Segmenting a digitized image includes providing a diffusion tensor field image, partitioning the image into 2 regions, each point being assigned to one of the regions, associating a level-set function with each region, the level-set function having a negative value in one region and a positive value in the other region, calculating a mean value of the diffusion tensor field over each of 2 regions, initializing an energy defined as a functional of level-set functions and diffusion tensor field, changing the region membership of each point in the image if the energy functional value decreases as a result of region membership change and updating said mean value of said diffusion tensor field over each of 2 regions, and obtaining a segmentation of the image when the magnitude of the change of the energy function value resulting from changing the region membership of a point is less then a predetermined threshold.