How to Combine Knowledge Graphs and AI Agents for Medical Diagnosis Prediction

Explore a groundbreaking multi-agent architecture that combines AI agents with knowledge graphs for medical diagnosis prediction in this 26-minute research presentation. Learn how researchers from Emory and Stanford Universities developed KERAP (Knowledge-Enhanced Reasoning Approach), an innovative system that uses three interacting agents working with knowledge graphs updated through fine-tuned Sentence BERT Named Entity Recognition tools. Discover how this approach enables accurate zero-shot diagnosis prediction using Large Language Models for clinical and professional healthcare applications. Understand the methodology behind predicting patients' future health risks based on historical medical data such as electronic health records (EHRs), and examine how structured knowledge representation enhances AI agent effectiveness in healthcare scenarios. Gain insights into the multi-stage reasoning process that makes this system particularly valuable for clinical decision-making and professional medical applications.