Medical Image Computing and Computer Assisted Intervention – MICCAI 2022: 25th International Conference, Singapore, September 18–22, 2022, Proceedings, Part II (Lecture Notes in Computer Science)

✍ Scribed by Linwei Wang; Qi Dou; P. Thomas Fletcher; Stefanie Speidel; Shuo Li

• Tongue: English
• Leaves: 802
• Category: Library

✦ Table of Contents

Preface
Organization
Contents – Part II
Computational (Integrative) Pathology
Semi-supervised Histological Image Segmentation via Hierarchical Consistency Enforcement
1 Introduction
2 Methodology
2.1 Basic Student Model for Supervised Learning
2.2 Hierarchical Consistency Enforcement (HCE) Module
2.3 Hierarchical Consistency Loss (HC-Loss)
2.4 Dataset and Evaluation Metrics
2.5 Implementation Details
3 Experiments and Results
3.1 Quantitative and Qualitative Comparison
3.2 Ablation Study of the Proposed Method
4 Conclusions
References
Federated Stain Normalization for Computational Pathology
1 Introduction
2 Related Work
2.1 Weight Aggregation
2.2 Stain Normalization
3 Method
3.1 Problem Statement
3.2 Network Architecture
3.3 Federated Learning
4 Evaluation
4.1 Dataset
4.2 Experimental Setup
4.3 Results
4.4 BottleGAN Architecture
5 Conclusion
References
DGMIL: Distribution Guided Multiple Instance Learning for Whole Slide Image Classification
1 Introduction
2 Method
2.1 Problem Formulation
2.2 Framework Overview
2.3 Self-supervised Masked Autoencoders for Feature Space Initialization
2.4 Cluster-Conditioned Feature Distribution Modeling
2.5 Pseudo Label-Based Feature Space Refinement
3 Experimental Results
3.1 Datasets
3.2 Implementation Details
3.3 Evaluation Metrics and Comparison Methods
3.4 Results
4 Ablation Study
5 Conclusions
References
ReMix: A General and Efficient Framework for Multiple Instance Learning Based Whole Slide Image Classification
1 Introduction
2 Method
2.1 Preliminary: MIL Formulation
2.2 ReMix
2.3 Intuitions on ReMix's Effectiveness
3 Experiments
3.1 Datasets and Metrics
3.2 Implementation Details
3.3 Comparisons
4 Conclusion
References
S3R: Self-supervised Spectral Regression for Hyperspectral Histopathology Image Classification
1 Introduction
2 Method
2.1 Spectral Regression Based on Low-Rank Prior
2.2 Model Based Spectral Regression
3 Experiments
3.1 Experimental Setup
3.2 Ablation Study
3.3 Comparison Between S3R and Other Methods
4 Conclusion
References
Distilling Knowledge from Topological Representations for Pathological Complete Response Prediction
1 Introduction
2 Methods
2.1 Persistent Homology
2.2 Topological Feature Representation and Its Computation
2.3 Distilling Betti Curves into DenseNet
3 Experiments
3.1 Dataset and Experimental Setup
3.2 Visualization of Betti Curves of Breast MRI Scan
3.3 Results of Breast Cancer Treatment Response Prediction
4 Conclusion
References
SETMIL: Spatial Encoding Transformer-Based Multiple Instance Learning for Pathological Image Analysis
1 Introduction
2 Methods
2.1 Overview of the Proposed Method
2.2 Position-Preserving Encoding
2.3 Transformer-Based Pyramid Multi-scale Fusion
2.4 Spatial Encoding Transformer
3 Experiments
4 Results and Discussion
5 Conclusion
References
Clinical-Realistic Annotation for Histopathology Images with Probabilistic Semi-supervision: A Worst-Case Study
1 Introduction
2 Method
3 Experiments and Results
4 Conclusion and Discussion
References
