Abstract
Purpose :
RMDA measured at 5° eccentricity assesses photoreceptor sustenance from the circulation. RMDA is slower in eyes with SDD, which start where rods are numerous at 3-6 mm (10.4°-20.8°). SDD’s role in AMD pathogenesis could be clarified if SDD were quantified over a larger extent than accessible with typical OCT volumes. Using supervised deep learning, we quantified SDD in 55°-wide (15.84 mm) posterior pole images to associate with vision tests.
Methods :
This cross-sectional analysis included persons aged ≥60 years from the Alabama Study on Early Age-Related Macular Degeneration 2 (ALSTAR2) baseline (NCT04112667). AMD status was determined using AREDS 9-step. A convolutional neural network model was trained on near-infrared reflectance (NIR) TIFF images. Ground truth (visible vs no visible SDD) was annotated manually by a trained grader and checked by an experienced grader. Training/validation included 20/15 eyes with SDD and 15/10 without SDD. Predicted and true datasets agreed strongly (Dice coefficient, 0.92). Inference was performed on all eyes, proofread to verify SDD segmentation using both NIR and OCT, and checked further with a map. Mean SDD area in sq mm for each AMD group was compared using linear regression adjusting for age. RMDA and other vision tests were measured as described (PMID 35861686) and associated with SDD area using Spearman correlations, adjusted for age.
Results :
A total of 428 eyes of 428 persons (normal, 218; early(e) AMD, 120; intermediate (i)AMD, 90) were assessed for SDD area and visual function. Mean SDD area differed significantly by AMD severity (p<0.0001): 0.16±0.87 (normal), 2.48±11.23 (eAMD), and 11.97±23.33 (iAMD). Increasing SDD area was correlated with longer (worse) RIT (r=0.27, p<0.0001). The only other significant correlations with SDD area were mesopic (r= -0.13, p=0.0171) and scotopic (r= -0.17, p=0.0007) light sensitivity.
Conclusions :
Use of 55° images allows visualization and quantification of SDD at eccentricities greater than those covered in standard 30° NIR images and OCT. Of vision tests examined, RMDA at 5°, i.e., closer to the fovea than where SDD typically first appear, is the function most strongly related to SDD area. SDD may indicate dysfunctions near the fovea, such as poor transfer to and from circulation, as also suggested by slow RMDA.
This abstract was presented at the 2024 ARVO Annual Meeting, held in Seattle, WA, May 5-9, 2024.