CGNet-assisted Automatic Vessel Segmentation for Optical Coherence Tomography Angiography.

Area of Science:

• Ophthalmology
• Medical Imaging
• Computer Vision

Background:

• Accurate segmentation of retinal optical coherence tomography angiography (OCTA) vasculature is crucial for diagnosing retinal diseases.
• Complex vascular structures and imaging artifacts present significant challenges for automated segmentation methods.

Purpose of the Study:

• To develop and validate a novel end-to-end deep learning framework, CGNet, for enhanced retinal OCTA vessel segmentation.
• To improve the accuracy and efficiency of automated vessel segmentation in OCTA images for clinical applications.

Main Methods:

• A three-stage U-shaped neural network incorporating channel and position attention (CPA) and graph reasoning network (GRN) modules was developed.
• The model utilizes a coarse stage for initial vessel confidence maps, a fine stage for refining micro-vasculatures, and a final refining stage for segmentation fusion.
• The end-to-end training scheme was evaluated on public datasets.

Main Results:

• CGNet achieved high performance, with an area under the ROC curve (AUC) of 94.29% and 85.62% on two distinct datasets.
• The proposed method demonstrated superior performance compared to existing state-of-the-art techniques.
• The integrated CPA and GRN modules contributed to improved operability and reduced complexity.

Conclusions:

• The proposed CGNet framework offers a robust and effective solution for retinal OCTA vessel segmentation.
• This advancement holds significant potential for improving the diagnosis and management of retinal diseases.
• The study provides a valuable tool for researchers and clinicians in the field of retinal imaging.