External Email - Use Caution
Hi, I would like to create a mask of the WM for fMRI analysis. I read a previous post in which it was suggested to get the wm segmentation and erode it, since the result will be exclusively WM. My question is why is this so? Why can I assume that getting the WM mask from the aseg file will almost certainly get me high probability WM, >0.9 for example? In most segmentation algorithms we get posterior probabilities at each voxel, so that we can then select voxels with posterior probabilities >0.90 for example. Why is freesurfer's WM segmentation so accurate that we do not need posterior probabilities? Are these very high probabilities somehow already coded in freesurfer's segmentation algorithm? Thanks, Ben
External Email - Use Caution
Perhaps erosion was suggested because of partial voluming in your fMRI voxels.
Peace,
Matt.
From: <freesurfer-bounces@nmr.mgh.harvard.edumailto:freesurfer-bounces@nmr.mgh.harvard.edu> on behalf of Ben M <fmri.mri@gmail.commailto:fmri.mri@gmail.com> Reply-To: Freesurfer support list <freesurfer@nmr.mgh.harvard.edumailto:freesurfer@nmr.mgh.harvard.edu> Date: Monday, August 6, 2018 at 9:27 PM To: "freesurfer@nmr.mgh.harvard.edumailto:freesurfer@nmr.mgh.harvard.edu" <freesurfer@nmr.mgh.harvard.edumailto:freesurfer@nmr.mgh.harvard.edu> Subject: [Freesurfer] WM segmentation
External Email - Use Caution
Hi, I would like to create a mask of the WM for fMRI analysis. I read a previous post in which it was suggested to get the wm segmentation and erode it, since the result will be exclusively WM. My question is why is this so? Why can I assume that getting the WM mask from the aseg file will almost certainly get me high probability WM, >0.9 for example? In most segmentation algorithms we get posterior probabilities at each voxel, so that we can then select voxels with posterior probabilities >0.90 for example. Why is freesurfer's WM segmentation so accurate that we do not need posterior probabilities? Are these very high probabilities somehow already coded in freesurfer's segmentation algorithm? Thanks, Ben
________________________________ The materials in this message are private and may contain Protected Healthcare Information or other information of a sensitive nature. If you are not the intended recipient, be advised that any unauthorized use, disclosure, copying or the taking of any action in reliance on the contents of this information is strictly prohibited. If you have received this email in error, please immediately notify the sender via telephone or return mail.
Hi Ben
our primary cortical segmentation of the wm is not probabilistic. In any case you want partial volume fractions I expect, not posterior probabilities. I would use mris_fill to create a mask of the WM from the ?h.white surfaces, if you want the most accuracy
cheers Bruce
On Tue, 7 Aug 2018, Ben M wrote:
External Email - Use Caution
Hi, I would like to create a mask of the WM for fMRI analysis. I read a previous post in which it was suggested to get the wm segmentation and erode it, since the result will be exclusively WM. My question is why is this so? Why can I assume that getting the WM mask from the aseg file will almost certainly get me high probability WM, >0.9 for example? In most segmentation algorithms we get posterior probabilities at each voxel, so that we can then select voxels with posterior probabilities >0.90 for example. Why is freesurfer's WM segmentation so accurate that we do not need posterior probabilities? Are these very high probabilities somehow already coded in freesurfer's segmentation algorithm? Thanks, Ben
External Email - Use Caution
Hi Bruce,
Thank you very much for your reply. The idea is to create a WM mask that I can use for the denoising step of resting state connectivity analysis. Do you mind if I ask you a few follow up questions?
1 - "our primary cortical segmentation of the wm is not probabilistic." How is the segmentation of the WM achieved in freesurfer then, if not probabilistic?
2 - "In any case you want partial volume fractions I expect, not posterior probabilities." Sorry, I'm not sure what you mean by partial volume fractions. In several papers I have read the WM is first thresholded (eg p >.9) to ensure that only WM voxels are included in the mask, which is also eroded. Are the values being thresholded partial volume fractions then, and not probabilities? But if segmentation is done on tissue probability maps (eg like in SPM I think), wouldn't you have probabilities? Or do you mean you have partial volume fractions just in freesurfer?
3 - "I would use mris_fill to create a mask of the WM from the ?h.white surfaces, if you want the most accuracy" I found more information about the command mri_fill than mris_fill, is there a difference between the two, or can I use both for the same thing?
