Hi
melodic represents the 20 'interesting' components as the set of IC
maps and associated time courses. You could load these two things
into e.g. matlab and re-calculate the their outer product to get a 4D
file which only contains the 20 components. Note that (i) you'd need
to run melodic with the --Ostats folder and then use melodic_oIC
together with melodic_mix or melodic_pca together with
melodic_dewhite (ii) this new outer product will have the time-course
mean removed at every voxel.
I'm not sure what you're setting out to do but just in case you're
planning to feed this 4D file into FEAT: the stats that feat then
spits out will _not_ be valid, i.e. t/z/p values become quite
meaningless, given that in the process of throwing out 60 of the
directions you're messing with the residual noise and the degrees-of-
freedom...
hope this helps
christian
On 22 Nov 2006, at 16:25, Jason Steffener wrote:
> Hello.
> I am looking into using Melodic on my data. Can it be
> used to just remove the "small" components and keep
> just the number of components it estimates from the
> data?
> If I have 80 time points the maximum number of
> components it can find is 80. If Melodic finds that 20
> of these are of interest can I just keep these 20 and
> remove the remaining 60 from my data?
>
> Thank you,
> Jason.
--
Christian F. Beckmann
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 222551 Fax: +44(0)1865 222717
|