Multiobjective Differential Evolution-Based Fuzzy Clustering for MR Brain Image Segmentation

The problem of classifying magnetic resonance (MR) brain images into different tissue classes has gained significant attention during the last couple of decades. In this regard, unsupervised clustering methods are used to group the pixels of images in the intensity space. Most of these clustering methods optimize a single clustering measure. However, to improve the clustering further multiple conflicting clustering measures can be used. In this chapter, the Multiobjective Differential Evolution based Fuzzy Clustering technique is used to optimize multiple clustering measures simultaneously. Here differential evolution is used as the underlying optimization technique and the cluster centers are encoded in a vector of differential evolution. In the final generation, it produces a set of nondominated solutions, from which the most promising one is selected based on a Minkowski Score as a final clustering solution. The proposed technique is applied on several simulated T1-weighted, T2-weighted, and proton density for normal and multiple sclerosis lesion magnetic resonance brain images. The performance of the differential evolution-based fuzzy clustering technique is demonstrated quantitatively by comparing it with other state-of-the-art methods. Apart from this, the segmented MR brain images produced by the proposed technique are also compared with the available ground truth information. Finally, a statistical significance test is conducted to judge the superiority of the results produced by the proposed technique.