Hi all, I have a few subjects who have angled cutting planes. I ran talairach <subjid> and rerun the mri_fill and there is a very angled cut. A reason for this may be that the pons have been removed during white matter segmentation. The brain images have the pons intact but the wm images don't. I've run recon-all stage1, segment_subject, and process volume from the csurf interface. They all remove the pons. I've also tracked the coordinates of the pons in the brain volumes and used those in the mri_fill command (-P <x> <y> <z>). Is there a reason for this and a way to fix it? Thank you, Thang Nguyen
Hi Thang,
running talairach just before running the fill won't necessarily help, as it will added the talairach.xfm transform to the orig/COR-.info, which the fill (mri_fill) doesn't read. You can either add it by hand to the wm/COR-.info file (copies from the orig), or you can specify it directly on the command line using mri_fill -xform \ $SUBJECTS_DIR/$subject/mri/transforms/talairach.xfm ...
cheers, Bruce
On Fri, 22 Oct 2004, Nguyen, Thang Q wrote:
Hi all, I have a few subjects who have angled cutting planes. I ran talairach <subjid> and rerun the mri_fill and there is a very angled cut. A reason for this may be that the pons have been removed during white matter segmentation. The brain images have the pons intact but the wm images don't. I've run recon-all stage1, segment_subject, and process volume from the csurf interface. They all remove the pons. I've also tracked the coordinates of the pons in the brain volumes and used those in the mri_fill command (-P <x> <y> <z>). Is there a reason for this and a way to fix it? Thank you, Thang Nguyen
Hi Thang,
A while ago Tosa told me some tricks for dealing with these issues. Make sure you are using a relatively recent version of mri_fill as he fixed some fairly serious bugs (late last year / early this year) in this area.
I found Tosa's tips to be quite useful so I will summarize them.
First, confirmed that the talairach xform you are getting is reasonable.
A good way to do this is to take your original images and transform them through the transform and create a new volume you can look at.
mri_convert --apply_transform mri/transforms/taliarch.xfm mri/orig/ orig_tal.mgh
This creates the file orig_tal.mgh which you can look at in tkmedit. (Open main volume and choose orig_tal.mgh)
If the transform is good, you should see your subject with his head slightly tilted forward. If it is bad, usually it is pretty obvious.... it will be flipped, upside down or whatever.
If the talairach is good, make sure you find the position of the pons (for your -P) is in volume scanner RAS coordinates. To do this you may need to go to View -> Information and check volume scanner. That new triplet will appear down below.
If your talaraich is screwed up, you might try using T1 images for mritotal instead of orig. mritotal is the NMI tool that finds the talairach and is run at the beginning of stage1.
I did this successfully once by:
mri_convert T1 T1/T1.mnc mritotal -verbose -debug -clobber -protocol icbm T1/T1.mnc transforms/talairach.xfm
(see bin/noarch/talaiarch2... this is how it gets run from recon-all)
Once you have done this recheck the talairach with the above procedure.
I had a question for Bruce. Not that I really understand the data flow of FS like you do, but I would be worried specifying the xform explicitly to mri_fill because what if some other tool tried to use that wm/COR-.info. It seems like you want to use a command like mri_add_xform_to_header to make sure all your data remains 100% coherent.
While on this subject I have wondered about exactly what this command does other than insert "xform /path/foo.xfm" to COR-.info In particular, why does it take the destination path twice as in the script talaraich2?
Finally, is the order of parameters in the COR-.info files important? For example, does it expect xform to be right after tr te and ti? That might be useful to know someday.
Thank you, Ray Fix
On Oct 22, 2004, at 6:42 PM, Bruce Fischl wrote:
Hi Thang,
running talairach just before running the fill won't necessarily help, as it will added the talairach.xfm transform to the orig/COR-.info, which the fill (mri_fill) doesn't read. You can either add it by hand to the wm/COR-.info file (copies from the orig), or you can specify it directly on the command line using mri_fill -xform \ $SUBJECTS_DIR/$subject/mri/transforms/talairach.xfm ...
