基于单心搏活动特征与BiLSTM-Attention模型的心律失常检测  

ARRHYTHMIA DETECTION BASED ON SINGLE HEARTBEAT ACTIVITY FEATURE AND BILSTM-ATTENTION MODEL

作　　者：李润川[1,2] 张行进[2,3] 王旭 陈刚[1,2,4] 冀沙沙[1,2] 王宗敏 Li Runchuan;Zhang Hangjin;Wang Xu;Chen Gang;Ji Shasha;Wang Zongmin(Research Institute of Industrial Technology,Zhengzhou University,Zhengzhou 450000,Henan,China;Cooperative Innovation Center of Internet Healthcare,Zhengzhou University,Zhengzhou 450000,Henan,China;State Key Laboratory of Mathematical Engineering and Advanced Computing,People's Liberation Army of China Information Engineering University,Zhengzhou 450003,Henan,China;Distance Learning School,Zhengzhou University,Zhengzhou 450000,Henan,China)

机构地区：[1]郑州大学产业技术研究院,河南郑州450000 [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450000 [3]解放军信息工程大学数学工程与先进计算国家重点实验室,河南郑州450003 [4]郑州大学远程教育学院,河南郑州450000

出　　处：《计算机应用与软件》2019年第10期145-150,共6页Computer Applications and Software

基　　金：国家重点研发计划项目(2017YFB1401200);兵团重点领域科技攻关计划项目(2018AB017);教育部科技发展中心“云数融合科教创新”基金资助课题(2017A11017)

摘　　要：为了更准确地检测心律失常,提出基于单心搏活动特征与BiLSTM-Attention模型的心律失常检测方法。采用MIT-BIH心律失常数据库对算法进行验证,用双正交小波变换去除噪声干扰;通过二进样条小波变换的模极大极小值对检测R波峰值位置,并提取QRS波群数据及RR间期;使用BiLSTM-Attention分类模型进行心搏识别。实验结果表明,N、S、V和F类心搏的灵敏度分别为99.76%、94.74%、97.53%、83.93%,阳性预测值分别为99.76%、94.03%、97.53%、87.04%,F1综合指标达到了99.40%,证明了该算法的有效性。In order to detect arrhythmia more accurately,this paper proposed arrhythmia detection method based on single heartbeat activity characteristics and BiLSTM-Attention model.We used MIT-BIH arrhythmia database to verify the algorithm,and biorthogonal wavelet transform was used to remove noise interference.Then,the peak position of R-wave was detected by the modulus minimax value pairs of binary spline wavelet transform,and we extracted QRS wave group data and RR interval.Finally,the BiLSTM-Attention classification model was used for heartbeat recognition.The experimental results show that the sensitivity of N,S,V and F type beats are 99.76%,94.74%,97.53%,83.93%,the positive predictive value are 99.76%,94.03%,97.53%,87.04%,and the comprehensive index of F1 reaches 99.40%,which proves the effectiveness of the proposed algorithm.

关 键 词：心律失常 单心搏活动特征 注意力机制 双向LSTM模型 心搏分类 

分 类 号：TP3[自动化与计算机技术—计算机科学与技术]