Abstract
Lung problems greatly affect people’s health worldwide, leading to many illnesses and fatalities every year. Chronic obstructive pulmonary disease (COPD) causes around responsible for about 3 million deaths each year, ranking as the third most deadly condition worldwide. Pneumonia, tuberculosis, and lung cancer exacerbate this problem. Early detection is crucial for quick treatment and better health results; yet, diagnosing accurately is challenging because of disease structures’ similarity and individual variations in the same condition. An automated deep learning-based system has been developed for categorizing lung diseases using chest X-ray images. DenseNet-121 model, famous for quick feature sharing and less repetition, helped improve performance through transfer learning. Images were resized to 128x128 pixels and then altered by rotating, flipping, and adjusting their contrast to improve how the model understands them. Sigmoid activation was utilized as the final layer in multi-label classification, fine-tuned with Adam optimizer at a learning rate set at 0.001 was used, along with binary cross-entropy as the loss function. The model scored nine-five. Twenty percent training accuracy and ninety-four. Eighty-nine percent validation accuracy after twenty epochs shows strong performance.
View Research Publication: