Ethical and philosophical issues in medical imaging, multimodal learning and fusion across scales for clinical decision support, and topological data analysis for biomedical imaging : 1st international workshop, EPIMI 2022, 12th international workshop, ML

This book constitutes the refereed joint proceedings of the 1 st International Workshop on Ethical & Philosophical Issues in Medical Imaging (EPIMI 2022); the 12 th International Workshop on Multimodal Learning and Fusion Across Scales for Clinical Decision Support (ML-CDS 2022) and the 2 nd International Workshop on Topological Data Analysis for Biomedical Imaging (TDA4BiomedicalImaging 2022), held in conjunction with the 25th International Conference on Medical Imaging and Computer-Assisted Intervention, MICCAI 2022, in Singapore, in September 2022. EPIMI includes five short papers about various humanistic aspects of medical image computing and computer-assisted interventions. The ML-CDS papers discuss machine learning on multimodal data sets for clinical decision support and treatment planning. The TDA papers focus on Topological Data Analysis: a collection of techniques and tools that have matured from an increasing interest in the role topology plays in machine learning and data science. EPIMI Preface EPIMI Organization ML-CDS Preface ML-CDS Organization TDA Preface TDA Organization Contents Ethical and Philosophical Issues in Medical Imaging Data Poisoning Attack and Defenses in Connectome-Based Predictive Models 1 Introduction 2 Motivation 3 Backdoor Data Poisoning Attacks 4 Proposed Defenses 5 Results 6 Discussion and Conclusions References Disproportionate Subgroup Impacts and Other Challenges of Fairness in Artificial Intelligence for Medical Image Analysis 1 Introduction 2 Material and Methods 3 Results 4 Discussion 4.1 Analysis of Empirical Results 4.2 Developing Fair Medical Image Analysis Tools 4.3 Fairness Considerations in Clinical Practice 5 Conclusion 6 Author's Note References Separable vs. End-to-End Learning: A Critical Examination of Learning Paradigms 1 Introduction 2 Nature as a Spectrum with Illustrative Examples 3 Values Arising from Technical Considerations 3.1 Robustness and Reusability 3.2 Dataset and Training Considerations 3.3 Automation and Workflow Considerations 3.4 Publishability Considerations 3.5 Medical Knowledge Discovery 4 Conclusions References .28em plus .1em minus .1emA 35-Year Longitudinal Analysis of Dermatology Patient Behavior Across Economic and Cultural Manifestations in Tunisia, and the Impact of Digital Tools 1 Introduction 2 35 Years of Dermatology Patient Behavior Across All Three Clinic Stages 2.1 Pre-visit Perception 2.2 In-Visit Perception 2.3 Post-visit Perception 3 The Historical Evolution of Digital Tools and Their Patient-Behavior Impact Across Clinical Stages 3.1 Social Media 3.2 ``Dr. Google'' 3.3 The First AI Tools 4 The Future: How Can Advanced Digital Tools Optimize Patient Behavior Across All Clinical Stages and Address Economic and Cultural Barriers? 5 Conclusion References User-Centered Design for Surgical Innovations: A Ventriculostomy Case Study 1 Introduction 2 User-Centered Design 3 Case Study 3.1 Research 3.2 Analysis 3.3 Design 3.4 Develop 3.5 Test 4 Discussion 5 Conclusion References Multimodal Learning and Fusion Across Scales for Clinical Decision Support Visually Aware Metadata-Guided Supervision for Improved Skin Lesion Classification Using Deep Learning*-4pt 1 Introduction 2 Proposed Method 2.1 Problem Formulation 2.2 Visual Feature Extraction Module 2.3 Meta Feature Extraction Module 2.4 Visual Attention Module 2.5 Global Meta Fusion Module 2.6 Classification Module 3 Experiments and Results 3.1 Datasets 3.2 Implementation Details 3.3 Ablation Study 3.4 Comparison with Other Methods 4 Discussion 5 Conclusion and Future Work References Predicting Osteoarthritis of the Temporomandibular Joint Using Random Forest with Privileged Information 1 Introduction 2 Methods 2.1 Data Acquisition and Preparation 2.2 Model Construction 2.3 Cross Validation and Evaluation 2.4 Post-hoc Feature Analysis 2.5 Implementation 3 Results and Discussion 3.1 Dataset Analysis 3.2 Feature Selection Analysis 3.3 Model Results 3.4 Feature Importance Based on Tree-Based Feature Transforms 4 Conclusion References Hybrid Network Based on Cross-Modal Feature Fusion for Diagnosis of Alzheimer’s Disease 1 Introduction 2 Methodology 2.1 Method Overview 3 Large Kernel Attention Module 3.1 CNN Branch 3.2 Transformer Branch 3.3 Modal Feature Fusion Block 3.4 Spatial-Channel Attention 4 Experiment and Results 4.1 Experiment Settings 4.2 Diagnosis Performance and Discussion 4.3 t-SNE Visualization and Heatmap 5 Conclusion References Topological Data Analysis for Biomedical Imaging Future Unruptured Intracranial Aneurysm Growth Prediction Using Mesh Convolutional Neural Networks 1 Introduction 2 Materials and Methods 2.1 Dataset 2.2 Methods 3 Results 4 Discussion 5 Conclusion References TDA-Clustering Strategies for the Characterization of Brain Organoids 1 Introduction 2 Methods 2.1 Resources 2.2 TDA 2.3 Clustering Feature Vectors 2.4 Quantitative Comparisons 3 Results 3.1 Qualitative 3.2 Quantitative 4 Discussion 5 Conclusion References Fetal Cortex Segmentation with Topology and Thickness Loss Constraints 1 Introduction 2 Method 3 Evaluation and Results 4 Conclusion References Author Index