Dear Jamil,
What you describe is a poor man's way of doing [functional] connectivity
analyses. I myself have gone this way once [before knowing better], although
it might have been okay for the specific regions I have chosen.
The way I did it was to simply export the time course of the region of
interest which I then fed into a new design - using the same images - as a
userspecified regressor. You can then do a +1 or -1 T-test and usually find
the global max to be in the source region [ROI].
The danger is that depending on the ROI you might actually get a lot of
noise for example from respiration or cardiac effects such as in brain
stem-near regions, and then get regions which are affected by that noise in
a similar way as your ROI (an example can be seen in a recent Brain paper
[hypothalamus]).
A recent paper by Fox and Snyder 2005 did a slightly more clever thing than
what I have described above: aside from the time course of interest they
also regressed out a time course taken from CSF and white-matter (if I
remember correctly), as effects of no interest if you wish (this has been
discussed on the list before [look for posting by myself, e.g.]).
Ideally, you should model cardiac noise (e.g. if your scanner can record
ECG) and respiration (e.g. respiration belt) or pursue an entirely different
approach to connectivity (as has been suggested in another reply).
If you DO wish to go ahead, Greicius et al. (PNAS 2003) and Fransson (HBM
2005) would be other examples.
Hope this helps a little,
Helmut Laufs
----- Original Message -----
From: "Jamil Zaki" <[log in to unmask]>
Sent: Monday, February 27, 2006 4:14 AM
Subject: Functional Connectivity based on Time Course Data
> Hi All --
>
> I've seen different groups conducting connectivity analyses by extracting
> time course data from particular regions and using these as "seeds" to
> assess covariance between extracted time series and other regions. Is
> there
> a methodology for doing this type of analysis in SPM?
>
> Thanks for all your help and best regards,
>
> Jamil Zaki
>
|