Purchase this article with an account.
Cristina Modak, Jianyan Huang, Jyotsna Maram, Ken Marion, Luisa Obregon, Moritz Niemeyer, Srinivas R Sadda, Vikas Chopra, Olivia L Lee; Comparison of Manual and Automated Grading Methods for Corneal Endothelial Cell Density Measurements by Specular Microscopy. Invest. Ophthalmol. Vis. Sci. 2014;55(13):2064.
Download citation file:
© ARVO (1962-2015); The Authors (2016-present)
To determine the reliability of automated specular microscopy grading compared to validated manual methods (Center >100 cell count; Flex Center < 100 cell count) and the factors that predict this reliability.
Central images from 58 control and 39 glaucoma eyes were captured with Konan’s non-contact specular microscope NP-9900. All images were analyzed by trained graders using the automated method (AutoTrace, AT) in Konan’s CellChek Analysis Software. The fully-automated version (AT-FA) was compared to the semi-automated (AT-SA) one, where cell size is manually selected. All images were also graded manually with the Center (C) method. Images with low cell count (<100 cells) were also graded with the Flex Center (FC) method. ECDs from the different methods were compared to each other and absolute error was used to assess variation. The effect of age, cell count, cell size variation, and cell size on ECD variation between methods was evaluated.
For manual methods, minimal difference was observed between C (885 cells/mm2) and FC (881cells/mm2) in all analyzed corneas (p>0.2). When AT was compared to manual (Table 1), minimal difference (p=0.32) was observe with AT-FA only in controls <40yrs, which showed highest numbers of small cell size and lowest cell size variation. In older controls and diseased eyes, which showed overall greater cell size variation and/or larger cells, ECD difference with AT-FA was significant (p<0.2) but could be significantly reduced with AT-SA (p=0.86) in corneas with uniformly larger cell size.
While good correlation was observed between different manual methods, suggesting that, with trained graders, Center method is still a reliable option for ECD determination even at low cell counts, our findings raise concerns for relying on automated methods for determining ECD values. The reliability of automated methods is particularly affected by cell size variation and maximum cell size (both of which often increase with age and disease); fully-automated analysis should be considered only in corneas of uniformly small cells. While the semi-automatic option may still be reliable in corneas with uniformly larger cells, manual methods are strongly recommended in eyes where cell size is not uniform.
This PDF is available to Subscribers Only