Deep Learning Approach for Diagnosis of Thoracic Diseases from Chest X-Ray Images

Abstract

The diseases that affect the heart, lungs, chest wall, esophagus, and diaphragm are known as thoracic diseases. Some examples of the thoracic disease are cystic fibrosis, emphysema, hernia, pneumothorax. These diseases can prove to be fatal if there is no timely diagnosis or proper treatment given to the patient. The diagnosis of these diseases in time is crucial to provide appropriate treatment and save a life. X-Ray is considered an effective and economical diagnostic tool to detect various pathological conditions. Fourteen different thoracic diseases namely Consolidation, Pneumothorax, Emphysema, Effusion, Pleural_thickening, Nodule, Infiltration, Hernia, Pneumonia, Cardiomegaly, Atelectasis, Fibrosis, Edema, and Mass will be detected from the X-Ray images. Each image has fourteen disease image labels which are text-mined using NLP (Natural Language Processing) techniques. The 121-layer Convolutional Neural Network (CNN), also known as DenseNet121 will be used as a means of classifying thoracic diseases of the Chest with modifications in the network of DenseNet121.