Cone packing measurements from confocal and split-detector adaptive optics images in human eyes

Date

2020-08

Journal Title

Journal ISSN

Volume Title

Publisher

Abstract

Purpose: Adaptive optics scanning laser ophthalmoscope (AOSLO) imaging has been used to calculate metrics of cone packing in healthy and diseased eyes. However, there is a lack of data comparing metric values obtained using different AOSLO imaging modalities, as well as the impact of different image analysis methods on these metrics. Here, we 1) calculate the longitudinal repeatability of confocal and split-detector AOSLO imaging, 2) compare cone density measurements made using different marking techniques, and 3) compare cone metrics calculated using different region of interest (ROI) sizes. Methods: AOSLO imaging was performed in 10 healthy individuals from the foveal center to 10° in 4 major meridians at baseline and after 12 months. Cone metrics were quantified from confocal and split-detector images over the same retinal patches and compared. Next, cones extracted from simulated and in vivo images from 5 healthy subjects were marked using different techniques. Cone densities were calculated and compared with known densities for simulated data. Coefficients of variation (CVs) were calculated for in vivo data. Finally, square ROIs of different sizes were extracted from simulated and in vivo cone mosaics. Cone metrics were compared between ROI sizes. Results: 1) The mean CVs of density for confocal and split-detector images were 8.4% and 6.2%, respectively. 2) Unbound densities when marking all cones fully inside and partially along the ROI borders were significantly greater (P<0.05) than bound densities. CVs for bound densities tended to be smaller than for unbound densities. 3) CVs for small ROIs were greater across all eccentricities and increased with increasing eccentricity for in vivo data. In simulated data, ROIs of 25μm x 25μm yielded values that were significantly lower than all larger ROI sizes and were closest to simulated values near the foveal center. Conclusions: The intersession repeatability data in healthy human eyes may be used in future longitudinal studies examining diseased eyes. Computing bound cone density provides measurements with greatest accuracy and least variability. For eccentricities close to the fovea, a 25μm x 25μm ROI size provides measurements with greatest accuracy while larger ROI sizes provide lower variability in the periphery.

Description

Keywords

Adaptive optics, Cone photoreceptors, Confocal imaging, Split detector imaging, Repeatability, Region of interest, Image analysis

Citation