This article proposes a multi-objective clustering ensemble method for medical image segmentation. The proposed method is called adaptive multi-objective archive-based hybrid scatter search (AMAHSS). It utilizes fuzzy clustering with optimization of three fitness functions: global fuzzy compactness of the clusters, fuzzy separation and symmetry distance-based cluster validity index. The AMAHSS enables the search strategy to explore intensively the search space with high-quality solutions and to move to unexplored search space when necessary. The best single solution is processed using the metaclustering algorithm. The proposed framework is designed to segment lung computed tomography images for candidate nodule detection. This candidate nodule will then be classified as cancerous or non-cancerous. The authors validate the method with standard k-means, fuzzy c-means and the multi-objective genetic algorithm with different postprocessing methods for the final solution. The results obtained from the benchmark experiment indicate that the method achieves up to 90% of the positive predictive rate.
Abstract. The aim of this paper is to propose and apply state-of-the-art multiobjective scatter search for solving image segmentation problem. The algorithm incorporates the concepts of Pareto dominance, external archiving, diversification and intensification of solutions. The multiobjective optimization method is Archive-based Hybrid Scatter Search (AbYSS) for image segmentation. It utilized fuzzy clustering method with optimization of two fitness functions, viz., the global fuzzy compactness of the clusters and the fuzzy separation. We have tested the methods on two types of grey scale images, namely SAR (synthetic aperture radar) image and CT scan (Computer Tomography) image. We then compared it with fuzzy c-means (FCM) and a popular evolutionary multiobjective evolutionary clustering named NSGA-II. The performance result for the proposed method is compatible with the existing methods.
scite is a Brooklyn-based organization that helps researchers better discover and understand research articles through Smart Citations–citations that display the context of the citation and describe whether the article provides supporting or contrasting evidence. scite is used by students and researchers from around the world and is funded in part by the National Science Foundation and the National Institute on Drug Abuse of the National Institutes of Health.