Brain MR image segmentation using differential evolution guided by RBF-kernel based FCM

Fuzzy C-Means and Possibilistic C-Means are two most used algorithms in the computer aided diagnosis systems(CAD) for image segmentation. Due to simple implementation, fast convergence and unsupervised procedure, these algorithms have been become widespread and the most desirable methods to employ in disparate image segmentation problems. However, emerging new imaging devices and different quality of images, unveil disability of these methods. Falling in local optima and high sensitivity to noise and outliers, have confronted the scholars with the new challenges in recent decades. This paper by employing a meta-heuristic algorithm, Differential Evolution, and kernel-based inner product norm metric has taken a new direction yet simple to overcome to initial configuration sensitivity and premature convergence in the conventional FCM and PCM. Proposed method succeeded to get to an accuracy about 94% true segmentation of brain tissue in presence of noise and intensity inhomogeneity.