Hierarchical-Graph RAG for Medical Research

Author(s): Michael Shapiro MD MSc

Originally published on Towards AI.

Leveraging hierarchical graph structures for effective retrieval

This member-only story is on us. Upgrade to access all of Medium.

In a previous article, I demonstrated how a basic Retrieval-Augmented Generation (RAG) model can help identify the most appropriate ICD-10-CM code for a patient note. But what if your goal is broader? Suppose you’re a researcher studying streptococcal infections in a hospital setting — how can you efficiently gather all relevant ICD-10-CM codes related to this condition? This article introduces a novel approach combining a Large Language Model (LLM) with a hierarchical graph-based algorithm to streamline this process. By leveraging the hierarchical nature of ICD-10-CM codes, this method enables the retrieval of a comprehensive list of relevant entries with minimal effort.

Hierarchical data structures are widespread across numerous disciplines. In biology, for instance, taxonomic classification organizes organisms from the broadest categories to the most specific: domain, kingdom, phylum, class, order, family, genus, and species. Medicine, too, relies heavily on hierarchical frameworks. Disease coding systems like ICD-10-CM exemplify this structure by grouping diseases from general categories to highly specific conditions.

This tutorial focuses on ICD-10-CM as a prime example of hierarchical data. Using this system, I will demonstrate how to build an efficient retrieval process that can uncover all relevant codes while… Read the full blog for free on Medium.

Join thousands of data leaders on the AI newsletter. Join over 80,000 subscribers and keep up to date with the latest developments in AI. From research to projects and ideas. If you are building an AI startup, an AI-related product, or a service, we invite you to consider becoming a sponsor.

Published via Towards AI