A Study on an Intelligent Diagnosis and Treatment Assistant System for Acupuncture in Diminished Ovarian Reserve Based on a Knowledge Graph
May 29th, 2026
A gynecology-specific acupuncture knowledge graph and decision-support system for diminished ovarian reserve (DOR) was developed using literature, classical texts, and clinical data. Clinical evaluation demonstrated 86.7% diagnostic agreement and high prescription scores, with statistically significant but limited inter-expert agreement, supporting its role as an auxiliary clinical decision-support tool.
Overview
This study developed a specialized acupuncture knowledge graph and decision-support system for diminished ovarian reserve (DOR) using 82 classical texts, literature, and clinical data. A hybrid approach combining large language models and manual annotation extracted knowledge to build a full graph and a gynecology subgraph. The system uses BiLSTM-CRF for entity recognition and dual-mode retrieval to automate syndrome differentiation and prescription generation. In a clinical evaluation of 90 DOR patients, the system achieved 86.7% diagnostic agreement with experts and a mean prescription appropriateness score of 4.65.
Key Study Components
Area of Science
- Acupuncture
- Artificial Intelligence
- Gynecology
Background
- Diminished ovarian reserve (DOR) is increasing in younger women.
- Acupuncture is used clinically for DOR but lacks standardized guidelines.
- Previous work developed a general intelligent acupuncture system.
- Knowledge graphs can structure traditional medicine knowledge.
Purpose of Study
- Develop a specialized knowledge graph and decision-support system for DOR.
- Use DOR as a clinical entry point for gynecological disease applications.
- Automate syndrome differentiation and acupuncture prescription generation.
- Evaluate system performance against expert assessments.
Methods Used
- Extracted knowledge from 82 classical acupuncture texts, Chinese and English literature, and clinical data.
- Used hybrid approach: large language models and manual annotation.
- Constructed full acupuncture knowledge graph and DOR-specific gynecology subgraph.
- Employed BiLSTM-CRF-based entity recognition.
- Implemented dual-mode retrieval for automated syndrome differentiation and prescription generation.
Main Results
- Full graph: 16,558 entities and 80,084 relations.
- Gynecology subgraph: 8,677 entities and 27,092 relations.
- In 90 DOR patients, system achieved 86.7% diagnostic agreement with experts.
- Mean prescription appropriateness score was 4.65.
- Expert inter-rater agreement: Kendall’s W ≈ 0.13 (P < 0.05), indicating limited but significant agreement.
Conclusions
- The system provides structured support for acupuncture-based syndrome differentiation in DOR.
- It can automate prescription generation with high appropriateness scores.
- It serves as an auxiliary clinical decision-support tool.
- Standardized acupuncture practice for DOR may be improved via AI integration.