End-to-End Learning for Image-Based Detection of Molecular Alterations in Digital Pathology
1 Our Method: k-Siamese Networks
2 Data
2.1 The CancerScout Colon Data
2.2 TCGA Data
3 Experiments and Results
3.1 MSI Prediction
3.2 Detecting Molecular Alterations
4 Discussion and Conclusion
References
S5CL: Unifying Fully-Supervised, Self-supervised, and Semi-supervised Learning Through Hierarchical Contrastive Learning
1 Introduction
2 Method
2.1 Description of S5CL
2.2 Augmentations
2.3 Competitive Methods and Implementation Details
3 Results and Discussions
3.1 Evaluation on the Colon Cancer Histology Dataset
3.2 Evaluation on the Leukemia Single-Cell Dataset
3.3 Ablation Study
4 Conclusion
References
Sample Hardness Based Gradient Loss for Long-Tailed Cervical Cell Detection
1 Introduction
2 System Framework
3 Gradient Libra Loss
4 Experiments
4.1 Dataset and Evaluation Metrics
4.2 Implementation Details
4.3 Benchmark Results
4.4 Performance Analysis
4.5 Ablation Study
5 Conclusion
References
Test-Time Image-to-Image Translation Ensembling Improves Out-of-Distribution Generalization in Histopathology
1 Introduction
2 Method
2.1 StarGanV2 Architecture
2.2 Test-Time Image-to-Image Translation Ensembling
3 Experiments
3.1 Datasets
3.2 Method Evaluation
3.3 Ablation Studies
4 Conclusion
References
Predicting Molecular Traits from Tissue Morphology Through Self-interactive Multi-instance Learning
1 Introduction
2 Methodology
3 Experiments and Results
4 Conclusion
References
InsMix: Towards Realistic Generative Data Augmentation for Nuclei Instance Segmentation
1 Introduction
2 Method
2.1 Foreground Augmentation with Morphology Constraints
2.2 Background Perturbation for Robustness Improvement
2.3 Smooth-GAN for Realistic Instance Augmentation
3 Experiments
4 Conclusion
References
Improved Domain Generalization for Cell Detection in Histopathology Images via Test-Time Stain Augmentation
1 Introduction
2 Method
2.1 Problem Formulation
2.2 Test-Time Stain Augmentation for Improved Domain Generalization
2.3 Fusion of Augmented Detection Results
2.4 Implementation Details
3 Results
3.1 Data Description and Experimental Settings
3.2 Evaluation of Cell Detection Results
4 Conclusion
References
Transformer Based Multiple Instance Learning for Weakly Supervised Histopathology Image Segmentation
1 Introduction
2 Method
3 Experiment
3.1 Dataset
3.2 Implementation
3.3 Comparisons
3.4 Ablation Study
4 Conclusion
References
GradMix for Nuclei Segmentation and Classification in Imbalanced Pathology Image Datasets
1 Introduction
2 Methodology
2.1 GradMix
2.2 Network Architecture and Optimization
3 Experiments and Results
3.1 Datasets
3.2 Implementation Details
3.3 Results and Discussions
4 Conclusions
References
Spatial-Hierarchical Graph Neural Network with Dynamic Structure Learning for Histological Image Classification
1 Introduction
2 Methodology
2.1 Multi-level Entities Extraction
2.2 Dynamic Structure Learning
2.3 Spatial-Hierarchical Graph Neural Network
3 Experiments
3.1 Clinical Datasets and Evaluation Protocols
3.2 Comparison with State-of-the-Arts
3.3 Ablation Study
3.4 Visualization of Proposed Framework
4 Conclusion
References
Gigapixel Whole-Slide Images Classification Using Locally Supervised Learning
1 Introduction
2 Method
2.1 Auxiliary Model
2.2 Optimization
3 Experiments and Discussion
3.1 Datasets
