Simulation and synthesis in medical imaging : 4th international workshop, SASHIMI 2019, held in conjunction with MICCAI 2019, Shenzhen, China, October 13, 2019 : proceedings

This book constitutes the refereed proceedings of the 4th International Workshop on Simulation and Synthesis in Medical Imaging, SASHIMI 2019, held in conjunction with MICCAI 2019, in Shenzhen, China, in October 2019. The 16 full papers presented were carefully reviewed and selected from 21 submissions. The contributions span the following broad categories in alignment with the initial call-for-papers: methods based on generative models or adversarial learning for MRI/CT/PET/microscopy image synthesis, image super resolution, and several applications of image synthesis and simulation for data augmentation, segmentation or lesion detection. Preface 6 Organization 8 Contents 9 Empirical Bayesian Mixture Models for Medical Image Translation 11 1 Introduction 11 2 Methods 12 3 Experiments and Results 16 3.1 MRI Contrast Translation 16 3.2 MRI to CT Translation 19 4 Conclusion 20 References 21 Improved MR to CT Synthesis for PET/MR Attenuation Correction Using Imitation Learning 23 1 Introduction 23 2 Methods 25 2.1 Multi-hypothesis Learning 25 2.2 Imitation Learning 26 2.3 Proposed Network Architecture 26 2.4 Implementation Details 27 3 Experimental Datasets and Materials 27 4 Experiments and Results 28 5 Discussion and Conclusion 29 References 31 Unpaired Multi-contrast MR Image Synthesis Using Generative Adversarial Networks 32 1 Introduction 32 2 Method 34 2.1 Loss Functions 34 2.2 Network Architecture 36 3 Experiments and Results 36 3.1 Dataset 36 3.2 Implementation Details 37 3.3 Quantitative Results 37 3.4 Qualitative Results 37 4 Conclusion 38 References 40 Unsupervised Retina Image Synthesis via Disentangled Representation Learning 42 1 Introduction 42 2 Methodology 44 2.1 Disentanglement of Structure Features and Appearance Features 44 2.2 Loss Functions 46 3 Experiments and Results 47 3.1 Dataset 47 3.2 Technique Details 47 3.3 Qualitative Analysis 47 3.4 Quantitative Analysis 48 4 Discussion and Conclusion 49 References 50 Pseudo-normal PET Synthesis with Generative Adversarial Networks for Localising Hypometabolism in Epilepsies 52 1 Introduction 53 2 Methods 53 2.1 Stage 1: Network Architecture 53 2.2 Stage 2: Identification of Hypometabolic Regions 55 3 Experiments and Results 55 3.1 PET Synthesis from T1 in Healthy Control Data 55 3.2 Identification of Hypometabolism in Epilepsy Patients 57 4 Discussion and Conclusion 58 References 60 Breast Mass Detection in Mammograms via Blending Adversarial Learning 62 1 Introduction 62 2 Methodology 64 2.1 Blending Adversarial Networks 64 2.2 Adversarial Seamless Blending Supervision 65 2.3 Mass Localization 67 3 Experiments and Results 67 4 Conclusion 70 References 70 Tunable CT Lung Nodule Synthesis Conditioned on Background Image and Semantic Features 72 1 Introduction 72 2 Method 73 2.1 GAN Architecture 74 2.2 Loss Functions and Training Strategy 75 3 Experiment and Result 76 4 Conclusion 79 References 79 Mask2Lesion: Mask-Constrained Adversarial Skin Lesion Image Synthesis 81 1 Introduction 81 2 Method 82 2.1 Method Overview 82 2.2 Segmentation Masks 83 2.3 Image-to-Image Translation Network 83 3 Data and Experimental Details 84 4 Results and Discussion 86 5 Conclusion 88 References 89 Towards Annotation-Free Segmentation of Fluorescently Labeled Cell Membranes in Confocal Microscopy Images 91 1 Introduction 91 2 Method 93 2.1 Label Generation 93 2.2 Image Synthesis 93 3 Experiments 94 3.1 No Correspondence (Dnaive) 94 3.2 Global Shape Correspondence (Dglobal) 94 3.3 Local Structure Correspondence (Dlocal) 95 4 Results 95 4.1 Image Quality Assessment 95 4.2 Training with Synthesized Data 96 5 Conclusion 98 References 98 Intelligent Image Synthesis to Attack a Segmentation CNN Using Adversarial Learning 100 1 Introduction 100 2 Related Work 101 3 Our Approach 102 3.1 Model for Generating Adversarial Attack 103 3.2 Learning 103 3.3 Implementation Details 104 4 Experiments and Results 104 4.1 Adversarial Examples 105 4.2 Attacking the Segmentation Model 106 5 Discussion and Conclusion 107 References 108 Physics-Informed Brain MRI Segmentation 110 1 Introduction 110 2 Methodology 111 2.1 MRI Simulation Methods 112 2.2 Physics-Aware CNNs for Image Segmentation 113 2.3 Network Architecture 113 3 Experiments 113 3.1 Creation of Physics-Based Gold Standard Segmentation 113 3.2 Datasets 114 3.3 Simulation Experiment and Results 115 3.4 Robustness to