Von: freesurfer-bounces@nmr.mgh.harvard.edu <freesurfer-bounces@nmr.mgh.harvard.edu> im Auftrag von Douglas N. Greve <dgreve@MGH.HARVARD.EDU>
Gesendet: Donnerstag, 15. Juli 2021 18:05
An: freesurfer@nmr.mgh.harvard.edu
Betreff: Re: [Freesurfer] FIR analysis and task regressor (continued) & slice-timing correction

The FIR analysis can only be used with a task. It sounds like you want to deconvolve the HRF with the pupil diameter? This can be done, but it is tricky (can't  just use FIR).  You'd have to create a special task regressor where you have multiple columns. The first column would be the current task regressor. The next column would be the current regressor shifted down one (put a 0 in the first slot). Keep adding regressors to cover about 20-26 seconds (each regressor covers one TR). Then the beta values for these regressors will be the FIR you want. To visualize, you'd have to set up an F contrast (mkcontrast-sess), and then load the ces.nii.gz file

        External Email - Use Caution        

Hi,

Back in February I already had some issues with FIR analysis using a task regressor. In the meantime, I switched to other analysis. But now we would want to try that again.
I attached the old mails below, but since the problem slightly changed, I'll describe them again:

I want to run an analysis with the normalized pupil size as task regressor. Because the pupil dilation response itself is already slow (peaking after ~1s), we thought it would be good to use FIR instead of assuming a certain shape of HRF.

I tried two different analysis:

1) mkanalysis-sess -fsd bold -stc siemens  -surface fsaverage lh -fwhm 5 -notask -fir 2 15 -TR 2 -polyfit 2 -analysis PupP_fir.lh -per-run -taskreg PupPAvg 1

2) same with "-event-related -nc 2 -paradigm paradigm.txt" instead of -notask.

With 1), the analysis does run without errors but I don't get to see a time course when plotting with tksurfer-sess and I don't know what exactly is plotted than (the maximum or the first time point or...?).

With 2) I arrive at a similar point as in February: when I try to plot the results, I get this error:

ERROR: could not determine file for ./S05/bold/PupP_firTask.lh/h

The h.nii.gz file is indeed not there. If I copy h-offset.nii.gz and rename the copy to h.nii.gz, tksurfer-sess opens including the time course viewer. But this shows does not show a time course.

So the question is:
How do I get time courses for this FIR analysis with a task regressor? Where is my mistake?

And a slightly off-topic question regarding task regressors & slice timing:

As far as I know, stc shifts all slices to TR/2 and with the default stimulus delay parameter, TR/2 is also added to the times in paradigm files.
But how does it work for task regressors? Do I have to give "corrected" values?
I.e. in my case I average the pupil size over TR in order to arrive at the desired number of time points. But should I do that around the original start of the sequence or around the "corrected" time (i.e. +TR/2)?

Many thanks in advance,

Laura

On 2/5/2021 2:08 AM, Doll, Laura wrote:

        External Email - Use Caution

Hi Doug,

Thanks for your response.
No, I did not demean the task regressor... do I have to?

Yea, probably a good idea

Exactly. When I try to run tksurfer-sess, it complains about the missing file "h". When I rename the "h-offset" file to "h", I can run tksurfer-sess and I do see a map. The time course viewer opens as well but it only shows one time point.

Not sure here (have not run FIR in a long time). I noticed that you have -notask on your mkanalysis command. Have you tried it without -notask?

I can also continue to run the group analysis. During "isxconcat-sess" it while again complain about the missing file but mri_glmfit and following commands still work. So I do get a group map but I am not sure what it shows.

Depends on the contrast, but I think it should create maps for each time point in the FIR window.

Best,
Laura

Did you demean the task regressor? Is the problem that you cannot run tksurfer-sess? Or are you reporting multiple problems?

On 2/3/2021 8:32 AM, Doll, Laura wrote:

             External Email - Use Caution

    Hi everyone,


    Is it possible to do an FIR analysis using a task regressor
    instead of (or additional to) a paradigm file? More precisely, we
    want to use pupil size as task regressor. It was recorded during
    an auditory detection task (event related).

    This works when I use "-spmhrf" or other HRF options.

    However, for FIR I cannot get the expected time course in the
    single subjects data.

    Tksurfer ends with an error because it cannot find the file h.nii.gz:


    ERROR: could not determine file for [...]/S01_3/bold/PupilFIR.lh/h


    This file indeed does not exist, there is only one called "h-offset".

    The commands I used until that point are (here neglecting the
    original task, but using -event-related etc instead of no task
    results in the same error): mkanalysis-sess   -fsd bold -stc up 
    -surface fsaverage lh -fwhm 5 -notask -fir 4 20 -TR 2 -polyfit 2
    -analysis PupilFIR.lh -per-run -taskreg Pupil 1

    selxavg3-sess -s S01_3 -analysis PupilFIR.lh

    tksurfer-sess -s S01_3 -analysis PupilFIR.lh -call -tksurfer


    Is there any solution for this?
    Or any reason why it should not be done that way?


    Best,
    Laura



_______________________________________________
Freesurfer mailing list
Freesurfer@nmr.mgh.harvard.edu
MailScanner has detected a possible fraud attempt from "secure-web.cisco.com" claiming to be https://mail.nmr.mgh.harvard.edu/mailman/listinfo/freesurfer

_______________________________________________
Freesurfer mailing list
Freesurfer@nmr.mgh.harvard.edu
MailScanner has detected a possible fraud attempt from "secure-web.cisco.com" claiming to be https://mail.nmr.mgh.harvard.edu/mailman/listinfo/freesurfer

_______________________________________________
Freesurfer mailing list
Freesurfer@nmr.mgh.harvard.edu
MailScanner has detected a possible fraud attempt from "secure-web.cisco.com" claiming to be https://mail.nmr.mgh.harvard.edu/mailman/listinfo/freesurfer