Redescending intuitionistic fuzzy clustering to brain magnetic resonance image segmentation

Article type: Research Article

Authors: Mújica-Vargas, Dante^{a; *}

Affiliations: Tecnológico Nacional de México/CENIDET, Cuernavaca, Morelos, México

Correspondence: [*] Corresponding author. Dante Mújica-Vargas. E-mail: [email protected].

Abstract: In brain medical imaging, magnetic resonance is an important and effective means to support the computer aided diagnosis. Notwithstanding, inherent conditions such as atypical information, artifacts and vaguely delimited boundaries between existing tissues can hinder the segmentation task. A popular method to carry out this process is through Fuzzy C-Means algorithm, as well as its variants. These include the Intuitionistic Fuzzy C-Means algorithm, which is found suitable for brain magnetic resonance image segmentation, since it incorporates the advantage of intuitionistic fuzzy sets theory to handle the uncertainty. Most clustering algorithms depend of customized hand-crafted features as well as an appropriate initialization process; this last aspect is a mandatory pre-requisite for convergence of the algorithm. In order to develop the brain image segmentation, in this paper we enhance the Intuitionistic Fuzzy C-Means performance by means of Robust Statistics. Explicitly, a non-parametric German-McClure Redescending M-Estimator is used at the initialization and clustering stages, it behaves such as a robust location estimator when the centroid vector is computed, and as a weighting when the membership matrix is updated. The fusion of both paradigms allows us to propose a clustering algorithm that develops efficiently the segmentation of magnetic resonance images, with the important merit of reduce the iteration required to converge. The robustness and effectiveness of this proposal is verified by experiments on simulated and real brain images.

Keywords: Brain MRI image segmentation, intuitionistic fuzzy C-means, German-McClure redescending M-estimator

DOI: 10.3233/JIFS-192005

Journal: Journal of Intelligent & Fuzzy Systems, vol. 39, no. 1, pp. 1097-1108, 2020