Hi,
I am trying to mask a non-brain tissue left by watershed based on the cortical surfaces. I found that all what I need to be left is in aparc+aseg.mgz, so I converted it to nifiti, along with brainmask.mgz and used FSL to binarize aparc+aseg and mask the brainmask. I found that at many places the results is too conservative and too much of the GM is missing, see the attached images, and particularly the temporal lobes.
Is there a better approach to what I am trying to do and what could be the explanation to the results I get.
Thanks,
Martin
This looks like a problem we had with an older version. What version of freesurfer are you running?
Martin Kavec wrote:
Hi,
I am trying to mask a non-brain tissue left by watershed based on the cortical surfaces. I found that all what I need to be left is in aparc+aseg.mgz, so I converted it to nifiti, along with brainmask.mgz and used FSL to binarize aparc+aseg and mask the brainmask. I found that at many places the results is too conservative and too much of the GM is missing, see the attached images, and particularly the temporal lobes.
Is there a better approach to what I am trying to do and what could be the explanation to the results I get.
Thanks,
Martin
Freesurfer mailing list Freesurfer@nmr.mgh.harvard.edu https://mail.nmr.mgh.harvard.edu/mailman/listinfo/freesurfer
Hi Dough,
this data were analyzed using 4.0.4. So indeed it not the latest, which I am running normally. From which version has this been fixed, so I can check, before re-runing. From which point should I rerun the analysis.
Thanks,
Martin
On Monday 06 April 2009 18:42:36 Douglas N Greve wrote:
This looks like a problem we had with an older version. What version of freesurfer are you running?
Martin Kavec wrote:
Hi,
I am trying to mask a non-brain tissue left by watershed based on the cortical surfaces. I found that all what I need to be left is in aparc+aseg.mgz, so I converted it to nifiti, along with brainmask.mgz and used FSL to binarize aparc+aseg and mask the brainmask. I found that at many places the results is too conservative and too much of the GM is missing, see the attached images, and particularly the temporal lobes.
Is there a better approach to what I am trying to do and what could be the explanation to the results I get.
Thanks,
Martin
I thought it was 4.0.3, so maybe that's not it. Can you look at the ribbon.mgz file? You can load it as a segmentation, ie,
tkmedit subject orig.mgz -seg ribbon.mgz
aparc+aseg is suppossed to inherit the cortex from ribbon.mgz
doug
On Mon, 6 Apr 2009, Martin Kavec wrote:
Hi Dough,
this data were analyzed using 4.0.4. So indeed it not the latest, which I am running normally. From which version has this been fixed, so I can check, before re-runing. From which point should I rerun the analysis.
Thanks,
Martin
On Monday 06 April 2009 18:42:36 Douglas N Greve wrote:
This looks like a problem we had with an older version. What version of freesurfer are you running?
Martin Kavec wrote:
Hi,
I am trying to mask a non-brain tissue left by watershed based on the cortical surfaces. I found that all what I need to be left is in aparc+aseg.mgz, so I converted it to nifiti, along with brainmask.mgz and used FSL to binarize aparc+aseg and mask the brainmask. I found that at many places the results is too conservative and too much of the GM is missing, see the attached images, and particularly the temporal lobes.
Is there a better approach to what I am trying to do and what could be the explanation to the results I get.
Thanks,
Martin
Doug,
ribbon (aparc+aseg) looks perfect, and corresponds to the surfaces ?h.white.
Thanks,
Martin
On Monday 06 April 2009 22:17:18 Doug Greve wrote:
I thought it was 4.0.3, so maybe that's not it. Can you look at the ribbon.mgz file? You can load it as a segmentation, ie,
tkmedit subject orig.mgz -seg ribbon.mgz
aparc+aseg is suppossed to inherit the cortex from ribbon.mgz
doug
On Mon, 6 Apr 2009, Martin Kavec wrote:
Hi Dough,
this data were analyzed using 4.0.4. So indeed it not the latest, which I am running normally. From which version has this been fixed, so I can check, before re-runing. From which point should I rerun the analysis.
Thanks,
Martin
On Monday 06 April 2009 18:42:36 Douglas N Greve wrote:
This looks like a problem we had with an older version. What version of freesurfer are you running?
Martin Kavec wrote:
Hi,
I am trying to mask a non-brain tissue left by watershed based on the cortical surfaces. I found that all what I need to be left is in aparc+aseg.mgz, so I converted it to nifiti, along with brainmask.mgz and used FSL to binarize aparc+aseg and mask the brainmask. I found that at many places the results is too conservative and too much of the GM is missing, see the attached images, and particularly the temporal lobes.
Is there a better approach to what I am trying to do and what could be the explanation to the results I get.
Thanks,
Martin
but you are using aparc+aseg to generate the mask, so if aparc+aseg is perfect, the mask should be as well. Do the mask and aparc+aseg match?
