Pneumonia Detection in Chest X - Ray Images Using Deep Learning and Grad-CAM++ Visualization

Pneumonia is a critical lung disease and the leading cause of death in children under five, responsible for 15 % of fatalities in this age group. Its severity can be exacerbated by pollution, overpopulation, and poor sanitation. Early detection is crucial for improving survival rates. Among imaging techniques, X-rays are preferred due to their low cost and simplicity. This study used the Pneumonia Chest X-ray dataset to classify regular and pneumonia-affected X-rays. Given that chest X-ray images contain numerous details, we proposed a pre-processing method for segmenting the lung area. Using the segmented images, we employed pre-trained models and trained them with the pre-processed images. In our experiments, EfficientNetB0 achieved the highest accuracy of 95.03% and an F-measure of 96.12%. Utilizing Grad-CAM++, we demonstrated that EfficientNetB0 and Inception-V3 effectively focus on pneumonia-infected lung regions, underscoring the potential of our approach for reliable pneumonia classification.