基于注意力残差模型的心律失常分类研究

Classification of arrhythmia based on attention residual model

心电图是诊断心脏病的重要手段。针对现有算法对样本较少的心律失常敏感性较低的问题,设计了一种残差结构与注意力机制结合的心电图分类算法,利用注意力机制有效地降低算法的漏诊率。首先利用残差块提取心电信号中的重要特征,并通过注意力机制学习特征之间的关系,最后根据提取的特征采用2层全连接层对心电图进行分类。实验结果表明:在基于病人的心律失常分类方式下算法达到了99.42%的准确率,敏感性和特异度分别为96.88%和99.69%;在基于患者特异性的分类方式下,算法达到了98.24%的准确率,90.03%的敏感性和97.95%的特异度。对于样本数量较少的心律失常类别,其敏感性最高提升了10%.说明残差网络与注意力机制相结合能有效地提高分类的敏感性,并且有着较好的分类性能。

The electrocardiogram(ECG)is an important way to diagnose heart disease.However,the imbalance of samples leads to the fact that the existing algorithms have low sensitivity for arrhythmia with fewer sample.Therefore,this paper proposed an electrocardiogram classification algorithm which combines the deep residual network and attention mechanism,to reduce the rate of missed diagnosis in an effective way.The features of ECG signal will be extracted by residual block,and the weight between the features will be learned through the attention mechanism.Then,it classified to signal into two full-connected layers according to the extracted features.The result shows that the accuracy,sensitivity,and specificity under intra-patient paradigm reach 99.42%,96.88%and 99.69%;and underpatient-specific paradigm the accuracy,sensitivity and specificity reach 98.24%,90.03%and 97.95%.Moreover,the sensitivity for the heart disease with fewer sample is increased by 10%.The process and the result of this algorithm provide solid evidence for the truth that the sensitivity of ECG classification will be effectively improved under the combination of residual network and attention mechanism.Also,the result may prove that this algorithm has a better performance of classification.