Purchase this article with an account.
BE McCarey, MJ Lynn, HF Edelhauser; Non-clinical Specular Microscopy: Accuaracy And Repeatability . Invest. Ophthalmol. Vis. Sci. 2002;43(13):3177.
Download citation file:
© ARVO (1962-2015); The Authors (2016-present)
Purpose:To assess the factors in the accuracy and repeatability of endothelial cell morphology analysis. Methods:The ROBO SP8000 Clinical Specular Microscopy (Konan Medical Corp.) was used to compare repeatability of cell analysis by Frame and Center Method. Deviations in cell density values with errors in centering the dots and omitting cells were measured. The accuracy of the Center Method was determined by recounting the same field of a perfect hexagon pattern and with specular micrographs of the human endothelium. The repeatability with various coefficient of variations (CV) and number of cells utilized per image were determined. This was accomplished by analyzing 16 sets of data with increasing number of cells (10 to 150 cells/image) from cells patterns with CV of 25 or 45. The homogeneity of the endothelial morphology was assessed by comparing image samples from various locations on the cornea. Image samples (n=13) were collected from within a 1 to 4 mm diameter zone of central cornea. One Way Analysis of Variance and the per cent difference ((CDmax -CDmin)/CDmax) x 100) were used to compare the effect of image sample location. Results:Frame Method has inherent weaknesses but when performed carefully it is equivalent to the Central Method (p=1.000). One hundred cells were analyzed with the Center Method then reanalyzed with 1 to 10 cells de-centered. The average errors ranged -0.2% to +0.1%. Omitting cells during the analysis had approximately -1% reduction in cell density per omitted cell. The Center Method was repeatable on a perfect hexagon and human endothelial cell pattern (p=1.000). The standard deviations for the mean cell density minimized and stabilized with ≷100 cell analyzed. The analysis also demonstrated cell-clustering effects at <100 cells analyzed. Data spread was less with CV=25 than CV=45. Within a 4 mm diameter zone the comparison of cell density among image samples had a p=0.0004; a 3 mm zone had p=0.002; a 2 mm zone had p=0.24; and 1 mm zone had p=0.28. Ten sequential central images had a p=0.98 and a per cent difference of 4%. Conclusion:The Konan non-contact specular microscope and software analysis is repeatable and accurate if performed with care. Based on this analysis, its use can be recommended for clinical trials.
This PDF is available to Subscribers Only