In today’s data-driven healthcare landscape, the rapid growth of medical data and the demand for accurate and timely diagnosis have accelerated the adoption of Artificial Intelligence (AI) and deep learning techniques. This study investigates the application of deep learning models for detecting pulmonary diseases from chest X-ray images. Several convolutional neural network (CNN) architectures, including ResNet50, MobileNetV2, Xception, DenseNet121, and a hybrid DenseNet121 + Vision Transformer (ViT) model, were implemented and evaluated using a dataset containing labeled chest X-ray images of conditions such as pneumonia, cardiomegaly, and pleural effusion.To improve model performance, advanced preprocessing techniques such as image augmentation, pixel normalization, and class imbalance handling were applied. These steps ensured better generalization across different disease categories. The models were evaluated using metrics including Accuracy, Precision, Recall, F1-Score, and AUC-ROC. Among all models, the hybrid DenseNet + ViT approach achieved the highest accuracy of 96%, demonstrating the effectiveness of combining convolutional and transformer-based methods.Model interpretability was enhanced using Grad-CAM, which provided visual explanations by highlighting important regions influencing predictions. This improved transparency and reliability in medical decision-making. Additionally, weighted loss functions addressed class imbalance, ensuring fair performance across all disease classes.This research underscores the transformative potential of deep learning in medical diagnostics and points toward the future of AI in healthcare, where AI systems work alongside healthcare professionals to enhance diagnostic accuracy and efficiency. The findings set the stage for future advancements, including the integration of multimodal data and the use of federated learning for privacy-preserving AI applications across institutions.