Enhanced Feature Fused Vision Transformer Framework for Multiclass Brain Tumor Classification Using MRI Images

Precise classification of brain tumors is essential for successful diagnosis and treatment planning in medical imaging. In this research, a new method using a Vision Transformer (ViT) with a feature infusion scheme to enhance classification accuracy is proposed. The new approach effectively extracts both local and global features to classify glioma, meningioma, pituitary tumors, and non-tumor instances. By combining ViT with feature fusion improves classification performance, overcoming the weak points of conventional methods. Such promising results verify the clinical application potential, providing more accurate and trustworthy diagnoses. Experimental evidence proves the performance excellence of the method, obtaining a remarkable accuracy of 99.5%, outperforming current techniques. Future work will proceed to optimize hyperparameters and increase the dataset to improve robustness and generalizability even further.