External Email - Use Caution
Hello FS experts,
I have a cohort of subjects for whom I created cortical thickness and T1w/T2w signal ratio surface maps, projected them to fsaverage’s surface, smoothed them, and then created average thickness and T1w/T2w maps using mris_preproc. Is there a way to statistically compare the two averaged surface maps (i.e., come up with a third surface map displaying correlation patterns between thickness and T1w/T2w signal across the surface)?
Thanks in advance!
Best, Panos
Try this
To regress include a per-vertex regressor (PVR, --pvr) in mri_glmfit. This procedure will essentially add a column to the design matrix for each vertex. Eg, if your FSGD file (my.fsgd) has two groups, then, by itself, the design matrix would be two columns, one for each group. You will pass mri_glmfit two maps (one with --y, the input), and one the PVR; each map will have a frame for each subject. For a given vertex, the PVR values for each subject will be extracted and added as a column to the design matrix; the design matrix will be fit to the y-values for that vertex, the contrast matrices will be applied, and p-values computed. The extra column in the design matrix needs to be accounted for in the contrast matrix. Eg, if you have two groups, then you might have a contrast of +1 -1. If you are going to treat the PVR as a nuisance variable, then you would add a 0 (so +1 -1 0). If the PVR is a regressor of interest, then you could use 0 0 +1. Multiple PVRs can be added to make more complicated designs and test more complicated PVR hypotheses. DOSS and DODS do not apply to PVR. All the same rules about design matrices do apply (eg, scaling and colinearity). Permutation does not work with PVR.
Eg, create two maps, thickness and curvature, with
mris_preproc --f my.fsgd --hemi lh --meas thickness --o thickness.lh.sm00.mgz
mris_preproc --f my.fsgd --hemi lh --meas curv --o curv.lh.sm00.mgz
When you create your contrast matrices, make sure to include an item for the PVR.
mri_glmfit --y thickness.lh.sm00.mgz --fsgd my.fsgd --C contrast.mtx --pvr curv.lh.sm00.mgz --o my.glm --surf fsaverage lh
On 8/10/2020 1:29 AM, Fotiadis, Panagiotis wrote:
External Email - Use Caution
Hello FS experts,
I have a cohort of subjects for whom I created cortical thickness and T1w/T2w signal ratio surface maps, projected them to fsaverage’s surface, smoothed them, and then created average thickness and T1w/T2w maps using mris_preproc. Is there a way to statistically compare the two averaged surface maps (i.e., come up with a third surface map displaying correlation patterns between thickness and T1w/T2w signal across the surface)?
Thanks in advance!
Best,
Panos
Freesurfer mailing list Freesurfer@nmr.mgh.harvard.edu https://mail.nmr.mgh.harvard.edu/mailman/listinfo/freesurfer
External Email - Use Caution
Hi Doug,
That’s extremely helpful, thank you! And just to verify:
1. Since I have 1 group and I just want to compare the two surface maps (without adjusting for other covariates like age/sex at the moment), my contrast matrix should be [1 0], right? 2. If I add other covariates (let’s say age for instance), then if I want to test the hypothesis that the two regional maps are sig. different from each other after regressing the effects of age, would my contrast matrix then be [1 0 0]?
Best, Panos
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, August 10, 2020 at 9:57 AM To: "freesurfer@nmr.mgh.harvard.edu" freesurfer@nmr.mgh.harvard.edu Subject: [External] Re: [Freesurfer] Comparing surface maps
Try this
To regress include a per-vertex regressor (PVR, --pvr) in mri_glmfit. This procedure will essentially add a column to the design matrix for each vertex. Eg, if your FSGD file (my.fsgd) has two groups, then, by itself, the design matrix would be two columns, one for each group. You will pass mri_glmfit two maps (one with --y, the input), and one the PVR; each map will have a frame for each subject. For a given vertex, the PVR values for each subject will be extracted and added as a column to the design matrix; the design matrix will be fit to the y-values for that vertex, the contrast matrices will be applied, and p-values computed. The extra column in the design matrix needs to be accounted for in the contrast matrix. Eg, if you have two groups, then you might have a contrast of +1 -1. If you are going to treat the PVR as a nuisance variable, then you would add a 0 (so +1 -1 0). If the PVR is a regressor of interest, then you could use 0 0 +1. Multiple PVRs can be added to make more complicated designs and test more complicated PVR hypotheses. DOSS and DODS do not apply to PVR. All the same rules about design matrices do apply (eg, scaling and colinearity). Permutation does not work with PVR.
