Characterization of Retinal Arteries by Adaptive Optics Ophthalmoscopy Image Analysis

Area of Science:

• Ophthalmology
• Medical Imaging
• Biomedical Engineering

Background:

• Retinal artery analysis is crucial for diagnosing vascular diseases.
• Accurate segmentation of retinal vessels is challenging but essential for biomarker quantification.

Purpose of the Study:

• To develop and validate a novel method for quantifying biomarkers from retinal artery segmentation in adaptive optics ophthalmoscopy images (AOO).
• To assess the utility of these biomarkers in a preliminary clinical study of small vessel disease.

Main Methods:

• A hybrid approach combining deep learning and knowledge-driven deformable models for precise vessel wall segmentation, with a focus on bifurcations.
• Derivation of biomarkers including junction coefficient, branching coefficient, and wall-to-lumen ratio (WLR).
• Validation using mean squared error (MSE) and assessment of reproducibility and bias.

Main Results:

• Achieved reliable and accurate segmentations with low MSE (1.75 ±1.24 pixels).
• Demonstrated high reproducibility across image acquisition and users, with no significant bias.
• Identified an increased WLR in patients with genetic small vessel disease compared to controls.

Conclusions:

• The Adaptive Optics Vessel analysis (AOV) method provides robust and accurate WLR estimation from AOO images.
• WLR shows promise as a sensitive biomarker for detecting small vessel diseases, particularly when vessel walls are well-contrasted.