Brain Tumor Segmentation using Vision Transformer (ViT)

EasyChair Preprint 15897

Abstract

This paper presents a novel approach to brain tumor segmentation using Vision Transformer (ViT) architecture. We propose a ViT-based model that leverages the power of self- attention mechanisms to accurately segment brain tumors from MRI images. Our method combines the strengths of transformer models with traditional convolutional neural networks to create a hybrid architecture optimized for medical image segmentation tasks. The model achieves 94.41% accuracy and a Dice coefficient of 0.3820 on the test set, demonstrating its effectiveness for tumor segmentation.

Keyphrases: Brain Tumor Segmentation, Medical Imaging, Vision Transformer, deep learning, neural networks

@booklet{EasyChair:15897,
  author    = {Aman Malik},
  title     = {Brain Tumor Segmentation using Vision Transformer (ViT)},
  howpublished = {EasyChair Preprint 15897},
  year      = {EasyChair, 2025}}