3.2 Implementation Details
3.3 Results
4 Conclusion
References
Whole Slide Cervical Cancer Screening Using Graph Attention Network and Supervised Contrastive Learning
1 Introduction
2 Method
2.1 Representative Patches Extraction and Ranking
2.2 WSI-Level Classification with GAT
3 Experimental Results
4 Conclusion
References
RandStainNA: Learning Stain-Agnostic Features from Histology Slides by Bridging Stain Augmentation and Normalization
1 Introduction
2 Methodology
3 Experiments
4 Conclusion
References
Identify Consistent Imaging Genomic Biomarkers for Characterizing the Survival-Associated Interactions Between Tumor-Infiltrating Lymphocytes and Tumors
1 Introduction
2 Method
3 Experiments and Results
4 Conclusion
References
Semi-supervised PR Virtual Staining for Breast Histopathological Images
1 Introduction
2 Method
2.1 Registration and Label Assignment
2.2 PR Virtual Staining Based on Semi-supervised GAN
3 Experiments
3.1 Experimental Setup
3.2 Comparative Experiment Results and Analysis
3.3 Ablation Experiment Results and Analysis
4 Conclusions
References
Benchmarking the Robustness of Deep Neural Networks to Common Corruptions in Digital Pathology
1 Introduction
2 Benchmark Design
2.1 Formulation
2.2 Corruption Setup
2.3 Metric Setup
3 Experiments
3.1 Experimental Setup
3.2 Experimental Results
4 Conclusion
References
Weakly Supervised Segmentation by Tensor Graph Learning for Whole Slide Images
1 Introduction
2 Method
2.1 Superpixel Classification with Sparse Point Annotations
2.2 Multiple Feature Extraction
2.3 Tensor Graph Learning
2.4 Credible Node Reweighting and Optimization
3 Experiments
3.1 Datasets
3.2 Setup
3.3 Analysis
4 Conclusion
References
Test Time Transform Prediction for Open Set Histopathological Image Recognition
1 Introduction and Related Work
2 Methodology
2.1 Open Set Recognition - Max over Softmax as a Strong Baseline
2.2 Decoupled Color-Appearance Data Augmentation
2.3 Test-Time Transform Prediction and Open Set Recognition
3 Experimental Analysis
3.1 Datasets and Open Set Splits
3.2 Implementation Details and Performance Evaluation
3.3 Results and Discussion
3.4 Conclusion and Future Work
References
Lesion-Aware Contrastive Representation Learning for Histopathology Whole Slide Images Analysis
1 Introduction
2 Methods
2.1 Contrastive Learning Baseline
2.2 Lesion Queue Construction
2.3 Queue Refinement Strategy
3 Datasets
4 Experiments and Results
4.1 Experimental Setup
4.2 Structural Verification
4.3 Comparisons with State-of-the-Art Methods
5 Conclusion
References
Kernel Attention Transformer (KAT) for Histopathology Whole Slide Image Classification
1 Introduction
2 Method
2.1 Pre-processing and Data Preparation
2.2 Kernel Attention Transformer (KAT)
3 Experiment and Result
3.1 Model Verification
3.2 Comparison with Other Methods
4 Conclusion
References
Joint Region-Attention and Multi-scale Transformer for Microsatellite Instability Detection from Whole Slide Images in Gastrointestinal Cancer
1 Introduction
2 Method
2.1 Region-Level Region-Attention and Multi-scale Transformer
2.2 WSI-Level Region-Attention and Multi-scale Transformer
3 Experiments
3.1 Results
4 Conclusion
References
Self-supervised Pre-training for Nuclei Segmentation
1 Introduction
2 Related Works
3 Methodology
3.1 Self-supervised Pre-training with Unannotated Dataset
