Multi-grade Brain Tumor Classification using VIT and LSTM

Abstract: Medical staff must identify brain tumors at an early stage for proper medical treatments to work. The research presents a deep learning model combining Vision Transformer (ViT) and LSTM for identifying between Glioma, Meningioma, Pituitary Tumor, and No Tumor cases. The spatial characteristics of MRI scans become visible to ViT at the same time that LSTM identifies sequential relationships within the data. The dataset receives both augmentation and normalization procedures before the model achieves high accuracy through training with categorical cross-entropy loss and Adam optimizer. Through its GUI interface the system performs real-time MRI classifying processes while its confusion matrix procedure checks model dependability. The method improves computerized healthcare diagnosis by helping radiologists deliver their work. Keywords: Brain tumors, Vision Transformer, Feature Extraction, Image Augmentation, Normalization