Segmentation of pneumothorax from chest radiographic images using deep learning

Abstract

Pneumothorax is a thoracic disease condition causing air to leak into the area newlinesurrounding the lungs and the walls of the chest. The leaked air then pushes the newlineoutside boundary of the lung and can result in the collapse of lungs. This may be a newlinecomplete collapse of lungs or its part. Pneumothorax can occur due to: blunt chest newlineinjury, lung disorders like pneumonia, lung cancer, bronchitis, pleural effusion etc. newlineIn pneumothorax treatment, a small tube is inserted between the ribs so as to newlinerelease the air trapped into the cavity. The lungs slowly decompresses on release of newlinethe air. The proper diagnose of the lungs disorder is crucial to avoid any serious harm newlineto mankind. It is difficult to diagnose pneumothorax by physical examination of a newlinepatient. The chest radiographic imaging is at present the widely accepted method for newlinescreening and diagnosis of pneumothorax. The reading of chest radiographs for newlinepresence of pneumothrax is to be done by expert radiologists and there is a shortage newlineof expert radiologists. An AI based deep neural network can detect pneumothorax newlinefrom chest X-ray image and can provide a confident diagnosis to non-radiologists. newlineSo, there is a requirement to design a model which can precisely identify the presence newlineof pneumothorax on chest radiographic images. newline