Purchase this article with an account.
Rigmor C Baraas, Hilde Rogeberg Pedersen, Stuart J. Gilson, Kenneth Knoblauch; Cone-to-RPE ratio profiles can predict foveal shape.. Invest. Ophthalmol. Vis. Sci. 2019;60(9):1775. doi: https://doi.org/.
Download citation file:
© ARVO (1962-2015); The Authors (2016-present)
The retinal pigment epithelium (RPE) is essential for appropriate development of the human retina, through promoting photoreceptor development and differentiation. RPE cell maturation is thought to play a critical role in defining foveal specialization and shape. Can this association be observed in foveal shape profiles from OCT imaging? The aim here was to try to model foveal shape from the cone and RPE cell density measurements of healthy adults.
Twenty-two healthy males and females, aged 15–69 yrs were imaged with high-resolution reflectance confocal, dark field and split-detector AOSLO and SD-OCT (Heidelberg Spectralis OCT2). The registered and averaged images were scaled for individual retinal magnification factor. Cone and RPE cells were identified via a semi-automatic algorithm along the nasal-temporal (0–6 deg) meridian to quantify RPE and cone density. Power functions and generalized linear additive models (GAM), which give smooth estimations of the density profiles, were fitted to the cell density data including the cone-to-RPE ratio to parameterize the eccentricity dependence across the central 6 degrees. Horizontal B-scans were segmented semi-automatically and a Difference-of-Gaussians model was fitted to the border between the inner limiting membrane and vitreous to parameterize foveal shape.
For all observers, the GAM approach gave a better description of the cell density profiles in yielding lower AIC values than a power function fit and a better estimate of foveal cone density profiles when counts of peak cone density were available (n=13). A linear mixed-effects model predicting foveal shape by RPE and cone densities indicated RPE to be the more important explanatory variable even if removal of the cone density term made the model significantly worse (χ2(1) = 15, p<0.001). When predicting foveal shape by either the cone-to-RPE ratio or by the two densities entering as independent additive components, the cone-to-RPE ratio profile provided the best predictor, yielding the lowest AIC values.
GAM functions provided a better account of the data than parametric models in generating lower prediction error and more accurate estimation of foveal cone densities. The results show that the inter-relationship between photoreceptors and RPE cells during development is preserved through to adulthood and each individual’s cone-to-RPE density profile mirrors their foveal shape.
This abstract was presented at the 2019 ARVO Annual Meeting, held in Vancouver, Canada, April 28 - May 2, 2019.
This PDF is available to Subscribers Only