Histologically guided metrics for semi-automated analysis of fundus autofluorescence (FAF) in aging and age-related macular degeneration (AMD)

Abstract

Purpose : The retinal pigment epithelium (RPE) is the major signal source for blue FAF. New histology of human RPE in AMD revealed variable cell morphologies, layer thickening, and re-distribution and loss of AF granules [1-4]. Guided by these findings, we tested whether automated analysis of FAF intensity variation could elucidate RPE health in aging and early AMD.

Methods : The Alabama Study of Early Age-Related Macular Degeneration prospectively tested functional biomarkers of incident AMD at 3 years [5]. Participants (n=544; 410 normal, 134 AMD, via AREDS 9-step scale) underwent multimodal imaging with FAF using a Spectralis without an internal AF reference [6]. In baseline and follow-up images, the vasculature and optic nerve head were masked semi-automatically to reduce variability from non-AF structures. In masked images, we computed standard deviation/ mean intensity (coefficient of variation, CV), previously used for Stargardt disease [7], within the ETDRS grid and sectors (central subfield, C1; inner ring, R3; outer ring; R6). Associations with age and AMD severity were examined by Spearman correlation coefficients (SCC). Changes in FAF metrics between baseline and follow-up were checked with general linear models.

Results : Table 1 shows that at baseline, CV was positively and significantly associated with age in C1, R6, and the whole grid. There was a nominal association of CV with age in R3. CV was positively and significantly associated with AREDS grade in C1, R3, and the whole grid. Table 2 shows that at 3 years, CV increased significantly with time for the entire cohort, for the ETDRS grid and all sectors.

Conclusions : In initial analyses of one histologically informed metric, we find that CV of FAF intensity in ETDRS sectors exhibits biologically plausible features suggestive of strong internal validity. These features include positive associations at baseline with age and AREDS grade, in a regionally specific manner (i.e., central 3 mm where Bruch’s membrane lipids are prominent), and an overall increase, although small, over time. Other metrics are also under consideration.
1. Rudolf. Ophthalmology. 2013;120:821.
2. Ach. IOVS. 2014;55:4832.
3. Ach. IOVS. 2015;56:3242.
4. Zanzottera. IOVS. 2015;56:3253.
5. Owsley. Ophthalmology. 2015;2015.
6. Delori. IOVS. 2011;52:9379.
7. Cideciyan. Hum Mol Genet. 2004;13:525.

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