PPI
- Paul Macey (admin)
How-to
Access from Stats GUI (right)
Choose any combination of VOI and fMRI
For each fMRI and VOI:
Go to 1st level SPM
E.g.: …\fMRIpreprocess\…\SPMboxcar_s6wcrLinear\
Create a VOI subfolder
E.g.: …\fMRIpreprocess\…\SPMboxcar_s6wcrLinear\VOI\Hipp L\
Get F-contrast for All Effects (calculate if necessary)
Run Results with no threshold (P = 1 => F = 0) and VOI as mask
batch file: job_1stlevelVOIresults4ppi
Extract Eigenvariate using VOI as mask, no adjustment
Saves in VOI_…mat
Move to VOI folder
Create PPI variable (SPM PPI)
“psychophysiologic interaction” option (as of 2020-09-11)
Saved in PPI_…mat in VOI folder
Create new SPM model with PPI variables as regressors
Saved in VOI folder
batch file: job_design_ppi
* Set mask threshold to 0.0001 (default 0.8)
Estiamte model
Create T contrast of 1st variable (interaction)
con_0001.nii
Copy file to group folder for 2nd level analysis (prepend filename with study name)
CSPM will create a new fMRI measure than can be analysed with the SPM. (Need to re-load).
References:
Chapter 36 of SPM manual
Andy’s Brain Book “Psychophysiological Interactions (PPI) in SPM”
Our notes
* Mask Threshold: The SUIT images, its doing the same thing as when i was running it by hand and it was NaN-ing half the image, the way i worked around this was setting the masking threshold to 0.0001 instead of 0.8 in the job design step, i spoke to Luke about it and he thought it was something to do with the fact we isolate the brainstem by using a mask in the initial analysis when SUIT is run, and so it needs to be set to a near zero integer for the entire con image to be included in the analyses when run by hand and not through the server if that makes sense. Here is an image of what it generates currently - would it be possible to make a change to the code so that when it runs a job design step it sets the masking threshold to 0.0001?
