Hello,
I think I need to amend my previous message. From looking at the
spm_peb_ppi.m file, I see the PPI.Y contains the original BOLD
eigenvariate, which is why the PPI.Y from the simple deconvolution and
psychophysiologic interaction would be the same.
I also see that PPI.xn contains the deconvolved neural signal. For my
345 scan design, PPI.xn contains 5520 values. How do use these values
to form the PPI and reconvolve with hrf, and ultimately do a
psychophysiological interaction on the event-related design?
Thanks,
Patrick Pruitt
On Tue, Jun 9, 2009 at 11:46 AM, Patrick Pruitt<[log in to unmask]> wrote:
> Hi all,
>
> My lab is interested in performing a PPI analysis on an event-related
> design (345 scans; TR=2s; 6 conditions; 96 trials; each modeled as
> 3s). I have been reading through the relevant posts on this listserv,
> as well as the Gitelman Neuroimage paper. I understand that I need to
> deconvolve the hrf before forming the PPI, then reconvolve, but am not
> clear on how to actually implement these concepts in SPM2.
>
> After setting up first-level analysis and grabbing my VOI, I tried a
> few different things. First I did a "simple deconvolution" using my
> VOI, which gave me a ppi file with a PPI.Y regressor. Then I tried
> doing a "psychophysiologic interaction", using my first-level SPM.mat
> and my VOI. The PPI.Y regressor from the psychophysiologic
> interaction and the PPI.Y regressor from the simple deconvolution
> contain the exact same values, which leads me to believe I'm missing
> something important.
>
> Does anything need to be changed during the first-level analysis or
> does that get modelled as usual? It seems what I'm missing are the
> steps in SPM2 to deconvolve the hrf and use that to form the PPI. I
> assume the reconvolution takes place when setting up the PPI analysis
> in "fMRI design", and selecting "hrf" as basis set under Hemodynamic
> Basis functions.
>
> I've tried to study up on this so as not to waste anybody's time, but
> I think I've hit a wall. Thanks for any help you can provide!
>
> Patrick Pruitt
>
|