Medical Image Computing and Computer Assisted Intervention – MICCAI 2021. 24th International Conference Strasbourg, France, September 27 – October 1, 2021 Proceedings Part VI

The eight-volume set LNCS 12901, 12902, 12903, 12904, 12905, 12906, 12907, and 12908 constitutes the refereed proceedings of the 24th International Conference on Medical Image Computing and Computer-Assisted Intervention, MICCAI 2021, held in Strasbourg, France, in September/October 2021.* The 531 revised full papers presented were carefully reviewed and selected from 1630 submissions in a double-blind review process. The papers are organized in the following topical sections: Part I: image segmentation Part II: machine learning - self-supervised learning; machine learning - semi-supervised learning; and machine learning - weakly supervised learning Part III: machine learning - advances in machine learning theory; machine learning - attention models; machine learning - domain adaptation; machine learning - federated learning; machine learning - interpretability / explainability; and machine learning - uncertainty Part IV: image registration; image-guided interventions and surgery; surgical data science; surgical planning and simulation; surgical skill and work flow analysis; and surgical visualization and mixed, augmented and virtual reality Part V: computer aided diagnosis; integration of imaging with non-imaging biomarkers; and outcome/disease prediction Part VI: image reconstruction; clinical applications - cardiac; and clinical applications - vascular Part VII: clinical applications - abdomen; clinical applications - breast; clinical applications - dermatology; clinical applications - fetal imaging; clinical applications - lung; clinical applications - neuroimaging - brain development; clinical applications - neuroimaging - DWI and tractography; clinical applications - neuroimaging - functional brain networks; clinical applications - neuroimaging – others; and clinical applications - oncology Part VIII: clinical applications - ophthalmology; computational (integrative) pathology; modalities - microscopy; modalities - histopathology; and modalities - ultrasound *The conference was held virtually. Preface Organization Contents – Part VI Image Reconstruction Two-Stage Self-supervised Cycle-Consistency Network for Reconstruction of Thin-Slice MR Images 1 Introduction 2 Method 2.1 Interpolation Network 2.2 The First-Stage SSL Based on Synthesized LR-HR Image Pairs 2.3 The Second-Stage SSL with Cycle-Consistency Constraint 3 Experiments 3.1 Dataset 3.2 Experimental Design 3.3 Implementation Details 3.4 Experimental Results 4 Conclusion References Over-and-Under Complete Convolutional RNN for MRI Reconstruction 1 Introduction 2 Methodology 3 Experiments and Results 4 Discussion and Conclusion References TarGAN: Target-Aware Generative Adversarial Networks for Multi-modality Medical Image Translation 1 Introduction 2 Methods 2.1 Proposed Framework 2.2 Training Objectives 3 Experiments and Results 3.1 Settings 3.2 Results and Analyses 4 Conclusion References Synthesizing Multi-tracer PET Images for Alzheimer's Disease Patients Using a 3D Unified Anatomy-Aware Cyclic Adversarial Network 1 Introduction 2 Methods 2.1 Evaluation with Human Data 3 Results 4 Conclusion References Generalised Super Resolution for Quantitative MRI Using Self-supervised Mixture of Experts 1 Introduction 2 Method 2.1 Data Description 2.2 Backbone Network Architecture 2.3 Self-supervised Mixture of Experts 3 Experiments 3.1 Implementation Details 3.2 Results 4 Discussion and Conclusion References TransCT: Dual-Path