Hello to the kind folks at FreeSurfer. I'm resending this e-mail as I'm thinking it may not have come through..
We have some remaining questions surround average subjects and group analysis:
1) In visualizing the surface of an average subject (such as when running
QDEC analyses) the brain surface seems to be "incompletely inflated"
compared to viewing a single subjects inflated hemisphere in tksurfer (major
sulci still show depressions on the average subjects surface). What is the
reason/explanation for this?
2) The main discussion I have seen on the validity of using an average
subject vs a custom one (M. Harms and Mr. Fischl) have centered around
registration. If I understand correctly, registration will align various
surfaces (and thereby influence which voxels fall into a particular
ROI/label), but will not substantively change the thickness values of the
surfaces in question. Is this correct? If this is so, it would appear that
the primary issue with using the buckner40 average subject would be the
mapping of vertices into various regions, but that this would NOT interfere
with generating accurate statistical maps of thickness measures for ones own
data set.
I can see that imperfect alignment of vertices across a group may introduce
errors in the averages for a given point. It seems make_average_subject
essentially does precisely this, so accurate registration for a particular
population would maximize the accuracy of any given vertices average. Is
this what making a custom subject atlas seeks to do: the make the
registration between subjects in an average as "linear" as possible to
maximize the accuracy of each averaged vertex?
If anyone could answer this issue for me/point me towards a more complete
resource I would very much appreciate it.
3) There has been an exchange between Graham Wideman, Mr. Fischl and Mr,
Greve about creating one's own average subject. Mr. Greve metioned adding
"hooks" into make_average_subject in order to make the generation of a
custom atlas more accessible (and then using -surfreg-to-subject in the
recon-all stream). Is there any documentation/further explanation of what
these hooks are and how to go about completing this process (can't find
anything on the wiki)?
Many thanks in advance!
-Mischka