Hello,
I'm trying to generate surface meshes for different brain regions that can be imported into a software like MeshLab or Blender (i.e. .STL, .PLY, .OBJ, etc.). I ran recon-all on my T1 images, and the parcellation/segmentation looks fine. However, I noticed that when I imported this data into 3DSlicer, the volumes of these regions did not match the volumes in the aseg.stats file. In some cases, the differences were quite substantial. I tried removing all smoothing filters in 3DSlicer but the outcome was the same.
Do you know what could be going wrong here? Alternatively, is there an easier way to generate surface meshes from the aparc or aseg atlases?
I'm using the ICBM 2009c Nonlinear Asymmetric template ( http://nist.mni.mcgill.ca/?p=904) as a generic brain volume (though alternative suggestions are welcome).
Thanks! Jonah
External Email - Use Caution
Could you clarify what kind of meshes you are working with? Are the meshes you are working with plane-like (parts of the pial surface), or are they closed (e.g., generating from running marching cubes on a set of cortex voxels)?
Also what is your intention, i.e., what do you want to do with the meshes?
The exact volume inside the mesh will always depend on the parameters of the mesh generation algorithm.
Best,
Tim
On April 23, 2020 at 1:14 AM "Peter, Jonah" jonahpeter@g.harvard.edu wrote:
Hello,
I'm trying to generate surface meshes for different brain regions that can be imported into a software like MeshLab or Blender (i.e. .STL, .PLY, .OBJ, etc.). I ran recon-all on my T1 images, and the parcellation/segmentation looks fine. However, I noticed that when I imported this data into 3DSlicer, the volumes of these regions did not match the volumes in the aseg.stats file. In some cases, the differences were quite substantial. I tried removing all smoothing filters in 3DSlicer but the outcome was the same.
Do you know what could be going wrong here? Alternatively, is there an easier way to generate surface meshes from the aparc or aseg atlases?
I'm using the ICBM 2009c Nonlinear Asymmetric template ( http://nist.mni.mcgill.ca/?p=904) as a generic brain volume (though alternative suggestions are welcome).
Thanks! Jonah
-- Jonah Peter Graduate Student in The Biophysics Program Harvard University P: 646-306-0848 E: jonahpeter@g.harvard.edu _______________________________________________ Freesurfer mailing list Freesurfer@nmr.mgh.harvard.edu https://mail.nmr.mgh.harvard.edu/mailman/listinfo/freesurfer
Hi Jonah
when we compute volumes we typically use a partial-volume model, which we have found improves accuracy and repeatability. That may be the source of your discrepancy
cheers Bruce On Wed, 22 Apr 2020, Peter, Jonah wrote:
Hello, I'm trying to generate surface meshes for different brain regions that can be imported into a software like MeshLab or Blender (i.e. .STL, .PLY, .OBJ, etc.). I ran recon-all on my T1 images, and the parcellation/segmentation looks fine. However, I noticed that when I imported this data into 3DSlicer, the volumes of these regions did not match the volumes in the aseg.stats file. In some cases, the differences were quite substantial. I tried removing all smoothing filters in 3DSlicer but the outcome was the same.
Do you know what could be going wrong here? Alternatively, is there an easier way to generate surface meshes from the aparc or aseg atlases?
I'm using the ICBM 2009c Nonlinear Asymmetric template (http://nist.mni.mcgill.ca/?p=904) as a generic brain volume (though alternative suggestions are welcome).
Thanks! Jonah
-- Jonah PeterGraduate Student in The Biophysics ProgramHarvard University P: 646-306-0848 E: jonahpeter@g.harvard.edu
Hi all,
Thanks for the response. I'm familiar with partial volume effects within the context of PET imaging, but wouldn't you expect these effects to be negligible in MRI for structures on the order of mm?
With respect to the mesh type, I'd like to get closed meshes for each segmentation/parcellation.
Is there a way to do this in freesurfer?
On Thu, Apr 23, 2020 at 10:44 AM Bruce Fischl fischl@nmr.mgh.harvard.edu wrote:
Hi Jonah
when we compute volumes we typically use a partial-volume model, which we have found improves accuracy and repeatability. That may be the source of your discrepancy
cheers Bruce On Wed, 22 Apr 2020, Peter, Jonah wrote:
Hello, I'm trying to generate surface meshes for different brain regions that
can
be imported into a software like MeshLab or Blender (i.e. .STL, .PLY,
.OBJ,
etc.). I ran recon-all on my T1 images, and the parcellation/segmentation looks fine. However, I noticed that when I imported this data into
3DSlicer,
the volumes of these regions did not match the volumes in the aseg.stats file. In some cases, the differences were quite substantial. I tried removing all smoothing filters in 3DSlicer but the outcome was the same.