Transformer for Low Dose Computed Tomography 1 Introduction 2 Method 2.1 TransCT 2.2 Loss Function 2.3 Implementation 3 Experiments 3.1 Ablation Study 4 Conclusion References IREM: High-Resolution Magnetic Resonance Image Reconstruction via Implicit Neural Representation 1 Introduction 2 Method 2.1 Image Spatial Normalization 2.2 Model Optimization 2.3 HR Image Reconstruction 3 Experiments 3.1 Data 3.2 Implementation Details 3.3 Results 4 Conclusion References DA-VSR: Domain Adaptable Volumetric Super-Resolution for Medical Images 1 Introduction 2 Domain Adaptable Volumetric Super-Resolution 2.1 Network Structure 2.2 Self-supervised Adaptation 3 Experiments 3.1 Implementation Details 3.2 Dataset 3.3 Ablation Study 3.4 Quantitative Evaluation 4 Conclusion References Improving Generalizability in Limited-Angle CT Reconstruction with Sinogram Extrapolation 1 Introduction and Motivation 2 Problem Formulation 3 Proposed Method 3.1 HQS-CG Algorithm 3.2 Dual-Domain Reconstruction Pipelines 4 Experimental Results 4.1 Datasets and Experimental Settings 4.2 Ablation Study 4.3 Quantitative and Qualitative Results Comparison 5 Conclusion References Fast Magnetic Resonance Imaging on Regions of Interest: From Sensing to Reconstruction 1 Introduction 2 Methods 2.1 Problem Statement 2.2 Adaptive Sampler 2.3 Deep Reconstructor 2.4 Training Strategy 3 Implementations 4 Experiments 4.1 Data 4.2 Results 5 Conclusions References InDuDoNet: An Interpretable Dual Domain Network for CT Metal Artifact Reduction 1 Introduction 2 Method 2.1 Optimization Algorithm 2.2 Overview of InDuDoNet 3 Experimental Results 3.1 Ablation Study 3.2 Performance Evaluation 4 Conclusion References Depth Estimation for Colonoscopy Images with Self-supervised Learning from Videos 1 Introduction 2 Methodology 2.1 Training Baseline Model with Synthetic Data 2.2 Self-supervision with Colonoscopy Videos 3 Experiments 3.1 Dataset and Implementation Details 3.2 Quantitative Evaluation 3.3 Qualitative Evaluation on Real Data 4 Conclusion References Joint Optimization of Hadamard Sensing and Reconstruction in Compressed Sensing Fluorescence Microscopy 1 Introduction 2 Background 2.1 Fluorescence Microscopy and Hadamard Sensing 2.2 Sensing and Reconstruction Optimization 3 Proposed Method 3.1 End-to-End Sensing and Reconstruction Scheme 3.2 Loss Function 3.3 Implementation 4 Experiments 4.1 Masks 4.2 Reconstruction Methods 5 Conclusion References Multi-contrast MRI Super-Resolution via a Multi-stage Integration Network 1 Introduction 2 Methodology 2.1 Overall Architecture 2.2 Multi-stage Integration Module 3 Experiments 4 Conclusion References Generator Versus Segmentor: Pseudo-healthy Synthesis 1 Introduction 2 Methods 2.1 Basic GVS Flowchart 2.2 Improved Residual Loss 2.3 Training a Segmentor with Strong Generalization Ability 3 Experiments 3.1 Implementation Details 3.2 Evaluation Metrics 3.3 Comparisons with Other Methods 3.4 Ablation Study 3.5 Results on LiTS Dataset 4 Conclusions References Real-Time Mapping of Tissue Properties for Magnetic Resonance Fingerprinting 1 Introduction 2 Methods 2.1 Problem Formulation 2.2 Proposed Framework 2.3 Sliding-Window Stacking of Spirals 2.4 Learned Density Compensation 2.5 Tissue Mapping via Agglomerated Neighboring Features 3 Experiments and Results 4 Conclusion References Estimation of High Framerate Digital Subtraction Angiography Sequences at Low Radiation Dose 1 Introduction and Related Work 2 Methods 2.1 Phase Decomposition Using Independent Component Analysis 2.2 Training and Optimization 2.3 Network Details 2.4 Final Composition 3 Results 4 Conclusion References RLP-Net: