Hello FreeSurfer Folks-
I've seen several threads in the archives discussing the expected inaccuracies in the pial, white, and orig surfaces running through subcortical structures. However, I've been unable to ascertain exactly how much and what kind of inaccuracy is acceptable. It makes sense to me that if the entirety of a structure, say the amygdala, were included or excluded in the pial surface, then it would be easy to adjust later measurements. However, what we see in all of our subjects is a random pattern of half in/half out, with weird shapes, when it comes to the surfaces running in and around subcortical structures. For example, take a look the image at this link.
http://www.tc.umn.edu/~norb0062/Amygdala.tiff
There you will see the how the hippocampus and amygdala are both half included in the pial surface. Item 1 has a strange guitar-shaped chunk taken out of the amygdala, and item 2 has a triangle and a peanut-shaped area removed. Are these the types of expected and accepted inaccuracies, or is there something afoul here that needs to be corrected?
I have one other question regarding the interactions of aseg and surface information. Item 3 in the image points out a section of apparent subdural matter that has been classified as grey matter in the aseg but has been excluded from the pial surface. When the final statistics are spit out of recon-all, is there a section of processing that says something like, "Based on the aseg values my grey matter volume is X, but since Y of that is outside of the pial surface I'll call my total volume X-Y"? Or, is this another item to that requires correction and re-running of recon-all?
Many thanks in advance,
Hi Jim,
you can actually run mris_make_surfaces with -fix_mtl and it will inhibit all deformations in these regions. I didn't enable that by default as I was afraid it would mean missing a bit of entorhinal. In general it doesn't really matter as the gray and white volumes computed from the surfaces will exclude amygdala and hippocampal voxels that are interior to the surface.
For your second question: we use the gray matter volume estimated from the surfaces instead of the aseg, as we find it to be more accurate than the voxel-based techniques.
cheers, Bruce
On Tue, 27 Nov 2007, James N. Porter wrote:
Hello FreeSurfer Folks-
I've seen several threads in the archives discussing the expected inaccuracies in the pial, white, and orig surfaces running through subcortical structures. However, I've been unable to ascertain exactly how much and what kind of inaccuracy is acceptable. It makes sense to me that if the entirety of a structure, say the amygdala, were included or excluded in the pial surface, then it would be easy to adjust later measurements. However, what we see in all of our subjects is a random pattern of half in/half out, with weird shapes, when it comes to the surfaces running in and around subcortical structures. For example, take a look the image at this link.
http://www.tc.umn.edu/~norb0062/Amygdala.tiff
There you will see the how the hippocampus and amygdala are both half included in the pial surface. Item 1 has a strange guitar-shaped chunk taken out of the amygdala, and item 2 has a triangle and a peanut-shaped area removed. Are these the types of expected and accepted inaccuracies, or is there something afoul here that needs to be corrected?
I have one other question regarding the interactions of aseg and surface information. Item 3 in the image points out a section of apparent subdural matter that has been classified as grey matter in the aseg but has been excluded from the pial surface. When the final statistics are spit out of recon-all, is there a section of processing that says something like, "Based on the aseg values my grey matter volume is X, but since Y of that is outside of the pial surface I'll call my total volume X-Y"? Or, is this another item to that requires correction and re-running of recon-all?
Many thanks in advance,
Hello Bruce-
From the wiki, I gather that I would rerun recon-all from -finalsurfs onward, as that is the step that calls make_final_surfaces. In order to insert the fix_mtl flag I'll have to run that section via a separate command line outside of recon-all. Are there any other subsequent steps that require manual intervention or nonstandard parameters, or would something like the code below do the work?
for S in $SUBJ; do for H in $HEMI; do mris_make_surfaces --noaparc --mgz -T1 --fix_mtl brain.finalsurfs $S $H done recon-all -s $S -cortribbon -autorecon3 done
Thanks,
Jim Porter Graduate Student Clinical Science & Psychopathology Research University of Minnesota
Bruce Fischl wrote:
Hi Jim,
you can actually run mris_make_surfaces with -fix_mtl and it will inhibit all deformations in these regions. I didn't enable that by default as I was afraid it would mean missing a bit of entorhinal. In general it doesn't really matter as the gray and white volumes computed from the surfaces will exclude amygdala and hippocampal voxels that are interior to the surface.
