Enhancing brain tumor detection in MRI with a rotation invariant Vision Transformer

Area of Science:

• Medical Imaging
• Artificial Intelligence
• Deep Learning

Background:

• The study introduces the Rotation Invariant Vision Transformer (RViT), a novel deep learning model.
• RViT is specifically designed for brain tumor classification using Magnetic Resonance Imaging (MRI) scans.

Purpose of the Study:

• To develop and evaluate a deep learning model for enhanced brain tumor classification.
• To improve the accuracy and robustness of tumor identification in MRI scans.

Main Methods:

• The RViT model incorporates rotated patch embeddings.
• This approach enhances the model's ability to handle diverse orientations in medical images.

Main Results:

• RViT achieved high performance metrics on the Brain Tumor MRI Dataset.
• Key results include sensitivity (1.0), specificity (0.975), F1-score (0.984), MCC (0.972), and accuracy (0.986).

Conclusions:

• RViT demonstrates superior performance compared to the standard Vision Transformer and other existing methods.
• The integration of rotational patch embeddings significantly improves brain tumor detection capabilities.
• RViT shows potential for advancing brain tumor detection and other complex medical imaging tasks.