Automatic estimation of endothelium cell density in donor corneas by means of Fourier analysis

One of the main clinical parameters expressing the health of a cornea is the cell density of its endothelium. This information is particularly important in an eye bank environment, where donor corneas are screened to assess their suitability as a human graft. Endothelium cell density is conventionally estimated by a long, tedious and error-prone manual counting procedure, performed by cornea experts on specimen images observed through an optical microscope. An alternative solution is proposed: a computer program that provides automatic estimation of cell density in donor corneas by analysing the spatial frequencies contained in the image. A circular band in the 2D discrete Fourier transform of the image is shown to contain the relevant information about the cell density. A system for extracting from this spatial frequency information an estimate of the cell density has been developed. A clinical evaluation of the proposed technique was performed on 18 corneas, where the densities provided by the proposed technique were compared with those manually obtained by two experts. The results showed an average percentage difference of 3% (maximum 19%), a value well within the measured inter-expert range of variability. The proposed automatic procedure confirmed its ability to estimate correctly corneal endothelium cell density.