Do you know what could be going wrong here? Alternatively, is there an easier way to generate surface meshes from the aparc or aseg atlases?
I'm using the ICBM 2009c Nonlinear Asymmetric template (http://nist.mni.mcgill.ca/?p=904) as a generic brain volume (though alternative suggestions are welcome).
Thanks! Jonah
-- Jonah PeterGraduate Student in The Biophysics ProgramHarvard University P: 646-306-0848 E: jonahpeter@g.harvard.edu
Freesurfer mailing list Freesurfer@nmr.mgh.harvard.edu https://mail.nmr.mgh.harvard.edu/mailman/listinfo/freesurfer
no, they are definitely not negligible. It depends on structure geometry, but I bet for something like the hippocampus a large fraction of voxels in it are partial volumed.
You can use mri_mc or mri_tessellate to generate a mesh. It will be closed but may not have a simple topology
On Thu, 23 Apr 2020, Peter, Jonah wrote:
Hi all, Thanks for the response. I'm familiar with partial volume effects within the context of PET imaging, but wouldn't you expect these effects to be negligible in MRI for structures on the order of mm?
With respect to the mesh type, I'd like to get closed meshes for each segmentation/parcellation.
Is there a way to do this in freesurfer?
On Thu, Apr 23, 2020 at 10:44 AM Bruce Fischl fischl@nmr.mgh.harvard.edu wrote: Hi Jonah
when we compute volumes we typically use a partial-volume model, which we have found improves accuracy and repeatability. That may be the source of your discrepancy cheers Bruce On Wed, 22 Apr 2020, Peter, Jonah wrote: > Hello, > I'm trying to generate surface meshes for different brain regions that can > be imported into a software like MeshLab or Blender (i.e. .STL, .PLY, .OBJ, > etc.). I ran recon-all on my T1 images, and the parcellation/segmentation > looks fine. However, I noticed that when I imported this data into 3DSlicer, > the volumes of these regions did not match the volumes in the aseg.stats > file. In some cases, the differences were quite substantial. I tried > removing all smoothing filters in 3DSlicer but the outcome was the same. > > Do you know what could be going wrong here? Alternatively, is there an > easier way to generate surface meshes from the aparc or aseg atlases? > > I'm using the ICBM 2009c Nonlinear Asymmetric > template (http://nist.mni.mcgill.ca/?p=904) as a generic brain volume > (though alternative suggestions are welcome). > > Thanks! > Jonah > > -- > Jonah PeterGraduate Student in The Biophysics ProgramHarvard University > P: 646-306-0848 > E: jonahpeter@g.harvard.edu > >_______________________________________________ Freesurfer mailing list Freesurfer@nmr.mgh.harvard.edu https://mail.nmr.mgh.harvard.edu/mailman/listinfo/freesurfer-- Jonah PeterGraduate Student in The Biophysics ProgramHarvard University P: 646-306-0848 E: jonahpeter@g.harvard.edu
Hi Bruce,
I tried mri_tessellate which worked well in that the surface seems to follow the aseg projections exactly. Does the volume of this mesh account for partial volume effects? Also, is there a need to run mri_pretess or some sort of smoothing procedure?
For reference, what I am trying to do is import these surface meshes into a Geant4 simulation for proton therapy, and want to make sure I'm using as realistic geometry as possible. Ideally I'd be able to generate a mesh for the whole brain as well as individual segmentations.
Thanks again for your help, Jonah
On Thu, Apr 23, 2020 at 5:50 PM Bruce Fischl fischl@nmr.mgh.harvard.edu wrote:
no, they are definitely not negligible. It depends on structure geometry, but I bet for something like the hippocampus a large fraction of voxels in it are partial volumed.
You can use mri_mc or mri_tessellate to generate a mesh. It will be closed but may not have a simple topology
On Thu, 23 Apr 2020, Peter, Jonah wrote:
Hi all, Thanks for the response. I'm familiar with partial volume effects within
the context of PET imaging, but
wouldn't you expect these effects to be negligible in MRI for structures
on the order of mm?
With respect to the mesh type, I'd like to get closed meshes for each
segmentation/parcellation.
Is there a way to do this in freesurfer?
On Thu, Apr 23, 2020 at 10:44 AM Bruce Fischl <
fischl@nmr.mgh.harvard.edu> wrote:
Hi Jonah when we compute volumes we typically use a partial-volume model,which we
have found improves accuracy and repeatability. That may be thesource of
your discrepancy cheers Bruce On Wed, 22 Apr 2020, Peter, Jonah wrote: > Hello, > I'm trying to generate surface meshes for different brainregions that can
> be imported into a software like MeshLab or Blender (i.e. .STL,.PLY, .OBJ,
> etc.). I ran recon-all on my T1 images, and theparcellation/segmentation
> looks fine. However, I noticed that when I imported this datainto 3DSlicer,
> the volumes of these regions did not match the volumes in theaseg.stats
> file. In some cases, the differences were quite substantial. Itried
> removing all smoothing filters in 3DSlicer but the outcome wasthe same.
