Abstract

COVID-19 is one of the most serious diseases caused by the SARS coronavirus. This is a fatal disease, and it becomes difficult to save the life of the infected person because it progresses very rapidly and starts in the lungs and causing damage to the entire body. With the help of advanced machine learning techniques, this disease can be detected early and with very little probability of error. In this paper, we have developed two deep transfer learning models, VGG 16 and MobileNetv2, to detect COVID-19. Both models were applied to datasets based on human lung CT scan images, and their performance was evaluated. To analyze and compare their performance, we have used several performance met- rics such as Prevalence, Null Error Rate, False DR, Negative PV, False OR, LR ratio (+), LR ratio (-), CSI, Accuracy, FM Index, BM, Diagnostic Ratio, MK and Critical Success Index, that have not been used in previous papers.