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

Description

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. Preface Organization Contents – Part IV Image Registration Medical Image Registration Based on Uncoupled Learning and Accumulative Enhancement 1 Introduction 2 Method 2.1 Overview 2.2 Uncoupled Spatial Encoder 2.3 Accumulative Warping Enhancement 2.4 Multi-window Loss 3 Experiments and Results 4 Conclusion References Atlas-based Segmentation of Intracochlear Anatomy in Metal Artifact Affected CT Images of the Ear with Co-trained Deep Neural Networks 1 Introduction 2 Method 2.1 Data 2.2 Learning to Register the Artifact-Affected Images and the Atlas Image with Assistance of the Paired Artifact-Free Images 2.3 Network Architecture 2.4 Evaluation 3 Experiments 4 Results 5 Summary References Learning Unsupervised Parameter-Specific Affine Transformation for Medical Images Registration*-8pt 1 Introduction 2 Methods 2.1 Parameter-Specific Affine Transformation 2.2 Cross-stitch Affine Network 3 Experiments and Results 4 Conclusion References Conditional Deformable Image Registration with Convolutional Neural Network 1 Introduction 2 Methods 2.1 Conditional Deformable Image Registration 2.2 Conditional Image Registration Module 2.3 Self-supervised Learning 3 Experiments 4 Conclusion References A Deep Discontinuity-Preserving Image Registration Network 1 Introduction 2 Method 3 Experiments and Results 4 Conclusion References End-to-end Ultrasound Frame to Volume Registration 1 Introduction 2 Problem Definition 3 Method 3.1 End-to-end Slice-to-Volume Registration 3.2 Implementation Details 4 Experiments and Results 4.1 Datasets and Experimental Setting 4.2 Results Evaluation 5 Conclusions References Cross-Modal Attention for MRI and Ultrasound Volume Registration 1 Introduction 2 Method 2.1 Cross-Modal Attention 2.2 Feature Extraction and Deep Registration Modules 2.3 Implementation Details 3 Experiments and Results 3.1 Dataset and Preprocessing 3.2 Experimental Results 4 Conclusion References Bayesian Atlas Building with Hierarchical Priors for Subject-Specific Regularization 1 Introduction 2 Background: Atlas Building with Fast LDDMM 3 Our Model: Bayesian Atlas Building with Hierarchical Priors 3.1 Model Inference 4 Experimental Evaluation 5 Conclusion References SAME: Deformable Image Registration Based on Self-supervised Anatomical Embeddings 1 Introduction 2 Method 2.1 Self-supervised Anatomical Embedding (SAM) 2.2 SAM-Affine and SAM-Coarse 2.3 SAM-VoxelMorph 3 Experiments 4 Conclusion References Weakly Supervised Registration of Prostate MRI and Histopathology Images 1 Introduction 2 Methods 2.1 Data Acquisition 2.2 Overview of Proposed Method 2.3 Registration Neural Network 2.4 Transformation Model 2.5 Loss Functions 2.6 Previous Method 2.7 Evaluation Metrics 2.8 Experimental Design 3 Results 3.1 Qualitative Results 3.2 Quantitative Results 4 Discussion and Conclusion References 4D-CBCT Registration with a FBCT-derived Plug-and-Play Feasibility Regularizer 1 Introduction 2 Method 2.1 Overview 2.2 Feasibility Descriptor of Respiratory Motion 2.3 Unsupervised Learning for DVF Estimation 3 Experiments and Results 3.1 Training of the Feasibility Descriptor 3.2 Training of the DVF Inference Network 3.3 Evaluation 3.4 Results 4 Discussion and Conclusion References Unsupervised Diffeomorphic Surface Registration and Non-linear Modelling 1 Introduction 2 Methods 2.1 Registration Model 2.2 CVAE Network 2.3 Objective Function 3 Experiments and Results 3.1 Implementation Details 3.2 Validation of Registration 3.3 Validation of Internal Probabilistic Deformation Model 4 Discussion and Conclusion References Learning Dual Transformer Network for Diffeomorphic Registration 1 Introduction 2 Proposed Method 2.1 Volumetric Image Embedding 2.2 Dual Transformer 2.3 Diffeomorphic Registration 2.4 Unsupervised Learning 3 Experiments 3.1 Qualitative Assessment 4 Conclusion References