> > Do you know what could be going wrong here? Alternatively, isthere an
> easier way to generate surface meshes from the aparc or asegatlases?
> > I'm using the ICBM 2009c Nonlinear Asymmetric > template (http://nist.mni.mcgill.ca/?p=904) as a generic brainvolume
> (though alternative suggestions are welcome). > > Thanks! > Jonah > > -- > Jonah PeterGraduate Student in The Biophysics ProgramHarvardUniversity
> P: 646-306-0848 > E: jonahpeter@g.harvard.edu > >_______________________________________________ Freesurfer mailing list Freesurfer@nmr.mgh.harvard.edu https://mail.nmr.mgh.harvard.edu/mailman/listinfo/freesurfer-- Jonah PeterGraduate Student in The Biophysics ProgramHarvard University P: 646-306-0848 E: jonahpeter@g.harvard.edu
Freesurfer mailing list Freesurfer@nmr.mgh.harvard.edu https://mail.nmr.mgh.harvard.edu/mailman/listinfo/freesurfer
HI Jonah
it depends how you compute volume. mris_volume for example will compute the volume of the interior without any partial volume correction.
And yes, I would expect you want to smooth the surface a bit. You might try the Taubin smoothing built into freeview as it won't shrink the surface, but you can also try mris_smooth
cheers Bruce
On Thu, 23 Apr 2020, Peter, Jonah wrote:
Hi Bruce, I tried mri_tessellate which worked well in that the surface seems to follow the aseg projections exactly. Does the volume of this mesh account for partial volume effects? Also, is there a need to run mri_pretess or some sort of smoothing procedure?
For reference, what I am trying to do is import these surface meshes into a Geant4 simulation for proton therapy, and want to make sure I'm using as realistic geometry as possible. Ideally I'd be able to generate a mesh for the whole brain as well as individual segmentations.
Thanks again for your help, Jonah
On Thu, Apr 23, 2020 at 5:50 PM Bruce Fischl fischl@nmr.mgh.harvard.edu wrote: no, they are definitely not negligible. It depends on structure geometry, but I bet for something like the hippocampus a large fraction of voxels in it are partial volumed.
You can use mri_mc or mri_tessellate to generate a mesh. It will be closed but may not have a simple topology On Thu, 23 Apr 2020, Peter, Jonah wrote: > Hi all, > Thanks for the response. I'm familiar with partial volume effects within the context of PET imaging, but > wouldn't you expect these effects to be negligible in MRI for structures on the order of mm? > > With respect to the mesh type, I'd like to get closed meshes for each segmentation/parcellation. > > Is there a way to do this in freesurfer? > > On Thu, Apr 23, 2020 at 10:44 AM Bruce Fischl <fischl@nmr.mgh.harvard.edu> wrote: > Hi Jonah > > when we compute volumes we typically use a partial-volume model, which we > have found improves accuracy and repeatability. That may be the source of > your discrepancy > > cheers > Bruce > On Wed, 22 Apr 2020, Peter, Jonah wrote: > > > Hello, > > I'm trying to generate surface meshes for different brain regions that can > > be imported into a software like MeshLab or Blender (i.e. .STL, .PLY, .OBJ, > > etc.). I ran recon-all on my T1 images, and the parcellation/segmentation > > looks fine. However, I noticed that when I imported this data into 3DSlicer, > > the volumes of these regions did not match the volumes in the aseg.stats > > file. In some cases, the differences were quite substantial. I tried > > removing all smoothing filters in 3DSlicer but the outcome was the same. > > > > Do you know what could be going wrong here? Alternatively, is there an > > easier way to generate surface meshes from the aparc or aseg atlases? > > > > I'm using the ICBM 2009c Nonlinear Asymmetric > > template (http://nist.mni.mcgill.ca/?p=904) as a generic brain volume > > (though alternative suggestions are welcome). > > > > Thanks! > > Jonah > > > > -- > > Jonah PeterGraduate Student in The Biophysics ProgramHarvard University > > P: 646-306-0848 > > E: jonahpeter@g.harvard.edu > > > >_______________________________________________ > Freesurfer mailing list > Freesurfer@nmr.mgh.harvard.edu > https://mail.nmr.mgh.harvard.edu/mailman/listinfo/freesurfer > > > > -- > Jonah PeterGraduate Student in The Biophysics ProgramHarvard University > P: 646-306-0848 > E: jonahpeter@g.harvard.edu > >_______________________________________________ Freesurfer mailing list Freesurfer@nmr.mgh.harvard.edu https://mail.nmr.mgh.harvard.edu/mailman/listinfo/freesurfer-- Jonah PeterGraduate Student in The Biophysics ProgramHarvard University P: 646-306-0848 E: jonahpeter@g.harvard.edu
Hi Bruce,
The issue is that I want the surface mesh itself to reflect the correct (partial volume corrected) volume as I will be simulating the passage of high energy particles through it. Do you think it would suffice to take the (smoothed) mesh and simply scale its volume in an application like blender? My thinking is that the partial volume effects are probably isotropic so a uniform scaling would leave the geometry unchanged.
