🤖 AI Summary
This study systematically reviews graph convolutional networks (GCNs) for electronic health record (EHR) modeling. Addressing key challenges—including difficulty in capturing complex clinical relationships, poor cross-institutional generalizability, and insufficient modeling of temporal dynamics in EHRs—the work first comprehensively categorizes EHR graph construction paradigms (e.g., diagnosis co-occurrence graphs, knowledge-enhanced graphs, temporal event graphs) and traces the evolution of GCN architectures. It establishes a unified taxonomy covering eight clinical tasks: disease prediction, drug response, hospitalization risk, among others. Integrating over 12 benchmark datasets and 30 representative models, the study identifies core challenges—cross-center transfer learning, dynamic graph representation, and model interpretability—and proposes future directions: multi-task GCN design, joint node-edge encoding, and clinically aligned interpretability frameworks. This work delivers a systematic methodology roadmap for medical graph neural network research.
📝 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.