Classification of COVID-19 and pleural effusion on chest radiographs using CNN fusion

Abstract

Coronovirus disease 2019 (COVID-19) is a very contagious respiratory infection caused by a coronavirus. Some of its common symptoms are fever, cough, headache, and loss of taste and smell. It may cause severe medical complications, such as pneumonia, blood clots, and acute respiratory distress syndrome. Though with low incidence in general, it can also result in pleural effusion. The use of deep learning architectures on diagnostic imaging has led researchers to detect and classify diseases of various kinds with performances comparable to the diagnostic accuracy of medical doctors. This paper uses deep learning architectures singly and jointly to classify COVID-19 and pleural effusion on chest radiographs. In general, when architectures are fused, the results are better than those obtained when each architecture is employed separately. However, such an improvement comes at the expense of system complexity. The classification results are the best when the architectures are fused with majority voting method.