A Novel Automated Method for the Objective Quantification of Retinal Layers Based on Spectral Domain Optical Coherence Tomography (SD-OCT) Imaging Reveals Sequential Changes in the Normal Retina with Age

Abstract

Purpose : The main limitation of SD-OCT imaging is that data obtained is qualitative and hence the diagnosis of retinal pathology is susceptible to interpreter bias. Quantitative, unbiased interpretation of the data is needed for objective definition of retinal layers. The purpose of this study was to develop a novel automated algorithm that objectively quantifies reflectivity of retinal layers from OCT images and to apply this algorithm to the investigation of changes that occur in the normal retina with age.

Methods : SD-OCT scans (Zeiss Cirrus HD-OCT 5000) were prospectively collected from 215 normal subjects (age 10–79) without clinical evidence of retinal pathology. Subjects with diabetes mellitus, myopia ≤ −6.00 diopters, and tilted OCT were excluded, leaving 165 (106 female; 59 male). A novel segmentation algorithm was first applied to extract twelve retinal layers using shape, intensity, and spatial information. A novel normalized reflectivity scale (NRS) ranging from 0 units (vitreous) to 1000 units (retinal pigment epithelium [RPE]) was then applied to the raw data. Statistical analysis employed a full factorial ANCOVA design with gender, foveal side (nasal/temporal), retinal layer, and continuous covariate age as factors.

Results : Normalized reflectivity varied significantly with age (F_1,3864=482.3; p<0.0001) and layer (F_11,3864=2350.5; p<0.0001) but not gender (F_1,3864=0.226; p=0.635) or foveal side (F_1,3864=0.159; p=0.690). The slope (change in NRS/yr) varied significantly by gender (F_1,3864=25.08; p<0.0001) and retinal layer (F_11,3864=18.67; p<0.0001). The interactions of foveal side with layer (F_11,3864=27.07; p<0.0001) and gender (F_1,3864=5.58; p=0.018) were significant. No other terms in the model were significant. NRS decreased with age overall, with the greatest declines seen in the ellipsoid zone (−3.91 NRS/yr) and NFL (−3.84 NRS/yr) (Figure 1). Normal aging had little to no effect on the RPE (−0.72 NRS/yr) or ELM (−0.95 NRS/yr).

Conclusions : A novel automated method enables quantitative analysis of the reflectivity changes in each retinal layer that occur with normal aging. Automatic quantification of SD-OCT data carries the promise to detect occult retinal pathology.

This is an abstract that was submitted for the 2016 ARVO Annual Meeting, held in Seattle, Wash., May 1-5, 2016.

 

Figure 1. Slopes showing the decline of reflectivity for the twelve layers with age (95% confidence intervals).