An automatic graph-based approach for artery/vein classification in retinal images.

Area of Science:

• Ophthalmology
• Medical Imaging
• Computer Vision

Background:

• Accurate classification of retinal vessels into arteries and veins (A/V) is crucial for diagnosing vascular changes.
• These changes are indicators of systemic diseases like diabetes and hypertension.
• Automating A/V classification aids in early detection and management of these conditions.

Purpose of the Study:

• To present an automatic approach for artery/vein (A/V) classification in retinal vasculature.
• To develop a method based on graph analysis of the retinal vascular tree.
• To improve the accuracy and efficiency of A/V classification for clinical applications.

Main Methods:

• Extraction of a graph representation from the retinal vasculature.
• Classification of graph nodes (intersection points) and graph links (vessel segments).
• Combination of graph-based labeling with intensity features for final A/V classification.

Main Results:

• The proposed method achieved high accuracy on public datasets: 88.3% (INSPIRE-AVR), 87.4% (DRIVE), and 89.8% (VICAVR).
• The automated A/V classification demonstrated superior performance compared to recent approaches.
• The method effectively classifies the entire vascular tree, including vessel segments and intersection points.

Conclusions:

• The developed graph-based method provides an accurate and efficient solution for automated retinal A/V classification.
• This technique has the potential to enhance the diagnosis of systemic diseases through retinal image analysis.
• The superior performance suggests its utility in clinical settings for automated vascular assessment.