Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Medical Applications of Graph Convolutional Networks Using Electronic Health Records: A Survey

Garrik Hoyt, Noyonica Chatterjee, Fortunato Battaglia, Paramita Basu

TL;DR

Problem: EHRs are heterogeneous and high-dimensional, and conventional models struggle to capture complex patient relationships. Approach: this survey compiles recent literature on Graph Convolutional Networks applied to EHR data, mapping domains, datasets, architectures, and hybrid pipelines. Contributions: it catalogs medical fields, datasets (e.g., MIMIC-III/IV), model hybrids (including BERT-based components), and key limitations with directions for future work. Significance: the synthesis provides researchers and practitioners with a consolidated view of how GCNs can inform medical decision making and highlights practical data sources and architectural patterns for deploying GCN-based EHR models.

Abstract

Graph Convolutional Networks (GCNs) have emerged as a promising approach to machine learning on Electronic Health Records (EHRs). By constructing a graph representation of patient data and performing convolutions on neighborhoods of nodes, GCNs can capture complex relationships and extract meaningful insights to support medical decision making. This survey provides an overview of the current research in applying GCNs to EHR data. We identify the key medical domains and prediction tasks where these models are being utilized, common benchmark datasets, and architectural patterns to provide a comprehensive survey of this field. While this is a nascent area of research, GCNs demonstrate strong potential to leverage the complex information hidden in EHRs. Challenges and opportunities for future work are also discussed.

Medical Applications of Graph Convolutional Networks Using Electronic Health Records: A Survey

TL;DR

Problem: EHRs are heterogeneous and high-dimensional, and conventional models struggle to capture complex patient relationships. Approach: this survey compiles recent literature on Graph Convolutional Networks applied to EHR data, mapping domains, datasets, architectures, and hybrid pipelines. Contributions: it catalogs medical fields, datasets (e.g., MIMIC-III/IV), model hybrids (including BERT-based components), and key limitations with directions for future work. Significance: the synthesis provides researchers and practitioners with a consolidated view of how GCNs can inform medical decision making and highlights practical data sources and architectural patterns for deploying GCN-based EHR models.

Abstract

Graph Convolutional Networks (GCNs) have emerged as a promising approach to machine learning on Electronic Health Records (EHRs). By constructing a graph representation of patient data and performing convolutions on neighborhoods of nodes, GCNs can capture complex relationships and extract meaningful insights to support medical decision making. This survey provides an overview of the current research in applying GCNs to EHR data. We identify the key medical domains and prediction tasks where these models are being utilized, common benchmark datasets, and architectural patterns to provide a comprehensive survey of this field. While this is a nascent area of research, GCNs demonstrate strong potential to leverage the complex information hidden in EHRs. Challenges and opportunities for future work are also discussed.

Paper Structure

This paper contains 20 sections, 4 figures, 2 tables.

Table of Contents

  1. Introduction
  2. Electronic Health Records
  3. Graph Neural Networks
  4. Graph Convolutional Networks
  5. Methods
  6. Search Strategy and Sources
  7. Eligibility Criteria
  8. Data Extraction, Synthesis, and Analysis
  9. Summary of Study Selection
  10. Results
  11. Overview of Study Characteristics
  12. Medical Fields
  13. Hybrid Models
  14. EHR Data Sources
  15. Discussion
  16. ...and 5 more sections

Figures (4)

  • Figure 1: Process diagram of how GCNs are used to extract insight from patient data. First, a graph representation of the data is constructed from the EHRs. Next, a GCN performs convolution operations to derive information from the relationships between neighboring nodes in the graph. The swirling in the representation of the GCN conveys the aggregation process performed on neighborhoods in graphs. Results can be used to make informed decisions and gain insight into potential causal factors.
  • Figure 2: PRISMA flowchart for the literature search and selection process for the review.
  • Figure 3: Distribution of included articles by year. The cutoff date for selection was April 27, 2024.
  • Figure 4: Distribution of the application of GCNs to various medical fields identified by the review.