For your second question: we use the gray matter volume estimated from the surfaces instead of the aseg, as we find it to be more accurate than the voxel-based techniques.
cheers, Bruce
On Tue, 27 Nov 2007, James N. Porter wrote:
Hello FreeSurfer Folks-
I've seen several threads in the archives discussing the expected inaccuracies in the pial, white, and orig surfaces running through subcortical structures. However, I've been unable to ascertain exactly how much and what kind of inaccuracy is acceptable. It makes sense to me that if the entirety of a structure, say the amygdala, were included or excluded in the pial surface, then it would be easy to adjust later measurements. However, what we see in all of our subjects is a random pattern of half in/half out, with weird shapes, when it comes to the surfaces running in and around subcortical structures. For example, take a look the image at this link.
http://www.tc.umn.edu/~norb0062/Amygdala.tiff
There you will see the how the hippocampus and amygdala are both half included in the pial surface. Item 1 has a strange guitar-shaped chunk taken out of the amygdala, and item 2 has a triangle and a peanut-shaped area removed. Are these the types of expected and accepted inaccuracies, or is there something afoul here that needs to be corrected?
I have one other question regarding the interactions of aseg and surface information. Item 3 in the image points out a section of apparent subdural matter that has been classified as grey matter in the aseg but has been excluded from the pial surface. When the final statistics are spit out of recon-all, is there a section of processing that says something like, "Based on the aseg values my grey matter volume is X, but since Y of that is outside of the pial surface I'll call my total volume X-Y"? Or, is this another item to that requires correction and re-running of recon-all?
Many thanks in advance,
Hi Jim,
yes, I think that should be fine
Bruce On Mon, 10 Dec 2007, James N. Porter wrote:
Hello Bruce-
From the wiki, I gather that I would rerun recon-all from -finalsurfs onward, as that is the step that calls make_final_surfaces. In order to insert the fix_mtl flag I'll have to run that section via a separate command line outside of recon-all. Are there any other subsequent steps that require manual intervention or nonstandard parameters, or would something like the code below do the work?
for S in $SUBJ; do for H in $HEMI; do mris_make_surfaces --noaparc --mgz -T1 --fix_mtl brain.finalsurfs $S $H done recon-all -s $S -cortribbon -autorecon3 done
Thanks,
Jim Porter Graduate Student Clinical Science & Psychopathology Research University of Minnesota
Bruce Fischl wrote:
Hi Jim,
you can actually run mris_make_surfaces with -fix_mtl and it will inhibit all deformations in these regions. I didn't enable that by default as I was afraid it would mean missing a bit of entorhinal. In general it doesn't really matter as the gray and white volumes computed from the surfaces will exclude amygdala and hippocampal voxels that are interior to the surface.
For your second question: we use the gray matter volume estimated from the surfaces instead of the aseg, as we find it to be more accurate than the voxel-based techniques.
cheers, Bruce
On Tue, 27 Nov 2007, James N. Porter wrote:
Hello FreeSurfer Folks-
I've seen several threads in the archives discussing the expected inaccuracies in the pial, white, and orig surfaces running through subcortical structures. However, I've been unable to ascertain exactly how much and what kind of inaccuracy is acceptable. It makes sense to me that if the entirety of a structure, say the amygdala, were included or excluded in the pial surface, then it would be easy to adjust later measurements. However, what we see in all of our subjects is a random pattern of half in/half out, with weird shapes, when it comes to the surfaces running in and around subcortical structures. For example, take a look the image at this link.
http://www.tc.umn.edu/~norb0062/Amygdala.tiff
There you will see the how the hippocampus and amygdala are both half included in the pial surface. Item 1 has a strange guitar-shaped chunk taken out of the amygdala, and item 2 has a triangle and a peanut-shaped area removed. Are these the types of expected and accepted inaccuracies, or is there something afoul here that needs to be corrected?
I have one other question regarding the interactions of aseg and surface information. Item 3 in the image points out a section of apparent subdural matter that has been classified as grey matter in the aseg but has been excluded from the pial surface. When the final statistics are spit out of recon-all, is there a section of processing that says something like, "Based on the aseg values my grey matter volume is X, but since Y of that is outside of the pial surface I'll call my total volume X-Y"? Or, is this another item to that requires correction and re-running of recon-all?
Many thanks in advance,
Hi, So if the volumes are estimated from the surfaces and not aseg.mgz, does this mean it is unnecessary to edit aseg.mgz? Or is it still necessary, but only for the subcortical structures? Is it best to make pial surface/wm.mgz edits and control points for grey and white matter, and aseg.mgz for the subcortical volumes? I hope I am being clear enough.
