Dear Anna,
First of all: YES it would be very nice if these results could be
replicated. It's been a few weeks since I have read the paper, but I
remember that their analysis was not quite straight forward.
A simple approach could be do extract a time series, say from the precuneus
and use that as a regressor in SPM. It's the poor man's way of doing a
functional connectivity analysis (see Greicius et al.PNAS 2003). Now - how
to obtain the timecourse in the first place? The raw data is probably not
what you want. So you could run an SPM analysis incorporating motion,
respiration, cardiac, ... regressors (e.g. see Lund 2005) to 'clean' the
data somewhat from factors that would lead to [false] correlation between
voxels. Of course, Fox et al. would have had to monitor all these.
Once you have done this, you could work on the 'residuals', i.e. write out
at new time series after the effects of the mentioned 'noise makers' have
been regressed out (do not ask me how exactly to do that in terms of coding,
but using the SPM machinery it should not be a long script). This you could
then treat as your new raw data and - using any of the ROI tools, e.g.
MarsBar - extract a 'raw' voxel time course and feed this back into an SPM
analysis with just that as a regressor.
Maybe, there are more sophisticated approaches, but for a simple mind (like
me), this would be an analysis I could follow.
By the way: if you read carefully, you will find that they have EEG data
available for that data set, which would be most interesting to have. At
least, they should give you the ECG trace which probably was recorded.
Depending on how that was done (bipolar, e.g.) you might be able to even see
respiration as slow "drifts" of that channel.
Good luck, I would be very happy to hear of any results/progress.
Best wishes,
Helmut
----- Original Message -----
From: "Maria Densmore" <[log in to unmask]>
Sent: Wednesday, October 12, 2005 9:04 PM
Subject: Resting Brain time series correlations
> Dear spm group,
>
> I've been asked to analyze a data set that was scanned with the same
> parameters
> as the paper: Fox et al "The human brain is intrisically organized into
> dynamic, anticorrelated functional networks", PNAS July5,2005
>
> This data set was not analyzed using SPM but seems to correlate time
> series data
> for the voxels or regions of interest during a resting state.
>
>
> My dilemma is:
>
> 1) I don't quite understand how to set my design matrix since there is no
> baseline condition to subtract the task from. What I do have is, one task
> (eyes
> closed)that continues throughout the entire 110 volumes acquired.
>
> 2) I'm also not sure if this translates to a PPI analysis in SPM or if
> it's
> something I won't be able to replicate using SPM.
>
>
> Any comments would be greatly appreciated,
> Thanks for your attention and time,
>
> Maria
>
> _____________________________________
> M. Densmore, Image Analyst
> Lawson Research Institute
> St. Joseph's Health Centre
> 268 Grosvenor Street
> London, Ont. N6A 4V2
>
> [log in to unmask]
> office: (519) 646 6100 ext 64864
> fax: (519) 646 6399
> _____________________________________
>
|