Dear Philip,
Have a look at how psycho-physiologic interactions (PPI) are implemented
in SPM:
https://www.fil.ion.ucl.ac.uk/spm/doc/spm12_manual.pdf#Chap:data:ppi
https://github.com/spm/spm12/blob/master/spm_peb_ppi.m#L443
The time series from ICA would here replace the average time course from
a ROI.
In the PPI tutorial in the SPM manual, the data from multiple sessions
were concatenated (and session specific effects manually modelled) but I
presume you could still model a session specific PPI then compute the
session average using a contrast.
Best regards,
Guillaume.
On 10/01/2022 22:11, Philip Deming wrote:
> ⚠ Caution: External sender
>
>
> Hello,
>
> I'm looking to identify regions within three large-scale networks
> (default mode, frontoparietal, salience) that were significantly related
> to stimulus onset in a task. One way to do this would be to first
> calculate an "average" time series for each network (using constrained
> independent component analysis), and then to calculate the interaction
> between the task effect and ICA (or "average") time series for each network.
>
> I have two questions about implementing this in SPM:
>
> First, what is the best way to calculate the interaction between the
> task effect (i.e., stimulus onset) and the ICA time series? Would I
> simply multiply the two input time series? For example, let's say I have
> one task effect and one ICA time series. Would I enter the following
> three time series into the first-level GLM?
>
> Stimulus onset boxcar function: [0 0 0 1 1 1 0.../last TR/]
> ICA time series: [.4 .2 .5 .6 .7 .9 .3.../last TR/]
> Stimulus*ICA Interaction: [0 0 0 .6 .7 .9 0.../last TR/]
>
> Second, the task is broken up into four runs. How would I specify
> different regressors for each run (i.e., different ICA time series files
> and interaction terms for each run)? Or would it be better to
> concatenate the EPIs and regressor files and run one first-level model
> with one session?
>
> Thanks for the help!
> Phil
--
Guillaume Flandin, PhD
Wellcome Centre for Human Neuroimaging
UCL Queen Square Institute of Neurology
London WC1N 3BG
|