OWAE-Net: COVID-19 detection from ECG images using deep learning and optimized weighted average ensemble technique

Abstract

COVID-19 is an infectious disease that has cost millions of lives all over the world. A faster and safer diagnosis of COVID-19 is highly desirable in order to stop its spread. An electrocardiogram (ECG) signal-based diagnosis has shown its potential in the diagnosis of cardiac, stroke, and COVID-19. In this study, an ensemble of three deep learning models are used for COVID-19 detection in ECG images for multi-class classification. The results obtained with the weighted average ensemble technique have been improved by using the grid search technique. For multi-class classification, an optimized weighted average ensemble (OWAE) model classifies the ECG images with an accuracy of 95.29%, an F1-score of 95.4%, a precision of 95.5%, and a recall of 95.3%. In case of binary classification, VGG-19, EfficientNet-B4, and DenseNet-121 performed comparatively well with 100% accuracy. These results show that deep learning can be used in the diagnosis of COVID-19 disease using ECG images. © 2022 The Author(s)