Comeal Endothelium cell segmentation and count using K-means and watershed algorithms

Abstract

Corneal endothelial cell segmentation and count is used to diagnose the cornea and validate the health of the eye. Loss of cells, due to damage of endothelial cells leads to dis-functioning of cornea and loss of vision. Unsupervised k-means algorithm, along with the watershed algorithm are used for segmentation of the individual cell areas and hence determine the cell count. Further, cornea related Polymegathism and Pleomorphism are calculated based on the individual cell areas and cell centroids. Algorithms provides a cell count accuracy of 96.87%, with Polymegathism and Pleomorphism.