Intergrader reproducibility of peripapillary retinal metrics from optical coherence tomographic angiography in community dwelling older adults

Abstract

Purpose : Optical coherence tomographic angiography (OCTA) was used to measure peripapillary retinal structure and perfusion in community dwelling older adults in the US based Eye Determinant of Cognition (EyeDOC) Study. Preliminary data with 164 scans showed that manual segmentation (MS) was needed for ~18% scans. We report intergrader reproducibility of OCTA parameters when MS was needed.

Methods : The Avanti/AngioVue OCT/OCTA system (Optovue, Inc., Fremont, CA, USA) was used to capture 4.5 mm x4.5mm optic disc scans. Scans were sent to the Casey Eye Institute for quality control and grading. Images with adequate quality were analyzed using a custom software. The analysis started with automatic segmentation of the disc boundaries and retinal layers. A grader then visually inspected the segmentation results and manually corrected errors if any were noted. The software then automatically calculates quantitative parameters for the overall 4-mm square analytic region as well as for the superior and inferior hemispheres. The nerve fiber layer (NFL) thicknesses was the structural parameter analyzed. The perfusion parameters are the capillary densities in the NFL plexus, and the superficial and deep vascular complexes. To assess measurement reproducibility, scans requiring MS were regraded by a different grader independently. Bland-Altman (B-A) plots, mean between-grading differences, relative absolute differences (RAD), and intra-class correlations (ICCs) were used to assess reproducibility. Scans with relatively large differences were qualitatively reviewed.

Results : Scans of 28 eyes (25 participants) required MS and contributed to intergrader assessment . The participants were 68% female, 56% black, with median age of 78 years. The median distance presenting acuity was 0.13 LogMAR (range -0.06 to 0.44). There was no significant mean bias between the two graders (Tab 1) and the median relative absolute differences were all <5%. The intraclass correlations ranged from 0.93 to 0.98. The BlandAltman plots showed no heteroscedasticity (Fig 1). Review of the few cases with relatively large difference found that relatively low scan quality was the main reason of intergrader discrepancy.

Conclusions : There was good intergrader reproducibility of the OCTA peripapillary retinal parameters when manual segmentation correction was needed.

This abstract was presented at the 2019 ARVO Annual Meeting, held in Vancouver, Canada, April 28 - May 2, 2019.