基于多特征融合的心搏类型识别研究  

Research on Heartbeat Type Recognition Based on Multi-feature Fusion

作　　者：李润川[1,2] 张行进[1,2] 陈刚[2] 姚金良 于婕 王宗敏[1,2] LI Runchuan;ZHANG Xingjin;CHEN Gang;YAO Jinliang;YU Jie;WANG Zongmin(School of Information Engineering,Zhengzhou University,Zhengzhou 450001,China;Cooperative Innovation Center of Internet Healthcare,Zhengzhou University,Zhengzhou 450001,China)

机构地区：[1]郑州大学信息工程学院,河南郑州450001 [2]郑州大学互联网医疗与健康服务河南省协同创新中心,河南郑州450001

出　　处：《郑州大学学报（工学版）》2021年第4期7-12,共6页Journal of Zhengzhou University（Engineering Science）

基　　金：国家重点研发计划项目(2017YFB1401200);兵团重点领域科技攻关项目(2018AB017)。

摘　　要：心律失常是一种常见的心电活动异常,严重的可能危及人的生命。为了能准确诊断心律失常,提出了一种新的方法,用于心律失常诊断中对心搏的识别分类。首先对原始心电信号进行去噪预处理,并根据R峰位置获得心搏段。然后提取235单心搏特征点、R波幅值、PR间期、QT间期、ST段和RR间期作为特征参数,并对比分析不同特征组合下分类的性能,选出最佳的特征组合,最后基于最佳特征组合使用KNN模型对心搏进行分类。在MIT-BIH心律失常数据库上进行实验,并根据ANSI/AAMI分类方法对MIT-BIH心律失常数据库中的3种心搏类型:正常或束支传导阻滞(N)、室上性异位搏动(S)、心室异位搏动(V)进行分类。实验结果显示:S类心搏的灵敏度为87.8%,阳性预测值为95.1%;V类心搏的灵敏度为96.6%,阳性预测值为98.2%,测得的平均准确率为99.2%。与其他心搏分类方法相比,所提的基于多特征融合与KNN模型的心搏分类方法提高了分类准确率,具有较高的灵敏度和阳性预测值,对临床决策具有重要价值。Arrhythmia is a common abnormality of cardiac electrical activity,which may seriously endanger human life. Therefore,in order to accurately diagnose arrhythmia,this paper presents a new method for the recognition and classification of heartbeat in the diagnosis of arrhythmia. This paper proposed a new method for the recognition and classification of heartbeat in the diagnosis of arrhythmia. Firstly,the original ECG signal was denoised and preprocessed,and the heartbeat segment was obtained according to the R peak position.Then 235 single heartbeat feature points,R wave amplitude,PR interval,QT interval,ST segment and RR interval as feature parameters,and the performance of classification under different feature combinations were comparatively analyzed to select the best feature combination. Finally,the KNN model was used to classify the heartbeat based on the best feature combination. In this paper,experiments on MIT-BIH arrhythmia database,and according to ANSI/AAMI classification,they were classified three types of heart beats: normal or bundle branch block( N),supraventricular ectopic beat( S),and ventricular ectopic beat( V). The results showed that the sensitivity and positive predictive value of S type heart beats were 87. 8% and 95. 1%,respectively.The sensitivity and positive predictive value of V type heart beats were 96. 6% and 98. 2%,respectively. The average accuracy of measurement was 99. 2%. Compared with other cardiac classification methods,the proposed cardiac classification method based on multi-feature fusion and KNN model could improve the classification accuracy,with higher sensitivity and positive predictive value,and it was of great significance for clinical decision-making.

关 键 词：心电图 多特征融合 KNN模型 心搏分类 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]