Construction of Longitudinally Consistent 4D Infant Cerebellum Atlases Based on Deep Learning 1 Introduction 2 Method 2.1 Deformable Atlas Construction Network (DACN) 2.2 Affine Atlas Rescaling Network (AARN) 3 Experiments 4 Conclusion References Nesterov Accelerated ADMM for Fast Diffeomorphic Image Registration 1 Introduction 2 Diffeomorphic Image Registration 3 Nesterov Accelerated ADMM 4 Experimental Results 5 Conclusion References Spectral Embedding Approximation and Descriptor Learning for Craniofacial Volumetric Image Correspondence 1 Introduction 2 Method 2.1 Volumetric Image Descriptor 2.2 Spectral Embedding Approximation 2.3 Spectral Map-Based Correspondence 3 Experiments 3.1 Qualitative Assessment 4 Conclusion References A Deep Network for Joint Registration and Parcellation of Cortical Surfaces 1 Introduction 2 Method 2.1 Network Architecture 2.2 Loss Functions 2.3 Training Strategy 3 Experiments and Results 3.1 Experimental Setting 3.2 Results 4 Conclusion References 4D-Foot: A Fully Automated Pipeline of Four-Dimensional Analysis of the Foot Bones Using Bi-plane X-Ray Video and CT 1 Introduction 2 Method 2.1 Overview of the Proposed Pipeline 2.2 Automated Segmentation and Landmark Detection 2.3 2D-3D Registration Incorporating Landmark Reprojection Error 3 Experiment and Results 3.1 Experimental Materials 3.2 Evaluation of Automated Segmentation and Landmark Detection 3.3 Evaluation of 2D-3D Registration Using Bone Phantom 3.4 Evaluation of 2D-3D Registration Using Images of Real Subjects 4 Discussion and Conclusion References Equivariant Filters for Efficient Tracking in 3D Imaging 1 Introduction 1.1 Previous Work 2 Method 2.1 Construction of Equivariant Convolutional Filters 2.2 Registration of Equivaritant Filters 2.3 Loss Functions and Implementations 3 Experiments 4 Discussion and Conclusion References Revisiting Iterative Highly Efficient Optimisation Schemes in Medical Image Registration 1 Introduction/Motivation 2 Method 3 Experiments 3.1 Datasets 4 Results 5 Discussion and Conclusion References Multi-scale Neural ODEs for 3D Medical Image Registration 1 Introduction 2 Method 2.1 Learn Registration Optimizer via Neural ODEs 2.2 Pretrained Feature Extraction Network 3 Experiments and Results 3.1 Experiment Setup 3.2 Results 4 Discussions and Conclusions References Image-Guided Interventions and Surgery Self-supervised Generative Adversarial Network for Depth Estimation in Laparoscopic Images 1 Introduction 2 Methodology 2.1 Overview 2.2 Network Architecture 2.3 Training Losses 3 Experiments and Results 3.1 Dataset 3.2 Evaluation Metrics, Baseline, and Implementation Details 3.3 Results 4 Conclusions References Personalized Respiratory Motion Model Using Conditional Generative Networks for MR-Guided Radiotherapy 1 Introduction 2 Methods 2.1 Model Building 2.2 Model Personalization and Application 3 Experimental Setup and Results 4 Conclusion References Multimodal Sensing Guidewire for C-Arm Navigation with Random UV Enhanced Optical Sensors Using Spatio-Temporal Networks 1 Introduction 2 Materials and Methods 2.1 Device Fabrication with UV Enhanced Random Gratings 2.2 Event Trajectory Generation of Wavelength Data 2.3 Shape and Flow Networks 2.4 Refinement Network 2.5 Implementation Details 3 Results 3.1 Experimental Setup and Training Data 3.2 Synthetic and In-Vivo Experiments 4 Conclusion References Image-to-Graph Convolutional Network for Deformable Shape Reconstruction from a Single Projection Image 1 Introduction 2 Methods 2.1 Dataset and Problem Definition 2.2 Image-to-Graph Convolutional Network 2.3 Loss Functions 3 Experiments 4 Conclusion References Class-Incremental Domain Adaptation with Smoothing and Calibration for Surgical Report Generation 1 Introduction 2 Proposed Method 2.1 Preliminaries 2.2 Feature Extraction 2.3 Captioning Model 3 Experiments 3.1 Dataset 3.2 Implementation Details 4 Results and Evaluation 5 Discussion and Conclusion References