Thanks, Chris
Bruce Fischl wrote:
Hi Jim,
yes, I think that should be fine
Bruce On Mon, 10 Dec 2007, James N. Porter wrote:
Hello Bruce-
From the wiki, I gather that I would rerun recon-all from -finalsurfs onward, as that is the step that calls make_final_surfaces. In order to insert the fix_mtl flag I'll have to run that section via a separate command line outside of recon-all. Are there any other subsequent steps that require manual intervention or nonstandard parameters, or would something like the code below do the work?
for S in $SUBJ; do for H in $HEMI; do mris_make_surfaces --noaparc --mgz -T1 --fix_mtl brain.finalsurfs $S $H done recon-all -s $S -cortribbon -autorecon3 done
Thanks,
Jim Porter Graduate Student Clinical Science & Psychopathology Research University of Minnesota
Bruce Fischl wrote:
Hi Jim,
you can actually run mris_make_surfaces with -fix_mtl and it will inhibit all deformations in these regions. I didn't enable that by default as I was afraid it would mean missing a bit of entorhinal. In general it doesn't really matter as the gray and white volumes computed from the surfaces will exclude amygdala and hippocampal voxels that are interior to the surface.
For your second question: we use the gray matter volume estimated from the surfaces instead of the aseg, as we find it to be more accurate than the voxel-based techniques.
cheers, Bruce
On Tue, 27 Nov 2007, James N. Porter wrote:
Hello FreeSurfer Folks-
I've seen several threads in the archives discussing the expected inaccuracies in the pial, white, and orig surfaces running through subcortical structures. However, I've been unable to ascertain exactly how much and what kind of inaccuracy is acceptable. It makes sense to me that if the entirety of a structure, say the amygdala, were included or excluded in the pial surface, then it would be easy to adjust later measurements. However, what we see in all of our subjects is a random pattern of half in/half out, with weird shapes, when it comes to the surfaces running in and around subcortical structures. For example, take a look the image at this link.
http://www.tc.umn.edu/~norb0062/Amygdala.tiff
There you will see the how the hippocampus and amygdala are both half included in the pial surface. Item 1 has a strange guitar-shaped chunk taken out of the amygdala, and item 2 has a triangle and a peanut-shaped area removed. Are these the types of expected and accepted inaccuracies, or is there something afoul here that needs to be corrected?
I have one other question regarding the interactions of aseg and surface information. Item 3 in the image points out a section of apparent subdural matter that has been classified as grey matter in the aseg but has been excluded from the pial surface. When the final statistics are spit out of recon-all, is there a section of processing that says something like, "Based on the aseg values my grey matter volume is X, but since Y of that is outside of the pial surface I'll call my total volume X-Y"? Or, is this another item to that requires correction and re-running of recon-all?
Many thanks in advance,
Freesurfer mailing list Freesurfer@nmr.mgh.harvard.edu https://mail.nmr.mgh.harvard.edu/mailman/listinfo/freesurfer
yes, although the white matter volume that comes from mris_wm_volume is affected by the aseg (since non-wm voxels that are interior to the ?h.white surface are removed from the volume calculation)
cheers, Bruce On Mon, 10 Dec 2007, Chris Watson wrote:
Hi, So if the volumes are estimated from the surfaces and not aseg.mgz, does this mean it is unnecessary to edit aseg.mgz? Or is it still necessary, but only for the subcortical structures? Is it best to make pial surface/wm.mgz edits and control points for grey and white matter, and aseg.mgz for the subcortical volumes? I hope I am being clear enough.
Thanks, Chris
Bruce Fischl wrote:
Hi Jim,
yes, I think that should be fine
Bruce On Mon, 10 Dec 2007, James N. Porter wrote:
Hello Bruce-
From the wiki, I gather that I would rerun recon-all from -finalsurfs onward, as that is the step that calls make_final_surfaces. In order to insert the fix_mtl flag I'll have to run that section via a separate command line outside of recon-all. Are there any other subsequent steps that require manual intervention or nonstandard parameters, or would something like the code below do the work?
for S in $SUBJ; do for H in $HEMI; do mris_make_surfaces --noaparc --mgz -T1 --fix_mtl brain.finalsurfs $S $H done recon-all -s $S -cortribbon -autorecon3 done
Thanks,
Jim Porter Graduate Student Clinical Science & Psychopathology Research University of Minnesota
Bruce Fischl wrote:
Hi Jim,
you can actually run mris_make_surfaces with -fix_mtl and it will inhibit all deformations in these regions. I didn't enable that by default as I was afraid it would mean missing a bit of entorhinal. In general it doesn't really matter as the gray and white volumes computed from the surfaces will exclude amygdala and hippocampal voxels that are interior to the surface.
