Hi Christian,
Thanks for your swift reply! Indeed, all the components are inside the
mask saved in the groupmelodic.ica folder. However, this mask seems to
be based on the structural images from each subject because the entire
standard brain falls within this mask. This could explain why the
spatial pattern of several components exceeded the boundaries of some of
the subjects' functional data... Another mask is saved in the parent
directory, though, which does seem to be the mean mask based on the
functional images (top of the brain is not included as expected). Why
isn't this mask being used, rather than the one in groupmelodic.ica?
Thanks!
Ilya.
-----Original Message-----
From: FSL - FMRIB's Software Library [mailto:[log in to unmask]] On
Behalf Of Christian F. Beckmann
Sent: Tuesday, November 13, 2007 5:48 PM
To: [log in to unmask]
Subject: Re: [FSL] mean group mask in melodic
Hi
the mask melodic uses does get saved in the groupmelodic.ica folder -
there certainly should not be any voxels within any of the components
outside this mask. Wrt fsl_glm : just use the -c option and supply
a .con file generated by the Feat or the Glm setup program. Have
forgotten to list -c in the command line help, will be added to the
next release
hth
christian
On 13 Nov 2007, at 16:23, Ilya Veer wrote:
> Hi,
>
> I have a question regarding multi-session temporal concatenation in
> Melodic. The top of the brain is not scanned in about one third of
> my subjects. A mean mask, including voxels that are present in all
> subjects, is correctly generated by melodic. However, several
> components found in the group ica do show a spatial pattern
> exceeding this mask. Does melodic not take the mask information into
> account? Should I therefore only look at components with spatial
> patterns within the borders of the mean mask or maybe use fslroi
> beforehand to exclude these voxels? Any suggestions would be greatly
> appreciated.
>
> Second, I tried the fsl_glm command to redo post-stats on the
> melodic subject modes (melodic_Smodes) using another design matrix
> and other contrasts. However, I could not find a way to include a
> contrast file. How does one do this using the fsl_glm command?
>
> Many thanks in advance!
>
> Ilya Veer
> Leiden Institute for Brain and Cognition
>
|