A Recursive Light Propagation Network for 3-D Virtual Refocusing 1 Introduction 2 Proposed Method 2.1 Recursive Light Propagation Network (RLP-Net) 2.2 Training RLP-Net 3 Experiments 3.1 Fluorescence Microscopy Dataset 3.2 Training Details 3.3 Evaluation Results 4 Conclusion References Noise Mapping and Removal in Complex-Valued Multi-Channel MRI via Optimal Shrinkage of Singular Values 1 Introduction 2 Methods 2.1 Redundancy in MR Data 2.2 Optimal Shrinkage of Singular Value and Noise Estimation 2.3 Noise Mapping and Removal 3 Experiments 3.1 Numerical Validation 3.2 In-Vivo High-Resolution Diffusion MRI 3.3 In-Vivo Human Lung MRI 4 Conclusion References Self Context and Shape Prior for Sensorless Freehand 3D Ultrasound Reconstruction 1 Introduction 2 Methodology 2.1 Online Self-supervised Learning for Context Consistency 2.2 Online Adversarial Learning for Shape Constraint 2.3 Differentiable Reconstruction Approximation 2.4 Loss Function 3 Experiments 4 Conclusion References Universal Undersampled MRI Reconstruction 1 Introduction 2 Methods 2.1 The Overall Framework 2.2 Anatomy-SPecific Instance Normalization (ASPIN) 2.3 Model Distillation 2.4 Network Training Pipeline 3 Experimental Results 3.1 Datasets and Network Configuration 3.2 Algorithm Comparison 3.3 Ablation Study 3.4 Model Complexity 4 Conclusion References A Neural Framework for Multi-variable Lesion Quantification Through B-Mode Style Transfer 1 Introduction 2 Method 2.1 BQI-Net Architecture 2.2 Training Details 3 Experiments 3.1 Numerical Simulation 3.2 Phantom, and Ex-Vivo Measurements 4 Conclusion References Temporal Feature Fusion with Sampling Pattern Optimization for Multi-echo Gradient Echo Acquisition and Image Reconstruction 1 Introduction 2 Method 2.1 Deep ADMM as Backbone 2.2 Temporal Feature Fusion Block 2.3 Sampling Pattern Optimization Block 3 Experiments 3.1 Data Acquisition and Preprocessing 3.2 Implementation Details and Ablation Study 3.3 Performance Comparison 4 Conclusion References Dual-Domain Adaptive-Scaling Non-local Network for CT Metal Artifact Reduction 1 Introduction 2 Method 2.1 Problem Formulation 2.2 The Proposed DAN-Net 3 Experiments 3.1 Dataset 3.2 Implementation Details 3.3 Comparison with State-of-the-Art Methods 3.4 Clinical Study 4 Ablation Study 5 Conclusion References Towards Ultrafast MRI via Extreme k-Space Undersampling and Superresolution 1 Introduction 2 Related Work 3 Method 3.1 Data Description 3.2 Models 4 Experiments 5 User Study 6 Discussion and Conclusion References Adaptive Squeeze-and-Shrink Image Denoising for Improving Deep Detection of Cerebral Microbleeds 1 Introduction 2 Adaptive Squeeze-and-Shrink Denoising 2.1 Overall Framework 2.2 Implicit Near-Optimal Sparse Representation 2.3 The Denoising Procedure 3 The Deep Detection of Cerebral Microbleeds 4 Gaussian White Noise Removal 5 Experiments on Real-World Data 5.1 Data 5.2 CMB Detection 6 Discussion and Conclusions References 3D Transformer-GAN for High-Quality PET Reconstruction 1 Introduction 2 Methodology 2.1 Architecture 2.2 Objective Functions 2.3 Training Details 3 Experiments and Results 4 Conclusion References Learnable Multi-scale Fourier Interpolation for Sparse View CT Image Reconstruction 1 Introduction 2 Main Body 3 Experiments 4 Conclusion References U-DuDoNet: Unpaired Dual-Domain Network for CT Metal Artifact Reduction 1 Introduction 2 Additive Property for Metal Artifacts 3 Methodology 3.1 Network Architecture 3.2 Dual-Domain Cyclic Learning 4 Experiment 4.1 Experimental Setup 4.2 Comparison on Simulated and Real Data 4.3 Ablation Study 5 Conclusion References Task Transformer