Purpose: To investigate the repeatability of the semiautomatic assessment of corneal endothelial cells and its association with the measurement area in the Topcon SP-2000P microscope and IMAGEnet system.
Methods: Specular microscopic images of 86 healthy subjects were captured and analyzed using the Topcon SP-2000P microscope and IMAGEnet system. The same images were analyzed twice, on separate days, by the same examiner using the built-in measurement tool of the IMAGEnet system. The measurement areas were defined with a frame mounted on a computer screen. Four different-sized measurement areas were chosen for the semiautomatic measurements: box A (5.4 × 13.9 cm(2)), box B (4 × 10 cm(2)), box C (4 × 7 cm(2)), and box D (2 × 5 cm(2)). Average cell size (ACS), endothelial cell density (ECD), coefficient of variance, and hexagonality were measured. Repeatability was assessed based on the limit of agreement (LOA).
Results: The means of ACS, ECD, and hexagonality were not statistically different across 4 measurement areas (analysis of variance, P > 0.05). The mean differences (bias) were modest for ACS (range, -1.9∼3.9 μm(2)), ECD (range, -27.2∼14.6 cells per square millimeter), coefficient of variance (range, -0.14∼1.00), and hexagonality (range, -1.3%∼6.8%). Limits of agreement (mean difference ± 1.96× SD) were greater in the measurements with smaller areas: limit of agreement values for ECD were 14.6 ± 99.6, -3.8 ± 101.1, -27.2 ± 179, and -15.8 ± 488 cells per square millimeter for boxes A, B, C, and D, respectively. Similar trends were found in the repeatability of ACS and hexagonality.
Conclusions: Repeatability is improved when larger measurement areas are chosen.