三种T波终点检测算法的对比研究——小波变换法、累积积分面积法及梯形面积法  

Comparative Study on the Three Algorithms of T-wave End Detection:Wavelet Method,Cumulative Points Area Method and Trapezium Area Method

作　　者：黎承涛[1,2] 张永亮[1] 何子军[1] 叶骏[1,2] 胡福松[1] 马祖长[1] 王敬志[1,3] 

机构地区：[1]中国科学院合肥智能机械研究所,合肥230031 [2]安徽大学电子信息工程学院,合肥230039 [3]中国科学技术大学自动化系,合肥230026

出　　处：《生物医学工程学杂志》2015年第6期1185-1190,1195,共7页Journal of Biomedical Engineering

基　　金：国家自然科学基金项目资助(61301059);国家科技支撑计划项目资助(2013BAH14F01)

摘　　要：本研究通过选取QT数据库中20例共3 569个心拍的心电数据,对三种不单纯依赖阈值的T波终点检测算法(小波法、累积积分面积法、梯形面积法)的检测性能进行对比分析,评估出最适宜于临床检测的T波终点检测算法。首先,基于小波变换的多尺度分析方法定位QRS波群及T波;然后,将T波区分为四种形态(正向、反向、双向:+-/-+),分别采用三种算法用于T波终点的检测;最后,文中提出一种基于T波形态的自适应选择T波终点检测算法,并对其进行实验验证。结果表明:该自适应选择方法相比单一的T波终点检测算法,有着更好的检测性能,其灵敏度、阳性预测度和时间差分别为98.93%、99.11%和(-2.33±19.70)ms。因此,根据T波形态自适应选择T波终点的检测算法有助于提高T波终点的检测效率。In order to find the most suitable algorithm of T-wave end point detection for clinical detection, we tested three methods, which are not just dependent on the threshold value of T-wave end point detection, i.e. wavelet method, cumulative point area method and trapezium area method, in PhysioNet QT database （20 records with 3 569 beats each）. We analyzed and compared their detection performance. First, we used the wavelet method to locate the QRS complex and T-wave. Then we divided the T-wave into four morphologies, and we used the three algorithms mentioned above to detect T-wave end point. Finally, we proposed an adaptive selection T-wave end point detection algorithm based on T-wave morphology and tested it with experiments. The results showed that this adaptive selection method had better detection performance than that of the single T-wave end point detection algorithm. The sensitivity, positive predictive value and the average time errors were 98.93%, 99.11% and （--2.33~19.70） ms, re- spectively. Consequently, it can be concluded that the adaptive selection algorithm based on T-wave morphology im proves the efficiency of T-wave end point detection.

关 键 词：T波终点 小波变换 累积积分面积法 梯形面积法 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]