基于马尔可夫转移场与改进MobileNetV2的心律失常分类方法

Markov transfer field combined with modified MobileNetV2 for arrhythmia classification

心律失常自动分类对心血管疾病的预防尤为重要,本研究提出一种基于马尔可夫转移场(MTF)和改进MobileNetV2网络的心律失常图像分类方法。首先将原始心电(ECG)信号进行预处理和数据增强,并通过MTF将处理后的ECG片段转变为具有时间关联性的二维图像。其次在MobileNetV2网络的模块中融入高效通道注意力。将正常心拍、左束支传导阻滞、右束支传导阻滞和起搏心拍4种类型的ECG信号通过改进MobileNetV2网络进行分类。结果表明改进MobileNetV2模型复杂度仅略高于原始MobileNetV2,在心律失常分类准确率上,比原始MobileNetV2网络提高0.89%,达到99.71%,实现了对4种不同类型的ECG信号的有效分类。

The automatic arrhythmia classification is critical for cardiovascular disease prevention.An approach for arrhythmia classification based on Markov transfer field(MTF)and modified MobileNetV2 network is presented.After preprocessing and data enhancement for the original electrocardiogram(ECG)signals,MTF maps the processed ECG segments into two-dimensional images with temporal correlation,and then a modified MobileNetV2 network which incorporates with efficient channel attention classifies the ECG signals of 4 types:normal beat,left bundle-branch block,right bundle-branch block,and paced beat.The results show that the modified MobileNetV2 is slightly more complex than the original MobileNetV2,and it has a classification accuracy of 99.71%,which is 0.89%higher than the original MobileNetV2,demonstrating that the proposed approach can achieve the effective arrhythmia classification.