Hi - just a quick point to add to Christian's detailed answer - if you put
a tXX.txt file from the MELODIC report subdirectory into FEAT as an EV,
don't forget to turn off Convolution and temporal filtering (and probably
also temporal derivative) for that EV (under Advanced).
Thanks, Steve.
Stephen M. Smith
Head of Image Analysis, FMRIB
Oxford University Centre for Functional MRI of the Brain
John Radcliffe Hospital, Headington, Oxford OX3 9DU, UK
+44 (0) 1865 222726 (fax 222717)
[log in to unmask] http://www.fmrib.ox.ac.uk/~steve
On Wed, 12 Sep 2001, Christian Beckmann wrote:
> Hi Marnie
>
> the easiest way of doing this is to use those MELODIC time courses that
> you think are attributable to artificial noise as 'regressors of no
> interest' in a standard GLM. To do this, you can use FEAT and add
> additional evs - using 1 column format for the ev and specify the file
> name of the component time course from the MELODIC report directory
> (someting like tXX.txt).
>
> You should add these regressors as the first ones in the model, followed
> by the regressors of interest. The latter set of ti,e courses should
> also be orthogonalised with respect to the first set (again you can do
> this easily by clicking the appropriate checkboxes inside the FEAT model
> setup GUI). I have been asked a similar question by various other people
> an the results appear to be very good.
>
> An alternative way of removing noise components is to use the fact that
> the model is additive. Thus, you can try to generate a 4D representation
> of the artificial noise and subtract this from the original data. This,
> however, is far more complicated and the sequence of pre-processing
> steps needs to be taken into account. I only have research versions of
> melodic that contain this feature but hope to be able to include this
> into the next release version. For the time being, the first option is
> easy to try and appears to work remarkably well on our simulations and
> on real data. There was a poster on this at last years HBM (in Sixth
> Int. Conf. on Functional Mapping of the Human Brain, page 614) that
> deals with artefact detection in FMRI (the a4 version of the poster is
> on my homepage)
>
> hope this helps
>
> --best
> Christian
>
> Marnie Shaw wrote:
> >
> > Hi Christian,
> > We have been testing MELODIC on our simulated data set and found it
> > performs very well. After observing how well it separates the noise
> > components from the signal, we would like to try to remove those
> > components from the data and re-analyse it. Is there currently any way to
> > do this in MELODIC?
> > thanks,
> > Marnie
> >
> > PS> Many thanks to you kindly folk at fmrib for making your fsl software
> > available to everyone!
> >
> > -------------------------
> > Marnie Shaw
> > Howard Florey Institute
> > University of Melbourne, 3010
> > Victoria, Australia
> > Telephone (+61) (03) 8344 3985
> > Fax (+61) (03) 93481707
> > -------------------------
>
> --
> 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 222704 Fax: +44(0)1865 222717 Mob: +44(0)7980
> 681852
>
|