An Integrated PneumoNet Model for the Early Detection of Pneumonic Lungs in Chest X-ray Radiographs using Pre-trained Xception, VGG16, and VGG19

Background: Pneumonia is a respiratory infection that inflames the air sacs in one or both lungs, potentially filling them with fluid. This condition leads to symptoms, such as cough with phlegm, fever, and difficulty in breathing. The accuracy of pneumonia detection via chest X-ray interpretation is heavily reliant on the experience and expertise of radiologists. In low-resource areas, the absence of radiologists and advanced diagnostic facilities hampers the timely and accurate diagnosis of pneumonia. Methods: This work proposes the use of deep learning models for efficient feature extraction and classification in the early prediction of pneumonia from chest X-ray images. This work employs convolutional neural networks (CNN) and four pre-trained models, MobileNetV2, VGG16, VGG19, and Xception. The performance of each model was evaluated, and the top-performing models, namely, Xception, VGG16, and VGG19, were combined to form an integrated model named PneumoNet, which was configured with 20 epochs, a batch size of 32, and a learning rate of 0.01. Results: The integrated PneumoNet model demonstrated improved performance in the early prediction of pneumonia. The model achieved significant enhancements in accuracy, precision, recall, F1- score, and loss values. Specifically, the model reached 100% accuracy, 1.00 precision, 1.00 recall, 1.00 F1-score, and a loss value of 1.32E-21. Conclusion: The implementation of the PneumoNet model has identified the most prominent outcomes for the early prediction of pneumonia in chest X-ray images. The configuration and optimization of the model have resulted in superior performance metrics, addressing the limitations of traditional diagnostic methods, particularly in low-resource environments.