Fuzzy Attention Integrated Lightweight Vision Transformer Image Processing for Brain Tumor MRI Classification

Brain tumors pose a significant threat to life, making early and accurate classification crucial. Many computer-aided tumor detection systems utilise Deep Learning methods like Convolutional Neural Networks(CNNs), and more recently Transfer Learning, which have shown great progress. Recently, Vision Transformers(ViTs) have emerged as a compelling alternative for image classification tasks, but remain under-explored in medical imaging due to computational demands and high data requirements; addressed through recent advancements in Lightweight ViT architectures. This study proposes a novel framework that leverages Lightweight ViTs, specifically MobileViT and DeiTTiny, integrated with a Fuzzy Attention Mechanism to categorise brain tumors into glioma, pituitary, and meningioma. The models achieve impressive accuracies of $\mathbf{9 8. 2 3 \%}$ and $\mathbf{9 5. 6 5 \%}$ respectively, while maintaining low computational overhead. Comparative analysis with traditional deep learning and transfer learning techniques further validates the model's effectiveness. The experimental results indicate that the proposed method outperforms state-of-the-art classification methods on the utilised dataset.