The Topcon SP 1000 and Image-NET systems. A comparison of four methods for evaluating corneal endothelial cell density.

The objective of this study was to compare four methods of counting corneal endothelial cell density using the Topcon SP 1000 microscope and Image-NET digital graphic software. Two independent observers quantified cell density from 63 endothelial photomicrographs using the various standardized techniques. The first method was that suggested for use with Topcon SP 1000, employing five different reference grids to compare subjectively with the endothelial cell density of the unknown sample. The other three methods involved the automated, semiautomated, and manual procedures developed for use with the Image-NET system software. The manual Image-NET system is presently considered to yield accurate cell morphology data. Confidence limits and standard errors of mean differences between values obtained by different methods were used to evaluate agreement and reproducibility of computerized methods. Sensitivity and specificity were calculated for two different threshold limits of endothelial cell density. Results showed that the automated Image-NET system is not reliable for clinical use because of its poor agreement with other methods and its lack of sensitivity and specificity at the selected threshold limits of endothelial cell density. The comparative Topcon method of using reference grids, although inexpensive and accurate for 500 cell/mm2 increments, was considered too imprecise for most clinical situations. The semiautomated Image-NET system, in half the analysis time required by the manual method, provided endothelial cell count estimates that were not clinically different from those obtained with manual counting.