Morphometry of organ cultured corneal endothelium using Voronoi segmentation

Abstract

The endothelial viability of eye-banked corneas must be assessed before transplantation, yet phase-contrast images of the endothelium of corneas stored in an Organ Culture system have a huge variety of endothelial appearance at different focal planes, due to stromal swelling and an evaluation technique that swells the cells and their intercellular spaces to make them visible. This makes them difficult to assess using common edge-based segmentation methodologies that have mostly been developed in vivo using specular microscopy. Additionally, as the endothelial cell architecture has radically different form at apical and basal poles, segmentation of only the apical view cannot describe the cell as a whole. Using custom software, this study instead defines the cell borders using a calibrated Voronoi diagram, a region-based image segmentation technique that uses a point cloud of cell centroids as the only input required to measure the size (polymegathism) and shape (pleomorphism) of the endothelial cells. Measurements included the range and variation in the cell area and density, the number of cell sides, and several novel shape descriptors. A dataset of endothelial cell measurements was compiled from 2000 images of 678 corneas comprising 354,998 cells, and validated by comparison to previous studies from a variety of tissue sources, imaging modalities, and analysis techniques. While there were some differences in cell size variation and pleomorphic composition than seen in some other studies, much of the data was directly comparable, demonstrating fast and accurate endothelial cell morphometry at a large scale for a wide range of routine organ cultured corneas.