Reproducibility of Automatic Endothelial Cell Analysis Performed With Non-Contact Specular Microscopy

Objective: To investigate the reproducibility of automatic endothelial cell analysis performed with non-contact specular microscopes. Method: Fifteen eyes of 15 case were included in the study. From each eye 2 consecutive endothelial images of central cornea obtained with the Topcon SP-2000P by the first observer. Firstly, first image of each eye was assessed by correcting the cell borders defined by automatic analysis system (IMAGEnet) and using standart frame. "Intra-operator reproducibility" was investigated by comparison of results of these analysis with the analysis results of uncorrected cell borders defined by the automatic analysis system (I), with the use of smaller frame rather than the standard frame (II) and with the analysis results of second image of each eye captured by the first observer (III). In the second phase, "inter-operator reproducibility" was investigated by comparison of the results of the first observer with the results obtained by assessment of the first image of each eye by a second observer. Findings: Both the assesment made for the intra-operator reproducibility and inter-operator reproducibility revealed a medium-low degree of correlation for the coefficient of variation and hexagonality rates while a high correlation was observed for the values of cell density and average cell area. Conclusion: In conclusion, assesments made with the non-contact specular microscopy and automatic analysis system have a substantially high rate of reproducibility when standard frame and corrected cell borders are used.