Validation of a new software system for automatic endothelial cell density calculation of donor corneas

Purpose: To validate the reliability of a new automatic system for endothelial cell density calculation. Methods: Endothelial cell densities of 100 donor corneas were manually counted (MCC) (two central images per cornea, three experienced investigators, fixed frame method on digitalized 200x images). The same two 200x images and two 100x images from these 100 corneas were automatically analysed (ACC) by a new software system. Results: MCC: mean 2376 cells/mm2 (range 1706 to 3098); mean deviation between the three investigators: 97 cells/mm2 (min. difference 5; max. 249). 1. ACC on 100x images: mean 2389 cells/mm2 (range 1692 to 3072); ACC compared to MCC: average difference 14,0 cells/mm2 (range – 333 to + 333), R2 0,92. 2. ACC on 200x images: mean 2350 cells/mm2 (range 1804 to 2996); ACC compared to MCC: average difference – 26,0 cells/mm2 (range – 324 to + 370), R2 0,91. 3. Mean values ACC on 100 and 200x images: 2370 cells/mm2 (range 1748 to 3015); ACC compared to MCC: average difference – 6,0 cells/mm2 (range – 328 to + 292), R2 0,93; overestimation of more than 150 cells/mm2 in 8%, underestimation of more than 150 cells/mm2 in 8% of all cases. In 67% deviation smaller than +/- 100 cells/mm2 . Conclusion: The ACC gives very reliable results and correlates highly with the MCC on 100 an 200x images.