摘要
目的基于CART(分类回归树)决策树与BP(反向传播)神经网络算法探讨蒋小敏教授治疗骨痹的中医辨证规律。方法收集蒋小敏教授2013年1月-2021年9月于江西中医药大学附属医院国医堂的门诊医案数据,通过Microsoft Excel 2019建立蒋小敏治疗骨痹的临床医案数据库。采用R 4.1.2软件rpart函数实现CART决策树模型的构建。采用SPSS Modeler 18.0软件,建立BP神经网络的中医诊断模型。结果共建立了气滞血瘀证、湿热阻络证和肝肾亏虚证3个中医辨证模型。CART决策树算法建模结果显示:气滞血瘀证、湿热阻络证和肝肾亏虚证的预测准确率分别为73.54%、84.56%、84.98%,各模型的曲线下面积(ACU)值分别为0.738、0.795、0.748。BP神经网络算法建模结果显示:气滞血瘀证、湿热阻络证和肝肾亏虚证的预测准确率分别为75.13%、87.69%、83.77%,AUC值分别为0.764、0.829、0.817。结论CART决策树和BP神经网络2种不同的算法在中医诊断建模上均具有一定的可行性,能够优势互补。基于真实世界临床医案,挖掘出骨痹症状与证型之间的关系特点,并初步预测了蒋小敏教授的辨证思路,对骨痹的辨证提供了科学依据。
Objective Based on CART(classification and regression tree) decision tree and BP(back propagation)neural network algorithm to explore the law of TCM syndrome differentiation in treating bone arthralgia by Professor Jiang Xiaomin.Methods The clinical case data of Jiang Xiaomin in the Affiliated Hospital of Jiangxi University of Traditional Chinese Medicine from January 2013 to September 2021 were collected,and the clinical case database of Jiang Xiaomin in the treatment of bone paralysis was established through Microsoft Excel 2019.CART decision tree model construction using R 4.1.2 software rpart function.SPSS Modeler 18.0 software was used to develop a BP neural network model for TCM diagnosis.Results A total of three TCM identification models were established for the Qi stagnation and blood stasis evidence,the damp-heat blocking ligament evidence and the liver and kidney deficiency evidence.The modeling results of CART decision tree algorithm showed that the prediction accuracy of Qi stagnation and blood stasis evidence,damp-heat blocking evidence and liver and kidney deficiency evidence were 73.54%,84.56% and 84.98%,respectively,and the area under the curve(ACU) values of each model were 0.738,0.795 and 0.748,respectively.The modeling results of BP neural network algorithm showed that the prediction accuracy of Qi stagnation and blood stasis evidence,damp-heat blocking evidence and liver and kidney deficiency evidence were 75.13%,87.69% and 83.77% with AUC values of 0.764,0.829 and 0.817,respectively.Conclusion The two different algorithms,CART decision tree and BP neural network,are both feasible in modeling TCM diagnosis and can complement each other's strengths.Based on real-world clinical cases,the relationship between bone paralysis symptoms and evidence patterns was uncovered and the preliminary prediction of Professor Jiang Xiaomin's identification ideas provided a scientific basis for the identification of bone paralysis.
作者
李人亮
张平
胡子毅
叶菁
易莹
张恒毅
刘婉婷
李洪
吴梦文
Li Renliang;Zhang Ping;Hu Ziyi;Ye Jing;Yi Ying;Zhang Hengyi;Liu Wanting;Li Hong;Wu Mengwen(School of Traditional Chinese Medicine,Jiangxi University of Chinese Medicine,Nanchang 330004,China;The Affiliated Hospital of Jiangxi University of Chinese Medicine,Nanchang 330006,China)
出处
《世界科学技术-中医药现代化》
CSCD
北大核心
2023年第1期401-412,共12页
Modernization of Traditional Chinese Medicine and Materia Medica-World Science and Technology
基金
江西省中医药管理局中医药中青年骨干人才培养计划(赣中医药科教字[2020]2号),负责人:胡子毅
江西省卫生健康委员会科技计划项目(202130558):消痛汤调控IL-17信号通路治疗急性痛风性关节炎的作用机制研究,负责人:胡子毅。
关键词
CART决策树
BP神经网络
蒋小敏
骨痹
骨关节炎
辨证规律
CART decision tree
BP neural network
Jiang Xiaomin
Bone paralysis
Osteoarthritis
Pattern of discernment
