Hi
On 16 Apr 2008, at 01:00, lingling pu wrote:
> Dr. Beckmann,
>
> I am trying to use the method you suggested in the previous message.
> I just want
> to make sure that I use the correct commands.
>
> So I have a 4D file which is a combination of selected IC maps in
> melodic_IC.
> Then I use:
>
> fslmaths 4Dfile -Tmean 4D_mean
> fslmaths 4Dfile -Tstd 4D_std
> fslmaths 4D_mean -div -Tstd 4D_std 4D_normalize
> mm --mask=mask --sdf=4D_normalize
>
That looks OK
> Another question is if I can not identify single task-related IC
> maps in melodic_IC, instead,
> I find several maps that might be related. Does it make sense to use
> the above normalizing method
> for each individual and then do a 2-level analysis on these
> normalized files from all the subjects?
>
No, that does not make sense - you can use the method described above
to do a pseudo-mixed-effects analysis of what you believe are
associated spatial maps across a population. If within a single run
you end up having multiple maps representing the signal then there
typically is a reason, e.g. consistent differences in the haemodynamic
response between voxels ending up in the different spatial maps. If
your study is a task activation study then you might want to try
tensor-ICA instead - if it's a resting experiment I suggest you use
the concatenation approach or the GIFT toolbox.
hth
christian
> Thanks,
> Lingling
>
>
> ----- Original Message ----
> From: Christian F. Beckmann <[log in to unmask]>
> To: [log in to unmask]
> Sent: Friday, April 11, 2008 6:09:06 AM
> Subject: Re: [FSL] question of the group-level analysis on the ICs
> from pICA
>
> Hi,
>
> you'd want to use the correct 3D volumes contained in the different
> melodic_IC files and combine these, e.g. using pseudo-mixed effects by
> averaging and dividing by their standard deviation. Use fslroi ti
> extract the relevant volumes first, then combine the different default
> mode maps using fslmarge into a single 4D file. Finally, use fslmaths
> to average, calculate the standard deviation and combine these two
> quantities into a new map. You can threshold this map using mm, the
> mixture-modeling tool.
> hope this helps
> Christian
>
>
>
>
> On 11 Apr 2008, at 04:39, Xi-Nian Zuo wrote:
> > Hi, FSLers
> >
> > I have a question about the second-level analysis of components
> > generated by
> > MELODIC. I performed pICA on individual fmri data and selected the
> > default
> > resting-state networks for everyone. What kind of data should be
> > used for
> > the following group analysis procedure? melodic_ICs.nii.gz or
> > thresh_zstat.nii.gz? i.e., original score or z-statistics value in
> the
> > spatial component? Any suggestion on this is appreciated very much.
> >
> > Best regards,
> > Xi-Nian
>
>
> between 0000-00-00 and 9999-99-99
|