Automated Retinal Vascular Topological Information Extraction From OCTA.

Area of Science:

• Ophthalmology
• Biomedical Engineering
• Medical Imaging

Background:

• The retinal vascular system is crucial for vision and is affected by diseases like glaucoma, diabetic retinopathy, and age-related macular degeneration.
• Accurate analysis of retinal vasculature is essential for diagnosing and monitoring these conditions.

Purpose of the Study:

• To develop and validate a comprehensive method for extracting detailed vascular information from wide-field optical coherence tomography angiography (OCTA) data.
• To improve the accuracy of retinal vessel segmentation and classification, and to enable quantitative morphometric analysis.

Main Methods:

• An integrated U-Net model was developed for simultaneous segmentation and classification of retinal arteries and veins from OCTA images.
• A topological information-based algorithm was implemented to correct misclassifications and improve vessel connectivity.
• Vessel morphometry across different branch orders (arteries/veins, arterioles/venules, capillaries) was extracted for comparative analysis.

Main Results:

• The U-Net model achieved high segmentation and classification IoU scores (0.7095±0.0224 for segmentation, 0.8793±0.1049 and 0.8928±0.0929 for artery/vein classification).
• The topological correction algorithm enhanced the overall IoU by an average of 8.29%, improving vessel network accuracy.
• Quantitative vessel morphometry data for various branch orders were successfully extracted, allowing for detailed comparisons.

Conclusions:

• The presented combination of U-Net segmentation and topological correction offers a robust approach for analyzing retinal vasculature in wide-field OCTA.
• This method enhances the accuracy of identifying and characterizing arterial and venous networks, crucial for understanding ocular diseases.
• The extracted vessel morphometry provides valuable quantitative insights for research and clinical applications in ophthalmology.