Unfortunately, it is worse than that - even ignoring the individual
variability issue (which should not be ignored), a small change in MNI
coordinate can jump from one bank of a sulcus to the other, so having only
the center coordinate of a cluster makes this a badly posed problem (and
thus explains why peak or center of gravity MNI coordinate reporting is not
nearly as useful as actual data files).
Coordinates of each vertex are contained in .surf.gii files, but different
files can represent different depths of cortex (gray/white boundary
("white"), csf/gray boundary ("pial"), halfway between ("midthickness),
etc), or other things entirely (inflated, sphere). What you could do to
get an idea of the problem, is to take a group of subjects, get the vertex
coordinates of the midthickness surface, and compute the euclidean distance
to each of your cluster centers. If you average these across subjects, I
suspect you will generally end up with 2 similarly low-distance spots for
each cluster center, on opposite sides of a sulcus (for consistent sulci,
anyway - if it is in or near a high-folding-variability region, one of the
two low spots may get blurred out of existence (or the two may be blurred
into one larger spot) because the folding patterns don't align, but it is
the folding patterns that determine where the MNI coordinate is close to
cortex). To make it easier to see this issue, you could transform these
distances into a gaussian kernel (matching the size of the original cluster
if you know it, or the amount of smoothing that was used, or...) and then
average that across subjects.
For a less-rigorous answer, there is a -surface-closest-vertex command in
wb_command that will simply find the closest vertex on the given surface
(you will need to separate your right and left coordinates), but as this
demonstration should show, MNI coordinate data is generally ambiguous to
begin with. Also note that group-average surfaces are missing a lot of
folding detail that individual subjects have, and while that may
incidentally make "closest vertex" more stable, it doesn't imply that it
makes the answer more correct.
Tim
On Thu, Mar 1, 2018 at 9:32 AM, Manasij Venkatesh <mana...@umd.edu> wrote:
> Hi,
>
> I have a set of ROI cluster center coordinates in MNI space. My goal is to
> create similar ROI clusters to use with HCP data. I understand there's no
> true equivalent in terms of surface grayordinates but what would be the
> best way to find their approximate position on the surface? Is this the
> information contained in the surf.gii files? Please let me know.
>
> Sincerely,
> Manasij
>
> _______________________________________________
> HCP-Users mailing list
> HCP-Users@humanconnectome.org
> http://lists.humanconnectome.org/mailman/listinfo/hcp-users
>
_______________________________________________
HCP-Users mailing list
HCP-Users@humanconnectome.org
http://lists.humanconnectome.org/mailman/listinfo/hcp-users
