Abstract
Purpose::
To compare two methods of evaluating lens opacity area from digitized fundus reflex images: planimetry (grader circumscribes the opacity; the computer calculates area) and estimation (grader estimates percent opacity by grid subfield; an algorithm calculates area).
Methods::
Twenty-seven color stereo fundus reflex photos were selected from the Age-Related Eye Disease Study (AREDS), a combined epidemiological study and clinical trial of nutritive supplements for AMD and cataract. Neitz photo grading had shown these eyes to have >5% posterior subcapsular (PSC) opacity in the central 5 mm diameter zone of the lens. The photos were digitized at high resolution, then displayed and analyzed with Topcon IMAGEnet. Spatial calibration assumed a 12 mm horizontal iris diameter, and a digital version of the AREDS lens grid was applied. Lenses were viewed in stereo using the "ScreenVu Stereoscope". Area of PSC opacity in the central zone was estimated manually with the grid method (AREDS Report No. 4, Am J Ophthalmol 2001; 131:167-175). For each eye, manual estimation was immediately followed by planimetry, in which PSC opacities were outlined with the IMAGEnet "Area" tool. Two lens evaluators independently graded the entire image set.
Results::
Within each evaluator, PSC opacity area by planimetry vs. estimation yielded intra-class correlation coefficients (ICC) of 0.97+ (nearly perfect agreement), with no evidence of systematic shift. Between evaluators, however, disagreements on opacity boundaries (reflected similarly in grader vs. grader plots from both techniques) lowered the ICC to ~ 0.7. Outliers that accounted for lower correlation had less stereo effect, complicating identification of opacities as PSC. Contrast enhancement and color separation are being explored to improve the grader’s ability to discern cataract boundaries. We will also present data on using planimetry vs. estimation for evaluating cortical opacities in digitized film images, and for PSC as well as cortical opacities in natively digital images.
Conclusions::
For determining area of PSC lens opacity, computerized planimetry on digital images produces results nearly identical to those from the manual estimation method. With digital images, planimetry may be a more efficient alternative.
Clinical Trial::
www.clinicaltrials.gov 1
Keywords: cataract • clinical (human) or epidemiologic studies: outcomes/complications • clinical (human) or epidemiologic studies: natural history