Advancing Brain Tumor Classification through Fine-Tuned Vision Transformers: A Comparative Study of Pre-Trained...

Area of Science:

• Medical Image Analysis
• Artificial Intelligence
• Computer Vision

Background:

• Brain tumor classification is critical for diagnosis and treatment.
• Traditional methods face challenges in accuracy and efficiency.
• Advanced deep learning models offer potential for improved performance.

Purpose of the Study:

• To evaluate the efficacy of pre-trained Vision Transformer (ViT) models for brain tumor image classification.
• To compare the performance of five ViT models (R50-ViT-l16, ViT-l16, ViT-l32, ViT-b16, ViT-b32) using a fine-tuning approach.
• To establish a new benchmark for brain tumor classification accuracy.

Main Methods:

• Utilized a dataset of 4855 training and 857 testing images across four tumor classes.
• Employed fine-tuning on five pre-trained Vision Transformer (ViT) models.
• Evaluated model performance using precision, recall, F1-score, accuracy, and confusion matrix metrics.

Main Results:

• The ViT-b32 model achieved the highest accuracy at 98.24%.
• All evaluated ViT models demonstrated strong performance in brain tumor classification.
• The fine-tuned ViT models surpassed existing methodologies in classification accuracy.

Conclusions:

• Vision Transformer (ViT) models show significant potential for accurate brain tumor image classification.
• The ViT-b32 model offers a highly effective solution for this task.
• This research provides a strong foundation for future advancements in AI-driven medical image analysis.