Purpose: Corneal endothelium plays an important role in the maintenance of corneal transparency and stability. The purpose of this study was to evaluate the performance of the TOPCON SP-2000P and IMAGEnet system in terms of (a) the difference in results between automated endothelial cell analysis and retraced cell analysis, (b) the differences in the endothelial cell analysis when using the first image and the clearest of three images taken with the SP-2000P, (c) the repeatability and reproducibility of the IMAGEnet system in the determination of corneal endothelial cell variables, and (d) the repeatability and reproducibility of the SP-2000P in capturing endothelial cell images for endothelial cell analysis.
Methods: Two experiments were performed. Twenty male subjects participated in the first experiment in which endothelial images were captured by examiner SWC and endothelial cell analysis was performed by two examiners, SWC (twice on two different days) and PC (once only). Nineteen male subjects participated in the second experiment in which endothelial cell images were captured by examiners SWC and PC on Visit 1 and by examiner SWC on Visit 2. Endothelial cell analysis was performed by examiner SWC only. The 95% limits of agreements (95%LA) and the intraclass correlation coefficient (ICC) between tested parameters were determined.
Results: Automated cell analysis significantly underestimated average cell size (ACS) and hexagonality, and overestimated endothelial cell density (ECD) and coefficient of variance of cell size (CV) (p < 0.05). No statistically significant differences were found in the endothelial cell variables between the first and the clearest images (p > 0.05). There were no statistically significant intra- or inter-examiner differences in any of the endothelial variables determined (p > 0.05). For a set of endothelial cell images, the reproducibility and repeatability of the endothelial cell variables determined with the IMAGEnet system were good, with ICC > 0.9. No significant inter-visit or inter-examiner differences were found (p > 0.05). Reliability was good for the determination of ECD and ACS (ICC > 0.9) but poor or average for the determination of CV and hexagonality (ICC: 0.4-0.8).
Conclusions: We suggest the use of the clearest image with the retraced method for endothelial cell analysis with the IMAGEnet system, as the variability in hexagonality results between the first and the clearest images can be clinically significant. For the same set of images captured by one examiner, the TOPCON IMAGEnet system was reliable in determining ECD and ACS, and was fairly reliable in determining hexagonality and CV. When images were captured on different days or by different examiners, the TOPCON SP-2000P with the IMAGEnet system gives repeatable and reproducible values for ECD and ACS only. Caution should therefore be exercised when using CV and hexagonality to monitor endothelial changes in comparative studies.