This project focuses on developing an intelligent, end-to-end radiology assistance system capable of automatically analyzing chest X-ray images to detect thoracic abnormalities using deep learning. The system integrates image classification, explainable AI visualization, and automated medical report generation to support radiologists in diagnosis, triage, and decision-making. A deep-learning model such as EfficientNet-B0 is trained on the CheXpert dataset, enabling it to predict multiple chest pathologies including pneumonia, pleural effusion, atelectasis, cardiomegaly, and more. The model performs multi-label classification, outputting probability scores for each disease. To ensure transparency and trust, the system incorporates Grad-CAM–based visual explanations, highlighting the specific lung regions contributing to each prediction.