Overestimation of Corneal Endothelial Cell Density by Selecting Smaller Areas to Count the Cells in Images of In Vivo Confocal Microscopy

Abstract

Purpose: Laser scanning in vivo confocal microscopy (LS-IVCM) is commonly used to measure corneal endothelial cell density (CECD). However, due to corneal applanation during imaging, it is common to see dark non-focused areas in LS-IVCM images which make it impossible to count all corneal endothelial cells in the image. It is unclear whether counting endothelial cells in smaller areas in LS-IVCM images will affect the measured CECD.

Methods: 32 endothelial images by LS-IVCM (Heidelberg Retina Tomograph 3 with Rostock Corneal Module; Heidelberg, Germany) with different endothelial cell densities were analyzed. In each image (160 000 µm²), endothelial cells were counted within the following areas: 80 000, 40 000, 20 000, 10 000, 5 000, and 2 500 µm². The cell numbers counted in these areas were then expressed as cell/mm², and the CECD were compared.

Results: The mean measured CECD was 2007 ± 750 cells/mm², 2109 ± 764 cells/mm², 2149 ± 770 cells/mm², 2290 ± 750 cells/mm², 2495 ± 758 cells/mm², and 2760 ± 854 cells/mm² for areas of 80 000 µm², 40 000 µm², 20 000 µm², 10 000 µm², 5 000 µm², and 2 500 µm², respectively; the differences between all these groups were statistically significant (all P<0.05). Compared to the CECD in 80 000-µm² area, the mean percentage of overestimation was 6.2 ± 6.7% for the 40 000-µm² area, 8.9 ± 11.4% for the 20 000-µm² area, 18.7 ± 21.1% for the 10 000-µm² area, 27.9 ± 23.6% for the 5 000-µm² area, and 45.6 ± 34.9% for the 2 500-µm² area. There were significant negative correlations between the measured CECD for the 80 000-µm² area and the percentage of overestimation in the smaller areas (R= -0.44 to -0.71, all P<0.05).

Conclusions: Counting endothelial cells in smaller areas of LS-IVCM images results in a significant overestimation of CECD.