体检人群肝脂肪病变者中医脉象信号的递归定量识别与分析

目的 运用研究非线性动力学的递归定量分析(recurrence quantification analysis, RQA)方法对体检人群肝脂肪病变者的脉象信号进行分析，探讨脉象信号非线性动力学特征对肝脂肪病变的识别价值。方法 运用ZY-Ⅰ型脉诊仪采集体检人群的脉象信号，根据腹部超声报告将体检人群分为肝脂肪病变组和非肝脂肪病变组；提取体检人群脉象信号RQA特征，并运用非参数检验分析两组人群脉象信号的RQA特征差异；基于脉象信号RQA特征，运用随机森林机器学习方法建立体检人群肝脂肪病的识别模型，并通过评价指标准确率、精确率、召回率、F1值、受试者工作特征曲线(receiver operating characteristic curve, ROC)及曲线下面积(area under the curve, AUC)评估模型识别性能。结果 肝脂肪病变组脉象信号RQA特征递归率、确定性、对角线长度的均值、递归熵、层状度、竖直/水平线段长度均值和最长竖直/水平线段长度均高于非肝脂肪病变组(P<0.05);基于脉象信号RQA特征建立的体检人群肝脂肪病变识别模型，其准确率为80.34%,精确率为...

Recurrence Quantification Identification and Analysis of Traditional Chinese Medicine Pulse Signals in the Health Check-up Population with Hepatic Steatosis

Objective: A nonlinear dynamical method, recurrence quantification analysis (RQA), was used to analyze the pulse signal of a check-up population with hepatic steatosis and to investigate the value of nonlinear dynamical characteristics of the pulse signal in identifying hepatic steatosis. Methods: The pulse signal was collected from the check-up population using ZY-I pulse diagnostic instrument, and the population was divided into hepatic fatty lesion group and non-hepatic fatty lesion group according to the abdominal ultrasound report; the RQA characteristics of the pulse signal of the check-up population were extracted, and the differences in RQA characteristics of the pulse signal between the two groups were analyzed by non-parametric test; based on the RQA characteristics of the pulse signal, the random forest (RF) algorithm was used to establish the identification model of hepatic fatty lesion in the population. Based on the RQA characteristics of the pulse signal, the RF algorithm was used to establish a variable identification model for hepatic adiposity in the check-up population, and the evaluation criteria including accuracy, precision, recall, F1-score, receiver operating characteristic curve (ROC) and area under the curve of ROC (AUC) were used to evaluate the identification effect of the model. Results: The RQA characteristics RR, DET, L, ENTR, LAM, TT, and Vmax were higher in the hepatic steatosis group than in the non-hepatic steatosis group (P<0.05); the accuracy of the identification model of hepatic steatosis in the check-up population based on the pulse signal RQA characteristics was 80.340%, the precision was 82.166%, the recall was 86.000%, the F1-score was 84.039%, and the AUC was 86.774%. Conclusion: Compared with the non-hepatic steatosis group, the pulse signal system of the hepatic steatosis group showed higher regularity, certainty and stability. The identification model of hepatic steatosis in the check-up population based on RQA characteristics can better distinguish the pulse signals of the hepatic steatosis group from those of the non-hepatic steatosis group, which can provide reference for the early warning and auxiliary diagnosis of hepatic steatosis.