Tumor detection in T1, T2, FLAIR and MPR brain images using a combination of optimization and fuzzy clustering improved by seed‐based region growing algorithm. Issue 1 (21st March 2017)

• Record Type: Journal Article
• Authors: Vishnuvarthanan, G.; Rajasekaran, M. Pallikonda; Vishnuvarthanan, N. Anitha; Prasath, T. Arun; Kannan, M.
• Is Part Of: International journal of imaging systems and technology. Volume 27:Issue 1(2017:Mar.)
• Page Start: 33
• Page End: 45
• Publication Date: 2017-03-21
• Subjects: MPSO‐based FCM -- seed‐based region growing -- tumor identification -- image segmentation -- mean square error -- peak signal to noise ratio -- Jaccard (Tanimoto) index -- dice overlap index; Imaging systems -- Periodicals; Image processing -- Periodicals; 621.367
• DOI: 10.1002/ima.22208
• Languages: English
• ISSNs: 0899-9457

Abstract

ABSTRACT: Tumor and Edema region present in Magnetic Resonance (MR) brain image can be segmented using Optimization and Clustering merged with seed‐based region growing algorithm. The proposed algorithm shows effectiveness in tumor detection in T1 ‐ w, T2 – w, Fluid Attenuated Inversion Recovery and Multiplanar Reconstruction type MR brain images. After an initial level segmentation exhibited by Modified Particle Swarm Optimization (MPSO) and Fuzzy C – Means (FCM) algorithm, the seed points are initialized using the region growing algorithm and based on these seed points; tumor detection in MR brain images is done. The parameters taken for comparison with the conventional techniques are Mean Square Error, Peak Signal to Noise Ratio, Jaccard (Tanimoto) index, Dice Overlap indices and Computational Time. These parameters prove the efficacy of the proposed algorithm. Heterogeneous type tumor regions present in the input MR brain images are segmented using the proposed algorithm. Furthermore, the algorithm shows augmentation in the process of brain tumor identification. Availability of gold standard images has led to the comparison of the suggested algorithm with MPSO‐based FCM and conventional Region Growing algorithm. Also, the algorithm recommended through this research is capable of producing Similarity Index value of 0.96, Overlap Fraction value of 0.97 and Extra Fraction value of 0.05, which are far better than the values articulated by MPSO‐based FCM and Region Growing algorithm. The proposed algorithm favors the segmentation of contrast enhanced images. © 2017 Wiley Periodicals, Inc. Int J Imaging Syst Technol, 27, 33–45, 2017