Reproduce “MedKGI: Iterative Differential Diagnosis with Medical Knowledge Graphs and Information-Guided Inquiring” on OpenResearch
Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.
Universal app for Apple silicon and Intel. Requires macOS 11 or later.
About this paper
Qipeng Wang, Rui Sheng, Yafei Li, Huamin Qu, Yushi Sun, Min Zhu
Researchers at Sichuan University and HKUST developed MedKGI, a framework for iterative differential diagnosis that integrates Large Language Models with a medical knowledge graph and an information-guided inquiry mechanism. This framework achieved superior diagnostic accuracy and efficiency across multiple medical consultation benchmarks by consistently grounding reasoning in verified knowledge and strategically selecting questions.
Key insights
- Grounding LLM reasoning in structured medical knowledge graphs is critical for preventing hallucinations and enhancing the factual accuracy of AI-assisted diagnosis.
- An information-theoretic approach to question selection, based on information gain, significantly improves the efficiency of diagnostic inquiry by targeting the most discriminative symptoms.
- Maintaining a consistent and structured diagnostic record across turns is essential for managing context, overcoming LLM memory limitations, and enabling coherent, iterative hypothesis refinement.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.