文献类型：期刊文献

中文题名：冠状动脉不同阻塞程度冠心病患者的中医脉图特征参数分析

英文题名：Analysis of characteristics of pulse-graph parameters in patients with different degree of coronary artery occlusion

作者：刘璐[1,2,3];张春柯[2,3];颜建军[4];郭睿[2,3];王忆勤[2,3];燕海霞[2,3];张叶青[5]

机构：[1]甘州区人民医院,张掖734000;[2]上海中医药大学基础医学院;[3]上海市健康辨识与评估重点实验室;[4]华东理工大学机械与动力工程学院;[5]上海市中医医院心内科

年份：2022

卷号：45

期号：8

起止页码：835

中文期刊名：北京中医药大学学报

外文期刊名：Journal of Beijing University of Traditional Chinese Medicine

收录：CSTPCD;;北大核心:【北大核心2020】；CSCD:【CSCD2021_2022】；

基金：国家自然科学基金面上项目(No.82074332);上海科学技术委员会生物医药领域项目(No.19441901100);上海健康辨识与评估重点实验室项目(No.21DZ2271000)。

语种：中文

中文关键词：冠心病;冠状动脉阻塞;脉诊;时域分析;多尺度熵分析;机器学习

外文关键词：coronary heart disease;coronary artery occlusion;pulse diagnosis;time-domain analysis;multiscale entropy analysis;machine leaning

摘要：目的基于脉图特征等信息建立冠心病患者冠状动脉不同阻塞程度评估模型,探讨中医脉图特征的临床诊断价值。方法采用Smart TCM-I型脉象仪采集脉象信息。根据冠状动脉造影检查报告,将531例冠心病及疑似冠心病患者按照冠状动脉不同阻塞程度分为冠状动脉非阻塞组、冠状动脉轻度阻塞组、冠状动脉中/重度阻塞组3组。运用时域分析法和多尺度熵(MSE)分析法提取不同组别脉图的时域特征和多尺度熵特征,并运用非参数检验的方法比较冠状动脉不同阻塞程度患者脉图特征参数的组间差异;基于脉图特征参数,运用随机森林(RF)机器学习算法建立冠状动脉不同阻塞程度评估模型。结果与冠状动脉非阻塞组相比,冠状动脉轻度阻塞组和冠状动脉中/重度阻塞组脉图时域特征主波峡/主波幅值比(h2/h1)、重博前波/主波幅值比(h3/h1)增大,差异具有统计学意义(P<0.05)。与冠状动脉非阻塞组及冠状动脉轻度阻塞组相比,冠状动脉中/重度阻塞组多尺度熵特征MSE1、MSE2、MSE3、MSE4、MSE5均减小,差异具有统计学意义(P<0.01)。本研究基于531例样本的脉图特征等信息建立了冠状动脉不同阻塞程度评估识别模型,当脉图时域特征和多尺度熵特征全部参与建模时,模型平均准确率最高(86.79%)。结论脉图检测技术对于冠状动脉危险事件的评估具有潜在的应用价值。
Objective To establish a model for assessing the degree of coronary artery obstruction based on features of their pulse-graph parameters,so as to explore the clinical diagnostic value of pulse diagnosis.Methods Smart TCM-I electropulsograph was used to collect pulse samples.According to degrees of coronary artery occlusion based on coronary angiography(CAG)result,531 patients with coronary heart disease(CHD)or suspicious CHD were divided into three groups:non-occlusion group,mild occlusion group and moderate/severe occlusion group.The time-domain features and MSE features of pulse samples were extracted by time-domain analysis and Multiscale entropy(MSE),and the differences in pulse features among different groups were compared with nonparametric test.Based on the features of pulse-graph parameters,the model for assessing the degree of coronary artery occlusion was established with the random forest(RF)machine leaning method.Results Compared with the non-occlusion group,h2/h1 and h3/h1 among time-domain features of pulse-graph were significantly increased in mild occlusion group and moderate/severe occlusion group(P<0.05).Compared with the non-occlusion group and mild occlusion group,MSE1,MSE2,MSE3,MSE4 and MSE5 were significantly decreased in moderate/severe occlusion group(P<0.01).Based on the characteristics of pulse-graph parameters of 531 cases,a model for assessing degree of coronary artery occlusion was established with a 86.79%average accuracy rate.Conclusion Pulse detection technology has potential application value in the assessment of coronary risk events.