Simulation studies of fuzzy clustering in the context of brain magnetic resonance imaging

An important problem in segmentation of brain magnetic resonance images (MRI) is partial volume averaging of different types of tissue. This manifests itself as an overlap of tissue groups in the image histogram space. Fuzzy clustering is an effective technique for separating groups having vague boundaries. The fuzzy C-means (FCM) algorithm has been used for this purpose yet its effectiveness in discerning group differences on the order of a few percent in MRIs is not known. In this report, we compare the effectiveness of the hard C-means, several variants of FCM, and three versions of a possibilistic clustering approach in separating three simulated clusters as boundary overlap is increased.<>