3.2 Fine-Tuning with Annotated Dataset
3.3 Implementations
4 Experiments
4.1 Dataset
4.2 Experimental Results
5 Conclusion
References
LifeLonger: A Benchmark for Continual Disease Classification
1 Introduction
2 LifeLonger Benchmark Definition
2.1 Multi-class Disease Datasets
2.2 Continual Learning Scenarios
2.3 Evaluation Criteria
3 Baseline Continual Learners
3.1 Implementation Details
4 Baseline Results
5 Conclusion
References
Unsupervised Nuclei Segmentation Using Spatial Organization Priors
1 Introduction
2 Related Work
3 Methodology
4 Experimental Configuration
4.1 Databases
4.2 Baselines
4.3 Implementation Details
5 Results and Discussion
6 Conclusion
References
Visual Deep Learning-Based Explanation for Neuritic Plaques Segmentation in Alzheimer's Disease Using Weakly Annotated Whole Slide Histopathological Images
1 Introduction
2 Methodology
2.1 Dataset Characteristics
2.2 Data Preparation
2.3 Deep Learning Architecture for Segmentation
2.4 Deep Learning Architecture for Visual Interpretation
3 Experiments and Results
3.1 Results from UNet Architecture
3.2 Visual Deep Learning Interpretation
4 Discussion and Conclusion
References
MaNi: Maximizing Mutual Information for Nuclei Cross-Domain Unsupervised Segmentation
1 Introduction
2 Related Works
2.1 Unsupervised Domain Adaptation
2.2 Contrastive Learning and Mutual Information
3 Methods
3.1 Problem Set-Up
3.2 Segmentation Loss and Mutual Information Maximization
3.3 Training
4 Experiments
4.1 Dataset and Implementation
4.2 Evaluation
4.3 Ablation Studies
5 Conclusion
References
Region-Guided CycleGANs for Stain Transfer in Whole Slide Images
1 Introduction
2 Related Work
3 Method
3.1 Region of Interest Discrimination
3.2 Library Generation for Region-Based Discrimination
3.3 Implementation and Training Details
4 Datasets
5 Experimental Results
5.1 Tile-Level Quantitative Results
5.2 Slide-Level Qualitative Results
6 Discussion
References
Uncertainty Aware Sampling Framework of Weak-Label Learning for Histology Image Classification
1 Introduction
2 Related Work
3 Methods
4 Experiments
4.1 Results
5 Conclusion
References
Local Attention Graph-Based Transformer for Multi-target Genetic Alteration Prediction
1 Introduction
2 Method
2.1 LA-MIL Framework
2.2 Local Attention Layer
3 Experiments
4 Results
5 Conclusion
References
Incorporating Intratumoral Heterogeneity into Weakly-Supervised Deep Learning Models via Variance Pooling
1 Introduction
1.1 Related Work
2 Incorporating Heterogeneity with Variance Pooling
2.1 Attention Mean Pooling Architecture
2.2 Attention Variance Pooling
3 Experiments with Survival Prediction
4 Interpretability and Biological Insights
4.1 Interpretability Visualizations
5 Conclusion
References
Prostate Cancer Histology Synthesis Using StyleGAN Latent Space Annotation
1 Introduction
2 Method
3 Experiments and Results
3.1 Patient Population
3.2 Tissue Processing
3.3 Tissue Annotation
3.4 GAN Training
3.5 Pathologist Annotation of Synthetic Histology
3.6 Generating Categorized Synthetic Images
4 Discussion
4.1 Limitations
5 Conclusions
References
Fast FF-to-FFPE Whole Slide Image Translation via Laplacian Pyramid and Contrastive Learning
1 Introduction
2 Methodology
2.1 Frequency Decomposition via Laplacian Pyramid
2.2 Contrastive Learning via Memory Bank
2.3 Optimization Objective
3 Experiment
3.1 Dataset and Implementation Details
