ABSTRACT
World has experienced a new potent challenge in the shape of Coronavirus disease 2019 (COVID-19). Rapid screening and detection of infected patients is important step in fighting against this disease, so that proper measures can be taken to stop it from further spreading. Majority of the countries who have been successful in controlling the disease, have done it through effective early detection. The same factor is very evident in the countries where COVID-19 has gone out of control that they were or are not successful in early detection of suspected patients. This paper presents an artificial intelligence-based approach to provide new screening approach to detect COVID-19 from X-ray images. More than thirty-five thousand local/international negative and positive corona X-ray images were obtained to train VGG-16 model. Proposed method has two classifiers, first classifier distinguishes between negative cases and other infected cases, second classifier identifies pneumonia and other infected cases. These other infected cases will be recognized as COVID-19. Experimental evaluation on different X-ray imaging were conducted where this method classified positive and negative cases very effectively. A comparative study with publicly available network such as COVID-NET is also carried out. Proposed method outperformed COVID-NET in all three major areas such as overall accuracy, sensitivity and specificity. Overall accuracy for our technique is 95.08%, while sensitivity and specificity values are 100% and 93.15% respectively, while overall accuracy, sensitivity, specificity values for COVID-NET are 52.36%, 86.79% and 27.39% respectively. © 2021 IEEE.