Dual vision Transformer-DSUNET with feature fusion for brain tumor segmentation

Area of Science:

• Medical Imaging
• Artificial Intelligence
• Oncology

Background:

• Brain tumors are a leading cause of cancer death, necessitating early diagnosis and effective treatment.
• Magnetic Resonance Imaging (MRI) is crucial for brain tumor diagnosis, but automated segmentation accuracy remains a challenge.
• Automating brain tumor segmentation is critical for diagnosis and therapy planning.

Purpose of the Study:

• To develop an automated model for precise brain tumor segmentation using multi-modal MRI data.
• To improve the accuracy and efficiency of differentiating brain tumors from healthy brain tissue.
• To leverage feature fusion techniques for enhanced segmentation performance.

Main Methods:

• The study proposes the Dual Vision Transformer-DSUNET model, incorporating dual vision and feature fusion.
• Multi-modal MRI data (T1, T2, T1Gd, FLAIR) from the BRATS 2020 dataset was utilized.
• The model captures heterogeneous tumor properties across imaging modalities.

Main Results:

• The Dual Vision Transformer-DSUNET model achieved high segmentation accuracy.
• Dice Coefficient scores were 91.47% (enhanced), 92.38% (core), and 90.88% (edema), with a cumulative score of 91.29%.
• The model demonstrated an overall accuracy of 99.93%, indicating robust performance.

Conclusions:

• The proposed Dual Vision Transformer-DSUNET architecture shows significant promise for accurate brain tumor segmentation.
• The model's high performance can enhance the early detection and management of brain diseases.
• This approach offers a reliable tool for improving diagnostic accuracy in neuro-oncology.