Medical Image Computing and Computer Assisted Intervention – MICCAI 2023: 26th International Conference, Vancouver, BC, Canada, October 8–12, 2023, ... V (Lecture Notes in Computer Science, 14224)

✍ Scribed by Hayit Greenspan; Anant Madabhushi; Parvin Mousavi; Septimiu Salcudean; James Duncan; Tanveer Syeda-Mahmood; Russell Taylor

• Tongue: English
• Leaves: 844
• Category: Library

✦ Table of Contents

Preface
Organization
Contents – Part V
Computer-Aided Diagnosis I
Automatic Bleeding Risk Rating System of Gastric Varices
1 Introduction
2 Methodology
2.1 Segmentation Module
2.2 Cross-Region Attention Module
2.3 Region Constraint Module
2.4 Network Training
3 GVBleed Dataset
4 Experiments
4.1 Implementation Details
4.2 Results Analysis
5 Conclusions
References
You Don't Have to Be Perfect to Be Amazing: Unveil the Utility of Synthetic Images
1 Introduction
2 Deep Generative Models and Evaluation Metrics
3 Experimental Settings and Parameters
4 Experimental Results
4.1 The Proposed Metrics Match with Human Perception
4.2 The Trade-Off Between Fidelity and Variety
4.3 What Kind of Synthetic Data is Desired by Downstream Tasks When Privacy is Not an Issue?
4.4 What Kind of Synthetic Data is Desired by Downstream Tasks When Privacy is an Issue?
5 Conclusion
References
SHISRCNet: Super-Resolution and Classification Network for Low-Resolution Breast Cancer Histopathology Image
1 Introduction
2 Methods
2.1 Super-Resolution Module
2.2 Classification Module
2.3 Loss Function
3 Experiment
4 Results and Discussion
4.1 The Results of Super-Resolution and Classification
4.2 Ablation Study of the SHISRCNet
5 Conclusion
References
cOOpD: Reformulating COPD Classification on Chest CT Scans as Anomaly Detection Using Contrastive Representations
1 Introduction
2 Method
2.1 Patch-Level Representations Using Contrastive Learning
2.2 Generative Models Operating on Representation Space
3 Experiment Setup
4 Results
5 Discussion
6 Conclusion
References
YONA: You Only Need One Adjacent Reference-Frame for Accurate and Fast Video Polyp Detection
1 Introduction
2 Method
2.1 Foreground Temporal Alignment
2.2 Background Dynamic Alignment
2.3 Cross-Frame Box-Assisted Contrastive Learning
3 Experiments
3.1 Quantitative and Qualitative Comparison
3.2 Ablation Study
4 Conclusion
References
Personalized Patch-Based Normality Assessment of Brain Atrophy in Alzheimer's Disease
1 Introduction
2 Methods
2.1 Brain Surface Segmentation
2.2 Patch Similarity Metric
2.3 Personalized Template Set
3 Results
3.1 Normality Assessment Experiments
3.2 CN vs MCI, AD Prediction Experiment
4 Conclusion
References
Patients and Slides are Equal: A Multi-level Multi-instance Learning Framework for Pathological Image Analysis
1 Introduction
2 Method
2.1 Overview
2.2 Slide-Patch Level MIL
2.3 Patient-Slide Level MIL
3 Experiments and Results
3.1 Dataset and Evaluation
3.2 Implementation Details
3.3 Comparisons and Results
4 Limitations
5 Conclusion
References
Liver Tumor Screening and Diagnosis in CT with Pixel-Lesion-Patient Network
1 Introduction
2 Method
2.1 Preliminary on Mask Transformer
2.2 Pixel-Lesion-Patient Network (PLAN)
3 Experiments
4 Conclusion
References
Self- and Semi-supervised Learning for Gastroscopic Lesion Detection
1 Introduction
2 Methodology
2.1 Hybrid Self-supervised Learning
2.2 Prototype-Based Pseudo-label Generation Method
3 Datasets
4 Experiments
5 Conclusion
References
DiffULD: Diffusive Universal Lesion Detection
1 Introduction
2 Method
2.1 Diffusion-Based Detector for Lesion Detection
