Deep learning to distinguish COVID-19 from other lung infections, pleural diseases, and lung tumors

Abstract

COVID-19 is a highly infectious respiratory disease caused by severe acute respiratory syndrome coronavirus 2. It can lead to cough and fever and in some cases severe pneumonia. It is generally detected by reverse-transcription polymerase chain reaction and computed tomography scans. However, as it is a lung disease, it has common symptoms with other respiratory diseases. This necessitates us to carefully differentiate COVID-19 from such diseases during the diagnosis. This work aims to do that with the help of several deep learning architectures and chest radiographs. It specifically focuses on differentiating COVID-19 from pneumonia, pleural effusion and lung mass. During this analysis, it is shown that we can differentiate COVID-19 from other respiratory diseases using various deep learning architectures. It is further shown that ResNet-18 architecture produces the best overall performance in three scenarios of experiments.