

浏览全部资源
扫码关注微信
1.中国中医科学院医学实验中心,北京市东城区东直门内南小街16号,100700
2.中国科学院自动化研究所
3.天津中科闻歌科技有限公司
4.中国中医科学院中医基础理论研究所
Received:18 November 2024,
Revised:2025-03-04,
Published:02 May 2025
移动端阅览
刘奕清,李迎,杨洪军等.基于大语言模型的中医智能预问诊系统的构建与验证[J].中医杂志,2025,66(09):895-900.
LIU Yiqing,LI Ying,YANG Hongjun,et al.Construction and Validation of a Large Language Model-Based Intelligent Pre-Consultation System for Traditional Chinese Medicine[J].Journal of Traditional Chinese Medicine,2025,66(09):895-900.
刘奕清,李迎,杨洪军等.基于大语言模型的中医智能预问诊系统的构建与验证[J].中医杂志,2025,66(09):895-900. DOI: 10.13288/j.11-2166/r.2025.09.005.
LIU Yiqing,LI Ying,YANG Hongjun,et al.Construction and Validation of a Large Language Model-Based Intelligent Pre-Consultation System for Traditional Chinese Medicine[J].Journal of Traditional Chinese Medicine,2025,66(09):895-900. DOI: 10.13288/j.11-2166/r.2025.09.005.
目的
2
基于大语言模型构建一种中医智能预问诊系统,以提高临床诊疗效率。
方法
2
采用DeepSpeed ZeRO-3分布式训练策略,基于YAYI 2-30B微调中医大语言模型;整理中医书籍和临床病案中的关系数据,设计带权无向图网络,建立智能体辨证推理模型。利用智能体协作架构实现中医大语言模型与辨证推理模型交互。应用Loss函数、BLEU-4和ROUGE-L分别从模型训练、文本生成质量和语言理解能力方面对中医大语言模型进行评估。设计含中医执业医师、执业中药师、中医症状术语识别和经络识别的专业知识测试集,评估系统的专业知识掌握水平;通过临床测试对比系统与主治医师在辨证准确率、问诊轮次和问诊时间等指标的差异。
结果
2
Loss函数显示经100 000次迭代后,训练损失值逐渐稳定于0.7±0.08,表明中医大语言模型已训练完成,且具有较好泛化能力。中医大语言模型在BLEU-4得分0.38,在ROUGE-L得分0.62,提示其自然语言翻译质量已达标。整理得到2715个症状术语、505种疾病与证型关系、1011种疾病与主症关系、1 303 600种不同症状之间关系,构建了智能体辨证推理模型。在中医执业医师、执业中药师、中医症状术语识别、经络识别模拟测试中的正确率分别达到94.09%、78.00%、87.50%、68.80%。临床测试中,该系统辨证准确率达88.33%,其问诊轮次、问诊时间均少于主治医师(
P
<0.01),提示该系统具备一定的预问诊能力。
结论
2
基于大语言模型的中医智能预问诊系统可以在一定程度上模拟临床中医医生的辨证思维,提高医生诊疗效率及患者就医体验。
Objective
2
To construct a large language model (LLM)-based intelligent pre-consultation system for traditional Chinese medicine (TCM) to improve efficacy of clinical practice.
Methods
2
A TCM large language model was fine-tuned using DeepSpeed ZeRO-3 distributed training strategy based on YAYI 2-30B. A weighted undirected graph network was designed and an agent-based syndrome differentiation model was established based on relationship data extracted from TCM literature and clinical records. An agent collaboration framework was developed to integrate the TCM LLM with the syndrome differentiation model. Model performance was comprehensively evaluated by Loss function, BLEU-4, and ROUGE-L metrics, through which training convergence, text generation quality, and language understanding capability were assessed. Professional knowledge test sets were developed to evaluate system proficiency in TCM physician licensure content, TCM pharmacist licensure content, TCM symptom terminology recognition, and meridian identification. Clinical tests were conducted to compare the system with attending physicians in terms of diagnostic accuracy, consultation rounds, and consultation duration.
