PneuVision: An Explainable AI-Enabled Deep LearningFramework for Chest X-Ray Severity Assessment

Medical imaging, particularly chest radiography, remains a critical tool for identifying lung-related conditions such as pneumonia, tuberculosis, and COVID-19. However, interpreting chest X-rays can be time-consuming and subject to variability among radiologists. This study introduces PneuVision, an explainable deep learning framework designed to classify chest X-ray images into four severity categories—Normal, Mild, Moderate, and Severe—while maintaining high interpretability. The system uses a pre-trained DenseNet-121 model fine-tuned on a custom-labeled dataset, where Kaggle samples served only as references for defining severity criteria. Grad-CAM visualization is employed to enhance transparency by highlighting lung regions that contribute to each prediction. Experimental findings demonstrate improved performance compared to standard CNN and ResNet models. By integrating explainable AI, PneuVision supports reliable and interpretable automated analysis for radiological applications.