Objectives: In eye banks, endothelial cell density (number of endothelial cells per surface unit, ECD) is the main criterion to validate the quality of corneal grafts. Stored corneas present numerous deep endothelial folds. Consequently only few endothelial areas are parallel to the microscope and parallax errors inevitably lead to overestimating ECD.

Aim: to present an innovative method of ECs count on the reconstructed 3D topography of the endothelium.

Materials and methods: Stacks of endothelial images were acquired using a Z motorized conventional transmitted light microscope. A depth map was obtained by researching maximized measurements of the focus in images (shape-from-focus method). Texture of the surface was built by summing image parts presenting the right focus leading to an all-in-focus image. By calculating first derivative in all directions of depth map, precise estimation of cells in folds surfaces could be found. Accuracy of slope detection was verified using artificial endothelial mosaics micro printed on a glass slide with known ECD. Slides were tilted with an angle of 20° and 45°. Gain in accuracy was calculated using human organ cultured donor corneas. For each cornea, the new 3D count was compared with the conventional 2D count of the same area as usually performed in eye banks. The mean difference between both counts and its 95% confidence interval was calculated.

Results: The 3D reconstruction eliminated the areas that were usually blurred and therefore increased the number of visible cells that could be counted in each field. The 3D reconstruction algorithms precisely detected the local slopes. The conventional 2D counts overestimated the ECD by a mean 8.5% _95%CI(5.5–11.5), because of parallax error.

Conclusion: The 3D endothelial cell counting is ready to be integrated in eye-banks in order to improve the accuracy of endothelial quality controls.