Hi,
Christian makes a very good point...ARRS (always remember the
relative scaling). There are several ways to do this in addition to
what Christian suggested. One could determing the scaling using a GLM
analysis on the ICA time courses. A perhaps better approach would be to
model the ICA time courses as zero-mean and standardize the time course
amplitudes (thus scaling the spatial map amplitudes). In both of the
latter cases, the ICA voxels are treated as something analogous to MR
signal strength, a physiologically meaningful thing, the averaging of
which makes sense.
Cheers,
Vince
> -----Original Message-----
> From: FSL - FMRIB's Software Library
> [mailto:[log in to unmask]] On Behalf Of Christian Beckmann
> Sent: Monday, December 30, 2002 3:13 AM
> To: [log in to unmask]
> Subject: Re: [FSL] ICA
>
>
> Hi all,
>
> just a quick note: It is important to perform the averaging
> on statistical
> images, i.e. posterior probability plots or Z-transformed IC
> maps. Raw IC
> maps are no good unless you want to compare apples with
> oranges. You'll
> need to be sure that the reltive scaling is meaningfull....
>
> cheers
> christian
>
> Christian F. Beckmann
> Address: Oxford University Centre for Functional
> Magnetic Resonance Imaging of the Brain,
> John Radcliffe Hospital, Headington, Oxford OX3 9DU, UK
> Email: [log in to unmask]
> http://www.fmrib.ox.ac.uk/~beckmann/
> Phone: +44(0)1865 222782 Fax: +44(0)1865 222717 Mob:
> +44(0)7811 189123
>
> On Fri, 20 Dec 2002, Vince Calhoun wrote:
>
> > hi appu....
> > i agree with christian...just an add on...we've also tried the
> > approach similar to what christian is suggesting (running
> separate ica
> > estimations)... except not in fsl...see: calhoun et al,
> neuroimage 14(5)
> > 2001 pp 1080-1088. hope this helps.
> >
> > cheers,
> > vince
> >
> > Christian Beckmann wrote:
> >
|