Fuzzy adaptive level set algorithm for brain tissue segmentation

This paper presents an image segmentation method based on fuzzy clustering and level set methods. The MRI image is firstly segmented with the fuzzy clustering method, and then the resulting fuzzy memberships are used to generate a new constraint function which guides the level-set curves evolution. Thanks to the new constraint term, the presented method is able to adaptively determine the directions of the curves evolution without any manual intervention. The new method does not depend excessively on the gradient information as the geodesic active contour method does; therefore it is less sensitive to the noise. The region information provided by the first step of the algorithm allows to improve the segmentation accuracy of the brain tissues. The experimental quantitative and qualitative analyses indicate the capability of the proposed method to segment the brain tissues with high accuracy comparing with the fuzzy clustering method.