Hi Shane,
It would be helpful to see the actual command(s) used to
get from MNI to TLRC space, and in particular, the exact
input and the interpolation method. It does not seem too
odd for significance at the edges of a cluster to erode
during such a transformation, since most neighbors would
be sub-threshold at the edges.
My general recommendation is to process data in the space
of your atlas of choice. Unnecessary resample steps, and
certainly of statistical data, should be avoided.
Regarding MNI vs TLRC spaces, it does not seem appropriate
to compare the MNI template to the TLRC atlas, particularly
since they are different types of objects (template vs.
atlas). While I agree that the TLRC atlas is not very good,
there is nothing wrong with using the TLRC space, i.e.
aligning to some TLRC template and then using a reasonable
atlas in that space.
That a TLRC brain is a little smaller than average is also
not a strong complaint if the new (MNI) template makes
brains larger than average. Neither size seems great, but
a little size scaling is not too important.
The real question is about having a good atlas to use.
Focus on that, and analyze your data in the space of that
atlas, using some appropriate template to get there. For
example, we have the Colin N27 dataset in TLRC, MNI and
MNI anat spaces. Those tend to make nice templates.
Again, it would help to see your commands. But as
suggested above, if you want to use a TLRC atlas, it is
probably better to align to a TLRC template to begin with.
- rick
