Based on our understanding of the longitudinal pipeline for 4.5, image edits for the -long runs come mostly from the template/base subj (with an optional flag for CPs to be copied from the template; http://www.mail-archive.com/freesurfer@nmr.mgh.harvard.edu/msg13256.html, http://www.mail-archive.com/freesurfer@nmr.mgh.harvard.edu/msg12895.html). In addition, our assumption is that pial edits will come via the template, as the -long brainmask, is a result of masking the -long T1.mgz by the template brainmask.mgz (which itself is the norm_template.mgz, a combination of every time point's norm.mgz [which is pre-edit stage]).
We've run into an issue where our -long runs are still in need of pial editing, despite edits having been done to both the cross sectional and template runs (see attached). However, it's also our understanding that we should not edit the -long runs and try to re-run (using -autorecon2-pial -autorecon3), as this will pollute the correction for intrasubject variation (per communication from Martin). How would you recommend to proceed? Thanks.
-Derin
Hi Derin,
as far as I can see your description is correct. Note: in your screenshots the base is shown in a different slice (but I believe that you probably scrolled through the image to make sure no skull is included in a different slice).
Actually in this situation, the long runs should not need editing, if the brainmask.mgz in the base is accurate. It is mapped to the longitudinals and used as a brainmask there.
Have you tried editing the brainmask in the longitudinal directory (or if done in the cross sectional you can simply copy it over as it is in the same space). Then pass -long .... and -autorecon2-pial -autorecon3? Not sure what this does or if it works, but it's worth a try. If it does not work, I guess we need to look at this data, to see what is going on.
Currently I am working on testing out the 5.0 fix and some other possible improvements...
Best, Martin
On Fri, 2010-09-03 at 13:54 -0500, Derin Cobia wrote:
Based on our understanding of the longitudinal pipeline for 4.5, image edits for the -long runs come mostly from the template/base subj (with an optional flag for CPs to be copied from the template; http://www.mail-archive.com/freesurfer@nmr.mgh.harvard.edu/msg13256.html, http://www.mail-archive.com/freesurfer@nmr.mgh.harvard.edu/msg12895.html). In addition, our assumption is that pial edits will come via the template, as the -long brainmask, is a result of masking the -long T1.mgz by the template brainmask.mgz (which itself is the norm_template.mgz, a combination of every time point's norm.mgz [which is pre-edit stage]).
We've run into an issue where our -long runs are still in need of pial editing, despite edits having been done to both the cross sectional and template runs (see attached). However, it's also our understanding that we should not edit the -long runs and try to re-run (using -autorecon2-pial -autorecon3), as this will pollute the correction for intrasubject variation (per communication from Martin). How would you recommend to proceed? Thanks.
-Derin
Freesurfer mailing list Freesurfer@nmr.mgh.harvard.edu https://mail.nmr.mgh.harvard.edu/mailman/listinfo/freesurfer
Hi Martin,
Thanks for the input. You are correct that the slices from the tp's to base are slightly off, this is because of the poor correspondence in slice numbering between them. And yes, there was still skull/dura several slices forward and back.
Are you sure that the -long runs don't need to be edited in this case? Even with clearly bad pial surfaces? I believe that the brainmask.mgz in the base is correct, however, my concern is that it is being poorly mapped to the longitudinals. There are multiple instances in this same subject where the brainmask.mgz differs between the -long and base; particularly, excess skull/dura in the -long runs, where it has been edited out in the base. What would cause this?
I like the idea of copying over the cross-sectional brainmask.mgz to initialize the longitudinal pipeline (as opposed to masking T1.mgz with norm_template.mgz). Is there a reason why this is not the default? This would make sense to me given the time/effort put into edits at tp1 - tpn, then they can just be pushed forward to the longitudinal runs. Maybe Nick or Doug could comment on it? I'll test it out and see. Thanks.
-Derin
On Sep 3, 2010, at 4:37 PM, Martin Reuter wrote:
Hi Derin,
as far as I can see your description is correct. Note: in your screenshots the base is shown in a different slice (but I believe that you probably scrolled through the image to make sure no skull is included in a different slice).
Actually in this situation, the long runs should not need editing, if the brainmask.mgz in the base is accurate. It is mapped to the longitudinals and used as a brainmask there.
Have you tried editing the brainmask in the longitudinal directory (or if done in the cross sectional you can simply copy it over as it is in the same space). Then pass -long .... and -autorecon2-pial -autorecon3? Not sure what this does or if it works, but it's worth a try. If it does not work, I guess we need to look at this data, to see what is going on.
Currently I am working on testing out the 5.0 fix and some other possible improvements...
Best, Martin
On Fri, 2010-09-03 at 13:54 -0500, Derin Cobia wrote:
Based on our understanding of the longitudinal pipeline for 4.5, image edits for the -long runs come mostly from the template/base subj (with an optional flag for CPs to be copied from the template; http://www.mail-archive.com/freesurfer@nmr.mgh.harvard.edu/msg13256.html, http://www.mail-archive.com/freesurfer@nmr.mgh.harvard.edu/msg12895.html). In addition, our assumption is that pial edits will come via the template, as the -long brainmask, is a result of masking the -long T1.mgz by the template brainmask.mgz (which itself is the norm_template.mgz, a combination of every time point's norm.mgz [which is pre-edit stage]).
