Key Finding
The knowledge-graph-enhanced question-answering system achieved perfect expert-rated scores for correctness and professionalism (5.00/5.00) while providing fully traceable answers with zero fabricated claims, significantly outperforming general-purpose AI alone across all evaluation dimensions (p<0.01).
Researchers in China have created a new digital tool to help doctors and patients access information about Jin San Zhen, a traditional Chinese acupuncture technique from southern China. This acupuncture method is commonly used to treat nervous system diseases, internal medical conditions, and pain disorders, but knowledge about it has been scattered across hundreds of journal articles and books, making it hard for practitioners to find reliable information quickly. The research team collected data from 191 clinical studies published between 2016 and 2025 and built a "knowledge graph"—a structured database connecting diseases, acupuncture points, and treatment plans. They combined this database with artificial intelligence to create a question-answering system that can provide accurate, evidence-based responses to clinical questions about Jin San Zhen treatments. When tested by two acupuncture experts on 60 questions, the new system scored significantly better than standard AI alone, achieving perfect scores for correctness and professionalism while providing answers that could be traced back to original research studies. The system never fabricated information, unlike general-purpose AI which can sometimes "hallucinate" false facts. This tool could help both experienced practitioners and those learning acupuncture to make better-informed treatment decisions by quickly accessing reliable, evidence-based information about which acupuncture points work best for specific conditions. If you're considering acupuncture treatment, seek a licensed acupuncturist certified by your state or national regulatory board.
This study constructed a knowledge graph (KG) for Jin San Zhen acupuncture therapy by systematically extracting data from 191 Chinese clinical studies (2016-2025) and four authoritative monographs. The final KG contained 921 nodes and 3,745 relations, representing over 80 diseases, 360+ acupoints, 55 core point combinations, and 298 treatment plans. Intrinsic evaluation against a stratified sample (n=149) demonstrated F1 scores of 0.952 for main acupoints and 0.859 for auxiliary acupoints. A retrieval-augmented generation (RAG) system combining the KG with large language models was developed and evaluated by two blinded TCM experts across 60 questions. The KG+LLM hybrid model significantly outperformed both template-based (p<0.01) and LLM-only approaches (p<0.01), achieving perfect mean scores for correctness (5.00) and professionalism (5.00), with complete traceability to source literature and zero fabricated claims. Clinical takeaway: This evidence-based digital tool offers practitioners rapid, accurate access to Jin San Zhen protocols with full source attribution.
Browse our directory of verified licensed practitioners near you.
Find a practitioner →📌 STING pathway overactivation drives excessive type I interferon expression and inflammatory cell infiltration in Sjögren's syndrome, suggesting STING inhibitors as potential therapeutic targets.
📌 The NLRP3 inflammasome, which regulates inflammatory cytokines IL-1β and IL-18, plays a critical role in inflammatory bowel disease pathophysiology and represents a promising target for pharmacological intervention.
📌 Healthcare providers working with children with autism showed positive attitudes and openness toward acupuncture as adjunct therapy despite limited prior knowledge, with laser acupuncture generating particular interest as a non-invasive option.