Variability of contour line alignment on sequential images with the Heidelberg Retina Tomograph

Abstract

•Background: The influence of the contour line alignment software algorithm on the variability of the Heidelberg Retina Tomograph (HRT) parameters remains unclear. •Methods: Nine discrete topographic images were acquired with the HRT from the right eye in six healthy, emmetropic subjects. The variability of topometric data obtained from the same topographic image, analyzed within different samples of images, was evaluated. A total of four mean topographic images was computed for each subject from: all nine discrete images (A), the first six of those images (B), the last six of those nine images (C), and the first three combined with the last three images (D). A contour line was computed on the mean topographic image generated from the nine discrete topographic images (A). This contour line was then applied to the three other mean topographic images (B, C, and D), using the contour line alignment in the HRT software. Subsequently, the contour line on the mean topographic images was applied to each of the discrete members of the particular images subsets used to compute the mean topographic image, and the topometric data for these discrete topographic images was computed successively for each subset. Prior to processing each subset, the contour line on the discrete topographic images was deleted. This strategy provided a total of three analyses on each discrete topographic image: as a member of the nine images (mean topographic image A), and as a member of two subsets of images (mean topographic image B, C, and/or D). The coefficient of variation (100×SD/mean) of the topographic parameters within those three analyses was calculated for each discrete topographic image in each subject (“intraimage” coefficient of variation). In addition, a coefficient of variation between the nine discrete topographic images (“interimage” coefficient of variation) was calculated. •Results: The “intraimage” and “interimage” variability for the various topographic parameters ranged between 0.03% and 3.10% and between 0.03% and 24.07% respectively. The “intraimage” coefficients of variation and “interimage” coefficients of variation correlated significant (r ²=0.77;P<0.0001). •Conclusion: A high “intraimage” variability, i.e. a high variability in contour line alignment between sequential images, might be an important source of test re-test variability between sequential images.