Lung Disease Classification Utilizing Convolutional Neural Network with CXR Imaging

Lung diseases present significant and widespread health challenges globally. Each year, a considerable portion of the world's population receives diagnoses for various lung conditions, affecting individuals of all ages, including adults, teenagers, children, smokers, and non-smokers alike. This classification underscores the status of lung diseases as a prominent cause of death and disability on a global scale. To enhance survival rates and prevent complications in the successful treatment of chest diseases, a Convolutional Neural Network (CNN) classification system was developed. This system focuses on the diagnosis of six distinct types of lung diseases-effusion, atelectasis, infiltration, nodule, mass and pneumothorax-through the analysis of chest X-rays. The dataset used for training the neural network model consisted of 8,125 sample images. The developed model demonstrated an impressive accuracy of 82.53% with a Cohen's kappa value of 0.788. This CNN classification model has practical utility and can be implemented on computer-based, mobile-based platforms, serving as supplementary tools for radiologists and doctors in the decision-making process during diagnostic evaluations of lung diseases.