Deep learning (DL) has increasingly been adopted to support automated medical diagnosis, particularly in radiological imaging where rapid and reliable interpretation is essential. In this study, a hybrid architecture integrating convolutional neural network (CNN), residual networks (ResNet), and densely connected networks (DenseNet) was developed to improve automated disease recognition in chest X-ray images. This unified framework was designed to capture shallow, residual, and densely connected representations simultaneously, thereby strengthening feature diversity and improving classification robustness relative to conventional singlemodel or dual-model approaches. The model was trained and evaluated using the ChestX-ray14 dataset, comprising more than 100,000 X-ray images representing 14 thoracic disease classes. Performance was assessed using established metrics, including accuracy, precision, recall, F1-score, and the area under the receiver operating characteristic curve (AUC-ROC). A classification accuracy of 92.5% was achieved, representing an improvement over widely used machine learning (ML) and contemporary DL baselines. To promote transparency and clinical interpretability, Gradient-weighted Class Activation Mapping (Grad-CAM) was incorporated, enhancing clinician confidence in model decisions. The findings demonstrate that DL-based diagnostic support systems can reduce diagnostic uncertainty, alleviate clinical workload, and facilitate timely decision-making in healthcare environments. The proposed hybrid model illustrates the potential of advanced feature-integration strategies to improve automated radiographic interpretation and underscores the importance of explainable artificial intelligence (XAI) in promoting trustworthy deployment of medical artificial intelligence (AI) technologies.