A Hybrid Deep Learning Model for Accurate Chest X-Ray Disease Classification

To begin with, the classification of chest X-rays is vital for automatically diagnosing respiratory diseases such as COVID-19, pneumonia, and other abnormalities. In our system, we operated custom convolutional neural networks (CNNs) and Vision Transformers (ViTs), including Tiny-ViT, SwinTransformer, FocalNet, and VOLO, to compare their effectiveness and for the classification task under limited data constraints in the Covid19, Pneumonia, and Normal Chest X-Ray Images dataset. Moreover, advanced data augmentation and regularization techniques have been used to improve the strength and inference of our system. Consequently, we attained an accuracy of <tex xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">$\mathbf{9 8. 2 9 \%}$</tex> on the custom CNN and <tex xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">$\mathbf{9 8 \%}$</tex> on the Vision transformer model VOLO, outperforming all other models. In addition, our selection of custom CNN and ViT models (VOLO-D1) was based on their intense feature extraction abilities and usefulness for transfer learning. Subsequently, we used FocalNet to replace selfattention (SA) with a focal modulation mechanism in our system's vision. Furthermore, we reached 98 %, surpassing the state of art images.