3.2 Results
3.3 Microsatellite Instability Prediction
4 Conclusion
References
Feature Re-calibration Based Multiple Instance Learning for Whole Slide Image Classification
1 Introduction
2 Methods
3 Experiments
4 Results and Discussion
5 Conclusion
References
Computational Anatomy and Physiology
Physiological Model Based Deep Learning Framework for Cardiac TMP Recovery
1 Introduce
2 Methodology
2.1 Physiological Model
2.2 Kalman Update
2.3 KFNet
3 Experiments
3.1 Experimental Settings
3.2 Experimental Results
4 Conclusions
References
DentalPointNet: Landmark Localization on High-Resolution 3D Digital Dental Models
1 Introduction
2 Methods
2.1 RPN for Dental Landmarks
2.2 Bounding Box Refinement Network
2.3 Implementation and Inference
3 Experiments
3.1 Data
3.2 Evaluation Methods
3.3 Results
4 Conclusion
References
Landmark-Free Statistical Shape Modeling Via Neural Flow Deformations
1 Introduction
2 Methodology
2.1 Neural Flow Deformer
2.2 Training
2.3 Inference
3 Experiments
3.1 Experimental Setup
3.2 Generality and Specificity
3.3 Osteoarthritis Classification
3.4 Shape Reconstruction
4 Conclusion
References
Learning Shape Distributions from Large Databases of Healthy Organs: Applications to Zero-Shot and Few-Shot Abnormal Pancreas Detection
1 Introduction
2 Methods
3 Experiments
4 Discussion and Conclusion
References
From Images to Probabilistic Anatomical Shapes: A Deep Variational Bottleneck Approach
1 Introduction
2 Methods
2.1 VIB-DeepSSM Formulation
2.2 Baseline Models in Comparison
3 Results
3.1 Left Atrium Dataset
3.2 Training Scheme
3.3 Accuracy Analysis
3.4 Uncertainty Calibration Analysis
4 Conclusion
References
Opthalmology
Structure-Consistent Restoration Network for Cataract Fundus Image Enhancement
1 Introduction
2 Methodology
2.1 Synthesized Cataract Set with Identical Structures
2.2 High-Frequency Components with Structure Consistency
2.3 SCR-Net Architecture
3 Experiments
4 Conclusion
References
Unsupervised Domain Adaptive Fundus Image Segmentation with Category-Level Regularization
1 Introduction
2 Methodology
2.1 Inter-domain Category Regularization
2.2 Intra-domain Category Regularization
2.3 Training Procedure
3 Experiments
4 Conclusion
References
Degradation-Invariant Enhancement of Fundus Images via Pyramid Constraint Network
1 Introduction
2 Methodology
2.1 Sequence of Low-Quality Images with the Same Content
2.2 Laplacian Pyramid Features
2.3 Feature Pyramid Constraint
3 Experiments
3.1 Implementation Details
3.2 Comparison and Ablation Study
4 Conclusion
References
A Spatiotemporal Model for Precise and Efficient Fully-Automatic 3D Motion Correction in OCT
1 Introduction
2 Methods
3 Experiments and Results
4 Discussion
References
DA-Net: Dual Branch Transformer and Adaptive Strip Upsampling for Retinal Vessels Segmentation
1 Introduction
2 Our Proposed Method
2.1 Framework Overview
2.2 Dual Branch Transformer Module: Local Patches Meet Global Context
2.3 Adaptive Strip Upsampling Block
2.4 Implementation Details
3 Experiments
3.1 Comparison with Advanced Methods
3.2 Ablation Study
4 Conclusion
5 Future Work
References
Visual Explanations for the Detection of Diabetic Retinopathy from Retinal Fundus Images
1 Introduction
2 Methods
2.1 Datasets
2.2 Plain, Robust and Ensemble Models
2.3 Generating Visual Counterfactual Explanations (VCEs)
2.4 Saliency Maps
