MAGR-TCM: a knowledge graph-powered multi-agent model for intelligent home health consultation and triage in TCM gynecology
BACKGROUND: Traditional Chinese Medicine (TCM) demonstrates distinct efficacy in managing gynecological disorders. However, accessing this specialized care in a home setting remains challenging due to the high cost of frequent hospital visits, privacy concerns, and a lack of immediate professional guidance. While general-purpose artificial intelligence offers a convenient alternative, the semantic gap between a patient's colloquial symptom descriptions and formal TCM diagnostic logic-compounded by the inherent 'hallucination' risks of large language models-prevents these tools from providing safe, actionable advice. To bridge this gap, this study proposes MAGR-TCM, an intelligent home health consultation and triage model. METHODS: Under expert supervision, a domain-specific knowledge graph was constructed, encompassing 10,231 entities and 32,051 relationships extracted from 741 high-quality clinical case records. Based on this foundation, we developed a Multi-Agent Retrieval-Augmented Generation (RAG) framework comprising four specialized agents designed to reconstruct the complete TCM 'Disease-Pattern-Therapeutic Principle-Formula-Medicinal' reasoning chain from informal user inquiries. This collaborative architecture simulates the expert diagnostic process to generate logical, evidence-based health guidance. RESULTS: Experimental results demonstrate that MAGR-TCM significantly outperforms all baseline models on both RAGAS automatic evaluation and expert human evaluation. The model achieved 100% sensitivity and specificity on the safety test set, further validating its practical utility in home health settings and its ability to distinguish self-treatable conditions from those requiring urgent hospital care. CONCLUSIONS: By transitioning from black-box predictions to transparent, safe, and explainable reasoning logic, MAGR-TCM offers a technological paradigm for patient-centered home health management. It effectively mitigates generative hallucinations, making professional TCM expertise safely accessible for women's daily self-care and preliminary triage.
Authors
- Yishan Li
- Fang Fang (ORCID: https://orcid.org/0000-0003-3784-3696)
- Jin Li (ORCID: https://orcid.org/0000-0002-0260-3169)
- Wei Xu (ORCID: https://orcid.org/0009-0005-4883-5132)
- Ming Lv
- Yalan Hu
- Xu Pan
Institutions
- Anhui University of Traditional Chinese Medicine (CN)
Publication Details
- Journal
- Chinese Medicine
- Published
- 2026-11-12
- DOI
- https://doi.org/10.1186/s13020-026-01485-y
- Primary Topic
- Artificial Intelligence in Healthcare and Education
- Type
- article
- Field-Weighted Citation Impact
- 0.00
Funders
- Anhui Provincial Quality Engineering Project