We've run into an issue where our -long runs are still in need of pial editing, despite edits having been done to both the cross sectional and template runs (see attached). However, it's also our understanding that we should not edit the -long runs and try to re-run (using -autorecon2-pial -autorecon3), as this will pollute the correction for intrasubject variation (per communication from Martin). How would you recommend to proceed? Thanks.
-Derin
Freesurfer mailing list 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
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.
Hi Derin,
the default behavior is to use a common brainmask for all timepoints instead of different ones (which is what happens when you take it from the cross sectionals). This is done to reduce variability.
Also there is not supposed to be a 'correspondence in slice numbering' between the base and the longitudinals. So far the longitudinals are in the same space as the cross sectionals NOT in the same space as the base. (this is likely to change in future releases, making much of this obsolete).
However the transforms are all there (in baseid/mri/transforms) and can be used to map anything of interest to the base space or vice versa from the base to the individual time points. Those registrations are highly accurate, but as always something can go wrong. You might want to check them e.g. by mapping the norm.mgz from the base to the specific time point (mri_convert -at baseid/mri/transforms/baseid_to_tpid.lta -rl <target-norm-mgz> baseid/mri/norm.mgz output.mgz
and look at them in tkmedit. The images should be nicely registered. The same transform is applied to the brainmask.mgz in the base to be mapped to each time point. If you remove voxels from the brainmask in the base, the corresponding voxels will be removed in the timepoints. If you don't see that: -either the transforms are incorrect (check above) -or you are not editing the right voxels
Another thing you can try is: map the edited brainmask from the cross sectionals to the base and compare to your edited base brainmask.
We plan to revisit the editing in the longitudinal processing anyway. I think it makes sense to create the brainmask in the base from the cross sectionals POST editing. That should also help to solve these problems.
Best, Martin
On Tue, 2010-09-07 at 16:58 -0500, Derin Cobia wrote:
Hi Martin,
Thanks for the input. You are correct that the slices from the tp's to base are slightly off, this is because of the poor correspondence in slice numbering between them. And yes, there was still skull/dura several slices forward and back.
Are you sure that the -long runs don't need to be edited in this case? Even with clearly bad pial surfaces? I believe that the brainmask.mgz in the base is correct, however, my concern is that it is being poorly mapped to the longitudinals. There are multiple instances in this same subject where the brainmask.mgz differs between the -long and base; particularly, excess skull/dura in the -long runs, where it has been edited out in the base. What would cause this?
I like the idea of copying over the cross-sectional brainmask.mgz to initialize the longitudinal pipeline (as opposed to masking T1.mgz with norm_template.mgz). Is there a reason why this is not the default? This would make sense to me given the time/effort put into edits at tp1 - tpn, then they can just be pushed forward to the longitudinal runs. Maybe Nick or Doug could comment on it? I'll test it out and see. Thanks.
-Derin
On Sep 3, 2010, at 4:37 PM, Martin Reuter wrote:
Hi Derin,
as far as I can see your description is correct. Note: in your screenshots the base is shown in a different slice (but I believe that you probably scrolled through the image to make sure no skull is included in a different slice).
Actually in this situation, the long runs should not need editing, if the brainmask.mgz in the base is accurate. It is mapped to the longitudinals and used as a brainmask there.
Have you tried editing the brainmask in the longitudinal directory (or if done in the cross sectional you can simply copy it over as it is in the same space). Then pass -long .... and -autorecon2-pial -autorecon3? Not sure what this does or if it works, but it's worth a try. If it does not work, I guess we need to look at this data, to see what is going on.
Currently I am working on testing out the 5.0 fix and some other possible improvements...
Best, Martin
On Fri, 2010-09-03 at 13:54 -0500, Derin Cobia wrote:
Based on our understanding of the longitudinal pipeline for 4.5, image edits for the -long runs come mostly from the template/base subj (with an optional flag for CPs to be copied from the template; http://www.mail-archive.com/freesurfer@nmr.mgh.harvard.edu/msg13256.html, http://www.mail-archive.com/freesurfer@nmr.mgh.harvard.edu/msg12895.html). In addition, our assumption is that pial edits will come via the template, as the -long brainmask, is a result of masking the -long T1.mgz by the template brainmask.mgz (which itself is the norm_template.mgz, a combination of every time point's norm.mgz [which is pre-edit stage]).
We've run into an issue where our -long runs are still in need of pial editing, despite edits having been done to both the cross sectional and template runs (see attached). However, it's also our understanding that we should not edit the -long runs and try to re-run (using -autorecon2-pial -autorecon3), as this will pollute the correction for intrasubject variation (per communication from Martin). How would you recommend to proceed? Thanks.
-Derin
Freesurfer mailing list 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
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.
freesurfer@nmr.mgh.harvard.edu
-
Derin Cobia -
Martin Reuter