Network for Joint MRI Reconstruction and Super-Resolution 1 Introduction 2 Method 2.1 Task Transformer Network 2.2 Task Transformer Module 3 Experiments 4 Conclusion References Conditional GAN with an Attention-Based Generator and a 3D Discriminator for 3D Medical Image Generation 1 Introduction 2 Method 3 Experimental Results 4 Conclusion References Multimodal MRI Acceleration via Deep Cascading Networks with Peer-Layer-Wise Dense Connections 1 Introduction 2 Problem Formulation 3 Proposed Method 4 Experiments 5 Conclusion References Rician Noise Estimation for 3D Magnetic Resonance Images Based on Benford's Law 1 Introduction 2 Methodology 3 Experimental Setting 4 Results 5 Conclusions References Deep J-Sense: Accelerated MRI Reconstruction via Unrolled Alternating Optimization 1 Introduction 2 System Model and Related Work 2.1 Deep Learning for MRI Reconstruction 3 Deep J-Sense: Unrolled Alternating Optimization 4 Experimental Results 4.1 Performance on Matching Test-Time Conditions 4.2 Robustness to Test-Time Varying Acceleration Factors 4.3 Robustness to Train-Time Varying ACS Size 5 Discussion and Conclusions References Label-Free Physics-Informed Image Sequence Reconstruction with Disentangled Spatial-Temporal Modeling 1 Introduction 2 Methodology 2.1 Physics-Informed Loss Function 2.2 Spatial Modeling via GCNN 2.3 Temporal Modeling via Neural ODEs 3 Experiments 3.1 Synthetic Experiments 3.2 Clinical Data 4 Conclusion References High-Resolution Hierarchical Adversarial Learning for OCT Speckle Noise Reduction 1 Introduction 2 Method 2.1 Proposed Model 2.2 Dataset 2.3 Evaluation Metrics 3 Experiments and Results 3.1 Implementation Details 3.2 Ablation Study 3.3 Comparison Study 4 Discussion 5 Conclusion References Self-supervised Learning for MRI Reconstruction with a Parallel Network Training Framework 1 Introduction 2 Methods 2.1 Mathematical Model of CS-MRI Reconstruction 2.2 Brief Recap of ISTA-Net 2.3 Proposed Self-supervised Learning Method 2.4 Implementation Details 3 Experiments and Results 4 Conclusion References Acceleration by Deep-Learnt Sharing of Superfluous Information in Multi-contrast MRI 1 Introduction 2 Method 3 Results 4 Discussion 5 Conclusion References Sequential Lung Nodule Synthesis Using Attribute-Guided Generative Adversarial Networks 1 Introduction 2 Proposed Method 2.1 Model Architecture 2.2 Loss Functions 3 Experimental Results 3.1 Dataset and Implementation 3.2 Analysis of Lung Nodule Synthesis and Computation Costs 3.3 Visual Turing Test 3.4 Ablation Study 4 Conclusion References A Data-Driven Approach for High Frame Rate Synthetic Transmit Aperture Ultrasound Imaging 1 Introduction 2 Methods 2.1 Theory Basis and Network Architecture 2.2 Training Configurations 2.3 Simulations and In-Vivo Experiments 2.4 Metrics 3 Results 4 Conclusion References Interpretable Deep Learning for Multimodal Super-Resolution of Medical Images 1 Introduction 2 Sparse Modelling for Image Reconstruction 3 Deep Unfolding 4 A Multimodal Convolutional Deep Unfolding Design for Medical Image Super-Resolution 5 Experiments 6 Conclusion References MRI Super-Resolution Through Generative Degradation Learning 1 Introduction 2 Methods 2.1 Theory 2.2 GDN-Based SRR 2.3 Materials 2.4 Experimental Design 3 Results 4 Discussion References Task-Oriented Low-Dose CT Image Denoising 1 Introduction 2 Method 2.1 WGAN for LDCT Denoising 2.2 Analysis of Task-Oriented Loss 2.3 Training Strategy 3 Experiments 3.1 Datasets 3.2 Segmentation Networks 3.3 Implementation Details 3.4 Enhancement on Task-Related Regions 3.5 Boosting Downstream Task Performance 4 Conclusion