Folks:
I'm trying to get a complete picture of what's entailed in creating our own atlas, including our own average reference sphere, so if there's a doc on this specific subject I'd love to see it.
Absent that, below are questions to fill in some gaps:
1. Normal spherical reg: mris_register's command-line params don't include all files it must read, so I'm looking for a list of all the files that it reads under the usual circumstances (ie: registration to ref sphere; registration to opposite hemi), and how it decides *which* files to read. (Amongst other discrepancies, when recon-all runs, mris_register doesn't seem to read ?h.curv, whereas various postings suggest that it does.)
2. Normal labeling with mris_ca_label: Presumably this only works if subjects are registered to the same average sphere mesh as was used to create the chosen GCS file, right? Yet mris_ca_label doesn't need to read that actual ?h.sphere.reg file because equivalent position-on-sphere info is contained in the GCS file... right?
3. Creating one's own average reference sphere: What progam(s) does one use to create the average ?h.sphere(.reg?) file (presumably from each subject's ?h.sphere file)? and average curvature file (ie: the tif file)? FWIW, it looks like mris_make_template is somehow involved here, but it's entire documentation consists of just "This program will add a template into an average surface" (what does that mean?), and the source code looks like it only produces the tif file containing average sulc and curv data. (From which I conclude the here, "template" refers to just that file.)
4. What, if any, is the relationship between .../freesurfer/average and .../freesurfer/subjects/fsaverage
Thanks,
Graham
On Wed, 16 Jan 2008, Graham Wideman wrote:
Folks:
I'm trying to get a complete picture of what's entailed in creating our own atlas, including our own average reference sphere, so if there's a doc on this specific subject I'd love to see it.
Absent that, below are questions to fill in some gaps:
- Normal spherical reg: mris_register's command-line params don't include
all files it must read, so I'm looking for a list of all the files that it reads under the usual circumstances (ie: registration to ref sphere; registration to opposite hemi), and how it decides *which* files to read. (Amongst other discrepancies, when recon-all runs, mris_register doesn't seem to read ?h.curv, whereas various postings suggest that it does.)
I think it recomputed the curvature internally instead of reading the file.
- Normal labeling with mris_ca_label: Presumably this only works if
subjects are registered to the same average sphere mesh as was used to create the chosen GCS file, right? Yet mris_ca_label doesn't need to read that actual ?h.sphere.reg file because equivalent position-on-sphere info is contained in the GCS file... right?
it needs to read the subject's ?h.sphere.reg, but the gcs is already in register with the atlas, so it doesn't need one for the .gcs (which is actually a bunch of info stored in different level icosahedral meshes)
- Creating one's own average reference sphere: What progam(s) does one use
to create the average ?h.sphere(.reg?) file (presumably from each subject's ?h.sphere file)? and average curvature file (ie: the tif file)? FWIW, it looks like mris_make_template is somehow involved here, but it's entire documentation consists of just "This program will add a template into an average surface" (what does that mean?), and the source code looks like it only produces the tif file containing average sulc and curv data. (From which I conclude the here, "template" refers to just that file.)
mris_make_template creates an "average" sphere.reg (actually a bunch of means and variances of different geometric features. To make a new gcs file use mris_ca_train.
- What, if any, is the relationship between
.../freesurfer/average and .../freesurfer/subjects/fsaverage
average is a catch-all for various "average" files, like volume statistics, tal atlases, etc..., fsaverage is the subject made from averaging our 40 manually labeled subjects
cheers, Bruce
Thanks,
Graham _______________________________________________ Freesurfer mailing list Freesurfer@nmr.mgh.harvard.edu https://mail.nmr.mgh.harvard.edu/mailman/listinfo/freesurfer
I've included hooks into make_average_subject and recon-all to make this much easier. make_average_subject will create the template (.tif) files from the subjects you specify, then you just run recon-all with -surfreg-to-subject averagesubjectname.
Bruce Fischl wrote:
On Wed, 16 Jan 2008, Graham Wideman wrote:
Folks:
I'm trying to get a complete picture of what's entailed in creating our own atlas, including our own average reference sphere, so if there's a doc on this specific subject I'd love to see it.
Absent that, below are questions to fill in some gaps:
- Normal spherical reg: mris_register's command-line params don't
include all files it must read, so I'm looking for a list of all the files that it reads under the usual circumstances (ie: registration to ref sphere; registration to opposite hemi), and how it decides *which* files to read. (Amongst other discrepancies, when recon-all runs, mris_register doesn't seem to read ?h.curv, whereas various postings suggest that it does.)
I think it recomputed the curvature internally instead of reading the file.
- Normal labeling with mris_ca_label: Presumably this only works if
subjects are registered to the same average sphere mesh as was used to create the chosen GCS file, right? Yet mris_ca_label doesn't need to read that actual ?h.sphere.reg file because equivalent position-on-sphere info is contained in the GCS file... right?
it needs to read the subject's ?h.sphere.reg, but the gcs is already in register with the atlas, so it doesn't need one for the .gcs (which is actually a bunch of info stored in different level icosahedral meshes)
- Creating one's own average reference sphere: What progam(s) does
one use to create the average ?h.sphere(.reg?) file (presumably from each subject's ?h.sphere file)? and average curvature file (ie: the tif file)? FWIW, it looks like mris_make_template is somehow involved here, but it's entire documentation consists of just "This program will add a template into an average surface" (what does that mean?), and the source code looks like it only produces the tif file containing average sulc and curv data. (From which I conclude the here, "template" refers to just that file.)
mris_make_template creates an "average" sphere.reg (actually a bunch of means and variances of different geometric features. To make a new gcs file use mris_ca_train.