Sorry for the newbie questions, I am just beginning using freesurfer.
Best Ben
On Tue, Aug 7, 2018 at 6:08 PM, Bruce Fischl fischl@nmr.mgh.harvard.edu wrote:
Hi Ben
our primary cortical segmentation of the wm is not probabilistic. In any case you want partial volume fractions I expect, not posterior probabilities. I would use mris_fill to create a mask of the WM from the ?h.white surfaces, if you want the most accuracy
cheers Bruce
On Tue, 7 Aug 2018, Ben M wrote:
External Email - Use CautionHi, I would like to create a mask of the WM for fMRI analysis. I read a previous post in which it was suggested to get the wm segmentation and erode it, since the result will be exclusively WM. My question is why is this so? Why can I assume that getting the WM mask from the aseg file will almost certainly get me high probability WM, >0.9 for example? In most segmentation algorithms we get posterior probabilities at each voxel, so that we can then select voxels with posterior probabilities >0.90 for example. Why is freesurfer's WM segmentation so accurate that we do not need posterior probabilities? Are these very high probabilities somehow already coded in freesurfer's segmentation algorithm? Thanks, Ben
Freesurfer mailing list Freesurfer@nmr.mgh.harvard.edu https://mail.nmr.mgh.harvard.edu/mailman/listinfo/freesurfer
The information in this e-mail is intended only for the person to whom it is addressed. If you believe this e-mail was sent to you in error and the e-mail contains patient information, please contact the Partners Compliance HelpLine at http://www.partners.org/complianceline . If the e-mail was sent to you in error but does not contain patient information, please contact the sender and properly dispose of the e-mail.
1. You can dig up our old 1999 recon papers if you are really interested. Lots of segmentation techniques are not probabilistic.
2. Again, we don't output probabilities. YOu are better off eroding as Matt suggested.
3. mris_fill fills the interior of a surface - nothing to do with mri_fill.
If all you want is a nuisance regressor just take the wm.mgz, or interior of the white surface and erode it once, then use those voxels Bruce
On Tue, 7 Aug 2018, Ben M wrote:
External Email - Use Caution
Hi Bruce,
Thank you very much for your reply. The idea is to create a WM mask that I can use for the denoising step of resting state connectivity analysis. Do you mind if I ask you a few follow up questions?
1 - "our primary cortical segmentation of the wm is not probabilistic." How is the segmentation of the WM achieved in freesurfer then, if not probabilistic?
2 - "In any case you want partial volume fractions I expect, not posterior probabilities." Sorry, I'm not sure what you mean by partial volume fractions. In several papers I have read the WM is first thresholded (eg p >.9) to ensure that only WM voxels are included in the mask, which is also eroded. Are the values being thresholded partial volume fractions then, and not probabilities? But if segmentation is done on tissue probability maps (eg like in SPM I think), wouldn't you have probabilities? Or do you mean you have partial volume fractions just in freesurfer?
3 - "I would use mris_fill to create a mask of the WM from the ?h.white surfaces, if you want the most accuracy" I found more information about the command mri_fill than mris_fill, is there a difference between the two, or can I use both for the same thing?
Sorry for the newbie questions, I am just beginning using freesurfer.
Best Ben
On Tue, Aug 7, 2018 at 6:08 PM, Bruce Fischl fischl@nmr.mgh.harvard.edu wrote: Hi Ben
our primary cortical segmentation of the wm is not probabilistic. In any case you want partial volume fractions I expect, not posterior probabilities. I would use mris_fill to create a mask of the WM from the ?h.white surfaces, if you want the most accuracy cheers Bruce On Tue, 7 Aug 2018, Ben M wrote: External Email - Use Caution Hi, I would like to create a mask of the WM for fMRI analysis. I read a previous post in which it was suggested to get the wm segmentation and erode it, since the result will be exclusively WM. My question is why is this so? Why can I assume that getting the WM mask from the aseg file will almost certainly get me high probability WM, >0.9 for example? In most segmentation algorithms we get posterior probabilities at each voxel, so that we can then select voxels with posterior probabilities >0.90 for example. Why is freesurfer's WM segmentation so accurate that we do not need posterior probabilities? Are these very high probabilities somehow already coded in freesurfer's segmentation algorithm? Thanks, Ben
Freesurfer mailing list Freesurfer@nmr.mgh.harvard.edu https://mail.nmr.mgh.harvard.edu/mailman/listinfo/freesurfer
The information in this e-mail is intended only for the person to whom it is addressed. If you believe this e-mail was sent to you in error and the e-mail contains patient information, please contact the Partners Compliance HelpLine at http://www.partners.org/complianceline . If the e-mail was sent to you in error but does not contain patient information, please contact the sender and properly dispose of the e-mail.
