Chest X-rays Analysis for Lung Disease Prediction and Interpretation Using Transfer Learning

Pneumonia, responsible for 16% of deaths among children under 5, is a severe lung infection primarily caused by bacteria or viruses. Diagnosis typically relies on manual examination of chest X-ray images. Leveraging computer vision and deep learning can expedite the diagnostic process, potentially saving lives. This study aims to enhance pneumonia diagnosis by introducing a robust image classifier using transfer learning techniques. We deployed popular convolutional neural network models, including VGG16 and ResNet50, for this purpose. Additionally, we utilized Gradient-weighted Class Activation Mapping (Grad-CAM) to assist doctors in result interpretation. Our dataset was sourced from the Guangzhou Women and Children's Medical Center. Our approach outperformed state-of-the-art methods with remarkable accuracies of 95.99% for ResNet50 and 96.53% for VGG16.