2014
EMSS 2014 Proceeding
Identification of endothelial cell morphology in cornea using evolution strategies
Authors: Lisa Obritzberger, Susanne Schaller, Viktoria Dorfer, Claudia Loimayr, Simone Hennerbichler, Stephan Winkler
Abstract
In this paper we describe an algorithm for identifying shapes of endothelial cells in images of human cornea. In order to analyze the quality of cornea samples that shall be transplanted, endothelial cells have to be counted and the morphology of cells has to be identified. The algorithm for shape detection presented in this paper is based on evolution strategies (ES) After several pre-processing steps an initial shape is fitted into the detected cells and repeatedly evaluated and mutated in order to create new candidates from which the best ones are promoted to the next generation. In the experimental section of this paper we analyze the performance of this approach for identifying cell morphologies in images of human cornea samples using different fitness functions.
I3M Scientific Sponsors
I3M Industrial Sponsors
I3M Media Sponsors
