A fine-tuned vision transformer based enhanced multi-class brain tumor classification using MRI scan imagery.

Area of Science:

• Medical Image Analysis
• Artificial Intelligence in Healthcare
• Oncology

Background:

• Brain tumors, abnormal cell growths, require early detection for effective treatment.
• Magnetic Resonance Imaging (MRI) is crucial for brain tumor diagnosis.
• Deep learning models have shown promise in analyzing medical images.

Purpose of the Study:

• To investigate the efficacy of novel Fine-Tuned Vision Transformer models (FTVTs) for brain tumor classification.
• To compare FTVTs against established deep learning models like ResNet50, MobileNet-V2, and EfficientNet-B0.
• To evaluate model performance using accuracy, recall, precision, and F1-score.

Main Methods:

• Utilized a dataset of 7,023 MRI scans categorized into glioma, meningioma, pituitary, and no tumor.
• Implemented and compared four FTVT models (FTVT-b16, FTVT-b32, FTVT-l16, FTVT-l32).
• Benchmarked FTVTs against ResNet50, MobileNet-V2, and EfficientNet-B0.

Main Results:

• FTVT models demonstrated superior performance in brain tumor classification.
• The FTVT-l16 model achieved the highest accuracy of 98.70%.
• Other FTVT models also showed high accuracies (96.87%–98.62%), outperforming established models.

Conclusions:

• Fine-Tuned Vision Transformer models are highly effective and robust for brain tumor classification using MRI data.
• FTVTs represent a significant advancement in AI-driven medical image processing for oncology.
• The study highlights the potential of FTVTs for improving diagnostic accuracy in neuroimaging.