Towards Precise Segmentation of Corneal Endothelial Cells

This article describes an algorithm for defining the precise, objective, repeatable and unambiguous segmentation of cells in images of the corneal endothelium. This issue is important for clinical purposes, because the quality of the grid cells is assessed on the basis of segmentation. Other solutions, including commercial software, do not always mark cell boundaries along lines of lowest brightness. The proposed algorithm is comprised of two parts. The first part determines the number of neighbors of less than or equal brightness to each image point in the input image, then a custom-made segmentation of the binary image is performed on the basis of the constructed map. Each of the 9 iterations of the segmentation considers a number of neighboring points equal to the iteration index, thinning them if they have equal or lower value than the analyzed input point, which allows the boundaries to be routed between cells through the darkest points, thus defining an objective and unambiguous selection of cellst.