Reliability of a video-based noncontact specular microscope for assessing the corneal endothelium

To assess the within- and between-examiner reliability of corneal endothelial cell density (ECD) measurements with 2 noncontact specular microscopes: a video-based system (SeaEagle) and a photographic system (Noncon Robo).

In 57 eyes of 57 subjects, 47 healthy subjects and 10 patients with corneal disease, ECD was measured in the center and at 4 paracentral areas with the video-based system (SeaEagle) and in the center with a photographic system (Noncon Robo). For the between-examiner reliability, a second examiner measured the central ECD with the SeaEagle device in 12 healthy subjects. The video-based system was also used to evaluate the agreement between the automated and the manually corrected (semiautomated) modes of repeated measurements.

For central ECD in all eyes, the test-retest reliability assessed by the intraclass correlation coefficient was 0.77, 0.99, and 0.91 for the automated SeaEagle, semiautomated SeaEagle, and Noncon Robo, respectively. Correlation was poor to good between the 3 measurement methods (range, 0.61-0.84). For between-examiner reliability, the upper (lower) limits of agreement were 200 (-384) and 149 (-92) cell/mm for the automated and semiautomated SeaEagle, respectively.

The best reliability of repeated measurements was found for the semiautomated SeaEagle. However, the semiautomated method showed significantly higher ECD values than the Noncon Robo, and this difference was highly dependent on ECD. Thus, we suggest that these methods should not be used interchangeably. The fully automated mode may need a further refinement of the cell detection algorithm.