DenseNet-121 for Intelligent Lung Disease Detection

Lung diseases can be fatal if not detected and treated immediately. They are responsible for over 1.6 million deaths worldwide as reported by WHO, making early detection and treatment crucial for improving this number. Chest X-Ray (CXR) imaging has emerged as a pivotal tool in diagnosing these diseases, given its widespread availability and relatively low cost. Although CXR scans are considered the most effective imaging technique in medical science, the identification and interpretation of diseases could be a challenging task for the physicians, particularly the identification of the affected area accurately. Much research has been done in the field of pulmonary diseases, but its clinical effectiveness is still an issue. This study aims to evaluate various techniques used in the diagnosis of lung diseases, analyze the most effective technique currently available, followed by identifying its limitations and drawbacks, and lastly propose a new model for lung disease detection that outperforms the existing models. An attempt has been made to achieve this objective by using machine learning algorithms such as CNN and DenseNet-121. This paper presents a comprehensive study on the application of DenseNet-121, a specific configuration of DenseNet, for the detection of lung diseases from CXR images. The proposed solution also involves the development of a web application that allows patients to upload their CXR images and get diagnosed. The stated web application uses DenseNet-121 to accurately predict any form of lung disease(s) through the CXR images. Resultantly, the proposed method achieved a mean AUC of 0.843. Further, the average outcome of the proposed approach is better as compared to the existing research approaches.