External Email - Use Caution
I would add that a properly eroded WM mask is not very effective for fMRI denoising at least relative to spatial and temporal ICA-based denoising. An uneroded mask behaves similarly to global signal regression, which removes both global noise and global neural signal.
Matt.
On 8/7/18, 4:22 PM, "freesurfer-bounces@nmr.mgh.harvard.edu on behalf of Bruce Fischl" <freesurfer-bounces@nmr.mgh.harvard.edu on behalf of fischl@nmr.mgh.harvard.edu> wrote:
- You can dig up our old 1999 recon papers if you are really interested.
Lots of segmentation techniques are not probabilistic.
- Again, we don't output probabilities. YOu are better off eroding as
Matt suggested.
- mris_fill fills the interior of a surface - nothing to do with
mri_fill.
If all you want is a nuisance regressor just take the wm.mgz, or interior of the white surface and erode it once, then use those voxels Bruce
On Tue, 7 Aug 2018, Ben M wrote:
External Email - Use CautionHi Bruce,
Thank you very much for your reply. The idea is to create a WM mask that I can use for the denoising step of resting state connectivity analysis. Do you mind if I ask you a few follow up questions?
1 - "our primary cortical segmentation of the wm is not probabilistic." How is the segmentation of the WM achieved in freesurfer then, if not probabilistic?
2 - "In any case you want partial volume fractions I expect, not posterior probabilities." Sorry, I'm not sure what you mean by partial volume fractions. In several papers I have read the WM is first thresholded (eg p >.9) to ensure that only WM voxels are included in the mask, which is also eroded. Are the values being thresholded partial volume fractions then, and not probabilities? But if segmentation is done on tissue probability maps (eg like in SPM I think), wouldn't you have probabilities? Or do you mean you have partial volume fractions just in freesurfer?
3 - "I would use mris_fill to create a mask of the WM from the ?h.white surfaces, if you want the most accuracy" I found more information about the command mri_fill than mris_fill, is there a difference between the two, or can I use both for the same thing?
Sorry for the newbie questions, I am just beginning using freesurfer.
Best Ben
On Tue, Aug 7, 2018 at 6:08 PM, Bruce Fischl fischl@nmr.mgh.harvard.edu wrote: Hi Ben
our primary cortical segmentation of the wm is not probabilistic.In any case you want partial volume fractions I expect, not posterior probabilities. I would use mris_fill to create a mask of the WM from the ?h.white surfaces, if you want the most accuracy
cheers Bruce On Tue, 7 Aug 2018, Ben M wrote: External Email - Use Caution Hi, I would like to create a mask of the WM for fMRI analysis.I read a previous post in which it was suggested to get the wm segmentation and erode it, since the result will be exclusively WM. My question is why is this so? Why can I assume that getting the WM mask from the aseg file will almost certainly get me high probability WM, >0.9 for example? In most segmentation algorithms we get posterior probabilities at each voxel, so that we can then select voxels with posterior probabilities >0.90 for example. Why is freesurfer's WM segmentation so accurate that we do not need posterior probabilities? Are these very high probabilities somehow already coded in freesurfer's segmentation algorithm? Thanks, Ben
Freesurfer mailing list Freesurfer@nmr.mgh.harvard.edu https://mail.nmr.mgh.harvard.edu/mailman/listinfo/freesurfer
The information in this e-mail is intended only for the person to whom it is addressed. If you believe this e-mail was sent to you in error and the e-mail contains patient information, please contact the Partners Compliance HelpLine at http://www.partners.org/complianceline . If the e-mail was sent to you in error but does not contain patient information, please contact the sender and properly dispose of the e-mail.
________________________________ The materials in this message are private and may contain Protected Healthcare Information or other information of a sensitive nature. If you are not the intended recipient, be advised that any unauthorized use, disclosure, copying or the taking of any action in reliance on the contents of this information is strictly prohibited. If you have received this email in error, please immediately notify the sender via telephone or return mail.
freesurfer@nmr.mgh.harvard.edu
-
Ben M -
Bruce Fischl -
Glasser, Matthew