Best, Jonah
On Thu, Apr 23, 2020 at 9:53 PM Bruce Fischl fischl@nmr.mgh.harvard.edu wrote:
HI Jonah
it depends how you compute volume. mris_volume for example will compute the volume of the interior without any partial volume correction.
And yes, I would expect you want to smooth the surface a bit. You might try the Taubin smoothing built into freeview as it won't shrink the surface, but you can also try mris_smooth
cheers Bruce
On Thu, 23 Apr 2020, Peter, Jonah wrote:
Hi Bruce, I tried mri_tessellate which worked well in that the surface seems to
follow the aseg projections
exactly. Does the volume of this mesh account for partial volume
effects? Also, is there a need to run
mri_pretess or some sort of smoothing procedure?
For reference, what I am trying to do is import these surface meshes
into a Geant4 simulation for proton
therapy, and want to make sure I'm using as realistic geometry as
possible. Ideally I'd be able to
generate a mesh for the whole brain as well as individual segmentations.
Thanks again for your help, Jonah
On Thu, Apr 23, 2020 at 5:50 PM Bruce Fischl fischl@nmr.mgh.harvard.edu
wrote:
no, they are definitely not negligible. It depends on structuregeometry,
but I bet for something like the hippocampus a large fraction ofvoxels in
it are partial volumed. You can use mri_mc or mri_tessellate to generate a mesh. It willbe closed
but may not have a simple topology On Thu, 23 Apr 2020, Peter, Jonah wrote: > Hi all, > Thanks for the response. I'm familiar with partial volumeeffects within the context of
PET imaging, but > wouldn't you expect these effects to be negligible in MRI forstructures on the order of
mm? > > With respect to the mesh type, I'd like to get closed meshes foreach
segmentation/parcellation. > > Is there a way to do this in freesurfer? > > On Thu, Apr 23, 2020 at 10:44 AM Bruce Fischl <fischl@nmr.mgh.harvard.edu> wrote:
> Hi Jonah > > when we compute volumes we typically use a partial-volumemodel, which we
> have found improves accuracy and repeatability. That maybe the source of
> your discrepancy > > cheers > Bruce > On Wed, 22 Apr 2020, Peter, Jonah wrote: > > > Hello, > > I'm trying to generate surface meshes for differentbrain regions that can
> > be imported into a software like MeshLab or Blender(i.e. .STL, .PLY, .OBJ,
> > etc.). I ran recon-all on my T1 images, and theparcellation/segmentation
> > looks fine. However, I noticed that when I imported thisdata into 3DSlicer,
> > the volumes of these regions did not match the volumesin the aseg.stats
> > file. In some cases, the differences were quitesubstantial. I tried
> > removing all smoothing filters in 3DSlicer but theoutcome was the same.
> > > > Do you know what could be going wrong here?Alternatively, is there an
> > easier way to generate surface meshes from the aparc oraseg atlases?
> > > > I'm using the ICBM 2009c Nonlinear Asymmetric > > template (http://nist.mni.mcgill.ca/?p=904) as ageneric brain volume
> > (though alternative suggestions are welcome). > > > > Thanks! > > Jonah > > > > -- > > Jonah PeterGraduate Student in The BiophysicsProgramHarvard University
> > P: 646-306-0848 > > E: jonahpeter@g.harvard.edu > > > >_______________________________________________ > Freesurfer mailing list > Freesurfer@nmr.mgh.harvard.edu >https://mail.nmr.mgh.harvard.edu/mailman/listinfo/freesurfer
> > > > -- > Jonah PeterGraduate Student in The Biophysics ProgramHarvardUniversity
> P: 646-306-0848 > E: jonahpeter@g.harvard.edu > >_______________________________________________ Freesurfer mailing list Freesurfer@nmr.mgh.harvard.edu https://mail.nmr.mgh.harvard.edu/mailman/listinfo/freesurfer-- Jonah PeterGraduate Student in The Biophysics ProgramHarvard University P: 646-306-0848 E: jonahpeter@g.harvard.edu
Freesurfer mailing list Freesurfer@nmr.mgh.harvard.edu https://mail.nmr.mgh.harvard.edu/mailman/listinfo/freesurfer
Hi Jonah
I'm not positive I understand, but I doubt partial volume effects are isotropic given the geometric complexity of a lot of brain structures. Maybe the easiest thing for you to do is to use something like SAMSEG in the upcoming V7 and try deforming the surface to pass through the p=0.5 isosurface. Not sure how hard that would be to implement, maybe others (Doug? Eugenio? Viviana?) can comment? Are there any specific structures you are interested in?
