Segmentation of multimodal medical volumes using evolutionary clustering

The Evolutionary Clustering (EC) algorithm presented in this paper is based on a (\mu,\lambda)-Evolution Strategy where the object variables of genotypes are the centers of clusters. In the experimental section we compare the segmentation obtained by the application of C-Means (CM) algorithm and two variants of EC to a simple data set consisting of a multimodal transverse slice from a MRI head acquisition volume. EC obtains more stable solutions than CM, and, as it can take into account the cardinality of clusters, dramatically improves the quality of segmentation results.