Discharge Summaries with GraphRAG: Context-Aware Generation Using Clinical Knowledge Graphs

Session Track: AI Engineering

Session Time: 04:00 am - 04:30 am November 6

Session description

Ashwin Murugappa will present a novel application of GraphRAG for generating high-quality, context-aware discharge summaries in a clinical setting. This work sits at the intersection of clinical informatics, graph databases, and LLMs, leveraging the power of Neo4j to structure and retrieve relevant patient context before generating summaries via LLMs. In modern hospitals, discharge summaries are time-consuming and highly variable in quality. Ashwin’s approach uses a Neo4j-based knowledge graph to represent key clinical entities (e.g., diagnoses, investigations, procedures, medications, etc.) and their temporal and relational context. By retrieving graph-based substructures relevant to a patient episode, GraphRAG enables more accurate, complete, and explainable document generation.