ML based prediction of COVID-19 diagnosis using statistical tests

Abstract

The first case of the novel Coronavirus disease (COVID-19), which is a respiratory disease, was seen in Wuhan city of China, in December 2019. From there, it spread to many countries and significantly affected human life. Deep learning, which is a very popular method today, is also widely used in the field of healthcare. In this study, it was aimed to determine the most suitable Deep Learning (DL) model for diagnosis of COVID-19. A popular public data set, which consists of 2482 scans was employed to select the best DL model. The success of the models was evaluated by using different performance evaluation metrics such as accuracy, sensitivity, specificity, precision, F1 score, kappa and AUC. According to the experimental results, it has been observed that DenseNet models, AdaGrad and NADAM optimizers are effective and successful. Also, whether there are statistically significant differences in each performance measure/score of the architectures by the optimizers was observed with statistical tests.