Early pleural effusion detection from respiratory diseases including COVID-19 via deep learning

Abstract

Pleural effusion is the build-up of excess fluid between the pleura layers around the lung. This fluid may be transudative or exudative. Pneumonia and cancer are common exudative causes of pleural effusion. Other causes include tuberculosis and recently discovered COVID-19. Physicians are able to diagnose pleural effusion through the use of chest radiographs. In this work, we propose, instead, the early detection of pleural effusion from tuberculosis, pneumonia, and COVID-19 diseases on chest radiographs using deep learning. The performance results show that the early detection of pleural effusion from pneumonia and tuberculosis have the highest accuracy. They further show that the deep learning architecture can distinguish bacterial pneumonia and COVID-19 diseases from pleural effusion the best.