On Mon, 6 Apr 2009, Martin Kavec wrote:
Doug,
ribbon (aparc+aseg) looks perfect, and corresponds to the surfaces ?h.white.
Thanks,
Martin
On Monday 06 April 2009 22:17:18 Doug Greve wrote:
I thought it was 4.0.3, so maybe that's not it. Can you look at the ribbon.mgz file? You can load it as a segmentation, ie,
tkmedit subject orig.mgz -seg ribbon.mgz
aparc+aseg is suppossed to inherit the cortex from ribbon.mgz
doug
On Mon, 6 Apr 2009, Martin Kavec wrote:
Hi Dough,
this data were analyzed using 4.0.4. So indeed it not the latest, which I am running normally. From which version has this been fixed, so I can check, before re-runing. From which point should I rerun the analysis.
Thanks,
Martin
On Monday 06 April 2009 18:42:36 Douglas N Greve wrote:
This looks like a problem we had with an older version. What version of freesurfer are you running?
Martin Kavec wrote:
Hi,
I am trying to mask a non-brain tissue left by watershed based on the cortical surfaces. I found that all what I need to be left is in aparc+aseg.mgz, so I converted it to nifiti, along with brainmask.mgz and used FSL to binarize aparc+aseg and mask the brainmask. I found that at many places the results is too conservative and too much of the GM is missing, see the attached images, and particularly the temporal lobes.
Is there a better approach to what I am trying to do and what could be the explanation to the results I get.
Thanks,
Martin
Hi Dough,
I took another subject, which was analyzed recently with the latest FS and the result of the masking is gook!
I found another issue in my experimenting!
Here is exactly what I did: 1. mri_convert -i brainmask.mgz -o brainmask.nii.gz 2. fslswapdim brainmask.nii.gz x z -y brainmask.axial.nii.gz (reslice to axial)
Now when I load brainmask.mgz (or brainmask.nii.gz) and brainmask.axial.nii.gz to tkmedit there is a 1 voxel shift in inferior to superior direction.
Can you reproduce that and do you know where it comes from?
Thanks,
Martin
On Monday 06 April 2009 23:13:48 Doug Greve wrote:
but you are using aparc+aseg to generate the mask, so if aparc+aseg is perfect, the mask should be as well. Do the mask and aparc+aseg match?
On Mon, 6 Apr 2009, Martin Kavec wrote:
Doug,
ribbon (aparc+aseg) looks perfect, and corresponds to the surfaces ?h.white.
Thanks,
Martin
On Monday 06 April 2009 22:17:18 Doug Greve wrote:
I thought it was 4.0.3, so maybe that's not it. Can you look at the ribbon.mgz file? You can load it as a segmentation, ie,
tkmedit subject orig.mgz -seg ribbon.mgz
aparc+aseg is suppossed to inherit the cortex from ribbon.mgz
doug
On Mon, 6 Apr 2009, Martin Kavec wrote:
Hi Dough,
this data were analyzed using 4.0.4. So indeed it not the latest, which I am running normally. From which version has this been fixed, so I can check, before re-runing. From which point should I rerun the analysis.
Thanks,
Martin
On Monday 06 April 2009 18:42:36 Douglas N Greve wrote:
This looks like a problem we had with an older version. What version of freesurfer are you running?
Martin Kavec wrote:
Hi,
I am trying to mask a non-brain tissue left by watershed based on the cortical surfaces. I found that all what I need to be left is in aparc+aseg.mgz, so I converted it to nifiti, along with brainmask.mgz and used FSL to binarize aparc+aseg and mask the brainmask. I found that at many places the results is too conservative and too much of the GM is missing, see the attached images, and particularly the temporal lobes.
Is there a better approach to what I am trying to do and what could be the explanation to the results I get.
Thanks,
Martin
Hi Martin, I don't know as I don't use fslswapdim on a regular basis. Sounds like the geometry is not being computed correctly in the new file. In any event, it seems more like an fslswapdim issue that a freesurfer one.
doug
Martin Kavec wrote:
Hi Dough,
I took another subject, which was analyzed recently with the latest FS and the result of the masking is gook!
I found another issue in my experimenting!
Here is exactly what I did:
- mri_convert -i brainmask.mgz -o brainmask.nii.gz
- fslswapdim brainmask.nii.gz x z -y brainmask.axial.nii.gz (reslice to
axial)
Now when I load brainmask.mgz (or brainmask.nii.gz) and brainmask.axial.nii.gz to tkmedit there is a 1 voxel shift in inferior to superior direction.
Can you reproduce that and do you know where it comes from?
Thanks,
Martin
On Monday 06 April 2009 23:13:48 Doug Greve wrote:
but you are using aparc+aseg to generate the mask, so if aparc+aseg is perfect, the mask should be as well. Do the mask and aparc+aseg match?