2.2 Training
2.3 Inference
2.4 Backbone Design
3 Experiments
3.1 Settings
3.2 Lesion Detection Performance
3.3 Ablation Study
4 Conclusion
References
Graph-Theoretic Automatic Lesion Tracking and Detection of Patterns of Lesion Changes in Longitudinal CT Studies
1 Introduction
2 Method
2.1 Problem Formalization
2.2 Lesion Matching Computation
2.3 Classification of Changes in Lesions and in Patterns of Lesion Changes
3 Experimental Results
4 Conclusion
References
Learning with Synthesized Data for Generalizable Lesion Detection in Real PET Images
1 Introduction
2 Methodology
2.1 Synthesized Data Augmentation
2.2 Patch Gradient Reversal
3 Experiments
4 Conclusion
References
Robust Exclusive Adaptive Sparse Feature Selection for Biomarker Discovery and Early Diagnosis of Neuropsychiatric Systemic Lupus Erythematosus
1 Introduction
2 Method
2.1 Sparse Coding Framework
2.2 Generalized Correntropic Loss
2.3 Generalized Correntropy-Induced Exclusive 2,1
3 Experimental Results and Conclusion
References
Multi-view Vertebra Localization and Identification from CT Images
1 Introduction
2 Methodology
2.1 DRR Multi-view Contrastive Learning
2.2 Single-view Vertebra Localization
2.3 Single-View Vertebra Identification
2.4 Multi-view Fusion
3 Experiments and Results
3.1 Dataset and Evaluation Metric
3.2 Implementation Details
3.3 Comparison with SOTA Methods
3.4 Ablation Study
4 Conclusion
References
Cluster-Induced Mask Transformers for Effective Opportunistic Gastric Cancer Screening on Non-contrast CT Scans
1 Introduction
2 Related Work
3 Methods
4 Experiments
4.1 Experimental Setup
4.2 Results
5 Conclusion
References
Detecting Domain Shift in Multiple Instance Learning for Digital Pathology Using Fréchet Domain Distance
1 Introduction
2 Methods
2.1 MIL Method
2.2 MIL Features
2.3 Fréchet Domain Distance
3 Datasets
4 Experiments
4.1 MIL Training
4.2 Domain Shift Quantification
5 Results
6 Discussion and Conclusion
References
Positive Definite Wasserstein Graph Kernel for Brain Disease Diagnosis
1 Introduction
2 Methods
2.1 Data and Preprocessing
2.2 Sliced Wasserstein Graph Kernel
2.3 Sliced Wasserstein Graph Kernel Based Learning
3 Experiments
3.1 Experimental Setup
3.2 Classification Results
3.3 Analysis on Wasserstein Distance
4 Conclusion
References
A Reliable and Interpretable Framework of Multi-view Learning for Liver Fibrosis Staging
1 Introduction
2 Methods
2.1 Subjective Logic for Uncertainty Estimation
2.2 Combination Rule
2.3 Global Representation Modeling
2.4 Training Paradigm
3 Experiments
3.1 Dataset
3.2 Implementation Details
3.3 Results
4 Conclusion
References
Utilizing Longitudinal Chest X-Rays and Reports to Pre-fill Radiology Reports
1 Introduction
2 Methods
3 Experiments and Results
4 Discussion and Conclusion
References
Parse and Recall: Towards Accurate Lung Nodule Malignancy Prediction Like Radiologists
1 Introduction
2 Method
2.1 Context Segmentation
2.2 Intra Context Parse
2.3 Inter Prototype Recall
2.4 Training Process of PARE
3 Experiment
3.1 Datasets and Implementation Details
3.2 Experiment Results
4 Conclusion
References
Privacy-Preserving Early Detection of Epileptic Seizures in Videos
1 Introduction
2 Proposed Method
2.1 Privacy Preserving Optical Flow Acquisition
2.2 Early Detection of Seizures in a Sample
3 Datasets and Experimental Results
3.1 In-House and Public Dataset
3.2 Training Implementation and Evaluation Metrics
3.3 Performance for Early Detection
3.4 Progressive V/s Direct Knowledge Distillation
4 Conclusion
References