cheers Bruce
On Thu, 23 Apr 2020, Peter, Jonah wrote:
Hi Bruce, The issue is that I want the surface mesh itself to reflect the correct (partial volume corrected) volume as I will be simulating the passage of high energy particles through it. Do you think it would suffice to take the (smoothed) mesh and simply scale its volume in an application like blender? My thinking is that the partial volume effects are probably isotropic so a uniform scaling would leave the geometry unchanged.
Best, Jonah
On Thu, Apr 23, 2020 at 9:53 PM Bruce Fischl fischl@nmr.mgh.harvard.edu wrote: HI Jonah
it depends how you compute volume. mris_volume for example will compute the volume of the interior without any partial volume correction. And yes, I would expect you want to smooth the surface a bit. You might try the Taubin smoothing built into freeview as it won't shrink the surface, but you can also try mris_smooth cheers Bruce On Thu, 23 Apr 2020, Peter, Jonah wrote: > Hi Bruce, > I tried mri_tessellate which worked well in that the surface seems to follow the aseg projections > exactly. Does the volume of this mesh account for partial volume effects? Also, is there a need to run > mri_pretess or some sort of smoothing procedure? > > For reference, what I am trying to do is import these surface meshes into a Geant4 simulation for proton > therapy, and want to make sure I'm using as realistic geometry as possible. Ideally I'd be able to > generate a mesh for the whole brain as well as individual segmentations. > > Thanks again for your help, > Jonah > > On Thu, Apr 23, 2020 at 5:50 PM Bruce Fischl <fischl@nmr.mgh.harvard.edu> wrote: > no, they are definitely not negligible. It depends on structure geometry, > but I bet for something like the hippocampus a large fraction of voxels in > it are partial volumed. > > You can use mri_mc or mri_tessellate to generate a mesh. It will be closed > but may not have a simple topology > > On Thu, 23 Apr 2020, Peter, Jonah > wrote: > > > Hi all, > > Thanks for the response. I'm familiar with partial volume effects within the context of > PET imaging, but > > wouldn't you expect these effects to be negligible in MRI for structures on the order of > mm? > > > > With respect to the mesh type, I'd like to get closed meshes for each > segmentation/parcellation. > > > > Is there a way to do this in freesurfer? > > > > On Thu, Apr 23, 2020 at 10:44 AM Bruce Fischl <fischl@nmr.mgh.harvard.edu> wrote: > > Hi Jonah > > > > when we compute volumes we typically use a partial-volume model, which we > > have found improves accuracy and repeatability. That may be the source of > > your discrepancy > > > > cheers > > Bruce > > On Wed, 22 Apr 2020, Peter, Jonah wrote: > > > > > Hello, > > > I'm trying to generate surface meshes for different brain regions that can > > > be imported into a software like MeshLab or Blender (i.e. .STL, .PLY, .OBJ, > > > etc.). I ran recon-all on my T1 images, and the parcellation/segmentation > > > looks fine. However, I noticed that when I imported this data into 3DSlicer, > > > the volumes of these regions did not match the volumes in the aseg.stats > > > file. In some cases, the differences were quite substantial. I tried > > > removing all smoothing filters in 3DSlicer but the outcome was the same. > > > > > > Do you know what could be going wrong here? Alternatively, is there an > > > easier way to generate surface meshes from the aparc or aseg atlases? > > > > > > I'm using the ICBM 2009c Nonlinear Asymmetric > > > template (http://nist.mni.mcgill.ca/?p=904) as a generic brain volume > > > (though alternative suggestions are welcome). > > > > > > Thanks! > > > Jonah > > > > > > -- > > > Jonah PeterGraduate Student in The Biophysics ProgramHarvard University > > > P: 646-306-0848 > > > E: jonahpeter@g.harvard.edu > > > > > >_______________________________________________ > > Freesurfer mailing list > > Freesurfer@nmr.mgh.harvard.edu > > https://mail.nmr.mgh.harvard.edu/mailman/listinfo/freesurfer > > > > > > > > -- > > Jonah PeterGraduate Student in The Biophysics ProgramHarvard University > > P: 646-306-0848 > > E: jonahpeter@g.harvard.edu > > > >_______________________________________________ > Freesurfer mailing list > Freesurfer@nmr.mgh.harvard.edu > https://mail.nmr.mgh.harvard.edu/mailman/listinfo/freesurfer > > > > -- > Jonah PeterGraduate Student in The Biophysics ProgramHarvard University > P: 646-306-0848 > E: jonahpeter@g.harvard.edu > >_______________________________________________ Freesurfer mailing list Freesurfer@nmr.mgh.harvard.edu https://mail.nmr.mgh.harvard.edu/mailman/listinfo/freesurfer-- Jonah PeterGraduate Student in The Biophysics ProgramHarvard University P: 646-306-0848 E: jonahpeter@g.harvard.edu
Hi Bruce,
I'm interested in generating meshes for the whole brain,and individual regions of the prefrontal cortex and the hippocampus.
