A TRANSFER LEARNING–BASED VGG-16 FRAMEWORK FOR AUTOMATED PNEUMONIA DETECTION FROM CHEST X-RAY IMAGES WITH REAL-TIME WEB DEPLOYMENT

Abstract

As part of our focus on enhancing respiratory diagnostic processes, we examine the use of advanced machine learning techniques to help automate pneumonia detection from chest X-rays. Our goal is to help radiologists manage their workloads by giving them screening tools to aid in the diagnosis of patients in resource-constrained, high patient volume environments. Using Python in Jupyter notebooks, we created a pneumonia classification model utilizing transfer learning combined with techniques of medical image analysis. The model was built using the Kaggle Chest X-ray Pneumonia dataset. This dataset included images of pediatric chest X-rays which were divided into two classes: Normal and Pneumonia. Additionally, we improved the model's robustness with data preprocessing and augmentation techniques. We used a VGG-16 model as a feature extractor, then modified the model to suit the needs of our classification task. We added a binary classification layer to the top of the model. The results of our experiments were very successful with a validation accuracy of 99.04% and a test accuracy of 98.28% along with strong precision, recall, and F1 scores. There was very little misclassification. Additionally, to improve accessibility, we created a simple web-based interface. This allows users to upload X-ray images of their chest and receive predictions along with a confidence score in real time. This research proves that a transfer learning framework based on VGG-16 is not only accurate, but also demonstrates the potential of AI assistance in pneumonia screening.