Hello,

I am new to gPPI in SPM, but I have run a similar model in FSL. I run an identical model in SPM and see what effect the deconvolution has on my results.
My input data for both analyses has already been preprocessed.


In FSL, I simply extracted the voxel timecourse from my preprocessed data (the voxel of interest) and used that as my physiological regressor.

In the PPPI setup, I have tried several things to replicate that, but so far to no avail.

--------------------
I read in one of these forums that I could directly set the P.VOI parameter to be a [NVols x 1] matrix with my timecourse, which would be ideal. But Whenever I try to do that I get an error when running PPPI.

Specifically, I set up the VOI matrix like so:



Then, I define the VOI parameter of the P structure

When I run the PPPI command, I get this error:


-------------------------

I have also tried using a mask .nii image file, and that runs, but when I look at the physiological timecourse afterwards it doesn't look anything like the original timecourse extracted from the data. They correlated at about 0. I have verified as many ways as I can that I am extracting the correct voxel.

I am getting the timecourse from the last column of the *_regressors.txt file that is written when PPPI is called, but I have gotten the same data from the SPM.mat file that is saved in the PPI directory as well.

For reference, here is the output from the PPPI command when it runs:






































I guess my main question is, assuming the mask is correct, why would the extracted timeseries not resemble the timeseries for that voxel from my input data at all?

And is there a way to circumvent the extraction altogether and directly define my VOI's timeseries data?


Thanks for any help you can provide.

Andrew Poppe

Hi Andrew,

Some very odd behavior in the scripts.

It should take in a numeric array without an issue, or at least get further than the error message its generating. What happens when you specify P.VOI, then run all(isnumeric(P.VOI))?

Do you get 1 or 0?

As for the mask, if you mask is a single voxel, then it should match FSL's timecourse -- are you using filtered_func.nii as the source in SPM and FSL?

If you send me your data, your mask, and the XYZ coordinate that you're trying to extract from, I can try and see what is going on with the program. I have some local tools for extracting data outside of the PPI toolbox and outside of the SPM.mat file.

Best,
Donald

Thank you for the quick reply!

When I run that command to test for numeric values, I get a 1.

I am actually using a somewhat odd dataset that actually represents the mean value of cortical parcels over time. I wouldn't expect that to have a bearing on this particular issue though. The data file is a nifti file that has the following dimensions:




In writing this, I just realized the answer might be that the analysis I ran in SPM prior to the PPPI may have altered the original data somehow even though it is already completely preprocessed. My unfamiliarity with SPM may have led me to leave on a filtering step or the like that altered the input data. Is that a possibility?

I have attached a zip file with the data file, my mask file, and the SPM.mat file from the GLM analysis. I'm trying to get data from the first voxel, 1,1,1 (1-indexed).

Thanks for your help!

Andrew