Corneal Endothelial Microscopy: Does a Manual Recognition of the Endothelial Cells Help the Morphometric Analysis Compared to a Fully Automatic Approach?

Abstract

This study investigated whether manual integration in the recognition of the endothelial cells produces different outcomes of morphometric parameters compared to a fully automatic approach. Eight hundred and ninety endothelial images, originally acquired by the Perseus Specular Microscope (CSO, Florence, Italy), from seven positions of right and left corneas were selected from the database of the Research Centre in Optics and Optometry at the University of Milano-Bicocca. For each image selected, two procedures of cell identification were performed by the Perseus: an automatic identification and a manual-integrated procedure to add potential additional cells with the available editing tool. At the end of both procedures, the endothelial cell density (ECD), coefficient of variation (CV), and hexagonality (HEX) of the mosaic were calculated. The HEX in the two procedures was significantly different for all comparisons (p < 0.001), but clinically negligible. No significant differences were found for the CV and ECD in the images of both eyes irrespective of the corneal position of acquisition (except for ECD in three corneal portions, p < 0.05). To conclude, it is possible to recognise a significantly higher number of cells using the manual-integrated procedure than it is using the fully automatic one, but this does not change the morphological parameters achieved.