Abstract
Purpose::
To determine whether lens opacity evaluations from digital fundus reflex (FR) images are comparable to film image evaluations.
Methods::
UTMB-Galveston (H.L.) recruited subjects with diabetic retinopathy (154 eligible eyes) to investigate whether ETDRS retinopathy severity level evaluated from digital color stereoscopic 30-degree fundus photos is comparable to that assessed from film. Subjects were chosen to capture the full range of retinopathy severity. We evaluated the FR images from this image set to investigate comparability of evaluating lens opacities in digital vs. film images. Film images were evaluated for cortical opacity to select a study sample of 50 eyes: 10 eyes with no cortical opacity, 10 with small amounts (up to ~5% total) of cortical opacity and 30 eyes with more extensive cortical opacity. Standardized procedures were developed to optimize illumination and contrast of the digital images. Evaluation was done masked to group, and film images were evaluated independently at least one week apart from corresponding digital images to minimize recall. Digital images were evaluated using IMAGEnet software. The AREDS lens opacity classification (AREDS report 4, AJO 2001; 131:167-175) was applied to both digital and film images, special focus to the central zone (5mm diameter circle).
Results::
The intraclass correlation coefficient (ICC) was 0.92 between digital and film images for cortical opacity area. There was no evidence of systematic difference between imaging media for presence/absence. The group with more extensive area had higher ICC (0.93), while the group with smaller area had less agreement (because differences are proportionally greater compared to absolute amount).
Conclusions::
This pilot study suggests that evaluations of cortical lens opacity from digital and film images are comparable. The correlation is about the same as observed for masked inter-evaluator comparisons between film FR images in AREDS 1 (AREDS report 4, AJO 2001; 131:167-175). This suggests that the variability we observed between digital and film images may be due to variability of the evaluator rather than differences between the imaging media.
Keywords: cataract • clinical (human) or epidemiologic studies: biostatistics/epidemiology methodology • clinical (human) or epidemiologic studies: outcomes/complications