A Review of U-Net Based Deep Learning Frameworks for MRI-Based Brain Tumor Segmentation.

Area of Science:

• Medical Imaging
• Artificial Intelligence
• Computational Biology

Background:

• Automated brain tumor segmentation from MRI is crucial for clinical applications.
• Deep learning, particularly the U-Net architecture, has emerged as a leading approach.
• Existing reviews lack a focused analysis of U-Net variants for brain tumor segmentation.

Purpose of the Study:

• To review and analyze 35 studies (2019-2025) on U-Net-based brain tumor segmentation.
• To examine the evolution of U-Net architectures from 2D to 3D and advanced variants.
• To synthesize results, evaluation criteria, and benchmark datasets in this field.

Main Methods:

• Systematic literature review of 35 studies published between 2019 and 2025.
• Analysis of U-Net architecture modifications and adaptations for brain tumor segmentation.
• Synthesis of reported results, evaluation metrics, and datasets (e.g., BRATS).

Main Results:

• U-Net architectures have shown significant effectiveness in brain tumor segmentation tasks.
• Various modifications enhance U-Net performance, addressing limitations of original 2D and 3D models.
• Standard evaluation criteria and benchmark datasets are crucial for model validation.

Conclusions:

• U-Net remains a dominant and evolving architecture for brain tumor segmentation.
• Future research should focus on model efficiency, generalization, multimodal data integration, and clinical translation.
• This review provides a comprehensive guide for researchers in the field.