cheers, Bruce
On Fri, 22 Oct 2004, Nguyen, Thang Q wrote:
Hi all, I have a few subjects who have angled cutting planes. I ran talairach <subjid> and rerun the mri_fill and there is a very angled cut. A reason for this may be that the pons have been removed during white matter segmentation. The brain images have the pons intact but the wm images don't. I've run recon-all stage1, segment_subject, and process volume from the csurf interface. They all remove the pons. I've also tracked the coordinates of the pons in the brain volumes and used those in the mri_fill command (-P <x> <y> <z>). Is there a reason for this and a way to fix it? Thank you, Thang Nguyen
Freesurfer mailing list Freesurfer@nmr.mgh.harvard.edu https://mail.nmr.mgh.harvard.edu/mailman/listinfo/freesurfer
FYI: a better way to check and correct your talairach registration is with tkregister2. Run it like:
tkregister2 --subject yoursubject --fstal
A window will come up allowing you to flip between the subject's anatomical and the talairach brain to check the registration. You can also change the registration interactively. Run with --help for more info.
doug
Ray Fix wrote:
Hi Thang,
A while ago Tosa told me some tricks for dealing with these issues. Make sure you are using a relatively recent version of mri_fill as he fixed some fairly serious bugs (late last year / early this year) in this area.
I found Tosa's tips to be quite useful so I will summarize them.
First, confirmed that the talairach xform you are getting is reasonable.
A good way to do this is to take your original images and transform them through the transform and create a new volume you can look at.
mri_convert --apply_transform mri/transforms/taliarch.xfm mri/orig/ orig_tal.mgh
This creates the file orig_tal.mgh which you can look at in tkmedit. (Open main volume and choose orig_tal.mgh)
If the transform is good, you should see your subject with his head slightly tilted forward. If it is bad, usually it is pretty obvious.... it will be flipped, upside down or whatever.
If the talairach is good, make sure you find the position of the pons (for your -P) is in volume scanner RAS coordinates. To do this you may need to go to View -> Information and check volume scanner. That new triplet will appear down below.
If your talaraich is screwed up, you might try using T1 images for mritotal instead of orig. mritotal is the NMI tool that finds the talairach and is run at the beginning of stage1.
I did this successfully once by:
mri_convert T1 T1/T1.mnc mritotal -verbose -debug -clobber -protocol icbm T1/T1.mnc transforms/talairach.xfm
(see bin/noarch/talaiarch2... this is how it gets run from recon-all)
Once you have done this recheck the talairach with the above procedure.
I had a question for Bruce. Not that I really understand the data flow of FS like you do, but I would be worried specifying the xform explicitly to mri_fill because what if some other tool tried to use that wm/COR-.info. It seems like you want to use a command like mri_add_xform_to_header to make sure all your data remains 100% coherent.
While on this subject I have wondered about exactly what this command does other than insert "xform /path/foo.xfm" to COR-.info In particular, why does it take the destination path twice as in the script talaraich2?
Finally, is the order of parameters in the COR-.info files important? For example, does it expect xform to be right after tr te and ti? That might be useful to know someday.
Thank you, Ray Fix
On Oct 22, 2004, at 6:42 PM, Bruce Fischl wrote:
Hi Thang,
running talairach just before running the fill won't necessarily help, as it will added the talairach.xfm transform to the orig/COR-.info, which the fill (mri_fill) doesn't read. You can either add it by hand to the wm/COR-.info file (copies from the orig), or you can specify it directly on the command line using mri_fill -xform \ $SUBJECTS_DIR/$subject/mri/transforms/talairach.xfm ...
cheers, Bruce
On Fri, 22 Oct 2004, Nguyen, Thang Q wrote:
Hi all, I have a few subjects who have angled cutting planes. I ran talairach <subjid> and rerun the mri_fill and there is a very angled cut. A reason for this may be that the pons have been removed during white matter segmentation. The brain images have the pons intact but the wm images don't. I've run recon-all stage1, segment_subject, and process volume from the csurf interface. They all remove the pons. I've also tracked the coordinates of the pons in the brain volumes and used those in the mri_fill command (-P <x> <y> <z>). Is there a reason for this and a way to fix it? Thank you, Thang Nguyen
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
freesurfer@nmr.mgh.harvard.edu
-
Bruce Fischl -
Doug Greve -
Nguyen, Thang Q -
Ray Fix