LGI Net: Enhancing local-global information interaction for medical image segmentation.

Area of Science:

• Medical Imaging
• Computer Vision
• Artificial Intelligence

Background:

• Medical image segmentation is crucial for identifying regions of interest and pathologies.
• Deep learning has advanced medical image segmentation, but challenges remain in integrating local and global information.
• Existing methods struggle with complex variations and irregular shapes in medical images.

Purpose of the Study:

• To propose LGI Net, a novel architecture for medical image segmentation.
• To enhance the fusion of local features and global contextual information.
• To improve inter-layer information exchange and channel interactions.

Main Methods:

• Developed LGI Net architecture with improved internal computation for local-global information interaction.
• Incorporated an ECA module to capture channel interplay and enhance inter-layer information exchange.
• Validated the method on Kvasir, ISIC, and X-ray datasets.

Main Results:

• LGI Net demonstrated superior accuracy and parameter efficiency compared to existing methods.
• Ablation studies confirmed the effectiveness of the proposed LGAF module.
• Experiments validated the network's capability in fusing local and global contextual information.

Conclusions:

• LGI Net offers an innovative approach to medical image segmentation.
• The architecture effectively addresses limitations in fusing local and global information.
• The study provides valuable insights for enhancing medical image segmentation accuracy and performance.