I have been doing my analysis using Feat and then displaying results
on inflated brains (generated using Freesurfer). I've run into what I
presume are very old problems -
First, because Feat smoothing is based on voxels in a 3d volume,
activation is spread across abutting grey matter edges. This leads to
the problem of trying to separate true and "ghost'' activation when
the results are displayed on a surface (I appreciate I can potentially
disambiguate the two by doing a separate run with no smoothing but
then that would require me to tamper with the smoothed images to
remove the "ghost" activation, something I'd rather not do).
Second, I'm interested in activation in some fairly small anatomical
areas (take MT+, the motion complex, as an example) and the location
of these areas vary relative to the rest of the brain. So when I do a
group analysis the same anatomical areas are not superimposed very
well and hence I don't find the activation I expect from inspection of
the individual analyses.
Now, one way round these two problems would be to switch from FSL to
Freesurfer for analysis. Freesurfer will do surface based smoothing
which should remove the ghosting problem. It also allows surface maps
of individual brains to be morphed to superimpose anatomical locations
in a group analysis. However, it would be a lot of effort to learn
Freesurfer and redo all my analyses, so, before doing that:
1. Are there some FSL solutions to these problems? For example, at
least in my head, it seems like it should be reasonably easy to do
some "smart" volume based smoothing that does not include voxels that
are separated by something other than grey matter. Has anyone written
a simple plugin "smart smoother" for Feat? Also, the new non-linear
registration might be expected to solve some of the registration
problems , maybe I just need to plug in the right parameters? In
which case, what are they?
2. If I need to go the Freesurfer route, does anyone have any
guidance re how to do this as easily, preferably sticking with Feat as
much as possible? I did find this
but the section on doing group based surface analyses first notes "FSL
tools have not been extensively tested on surface-based data" which
seems to be pushing me away from Feat, and then after a rather terse
set of Freesurfer instructions suggests consulting the Freesurfer
wiki. I interpret this as "now go and learn Freesurfer"!
Any guidance would be much appreciated...