Pyramid-Net: Intra-layer Pyramid-Scale Feature Aggregation Network for Retinal Vessel Segmentation.

Area of Science:

• Ophthalmology
• Medical Imaging
• Computer Vision

Background:

• Retinal vessel segmentation is crucial for diagnosing eye diseases and discovering biomarkers.
• Current methods use inter-layer feature aggregation, which has limitations in segmenting thin vessels.
• Limited fusion of multi-scale features in existing approaches can hinder segmentation performance.

Purpose of the Study:

• To propose Pyramid-Net, a novel deep learning model for accurate retinal vessel segmentation.
• To introduce intra-layer feature aggregation to improve the segmentation of fine retinal structures.
• To enhance segmentation performance by addressing limitations of inter-layer feature aggregation.

Main Methods:

• Developed Pyramid-Net featuring intra-layer pyramid-scale aggregation blocks (IPABs).
• IPABs enable multi-scale feature fusion within each layer, combining higher, lower, and current scale branches.
• Incorporated pyramid inputs enhancement, deep pyramid supervision, and pyramid skip connections.

Main Results:

• Pyramid-Net demonstrated improved segmentation performance, particularly for thin retinal vessels.
• The model achieved state-of-the-art results on three public datasets: DRIVE, STARE, and CHASE-DB1.
• Pyramid-Net offers a significant reduction in computational cost compared to existing methods.

Conclusions:

• Pyramid-Net effectively enhances retinal vessel segmentation accuracy, especially for challenging thin vessels.
• The proposed intra-layer aggregation strategy surpasses traditional inter-layer methods.
• Pyramid-Net presents an efficient and high-performing solution for retinal image analysis.