CC-Net: A cross-hierarchical context-aware network for medical image segmentation.

Area of Science:

• Medical Image Analysis
• Computer Vision
• Artificial Intelligence

Background:

• Accurate medical image segmentation is vital for disease diagnosis and treatment.
• Existing U-shaped networks struggle with integrating shallow structural and deep semantic features, hindering precise boundary delineation.

Purpose of the Study:

• To propose a novel Cross-Hierarchical Context-Aware Network (CC-Net) for enhanced medical image segmentation.
• To improve the integration of multi-level features for more accurate delineation of complex lesion boundaries and fine-grained structures.

Main Methods:

• CC-Net features Cross-Hierarchical Feature Aggregation (CFA) to model relationships across feature levels.
• Global Feature Aggregation (GFA) integrates multi-level features for unified global representation.
• Cross-Branch Semantic Supplement (CSS) and Enhanced Feature (EF) modules refine feature discrimination and noise suppression.

Main Results:

• CC-Net consistently outperformed state-of-the-art methods on six public medical image datasets.
• The proposed modules effectively integrated shallow and deep features, improving segmentation accuracy.
• Experiments validated CC-Net's superior performance across all evaluation metrics.

Conclusions:

• CC-Net offers a significant advancement in medical image segmentation by effectively leveraging cross-hierarchical contextual information.
• The network architecture addresses limitations of existing methods, providing more accurate and robust lesion segmentation.
• The proposed approach holds promise for improving early disease diagnosis and treatment planning.