Results
2
After 100 000 iterations, the training loss value was gradually stabilized at about 0.7±0.08, indicating that the TCM-LLM has been trained and has good generalization ability. The TCM-LLM scored 0.38 in BLEU-4 and 0.62 in ROUGE-L, suggesting that its natural language processing ability meets the standard. We obtained 2715 symptom terms, 505 relationships between diseases and syndromes, 1011 relationships between diseases and main symptoms, and 1 303 600 relationships among different symptoms, and constructed the Agent of syndrome differentiation model. The accuracy rates in the simulated tests for TCM practitioners, licensed pharmacists of Chinese materia medica, recognition of TCM symptom terminology, and meridian recognition were 94.09%, 78.00%, 87.50%, and 68.80%, respectively. In clinical tests, the syndrome differentiation accuracy of the system reached 88.33%, with fewer consultation rounds an
d shorter consultation time compared to the attending physicians (
P
<0.01), suggesting that the system has a certain pre- consultation ability.
Conclusion
2
The LLM-based intelligent TCM pre-diagnosis system could simulate diagnostic thinking of TCM physicians to a certain extent. After understanding the patients' natural language, it collects all the patient's symptom through guided questioning, thereby enhancing the diagnostic and treatment efficiency of physicians as well as the consultation experience of the patients.
侯佳音 , 徐博 , 何萍 . 基于AI的诊前预问诊系统的设计与应用 [J]. 中国数字医学 , 2024 , 19 ( 7 ): 17 - 22 .
国家中医药管理局 , 国家数据局 . 国家中医药管理局 国家数据局关于印发《关于促进数字中医药发展的若干意见》的通知 [EB/OL]. ( 2024-07-19 )[ 2024-11-30 ]. https://www.gov.cn/zhengce/zhengceku/202408/content_6968519.htm https://www.gov.cn/zhengce/zhengceku/202408/content_6968519.htm .
侯春蕾 , 许颖 , 崔延婕 , 等 . 2型糖尿病中医问诊量表的研制及临床应用概述 [J]. 世界科学技术-中医药现代化 , 2021 , 23 ( 2 ): 396 - 401 .
刘国萍 , 王忆勤 , 董英 , 等 . 中医心系问诊量表的研制及评价 [J]. 中西医结合学报 , 2009 , 7 ( 1 ): 20 - 24 .
王芬 , 刘铜华 , 丁雷 , 等 . 基于概率图的中医动态交互问诊与智能辨证的数学模型的探索 [J]. 世界科学技术-中医药现代化 , 2023 , 25 ( 10 ): 3370 - 3376 .
LIU GP , YAN JJ , WANG YQ , et al . Application of multilabel learning using the relevant feature for each label in chronic gastritis syndrome diagnosis [J]. Evid Based Complement Alternat Med , 2012 , 135387 . doi: 10.1155/2012/135387 http://dx.doi.org/10.1155/2012/135387 .
黄仲羽 , 刘凤斌 . 基于多维计算机自适应测试的中医证候量化方法学探索 [J]. 时珍国医国药 , 2018 , 29 ( 1 ): 234 - 236 .
佘楷杰 , 袁艿君 , 马庆宇 , 等 . 机器学习驱动中医诊断智能化的发展现状、问题及解决路径 [J]. 中国中医基础医学杂志 , 2024 , 30 ( 3 ): 398 - 406 .
TAN S , XIN X , WU D . ChatGPT in medicine: prospects and challenges: a review article [J]. Int J Surg , 2024 , 110 ( 6 ): 3701 - 3706 .
柏灿 , 王洁 . 人工智能大语言模型在中医药领域的应用 [J]. 西昌学院学报(自然科学版) , 2024 , 38 ( 2 ): 62 - 69 .
XIAO X , WANG Y , XU N , et al . YAYI-UIE: A chat-enhanced instruction tuning framework for universal information extraction [EB/OL]. ( 2023-12-24 )[ 2024-10-16 ]. http://arxiv.org/abs/2312.15548 http://arxiv.org/abs/2312.15548 .
