摘要
目的基于BP神经网络方法,建立冠状动脉临界病变患者证候要素及其常见组合的中医辨证诊断模型。方法采用临床流行病学研究方法,多中心收集814例冠状动脉临界病变患者中医四诊信息,由专家对证候要素及其组合进行判定,采用二进制位标识数据,“有”赋值为“1”,“无”赋值为“0”,建立数据库。基于BP神经网络构建冠状动脉临界病变患者证候要素及其常见组合的中医辨证诊断模型,随机从数据库中抽取70%数据作为训练集以建立模型,30%数据作为验证集用于验证模型准确性,同时计算模型受试者工作特征曲线下面积(AUC)以评估模型优劣。结果建立血瘀、气滞、痰浊、热蕴、气虚、阴虚和阳虚7种证候要素的BP神经网络中医辨证诊断模型,训练集准确率均超过90%,验证集准确率均超过85%,训练集平均准确率为94.51%,验证集平均准确率为88.29%,模型平均AUC为0.953。对5种常见证候要素组合形式气虚+血瘀、气虚+气滞、气滞+血瘀、气虚+阴虚、气虚+血瘀+气滞构建BP神经网络中医辨证诊断模型,训练集准确率均超过98%,验证集准确率均超过90%,训练集平均准确率为99.47%,验证集平均准确率为94.34%,模型平均AUC为0.996。结论通过BP神经网络算法,可形成准确率较高且符合临床实际的证候要素及其常见组合的中医辨证诊断模型,为规范冠状动脉临界病变中医证候诊断标准提供客观依据。
Objective To establish the TCM syndrome differentiation and diagnosis model of syndrome elements and their common combinations in patients with borderline coronary lesion(BCL)based on the method of back propagation(BP)neural network.Methods Using the method of clinical epidemiological research,the TCM four diagnostic information of 814 patients with BCL was collected from multi-centers,and the syndrome elements and their combinations were judged by the experts.Binary bits were used to identify data.If it was“yes”,the value was set to“1”,and if it was“no”,the value was set to“0”,and then these data were used to establish the database.Based on the BP neural network,the syndrome differentiation and diagnosis model of syndrome elements and their common combinations in patients with BCL were established.70%of the data were randomly extracted from the database as the training set to build the model,30%of the data as the test set to verify the accuracy of the model,and the area under curve(AUC)of the model was calculated to evaluate the pros and cons of the model.Results The BP neural network models for TCM syndrome differentiation and diagnosis of seven syndrome elements including blood stasis,qi stagnation,phlegm,heat excess,qi deficiency,yin deficiency and yang deficiency were developed.The accuracy of all training sets was above 90%,and the accuracy of all test sets was above 85%.The average accuracy of the training sets was 94.51%,the average accuracy of the test sets was 88.29%,and the average AUC of the models was 0.953.The BP neural network models of TCM syndrome differentiation and diagnosis for the five common combinations of syndrome elements were established,including qi deficiency+blood stasis,qi deficiency+qi stagnation,qi stagnation+blood stasis,qi deficiency+yin deficiency and qi deficiency+blood stasis+qi stagnation.The accuracy of all training sets was above 98%,and the accuracy of all tests sets was above 90%.The average accuracy of the training sets was 99.47%,the average accuracy of the test sets was 94.34%,and the average AUC of the models was 0.996.Conclusion Through the BP neural network algorithm,the TCM syndrome differentiation and diagnosis model of syndrome elements and their common combinations with high accuracy and in line with clinical reality can be formed,which would provide an objective basis for standardizing the diagnosis of TCM syndromes in BCL.
作者
刘超
高嘉良
董艳
黄信生
林飞
李易
张振鹏
李军
王阶
LIU Chao;GAO Jialiang;DONG Yan;HUANG Xinsheng;LIN Fei;LI Yi;ZHANG Zhenpeng;LI Jun;WANG Jie(Beijing University of Chinese Medicine,Beijing 100029,China;Guang’anmen Hospital,China Academy of Chinese Medical Sciences,Beijing 100053,China;Beijing Anzhen Hospital,Capital Medical University,Beijing 100029,China;The First Affiliated Hospital of Xinxiang Medical University,Xinxiang 453199,China;Yunnan Provincial Hospital of Traditional Chinese Medicine,Kunming 650021,China)
出处
《中国中医药信息杂志》
CAS
CSCD
2021年第3期104-110,共7页
Chinese Journal of Information on Traditional Chinese Medicine
基金
国家自然科学基金(81673847)
国家中医药管理局国家中医临床研究基地业务建设科研专项(JDZX2015248)
国家中医药管理局中医药传承与创新“百千万”人才工程(岐黄学者)——国家中医药领军人才支持计划项目(2018年)。
关键词
BP神经网络
冠状动脉临界病变
证候要素
常见组合
辨证诊断模型
BP neural network
borderline coronary lesion
syndrome elements
common combinations
syndrome differentiation and diagnosis model
作者简介
通讯作者:王阶,E-mail:wangjie0103@126.com。
