Surface-deformation techniques had been commonly utilized to attempt to generate accurate models of various boundaries from MRI (Davatzikos and Prince, 1995; Davatzikos et al., 1996; MacDonald et al., 2000). Unfortunately, it was difficult to directly use these techniques to construct surface representations with the requisite degree of accuracy for a number of reasons. First, isointensity surfaces, which assume that the MRI intensity of the gray/white and pial surface interfaces are constant over space, do not generate accurate enough surfaces. This is due to variance in the histological makeup of cortical gray matter and the subjacent white matter as well as acquisition artifacts such as variable RF penetration, dielectric resonance and nonuniform receive coil sensitivity profiles that all conspire to cause the MR intensity of a given tissue type to vary over space. For example, cortical gray matter in motor cortex is significantly brighter than in frontal cortex, due to its higher myelin content (which can be quantified with T1 mapping in MRI (Fischl et al., 2004a)).
