there is another problem that was previously discussed on the list - when
because all out-of-brain voxels are coded as NaN, the values in those voxels
neighboring these outer voxels become NaN after interpolation. I wrote a
program that will change those NaN's just outside the brain to zeros (think
of it as a crunchy shell of zeros for the brain) - I think I posted it in a
previous message, but if you can't find it there I can dig it up and mail it
to you.
cheers
russ
andreas joachim bartsch wrote:
> Dear group,
> it has been suggested previously that calculating non-normalized
> .con-Images and normalizing them by the parameters derived from the
> normalization of, e.g. coregistered anatomicals may suffice to finally
> use the resulting n-.con-Images in a rfx-type of analysis. The advantage
> would be less computing and storage. However, as I understand
> normalization involves smoothing which will affect the .con's
> differently in different subjects, right? Any opinion on the impetus
> that may exert, any advice? MAny regards- andreas
|