Real-Time Rotated Convolutional Descriptor for Surgical Environments 1 Introduction 2 Related Works 3 Method 3.1 Network Structure 3.2 Matching 3.3 Model Training 4 Experiments 4.1 Model Accuracy 4.2 Model Speed 5 Conclusion References Surgical Instruction Generation with Transformers 1 Introduction 2 Methodology 2.1 Encoder-Decoder with Transformer Backbone 2.2 Reinforcement Learning 3 Evaluation 3.1 Experimental Settings 3.2 Implementation and Training Details 4 Results and Discussion 4.1 Comparison with the State-of-the-Art 4.2 Effects of Reinforcement Learning 4.3 Limitations and Challenges 5 Conclusion References Adversarial Domain Feature Adaptation for Bronchoscopic Depth Estimation 1 Introduction 2 Method 2.1 Supervised Depth Image and Confidence Map Estimation 2.2 Unsupervised Adversarial Domain Feature Adaptation 3 Experiments 4 Conclusion References 2.5D Thermometry Maps for MRI-Guided Tumor Ablation 1 Introduction 2 Material and Method 2.1 Image Acquisition 2.2 2.5D Thermometry Reconstruction 2.3 Evaluation 3 Results 4 Discussion and Conclusion References Detection of Critical Structures in Laparoscopic Cholecystectomy Using Label Relaxation and Self-supervision 1 Introduction 2 Methods 2.1 Critical Structures Identification via Label Relaxation 2.2 Pseudo-label Self-supervision 2.3 Implementation 3 Experiments and Results 3.1 Data and Training 3.2 Ablation Study 3.3 Qualitative Performance Across Surgery 3.4 Surgeon Preference 4 Discussion and Conclusion 4.1 Heatmaps Improve Accuracy, but Can Impair Visualisation 4.2 Self-supervision Particularly Helps in Difficult Videos 4.3 Conclusion References EMDQ-SLAM: Real-Time High-Resolution Reconstruction of Soft Tissue Surface from Stereo Laparoscopy Videos 1 Background 2 Method 2.1 EMDQ Tracking 2.2 g2o-based Graph Optimization 2.3 GPU-Based Dense Mosaicking and MBB Texture Blending 3 Results 4 Conclusion References Efficient Global-Local Memory for Real-Time Instrument Segmentation of Robotic Surgical Video 1 Introduction 2 Methodology 2.1 Dual-Memory Architecture 2.2 Efficient Local Temporal Aggregation 2.3 Active Global Temporal Aggregation 3 Experiments 4 Conclusion References C-Arm Positioning for Spinal Standard Projections in Different Intra-operative Settings 1 Introduction 2 Methods 2.1 Mobile C-arm Device 2.2 Training Data Simulation 2.3 Generation of Ground Truth Segmentations and Landmarks 2.4 K-Wire and Screw Simulation 2.5 Pose Regression Framework 2.6 Validation Data 3 Experiments and Results 3.1 Accuracy and Robustness Analysis 3.2 Generalization Analysis: DRRs to Real X-Rays Without Metal 3.3 Analysis of Domain Adaptation: DRRs to Real X-Ray with Spinal Implants 4 Discussion and Conclusion References Quantitative Assessments for Ultrasound Probe Calibration 1 Introduction 2 Methods 3 Experiments and Results 4 Discussion and Conclusion References Intra-operative Update of Boundary Conditions for Patient-Specific Surgical Simulation 1 Introduction 2 Method 3 Experiments and Results 3.1 Synthetic Adipose Tissue Manipulation 3.2 Real Adipose Tissue Manipulation 4 Discussion and Conclusion References Deep Iterative 2D/3D Registration 1 Introduction 2 Methods 2.1 Background 2.2 DL-Based Update Step Prediction 2.3 Loss Function 3 Experiments and Results 3.1 Data 3.2 Training 3.3 Evaluation 3.4 Results 4 Discussion and Conclusion References hSDB-instrument: Instrument Localization Database for Laparoscopic and Robotic Surgeries 1 Introduction 2 Data Collection 2.1 Laparoscopic Cholecystectomy 2.2 Robotic Gastrectomy for Gastric Cancer 3 Statistics of hSDB-instrument Dataset 4 Baseline Localization Performances 5 Conclusion References Co-generation and Segmentation for Generalized Surgical Instrument Segmentation on Unlabelled Data 1 Introduction 2 Methods 2.1 Network Details 2.2 Training Strategy 2.3 Loss Functions 3 Experiments 4 Results and Discussion 5 Conclusion References Surgical Data Science E-DSSR: Efficient