KG4Diagnosis: Enhancing Medical AI Diagnosis
A hierarchical multi-agent framework with knowledge graph integration
KG4Diagnosis introduces a two-tier LLM framework that mirrors real medical systems, combining AI agents with robust medical knowledge graphs for improved diagnostic accuracy.
- Hierarchical approach: General Practitioner agent for initial assessment with Specialist agents for domain expertise
- Knowledge graph enhancement: Automated construction integrating 362 common diseases across specialties
- Real-world medical simulation: Models actual healthcare systems with referral workflows
- Enhanced diagnostic accuracy: Knowledge integration improves LLM diagnostic capabilities while maintaining explainability
This research addresses critical healthcare challenges by providing a systematic framework for LLM-based diagnosis that maintains specialized expertise while handling complex medical scenarios.