Machine Learning for Multimodal Healthcare Data : First International Workshop, ML4MHD 2023, Honolulu, Hawaii, USA, July 29, 2023, Proceedings / edited by Andreas K. Maier, Julia A. Schnabel, Pallavi Tiwari, Oliver Stegle.

Format:

Book

Contributor:

Maier, Andreas (Andreas K.), editor.

Series:

Lecture Notes in Computer Science, 1611-3349 ; 14315

Language:

English

Subjects (All):

Medical informatics.

Health Informatics.

Local Subjects:

Health Informatics.

Physical Description:

1 online resource (200 pages)

Edition:

1st ed. 2024.

Place of Publication:

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

Summary:

This book constitutes the proceedings of the First International Workshop on Machine Learning for Multimodal Healthcare Date, ML4MHD 2023, held in Honolulu, Hawaii, USA, in July 2023. The 18 full papers presented were carefully reviewed and selected from 30 submissions. The workshop's primary objective was to bring together experts from diverse fields such as medicine, pathology, biology, and machine learning. With the aim to present novel methods and solutions that address healthcare challenges, especially those that arise from the complexity and heterogeneity of patient data.

Contents:

Intro

Preface

Organization

Contents

Death Prediction by Race in Colorectal Cancer Patients Using Machine Learning Approaches

1 Introduction

2 Materials and Methods

2.1 Data

2.2 Pre-processing

2.3 Machine Learning Methods

3 Results and Discussion

4 Conclusions

References

Neural Graph Revealers

2 Related Methods

3 Neural Graph Revealers

3.1 Representation

3.2 Optimization

3.3 Modeling Multi-modal Data

3.4 Representation as a Probabilistic Graphical Model

4 Experiments

4.1 Learning Gaussian Graphical Models

4.2 Infant Mortality Data Analysis

5 Conclusions, Discussion and Future Work

6 NGR Design Choices

Multi-modal Biomarker Extraction Framework for Therapy Monitoring of Social Anxiety and Depression Using Audio and Video

2 Data

3 Audio Biomarkers

3.1 Prosody

3.2 PAD Model

4 Video Biomarkers

4.1 Face Emotion Recognition

5 Results

5.1 Audio Biomarkers

5.2 Video Biomarkers

6 Discussion

7 Conclusion

RobustSsF: Robust Missing Modality Brain Tumor Segmentation with Self-supervised Learning-Based Scenario-Specific Fusion

2 Method

2.1 Overall Architecture

2.2 Self-supervised Learning-Based Scenario-Specific Fusion

2.3 Lifelong Learning Strategy (LLS)

2.4 Total Loss

3 Experimental Results

3.1 Experimental Setup

3.2 Results and Analysis

4 Conclusion

Semi-supervised Cooperative Learning for Multiomics Data Fusion

2 Approach

2.1 Cooperative Learning

2.2 Semi-supervised Cooperative Learning

3 Experiments

3.1 Simulated Studies

3.2 Real Data Example

Exploiting Partial Common Information Microstructure for Multi-modal Brain Tumor Segmentation

