Boundary-enhanced local-global collaborative network for medical image segmentation.

Area of Science:

• Medical Imaging
• Computer Vision
• Artificial Intelligence

Background:

• Medical image segmentation is crucial for diagnosis and treatment but faces challenges like class imbalance and indistinct boundaries.
• Precisely segmenting small regions of interest (ROIs) remains a significant hurdle in medical image analysis.
• Existing methods struggle with the complexities of accurately delineating small or poorly defined ROIs.

Purpose of the Study:

• To propose a novel network, BELGNet, for enhanced medical image segmentation.
• To improve the precise segmentation of small ROIs in class-imbalanced medical imaging datasets.
• To develop a method that effectively integrates local and global features while emphasizing boundary information.

Main Methods:

• A local-global collaborative encoder using attention fusion to integrate CNN-based local features and Mamba-based global features.
• A boundary information-enhanced decoder incorporating boundary attention modules to refine segmentation details.
• Implementation of BELGNet leveraging both local and global feature extraction with specific attention mechanisms.

Main Results:

• BELGNet demonstrated superior performance on various public class-imbalanced medical image segmentation datasets.
• The proposed network effectively addressed the challenges of segmenting small ROIs and indistinct boundaries.
• Experimental results show BELGNet outperforming existing state-of-the-art methods in segmentation accuracy.

Conclusions:

• BELGNet offers a robust solution for medical image segmentation, particularly in challenging class-imbalanced scenarios.
• The integration of local-global features and boundary enhancement significantly improves segmentation precision.
• The proposed approach advances the capabilities of automated medical image analysis for clinical applications.