Eg, create two maps, thickness and curvature, with
mris_preproc --f my.fsgd --hemi lh --meas thickness --o thickness.lh.sm00.mgz
mris_preproc --f my.fsgd --hemi lh --meas curv --o curv.lh.sm00.mgz
When you create your contrast matrices, make sure to include an item for the PVR.
mri_glmfit --y thickness.lh.sm00.mgz --fsgd my.fsgd --C contrast.mtx --pvr curv.lh.sm00.mgz --o my.glm --surf fsaverage lh
On 8/10/2020 1:29 AM, Fotiadis, Panagiotis wrote:
External Email - Use Caution Hello FS experts,
I have a cohort of subjects for whom I created cortical thickness and T1w/T2w signal ratio surface maps, projected them to fsaverage’s surface, smoothed them, and then created average thickness and T1w/T2w maps using mris_preproc. Is there a way to statistically compare the two averaged surface maps (i.e., come up with a third surface map displaying correlation patterns between thickness and T1w/T2w signal across the surface)?
Thanks in advance!
Best, Panos
_______________________________________________
Freesurfer mailing list
Freesurfer@nmr.mgh.harvard.edumailto:Freesurfer@nmr.mgh.harvard.edu
https://mail.nmr.mgh.harvard.edu/mailman/listinfo/freesurfer
If you want to compare the maps, then the contrast would be [0 1] Imagine a scatter plot at each vertex with the values for one map on the x axis and the values for the other map on the y You then fit a straight line (offset and slope). The sig test will be for 0*offset+1*slope = slope
If you add a covariate, then the contrast would be [0 0 1]
On 8/11/2020 3:13 AM, Fotiadis, Panagiotis wrote:
External Email - Use Caution
Hi Doug,
That’s extremely helpful, thank you! And just to verify:
- Since I have 1 group and I just want to compare the two surface maps (without adjusting for other covariates like age/sex at the moment), my contrast matrix should be [1 0], right?
- If I add other covariates (let’s say age for instance), then if I want to test the hypothesis that the two regional maps are sig. different from each other after regressing the effects of age, would my contrast matrix then be [1 0 0]?
Best,
Panos
*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, August 10, 2020 at 9:57 AM *To: *"freesurfer@nmr.mgh.harvard.edu" freesurfer@nmr.mgh.harvard.edu *Subject: *[External] Re: [Freesurfer] Comparing surface maps
Try this
To regress include a per-vertex regressor (PVR, --pvr) in mri_glmfit. This procedure will essentially add a column to the design matrix for each vertex. Eg, if your FSGD file (my.fsgd) has two groups, then, by itself, the design matrix would be two columns, one for each group. You will pass mri_glmfit two maps (one with --y, the input), and one the PVR; each map will have a frame for each subject. For a given vertex, the PVR values for each subject will be extracted and added as a column to the design matrix; the design matrix will be fit to the y-values for that vertex, the contrast matrices will be applied, and p-values computed. The extra column in the design matrix needs to be accounted for in the contrast matrix. Eg, if you have two groups, then you might have a contrast of +1 -1. If you are going to treat the PVR as a nuisance variable, then you would add a 0 (so +1 -1 0). If the PVR is a regressor of interest, then you could use 0 0 +1. Multiple PVRs can be added to make more complicated designs and test more complicated PVR hypotheses. DOSS and DODS do not apply to PVR. All the same rules about design matrices do apply (eg, scaling and colinearity). Permutation does not work with PVR.
Eg, create two maps, thickness and curvature, with
mris_preproc --f my.fsgd --hemi lh --meas thickness --o thickness.lh.sm00.mgz
mris_preproc --f my.fsgd --hemi lh --meas curv --o curv.lh.sm00.mgz
When you create your contrast matrices, make sure to include an item for the PVR.
mri_glmfit --y thickness.lh.sm00.mgz --fsgd my.fsgd --C contrast.mtx --pvr curv.lh.sm00.mgz --o my.glm --surf fsaverage lh
On 8/10/2020 1:29 AM, Fotiadis, Panagiotis wrote:
* External Email - Use Caution * Hello FS experts, I have a cohort of subjects for whom I created cortical thickness and T1w/T2w signal ratio surface maps, projected them to fsaverage’s surface, smoothed them, and then created average thickness and T1w/T2w maps using mris_preproc. Is there a way to statistically compare the two averaged surface maps (i.e., come up with a third surface map displaying correlation patterns between thickness and T1w/T2w signal across the surface)? Thanks in advance! Best, Panos _______________________________________________ Freesurfer mailing list Freesurfer@nmr.mgh.harvard.edu <mailto:Freesurfer@nmr.mgh.harvard.edu> https://mail.nmr.mgh.harvard.edu/mailman/listinfo/freesurfer
Freesurfer mailing list Freesurfer@nmr.mgh.harvard.edu https://mail.nmr.mgh.harvard.edu/mailman/listinfo/freesurfer
External Email - Use Caution
Ah I see. Great, thanks Doug!!
