Multi-Disease Diagnostic Framework Using VGG19-Based Deep Visual Feature Extraction from Medical Images

Effective clinical decision-making depends on the early and accurate diagnosis of thoracic illnesses. In order to analyze chest X-ray and CT pictures, this work proposes a deep learning-based multi- disease diagnosis system that uses transfer learning with the VGG19 convolutional neural network. The suggested model is trained on publically accessible datasets that have undergone extensive preprocessing, such as data augmentation, scaling, and standardization to improve robustness. According to experimental findings, training accuracy steadily increases to about 70%, while validation accuracy stays steady at about 50%, suggesting that dataset complexity and inter-class similarity limit generalization. While the validation loss exhibits minor variations, indicating the existence of moderate overfitting, the training loss generally displays a decreasing trend. In spite of this, the framework successfully acquires discriminative characteristics pertinent to the categorization of lung diseases. The suggested method supports dependable and data-driven healthcare decision-making by showcasing the potential of transfer learning for medical imaging applications and offering a scalable basis for AI-assisted diagnostic systems.