References Revisiting Contour-Driven and Knowledge-Based Deformable Models: Application to 2D-3D Proximal Femur Reconstruction from X-ray Images 1 Introduction and Related Work 2 Method and Material 3 Results and Discussion 4 Conclusion References Memory-Efficient Learning for High-Dimensional MRI Reconstruction 1 Introduction 2 Methods 2.1 Memory-Efficient Learning 2.2 Memory-Efficient Learning for MoDL 2.3 Training and Evaluation of Memory-Efficient Learning 3 Results 4 Conclusions References SA-GAN: Structure-Aware GAN for Organ-Preserving Synthetic CT Generation 1 Introduction 2 Method 2.1 Global Stream in SA-GAN 2.2 Segmentation Network in the Local Stream 2.3 Organ Style Transfer with AdaON 3 Experiments and Results 4 Conclusion References Clinical Applications - Cardiac Distortion Energy for Deep Learning-Based Volumetric Finite Element Mesh Generation for Aortic Valves 1 Introduction 2 Methods 2.1 Template Deformation-Based Mesh Generation 2.2 Distortion Energy (Larap) 2.3 Weighted Larap (Lwarap) 3 Experiments and Results 3.1 Data Acquisition and Preprocessing 3.2 Implementation Details 3.3 Spatial Accuracy and Volumetric Mesh Quality 3.4 FE Stress Analysis During Valve Closure 3.5 Limitations and Future Works 4 Conclusion References Ultrasound Video Transformers for Cardiac Ejection Fraction Estimation 1 Introduction 2 Method 3 Experimentation 4 Conclusion References EchoCP: An Echocardiography Dataset in Contrast Transthoracic Echocardiography for Patent Foramen Ovale Diagnosis 1 Introduction 2 The EchoCP Dataset 2.1 Data Characteristics 2.2 PFO Diagnosis and Evaluation Protocol 3 Experiments of Baseline Method 4 Results and Analysis 5 Conclusion References Transformer Network for Significant Stenosis Detection in CCTA of Coronary Arteries 1 Introduction 2 Method 2.1 Semantic Feature Extraction for Local Cubic Volumes 2.2 Transformer Structure for Global Sequence Analysis 3 Experiment 3.1 Dataset 3.2 Experimental Results 4 Conclusion References Training Automatic View Planner for Cardiac MR Imaging via Self-supervision by Spatial Relationship Between Views 1 Introduction 2 Methods 3 Experiments 4 Conclusion References Phase-Independent Latent Representation for Cardiac Shape Analysis 1 Introduction 2 Methodology 2.1 Pre-processing Pipeline 2.2 Graph Representation of the LA 2.3 Design of Fusion and Classification Loss Function 3 Synthetic Experiments 3.1 Noisy Labels 4 Application to LAA Graphs 5 Conclusion References Cardiac Transmembrane Potential Imaging with GCN Based Iterative Soft Threshold Network 1 Introduction 2 Methodology 2.1 GISTA-Net Architecture 2.2 Implementation of Graph Convolution Network 3 Experiments 3.1 Ectopic Pacing Experiment 3.2 Myocardial Infarction Experiment 3.3 Cardiac Activation Sequence Reconstruction 4 Discussion 5 Conclusion References AtrialGeneral: Domain Generalization for Left Atrial Segmentation of Multi-center LGE MRIs 1 Introduction 2 Methodology 2.1 Image Segmentation Models 2.2 Domain Generalization Models 3 Materials 3.1 Data Acquisition and Pre-processing 3.2 Gold Standard and Evaluation 3.3 Implementation 4 Experiment 4.1 Comparisons of Different Semantic Segmentation Networks 4.2 Comparisons of Post- and Pre-ablation LGE MRI 4.3 Comparisons of Different Generalization Models 5 Conclusion References TVnet: Automated Time-Resolved Tracking of the Tricuspid Valve Plane in MRI Long-Axis Cine Images with a Dual-Stage Deep Learning Pipeline 1 Introduction 2 Methods 2.1 Imaging Data and Manual Annotation 2.2 Dual-Stage Residual Neural Network 2.3 Evaluation and Statistical Analysis 3 Experiments and Results 3.1 