- What, if any, is the relationship between
.../freesurfer/average and .../freesurfer/subjects/fsaverage
average is a catch-all for various "average" files, like volume statistics, tal atlases, etc..., fsaverage is the subject made from averaging our 40 manually labeled subjects
cheers, Bruce
Thanks,
Graham _______________________________________________ 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
Doug and Bruce:
Bruce says:
mris_make_template creates an "average" sphere.reg (actually a bunch of means and variances of different geometric features. To make a new gcs file use mris_ca_train.
Side point -- the "bunch of means and variances"... that sounds more like the template or GCS file than the average sphere.reg, which I thought was a mesh, right?)
Doug says:
I've included hooks into make_average_subject and recon-all to make this much easier. make_average_subject will create the template (.tif) files from the subjects you specify, then you just run recon-all with -surfreg-to-subject averagesubjectname.
Are these two suggestions alternates or complementary in some way? A major gap for me is this:
Supposing we start with 10 subjects, to be used in creating a new average sphere, tif and gcs for labels, later to be used for many more subjects. These initial 10 subjects are processed by freesurfer, and hence have all their surfaces, (along with ?h.sphere.reg and annot files, courtesy of the default average sphere, tif and label gcs --- which we don't want to use ultimately.)
What then are the steps to create a *new* average sphere and tif?
My assumption has been that some program reads the 10 individuals' ?h.sphere (not reg) files, plus their curv and sulc files, and by clever average, warp and whatever on spherical surface, it produces an average ?h.sphere along with corresponding average curv and sulc tif file. I'm expecting these to be done in one step, because without a curv/sulc tif file the new average sphere is meaningless I think.
I'm specifically *not* expecting that the process involves averaging the individual white surfaces (possibly including talairaching), followed by inflating the average white surface to a sphere. And because I have the impression that this is the kind of thing that make_average_surfaces (called by make_average_subject) does, I'm resistant to thinking that make_average_subject is part of the answer here.
So... I'm all ready to think that mris_make_template is the program to do the job... but I don't see anywhere in the source code where it writes out a surface.
(FWIW I do see that mris_ca_train is used to create the GCS file -- I don't have any questions on that part.)
And finally: Bruce wrote:
average is a catch-all for various "average" files, like volume statistics, tal atlases, etc..., fsaverage is the subject made from averaging our 40 manually labeled subjects
...so I'm getting the impression that some of the {fshome}/average files are not related to the fsaverage subjects (such as mni 305 files), and some perhaps are (like lh.average.curvature.filled.buckner40.tif)... and some others... not sure.
On this point, mainly I'm trying to understand what common sphere and tif file are used in a normal recon-all run, and where they came from (and trying to follow recon-all call's args into mris_register is puzzling).
Thanks,
Graham
Hi Graham,
no, the .tif file has 3 sets of means and variances - the curvature of the inflated surface, sulc and curv, and each one is matched in a multi-scale fashion.
The creation of a new .tif is usually to start with a tif from just one individual, register everyone to it, recreate, reregister and so on (usually twice is enough). Then the .gcs is build in the final space.
cheers, Bruce
On Fri, 18 Jan 2008, Graham Wideman
wrote:
Doug and Bruce:
Bruce says:
mris_make_template creates an "average" sphere.reg (actually a bunch of means and variances of different geometric features. To make a new gcs file use mris_ca_train.
Side point -- the "bunch of means and variances"... that sounds more like the template or GCS file than the average sphere.reg, which I thought was a mesh, right?)
Doug says:
I've included hooks into make_average_subject and recon-all to make this much easier. make_average_subject will create the template (.tif) files from the subjects you specify, then you just run recon-all with -surfreg-to-subject averagesubjectname.
Are these two suggestions alternates or complementary in some way? A major gap for me is this:
Supposing we start with 10 subjects, to be used in creating a new average sphere, tif and gcs for labels, later to be used for many more subjects. These initial 10 subjects are processed by freesurfer, and hence have all their surfaces, (along with ?h.sphere.reg and annot files, courtesy of the default average sphere, tif and label gcs --- which we don't want to use ultimately.)
What then are the steps to create a *new* average sphere and tif?
My assumption has been that some program reads the 10 individuals' ?h.sphere (not reg) files, plus their curv and sulc files, and by clever average, warp and whatever on spherical surface, it produces an average ?h.sphere along with corresponding average curv and sulc tif file. I'm expecting these to be done in one step, because without a curv/sulc tif file the new average sphere is meaningless I think.
I'm specifically *not* expecting that the process involves averaging the individual white surfaces (possibly including talairaching), followed by inflating the average white surface to a sphere. And because I have the impression that this is the kind of thing that make_average_surfaces (called by make_average_subject) does, I'm resistant to thinking that make_average_subject is part of the answer here.
So... I'm all ready to think that mris_make_template is the program to do the job... but I don't see anywhere in the source code where it writes out a surface.
(FWIW I do see that mris_ca_train is used to create the GCS file -- I don't have any questions on that part.)
And finally: Bruce wrote:
average is a catch-all for various "average" files, like volume statistics, tal atlases, etc..., fsaverage is the subject made from averaging our 40 manually labeled subjects
...so I'm getting the impression that some of the {fshome}/average files are not related to the fsaverage subjects (such as mni 305 files), and some perhaps are (like lh.average.curvature.filled.buckner40.tif)... and some others... not sure.
On this point, mainly I'm trying to understand what common sphere and tif file are used in a normal recon-all run, and where they came from (and trying to follow recon-all call's args into mris_register is puzzling).
Thanks,
Graham
freesurfer@nmr.mgh.harvard.edu