Dynamic Surgical Scene Reconstruction with Transformer-Based Stereoscopic Depth Perception 1 Introduction 2 Related Work 3 Method 3.1 Light-Weight Stereo Depth Estimation and Tool Segmentation 3.2 Dynamic Reconstruction 4 Experiments 4.1 Experimental Setting 4.2 Qualitative Result 4.3 Quantitative Evaluation and Analysis 5 Conclusion References CataNet: Predicting Remaining Cataract Surgery Duration 1 Introduction 2 Approach 2.1 Model 2.2 Training Objectives 3 Experiments 3.1 Training and Test Data 3.2 Implementation and Baseline Methods 3.3 Results 4 Conclusion References Task Fingerprinting for Meta Learning in Biomedical Image Analysis 1 Introduction 2 Methods 2.1 Task Fingerprinting 2.2 Data 2.3 Experimental Design 3 Results 4 Discussion References Acoustic-Based Spatio-Temporal Learning for Press-Fit Evaluation of Femoral Stem Implants 1 Introduction 2 Materials and Method 2.1 Data Pre-processing and Spatio-Temporal Model for Press-Fit Evaluation 2.2 Experimental Setup and Data Generation 3 Results and Evaluation 3.1 Model Performance 3.2 Comparison with Non-sequence Data 4 Discussion 5 Conclusion References Surgical Planning and Simulation Deep Simulation of Facial Appearance Changes Following Craniomaxillofacial Bony Movements in Orthognathic Surgical Planning 1 Introduction 2 Method 2.1 Bony Movement Vector Estimation 2.2 FC-Net 3 Experimental Results 4 Conclusions References A Self-supervised Deep Framework for Reference Bony Shape Estimation in Orthognathic Surgical Planning 1 Introduction 2 Methods 3 Experimental Results 4 Discussion and Conclusion References DLLNet: An Attention-Based Deep Learning Method for Dental Landmark Localization on High-Resolution 3D Digital Dental Models 1 Introduction 2 Methods 2.1 High-Level Feature Extraction 2.2 Feature Fusion and Attention Heatmap 2.3 Implementation and Inference 3 Experiments 3.1 Data 3.2 Comparison Methods 3.3 Results 4 Conclusion References Personalized CT Organ Dose Estimation from Scout Images 1 Introduction 2 Patient-Specific Organ Dose from Scouts 3 Organ Dose Estimation with CT Scans 3.1 Monte Carlo Dose Estimation 3.2 Multi-organ Segmentation in Chest-Abdomen-Pelvis CT 4 Experimental Evaluations 4.1 Materials 4.2 Implementations 4.3 Results and Discussion 5 Conclusions References High-Particle Simulation of Monte-Carlo Dose Distribution with 3D ConvLSTMs 1 Introduction 2 Related Work 3 Methodology 3.1 Formulation of the Monte-Carlo Progressive Denoising Task 3.2 LSTM and ConvLSTM Cells 3.3 Proposed 3DConvLSTM MC Denoiser 4 Dataset Construction 4.1 Implementation Details 5 Experimental Results 6 Conclusions References Effective Semantic Segmentation in Cataract Surgery: What Matters Most? 1 Introduction 2 Materials and Methods 2.1 Data 2.2 Network Architectures 2.3 Loss Functions 2.4 Addressing Class Imbalance 2.5 Training Schedule 2.6 Implementation Details 3 Results and Discussion 4 Conclusion References Facial and Cochlear Nerves Characterization Using Deep Reinforcement Learning for Landmark Detection*-6pt 1 Introduction 2 Data 3 Methods 4 Results 5 Discussion and Conclusion References Patient-Specific Virtual Spine Straightening and Vertebra Inpainting: An Automatic Framework for Osteoplasty Planning*-10pt 1 Introduction 2 Methodology 3 Experimental Setup 4 Results 5 Discussion and Future Work References A New Approach to Orthopedic Surgery Planning Using Deep Reinforcement Learning and Simulation 1 Introduction 2 Methods 2.1 Analytical Representation of the Anatomy and Intervention 2.2 Environment 2.3 DRL Method Training and Evaluation 3 Results and Discussion 4 Conclusion References Whole Heart Mesh Generation for Image-Based Computational Simulations by Learning Free-From Deformations 1 Introduction 2 Methods 3 Experiments and Results 4 Conclusion References Automatic Path Planning for Safe Guide Pin Insertion in PCL Reconstruction Surgery 1 Introduction 2 Methods 2.1 Learning-Based Extraction of Anatomical