I. Introduction
Cortical parcellation is a key tool that supplies regions of interests (ROIs) in cortical morphological studies. Regarding anatomical/functional associations, the cortex can be subdivided into non-overlapping sub-regions by tracing regional boundaries defined by parcellation protocols [1]–[3]. Nevertheless, earlier cortical parcellation was a tedious and error-prone task as often handcrafted by experts, which requires tremendous human resources. This necessitates an automatic parcellation technique to reduce such extensive manual effort significantly. An initial attempt was to employ label fusion, in which the reference labels are aggregated on a target after surface registration [4]. This approach yet requires non-rigid warping [5]–[7] of all references to a target, and the label fusion needs to be carefully designed to incorporate anatomical variability across the references.