Sparse transformer and multipath decision tree: a novel approach for efficient brain tumor classification.

Area of Science:

• Medical Imaging
• Artificial Intelligence
• Oncology

Background:

• Early brain tumor classification is crucial for effective treatment.
• Automated classification algorithms struggle with tumor diversity and high-resolution medical images.
• Existing models like Swin Transformer face challenges with small datasets and computational demands.

Purpose of the Study:

• To develop a novel model for accurate and efficient brain tumor classification.
• To address the limitations of existing methods in handling diverse tumor types and computational complexity.
• To improve the suitability of automated classification for resource-constrained medical settings.

Main Methods:

• Introduction of SparseSwinMDT, a hybrid model combining sparse token representation and multipath decision trees.
• Leveraging Swin Transformer architecture for high-resolution image analysis.
• Integration of sparse representation to mitigate challenges with small datasets and computational load.

Main Results:

• SparseSwinMDT achieved a classification accuracy of 99.47%.
• The model significantly outperformed existing brain tumor classification methods.
• Demonstrated a reduction in computation time compared to conventional approaches.

Conclusions:

• SparseSwinMDT offers a highly accurate and computationally efficient solution for brain tumor classification.
• The model's performance makes it ideal for deployment in resource-limited medical environments.
• This advancement holds significant potential for improving diagnostic workflows in neuro-oncology.