It seems that freesurfer doesn't generate meshes with physical volumes corrected for partial volume effects, but instead performs an a posteriori calculation that is output to the aseg.stats file (is this correct?). What I meant before was: do you think it would be reasonable to take the surface meshes and simply scale down their volumes such that the physical mesh volume then corresponds to the partial volume corrected volume?
As an example, suppose the actual physical volume of the hippocampal mesh is 5000 mm3, whereas the aseg.stats lists the volume as 4000 mm3 after correcting for partial volume effects. Would it be correct to just scale the mesh volume to 4000 mm3?
Thanks again, Jonah
On Fri, Apr 24, 2020 at 10:35 AM Bruce Fischl fischl@nmr.mgh.harvard.edu wrote:
Hi Jonah
I'm not positive I understand, but I doubt partial volume effects are isotropic given the geometric complexity of a lot of brain structures. Maybe the easiest thing for you to do is to use something like SAMSEG in the upcoming V7 and try deforming the surface to pass through the p=0.5 isosurface. Not sure how hard that would be to implement, maybe others (Doug? Eugenio? Viviana?) can comment? Are there any specific structures you are interested in?
cheers Bruce
On Thu, 23 Apr 2020, Peter, Jonah wrote:
Hi Bruce, The issue is that I want the surface mesh itself to reflect the correct
(partial volume corrected)
volume as I will be simulating the passage of high energy particles
through it. Do you think it would
suffice to take the (smoothed) mesh and simply scale its volume in an
application like blender? My
thinking is that the partial volume effects are probably isotropic so a
uniform scaling would leave the
geometry unchanged.
Best, Jonah
On Thu, Apr 23, 2020 at 9:53 PM Bruce Fischl fischl@nmr.mgh.harvard.edu
wrote:
HI Jonah it depends how you compute volume. mris_volume for example willcompute the
volume of the interior without any partial volume correction. And yes, I would expect you want to smooth the surface a bit. Youmight try
the Taubin smoothing built into freeview as it won't shrink thesurface,
but you can also try mris_smooth cheers Bruce On Thu, 23 Apr 2020, Peter, Jonah wrote: > Hi Bruce, > I tried mri_tessellate which worked well in that the surfaceseems to follow the aseg
projections > exactly. Does the volume of this mesh account for partial volumeeffects? Also, is there a
need to run > mri_pretess or some sort of smoothing procedure? > > For reference, what I am trying to do is import these surfacemeshes into a Geant4
simulation for proton > therapy, and want to make sure I'm using as realistic geometryas possible. Ideally I'd be
able to > generate a mesh for the whole brain as well asindividual segmentations.
> > Thanks again for your help, > Jonah > > On Thu, Apr 23, 2020 at 5:50 PM Bruce Fischl <fischl@nmr.mgh.harvard.edu> wrote:
> no, they are definitely not negligible. It depends onstructure geometry,
> but I bet for something like the hippocampus a largefraction of voxels in
> it are partial volumed. > > You can use mri_mc or mri_tessellate to generate a mesh.It will be closed
> but may not have a simple topology > > On Thu, 23 Apr 2020, Peter, Jonah > wrote: > > > Hi all, > > Thanks for the response. I'm familiar with partialvolume effects within the
context of > PET imaging, but > > wouldn't you expect these effects to be negligible inMRI for structures on the
order of > mm? > > > > With respect to the mesh type, I'd like to get closedmeshes for each
> segmentation/parcellation. > > > > Is there a way to do this in freesurfer? > > > > On Thu, Apr 23, 2020 at 10:44 AM Bruce Fischl <fischl@nmr.mgh.harvard.edu> wrote:
> > Hi Jonah > > > > when we compute volumes we typically use apartial-volume model, which we
> > have found improves accuracy and repeatability.That may be the source of
> > your discrepancy > > > > cheers > > Bruce > > On Wed, 22 Apr 2020, Peter, Jonah wrote: > > > > > Hello, > > > I'm trying to generate surface meshes fordifferent brain regions that can
> > > be imported into a software like MeshLab orBlender (i.e. .STL, .PLY,
.OBJ, > > > etc.). I ran recon-all on my T1 images, and theparcellation/segmentation
> > > looks fine. However, I noticed that when Iimported this data into
3DSlicer, > > > the volumes of these regions did not match thevolumes in the aseg.stats
> > > file. In some cases, the differences were quitesubstantial. I tried
> > > removing all smoothing filters in 3DSlicer butthe outcome was the same.
