Fig. 3.
Performance statistics. (a) Proportion of runs that resulted in the segmentation that was close to the best (max likelihood) segmentation. (b) Null hypothesis distribution for the number of voxels that showed perfect agreement across all subjects.