Fuzzy Entropy k-Plane Clustering Method and Its Application to Medical Image Segmentation

Abstract

AbstractMRI images are complex, and the data distribution of tissues in MRI are non-spherical and overlapping in nature. Plane-based clustering methods are more efficient in comparison to centroid based clustering for non-spherical data, and soft clustering methods can efficiently handle the overlapping nature by representing clusters in terms of fuzzy sets. In this paper, we propose fuzzy entropy k-plane clustering (FEkPC), which incorporates the fuzzy partition entropy term with a fuzzy entropy parameter in the optimization problem of the conventional kPC method. The fuzzy entropy parameter controls the degree of fuzziness, the same as the fuzzifier parameter in the fuzzy clustering method. The fuzzy entropy terms try to minimize averaged non-membership degrees in the cluster. The performance of the proposed method has been evaluated over three publicly available MRI datasets: one simulated and two real human brain MRI datasets. The experimental results show that the proposed FEkPC method outperforms other state-of-the-art methods in terms of ASA and Dice Score. KeywordsFuzzy setsFuzzy entropy measurek-plane clusteringFuzzy clusteringMRI image segmentation