基于“病脉结合”的2种不同疾病滑脉脉诊信息特征分类研究

Study on the Classification of Diagnostic Information of Rolling Pulse in Two Diseases Based on“Integration of Disease and Pulse”

目的研究同为滑脉的两种不同疾病患者脉诊信息的变化特征,为传统脉象可再分提供依据。方法采用"中医脉诊信息采集系统"分别采集同为滑脉慢性阻塞性肺病(COPD)患者与脂肪肝(FLD)患者各50例的左右手寸、关、尺部位的脉诊信息特征,应用MATLAB脉诊信息GUI分析程序进行脉诊信息预处理、193个参数提取和主成分分析、LS识别、Lasso3种不同分析方法的分类判决,分析2类人群脉诊信息之间的差异以及各自的信息特征。结果同为滑脉的2种疾病患者的脉诊信息之间存在很明显的差异,该差异在采用3种不同的数学方法分别进行统计的时具有较强的一致性。2种疾病患者的脉诊信息分析中,无监督学习方法分类识别的准确概率为67%,有监督学习的准确概率达79%-92%。2种疾病患者的脉诊信息之间的差异呈现规律性和部位特异性。结论滑脉在不同疾病状态下脉诊信息之间具有显著差异,可以进行再分类。

Objective To study the change characteristics of rolling pulse information in different diseases so as to provide evidence for the re- categorization of traditional pulse condition. Methods TCM pulse diagnosis information collection system was adopted to collect the pulse information characteristics on cun,guan and chi regions of hands in each of the 50 patients of chronic obstructive pulmonary disease( COPD) and fatty liver disease( FLD) with rolling pulse. GUI analyzing process of MATLAB pulse diagnostic information was used for the preprocessing of pulse diagnostic information,extraction of 193 parameters and cauterization with different methods such as principal component analysis,LS identification and Lasso3. The differences in pulse diagnostic information and individual information characteristics were analyzed between the two groups. Results The differences were very apparent in the pulse diagnostic information between the two groups of rolling pulse. These differences were high in consistency with 3 different mathematics methods adopted. In the analysis,the accurate probability of categorization identification was 67% with unsupervised learning method and was 79% to 92% with supervised learning method. The differences between the two groups presented regularity and regional specificity. Conclusion In the condition of different diseases,thepulse diagnostic information of rolling pulse is significantly different and can be re- categorized.