Modified fuzzy c -means algorithm for segmentation of T1–T2-weighted brain MRI

• Published in: Journal of computational and applied mathematics Vol. 235; no. 6; pp. 1578 - 1586
• Main Authors: Ramathilagam, S., Pandiyarajan, R., Sathya, A., Devi, R., Kannan, S.R.
• Format: Journal Article
• Language: English
• Published: Kidlington Elsevier B.V 15.01.2011; Elsevier
• Subjects: Brain MRI; Clustering; Data analysis; Exact sciences and technology; FCM; Mathematical analysis; Mathematics; Modified FCM; Numerical analysis; Numerical analysis. Scientific computation; Partial differential equations; Sciences and techniques of general use; Segmentation; FCM; Data analysis; Modified FCM; Brain MRI; Segmentation; Clustering; Iteration; Algorithm; Implementation; Experimental result; Numerical analysis; Applied mathematics; Resonance; Comparative study
• ISSN: 0377-0427; 1879-1778
• DOI: 10.1016/j.cam.2010.08.033

In the past decades, fuzzy segmentation methods, especially the fuzzy c -means algorithms, have been widely used in the segmentation of brain medical images, because they can preserve more information from the original image than other segmentation methods. This paper introduces a modified robust fuzzy c -means algorithm with special weighted bias estimation for segmentation of brain magnetic resonance image. In general, the intensity inhomogeneities are endorsed to imperfections in the radio-frequency coils or to the problems connected with the image acquisition. The intensity inhomogeneities are seriously affecting the segmentation process when computer assisted segmentation algorithms used in differentiating borders between the tissues in medical images. The proposed algorithm of this paper is considered the fact in the standard fuzzy c -means with no spatial context information, which makes it sensitive to noise. Hence the proposed method is capable to deal with the intensity inhomogeneities and noised image effectively. Further, to reduce the number of iterations, the proposed modified robust algorithm initializes the centroid using dist-max initialization algorithm before the execution of algorithm iteratively. Experimental results carried out through the comparative studies of segmentation results of the proposed method and the segmentation results of other existed algorithms when these algorithms implemented on brain magnetic resonance image and benchmark lung cancer dataset. The superiority of the proposed method has shown from the experimental results.