> > > > > > Do you know what could be going wrong here?Alternatively, is there an
> > > easier way to generate surface meshes from theaparc or aseg atlases?
> > > > > > I'm using the ICBM 2009c Nonlinear Asymmetric > > > template (http://nist.mni.mcgill.ca/?p=904) asa generic brain volume
> > > (though alternative suggestions are welcome). > > > > > > Thanks! > > > Jonah > > > > > > -- > > > Jonah PeterGraduate Student in The BiophysicsProgramHarvard University
> > > P: 646-306-0848 > > > E: jonahpeter@g.harvard.edu > > > > > >_______________________________________________ > > Freesurfer mailing list > > Freesurfer@nmr.mgh.harvard.edu > >https://mail.nmr.mgh.harvard.edu/mailman/listinfo/freesurfer
> > > > > > > > -- > > Jonah PeterGraduate Student in The BiophysicsProgramHarvard University
> > P: 646-306-0848 > > E: jonahpeter@g.harvard.edu > > > >_______________________________________________ > Freesurfer mailing list > Freesurfer@nmr.mgh.harvard.edu >https://mail.nmr.mgh.harvard.edu/mailman/listinfo/freesurfer
> > > > -- > Jonah PeterGraduate Student in The Biophysics ProgramHarvardUniversity
> P: 646-306-0848 > E: jonahpeter@g.harvard.edu > >_______________________________________________ Freesurfer mailing list Freesurfer@nmr.mgh.harvard.edu https://mail.nmr.mgh.harvard.edu/mailman/listinfo/freesurfer-- Jonah PeterGraduate Student in The Biophysics ProgramHarvard University P: 646-306-0848 E: jonahpeter@g.harvard.edu
Freesurfer mailing list Freesurfer@nmr.mgh.harvard.edu https://mail.nmr.mgh.harvard.edu/mailman/listinfo/freesurfer
Hi Jonah
no, I don't think that would be very accurate. In our cortical surface stream we do partial volume modeling in the surface deformation so that the surface settles at the right point between gray/white and gray/csf. We never wrote anything comparable for the aseg structs like hippocampus, but a quick-and-dirty solution would be to settle at the 0.5 posterior probablity level of something like samseg
cheers Bruce
On Fri, 24 Apr 2020, Peter, Jonah wrote:
Hi Bruce, I'm interested in generating meshes for the whole brain,and individual regions of the prefrontal cortex and the hippocampus.
It seems that freesurfer doesn't generate meshes with physical volumes corrected for partial volume effects, but instead performs an a posteriori calculation that is output to the aseg.stats file (is this correct?). What I meant before was: do you think it would be reasonable to take the surface meshes and simply scale down their volumes such that the physical mesh volume then corresponds to the partial volume corrected volume?
As an example, suppose the actual physical volume of the hippocampal mesh is 5000 mm3, whereas the aseg.stats lists the volume as 4000 mm3 after correcting for partial volume effects. Would it be correct to just scale the mesh volume to 4000 mm3?
Thanks again, Jonah
On Fri, Apr 24, 2020 at 10:35 AM Bruce Fischl fischl@nmr.mgh.harvard.edu wrote: Hi Jonah
I'm not positive I understand, but I doubt partial volume effects are isotropic given the geometric complexity of a lot of brain structures. Maybe the easiest thing for you to do is to use something like SAMSEG in the upcoming V7 and try deforming the surface to pass through the p=0.5 isosurface. Not sure how hard that would be to implement, maybe others (Doug? Eugenio? Viviana?) can comment? Are there any specific structures you are interested in? cheers Bruce On Thu, 23 Apr 2020, Peter, Jonah wrote: > Hi Bruce, > The issue is that I want the surface mesh itself to reflect the correct (partial volume corrected) > volume as I will be simulating the passage of high energy particles through it. Do you think it would > suffice to take the (smoothed) mesh and simply scale its volume in an application like blender? My > thinking is that the partial volume effects are probably isotropic so a uniform scaling would leave the > geometry unchanged. > > Best, > Jonah > > On Thu, Apr 23, 2020 at 9:53 PM Bruce Fischl <fischl@nmr.mgh.harvard.edu> wrote: > HI Jonah > > it depends how you compute volume. mris_volume for example will compute the > volume of the interior without any partial volume correction. > > And yes, I would expect you want to smooth the surface a bit. You might try > the Taubin smoothing built into freeview as it won't shrink the surface, > but you can also try mris_smooth > > cheers > Bruce > > > On Thu, 23 Apr 