摘要
目的:探讨基于压力脉搏波与光电容积脉搏波信息融合的脉象参数在动脉粥样硬化性心血管疾病(ASCVD)发病风险分层评估中的应用价值,为ASCVD发病风险评估提供新的思路与方法.方法:运用压力-光电多源传感器脉象仪采集ASCVD不同发病风险人群的桡动脉压力脉搏波和指端光电容积脉搏波,基于压力脉搏波提取时域特征参数,融合光电容积脉搏波提取血液动力学特征参数;运用非参数检验对ASCVD不同风险人群时域特征和血液动力学特征进行组间比较;基于不同的特征组合,运用随机森林(RF)算法分别建立不同的ASCVD发病风险分层评估模型,计算模型评价准则准确率、精准率、召回率、F1-score,比较模型的性能.结果:ASCVD不同风险分层人群时域特征参数T、T1、T4、T5、H3/H1、H4/H1、T1/T、T4/T、W1、W2、W1/T、W2/T和血液动力学参数R、L、C、Pm组间比较差异具有统计学意义(P＜0.01);运用RF算法,基于时域特征和血液动力学特征建立的ASCVD发病风险评估模型获得较好的性能,其准确率为82.05％、平均召回率为80.95％、平均精确率为80.69％、平均F1-score为80.62％.结论:基于压力脉搏波和光电容积脉搏波信息融合可以获取更丰富的心血管信息,提高ASCVD发病风险分层模型的性能,基于多源信息融合的脉诊检测技术有望为ASCVD发病风险评估与监测提供新的工具.
Abstract
Objective:To explore the application value of pulse signal parameters extracted by fusing pressure pulse wave and photoconductive pulse wave in the stratified assessment of the incidence risk of atherosclerotic cardiovascular disease(ASCVD),and to provide a new idea and method for the risk assessment of ASCVD.Methods:The radial artery pressure pulse wave and fingertip photoconductive pulse wave of different risk groups of ASCVD were collected by pressure photoelectric multi-sensor pulse detection equipment.The time-domain parameters were extracted based on the pressure pulse wave,and the hemodynamic parameters were extracted by fusing photoconductive pulse wave;Non parametric test was used to compare the time-domain parameters and hemodynamic parameters different risk groups of ASCVD;Based on different pulse signal feature combinations,the random forest(RF)algorithm was used to establish the ASCVD incidence risk hierarchical assessment model,calculate the accuracy,precision,recall and F1 score of the model,and comprehensively compare the performance of different models.Results:The time domain parameters T,T1,T4,T5,H3/H1,H4/H1,T1/T,T4/T,W1,W2,W1/T,W2/T and hemodynamic parameters R,L,C,Pm of ASCVD with different risk stratification were significantly different(P＜0.01).The RF algorithm was used to establish an ASCVD incidence risk assessment model based on the parameters of pulse signals.When the parameters of time domain and hemodynamics jointly participate in the model construction,the performance of the model was optimal:its accuracy rate was 82.05％,the average recall rate was 80.95％,the average precision rate was 80.69％,and the average F1 score was 80.62％.Conclusion:The characteristics of pulse signals extracted from pressure pulse wave and photoconductive pulse wave can reflect certain cardiovascular information;Pulse diagnosis and detection technology based on multi-source information can be used to obtain richer cardiovascular information and improve the performance of the risk stratification model of ASCVD,and pulse diagnosis detection technology based on multi-source information fusion is promising provide new tools for risk assessment and monitoring of the incidence of ASCVD.
基金项目
国家自然科学基金面上项目(82074332)
上海科学技术委员会医疗器械领域项目(19441901100)
国家科技期刊平台
NSTL
万方数据