Trustworthy Artificial Intelligence for Healthcare : Second International Workshop, TAI4H 2024, Jeju, South Korea, August 4, 2024, Proceedings / edited by Hao Chen, Yuyin Zhou, Daguang Xu, Varut Vince Vardhanabhuti.

Format:

Book

Contributor:

Chen, Hao, editor.

Series:

Lecture Notes in Computer Science, 1611-3349 ; 14812

Language:

English

Subjects (All):

Machine learning.

Image processing--Digital techniques.

Image processing.

Computer vision.

Machine Learning.

Computer Imaging, Vision, Pattern Recognition and Graphics.

Local Subjects:

Machine Learning.

Computer Imaging, Vision, Pattern Recognition and Graphics.

Physical Description:

1 online resource (180 pages)

Edition:

1st ed. 2024.

Place of Publication:

Cham : Springer Nature Switzerland : Imprint: Springer, 2024.

Summary:

This book constitutes the proceedings of Second International Workshop on Trustworthy Artificial Intelligence for Healthcare, TAI4H 2024, held in Jeju, South Korea, in August 2024, in conjunction with the International Joint Conference on Artificial Intelligence, IJCAI 2024. The 13 full papers included in this book were carefully reviewed and selected from 21 submissions. They focus on trustworthy artificial intelligence, healthcare, generalization, explainability, fairness, privacy, multi-modal fusion, foundation models. .

Contents:

Intro

Preface

Organization

Contents

AI Trustworthy Challenges in Drug Discovery

1 Introduction

2 Fundamentals

3 Methodology

4 Experimental Result on Four Dimensions

4.1 Fairness

4.2 Robustness

4.3 Privacy

4.4 Explainability

5 Enhancement and Future Work

6 Conclusion

References

ODR3DNet: Omni-Dimension Dynamic Residual 3D Net for Pulmonary Nodule Detection

2 Related Work

3 Method

3.1 Encoder-Decoder Structure

3.2 OD3D-Based Feature Extraction Module

3.3 Classification Module

3.4 Loss Functions

4 Experiments

4.1 Dataset

4.2 Experimental Setting

4.3 Comparative Experiment and Analysis

4.4 Ablation Experiment and Analysis

5 Conclusion

Knowledge Injected Multimodal Irregular EHRs Model for Medical Prediction

3.1 Problem Setup

3.2 Domain Knowledge Injection

3.3 In-memory Knowledge Injection

3.4 Cross Modal Fusion

3.5 Clinical Prediction

4.1 Experiment Setup

5 Results

FusionINN: Decomposable Image Fusion for Brain Tumor Monitoring

2 Method

2.1 INN-Based Decomposable Image Fusion

2.2 INN Architecture

2.3 Unsupervised Learning

3 Results and Discussion

4 Conclusion

Stochastic Featurization for Active Learning

3 Proposed Method

3.1 Problem Definition

3.2 Stochastic Featurization for AL

3.3 SFAL for Downstream Tasks

4 Experimental Setup

4.1 Datasets

4.2 Baselines

4.3 Evaluation Measures

4.4 Settings

5 Results and Discussion

5.1 Text Classification

5.2 NER

Human-in-the-Loop Chest X-Ray Diagnosis: Enhancing Large Multimodal Models with Eye Fixation Inputs.

1 Introduction

2.1 Tasks

2.2 Eye Fixation Dataset

2.3 Prompt

2.4 Baseline Models

2.5 Evaluation

2.6 Training

3.1 Fixation Prompts Evaluation

3.2 Finetuned Model Evaluation

3.3 Most Common Response Analysis

Exploring Vision Language Pretraining with Knowledge Enhancement via Large Language Model

2.1 Medical Vision-Language Pre-training (MVLP)

2.2 Medical Knowledge Enhancement

3.1 Triplet Extraction

3.2 Architecture

3.3 Knowledge-Enhanced Triplet Encoding with MedPALM 2

3.4 Fusion Module

4 Experiment

4.2 Implementation

4.3 Comparison with State-of-the-Art Methods

5 Ablation Study

Evaluating How Explainable AI Is Perceived in the Medical Domain: A Human-Centered Quantitative Study of XAI in Chest X-Ray Diagnostics

3 User Study Design and Methodology

3.1 XAI Methods

3.2 Pre-study Design and Participant Recruitment

3.3 Study Components

3.4 Hypotheses

3.5 User Study Scenario

4 Analysis and Results

4.1 Predictive and XAI Model Setup

4.2 User Study Evaluation Analysis

A Appendix

SnapSeg: Training-Free Few-Shot Medical Image Segmentation with Segment Anything Model

2.1 Few-Shot Learning

2.2 Few-Shot Semantic Segmentation

2.3 Medical Image Few-Shot Segmentation

3.2 Revisiting of Segment Anything Model

3.3 Overall Framework

3.4 Object Proposal Set and Multi-level Features Generator

3.5 Multi-level Similarity Analyser

3.6 Dual Perspective Predictor

4.1 Dataset and Pre-processing

4.2 Evaluation.

4.3 Implementation Details

4.4 Quantitative and Qualitative Results

4.5 Ablation Study

Assessing the Generalizability of Cancer Prognosis Models: Breast and Colon Cancer Case Studies

2 Literature Review on the Generalizability of Machine Learning Models

3 Experimental Setup

3.1 Algorithms

3.2 Datasets

4 Empirical Comparison

4.1 Data Preprocessing

4.2 Hyperparameter Tuning

4.3 Results and Analysis

5 Conclusions

SISU: A Holistic Self-training Framework on Semi-supervised White Blood Cell Segmentation

2 Related Works

3.1 Stage 1: Feature Perturbed Cross-View Co-training Scheme

3.2 Stage 2: Weak-to-Strong Consistency Self-training

3.3 Stage 3: All Samples Supervised Re-training

4.2 Settings

5.1 Performance Comparison Between SISU and Other Approaches

5.2 Performance on the Supervised Training Phase with the Incorporation of FixMatch

5.3 Qualitative Results of Prediction Masks from Proposed Method

Optimizing Foundation Models for Histopathology: A Continual Learning Approach to Cancer Detection

2 Materials and Methods

2.1 Dataset Description

2.2 Data Preprocessing

2.3 Model Training and Evaluation

3.1 Hardware Configuration

3.2 Software Configuration

3.3 Foundation Model Architecture

3.4 Rationale for Selecting EfficientNet B0

3.5 Training Protocols

4 Experiments and Results

4.1 Experiment 1: Baseline Training

4.2 Experiment 2: Incremental Training with Varying Number of Layers

4.3 Experiment 3: Trade-Off Between Accuracy Improvement and Training Time

5 Discussion

References.

RF-Lung-DR: Integrating Biological and Drug SMILES Features in a Random Forest-Based Drug Response Predictor for Lung Cancer Cell Lines

2.1 Data Collection and Processing

2.2 Model Constructing

2.3 Evaluation Metrics

3 Results

3.1 Random Forest as the Most Effective Algorithm for DRP in Lung Cancer Cell Lines

3.2 Application of Feature Selection Techniques to Enhance RF-Lung-DR Performance

3.3 Effectiveness of Drug SMILES Features on DRP Model Performance

3.4 Identifying Key Biological Features in the DRP Model for Clinical Applications in Lung Cancer Cell Lines

4 Discussions

4.1 Clinical Implications of RF-Lung-DR for Enhancing Personalized Medicine in Lung Cancer Treatment

4.2 Future Directions of DRP Model to Aid Cancer Treatment

4.3 Limitations of This Study

Author Index.

Notes:

Includes bibliographical references and index.

ISBN:

3-031-67751-X