Background: Fundus autofluorescence (AF) imaging using confocal scanning laser ophthalmoscopy (cSLO) has been shown to be superior to fundus photography or angiography for delineating areas of geographic atrophy (GA) in retinal pigment epithelium (RPE) and for recording variation over time. We have evaluated a method for automated computerized detection and quantitation of RPE atrophy.
Methods: AF images in vivo were recorded with a confocal scanning laser ophthalmoscope (exc. 488 nm, em. >500 nm; Heidelberg Retina Angiograph). The intensity of AF in atrophic areas was markedly decreased. Two independent readers analysed these areas in 24 right eyes manually by outlining GA areas using a mouse-driven arrow (method A) and automatically by image analysis software (Global Lab Image/2) after subjective adjustment of thresholding (method B). Agreement between observers and between methods A and B was assessed by the Bland-Altman design for method-comparison studies.
Results: Larger areas were measured using method A than B by both readers (agreement A/B: reader 1 mean difference 1.04 mm, 95% CI [0.66,1.42]; reader 2 mean difference 0.62 mm, 95% CI [0.43,0.81]). The agreement between the readers was mean difference 0.39 mm (95% CI [0.02,0.76]) for A and mean difference -0.03 mm (95% CI [-0.23,0.18]) for B. Features making the delineation of borders of GA difficult included large choroidal vessels with autofluorescent properties in the GA area and media opacities.
Conclusions: Fundus AF cSLO imaging provides a reliable means to delineate areas of GA. The automated image analysis allows more accurate detection and quantitative documentation of atrophic areas than manual outlining. This method will be useful in longitudinal natural history studies and for monitoring effects of future therapeutic interventions to slow down GA progression in patients with advanced atrophic ARMD and other retinal diseases associated with outer retinal atrophy.