A High-Accuracy Brain Tumor Classification Method Based onVIT in the Context of Small Sample Size

This paper proposes a brain tumor classification method based on VIT (Vision Transformer), which achieves high accuracy in the context of small sample size by classifying and recognizing MRI images. The study utilizes the VIT model to capture the relationship between different regions in the image using self-attention mechanism and enhances the diversity of the dataset through data augmentation techniques. Experimental results demonstrate that this method achieves 100% accuracy in brain tumor image classification tasks, outperforming traditional convolutional neural network models such as VGG16, Alex-Net, ResNet-50, and Inception-V3. This research is of significant importance for automated identification of brain tumor images and holds promise for improving diagnostic efficiency, treatment outcomes, and patient survival rates.