What is lost by digitizing stereoscopic fundus color slides for macular grading in age-related maculopathy and degeneration?

Abstract

Objective To compare 35-mm stereoscopic slide transparencies with digitized nonstereoscopic images (resolution 1024×768 pixels) for grading abnormalities in age-related maculopathy (ARM) and age-related macular degeneration (AMD). Design Comparative observational case series. Participants Twenty-five patients (50 eyes) with ARM and/or AMD. Methods Twenty-five patients with ARM/AMD in at least 1 eye were randomly selected from a large ongoing collection of clinical data and DNA in a tertiary referral United Kingdom population. Retinal photography was performed with mydriasis using the Zeiss FF-series 30° fundus camera on Ektachrome slide transparency film. The images were centered on the macula. The color transparencies were then digitized. The grading process has been set up based on the International ARM Epidemiology Study Group. All images were independently graded by 3 retinal specialists. Both κ statistics and exact agreement were calculated to assess agreement between and within observers and between the 2 master copies derived from the gradings of the color slides and digitized images. Main outcome measure Agreement between the 2 master copies derived from the gradings obtained from stereoscopic slide transparencies and digitized nonstereoscopic images. Results For small hard and intermediate soft drusen, agreement ranged between 77% and 91% (κ, 0.56–0.72) and 83% and 93% (κ, 0.31–0.64), respectively, for the 3 macular subfields. Agreement for the presence of hyperpigmentation was 12% to 56% (κ, 0.00–0.27). Agreement was 94% to 96% (κ, 0.80–0.82) for the presence of geographic atrophy and 93% (κ, 0.78) for the area covered. For the presence of choroidal neovascularization (CNV), agreement was 94% to 98% (κ, 0.81–0.88), and it was 95% (κ, 0.83) for the area covered. For individual features of CNV, exact agreement was 88% to 96% (κ, 0.22–0.49). In 3 cases of geographic atrophy and 2 cases of CNV, the lesion was missed on digitized images. Conclusions Because of the close agreement for most categories between the grading of stereoscopic color slides and digitized images, digitized nonstereoscopic color images prove to be useful for grading ARM and AMD.