Dear Jack,
I can answer some of your questions, though Christian and others will be
better placed to answer your questions about effective connectivity.
|The way I figure it is that practically the temporal evolution (fMRI) of a
|highly significant voxel (or any voxel) is used as covariate of interest.
|Is this true ?
The critical aspect is to examine the covariation of the voxel of
interest with the rest of the brain under two different experimental
conditions. It is the difference in the covariation of that voxel
with other brain areas under the different experimental conditions that
constitutes the 'psychophysiological interaction'.
|1- What is the easiest way to get this vector using SPM99b ?
Go to the results section and plot a voxel of interest. In the Matlab
workspace window the variable 'y' represents the time series of activity
in that voxel, corrected for the effects of no interest. Sarah Blakemore
and John Morris wrote a 'how to' faq for constructing a
psychophysiological interaction which is very helpful; it's archived at
http://www.mailbase.ac.uk/lists/spm/1998-11/0094.html
|2- Is this vector supposed to be corrected (whipping out the part of
|covariate of no interest) => how to do it (without programming a new matlab
|function :)) ?
Yes, and mean corrected (see recipe above for 'how to')
|3- Wouldn't it be more interesting to use a cluster ? (may be a good
|smoothing can do the same)
Karl's spm_regions.m function archived at
http://www.mailbase.ac.uk/lists/spm/1999-07/0029.html will allow
extraction of an ROI for this purpose (and calculate the first
eigenvariate of the ROI which may be even better)
|4- One can suspect a functional connectivity to be displayed within a task,
|but not with an other. Let say you which to see the effect of one factor on
|a fMRI acquisition with 3 conditions : Rest, Active1, Active2. How will you
|put your vector in ?
My apologies if I have misunderstood, but I think the question you
are interested in is 'which brain regions correlate more with activity
in ROI X under task Active1 compared to task Active2'. To estimate this
you need at least three vectors (regressors). First, the physiological
activity in ROI X. Second, a vector representing the psychological
difference of interest, which in this case would be a vector set to 1
during Active1 scans, -1 during Active2 scans, and 0 elsewhere. Third, a
regressor representing the interaction of the psychological and
physiological factors (just multiply the first two vectors together).
Everything should be mean corrected and other experimental factors of no
interest modelled in the usual way.
The third vector represents the 'psychophysiological interaction' of
interest, the other two being confounds. The SPM{t} of this third
regressor will represent areas showing a significant 'psychophysiological'
interaction.
|5- The way I figure it could only be name functional connectivity, not
|effective, since no model is imposed. True ? Thus to do effective connec
|tivity, one should plot the intensity of one voxel against the one of
|another that have to be specified (another activated regions for example).
I would defer to others on the subject of effective connectivity. My
understanding is that by imposing a model, an analysis of effective
connectivity allows an estimation of the relative contribution of multiple
other ROIs to evoked activity in a voxel of interest. In the analyses
discussed here, only the contribution of a single physiological source of
activity (in ROI X) is considered.
best wishes,
Geraint
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
|