Brain image segmentation using kernelized fuzzy c-means clustering

Segmentation of brain tissues is an important but inherently challenging task. Different brain tissues may have similar grayscale values and the intensity of a brain tissue may be confused with that of another one. The study accordingly develops a KFCM method based on kernelized fuzzy c-means clustering with ICA analysis for extracting regions of interest in MRI brain images. Through ICA, three independent components are then extracted from multimodal medical images. Since MRI signals can be regarded as a combination of the signals from brain matters, ICA can be used for contrast enhancement of MRI images. Then, the three independent components are utilized as inputs by FCM algorithm to extract different brain tissues. Relying on the decomposition of a multivariate signal into independent non-Gaussian components, the proposed method can achieve greater reliability in both theory and practice. Experiments show that the proposed method can accurately extract the complicated shapes of brain tissues and remains robust against various types of noise.