Purchase this article with an account.
KS Elsner, R Whittock, J Alexander, R Rees, B Madjarov; Grading Drusen from Digital Images . Invest. Ophthalmol. Vis. Sci. 2002;43(13):2503.
Download citation file:
© ARVO (1962-2015); The Authors (2016-present)
Purpose:To evaluate the agreement between grading drusen from digital images on a calibrated monitor versus grading drusen from film-based images on a lightbox. Methods:15 patients (30 eyes) were randomly selected from the Complications of Age-Related Macular Degeneration Prevention Trial (CAPT). The higher quality frame from each baseline F2 color stereoscopic pair was scanned at 1000 dpi (Nikon Super CoolScan 4000), 24 bits/pixel to an image size of 1100x945. A digitally generated template of the standard fields for grading age-related maculopathy was overlaid onto each digital image using custom developed software. For digital image grading, templated images were viewed on a calibrated monitor with XGA resolution. For film-based grading, templated images were re-rendered as 35 mm slides and viewed stereoscopically with a Donaldson 5X viewer on a lightbox. Two experienced graders independently graded all digital and film-based images using an ordinal scale as specified by the CAPT Reading Center protocol. Images were evaluated in 3 subfields for number of drusen ≥125 microns, largest drusen size, area covered by drusen and drusen confluence, as well as globally for predominant drusen size, area covered by drusen and presence of reticular pseudodrusen. Adjudicated monitor gradings were compared to adjudicated lightbox gradings using percent exact agreement and the weighted kappa statistic Κwt. Results:For the seven drusen questions, agreement ranged from fair to excellent. Agreement for the gradings in each subfield of number of drusen was substantial to excellent (73 - 100%; Κwt = 0.73 - 1.00) and moderate to excellent for largest drusen size (60 - 100%; Κwt = 0.48 - 1.00), area covered by drusen (40 - 93%; Κwt = 0.00 - 0.93), and drusen confluence (73 - 93%; Κwt = 0.44 - 0.92). Agreement was perfect for reticular pseudodrusen (100%; Κwt = 1.00), but fair to moderate for predominant drusen size (73%; Κwt = 0.16 - 0.58) and global drusen area (73 - 87%; Κwt = 0.38 - 0.44). Conclusion:Grading of drusen on a calibrated color monitor yielded high concordance with 'gold standard' lightbox grading, although discrepancies on some features warrant further investigation. Future research studies may be able to rely exclusively on the digital environment for grading drusen in patients with macular degeneration.
This PDF is available to Subscribers Only