2020, Peter, Jonah wrote: > > > Hi Bruce, > > I tried mri_tessellate which worked well in that the surface seems to follow the aseg > projections > > exactly. Does the volume of this mesh account for partial volume effects? Also, is there a > need to run > > mri_pretess or some sort of smoothing procedure? > > > > For reference, what I am trying to do is import these surface meshes into a Geant4 > simulation for proton > > therapy, and want to make sure I'm using as realistic geometry as possible. Ideally I'd be > able to > > generate a mesh for the whole brain as well as individual segmentations. > > > > Thanks again for your help, > > Jonah > > > > On Thu, Apr 23, 2020 at 5:50 PM Bruce Fischl <fischl@nmr.mgh.harvard.edu> wrote: > > no, they are definitely not negligible. It depends on structure geometry, > > but I bet for something like the hippocampus a large fraction of voxels in > > it are partial volumed. > > > > You can use mri_mc or mri_tessellate to generate a mesh. It will be closed > > but may not have a simple topology > > > > On Thu, 23 Apr 2020, Peter, Jonah > > wrote: > > > > > Hi all, > > > Thanks for the response. I'm familiar with partial volume effects within the > context of > > PET imaging, but > > > wouldn't you expect these effects to be negligible in MRI for structures on the > order of > > mm? > > > > > > With respect to the mesh type, I'd like to get closed meshes for each > > segmentation/parcellation. > > > > > > Is there a way to do this in freesurfer? > > > > > > On Thu, Apr 23, 2020 at 10:44 AM Bruce Fischl <fischl@nmr.mgh.harvard.edu> wrote: > > > Hi Jonah > > > > > > when we compute volumes we typically use a partial-volume model, which we > > > have found improves accuracy and repeatability. That may be the source of > > > your discrepancy > > > > > > cheers > > > Bruce > > > On Wed, 22 Apr 2020, Peter, Jonah wrote: > > > > > > > Hello, > > > > I'm trying to generate surface meshes for different brain regions that can > > > > be imported into a software like MeshLab or Blender (i.e. .STL, .PLY, > .OBJ, > > > > etc.). I ran recon-all on my T1 images, and the parcellation/segmentation > > > > looks fine. However, I noticed that when I imported this data into > 3DSlicer, > > > > the volumes of these regions did not match the volumes in the aseg.stats > > > > file. In some cases, the differences were quite substantial. I tried > > > > removing all smoothing filters in 3DSlicer but the outcome was the same. > > > > > > > > Do you know what could be going wrong here? Alternatively, is there an > > > > easier way to generate surface meshes from the aparc or aseg atlases? > > > > > > > > I'm using the ICBM 2009c Nonlinear Asymmetric > > > > template (http://nist.mni.mcgill.ca/?p=904) as a generic brain volume > > > > (though alternative suggestions are welcome). > > > > > > > > Thanks! > > > > Jonah > > > > > > > > -- > > > > Jonah PeterGraduate Student in The Biophysics ProgramHarvard University > > > > P: 646-306-0848 > > > > E: jonahpeter@g.harvard.edu > > > > > > > >_______________________________________________ > > > Freesurfer mailing list > > > Freesurfer@nmr.mgh.harvard.edu > > > https://mail.nmr.mgh.harvard.edu/mailman/listinfo/freesurfer > > > > > > > > > > > > -- > > > Jonah PeterGraduate Student in The Biophysics ProgramHarvard University > > > P: 646-306-0848 > > > E: jonahpeter@g.harvard.edu > > > > > >_______________________________________________ > > Freesurfer mailing list > > Freesurfer@nmr.mgh.harvard.edu > > https://mail.nmr.mgh.harvard.edu/mailman/listinfo/freesurfer > > > > > > > > -- > > Jonah PeterGraduate Student in The Biophysics ProgramHarvard University > > P: 646-306-0848 > > E: jonahpeter@g.harvard.edu > > > >_______________________________________________ > Freesurfer mailing list > Freesurfer@nmr.mgh.harvard.edu > https://mail.nmr.mgh.harvard.edu/mailman/listinfo/freesurfer > > > > -- > Jonah PeterGraduate Student in The Biophysics ProgramHarvard University > P: 646-306-0848 > E: jonahpeter@g.harvard.edu > >_______________________________________________ Freesurfer mailing list Freesurfer@nmr.mgh.harvard.edu https://mail.nmr.mgh.harvard.edu/mailman/listinfo/freesurfer-- Jonah PeterGraduate Student in The Biophysics ProgramHarvard University P: 646-306-0848 E: jonahpeter@g.harvard.edu
freesurfer@nmr.mgh.harvard.edu
-
Bruce Fischl -
Peter, Jonah -
Tim Schäfer