基于主成分分析和多种回归的肝硬化患者脉诊信息特征分析研究

目的:分析肝硬化患者的脉诊信息特征。方法:分别采集肝硬化患者的脉诊信息并与正常人对照进行分析。在对脉诊信息预处理后选取连续的多个周期的脉搏波作为研究对象,对各个周期进行12次谐波的最小二乘(LS)回归并建立数学模型,获取193个特征参数,根据上述参数对2类人群的脉诊信息进行特征分析,分别进行主成分分析和LS回归与Lasso回归的分类判决,获得肝硬化患者脉诊信息的变化特征。结果:肝硬化患者的脉诊信息与正常人比较有较明显的差异,主成分分析的判决准确率为81%,7参数拟合的回归判决准确率为98.77%,Lasso回归判决准确率为89%。结论:肝硬化患者机体的病理变化可以在桡动脉脉诊信息中得到体现并在诊脉的不同部位呈现较强的特征性,该特征经过不同的数学方法得到印证。本研究为中医"独取寸口"的诊脉方法提供了一定的科学依据,并在一定程度上为中医分寸、关、尺不同部位的诊脉方式提供了科学基础。

Research of features in the pulse waves of cirrhosis patients based on principal component analysis and regression analysis

Objective: To study the features in the pulse waves of cirrhosis patients.Methods: To collect the pulse waves of Cirrhosis patients and compare them with normal volunteers'.We choose continuous cycles waves with different period length and then do the LSQ regression by 12 harmonics fitting which are correspond with every cycle lengths to build mathematical model and extract 193 parameters after pretreatment,then do principal component analysis and regression for classification and identification to find the...