Using Chest Scans for the Diagnosis of COVID-19

Abstract

COVID-19 is a viral infection that affects the human respiratory system. Since the victims of COVID-19 are increasing globally, this contagious disease is characterized as a pandemic by the World Health Organization (WHO). Laboratory testing is not considered effective for COVID-19 patients due to an increase in their absolute number. In this study, we proposed XCov-Net which is a deep learning model that can diagnose patients suffering from corona using chest X-ray scans. XCov-Net model is formed by the combination of the Xception model and XCov block. The proposed model performs multiclass and binary classification. The accuracy obtained in multiclass classification is 0.948 and 0.987 for four classes (COVID-19 cases, Normal cases, cases of Pneumonia caused by bacteria, cases of Pneumonia caused by a virus) and three classes (COVID-19 cases, Normal cases, Pneumonia cases) respectively. For two class classification, the model performed outstandingly good with accuracy recorded as 1.