基于卷积神经网络的ECG心律失常分类研究

Classification of ECG arrhythmias based on convolutional neural network

基于心电信号进行心律失常自动检测和分类识别研究,辅助临床医生进行心血管相关疾病诊断。采用MIT-BIH数据库作为数据源,对该数据库心电数据进行小波分解与重构去噪后,构建卷积神经网络模型,结合Adam优化器,并优化丢弃值、训练步数和批大小三个超参数来优化模型,使用准确率、灵敏性和正预测率三个指标评价模型性能。实验结果表明,模型实现心律失常五分类的整体准确率大于99%,与现有模型性能相比,准确率提升1.2%。

Automatic detection and classification of arrhythmias based on ECG signals was used to assist clinicians in the diagnosis of cardiovascular diseases.MIT-BIH database was used as the data source,and the convolutional neural network model was constructed after wavelet decomposition and de-noising reconstruction of the ECG data of the database.Adam optimizer was selected to optimize the model by optimizing the three super parameters of the discard value,the number of training steps and the batch size.The accuracy,sensitivity and positive prediction rate were used to evaluate the model performance.The evaluation result shows that the overall accuracy of the model to achieve five classifications of arrhythmia is 99%,which is 1.2%higher than the performance of the existing model.