Near-Infrared Reflectance Imaging for Quantification of Atrophy Associated with Age-Related Macular Degeneration

Abstract

Purpose: To compare measurements of area of geographic atrophy (GA) in dry age-related macular degeneration (AMD) obtained by fundus autofluorescence (FAF) to those obtained by near-infrared reflectance (NIR).

Design: Interrater reliability analysis.

Methods: Ninety-seven confocal NIR images (Heidelberg HRA + Spectralis) and FAF images from 97 patients/eyes with GA with dry AMD were collected retrospectively from existing anonymized Doheny Image Reading Center datasets. Two masked reading center graders (N.S., J.S.) independently and blindly performed manual segmentation of the GA lesions on each NIR and FAF image using GNU Image Manipulation Program software (version 2.8.22). GA on NIR/FAF images was defined in accordance to recently published Classification of Atrophy Meeting criteria as sharply demarcated hyperreflective regions ≥250 μm in diameter. The difference and point-to-point correspondence between gradings in GA area measurements between NIR and FAF were assessed by mean difference, overlap ratio, and Dice similarity coefficient.

Results: Among the 97 eyes with dry AMD, the mean GA area was 7.62 ± 7.77 mm² from FAF images and 7.65 ± 7.83 mm² from NIR, with a mean nonsignificant difference of 0.31 ± 0.55 mm² (2-tailed t test, P = .65). The overlap ratio in the segmented GA lesion between modalities was 0.84 ± 0.28 with a Dice similarity coefficient of 0.87 ± 0.27. Intermodal reliability was high (intraclass correlation coefficient = 0.998, P < .01). Of note, in 5 cases (5.2%), the GA lesion could be identified on the FAF image but not on the NIR image, translating into a sensitivity of 94.8%.

Conclusions: GA lesions in dry AMD can be identified and quantified reliably using NIR images in many cases, though eyes with a thin choroid resulting in isoreflective GA lesions may be challenging. NIR imaging is comfortable for patients and is commonly obtained along with OCT, and therefore NIR-based GA assessment may be a useful surrogate in clinical settings.