply HiSD on in vivo CTA data in Chapter 5, the problem of differentiating calcification from contrastenhanced lumen had to be addressed. This differentiation is required in order to allow a restoration of the whole calcification volume while avoiding deconvolution artifacts in the contrast-enhanced vessels. To achieve this, 3D vessel segmentation is performed prior to applying HiSD using a method introduced by Manniesing et al. In our study, fully automated lumen segmentation was successfully achieved in 13 out of 15 plaque images, and in two cases (minor) manual editing was required. Qualitative and quantitative analyses presented in Chapters 4 and 5 show that HiSD considerably improves the imaging of small high-density structures in phantom, in vitro, and in vivo enhanced CT images. After applying HiSD calcification blur is reduced both in the transverse and longitudinal direction, while noise amplification and edge-ringing artifacts are avoided in the surrounding low-density tissues. Consequently, calcification overrepresentation is decreased and calcification SNR is increased, resulting in improved visualization. With HiSD geometric measurements are significantly improved; reducing overestimation in the calcification volume, and reducing dependency of measurements on the quantification threshold. In Chapter 4, errors on the in vitro calcification area and volume measurements are

on average reduced by 31% and 44%, respectively, after applying HiSD. In Chapter 5, both volume overestimation obtained with low quantification thresholds values (