Performance Analysis of Fuzzy C-Means Clustering Methods for MRI Image Segmentation

MRI is one of the most eminent medical imaging techniques and segmentation is a critical stage in investigation of MRI images. FCM is most usually used techniques for image segmentation of medical image applications because of its fuzzy nature, where one pixel can belong to multiple clusters and which lead to better performance than crisp methods. Conventional FCM fail to perform well enough in the presence of noise and intensity inhomogeneity in MRI images. Various FCM variations like BCFCM, PFCM, SFCM, FLICM, MDFCM, FCM_S1, FCM_S2, TEFCM, RFCMK, WIPFCM and KWFLICM, have been proposed to overcome these predicament by using the spatial statistics available in the images. In this paper all these techniques, used for segmentation, are implemented and compared in terms of two classes of cluster validity functions, fuzzy partition and feature structure, on the basis of their performance for the noisy MRI images. All these FCM variants are analyzed in terms of Partition Coefficient, Partition Entropy, Time Complexity and Segmentation Accuracy.