Page 84 - AIH-2-1

Artificial Intelligence in Health ViT for Glioma Classification in MRI

6. Conclusion the implementation, please contact Mr. Saneru
Wickramasinghe (saneruw@gmail.com).
Herein, the ViT architecture was studied for MRI image
classification, focusing on glioma. To address the issues of References
data scarcity and class imbalance, ViT was pretrained using
the CIFAR-10 dataset and fine-tuned using the BraTS 1. Pal A, Chaturvedi A, Garain U, Chandra A, Chatterjee R.
2015 dataset. The fine-tuned ViT could accurately and Severity Grading of Psoriatic Plaques using Deep CNN Based
effectively identify glioma compared with the popular CNN Multi-task Learning. Mexico: ICPR; 2016.
architecture. Moreover, the effects of the patch resolution doi: 10.1109/ICPR.2016.7899846
on the overall performance accuracy and the behavior of 2. Wang G. A perspective on deep imaging. IEEE Access.
the loss curve were discussed. Overall, this study proposed 2016;4:8914-8924.
a feasible and resource-optimized solution for the early doi: 10.1109/ACCESS.2016.2624938
detection and better prognosis of brain tumors. Further
research is required to improve the predictions of the model 3. Ker J, Wang L, Rao J, Lim T. Deep learning applications in
while making the results understandable with explainable medical image analysis. IEEE Access. 2018;6:9375-9389.
AI techniques for the advancement of automated systems doi: 10.1109/ACCESS.2017.2788044
for brain tumor detection and diagnosis.
4. Kabir Anaraki A, Ayati M, Kazemi F. Magnetic resonance
Acknowledgments imaging-based brain tumor grades classification and grading
via convolutional neural networks and genetic algorithms.
None. Biocybern. Biomed. Eng. 2019;39(1):63-74.
Funding doi: 10.1016/j.bbe.2018.10.004
5. Kaldera HNTK, Gunasekara SR, Dissanayake MB. Brain
None.
Tumor Classification and Segmentation Using Faster
Conflict of interest R-CNN. In: Proceedings ASET. United States: IEEE; 2019.
doi: 10.1109/ICASET.2019.8714263
The authors declare that they have no competing interests.
6. Vaswani A, Shazeer N, Parmar N, et al. Attention is all you
Author contributions need. In: Advances in Neural Information Processing Systems.
United States: The MIT Press; 2017. p. 5998-6008.
Conceptualization: Maheshi B. Dissanayake
Formal analysis: All authors doi: 10.48550/arXiv.1706.03762
Investigation: All authors 7. Menze BH, Jakab A, Bauer S, et al. The multimodal brain
Methodology: All authors tumor image segmentation benchmark (BRATS). IEEE
Writing – original draft: All authors Trans Med Imaging. 2014;34(10):1993-2024.
Writing – review & editing: Maheshi B. Dissanayake doi: 10.1109/TMI.2014.2377694

Ethics approval and consent to participate 8. Alsaif H, Guesmi R, Alshammari BM, et al. A novel
data augmentation-based brain tumor detection using
The data collection was not part of this research. We use convolutional neural network. Appl Sci. 2022;12(8):3773.
publicly available BRATS dataset. Ethical clearance had
already been obtained before the upload of the medical doi: 10.3390/app12083773
dataset BRATS onto the public domain by Menze BH 9. Pan X, Ge C, Lu R, et al. On the Integration of Self-Attention
et. al. 7 and Convolution. United States: IEEE/CVF; 2022. p. 815-825.

Consent for publication doi: 10.1109/CVPR52688.2022.0008
10. Devlin J, Chang MW, Lee K, Toutanova K. Pre-training
Not applicable.
of Deep Bidirectional Transformers for Language
Availability of data Understanding; 2018.
doi: 10.48550/arXiv.1810.04805
The data utilized in this research are publicly available.
The authors have released the code on the GitHub page 11. Dosovitskiy A, Beyer L, Kolesnikov A, et al. An image is
(https://github.com/SaneruW/Vision-transformers-for- worth 16x16 words: Transformers for image recognition at
glioma-classifications-using-T1-magnetic-resonance- scale; 2020.
images). Regarding materials or details related to doi: 10.48550/arXiv.2010.11929

Volume 2 Issue 1 (2025) 78 doi: 10.36922/aih.4155