Uncertainty for Safe Utilization of Machine Learning in Medical Imaging, and Graphs in Biomedical Image Analysis Second International Workshop, UNSURE 2020, and Third International Workshop, GRAIL 2020, Held in Conjunction with MICCAI 2020, Lima, Peru, Oc

This book constitutes the refereed proceedings of the Second International Workshop on Uncertainty for Safe Utilization of Machine Learning in Medical Imaging, UNSURE 2020, and the Third International Workshop on Graphs in Biomedical Image Analysis, GRAIL 2020, held in conjunction with MICCAI 2020, in Lima, Peru, in October 2020. The workshops were held virtually due to the COVID-19 pandemic. For UNSURE 2020, 10 papers from 18 submissions were accepted for publication. They focus on developing awareness and encouraging research in the field of uncertainty modelling to enable safe implementation of machine learning tools in the clinical world. GRAIL 2020 accepted 10 papers from the 12 submissions received. The workshop aims to bring together scientists that use and develop graph-based models for the analysis of biomedical images and to encourage the exploration of graph-based models for difficult clinical problems within a variety of biomedical imaging contexts. Additional Volume Editors 6 Preface UNSURE 2020 7 Organization 8 Preface GRAIL 2020 9 Organization 11 Contents 12 UNSURE 2020 15 Image Registration via Stochastic Gradient Markov Chain Monte Carlo 16 1 Introduction 16 2 Registration Model 17 3 Variational Inference 19 4 Stochastic Gradient Markov Chain Monte Carlo 19 5 Experiments 20 6 Discussion 23 7 Conclusion 23 References 24 RevPHiSeg: A Memory-Efficient Neural Network for Uncertainty Quantification in Medical Image Segmentation 26 1 Introduction 26 2 Methods 28 2.1 PHiSeg 28 2.2 Reversible Architectures 29 2.3 RevPHiSeg 29 3 Experimental Results 30 3.1 Evaluation Metrics 31 3.2 Datasets 32 3.3 Experimental Setup 33 3.4 Experimental Results 33 4 Discussion and Conclusion 34 References 34 Hierarchical Brain Parcellation with Uncertainty 36 1 Introduction 36 2 Methods 37 2.1 Flat Parcellation 38 2.2 Hierarchical Parcellation 38 2.3 Hierarchical Uncertainty 39 2.4 Architecture and Implementation Details 40 3 Experiments and Results 40 3.1 Data 40 3.2 Experiments 41 3.3 Results and Discussion 41 4 Conclusions 43 References 44 Quantitative Comparison of Monte-Carlo Dropout Uncertainty Measures for Multi-class Segmentation 45 1 Introduction 46 2 Methods 46 2.1 MC Dropout 46 2.2 Uncertainty Metrics 47 2.3 Evaluation 48 3 Experiments 50 4 Discussion and Conclusion 52 References 53 Uncertainty Estimation in Landmark Localization Based on Gaussian Heatmaps 55 1 Introduction 55 2 Heatmap Regression for Dataset-Based Uncertainty 56 3 Heatmap Fitting for Image-Based Uncertainty 57 4 Experimental Setup 58 5 Results and Discussion 59 6 Conclusion 63 References 63 Weight Averaging Impact on the Uncertainty of Retinal Artery-Venous Segmentation 65 1 Introduction 65 2 Data 67 3 Bayesian AV Classification 67 3.1 Baseline 67 3.2 Stochastic Weight Averaging 68 3.3 Stochastic Weight Averaging Gaussian 68 4 Experiments and Results 69 4.1 Description of Experiments 69 4.2 Performance of the Networks 70 4.3 Conclusions 72 References 73 Improving Pathological Distribution Measurements with Bayesian Uncertainty 74 1 Introduction 75 2 Method 76 2.1 Histopathological Measurements 76 2.2 Uncertainty Estimation 77 2.3 Datasets 77 2.4 Tissue Segmentation 78 3 Experiment Results 78 4 Conclusion 81 References 82 Improving Reliability of Clinical Models Using Prediction Calibration 84 1 Introduction 84 2 Prediction Calibration in Deep Models 85 3 Model Evaluation Using Reliability Plots 86 4 A New Prediction Calibration Objective 87 5 Experiments 89 5.1 Dataset and Problem Description 89 5.2 Model Design 89 5.3 Results 90 6 Conclusions 91 References 92 Uncertainty