Agreement of Measurements from Red-free Fundus Photographs, Fluorescein Angiograms, and Indocyanine Green Angiograms

Abstract

Abstract: : Purpose: To assess the agreement of measurements taken from three different imaging technologies in unmagnified and magnified form. Interobserver agreement of measurements is also analyzed. Methods: Images were randomly chosen from patients who had red-free fundus photography (RF), IVFA, and ICG done during a single visit to the Queen’s University Department of Ophthalmology. Twenty five sets of fundus images were examined from 25 patients. I mages were stored as TIFF files. Three measurements were taken from each image: 1) vertical disc diameter, 2) arterial diameter crossing the disc margin, 3) venous diameter crossing the disc margin. Measurement were converted to a ratio based on the venous diameter of the image in order to account for magnification differences between images. Two such ratio measurements were recorded for each magnified and unmagnified image for purposes of statistical comparison. Measurements from unmagnified TIFF images were taken using the Topcon 2000 software. To measure magnified images, each TIFF file was exported to Adobe Photoshop 7.0 and was then magnified 200%. Correlation was determined using the Pearson correlation coefficient. Confidence intervals for correlation data were determined after a Fischer transformation was performed. Agreement was illustrated by plotting the difference between two methods (or observers) against their mean. Assuming a best correlation of 0.9, to detect a worst correlation of 0.3 with a ß=0.2 and α=0.05, a sample size of N=22 was needed. Results: Correlation of RF and IVFA measurements was highest: 36% when measuring arteries, 73% discs. Correlation of RF and ICG measurements were lowest: 12% when measuring arteries, 39% discs. Inter-observer correlation of artery measurements was best for IVFA (70%), but poor for RF (20%) and ICG (27%). Inter-observer correlation of disc measurements was good for RF (75%) and IVFA (83%), but poor for ICG (36%). Correlation of measurements between imaging modalities is improved with magnification, particularly for arterial measurements, however correlation remains poor overall. This is also true of inter-observer correlation. Conclusions: Measurements taken from red-free photographs, IVFA images, and ICG images show generally poor agreement, even with 200% magnification. This is also true for inter-observer agreement of such measurements.

Keywords: retina • reading