Retinal artery-vein caliber grading using color fundus imaging

Area of Science:

• Ophthalmology
• Medical Imaging
• Cardiovascular Disease Research

Background:

• Retinal vessel caliber is a key predictor of cardiovascular diseases (CVDs).
• Current methods for measuring Central Retinal Artery Equivalent (CRAE) and Central Retinal Vein Equivalent (CRVE) are manual or semi-automatic, leading to time inefficiency and potential inaccuracies.
• Automated, efficient, and reproducible grading of retinal vessel caliber is crucial but technically challenging.

Purpose of the Study:

• To introduce a novel "edge-based" vessel tracking technique for automated retinal vessel caliber measurement.
• To accurately quantify CRAE and CRVE using the proposed method.
• To evaluate the accuracy and reproducibility of the new technique compared to manual and semi-automatic methods.

Main Methods:

• An "edge-based" vessel tracking algorithm was developed to measure retinal arteriolar and venular caliber.
• The method calculated Central Retinal Artery Equivalent (CRAE) and Central Retinal Vein Equivalent (CRVE).
• Performance was assessed by comparing measurements against manual grading for accuracy and against a semi-automatic method for overall accuracy and reproducibility.

Main Results:

• The proposed technique achieved an average cross-sectional width measurement accuracy of 96.23% compared to manual grading.
• High correlations were observed for overall vessel caliber measurement accuracy: 0.85 for CRAE and 0.92 for CRVE when compared to a semi-automatic method.
• Excellent intra-grader reproducibility was demonstrated, with correlation coefficients of 0.881 for CRAE and 0.875 for CRVE.

Conclusions:

• The developed "edge-based" vessel tracking method provides an accurate and highly reproducible automated approach for measuring retinal vessel caliber (CRAE and CRVE).
• This technique addresses the limitations of current manual and semi-automatic methods, offering a time-efficient solution for CVD risk assessment.
• The high accuracy and reproducibility suggest significant potential for clinical application in cardiovascular health monitoring.