This is actually a very obscure feature of freesurfer that takes some explaining. In FS, all p-values are computed with a two-sided test (so the sig.mgh represents a signed p-value). When you pass --sign pos to mri_surfcluster, it sees that you want to do a signed test instead of an unsigned test, so it essentially changes the sig.mgh internally to create unsigned p-values. Signed is more powerful than unsigned so you can get more clusters than the equivalent unsigned test. But when you look at the sig.mgh in freeview, you are looking at unsigned p-values. You could change the sig.mgh to signed p-values and then it should match what you are seeing in the summary file. Or you can simply change the threshold in freeview from 2 (= -log10(.01)) to 1.7 (= -log10(.02)). Sorry, I know this is confusing...

On 2/22/2023 8:59 PM, Smith, Christine wrote:

        External Email - Use Caution

Hi Yujing,

Thank you for explaining that.

I believe that mri_surfcluster is the correct output, in that it is carrying out a one-sided test of significance at 2 (Sig = -log(pvalue), i.e., p < 0.01).

For the Configure options, is there a similar option to allow the threshold to be 2 for a one-sided test?

We have experimented around by setting very high (max) and very low (min) values and we still weren't able to get the clusters identified by mri_surfclust to show up on Freeview, even when we reduce the threshold to account for the fact that it is a two-sided test.

Christine

--

Christine N. Smith, Ph.D.

Research Service, VA San Diego Healthcare System
Department of Psychiatry, University of California San Diego

VA San Diego Healthcare System

VMRF Building 13

3350 La Jolla Village Drive (151A)

San Diego, CA 92161

T: (858) 552-8585; then press *7128

From: freesurfer-bounces@nmr.mgh.harvard.edu <freesurfer-bounces@nmr.mgh.harvard.edu> on behalf of Huang, Yujing <YHUANG43@mgh.harvard.edu>
Sent: Wednesday, February 22, 2023 1:02 PM
To: Freesurfer support list <freesurfer@nmr.mgh.harvard.edu>
Subject: Re: [Freesurfer] different clusters in mri_surfcluster output versus viewing in Freeview

Hi Christine,

Sig = -log(pvalue)

In freeview, the overlay thresholds can be adjusted by clicking ‘Configure’ button (under the Overlay dropdown menu).

For mri_surfcluster (MailScanner has detected a possible fraud attempt from "secure-web.cisco.com" claiming to be https://surfer.nmr.mgh.harvard.edu/fswiki/mri_surfcluster), I think you can try to adjust these options to match freeview: --thmin <>, --thmax <>, --thsign <>, --minarea <>, --no-adjust.

Best,

Yujing

From: freesurfer-bounces@nmr.mgh.harvard.edu <freesurfer-bounces@nmr.mgh.harvard.edu> On Behalf Of Smith, Christine
Sent: Tuesday, February 21, 2023 6:01 PM
To: freesurfer@nmr.mgh.harvard.edu
Subject: [Freesurfer] different clusters in mri_surfcluster output versus viewing in Freeview

        External Email - Use Caution

Hello,

I am getting a mismatch of clusters when we view cluster results in freeview versus ask for a table of the significant clusters.

I am performing a GLM analysis on Freesurfer to analyze cortical thickness (mri_glmfit). When viewing the significance map with freeview, we set the minimum threshold to 2 for the sig.mgh file.

We also used the same threshold when creating a summary table of the clusters using mri_surfcluster.

mri_surfcluster --in {$dir}/visuospatial_MCI_correlation/sig.mgh --hemi left --thmin 2 --thsign pos --minarea 10 --subject fsaverage --surf inflated --annot BN_Atlas --sum {$dir}/visuospatial_MCI_correlation/left.thickness.clustersummary_BN.txt

We were hoping to use the summary text file to identify the clusters on freeview and their characteristics (e.g., size, x, y, z ccordinates). Unfortunately, the regions that we see for both methods don’t match as there are some clusters that we see on the summary text file that we don’t see on freeview and vice versa. How can we make the output of these two methods match up? Note that the issue is not the minimum cluster size in the summary table. When we put that to 1 we still get a mismatch.

Please note that in the mri_surfcluster program we can indicate that we only want positive results. We don't see a way to do this in freeview or a way for that to affect the p value in freeview.

Is this possibly the issue? If so, how can we make a picture in freeview that matches our summary table?

Regards,

Christine

--

Christine N. Smith, Ph.D.

Research Service, VA San Diego Healthcare System
Department of Psychiatry, University of California San Diego

VA San Diego Healthcare System

VMRF Building 13

3350 La Jolla Village Drive (151A)

San Diego, CA 92161

T: (858) 552-8585; then press *7128
_______________________________________________
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