Structures 2.2 Geometric Path Planning 2.3 Dataset and Training 3 Results 4 Discussion and Conclusion References Improving Hexahedral-FEM-Based Plasticity in Surgery Simulation 1 Motivation 2 Background 3 Methodology 4 Results and Discussion References Rapid Treatment Planning for Low-dose-rate Prostate Brachytherapy with TP-GAN 1 Introduction 2 Methods 2.1 Dataset 2.2 Seed Planning with TP-GAN 2.3 Post Processing Stage and Fine-Tuning with Simulated Annealing 3 Results and Discussions 3.1 Performance Analysis 3.2 Ablation Study 4 Conclusion References Surgical Skill and Work Flow Analysis Trans-SVNet: Accurate Phase Recognition from Surgical Videos via Hybrid Embedding Aggregation Transformer 1 Introduction 2 Method 2.1 Transformer Layer 2.2 Video Embedding Extraction 2.3 Hybrid Embedding Aggregation 3 Experiments 4 Conclusion References OperA: Attention-Regularized Transformers for Surgical Phase Recognition 1 Introduction 1.1 Related Work 2 Methodology 2.1 Sequential Transformer Network 2.2 Normalized Frame-Wise Attention 2.3 Attention Regularization 3 Experimental Setup 4 Results and Discussion 5 Conclusion References Surgical Workflow Anticipation Using Instrument Interaction 1 Introduction 2 Methodology 2.1 Task Formulation 2.2 Network Architecture 2.3 Instrument Interaction Module 2.4 Multi-stage Temporal Convolutional Network 3 Experiment Setup 3.1 Datasets and Preprocessing 3.2 Evaluation Metrics 4 Results and Discussions 4.1 Effect of IIM and Stages in MSTCN 4.2 Anticipation Results 4.3 Limitations 5 Conclusion References Multi-view Surgical Video Action Detection via Mixed Global View Attention*-10pt 1 Introduction 2 Related Work 3 Dataset 4 Model Architecture 4.1 Single-View RNN Model 4.2 Smart Fusion Model 5 Experimental Validation 6 Conclusion References Interhemispheric Functional Connectivity in the Primary Motor Cortex Distinguishes Between Training on a Physical and a Virtual Surgical Simulator 1 Introduction 2 Materials and Methods 2.1 Subjects and Experimental Design 2.2 Equipment 2.3 Data Processing for Oxy-Hemoglobin Time-Series 2.4 Mean and the Coefficient of Variation of Functional Connectivity Metrics from fNIRS HbO2 Time Series and FLS Performance Scores 3 Results 3.1 Physical Versus Virtual Simulator Effect on the Functional Connectivity 3.2 Statistical Analysis on Physical Versus Virtual Simulator Effect on the Individual Inter-regional Functional Connectivity 3.3 Brain – Behavior Correspondence 4 Discussion 5 Conclusion References Surgical Visualization and Mixed, Augmented and Virtual Reality Image-Based Incision Detection for Topological Intraoperative 3D Model Update in Augmented Reality Assisted Laparoscopic Surgery 1 Introduction 2 Related Work 3 Methodology 3.1 Overview 3.2 Image-Based Incision Detection 3.3 Image to Model Incision Transfer 3.4 Topological Model Update 3.5 3D-2D Registration 4 Results 4.1 Image-Based Incision Detection 4.2 Ex-Vivo Registration 5 Discussion and Future Work References Using Multiple Images and Contours for Deformable 3D-2D Registration of a Preoperative CT in Laparoscopic Liver Surgery 1 Introduction 2 Background 3 Methodology 3.1 Multi-View Rigid Base (MV-B) 3.2 Multi-View Rigid with Inter-image Correspondences (MV-C) 3.3 Multi-View Deformable Correspondences (MV-D) 4 Experimental Results 4.1 Rigidly-Related Views 4.2 Non-rigidly-related Views 5 Conclusions References SurgeonAssist-Net: Towards Context-Aware Head-Mounted Display-Based Augmented Reality for Surgical Guidance 1 Introduction 2 Methods 2.1 SurgeonAssist-Net: Surgical Task Prediction 2.2 Integrating SurgeonAssist-Net for Online Inference 2.3 Cholec80 Dataset 2.4 User-Centric Surgical Tasks Dataset 3 Results and Discussion 3.1 Cholec80: Surgical Task Prediction on a Benchmark Dataset 3.2 User-Centric Surgical Tasks Dataset: Task Prediction and Online Performance on the HoloLens 2 3.3 Clinical Significance 4 Conclusions and Future Work References Author Index