1 Introduction.

2 Related Works

2.1 Medical Image Segmentation Approaches

2.2 HGR Correlation in Multi-modal Learning

3 Background

3.1 Brain Tumor Segmentation

3.2 HGR Maximal Correlation

3.3 Soft-HGR

4 Identifying Partial Common Information

4.1 Masked Correlation Maximization

4.2 Learning Microstructure via PCI-Mask Update

5 System Design

5.1 Model Learning

5.2 Model Design

6 Experiments

6.1 Datasets

6.2 Data Preprocessing and Environmental Setup

6.3 Evaluation Metrics

6.4 Main Results

6.5 Ablation Experiments

A Proof of Theorem 1

B Proof of Lemma 1

C Algorithms

C.1 Masked Maximal Correlation Loss

C.2 Routine: Truncation Function

D Supplementary Experiments

D.1 Implementation Details and Hyperparameters

D.2 Experimental Results on BraTS-2015 Dataset

Multimodal LLMs for Health Grounded in Individual-Specific Data

2 Methods

2.1 LLMs with Tabular Data

2.2 Multimodal LLMs for Health: HeLM

2.3 UK Biobank Dataset Preparation

3.1 Quantifying Disease Risk Using Zero-Shot, Few-Shot, and Soft-Prompt Tuning

3.2 Encoding Quantitative Data Using HeLM

3.3 Estimating Asthma Risk Using Multiple Modalities

3.4 Using HeLM for Out-of-Distribution Traits

3.5 Natural Language Generation

4 Discussion

Speed-of-Sound Mapping for Pulse-Echo Ultrasound Raw Data Using Linked-Autoencoders

2.1 Network Architecture, Training

2.2 Network Architecture, Inference

3 Results

3.1 Training, Linked Autoencoder

3.2 Inference, Simulated Data

3.3 Inference, Measured Data

4 Discussion and Conclusion

HOOREX: Higher Order Optimizers for 3D Recovery from X-Ray Images

2.1 BOSS Model

2.2 Regression Network.

2.3 Second Order In-the-Loop Optimization

2.4 Loss Functions

3 Experiments and Results

3.1 Data

3.2 Experimental Setup

3.3 Results

GastroVision: A Multi-class Endoscopy Image Dataset for Computer Aided Gastrointestinal Disease Detection

2 Related Work

3 GastroVision

3.1 Dataset Details

3.2 Dataset Acquisition, Collection and Construction

3.3 Suggested Metrics

4 Experiments and Results

4.1 Implementation Details

4.2 Technical Validation

4.3 Limitation of the Dataset

5 Conclusion

MaxCorrMGNN: A Multi-graph Neural Network Framework for Generalized Multimodal Fusion of Medical Data for Outcome Prediction

1.1 Related Works

1.2 Our Contributions

2 MaxCorrMGNN Formulation for Multimodal Fusion

2.1 Patient-Modality Multi-layered Graph

2.2 HGR Maximal Correlations for Latent Multi-graph Learning

2.3 Multi-graph Neural Network

2.4 End-to-End Learning Through Task Supervision

3.1 Data and Preprocessing

3.2 Evaluation Metrics

3.3 Baseline Comparisons

3.4 Outcome Prediction Performance

SIM-CNN: Self-supervised Individualized Multimodal Learning for Stress Prediction on Nurses Using Biosignals

2.1 Time Series Prediction with 1D CNNs

2.2 Multimodal Fusion and Machine Learning

2.3 Personalized Patient-Specific Methods

2.4 Self-supervised Learning

3 Methods

3.1 Dataset Description

3.2 Data Preprocessing

3.3 Model Architecture

3.4 SIM-CNN

4 Results

4.1 Evaluation Metrics

4.2 Model Performance

5 Discussion

5.1 Limitations and Future Work

6 Conclusion

References.

InterSynth: A Semi-Synthetic Framework for Benchmarking Prescriptive Inference from Observational Data

1.1 Evaluating the Fidelity of Prescriptive Inference

2 Problem Setup

2.1 Assumptions

2.2 Ground Truth Modelling

3 Multimodal Prescriptive Ground Truth Modelling Using Neuroimaging Data

3.1 Observed Clinical Deficit Modelling

3.2 Unobserved Treatment Target Modelling

3.3 Prescriptive Ground Truth

4 Defining Observational Data Conditions

4.1 Modelling Assignment Bias

4.2 Modelling Interventional Effects

4.3 Ground Truth Compilation

5 Lesion Phenotype Modelling

5.1 Raw Lesion Data

5.2 Disconnectome Representations

5.3 Succinct Latent Representations

5.4 Atlas-Based Representations

6 End-to-End Framework Functionality

7 Discussion

7.1 Observational Vs RCT Ground Truths

8 Conclusion

Author Index.

Notes:

Includes bibliographical references and index.

Other Format:

Print version: Maier, Andreas K. Machine Learning for Multimodal Healthcare Data

ISBN:

3-031-47679-4