Punctate White Matter Lesion Segmentation in Preterm Infants Powered by Counterfactually Generative Learning
1 Introduction
2 Method
2.1 Tissue Segmentation Module
2.2 Classification Module and Counterfactual Map Generator
2.3 PWML Segmentation Module
3 Experiments and Results
3.1 Dataset and Experimental Setting
3.2 Results
4 Conclusion
References
Discovering Brain Network Dysfunction in Alzheimer's Disease Using Brain Hypergraph Neural Network
1 Introduction
2 Methodology
2.1 Hypergraph Construction
2.2 Hypergraph Convolution
2.3 Optimization
3 Experiments
3.1 Dataset Description and Experimental Settings
3.2 Evaluating Diagnostic Capability on Amyloid-PET and FDG-PET Data
3.3 Evaluating the Statistical Power of Identifying Brain Network Dysfunction in AD
4 Conclusion
References
Improved Prognostic Prediction of Pancreatic Cancer Using Multi-phase CT by Integrating Neural Distance and Texture-Aware Transformer
1 Introduction
2 Methods
2.1 Texture-Aware Vision Transformer: Combination of CNN and Transformer
2.2 Neural Distance: Positional and Structural Information Between PDAC and Vessels
3 Experiments
4 Conclusion
References
Contrastive Feature Decoupling for Weakly-Supervised Disease Detection
1 Introduction
2 Related Work
2.1 Disease Detection
2.2 Contrastive Learning
3 Method
3.1 Memory Bank Construction
3.2 Contrastive Feature Decoupling
3.3 Regularization
4 Experiments
4.1 Dataset and Metric
4.2 Implementation Details
4.3 Comparison Results
4.4 Ablation Study
5 Conclusion
References
Uncovering Heterogeneity in Alzheimer's Disease from Graphical Modeling of the Tau Spatiotemporal Topography
1 Introduction
2 Method
2.1 Reeb Graph Analysis for Pathology Detection
2.2 Directed Graph Construction for Spatiotemporal Subtyping
3 Experiments and Results
3.1 Synthetic Experiments
3.2 Real Data Experiments
4 Conclusions
References
Text-Guided Foundation Model Adaptation for Pathological Image Classification
1 Introduction
2 Related Work
3 Methodology
3.1 Connecting Text and Imaging
3.2 Learning Visual Prompt
4 Experimental Settings
5 Results
6 Conclusion
References
Multiple Prompt Fusion for Zero-Shot Lesion Detection Using Vision-Language Models
1 Introduction
2 Related Work
3 Method
3.1 Preliminaries
3.2 Language Syntax Based Prompt Fusion
3.3 Ensemble Learning Based Fusion
4 Experiments and Results
4.1 Experimental Settings
4.2 Results
5 Conclusion
References
Reversing the Abnormal: Pseudo-Healthy Generative Networks for Anomaly Detection
1 Introduction
2 Background
3 Method
4 Experiments
4.1 Reversing Synthetic Anomalies
4.2 Ischemic Stroke Lesion Segmentation on T1w Brain MRI
5 Discussion
References
What Do AEs Learn? Challenging Common Assumptions in Unsupervised Anomaly Detection
1 Introduction
2 Background
2.1 Unsupervised Anomaly Detection: Assumptions
2.2 Auto-Encoders: Challenges
3 MorphAEus: Deformable Auto-encoders
4 Pathology Detection on Chest X-rays
5 Discussion
References
SwIPE: Efficient and Robust Medical Image Segmentation with Implicit Patch Embeddings
1 Introduction
2 Methodology
2.1 Image Encoding and Patch Embeddings
2.2 Implicit Patch Decoding
2.3 Training SwIPE
3 Experiments and Results
3.1 Datasets, Implementations, and Baselines
3.2 Study 1: Performance Comparisons
3.3 Study 2: Robustness to Data Shifts
3.4 Study 3: Model Efficiency and Data Efficiency
3.5 Study 4: Component Studies and Ablations
4 Conclusions
References
Smooth Attention for Deep Multiple Instance Learning: Application to CT Intracranial Hemorrhage Detection
1 Introduction
2 Methods
2.1 Problem Formulation