Estimation in Medical Image Denoising with Bayesian Deep Image Prior 94 1 Introduction 94 2 Related Work 96 3 Methods 97 3.1 Aleatoric Uncertainty with Deep Image Prior 97 3.2 Epistemic Uncertainty with Bayesian Deep Image Prior 98 3.3 Calibration of Uncertainty 98 4 Experiments 99 5 Results 100 6 Discussion and Conclusion 101 A Appendix 102 A.1 Additional Figures 102 A.2 Additional Tables 104 A.3 SGLD with Step Size Decay 105 A.4 Downsampling 106 A.5 Link Between Poisson Distribution and Normal Distribution 107 References 108 Uncertainty Estimation for Assessment of 3D US Scan Adequacy and DDH Metric Reliability 110 1 Introduction 111 2 Methods 112 2.1 Materials and Experimental Setup 112 2.2 Bone Segmentation and Metric Extraction 112 2.3 Uncertainty Estimation 114 3 Results and Discussion 114 4 Conclusions 117 References 117 GRAIL 2020 119 Clustering-Based Deep Brain MultiGraph Integrator Network for Learning Connectional Brain Templates 120 1 Introduction 121 2 Proposed Method 123 3 Results and Discussion 128 4 Conclusion 130 References 130 Detection of Discriminative Neurological Circuits Using Hierarchical Graph Convolutional Networks in fMRI Sequences 132 1 Introduction 133 2 Method 134 2.1 Proposed HD-GCN for Classification 135 2.2 Visualization of Affected Neurological Circuits 136 3 Experiments 137 3.1 Data 137 3.2 Experimental Results 138 4 Conclusion 140 References 141 Graph Matching Based Connectomic Biomarker with Learning for Brain Disorders 142 1 Introduction 142 2 Methods 144 2.1 Graph Edit Distance 144 2.2 Edit Cost Parameter Estimation 145 2.3 Interpretation of Dysfunction Coefficients 146 3 Experiments 146 3.1 Dataset and Preprocessing 146 3.2 Experimental Setup 147 3.3 Results and Discussions 148 4 Conclusion 150 References 151 Multi-scale Profiling of Brain Multigraphs by Eigen-Based Cross-diffusion and Heat Tracing for Brain State Profiling 153 1 Introduction 154 2 Proposed Eigen-Based Cross-diffusion and Heat Tracing of Brain Multigraphs 155 3 Results and Discussion 159 4 Conclusion 161 References 162 Graph Domain Adaptation for Alignment-Invariant Brain Surface Segmentation 163 1 Introduction 163 2 Method 166 2.1 Spectral Embedding of Brain Graphs 166 2.2 Graph Domain Adaptation 167 2.3 Network Architecture 168 3 Results 168 3.1 Effect of on Parcellation 169 3.2 Comparison with the State-of-the-art 170 4 Conclusion 172 References 173 Min-Cut Max-Flow for Network Abnormality Detection: Applicationpg to Preterm Birth 175 1 Introduction 175 2 Methods 176 2.1 Data 176 2.2 Tractography and Networks Extraction 177 2.3 Graph Similarity Measure 178 2.4 Graph Cut Optimisation for the Detection of Abnormal Connectivity 178 3 Results 181 4 Discussion 183 References 184 Geometric Deep Learning for Post-Menstrual Age Prediction Based on the Neonatal White Matter Cortical Surface 185 1 Introduction 185 2 Background 187 3 Methods 188 4 Experiments 192 4.1 Data 192 4.2 Implementation 192 5 Results 193 6 Conclusion and Discussions 194 References 195 The GraphNet Zoo: An All-in-One Graph Based Deep Semi-supervised Framework for Medical Image Classification 198 1 Introduction 199 2 GraphNet Zoo: An All-in-One Framework 200 3 Experimental Results 203 4 Conclusion 206 References 207 Intraoperative Liver Surface Completion with Graph Convolutional VAE 209 1 Introduction 209 2 Methods 210 2.1 Shape Generator 211 2.2 Shape Completion 213 3 Results 214 4 Conclusion 216 References 217 HACT-Net: A Hierarchical Cell-to-Tissue Graph Neural Network for Histopathological Image Classification 219 1 Introduction 220 2 Methods 221 2.1 Representation 221 2.2 HACT Graph Neural Networks (HACT-Net) 223 3 Experimental Results 224 3.1 Dataset 224 3.2 Implementation 225 3.3 Discussion 227 4 Conclusion 229 References 229 Author Index 231 Front Matter ....Pages i-xvii Front Matter ....Pages 1-1 Image Registration via Stochastic Gradient Markov Chain Monte Carlo (Daniel Grzech, Bernhard Kainz, Ben Glocker, Loïc le Folgoc)....Pages 3-12 RevPHiSeg: A Memory-Efficient Neural Network for Uncertainty Quantification in Medical Image Segmentation (Marc Gantenbein, Ertunc Erdil, Ender Konukoglu)....Pages 13-22 Hierarchical Brain Parcellation with Uncertainty (Mark S. Graham, Carole H. Sudre, Thomas Varsavsky, Petru-Daniel Tudosiu, Parashkev Nachev, Sebastien Ourselin et al.)