Dear Douglas,

 

Very good to know this is possible, thank you!

 

 

 

If you just have a simple design matrix (not doing any temporal whitening or anything else other than a standard GLM), you can use mri_glmfit. Just pass the design matrix as a text file via --X (instead of using an FSGD). You'll have to pass contrast matrices as well.

        External Email - Use Caution        

Dear Douglas,

Yes, I mean fMRI. I am currently using MATLAB to fit a GLM to the timeseries and then export the p-values as a MGH file of the cortical surface.

Best,

 

Leonardo Tozzi, MD, PhD

Williams PanLab | Postdoctoral Researcher

Stanford University | 401 Quarry Rd

ltozzi@stanford.edu | (650) 5615738

 

Connect with us!

    

 

 

 

From: <freesurfer-bounces@nmr.mgh.harvard.edu> on behalf of "Douglas N. Greve" <dgreve@mgh.harvard.edu>
Reply-To: Freesurfer support list <freesurfer@nmr.mgh.harvard.edu>
Date: Monday, May 10, 2021 at 9:59 AM
To: "freesurfer@nmr.mgh.harvard.edu" <freesurfer@nmr.mgh.harvard.edu>
Subject: Re: [Freesurfer] Cluster correction at subject level using -log10(p) values

 

when you say time series, do you mean like fMRI? What are you using to do the time series analysis?

On 5/10/2021 12:56 PM, Leonardo Tozzi wrote:

        External Email - Use Caution        

Dear Douglas,

 

The reason is that I am running a GLM at the subject level on 4D fMRI timeseries. Thus, I obtain a parameter estimate on each vertex, which I can then enter in a contrast to obtain a p-value per vertex. These are the p-values I want to correct.

From what I understand, the mri_glmfit does not work on timeseries data, am I wrong in this? So at the moment I am fitting the GLM in matlab and using the mcsim to correct the p-values I obtain.

Best,

 

Leonardo Tozzi, MD, PhD

Williams PanLab | Postdoctoral Researcher

Stanford University | 401 Quarry Rd

ltozzi@stanford.edu | (650) 5615738

 

Connect with us!

    

 

 

 

From: <freesurfer-bounces@nmr.mgh.harvard.edu> on behalf of "Douglas N. Greve" <dgreve@mgh.harvard.edu>
Reply-To: Freesurfer support list <freesurfer@nmr.mgh.harvard.edu>
Date: Monday, May 10, 2021 at 8:32 AM
To: "freesurfer@nmr.mgh.harvard.edu" <freesurfer@nmr.mgh.harvard.edu>
Subject: Re: [Freesurfer] Cluster correction at subject level using -log10(p) values

 

Is there a reason that you are not using mri_glmfit-sim to do this? It is fairly complicated, and so I spent a lot of time writing mri_glmfit-sim to make sure it does the right thing.

On 4/29/2021 6:16 PM, Leonardo Tozzi wrote:

        External Email - Use Caution        

Dear Experts,

 

I am running an analysis where I have -log10(p) values on the cortical surface of some subjects stored as a MGH file.

I would now like to run a cluster level correction on these data at the level of each individual subject using mri_mcsim followed by mri_surfcluster and I would like to constrain my inference to a ROI.

I have a few questions. First of all, is it correct that I should to a -log10(p) transform on my p values before entering them in mri_surfcluster?

Second, will the csd generated by mcsim be valid to feed in mri_surfcluster since the values are a -log10(p) transformed? Does mcsim -log transform the gaussian noise it uses at some point or is this not necessary?

My calls to the two methods are as follows (for one hemisphere):

 

First the mcsim:

mri_mcsim --o “${subject}_mcsim_lh_${roi}" --base mc-z --surface ${subject} lh --nreps 10000 --label $roi

 

Then the mri_surfcluster:

mri_surfcluster --in "/${subject}_lh_${roi}_neglog10p_signed.mgh" \

--csd "${subject}_mcsim_lh_${roi}/fwhm05/abs/th30/mc-z.csd" \

--clabel "${subject}/label/lh.${roi}.label" \

--cwsig "${subject}_lh_${roi}_neglog10p_abs_clu05.mgh" \

--vwsig "${subject}_lh_${roi}_neglog10p_abs_vox05.mgh" \

--sum "${subject}_lh_${roi}_neglog10p_abs_summary.txt" \

--thmin 3 \

--surf white \

--subject ${subject } \

--hemi lh \

--ocn “${subject}_lh_${roi}_neglog10p_abs_clu05_clunum.mgh" \

--cwpvalthresh 0.05

 

In the call of mri_surfcluster, I want a cluster forming threshold of p=0.001, which would correspond to -logp(0.001)=3. Is it correct that the argument --thmin should be expressed as -log10(p) but the parameter cwpvalthresh should be p?

The reason why I think I did something wrong is that I get a lot of clusters from this analysis, sometimes >15 per individual. This makes me think that the correction might not be as stringent as I would like to.

Thank you,

 

Leonardo Tozzi, MD, PhD

Williams PanLab | Postdoctoral Researcher

Stanford University | 401 Quarry Rd

ltozzi@stanford.edu | (650) 5615738

 

Connect with us!

    

 

 

_______________________________________________

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