As part of a current study evaluating automated surface-based and volume-based brain image registration methods, I have made use of manually defined surface-based and volume-based labels as silver standards. This has given me an opportunity to speak with advocates of surface-based manual labeling (parcellation) who make compelling arguments about the relatively easy learning curve and visual evaluation capabilities when working with surfaces. I would appreciate hearing people's opinions regarding the differences between manual labeling of image volumes and surfaces.

In particular, one person raised a comment about how in the surface representation the concept of the 'dividing plane' would not exist, and would be replaced by some sort of vertex-vertex cuts in the surface view. Does anyone have a position on how to deal with a labeling protocol's dividing planes when labeling brain surfaces?

cheers,
@rno