HRIDYA RAJ P(STM21CS034)
SAYANTH K(STM21CS054)
SIVANTH P K(STM21CS059)
VYSHAGH C(STM21CS065)

This system introduces a novel automated multi-classification approach for identifying lung abnormalities through chest X-ray and CT scan images. It tackles key challenges in medical imaging such as class imbalance, computational complexity, and the necessity for explainable AI to aid clinical decision-making. The project makes several significant contributions: It employs data augmentation techniques, specifically VAE, to address class imbalance and enhance classification performance. Extensive image preprocessing methods are applied to improve image quality while preserving crucial diagnostic information. The modified VGG-19 model is developed and optimized for chest X-
ray and CT scan datasets, demonstrating high accuracy and efficiency in detecting lung abnormalities. Comprehensive comparisons with transfer learning networks are provided to assess modified VGG-19’s precision and training time. Moreover, GradCAM is utilized to visualize the model’s classification process, supporting clinicians with explainable AI techniques. The results of the study showcase the effectiveness of the proposed framework in accurately identifying various lung diseases, including COVID- 19, Pneumonia, Emphysema, Fibrosis, Lung cancer, Tuberculosis across both imaging modalities. This demonstrates the system’s potential to enhance diagnostic accuracy and facilitate timely interventions, ultimately contributing to improved patient outcomes.