Fig. 3.

Summary of the parcellation results of the first (A) and second (B) “pass” of the algorithm. The depicted results constitute winner-takes-all maps. The modules constituting each cluster are forming a probabilistic map (see Fig. 4). These probabilistic maps are combined to produce the depicted winner-takes-all maps by assigning each voxel a unique integer corresponding to the cluster exhibiting the highest probability in this voxel. Subsequently, each cluster is coded with a unique color and named arbitrarily C1, C2, …, C19. This cluster-wise color scheme is also followed in Figs. 4 and 5. Spatial location of the clusters dictates their color “family”: dorsal motor/premotor clusters are color coded with shades of green, ventral premotor clusters with brown/orange, and prearcuate clusters with blue/violet. Borders around clusters in A indicate those that were further subdivided on the basis of the second-pass results depicted in B (see results).