Extraction of topological features for the segmentation of endothelial cells in corneal images

The purpose of this work is to experiment the functionality of customized matched filters to detect the cells' borders in corneal endothelium images. The image segmentation is a first important step to achieve parameters regarding the endothelial cells' shape that is an important index of the endothelium health state. On a set of 30 images, three filtering operations and a classification with ROC analysis have been performed to evaluate the filters' performances and the best image of the set