COVID-19 Chest X-Ray Classification Using Residual Network

In 2019, the Covid-19 pandemic has spread across the globe and causing significant disruptions to daily life. Those who have tested positive for Covid-19 may experience long-term respiratory problems as the virus can damage the lungs. Specifically, patients who have recovered from Covid-19 may develop white spots on their lungs. This can be difficult to distinguish from normal lung tissue. Consequently, researchers have conducted extensive studies on image classification of Covid-19 chest x-rays, which has become a popular topic of investigation over the past two years. In this research, four datasets were utilized for image classification including COVID-19 Radiography, Chest X-ray, COVID-19, and CoronaHack datasets. All these datasets were sourced from Kaggle. The pre-trained ResNet152 model was used in conjunction with a transfer learning technique. Results indicated that the pre-trained ResNet152 with early stopping provided the highest accuracy among the techniques tested. In this research, the COVID-19 Radiography dataset achieved an accuracy of 95.61%, while the Chest X-ray dataset achieved an accuracy of 97.59%. CoronaHack dataset and COVID-19 X-ray dataset achieved accuracies of 93.59% and 100%, respectively.