Acquisition Parameters 115 3.5 Application to a Data Bridging Study 116 4 Discussion and Conclusions 118 References 119 3D Medical Image Synthesis by Factorised Representation and Deformable Model Learning 120 1 Introduction 120 2 Related Work 122 3 Proposed Approach 123 4 Experiment Details 125 4.1 Data and Pre-processing 125 4.2 Training Details 126 5 Results 126 6 Conclusion and Discussion 128 References 128 Cycle-Consistent Training for Reducing Negative Jacobian Determinant in Deep Registration Networks 130 1 Introduction 130 2 Related Work 132 3 Proposed Methods 133 3.1 Cycle Consistent Design 133 4 Experiment 134 4.1 Dataset 134 5 Conclusion 138 References 138 iSMORE: An Iterative Self Super-Resolution Algorithm 140 1 Introduction 141 2 Method 142 2.1 2D iSMORE 142 2.2 3D iSMORE and a New 3D Network 143 2.3 Modifications for MRI and Two-Photon Fluorescence Microscopy 144 3 Experiments 145 3.1 2D iSMORE on MRI from 3D Protocols 145 3.2 3D iSMORE on Two-Photon Fluorescence Microscopy 146 4 Conclusion and Discussion 148 References 149 An Optical Model of Whole Blood for Detecting Platelets in Lens-Free Images 150 1 Introduction 150 2 Optical Model 151 3 Platelet Detection 154 4 Testing and Validation 155 5 Conclusions 157 References 160 Evaluation of the Realism of an MRI Simulator for Stroke Lesion Prediction Using Convolutional Neural Network 161 1 Introduction 161 2 Methods 162 2.1 Clinical MRI 162 2.2 Description of the Perfusion MRI Simulator 163 2.3 Integration of the Arterial Flow Variability 163 2.4 Integration of the Contrast Agent Time of Transport 165 2.5 Evaluation of the Proposed Improvements 165 3 Experiments and Results 168 4 Conclusion and Perspectives 169 References 169 Author Index 171 Front Matter ....Pages i-x Empirical Bayesian Mixture Models for Medical Image Translation (Mikael Brudfors, John Ashburner, Parashkev Nachev, Yaël Balbastre)....Pages 1-12 Improved MR to CT Synthesis for PET/MR Attenuation Correction Using Imitation Learning (Kerstin Kläser, Thomas Varsavsky, Pawel Markiewicz, Tom Vercauteren, David Atkinson, Kris Thielemans et al.)....Pages 13-21 Unpaired Multi-contrast MR Image Synthesis Using Generative Adversarial Networks (Muhammad Sohail, Muhammad Naveed Riaz, Jing Wu, Chengnian Long, Shaoyuan Li)....Pages 22-31 Unsupervised Retina Image Synthesis via Disentangled Representation Learning (Kang Li, Lequan Yu, Shujun Wang, Pheng-Ann Heng)....Pages 32-41 Pseudo-normal PET Synthesis with Generative Adversarial Networks for Localising Hypometabolism in Epilepsies (Siti Nurbaya Yaakub, Colm J. McGinnity, James R. Clough, Eric Kerfoot, Nadine Girard, Eric Guedj et al.)....Pages 42-51 Breast Mass Detection in Mammograms via Blending Adversarial Learning (Chunze Lin, Ruixiang Tang, Darryl D. Lin, Langechuan Liu, Jiwen Lu, Yunqiang Chen et al.)....Pages 52-61 Tunable CT Lung Nodule Synthesis Conditioned on Background Image and Semantic Features (Ziyue Xu, Xiaosong Wang, Hoo-Chang Shin, Holger Roth, Dong Yang, Fausto Milletari et al.)....Pages 62-70 Mask2Lesion: Mask-Constrained Adversarial Skin Lesion Image Synthesis (Kumar Abhishek, Ghassan Hamarneh)....Pages 71-80 Towards Annotation-Free Segmentation of Fluorescently Labeled Cell Membranes in Confocal Microscopy Images (Dennis Eschweiler, Tim Klose, Florian Nicolas Müller-Fouarge, Marcin Kopaczka, Johannes Stegmaier)....Pages 81-89 Intelligent Image Synthesis to Attack a Segmentation CNN Using Adversarial Learning (Liang Chen, Paul Bentley, Kensaku Mori, Kazunari Misawa, Michitaka Fujiwara, Daniel Rueckert)....Pages 90-99 Physics-Informed Brain MRI Segmentation (Pedro Borges, Carole Sudre, Thomas Varsavsky, David Thomas, Ivana Drobnjak, Sebastien Ourselin et al.)....Pages 100-109 3D Medical Image Synthesis by Factorised Representation and Deformable Model Learning (Thomas Joyce, Sebastian Kozerke)....Pages 110-119 Cycle-Consistent Training for Reducing Negative Jacobian Determinant in Deep Registration Networks (Dongyang Kuang)....Pages 120-129 iSMORE: An Iterative Self Super-Resolution Algorithm (Can Zhao, Seoyoung Son, Yongsoo Kim, Jerry L. Prince)....Pages 130-139 An Optical Model of Whole Blood for Detecting Platelets in Lens-Free Images (Benjamin D. Haeffele, Christian Pick, Ziduo Lin, Evelien Mathieu, Stuart C. Ray, René Vidal)....Pages 140-150 Evaluation of the Realism of an MRI Simulator for Stroke Lesion Prediction Using Convolutional Neural Network (Noëlie Debs, Méghane Decroocq, Tae-Hee Cho, David Rousseau, Carole Frindel)....Pages 151-160 Back Matter ....Pages 161-162