2.5 Model Evaluation
3 Results
3.1 Ensembling Plain and Adversarially Trained DNNs
3.2 VCEs as an Alternative to Saliency Maps
3.3 Sparsity Versus Realism of VCEs
3.4 VCEs for Different Budgets
4 Discussion
References
Multidimensional Hypergraph on Delineated Retinal Features for Pathological Myopia Task
1 Introduction
2 Method
2.1 Feature Extraction
2.2 Hypergraph Learning
3 Experiments
3.1 Dataset
3.2 Inputs Setup
3.3 Hypergraph Learning and Evaluation
3.4 Results
4 Conclusion
References
Unsupervised Lesion-Aware Transfer Learning for Diabetic Retinopathy Grading in Ultra-Wide-Field Fundus Photography
1 Introduction
2 Methodology
2.1 Overall Architecture
2.2 Adversarial Lesion Generation Module
2.3 Lesion External Attention Module
3 Experimental Results
3.1 Data Description
3.2 Implementation Details
3.3 Ablation Study
3.4 Comparisons with State-of-the-Art Methods
3.5 Effect of Number of CFP Images
4 Conclusion
References
Local-Region and Cross-Dataset Contrastive Learning for Retinal Vessel Segmentation
1 Introduction
2 Local-Region and Cross-Dataset Contrastive Learning
2.1 Overview of the Proposed Method
2.2 Local-Region Contrastive Learning
2.3 Cross-Dataset Contrastive Learning
3 Experiments
3.1 Dataset and Implementation
3.2 Ablation Study
3.3 Comparison with Other Advanced Methods
4 Conclusion
References
Y-Net: A Spatiospectral Dual-Encoder Network for Medical Image Segmentation
1 Introduction
2 Related Work
3 Method
3.1 Segmentation Framework
3.2 Spectral Encoder Components
3.3 Losses
4 Experiments
4.1 Experimental Setup
4.2 Results
4.3 Discussions and Conclusion
References
Camera Adaptation for Fundus-Image-Based CVD Risk Estimation
1 Introduction
2 Method
2.1 Cross-Laterality Feature Alignment Pre-training
2.2 Self-attention Camera Adaptor Module
3 Experiment
3.1 Dataset
3.2 Implementation Details and Metrics
3.3 Ablation Study
3.4 Quantitative Analysis
4 Conclusion
References
Opinions Vary? Diagnosis First!
1 Introduction
2 Methodology
2.1 Motivation
2.2 Learning DiagFirstGT
2.3 ExpG
3 Experiment
3.1 Experimental Settings
3.2 Comparing with SOTA
3.3 Ablation Study on ExpG
4 Conclusion
References
Learning Self-calibrated Optic Disc and Cup Segmentation from Multi-rater Annotations
1 Introduction
2 Theoretical Premises
3 Methodology
3.1 ConM
3.2 DivM
3.3 Supervision
4 Experiment
4.1 Implement Details
4.2 Main Results
4.3 Compared with SOTA
4.4 Ablation Study
5 Conclusion
References
TINC: Temporally Informed Non-contrastive Learning for Disease Progression Modeling in Retinal OCT Volumes
1 Introduction
2 Method
3 Experiments and Results
4 Discussion and Conclusion
References
DRGen: Domain Generalization in Diabetic Retinopathy Classification
1 Introduction
2 Related Work
2.1 DG in Medical Imaging
2.2 Stochastic Weight Averaging Densely
2.3 Fishr
3 Methodology
3.1 Baseline Adoption
3.2 Stochastic Weighted Domain Invariance
4 Implementation Details
4.1 Model Selection Method
5 Experiments and Results
6 Discussion
7 Conclusion
References
Frequency-Aware Inverse-Consistent Deep Learning for OCT-Angiogram Super-Resolution
1 Introduction
2 Method
2.1 Frequency-Aware Based Restoration and Degradation
2.2 Inverse-Consistency via CycleGAN
2.3 Loss Function and Optimization
3 Experiments
3.1 Dataset and Pre-processing
3.2 Implementation Details
3.3 Comparison with State-of-the-Art Methods