2.2 Attention-Based Multiple Instance Learning Pooling
2.3 Modeling Correlation Through the Attention Mechanism
2.4 SA-DMIL Model Description
3 Experimental Design
3.1 Data and Data Preprocessing
3.2 Experimental Settings
4 Results and Discussion
4.1 Hyperparameters Tuning
4.2 Smooth Attention MIL vs. Other MIL Methods
4.3 Visualizing Smooth Regularizing Effects at Slice Level
5 Conclusion
References
DCAug: Domain-Aware and Content-Consistent Cross-Cycle Framework for Tumor Augmentation
1 Introduction
2 Method
2.1 Problem Definition
2.2 Domain-Aware Contrastive Learning for Domain Adaptation
2.3 Cross-Domain Consistency Learning for Content Preservation
2.4 Loss Function
3 Experiments
3.1 Datasets and Implementation Details
3.2 Comparison with State-of-the-Art Methods
3.3 Significant in Improving Existing Tumor Augmentation Methods
4 Conclusion
References
Beyond the Snapshot: Brain Tokenized Graph Transformer for Longitudinal Brain Functional Connectome Embedding
1 Introduction
2 Method
2.1 Problem Definition
2.2 Graph Invariant and Variant Embedding (GIVE)
2.3 Brain Informed Graph Transformer Readout (BIGTR)
3 Experiments
4 Conclusion
References
CARL: Cross-Aligned Representation Learning for Multi-view Lung Cancer Histology Classification
1 Introduction
2 Methodology
2.1 Architecture Overview
2.2 Cross-View Representation Alignment Learning
2.3 View-Specific Representation Learning
2.4 Histologic Subtype Classification
2.5 Network Optimization
3 Experiments and Results
3.1 Dataset
3.2 Implementation Details
3.3 Results
4 Conclusion
References
Style-Based Manifold for Weakly-Supervised Disease Characteristic Discovery
1 Introduction
2 Related Work
3 Methods
4 Experiment, Results and Discussion
5 Conclusion
References
COVID-19 Pneumonia Classification with Transformer from Incomplete Modalities
1 Introduction
2 Methodology
2.1 Architecture Overview
2.2 Feature Fusion Layer
2.3 Feature Matrix Dropout Layer
2.4 Transformer Layer
2.5 Dual X-Ray Attention
3 Experiments
3.1 Dataset
3.2 Implementation Details
3.3 Results
3.4 Ablation Study
4 Conclusion
References
Enhancing Breast Cancer Risk Prediction by Incorporating Prior Images
1 Introduction
2 Method
2.1 Risk Prediction
2.2 Architecture Overview
2.3 Incorporating Prior Mammograms
3 Experiments
3.1 Dataset
3.2 Evaluation
3.3 Implementation Details
3.4 Results
4 Conclusion
References
Uncertainty Inspired Autism Spectrum Disorder Screening
1 Introduction
2 Uncertainty Inspired ASD Screening
2.1 Uncertainty Guided Training
2.2 Uncertainty Guided Personalized Diagnosis
3 Experiments
3.1 Dataset and Experimental Settings
3.2 Comparison with State-of-the-Art
3.3 Ablation Study
4 Conclusion
References
Rad-ReStruct: A Novel VQA Benchmark and Method for Structured Radiology Reporting
1 Introduction
2 Methodology
2.1 Rad-ReStruct Dataset
2.2 Hierarchical Visual Question Answering
3 Experiments and Results
4 Discussion and Conclusion
References
Xplainer: From X-Ray Observations to Explainable Zero-Shot Diagnosis
1 Introduction
2 Methodology
2.1 Model Overview
2.2 Prompt Engineering
3 Experiments and Results
4 Conclusion
References
Towards Generalizable Diabetic Retinopathy Grading in Unseen Domains
1 Introduction
2 Methodology
3 Experiments
4 Conclusion
References
Boosting Breast Ultrasound Video Classification by the Guidance of Keyframe Feature Centers
1 Introduction
2 Related Works
3 Methodology
3.1 Video and Image Classification Network
3.2 Training with Coherence Loss
3.3 Total Training Loss
4 Experiments
