53 Fuzzy C-Means Clustering with GSO Based Centroid Initialization for Brain Tissue Segmentation in MRI Head Scans T. Kalaiselvi, P. Nagaraja and Z. Abdul Basith Department of Computer Science and Applications, The Gandhigram Rural Institute – Deemed University, Gandhigram, Tamil Nadu, India kalaiselvi.gri@gmail.com, pnagaraja02@gmail.com, abdbasith93@gmail.com Abstract The proposed work is a glowworm swarm optimization (GSO) based centroid initialization for image segmentation using the fuzzy c-means clustering (FCM). FCM is one such soft segmentation technique applicable for MRI brain tissue segmentation. In the conventional FCM, the centroids are initialized randomly. This leads to increase the processing time to reach the optimal solution. In order to accelerate the segmentation process, GSO is used to initialize the centroids of required clusters. The quantitative measures of results were compared using the metrics are number of iterations and processing time. The number of iterations and processing of proposed method take minimum value while compared to conventional FCM. The proposed method is very efficient and faster than conventional FCM for brain tissue segmentation from T2-weighted head scans. Keyword: Brain Tissue Segmentation, Clustering, Fuzzy c-means, Optimization. 1. Introduction Image segmentation plays an important role in most of image analyzing tasks. It is the process of partitioning a digital image into non-overlapped homogeneous regions with respect to some characteristics, such as gray value, motion, texture, etc [1] [2]. The diagnostic capability of medical experts improved significantly with the arrival of medical imaging techniques such as computed tomography (CT), positron emission tomography (PET), magnetic resonance (MR) images and single photon emission computed tomography (SPECT). An MRI image provides information about the human soft tissue anatomy. MRI head scans are classified into two types: classifying tissues, anatomical structures. It comprised into different tissue classes which contain four major regions, namely gray matter (GM), white matter (WM), cerebrospinal fluid (CSF), and background (BG) [3] [4]. In brain diagnostic system, segmentation is essential to study many brain disorders [5]. Segmentation techniques are classified into region-based methods [6] [7] and edge- based methods [8]. The proposed work focused on the region based approach using FCM clustering (soft clustering), instead of hard clustering strategies. FCM clustering algorithm is a soft segmentation technique that preserves more information from the given input images. It is proved to be the best method for the anisotropic nature of volumes. FCM is more appropriate for these types of volumes and abnormal images while compared to the k-means and expectation-maximization clustering [9]. The blind application of the conventional FCM algorithm is very sensitive to noise and the efficiency of FCM highly depends on the initialization step, because the iterative process easily falls into a locally optimal solution for image segmentation. Clustering with swarm-based method is emerging as an alternative to more conventional clustering methods [10]. Genetic algorithm (GA), ant colony optimization (ACO), simulated annealing (SA), tabu search (TS), particle swarm optimization (PSO) and their hybrid methods are widely applied to the cluster analysis, and highly attractive in many important clustering applications [11-14]. Glowworm swarm optimization (GSO) is a new swarm intelligence algorithm, was proposed by Krishnanand and Ghose in 2005 [15]. This algorithm was inspired the phenomenon that the glow attracts mates. The brighter the glow, more is the attraction. Each agent in the swarm decides its direction of movement by the strength of the signal picked up from its neighbors. Therefore, we use the glowworm metaphor to represent the underlying principles of our optimization approach. Therefore, it can avoid the influence of the initial condition. There are minimum papers to the clustering analysis based on the GSO algorithm in current literature. Aljarah and Ludwig proposed a GSO algorithm, for formulating the clustering problem as a multimodal optimization problem to extract the optimal centroids based on glowworms’ movement [16]. This proposed GSO algorithm for clustering can discover the numbers of clusters without needing to provide the number in advance. Experimental results of GSO based clustering used several datasets namely iris, ecoli, glass, balance, seed and two artificial data sets namely: mouse and vary Computational Methods, Communication Techniques and Informatics ISBN: 978-81-933316-1-3 53