Variability in digital assessment of cortical and posterior subcapsular cataract

To identify validity of the standardised Nidek EAS-1000 retroillumination image analysis, images of 450 consecutive patients were analysed for the standard 6.5 mm and for the maximal pupil size. The software allows for separation of cortical and posterior subcapsular opacities and defines threshold for cataract automatically at 12% below the brightest point of the histogram of pixel luminescence. The results were compared with clinical Wilmer cataract grading. Correlation between clinical and digital assessment was 0.48(*) for cortical opacities in maximal pupil size, 0.47(*) in 6.5 mm pupil size analyses, and 0.71(*) for posterior subcapsular opacities ((*)p < 0.001). In 24.6% of maximal pupil size analyses and in 11.7% of standard pupil size analyses standardisation revealed confounding features, such as other opacities of media, refractive shadows etc., that masqueraded as cataract and interfered with the cortical opacity measurements. Automatic standardized analysis has reduced many sources of observer variation (level of illumination, pupil size and threshold adjustment), but the revealed range of confounding opacities and artifacts still requires observer interpretation.