4.1 Implementation Details
4.2 Comparison with Video Models
4.3 Ablation Study
4.4 Visual Analysis
5 Conclusion
References
Learning with Domain-Knowledge for Generalizable Prediction of Alzheimer's Disease from Multi-site Structural MRI
1 Introduction
2 Methods
2.1 Patch-Free 3D Feature Extractor
2.2 Global Average Pooling
2.3 Domain-Knowledge Encoding
3 Experiments and Results
3.1 Data Description
3.2 Implementation Details
3.3 Performance Evaluation
4 Discussion
5 Conclusion
References
GSDG: Exploring a Global Semantic-Guided Dual-Stream Graph Model for Automated Volume Differential Diagnosis and Prognosis
1 Introduction and Related Works
2 Method
2.1 Problem Statement
2.2 Constructing Super-Nodes
2.3 Bi-level Adjacency Matrices
2.4 Dual-Stream Graph Classifier
3 Experiments and Discussion
3.1 Dataset and Pre-processing
3.2 Differential Diagnosis, Prognosis and Weakly-Supervised Localization
3.3 Visualization of Grouping and Slice Localization
3.4 Ablation Study
4 Conclusion
References
Improving Image-Based Precision Medicine with Uncertainty-Aware Causal Models
1 Introduction
2 Methods
2.1 Background on Individual Treatment Effect Estimation
2.2 Probabilistic Model of Individual Treatment Effects
2.3 Evaluating Probabilistic Predictions
3 Experiments and Results
3.1 Dataset
3.2 Evaluation of Factual Predictions and Uncertainty Estimation
3.3 Uncertainty for Individual Treatment Recommendations
3.4 Uncertainty for Clinical Trial Enrichment
4 Conclusion
References
Diversity-Preserving Chest Radiographs Generation from Reports in One Stage
1 Introduction
2 Method
2.1 Fidelity of Generated X-Rays
2.2 Diversity of Generated X-Rays
2.3 Learning Objectives and Training Process
3 Experiments and Results
3.1 Datasets and Experimental Settings
3.2 Results and Analysis
4 Conclusion
References
Contrastive Masked Image-Text Modeling for Medical Visual Representation Learning
1 Introduction
2 Method
2.1 Cross-Modal Contrastive Learning with Masked Images
2.2 Masked Image-Text Modeling
2.3 Training Procedures and Implementation Details
3 Experiments
4 Conclusion
References
Adjustable Robust Transformer for High Myopia Screening in Optical Coherence Tomography
1 Introduction
2 Method
2.1 Modified Vision Transformer for Screening
2.2 Shifted Subspace Transition Matrix
3 Experiments
3.1 Dataset
3.2 Comparison Experiments and Ablations
3.3 Adjustable Evaluation and Uncertainty Evaluation
4 Conclusion
References
Improving Outcome Prediction of Pulmonary Embolism by De-biased Multi-modality Model
1 Introduction
2 Bias in Survival Prediction
3 De-biased Survival Prediction Model
4 Experiment
4.1 Results
5 Discussions and Conclusions
References
Recruiting the Best Teacher Modality: A Customized Knowledge Distillation Method for if Based Nephropathy Diagnosis
1 Introduction
2 Dataset Establishment and Analysis
2.1 Dataset Establishment
2.2 Dataset Analysis
3 Proposed Method
3.1 Nephropathy Diagnosis Network
3.2 Customized Recruitment Module
3.3 Multi-level Knowledge Distillation
4 Experiment
4.1 Experimental Settings
4.2 Evaluation on Knowledge Distillation
4.3 Comparisons with the State-of-the-Art Models
4.4 Ablation Study
5 Conclusion
References
Text-Guided Cross-Position Attention for Segmentation: Case of Medical Image
1 Introduction
2 Methods
2.1 Configuration of Text-Image Encoder and Decoder
2.2 Text-Guided Cross Position Attention Module
3 Experiments
3.1 Setup
3.2 Segmentation Performance
3.3 Ablation Study
3.4 Application: Deep-Learning Based Disease Diagnosis
4 Conclusion
References
