Classification of COVID-19 electrocardiograms by using hexaxial feature mapping and deep learning

1,937 distinct patients' paper-based ECG report images from a publicly available dataset, including 250 COVID-19 patients, 77 myocardial infarction (MI) patients, 548 patients with abnormal heartbeats (recovered from COVID-19 or MI), 203 patients with MI history, and 859 people without any cardiac findings.

Automatic diagnosis of COVID-19 using a novel Convolutional Neural Network (CNN) architecture trained on hexaxial feature mapping images generated from 12-lead paper-based ECG data using the Gray-Level Co-Occurrence Matrix (GLCM) method.

Diagnostic accuracy and F1-Score for COVID-19 classification (COVID-19 vs. No-Findings, and COVID-19 vs. all Negative findings).