Implementation 3.2 Annotation Accuracy 4 Discussion and Conclusion References Clinical Applications - Vascular Deep Open Snake Tracker for Vessel Tracing 1 Introduction 2 Methods 2.1 Deep Open Curve Snake 2.2 Curve Proposal from Centerline Segmentation 2.3 Deep Snake Tracing 2.4 Global Tree Construction 3 Experimental Settings and Results 3.1 Datasets 3.2 Evaluation Metrics 3.3 Evaluation on BRAVE 3.4 Ablation Study 3.5 Adaptability of DOST on Other Datasets 4 Discussions and Conclusion References MASC-Units:Training Oriented Filters for Segmenting Curvilinear Structures 1 Introduction 2 Related Works 3 Methods 3.1 Rotatable MAC Unit and Response Shaping 3.2 Filter Re-use 3.3 Initialization Strategy 3.4 Multi-scale Processing with Pyramids 4 Experiments 5 Conclusion References Vessel Width Estimation via Convolutional Regression 1 Introduction 2 Method 2.1 Vessel Width Label Generation Method 2.2 Vessel Width Estimation Network 3 Dataset 3.1 Retinal Vessel Dataset for Width Estimation 3.2 Coronary Artery Dataset for Width Estimation 4 Experiment 4.1 Retinal Vessel Width Estimation 4.2 Coronary Artery Width Estimation 5 Conclusion References Renal Cell Carcinoma Classification from Vascular Morphology 1 Introduction 2 Related Works 2.1 Histopathological Images Dataset 2.2 Histopathological Images Classification 2.3 Hand-Crafted Features 3 Dataset 3.1 Dataset Building 3.2 VRCC200 4 Vascular Network Feature 4.1 Hand-Crafted Features 4.2 Deep Learning Feature 5 Experiments 5.1 Skeleton Features and Lattice Features Analysis 5.2 Vascular-Based RCC Classification Benchmark 6 Conclusion References Author Index The eight-volume set LNCS 12901, 12902, 12903, 12904, 12905, 12906, 12907, and 12908 constitutes the refereed proceedings of the 24th International Conference on Medical Image Computing and Computer-Assisted Intervention, MICCAI 2021, held in Strasbourg, France, in September/October 2021.* The 542 revised full papers presented were carefully reviewed and selected from 1809 submissions in a double-blind review process. The papers are organized in the following topical sections: Part I: image segmentation Part II: machine learning - self-supervised learning; machine learning - semi-supervised learning; and machine learning - weakly supervised learning Part III: machine learning - advances in machine learning theory; machine learning - domain adaptation; machine learning - federated learning; machine learning - interpretability / explainability; and machine learning - uncertainty Part IV: image registration; image-guided interventions and surgery; surgical data science; surgical planning and simulation; surgical skill and work flow analysis; and surgical visualization and mixed, augmented and virtual reality Part V: computer aided diagnosis; integration of imaging with non-imaging biomarkers; and outcome/disease prediction Part VI: image reconstruction; clinical applications - cardiac; and clinical applications - vascular Part VII: clinical applications - abdomen; clinical applications - breast; clinical applications - dermatology; clinical applications - fetal imaging; clinical applications - lung; clinical applications - neuroimaging - brain development; clinical applications - neuroimaging - DWI and tractography; clinical applications - neuroimaging - functional brain networks; clinical applications - neuroimaging - others; and clinical applications - oncology Part VIII: clinical applications - ophthalmology; computational (integrative) pathology; modalities - microscopy; modalities - histopathology; and modalities - ultrasound *The conference was held virtually.