Visual-Attribute Prompt Learning for Progressive Mild Cognitive Impairment Prediction
1 Introduction
2 Methodology
2.1 Network Architecture
2.2 Knowledge Transfer with Multi-modal Prompt Learning
2.3 Global Prompt for Better Visual Prompt Learning
3 Experiments and Results
3.1 Datasets and Implementation Details
3.2 Comparison with the State-of-the-Art Methods
3.3 Ablation Study and Investigation of Hyper-parameters
4 Conclusion
References
Acute Ischemic Stroke Onset Time Classification with Dynamic Convolution and Perfusion Maps Fusion
1 Introduction
2 Methodology
2.1 Dynamic Convolution Feature Extraction Network
2.2 Multi-map Fusion Module
2.3 Multi-head Pooling Attention
3 Experiments
3.1 Experimental Configuration
3.2 Experimental Results and Analysis
4 Conclusion
References
Self-supervised Learning for Endoscopic Video Analysis
1 Introduction
2 Background and Related Work
3 Self-supervised Learning for Endoscopy
3.1 Masked Siamese Networks
3.2 Private Datasets
4 Experiments
4.1 Results and Discussion
4.2 Ablation Study
5 Conclusion
References
Fast Non-Markovian Diffusion Model for Weakly Supervised Anomaly Detection in Brain MR Images
1 Introduction
2 Method
2.1 Non-Markovian Diffusion Model with Hybrid Condition
2.2 Accelerated Encoding and Sampling
3 Experiment
3.1 Dataset and Evaluation Metric
3.2 Implementation Details
3.3 Comparison with State-of-the-Art Methods
3.4 Ablation Study
4 Conclusion and Discussion
References
Self-supervised Polyp Re-identification in Colonoscopy
1 Introduction
2 Methods
2.1 Single-Frame Representation for ReID
2.2 Multi-view Tracklet Representation for ReID
3 Experiments
3.1 ReID Standalone Evaluation
3.2 ReID for CADx
4 Conclusions
References
A Multimodal Disease Progression Model for Genetic Associations with Disease Dynamics
1 Introduction
2 Method
2.1 A Generic Mixed-Effects Geometric Model
2.2 Covariate Association and Statistical Framework
3 Evaluation, Clinical Results and Discussion
3.1 Simulated Data
3.2 Multimodal Clinical Data
4 Conclusion
References
Visual Grounding of Whole Radiology Reports for 3D CT Images
1 Introduction
2 Related Work
3 Methods
3.1 Problem Formulation
3.2 Anatomical Segmentation
3.3 Report Structuring
3.4 Anomaly Localization
4 Dataset and Implementation Details
4.1 Clinical Data
4.2 Implementation Details
5 Experiments
5.1 Evaluation Metrics
5.2 Results
6 Conclusion
References
Identification of Disease-Sensitive Brain Imaging Phenotypes and Genetic Factors Using GWAS Summary Statistics
1 Introduction
2 Method
2.1 DMTSCCA
2.2 Summary-DMTSCCA (S-DMTSCCA)
3 Experiment and Results
3.1 Study on the ADNI Dataset
3.2 Application to Summary Statistics from Brain Imaging GWAS
4 Conclusion
References
Revisiting Feature Propagation and Aggregation in Polyp Segmentation
1 Introduction
2 Method
3 Experiments
4 Conclusion
References
MUVF-YOLOX: A Multi-modal Ultrasound Video Fusion Network for Renal Tumor Diagnosis
1 Introduction
2 Methods
2.1 Overview of Framework
2.2 Dual-Attention Strategy for Multimodal Fusion
2.3 Video-Level Decision Generation
3 Experimental Results
3.1 Materials and Implementations
3.2 Ablation Study
3.3 Comparison with Other Methods
4 Conclusions
References
Unsupervised Classification of Congenital Inner Ear Malformations Using DeepDiffusion for Latent Space Representation
1 Introduction
2 Data
3 Methods
3.1 Anatomical Representation
3.2 Deep Diffusion Algorithm
3.3 Implementation
4 Results
5 Conclusion
References
How Does Pruning Impact Long-Tailed Multi-label Medical Image Classifiers?