LUO Y , KONG Q , XU N , et al . YAYI 2: Multilingual open-source large language models [EB/OL].( 2023-12-22 )[ 2024-10-16 ]. http://arxiv.org/abs/2312.14862 http://arxiv.org/abs/2312.14862 .
刘阳 . 智能体交互模式大模型时代的新型交互范式 [J]. 通信世界 , 2024 ( 13 ): 26 - 29 .
薛清录 . 中国中医古籍总目 [M]. 上海 : 上海辞书出版社 , 2007 .
严世芸 . 中国医籍通考 [M]. 上海 : 上海中医学院出版社 , 1990 .
丹波元胤 . 中国医籍考 [M]. 北京 : 人民卫生出版社 , 1956 .
KUMAR A , KHARADI A , SINGH D , et al . Automatic question-answer pair generation using Deep Learning [EB/OL]. ( 2021-09-02 )[ 2025-02-01 ]. https://ieeexplore.ieee.org/xpl/conhome/9544088/proceeding https://ieeexplore.ieee.org/xpl/conhome/9544088/proceeding .
HANINDHITO B , PATEL B , JOHN LK . Bandwidth characterization of DeepSpeed on distributed large language model training [EB/OL]. ( 2024-05-05 ) [ 2025-02-01 ]. https://ieeexplore.ieee.org/xpl/conhome/10589923/proceeding https://ieeexplore.ieee.org/xpl/conhome/10589923/proceeding .
吴勉华 , 石岩 . 中医内科学 [M]. 5版 . 北京 : 中国中医药出版社 , 2021 .
陈红风 . 中医外科学 [M]. 5版 . 北京 : 中国中医药出版社 , 2021 .
冯晓玲 , 张婷婷 . 中医妇科学 [M]. 5版 . 北京 : 中国中医药出版社 , 2021 .
赵霞 , 李新民 . 中医儿科学 [M]. 5版 . 北京 : 中国中医药出版社 , 2021 .
赵宗耀 , 何安宇 , 张宇初 , 等 . 论症状的性质及其对症状标准化研究的意义 [J]. 中医杂志 , 2022 , 63 ( 19 ): 1801 - 1805 .
王志国 . 病状术语规范化基础 [M]. 北京 : 人民卫生出版社 , 2015 .
姚乃礼 . 中医症状鉴别诊断学 [M]. 2版 . 北京 : 人民卫生出版社 , 2005 .
全国中医标准化技术委员会 . 中医临床名词术语 第1部分: 内科学 [S]. 北京 : 中国标准出版社 , 2023 .
苑中旭 , 李理 , 何凡 , 等 . 融合思维链与知识图谱的中医问答模型 [J/OL]. 计算机工程与应用 .( 2024-10-15 )[ 2025-02-12 ]. https://link.cnki.net/urlid/11.2127. TP.20241015.1111.006 https://link.cnki.net/urlid/11.2127.TP.20241015.1111.006 .
DEEPSEEK-AI , LIU A , FENG B , et al . DeepSeek-V2: A strong, economical, and efficient mixture-of-experts language model [EB/OL]. ( 2024-05-07 )[ 2024-10-17 ]. https://arxiv.org/abs/2405.04434 https://arxiv.org/abs/2405.04434 .
中华医学会消化病学分会 . 中国慢性胃炎共识意见(2017年,上海) [J]. 胃肠病学 , 2017 , 22 ( 11 ): 670 - 687 .
中国中西医结合学会消化系统疾病专业委员会 . 慢性非萎缩性胃炎中西医结合诊疗共识意见(2017年) [J]. 中国中西医结合消化杂志 , 2018 , 26 ( 1 ): 1 - 8 .
中华中医药学会 . 2023年度中医药重大科学问题、工程技术难题和产业技术问题 [J]. 中医杂志 , 2023 , 64 ( 14 ): 1405 - 1421 .
丁勇 , 蔡秀军 , 薛翀 , 等 . 基于大模型的智能预问诊系统创新与实践 [J]. 中国信息化 , 2024 ( 6 ): 68 - 69 .
0
Views
0
下载量
2
CSCD
Publicity Resources
Related Articles
Related Author
Related Institution
京公网安备11010102007558号