Endothelial cell analysis using the Topcon SP-1000 non-contact specular microscope and IMAGEnet system

Introduction: A healthy corneal endothelium is essential for corneal transparency. The appearance of the corneal endothelium can be assessed quantitatively using the specular microscope combined with computer technology. Methods: Images of the corneal endothelium of 20 eyes from 12 subjects, who had normal ocular health, were captured and analysed using the Topcon SP-1000, connected to the IMAGEnet system. Three methods of endothelial analysis (untreated, blocking and retracing) were used on each captured image. In the untreated method, the results of the analyses were automatically provided by the computer once the image was captured. For the blocking method, incorrectly traced or ill-defined cells were blocked out by the operator manually and the endothelial variables were recalculated by the computer. For the retracing method, incorrectly traced cell borders were erased and reconstructed manually and the endothelial variables were recalculated. Results: The results obtained from the three methods of analysis were significantly different from each other (Repeated Measures Analysis of Variance, p < 0.05). Untreated results showed the highest coefficient of variation (mean ± SD: 60.1 ± 9.7 per cent) and endothelial cell density (3424.5 ± 227.7 cells/mm2) and the lowest hexagonality (38.8 ± 6.2 per cent). The results obtained using the retracing method show the lowest CV (30.2 ± 4.1 per cent) and ECD (2893.9 ± 257.2 cells/mm2) and the highest hexagonality (61.3 ± 11.4 per cent). CV. ECD and hexagonality using the blocking method were 34.0 ± 5.2 per cent, 3013.3 ± 240.8 cells/mm2and 48.5 ± 7.4 per cent, respectively. Conclusion: The method of retracing cell borders was more time-consuming than when untreated data was used for analysis. Nevertheless, clinicians should consider using the retracing method of analysis since the results using untreated data provided by the IMAGEnet system were significantly different from the results using retraced data, and therefore were less accurate.