1 Introduction
2 Methods
2.1 Preliminaries
2.2 Assessing the Impact of Pruning
2.3 Pruning-Identified Exemplars (PIEs)
3 Results
3.1 What is the Overall Effect of Pruning?
3.2 Which Diseases are Most Vulnerable to Pruning and Why?
3.3 How Does Disease Co-occurrence Influence Class Forgettability?
3.4 What Do Pruning-Identified CXRs have in Common?
4 Discussion and Conclusion
References
Multimodal Deep Fusion in Hyperbolic Space for Mild Cognitive Impairment Study
1 Introduction
2 Related Work
3 Methods
3.1 Data Description and Preprocessing
3.2 Preliminary
3.3 Functional Profile Learning in Hyperbolic Space
3.4 Multimodal Fusion by HGCN
4 Results
4.1 Experimental Setting
4.2 Classification Performance Comparison
4.3 Ablation Study
4.4 Feature Representation
5 Conclusion
References
Hierarchical Vision Transformers for Disease Progression Detection in Chest X-Ray Images
1 Introduction
2 Methodology
3 Experiments
3.1 Implementation Details
3.2 Dataset
3.3 Baselines
3.4 Experimental Results
3.5 Ablations on CheXRelFormer Architecture Components
3.6 Qualitative Analysis
4 Conclusion
References
Improved Flexibility and Interpretability of Large Vessel Stroke Prognostication Using Image Synthesis and Multi-task Learning
1 Introduction
2 Method
2.1 Dataset and Pre-processing
2.2 Models
3 Experiments and Results
3.1 Results of Image Synthesis and Prognostic Prediction
4 Conclusion
References
Transformer-Based Tooth Segmentation, Identification and Pulp Calcification Recognition in CBCT
1 Introduction
2 The Proposed Method
2.1 The Proposed Framework
2.2 Tooth Segmentation and Identification Module
2.3 Calcification Recognition Module and Tooth Instance Correlation Block
2.4 Triple Loss and Total Loss Function
2.5 Implementation
3 Experimental Results
3.1 Clinical Data, Experimental Setup and Evaluation Metric
3.2 Performance Evaluation
4 Conclusion
References
Treatment Outcome Prediction for Intracerebral Hemorrhage via Generative Prognostic Model with Imaging and Tabular Data
1 Introduction
2 Method
2.1 Formulation and Motivation
2.2 Generative Prognostic Model
2.3 Variational Distributions Combination
3 Experiment
3.1 Dataset and Experimental Setup
3.2 Experiment Results
4 Conclusion
References
Open-Ended Medical Visual Question Answering Through Prefix Tuning of Language Models
1 Introduction
2 Related Works
3 Methodology
3.1 Problem Statement
3.2 Model Architecture
3.3 Parameter-Efficient Strategies for Fine-Tuning the Language Model
4 Experimental Setup
5 Results
6 Conclusion
References
Flexible Unfolding of Circular Structures for Rendering Textbook-Style Cerebrovascular Maps
1 Introduction
2 Methods
2.1 Data
2.2 Rationale
2.3 Measuring Distortions
2.4 ARAP Vessel Unfolding
2.5 Merging and Image Assembly
2.6 Evaluation
3 Results
4 Discussion and Conclusion
References
Dynamic Curriculum Learning via In-Domain Uncertainty for Medical Image Classification
1 Introduction
2 Related Work
3 Method
3.1 Overview
3.2 Dynamic Difficulty Measurer
3.3 Uncertainty-Aware Sampling Pacing Function
3.4 Loss Function
4 Experiment
4.1 Dataset and Experimental Setup
4.2 Experimental Results
5 Conclusion
References
Joint Prediction of Response to Therapy, Molecular Traits, and Spatial Organisation in Colorectal Cancer Biopsies
1 Introduction
2 Methods
3 Experiments
4 Conclusion
References
Distributionally Robust Image Classifiers for Stroke Diagnosis in Accelerated MRI
1 Introduction
2 Methodology
2.1 Distributionally Robust Learning
2.2 DRL for Deep Stroke Diagnosis Networks
3 Experiments
3.1 Experimental Materials and Settings
3.2 Results
4 Conclusions
References
M&M: Tackling False Positives in Mammography with a Multi-view and Multi-instance Learning Sparse Detector
1 Introduction
2 M&M: A Multi-view and MIL System
2.1 Sparse R-CNN with Dual Classification Heads
2.2 Multi-view Reasoning
2.3 Multi-instance Learning
3 Experiments
4 Discussion and Conclusion
References
Convolving Directed Graph Edges via Hodge Laplacian for Brain Network Analysis
1 Introduction
2 Related Work
3 Preliminaries: Simplicial Complex Representation
4 Proposed Method
4.1 Hodge Laplacian of Brain Network Data
4.2 Convolving Graph Edges via Hodge Laplacian L1
4.3 Interpretability of the Connectomes in Brain Dysfunction
5 Experiments
5.1 Dataset and Experimental Settings
5.2 Experimental Results
5.3 Interpretation of AD via Trained Hodge-GNN
6 Conclusion
References
Author Index