Abstract
Purpose :
The conjunctival vasculature is known to play an important role in ocular surface diseases. Recently, anterior segment-optical coherence tomography angiography (AS-OCTA) has proven to be useful in assessing conjunctival vessels. Given the novelty of this technology, artifacts can interfere with automated analysis. We aim to establish a semi-automated methodology for quantification of conjunctival vessel density (VD) to overcome them.
Methods :
Using the Avanti XR (Optovue, Fremont, CA) system, AS-OCTA images of the nasal and temporal conjunctiva of each eye were acquired from patients with various ocular surface diseases and normal controls. Poor signal strength index images (<20) were excluded and the eyelids and tear film menisci were removed from images. In one group, raw images were processed with four different automated local thresholds (ALT) prior to VD assessment using Vessel Analysis plugin (Fiji). In another group, the same raw images were manually segmented to remove before processing through the same process as previously described. The VDs of each ALT were compared between the two groups using interclass correlation coefficient (ICC) to establish the agreement of the automated methods with the manual segmentation using SPSS (SPSS Inc, Chicago, IL).
Results :
A total of 32 images of 18 eyes (5 controls, 3 ocular allergy, 4 dry eye, and 6 contact lens wearers) were analyzed. The mean VD with mean, median, phansalkar and sauvola ALT were 42.8%, 49.5%, 41.1% and 60.8%, respectively in the unsegmented group and 41.2%, 49.3%, 39.0% and 47.3%, respectively in the segmented group. The ICC was 0.59 (p<0.001), 0.69 (p<0.001), 0.57 (p<0.001) and 0.15 (p=0.05) in the mean, median, phansalkar and sauvola ALT groups, respectively.
Conclusions :
Semi-automated analysis using median ALT showed a good agreement with manual segmentation and can be used to measure conjunctival VD on OCTA. This method can be useful for future prospective studies as it is reliable and faster than manual segmentation.
This is an abstract that was submitted for the 2018 ARVO Annual Meeting, held in Honolulu, Hawaii, April 29 - May 3, 2018.