Abstract

Info

Title:: Arrhythmia classification method based on ResDCGAN and improved residual network

Author(s):: ZHANG Peiling; ZHANG Shuo; School of Physics and Electronic Information Engineering, Henan Polytechnic University, Jiaozuo 454003, China

Keywords:: Keywords: arrhythmia classification data balancing ResDCGAN coordinate attention ResNet34

Abstract:: Abstract: Arrhythmia is a common cardiovascular disease, and its diagnosis primarily relies on electrocardiograms (ECG). Leveraging computational techniques to achieve automatic arrhythmia classification can avoid human errors while improving diagnostic efficiency. An arrhythmia classification method based on a residual-structured deep convolutional generative adversarial network (ResDCGAN) and an improved ResNet34 is proposed to address the class imbalance in ECG data and the limitation of one-dimensional ECG processing. Specifically, the one-dimensional ECG signals are firstly denoised using variational mode decomposition. These denoised signals are then converted into two-dimensional Gramian angular summation field images. Subsequently, the proposed ResDCGAN is employed for data balancing, and finally, arrhythmia classification is carried out using a ResNet34 enhanced with coordinate attention. Experimental tests on the MIT-BIH arrhythmia database show that the proposed method achieves improvements of 0.22%, 1.60%, 1.89%, and 1.73% in accuracy, precision, recall rate, and F1-score, respectively, obtaining an accuracy of 99.66%. These results fully demonstrate the effectiveness of the proposed method, providing an effective solution for ECG data balancing and arrhythmia classification.