....Pages 23-31 Quantitative Comparison of Monte-Carlo Dropout Uncertainty Measures for Multi-class Segmentation (Robin Camarasa, Daniel Bos, Jeroen Hendrikse, Paul Nederkoorn, Eline Kooi, Aad van der Lugt et al.)....Pages 32-41 Uncertainty Estimation in Landmark Localization Based on Gaussian Heatmaps (Christian Payer, Martin Urschler, Horst Bischof, Darko àtern)....Pages 42-51 Weight Averaging Impact on the Uncertainty of Retinal Artery-Venous Segmentation (Markus Lindén, Azat Garifullin, Lasse Lensu)....Pages 52-60 Improving Pathological Distribution Measurements with Bayesian Uncertainty (Ka Ho Tam, Korsuk Sirinukunwattana, Maria F. Soares, Maria Kaisar, Rutger Ploeg, Jens Rittscher)....Pages 61-70 Improving Reliability of Clinical Models Using Prediction Calibration (Jayaraman J. Thiagarajan, Bindya Venkatesh, Deepta Rajan, Prasanna Sattigeri)....Pages 71-80 Uncertainty Estimation in Medical Image Denoising with Bayesian Deep Image Prior (Max-Heinrich Laves, Malte Tölle, Tobias Ortmaier)....Pages 81-96 Uncertainty Estimation for Assessment of 3D US Scan Adequacy and DDH Metric Reliability (Arunkumar Kannan, Antony Hodgson, Kishore Mulpuri, Rafeef Garbi)....Pages 97-105 Front Matter ....Pages 107-107 Clustering-Based Deep Brain MultiGraph Integrator Network for Learning Connectional Brain Templates (Uğur Demir, Mohammed Amine Gharsallaoui, Islem Rekik)....Pages 109-120 Detection of Discriminative Neurological Circuits Using Hierarchical Graph Convolutional Networks in fMRI Sequences (Xiaodan Xing, Lili Jin, Qinfeng Li, Lei Chen, Zhong Xue, Ziwen Peng et al.)....Pages 121-130 Graph Matching Based Connectomic Biomarker with Learning for Brain Disorders (Rui Sherry Shen, Jacob A. Alappatt, Drew Parker, Junghoon Kim, Ragini Verma, Yusuf Osmanlıoğlu)....Pages 131-141 Multi-scale Profiling of Brain Multigraphs by Eigen-Based Cross-diffusion and Heat Tracing for Brain State Profiling (Mustafa Sağlam, Islem Rekik)....Pages 142-151 Graph Domain Adaptation for Alignment-Invariant Brain Surface Segmentation (Karthik Gopinath, Christian Desrosiers, Herve Lombaert)....Pages 152-163 Min-Cut Max-Flow for Network Abnormality Detection: Application to Preterm Birth (Hassna Irzan, Lucas Fidon, Tom Vercauteren, Sebastien Ourselin, Neil Marlow, Andrew Melbourne)....Pages 164-173 Geometric Deep Learning for Post-Menstrual Age Prediction Based on the Neonatal White Matter Cortical Surface (Vitalis Vosylius, Andy Wang, Cemlyn Waters, Alexey Zakharov, Francis Ward, Loic Le Folgoc et al.)....Pages 174-186 The GraphNet Zoo: An All-in-One Graph Based Deep Semi-supervised Framework for Medical Image Classification (Marianne de Vriendt, Philip Sellars, Angelica I. Aviles-Rivero)....Pages 187-197 Intraoperative Liver Surface Completion with Graph Convolutional VAE (Simone Foti, Bongjin Koo, Thomas Dowrick, João Ramalhinho, Moustafa Allam, Brian Davidson et al.)....Pages 198-207 HACT-Net: A Hierarchical Cell-to-Tissue Graph Neural Network for Histopathological Image Classification (Pushpak Pati, Guillaume Jaume, Lauren Alisha Fernandes, Antonio Foncubierta-Rodríguez, Florinda Feroce, Anna Maria Anniciello et al.)....Pages 208-219 Correction to: Graph Matching Based Connectomic Biomarker with Learning for Brain Disorders (Rui Sherry Shen, Jacob A. Alappatt, Drew Parker, Junghoon Kim, Ragini Verma, Yusuf Osmanlıoğlu)....Pages C1-C1 Back Matter ....Pages 221-222