A fast spatial constrained fuzzy kernel clustering algorithm for MRI brain image segmentation
Liang Liao, Lin Tu-sheng · 2007
A fast kernel clustering algorithm incorporating spatial constraints is proposed for segmenting MRI (magnetic resonance imaging) brain images. The algorithm called FKFCM (fast kernel based fuzzy c-means clustering) is implemented by using a kernel technique, which can improve the separability of clustered data, for segmenting MRI images. The kernel technique implicitly maps input data to a higher dimensional kernel space and therefore transforms a nonlinear segmenting problem to a linear one. Because the clustering performed in a kernel space is generally computational consuming, a fast clustering scheme is implemented to speed up the computation. Experimental results on synthetic image, digital phantom and real clinical data indicate the proposed algorithm is effective for segmenting MRI images corrupted by noise and intensity inhomogeneity, and usually outperforms the corresponding conventional methods.