Detecting COVID-19 from ECG images using deep learning methods

The area to affect the whole world and declared as global sources by WHO on March 11, 2020, Covid-19 diseases mainly affect Multiple organ systems, mainly large and heart. Isolation and early initiation of treatment is essential to minimize exit damage and death during periods of time. Measurement methods to be used in instrument inspection should consist of new methods due to the long measurement end, along with incorrect results. Covid-19 status can be detected as Abnormal heartbeat, Myocardial infarction, History of Myocardial Infarction and Normal. Disease detections traditionally made by specialist doctors in the field can lead to ill-treatment caused by human error. In this thesis, deep learning methods were used to diagnose with ECG. The aim of this study is to propose a new approach with high success rate for the detection of diseases using ECG images and to analyze detailed test results. A publicly available dataset containing 5-class ECG images was used in this study. Training and testing processes were carried out using the EfficientNetB0 convolutional neural network architecture. Afterwards, the results were analyzed in detail, graphs were drawn and the results were compared with other studies in the literature. The proposed multi-class classification architecture offers 99.13% accuracy. With the success achieved, it was superior to other studies in the literature. This study will contribute to the rapid and reliable detection of 5 different findings that can be detected from ECG images and to more accurate treatment of patients. Keywords: Covid-19, Classification, EfficientNet, Electrocardiogram, ECG, Deep Learning, Convolutional Neural Network