Hi,
Michael is right in that you cannot compare results in missing slices,
but
FEAT and MELODIC will correctly take this into account. It is better
not
to delete any slices from the data, in case there is movement and one
(or more) slices should get moved into the common, overlapping space.
If you just run the analyses as normal then you get (a) first level
results
in the native space (whatever size that is) and (b) higher level results
in the MNI152 space, but only for the portion of that space where there
was good data from all subjects. Therefore if some subjects have
few slices, this will result in the higher level analysis only showing
results that are common over those slices and excluding the extra
slices from the others, but done after registration. You can see how
many subjects contributed to each voxel's data in the higher level
analysis by looking at the mask images, which are shown in the
output.
All the best,
Mark
On 31 Aug 2010, at 06:32, Michael T Rubens wrote:
> well, obviously you can't directly compare the slices that the
> latter subject is missing. A simple solution is to just get rid of
> the 4 slices to make the dims equal.
>
> cheers,
> Michael
>
> On Mon, Aug 30, 2010 at 6:48 PM, Jingqi Ao <[log in to unmask]> wrote:
> Dear FSL experts:
>
> I have two questions about FSL group analysis:
>
> If volume sizes (function image) of subjects within one group are
> different (for example, one subject is 64x64x35 and another subject
> is 64x64x31), how FSL processes those datasets?
>
> Whether FEAT (GLM) and MELODIC (ICA) process different size of
> volumes in the same way or not?
>
> Thank you very much.
>
> Jingqi
>
>
>
> --
> Research Associate
> Depts. of Neurology & Physiology
> University of California, San Francisco
|