Online augmentation for quality improvement of neural networks for classification of single-channel electrocardiograms

Abstract

Currently, on the market, there are mobile devices that are capable of reading a person’s single-lead electrocardiogram (ECG). These ECGs can be used to solve problems of determining various diseases. Neural networks are onearameters of augmentations of the approaches to solving such problems. In this paper, the usage of online augmentation during the training of neural networks was proposed to improve the quality of the ECGs classification. The possibility of using various types of online augmentations was explored. The most promising methods were highlighted. Experimental studies showed that the quality of the classification was improved for various tasks and various neural network architectures.