A computer-implemented method is provided for generating patient models from multiple imaging contrast sources which may be, for example from magnetic resonance imaging without computed tomography. The method includes: acquiring multiple sets of image data representing a volume of a patient using magnetic resonance imaging, wherein each set of image data is acquired in a different manner so as to create contrast amongst tissue types of the patient; classifying tissue in each voxel in the volume using the multiple sets of image data as input to a classification algorithm; and generating a patient model for the volume from probability distributions of the classes of tissue as derived from the classification algorithm.