This is an application of the Vision Transformer for brain tumor detection through the use of hyperspectral images with a limited number of bands (i.e. using the SLIM Brain Database).
💡 An introduction to the ViT can be found here: An Image is Worth 16x16 Words: Transformers for Image Recognition at Scale
💡 Parts of this work were inspired from: SpectralFormer: Rethinking Hyperspectral Image Classification with Transformers
Table of Contents
- Pytorch 2.3.1 (+ all the packages in the requirements.txt file)
- Cuda 12.4 for parallelization
- data_analysis/ - Contains some scripts and notebooks used to study the composition of the SLIM Brain Database.
- experiments/ - There are the code used to evaluate three experiments concerning the intra- and inter- partients classification. Moreover, there are the models trained as result from each experiment (pth format).
- vit/ - Contains the ViT model used and the script used to run the hyperparameter optimization.
🎯 All the results obtained from the tests are explained in the following paper: Vision Transformer for Brain Tumor Detection Using Hyperspectral Images with Reduced Spectral Bands.
@article{Ragusa2025,
author={Ragusa, Domenico and Gazzoni, Marco and Torti, Emanuele and Marenzi, Elisa and Leporati, Francesco},
journal={IEEE Access},
title={Vision Transformer for Brain Tumor Detection Using Hyperspectral Images With Reduced Spectral Bands},
year={2025},
volume={13},
pages={121704-121719},
doi={10.1109/ACCESS.2025.3588001}}