Wave-Net: A lightweight deep network for retinal vessel segmentation from fundus images

Area of Science:

• Ophthalmology
• Medical Imaging
• Computer Vision

Background:

• Accurate retinal vessel segmentation is crucial for diagnosing eye diseases.
• Thin vessel segmentation is challenging due to image noise, lesions, and poor contrast.
• Deep learning shows promise but struggles with semantic information loss and limited receptive fields.

Purpose of the Study:

• To propose a novel lightweight segmentation network, Wave-Net, for precise retinal vessel segmentation.
• To address challenges in thin vessel segmentation and improve overall accuracy.
• To develop an automated method aiding clinical diagnosis and treatment planning.

Main Methods:

• Introduced Wave-Net, a lightweight deep learning model for retinal vessel segmentation.
• Developed a detail enhancement and denoising (DED) block to improve thin vessel segmentation and mitigate semantic gap.
• Proposed a multi-scale feature fusion (MFF) block to address limited receptive fields and fuse cross-scale contexts.

Main Results:

• Wave-Net achieved excellent performance in retinal vessel segmentation.
• The model demonstrated superior segmentation ability for thin vessels compared to advanced methods.
• Wave-Net maintained a lightweight network design while achieving high accuracy.

Conclusions:

• Wave-Net offers a precise and efficient solution for retinal vessel segmentation.
• The proposed DED and MFF blocks effectively enhance segmentation accuracy, particularly for thin vessels.
• This automated method can assist clinicians in diagnosing eye diseases.