Development of a Hybrid Deep Learning System for Multi-Class Lung Disease Detection

Chest X-ray imaging is a critical component of modern medical imaging research. It facilitates rapid disease diagnosis that can address the limitations of interpretations that are manual. The chest X-ray images can be used to diagnose different spectrum of diseases which includes pneumonia, COVID-19, and tuberculosis but variability of clinical presentations gives notable challenges for traditional image processing methods. Recent research made use of Convolutional Neural Networks (CNNs). transfer learning and vision transformer (ViT) architectures. New emerging works suggest the integration of two or more methods in the above can enhance model transparency and usability in clinical applications. Despite the progress above, challenges that include handling imbalanced data, improving generalization across datasets, and the balancing of performance with interpretability are still in existence. This research proposes an hybrid deep learning framework that uses state-of-the-art transfer learning models which are ResNet50 and Transfer Learning together with data augmentation, class weighting, and evaluation strategies for robust multi-class X-ray classification. The new model (Xramodel) has been benchmarked against existing studies, and evaluations were done to demonstrate its effectiveness. The work contributes to the advancement of AI-driven diagnostic support systems to enhance clinical decision-making and improve patient outcomes.