Deep Learning for ECG Image Analysis: A Lightweight Approach for Covid-19 Diagnosis

Abstract

Since late 2019, Covid-19 has broken out causing immense pressure on healthcare systems worldwide. Fast detection of Covid-19 has become crucial in controlling and slow-pacing the virus outbreak. Innovative methods that are cheap, fast, and accurate for Covid-19 detection are of high importance to aid in the efforts of containment of the disease. In this study a novel method is proposed for Covid-19 detection through analysis of ECG image records. Three models are introduced for three classification schemas, Normal vs Covid-19, Covid-19 vs non Covid-19, Normal vs Covid-19 vs Abnormal HeartBeat. An overall accuracy of 98.6 %, 99 %, and 90 % respectively is achieved. Automatic detection of Covid-19 with computer aided systems using ECG images is achievable and very promising.