<![CDATA[Pneumonia ranks among the deadliest respiratory infections, particularly affecting young children, where rapid and precise detection proves essential for prompt intervention and averting severe outcomes. This research examines the deployment of three leading deep learning models VGG16, MobileNetV2, and ResNet50. This research categorizes pediatric chest X-rays into normal, bacterial pneumonia, and viral pneumonia classes using a dataset of children's chest radiographs (ages 1-5 years) from Guangzhou Hospital, enhanced through preprocessing, normalization, and augmentation techniques to boost model robustness. Model performance was assessed via accuracy, precision, recall, specificity, F1-score, G-Mean, and AUC metrics for thorough evaluation, revealing accuracies between 79% and 82% with VGG16 leading, followed by ResNet50 and MobileNetV2. Grad-CAM visualizations effectively highlighted key diagnostic regions on X-rays, offering clinicians valuable insights into infection sites identified by the models, and advancing automated chest X-ray systems for precise bacterial/viral pneumonia detection with supportive interpretability.]]>
Copyrights © 2026
