结合卷积神经网络与注意力机制的多域特征融合ECG心率失常分类

CNN+Attention Multi-Domain Feature Fusion ECG Heart Rate Disorders Classification

心率失常是心血管疾病诊断的重要手段,其自动分类具有重要的临床意义。为了提高心率失常分类的准确性,结合一维卷积神经网络(Convolutional Neural Networks,CNN)和注意力机制(Attention)提出了一种CNN+Attention的深度学习模型,使用CNN提取心电信号的一维时域特征。针对一维时序心电信号时域特征表征能力有限的问题,使用短时傅里叶变换(Short-Time Fourier transform,STFT)将心电信号变换到时频域,通过Attention提取心电信号的时频域全局相关依赖关系,将时域与时频域特征融合对5种类型心电信号进行分类。在MIT-BIH数据集上验证了模型的有效性,所提模型对5种类型心电信号的平均分类准确率、精准率、召回率、灵敏度以及F1_Score分别为99.72%、98.55%、99.46%、99.90%以及99.00%。与已有先进方法对比,验证了所提模型具有先进的性能表现。

ECG arrhythmias are an important tool for the diagnosis of cardiovascular diseases,and their automatic classification is of great clinical importance.In order to improve the accuracy of ECG arrhythmias classification,a deep learning model of Convolutional Neural Networks(CNN)+Attention is proposed by combining one-dimensional convolutional neural network and attention mechanism,using CNN to extract one-dimensional time-domain features of ECG signals.To address the problem of limited ability to characterize the time-domain features of onedimensional temporal ECG signals,the Short-time Fourier Transform(STFT)is used to transform the ECG signals into the time-frequency domain,and the global correlation dependencies of the ECG signals in the time-frequency domain are extracted by Attention,and the time-domain and timefrequency domain features are fused to classify five types of ECG signals.The effectiveness of the model was verified on the MIT-BIH dataset.The average classification accuracy,precision,recall,sensitivity and F1_Score of the proposed model for the five types of ECG signals were 99.72%,98.55%,99.46%,99.90%and 99.00%,respectively.The comparison with existing advanced methods validates the advanced performance of the proposed model.