Deep Learning Approaches for Viral Infectious Disease Detection From Chest X-Ray Image Datasets

Viral infectious diseases, such as COVID-19, influenza, HIV, and Ebola, spread through air, liquids, or vectors and cause global health crises. The control of disease outbreaks relies on vaccination, antiviral medications, and surveillance systems. Nevertheless, the emergence of new variants and instances of transmission between species requires constant alertness. In recent years, convolutional neural networks (CNNs) with folding pieces for image classification have performed extremely well in early and accurate disease diagnosis. The study aims to develop a more optimized novel model to enhance the classification accuracy based on the chest X-ray image dataset of the COVID-19 disease. For this, the model was trained using various data extension methods and examined together with the compilation of transfer learning using the VGG19 architecture. The evaluation parameters such as accuracy, F1-score, recall, precision, and confusion matrices, are used to test the performance of the developed model. The outcome shows that the optimized VGG19 model and the combined data augmentation strategy performed better than the individual augmentation models and the baseline. For COVID-19, the improved VGG19 model produced the best results: 96% accuracy, 99% precision, 91% normal, 99% virus, 98% COVID recall, 100% normal, 90% virus, and 99% F1-score for COVID-19, 95% normal, 94% virus. This robust result demonstrated the usefulness of transfer learning in improving model performance for medical image classification. The insights suggest that the strategic use of combined augmentation with transfer learning significantly increases the accuracy and robustness of the CNN model in medical image classification tasks, offering a reliable tool for aiding in diagnosis.