Automated quantification of choriocapillaris anatomical features in ultrahigh-speed optical coherence tomography angiograms

In vivo visualization and quantification of choriocapillaris vascular anatomy is a fundamental step in understanding the relation between choriocapillaris degradation and atrophic retinopathies, including geographic atrophy. We describe a process utilizing ultrahigh-speed swept-source optical coherence tomography and a custom-designed “local min-max normalized masking” algorithm to extract in vivo anatomical metrics of the choriocapillaris. We used a swept-source optical coherence tomography system with a 1.6 MHz A-scan rate to image healthy retinas. With the postprocessing algorithm, we reduced noise, optimized visibility of vasculature, and skeletonized the vasculature within the images. These skeletonizations were in 89% agreement with those made by skilled technicians and were, on average, completed in 18.6 s as compared to the 5.6 h technicians required. Anatomy within the processed images and skeletonizations was analyzed to identify average values (mean±SD) of flow void radius (9.8±0.7 µm), flow void area (749±110 µm2), vessel radius (5.0±0.3 µm), branch-point to branch-point vessel length (26.8±1.1 µm), and branches per branch-point (3.1±0.1). To exemplify the uses of this tool a retina with geographic atrophy was imaged and processed to reveal statistically significant (p<0.05) increases in flow void radii and decreases in vessel radii under atrophic lesions as compared to atrophy-free regions on the same retina. Our results demonstrate a new avenue for quantifying choriocapillaris anatomy and studying vasculature changes in atrophic retinopathies.