基于卷积神经网络的心电图心博识别  

ECG heartbeat recognition based on convolution neural network

作　　者：王自强 刘洪运[2] 石金龙[2] 王卫东[2] WANG Ziqiang;LIU Hongyun;SHI Jinlong;WANG Weidong(School of Biological Science and Medical Engineering,Beihang University,Beijing 100191,China;Department of Medical Engineering Support Center,Chinese PLA General Hospital,Beijing 100853,China)

机构地区：[1]北京航空航天大学生物与医学工程学院,北京100191 [2]中国人民解放军总医院医学工程保障中心,北京100853

出　　处：《中国医学物理学杂志》2019年第8期938-944,共7页Chinese Journal of Medical Physics

基　　金：国家自然科学基金（61701540）;国家重点研发计划（2016YFC1305703A）;解放军总医院转化医学基金项目（2016TM-042）

摘　　要：心电图分析是诊断心律失常的最主要手段,心搏类型是诊断心律失常的重要信息,心搏的自动识别也是心律失常自动诊断的重要步骤。本研究尝试采用卷积神经网络自动识别心搏类型,使用的心电数据来源于MIT-BIH心律失常数据库,将心电信号的形态特征作为输入,采用端对端学习的网络结构。经过十折交叉验证测试。本研究网络识别13种心搏类型的平均准确率为99.24%,特异度达到99.59%。对于叠加不超过0.4 mV随机噪声的心电信号,本研究网络的识别准确率为99.07%。此外,将数据库的其中一个病人作为实测数据,得到的阳性预测为99.19%。研究结果表明文章的网络能自动学习输入特征,准确识别较多种类心搏且对噪声具有鲁棒性,为接下来的心律失常自动诊断提供可靠基础,同时也可能为基于心电图分析的其他相关诊断提供辅助决策支持。Electrocardiogram(ECG)analysis is the main method for diagnosing arrhythmia.The type of heartbeat plays a key role in the diagnosis of arrhythmia,and the automatic recognition of heartbeat is an important step in the automatic diagnosis of arrhythmia.Herein the convolution neural network is used to automatically identify the type of heartbeat.The ECG data used in this study are derived from MIT-BIH arrhythmia database.The morphological features of ECG are taken as inputs and an endto-end learning network structure is adopted.The results of 10-fold cross-validation test show that the average accuracy and specificity of the proposed network for the recognition of 13 types of heartbeats are 99.24%and 99.59%,respectively,and the recognition accuracy of the proposed network is 99.07%when adding random noise which is less than 0.4 mV.In addition,taking the clinical data of one of the patients from the database as the measured data,a positive predictive value of 99.19%can be obtained.The results prove that the proposed network can automatically learn the input features,accurately identify a variety of heartbeats,and has a good robustness to noise,which provides a reliable basis for the subsequent automatic diagnosis of arrhythmia,and may also provide decision-making support for other diagnoses based on ECG analysis.

关 键 词：心律失常 心搏类型识别 卷积神经网络 

分 类 号：R318[医药卫生—生物医学工程] TP391[医药卫生—基础医学]