| 基于单心搏活动特征与BiLSTM-Attention模型的心律失常检测 |
| |
| 引用本文： | 李润川,张行进,王旭,陈刚,冀沙沙,王宗敏.基于单心搏活动特征与BiLSTM-Attention模型的心律失常检测[J].计算机应用与软件,2019,36(10):145-150. |
| |
| 作者姓名： | 李润川 张行进 王旭 陈刚 冀沙沙 王宗敏 |
| |
| 作者单位： | 郑州大学产业技术研究院 河南郑州450000;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;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;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;郑州大学产业技术研究院 河南郑州450000;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;郑州大学远程教育学院 河南郑州450000 |
| |
| 基金项目： | 国家重点研发计划;兵团重点领域科技攻关计划;创新基金 |
| |
| 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。
|
| 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 |
ARRHYTHMIA DETECTION BASED ON SINGLE HEARTBEAT ACTIVITY FEATURE AND BILSTM-ATTENTION MODEL |
| |
| Li Runchuan,Zhang Hangjin,Wang Xu,Chen Gang,Ji Shasha,Wang Zongmin.ARRHYTHMIA DETECTION BASED ON SINGLE HEARTBEAT ACTIVITY FEATURE AND BILSTM-ATTENTION MODEL[J].Computer Applications and Software,2019,36(10):145-150. |
| |
| Authors: | Li Runchuan Zhang Hangjin Wang Xu Chen Gang Ji Shasha Wang Zongmin |
| |
| Affiliation: | (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) |
| |
| Abstract: | 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. |
| |
| Keywords: | Arrhythmia Characteristics of single heartbeat activity Attention mechanism Bidirectional LSTM model Heartbeat classification |
| 本文献已被 维普 万方数据 等数据库收录！ |
|