For your second question: we use the gray matter volume estimated from the surfaces instead of the aseg, as we find it to be more accurate than the voxel-based techniques.
cheers, Bruce
On Tue, 27 Nov 2007, James N. Porter wrote:
Hello FreeSurfer Folks-
I've seen several threads in the archives discussing the expected inaccuracies in the pial, white, and orig surfaces running through subcortical structures. However, I've been unable to ascertain exactly how much and what kind of inaccuracy is acceptable. It makes sense to me that if the entirety of a structure, say the amygdala, were included or excluded in the pial surface, then it would be easy to adjust later measurements. However, what we see in all of our subjects is a random pattern of half in/half out, with weird shapes, when it comes to the surfaces running in and around subcortical structures. For example, take a look the image at this link.
http://www.tc.umn.edu/~norb0062/Amygdala.tiff
There you will see the how the hippocampus and amygdala are both half included in the pial surface. Item 1 has a strange guitar-shaped chunk taken out of the amygdala, and item 2 has a triangle and a peanut-shaped area removed. Are these the types of expected and accepted inaccuracies, or is there something afoul here that needs to be corrected?
I have one other question regarding the interactions of aseg and surface information. Item 3 in the image points out a section of apparent subdural matter that has been classified as grey matter in the aseg but has been excluded from the pial surface. When the final statistics are spit out of recon-all, is there a section of processing that says something like, "Based on the aseg values my grey matter volume is X, but since Y of that is outside of the pial surface I'll call my total volume X-Y"? Or, is this another item to that requires correction and re-running of recon-all?
Many thanks in advance,
Freesurfer mailing list Freesurfer@nmr.mgh.harvard.edu https://mail.nmr.mgh.harvard.edu/mailman/listinfo/freesurfer
Jim,
There is a way to insert special flags into the recon-all command stream for a particular binary. See the section 'EXPERT OPTIONS FILE' in recon-all --help. So you would create a file called expert.opts containing:
mris_make_surfaces -fix_mtl
and then add the string '-expert expert.opts' to recon-all.
You could add -dontrun to recon-all to confirm that it is grabbing that change.
Nick
On Mon, 2007-12-10 at 13:32 -0600, James N. Porter wrote:
Hello Bruce-
From the wiki, I gather that I would rerun recon-all from -finalsurfs onward, as that is the step that calls make_final_surfaces. In order to insert the fix_mtl flag I'll have to run that section via a separate command line outside of recon-all. Are there any other subsequent steps that require manual intervention or nonstandard parameters, or would something like the code below do the work?
for S in $SUBJ; do for H in $HEMI; do mris_make_surfaces --noaparc --mgz -T1 --fix_mtl brain.finalsurfs $S $H done recon-all -s $S -cortribbon -autorecon3 done
Thanks,
Jim Porter Graduate Student Clinical Science & Psychopathology Research University of Minnesota
Bruce Fischl wrote:
Hi Jim,
you can actually run mris_make_surfaces with -fix_mtl and it will inhibit all deformations in these regions. I didn't enable that by default as I was afraid it would mean missing a bit of entorhinal. In general it doesn't really matter as the gray and white volumes computed from the surfaces will exclude amygdala and hippocampal voxels that are interior to the surface.
For your second question: we use the gray matter volume estimated from the surfaces instead of the aseg, as we find it to be more accurate than the voxel-based techniques.
cheers, Bruce
On Tue, 27 Nov 2007, James N. Porter wrote:
Hello FreeSurfer Folks-
I've seen several threads in the archives discussing the expected inaccuracies in the pial, white, and orig surfaces running through subcortical structures. However, I've been unable to ascertain exactly how much and what kind of inaccuracy is acceptable. It makes sense to me that if the entirety of a structure, say the amygdala, were included or excluded in the pial surface, then it would be easy to adjust later measurements. However, what we see in all of our subjects is a random pattern of half in/half out, with weird shapes, when it comes to the surfaces running in and around subcortical structures. For example, take a look the image at this link.
http://www.tc.umn.edu/~norb0062/Amygdala.tiff
There you will see the how the hippocampus and amygdala are both half included in the pial surface. Item 1 has a strange guitar-shaped chunk taken out of the amygdala, and item 2 has a triangle and a peanut-shaped area removed. Are these the types of expected and accepted inaccuracies, or is there something afoul here that needs to be corrected?
I have one other question regarding the interactions of aseg and surface information. Item 3 in the image points out a section of apparent subdural matter that has been classified as grey matter in the aseg but has been excluded from the pial surface. When the final statistics are spit out of recon-all, is there a section of processing that says something like, "Based on the aseg values my grey matter volume is X, but since Y of that is outside of the pial surface I'll call my total volume X-Y"? Or, is this another item to that requires correction and re-running of recon-all?
Many thanks in advance,
Freesurfer mailing list Freesurfer@nmr.mgh.harvard.edu https://mail.nmr.mgh.harvard.edu/mailman/listinfo/freesurfer
freesurfer@nmr.mgh.harvard.edu
-
Bruce Fischl -
Chris Watson -
James N. Porter -
Nick Schmansky