On Mon, 6 Apr 2009, Martin Kavec wrote:
Doug,
ribbon (aparc+aseg) looks perfect, and corresponds to the surfaces ?h.white.
Thanks,
Martin
On Monday 06 April 2009 22:17:18 Doug Greve wrote:
I thought it was 4.0.3, so maybe that's not it. Can you look at the ribbon.mgz file? You can load it as a segmentation, ie,
tkmedit subject orig.mgz -seg ribbon.mgz
aparc+aseg is suppossed to inherit the cortex from ribbon.mgz
doug
On Mon, 6 Apr 2009, Martin Kavec wrote:
Hi Dough,
this data were analyzed using 4.0.4. So indeed it not the latest, which I am running normally. From which version has this been fixed, so I can check, before re-runing. From which point should I rerun the analysis.
Thanks,
Martin
On Monday 06 April 2009 18:42:36 Douglas N Greve wrote:
This looks like a problem we had with an older version. What version of freesurfer are you running?
Martin Kavec wrote:
> Hi, > > I am trying to mask a non-brain tissue left by watershed based on the > cortical surfaces. I found that all what I need to be left is in > aparc+aseg.mgz, so I converted it to nifiti, along with brainmask.mgz > and used FSL to binarize aparc+aseg and mask the brainmask. I found > that at many places the results is too conservative and too much of > the GM is missing, see the attached images, and particularly the > temporal lobes. > > Is there a better approach to what I am trying to do and what could > be the explanation to the results I get. > > Thanks, > > Martin >
Hi all,
I have two problems here.
1. talairach.xfm By browsing through the mailing list, it seems that talairach.xfm was not used much in the recon-all process which is quite different from the table given at https://surfer.nmr.mgh.harvard.edu/fswiki/ReconAllFilesVsSteps. We did find many subjects with bad talairach transformation but reasonable segmentation and surface reconstruction (version 3.05). Are the results (volumes and cortical thickness) for those subjects still trustworthy? Can you kindly clarify the importance of talairach.xfm in the pipeline? And since the new method of eTIV uses this transformation, is there any way to ensure manual registration is fine enough to be used for this estimation, if I have to manually register those subjects mentioned above?
2. recon-all -long We processed our data with the longitudinal pipeline. So far we only found 6 subjects failed among 200 subjects (tp1 corrected). All these 6 subjects seemed to have problem during the registration from tp2 to tp1. I used the command "tkmedit -f tp2.long.tp1/mri/orig.mgz -aux tp2.long.tp1/mri/orig_tp1_to_tp2.long.tp1.mgz" and realized they were not aligned. I have tried different cost functions using "fsl_rigid_registration". All of them gave the same result. Is there any way I can do the manual registration? I used "fsl_rigid_registration" because no matter which cost function I specified to recon-all -long, it would call "fsl_rigid_registration" with corratio as the cost function.
Thanks in advance, Zheng Hui
Hi Zheng,
I'll leave 2 for Nick. As for 1, we don't use the tal very much, and a bad tal will not affect your thickness estimates. The newer versions rarely have a bad tal since we changed over to Avi Synder's code (which also fixes the problems we had with eTIV). You can manually correct it using tkregister2, although this is probably only needed if you want to report tal coords (or average in tal space).
cheers, Bruce
On Wed, 8 Apr 2009, gmszh wrote:
Hi all,
I have two problems here.
- talairach.xfm By browsing through the mailing list, it seems that talairach.xfm was
not used much in the recon-all process which is quite different from the table given at https://surfer.nmr.mgh.harvard.edu/fswiki/ReconAllFilesVsSteps. We did find many subjects with bad talairach transformation but reasonable segmentation and surface reconstruction (version 3.05). Are the results (volumes and cortical thickness) for those subjects still trustworthy? Can you kindly clarify the importance of talairach.xfm in the pipeline? And since the new method of eTIV uses this transformation, is there any way to ensure manual registration is fine enough to be used for this estimation, if I have to manually register those subjects mentioned above?
- recon-all -long We processed our data with the longitudinal pipeline. So far we only
found 6 subjects failed among 200 subjects (tp1 corrected). All these 6 subjects seemed to have problem during the registration from tp2 to tp1. I used the command "tkmedit -f tp2.long.tp1/mri/orig.mgz -aux tp2.long.tp1/mri/orig_tp1_to_tp2.long.tp1.mgz" and realized they were not aligned. I have tried different cost functions using "fsl_rigid_registration". All of them gave the same result. Is there any way I can do the manual registration? I used "fsl_rigid_registration" because no matter which cost function I specified to recon-all -long, it would call "fsl_rigid_registration" with corratio as the cost function.
Thanks in advance, Zheng Hui
Freesurfer mailing list Freesurfer@nmr.mgh.harvard.edu https://mail.nmr.mgh.harvard.edu/mailman/listinfo/freesurfer
freesurfer@nmr.mgh.harvard.edu