Hi, Christian's semi-travelling right now so I'll take a stab:
On 23 Aug 2009, at 11:29, Philipp G. Saemann wrote:
> Dear Christian,
>
> after temporal concatenation group ICA, use of melodic_mix time
> courses to derive
> individual resting networks and voxelwise comparison of these
> between two group we see
> robust focal effects between the conditions.
Note that for some datasets/questions, using melodic_mix is a little
dangerous in this context - see Christian's OHBM abstract on why dual-
regression is preferable to back-projection (the latter is what you
get in melodic_mix). If you'd like to beta-test the dual-regression
approach let me know and we can send you the stuff.
> My concern/hypothesis now is that 'simple' lower general network
> fluctuation amplitudes in
> one of the conditions leads to this (especially as the major nodes
> are affected).
>
> Question: is there a chance to use the melodid_mix file to derive an
> amplitude measure for
> each subject? E. g. by using FFT power? Or is the melodic_mix file
> changed by normalisation
> steps that no absolute amplitude information is contained any more?
The spatial maps are variance normalised in the core ICA in MELODIC -
so _yes_ there is relevant amplitude information in the timecourses.
However see the above comments - using the timecourses output by the
first stage of dual-regression would be safer than using melodic_mix.
Also, in general, the dual-reg script shows how to separate 'shape'
from 'shape-and-amplitude' effects in the spatial maps.
Cheers.
>
> Would you have an alternative suggestion on how to define the
> average percentage BOLD
> fluctuation of a resting network?
>
> Thanks a lot again!
> Philipp
>
---------------------------------------------------------------------------
Stephen M. Smith, Professor of Biomedical Engineering
Associate Director, Oxford University FMRIB Centre
FMRIB, JR Hospital, Headington, Oxford OX3 9DU, UK
+44 (0) 1865 222726 (fax 222717)
[log in to unmask] http://www.fmrib.ox.ac.uk/~steve
---------------------------------------------------------------------------
|