Abstract
Abstract: :
Purpose: To determine the presence of bias, measurement variation, and correlation statistics of quantitative measurements of drusen using the Drusen AnalyzerTM. Methods: Fourteen digital fundus images of 14 eyes with multiple drusen from age–related macular degeneration were analyzed with the A.R.T.T.–Iridex Drusen AnalyzerTM. The system quantitates total drusen area and drusen size distribution in a standarized manner in a specific region of regard, using a variable detection threshold. This threshold is optimized according to the operator’s preference during interactions with the algorithm. TIFF images were analyzed by 2 readers in a masked manner. Each image was presented in a random order and analyzed 10 separate times by each reader. Three measures were evaluated including 1) the presence or absence of a measurement bias in the readings, 2) the correlation of the results between the two readers regarding specific drusen parameters, and 3) the inter–class correlation coefficient (ICC) to determine whether the readers were internally consistent with their measurements. Results: The readers’ determination of total drusen area and relative drusen area were not significantly different from one another. The mean total drusen area difference between readers was 0.093 + 0.128 mm2 when the mean drusen area across all 14 images was 0.45 mm2. Regarding drusen distribution, there was no statistically significant difference except when tiny drusen (less than 63 microns in diameter) were considered. The variables of drusen area, relative drusen area, and drusen distribution were highly correlated, with a correlation coefficent between readers of 0.89 to 0.96 (P < 0.0001). The inter–class correlation coefficent with reader 1 was 0.90 and was 0.80 with reader 2. Conclusion: Semi–automated drusen analysis allows reproducible quantification of drusen characteristics. The results are highly correlated between readers and consistent between readings by the same reader. Such a quantitative analysis of drusen may prove invaluable when characterizing changes of drusen over time.
Keywords: drusen • age–related macular degeneration • imaging/image analysis: clinical