Best, Panos
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: Tuesday, August 11, 2020 at 10:37 AM To: "freesurfer@nmr.mgh.harvard.edu" freesurfer@nmr.mgh.harvard.edu Subject: Re: [Freesurfer] [External] Re: Comparing surface maps
If you want to compare the maps, then the contrast would be [0 1] Imagine a scatter plot at each vertex with the values for one map on the x axis and the values for the other map on the y You then fit a straight line (offset and slope). The sig test will be for 0*offset+1*slope = slope
If you add a covariate, then the contrast would be [0 0 1]
On 8/11/2020 3:13 AM, Fotiadis, Panagiotis wrote:
External Email - Use Caution Hi Doug,
That’s extremely helpful, thank you! And just to verify:
1. Since I have 1 group and I just want to compare the two surface maps (without adjusting for other covariates like age/sex at the moment), my contrast matrix should be [1 0], right? 2. If I add other covariates (let’s say age for instance), then if I want to test the hypothesis that the two regional maps are sig. different from each other after regressing the effects of age, would my contrast matrix then be [1 0 0]?
Best, Panos
From: freesurfer-bounces@nmr.mgh.harvard.edumailto:freesurfer-bounces@nmr.mgh.harvard.edu on behalf of "Douglas N. Greve" dgreve@mgh.harvard.edumailto:dgreve@mgh.harvard.edu Reply-To: Freesurfer support list freesurfer@nmr.mgh.harvard.edumailto:freesurfer@nmr.mgh.harvard.edu Date: Monday, August 10, 2020 at 9:57 AM To: "freesurfer@nmr.mgh.harvard.edu"mailto:freesurfer@nmr.mgh.harvard.edu freesurfer@nmr.mgh.harvard.edumailto:freesurfer@nmr.mgh.harvard.edu Subject: [External] Re: [Freesurfer] Comparing surface maps
Try this
To regress include a per-vertex regressor (PVR, --pvr) in mri_glmfit. This procedure will essentially add a column to the design matrix for each vertex. Eg, if your FSGD file (my.fsgd) has two groups, then, by itself, the design matrix would be two columns, one for each group. You will pass mri_glmfit two maps (one with --y, the input), and one the PVR; each map will have a frame for each subject. For a given vertex, the PVR values for each subject will be extracted and added as a column to the design matrix; the design matrix will be fit to the y-values for that vertex, the contrast matrices will be applied, and p-values computed. The extra column in the design matrix needs to be accounted for in the contrast matrix. Eg, if you have two groups, then you might have a contrast of +1 -1. If you are going to treat the PVR as a nuisance variable, then you would add a 0 (so +1 -1 0). If the PVR is a regressor of interest, then you could use 0 0 +1. Multiple PVRs can be added to make more complicated designs and test more complicated PVR hypotheses. DOSS and DODS do not apply to PVR. All the same rules about design matrices do apply (eg, scaling and colinearity). Permutation does not work with PVR.
Eg, create two maps, thickness and curvature, with
mris_preproc --f my.fsgd --hemi lh --meas thickness --o thickness.lh.sm00.mgz
mris_preproc --f my.fsgd --hemi lh --meas curv --o curv.lh.sm00.mgz
When you create your contrast matrices, make sure to include an item for the PVR.
mri_glmfit --y thickness.lh.sm00.mgz --fsgd my.fsgd --C contrast.mtx --pvr curv.lh.sm00.mgz --o my.glm --surf fsaverage lh
On 8/10/2020 1:29 AM, Fotiadis, Panagiotis wrote:
External Email - Use Caution Hello FS experts,
I have a cohort of subjects for whom I created cortical thickness and T1w/T2w signal ratio surface maps, projected them to fsaverage’s surface, smoothed them, and then created average thickness and T1w/T2w maps using mris_preproc. Is there a way to statistically compare the two averaged surface maps (i.e., come up with a third surface map displaying correlation patterns between thickness and T1w/T2w signal across the surface)?
Thanks in advance!
Best, Panos
_______________________________________________
Freesurfer mailing list
Freesurfer@nmr.mgh.harvard.edumailto:Freesurfer@nmr.mgh.harvard.edu
https://mail.nmr.mgh.harvard.edu/mailman/listinfo/freesurfer
_______________________________________________
Freesurfer mailing list
Freesurfer@nmr.mgh.harvard.edumailto:Freesurfer@nmr.mgh.harvard.edu
https://mail.nmr.mgh.harvard.edu/mailman/listinfo/freesurfer
freesurfer@nmr.mgh.harvard.edu
-
Douglas N. Greve -
Fotiadis, Panagiotis