4 Conclusion
References
A Multi-task Network with Weight Decay Skip Connection Training for Anomaly Detection in Retinal Fundus Images
1 Introduction
2 Method
2.1 Weight Decay Skip Connection Training
2.2 HOG Prediction as an Auxiliary Task
2.3 Anomaly Detection
3 Experiments
3.1 Dataset and Implementation Details
3.2 Ablation Study
3.3 Comparison to State-of-the-Art Methods
4 Conclusion
References
Multiscale Unsupervised Retinal Edema Area Segmentation in OCT Images
1 Introduction
2 Methods
2.1 Primary DCCS
2.2 Unsupervised Image Clustering
2.3 Pseudo-Mask-Guided Pixel-Wise Segmentation
3 Experiments
3.1 Materials and Experimental Settings
3.2 Comparison Study
3.3 Ablation Study
4 Conclusion
References
SeATrans: Learning Segmentation-Assisted Diagnosis Model via Transformer
1 Introduction
2 Methodology
2.1 Asymmetric Multi-scale Interaction
2.2 SeA-Block
3 Experiment
3.1 Diagnosis Tasks
3.2 Experimental Settings
3.3 Main Results
3.4 Comparing with SOTA
3.5 Ablation Study
4 Conclusion
References
Screening of Dementia on OCTA Images via Multi-projection Consistency and Complementarity
1 Introduction
2 Method
2.1 The Consistency and Complementarity Attention (CsCp)
2.2 The Cross-View Fusion (CVF)
2.3 Objective Function
3 Experiments
3.1 Dementia Experiment
3.2 Extended Experiment
4 Conclusion
References
Noise Transfer for Unsupervised Domain Adaptation of Retinal OCT Images
1 Introduction
2 Proposed Method
2.1 Singular Value Decomposition-Based Noise Adaptation (SVDNA)
2.2 Training the Segmentation Network with SVDNA
3 Results
3.1 Data
3.2 Evaluation
4 Discussions and Limitations
References
Long-Tailed Multi-label Retinal Diseases Recognition via Relational Learning and Knowledge Distillation
1 Introduction
2 Method
2.1 Framework Overview
2.2 Multi-task Pre-training
2.3 Region-Based Attention
2.4 Relational Subsets Generation
2.5 Knowledge Distillation
2.6 Implementation Details
3 Experiments
3.1 Datasets and Evaluation Metrics
3.2 Quantitative Performance
3.3 Ablation Study
4 Conclusion
References
Fetal Imaging
Weakly Supervised Online Action Detection for Infant General Movements
1 Introduction
2 Methodology
2.1 Local Feature Extraction Module
2.2 Clip-Level Pseudo Labels Generating Branch
2.3 Online Action Modeling Branch
2.4 Training and Inference
3 Experiments
3.1 Main Results
3.2 Ablation Study
4 Conclusion
References
Super-Focus: Domain Adaptation for Embryo Imaging via Self-supervised Focal Plane Regression
1 Introduction
2 Related Work
3 Methods
4 Experiments
4.1 Qualitative Evaluation of Generated Image Quality
4.2 Impact on Embryo Grading
4.3 Impact on Cell Instance Segmentation
4.4 Impact of the Training Dataset
5 Conclusions
References
SUPER-IVIM-DC: Intra-voxel Incoherent Motion Based Fetal Lung Maturity Assessment from Limited DWI Data Using Supervised Learning Coupled with Data-Consistency
1 Introduction
2 Method
2.1 The ``Intra-voxel Incoherent Motion'' Model of DWI
2.2 SUPER-IVIM-DC
2.3 Implementation Details
2.4 Evaluation Methodology
3 Results
4 Conclusions
References
Automated Classification of General Movements in Infants Using Two-Stream Spatiotemporal Fusion Network
1 Introduction
2 Related Work
3 Proposed GMs Classification Method
3.1 Preprocessing Networks
3.2 Motion Classification Network with Two-Stream Architecture
4 Experiments
5 Conclusion
References
Author Index