Enhancing Pneumonia Detection from Chest Radiographs Through a VGG-16-based Deep Learning Approach

Pneumonia is a significant respiratory disease with high global burdens, especially in resource limited settings where access to specialized radiology is restricted. Early and reliable diagnosis is essential for effective clinical intervention, yet manual interpretation of chest X-ray images is often time-consuming and subject to inter-observer variability. This framework employs deep learning for automated pneumonia detection using chest X-ray images, leveraging transfer learning with a pre-trained VGG-16 model and a custom DNN classifier that incorporates batch normalization and dropout layers to ensure stable training and prevent overfitting. The model achieved an accuracy of 92.79%, precision of 94.12%, recall of 94.36%, an F1-score of 94.24%, and an AUC of 0.98 on the public Chest X-Ray images (Pneumonia) dataset published on Kaggle, outperforming several state-of-the-art CNN methods. These performance metrics indicate that the proposed method exceeds several existing convolutional neural network-based techniques reported in contemporary studies. To enhance clinical transparency, Gradient weighted Class Activation Mapping (Grad-CAM) was utilized to visualize salient regions contributing to the model’s predictions, thereby improving interpretability and supporting potential clinical adoption. The results demonstrate that the framework is effective, computationally efficient, and capable of providing reliable diagnostic support. Its design makes it suitable for integration into real-time clinical decision support systems and telemedicine platforms, particularly in low-resource healthcare environments where rapid and accurate diagnostic tools are urgently needed.