Ah, are you running the latest melodic? I think Christian had fixed that a while ago (am i right, Christian?).
Aside from that, you are absdolutely right in using a fake mxing matrix.
Von: FSL - FMRIB's Software Library [[log in to unmask]] im Auftrag von Gael Varoquaux [[log in to unmask]]
Gesendet: Montag, 3. August 2009 11:04
An: [log in to unmask]
Betreff: Re: [FSL] Melodic mixture model and thresholding
On Mon, Aug 03, 2009 at 09:27:24AM +0200, Gael Varoquaux wrote:
> On Sat, Aug 01, 2009 at 12:35:47PM +0100, Steve Smith wrote:
> > Hi - close - and sorry the Usage isn't clearer:
> > Yes you need the -i and --mix options to be set, as you have done - but
> > they are then ignored (as you probably guessed). Then you need to add
> > the report option and may as well add verbose:
> > -v --report
> > That should do the trick,
> So, for fututre reference, the trick was to create a fake mixing matrix.
> I created an identity matrix of the right size, that I saved in
> 'fake_mixing.dat', then I could run:
> melodic -i canica_ic.nii.gz --ICs=canica_ic --mix=fake_mixing.dat -o melodic_mix -v --report
> And now, I is nicely applying mixing model to my ICs.
I am figthing the fact that Melodic reorders my ICs. I have to do
matching with the non-thresholded ones to figure out which one is which,
and my simple heuristic fails once in a while. Any idea how I could for
melodic not to reorder?