Free surfers --
I am collecting data from dogs. Any advice regarding passing this data through the system? I have a couple of specific questions:
1) Should I acquire data with voxel sizes smaller than 1mm, given that the brain is much smaller? How would I tell free-surfer to operate with this higher resolution data?
2) So far, automated skull stripping has been a failure, even with definition of a "seed point" in the white matter. The routine strips away the brain around the seed point, leaving a large chunk of scalp and muscle with a bit of cortex attached. Any thoughts on how to get this going?
Thanks in advance for any help,
Geoff
--
Geoffrey Karl Aguirre, M.D., Ph.D. University of Pennsylvania Center for Cognitive Neuroscience 3815 Walnut Street Fax: (215) 898-1982 Philadelphia, PA 19104-6196 mailto:aguirre@neuro.med.upenn.edu http://ccn.upenn.edu/~aguirre
Dear Geoff,
You could mri_convert raw volumes that label voxel sizes twice than they really are. To my experience, this overcomes quite some problems.
Hope this helps, Denis
Le 15 avr. 04, à 05:18, Geoffrey Aguirre a écrit :
Free surfers --
I am collecting data from dogs. Any advice regarding passing this data through the system? I have a couple of specific questions:
- Should I acquire data with voxel sizes smaller than 1mm, given that
the brain is much smaller? How would I tell free-surfer to operate with this higher resolution data?
- So far, automated skull stripping has been a failure, even with
definition of a "seed point" in the white matter. The routine strips away the brain around the seed point, leaving a large chunk of scalp and muscle with a bit of cortex attached. Any thoughts on how to get this going?
Thanks in advance for any help,
Geoff
--
Geoffrey Karl Aguirre, M.D., Ph.D. University of Pennsylvania Center for Cognitive Neuroscience 3815 Walnut Street Fax: (215) 898-1982 Philadelphia, PA 19104-6196 mailto:aguirre@neuro.med.upenn.edu http://ccn.upenn.edu/~aguirre
Freesurfer mailing list Freesurfer@nmr.mgh.harvard.edu http://mail.nmr.mgh.harvard.edu/mailman/listinfo/freesurfer
I've never seen a dog brain, but macaque brain requires .3mm or so voxels. Various pieces may not work (e.g. cutting planes will certainly fail), but there should be manual interventions for any pieces that fail.
cheers, Bruce
On Wed, 14 Apr 2004, Geoffrey Aguirre wrote:
Free surfers --
I am collecting data from dogs. Any advice regarding passing this data through the system? I have a couple of specific questions:
- Should I acquire data with voxel sizes smaller than 1mm, given that
the brain is much smaller? How would I tell free-surfer to operate with this higher resolution data?
- So far, automated skull stripping has been a failure, even with
definition of a "seed point" in the white matter. The routine strips away the brain around the seed point, leaving a large chunk of scalp and muscle with a bit of cortex attached. Any thoughts on how to get this going?
Thanks in advance for any help,
Geoff
--
Geoffrey Karl Aguirre, M.D., Ph.D. University of Pennsylvania Center for Cognitive Neuroscience 3815 Walnut Street Fax: (215) 898-1982 Philadelphia, PA 19104-6196 mailto:aguirre@neuro.med.upenn.edu http://ccn.upenn.edu/~aguirre
Freesurfer mailing list Freesurfer@nmr.mgh.harvard.edu http://mail.nmr.mgh.harvard.edu/mailman/listinfo/freesurfer
I can speak of higher resolution images for recon processing.
Currently, "recon-all" script can take the option of "-cm" (conform to min voxel size). This may end up creating more slices than 256, but freesurfer should be able to handle as long as there are enough memory left. Make sure that you have 2 Gbyte of memory and 2 GByte of swap space. Here, we are approaching the 32 bit memory address limit. If you run out of memory, try reducing the number of slices. Freesurfer "recon" routines should be able to handle slices different from 256. (Exception: tkregister2 cannot handle more than 256 slices. I "cannot" speak for fsfast.)
Make sure that the direction cosines for animal brains match those of the human, i.e. x_(r,a,s), y_(r,a,s), z_(r,a,s), c_(r,a,s) are correct in the human brain sense. You may have to hand edit these values.
Too many freesurfer routines "assume" the uniform voxel sizes and thus the volume must be conformed to uniform voxel sizes.
You may see freesurfer routines fail on the high resolution images because of the highly grainy nature of the high resolution images. Therefore, you may have to run each routine by hand with additional options.
You may try "cheating" freesurfer routines by changing the voxel size to be 1 mm, even though the actual size is less than 1mm (i.e. editing orig/COR-.info information).
Tosa
Yasunari Tosa, Ph.D. R&D Engineer email: tosa@nmr.mgh.harvard.edu Mass. General Hospital NMR Center phone: Building 149 13th Street fax : Charlestown, MA 02129
Geoffrey Aguirre wrote:
Free surfers --
I am collecting data from dogs. Any advice regarding passing thisdata through the system? I have a couple of specific questions:
- Should I acquire data with voxel sizes smaller than 1mm, given that
the brain is much smaller? How would I tell free-surfer to operate with this higher resolution data?
- So far, automated skull stripping has been a failure, even with
definition of a "seed point" in the white matter. The routine strips away the brain around the seed point, leaving a large chunk of scalp and muscle with a bit of cortex attached. Any thoughts on how to get this going?
Thanks in advance for any help,
Geoff
--
Geoffrey Karl Aguirre, M.D., Ph.D. University of Pennsylvania Center for Cognitive Neuroscience 3815 Walnut Street Fax: (215) 898-1982 Philadelphia, PA 19104-6196 mailto:aguirre@neuro.med.upenn.edu http://ccn.upenn.edu/~aguirre
Freesurfer mailing list Freesurfer@nmr.mgh.harvard.edu http://mail.nmr.mgh.harvard.edu/mailman/listinfo/freesurfer
On Apr 15, 2004, at 11:16 AM, Yasunari Tosa wrote:
Currently, "recon-all" script can take the option of "-cm" (conform to min voxel size). This may end up creating more slices than 256, but freesurfer should be able to handle as long as there are enough memory left. Make sure that you have 2 Gbyte of memory and 2 GByte of swap space. Here, we are approaching the 32 bit memory address limit.
{snip}
This 2gig, 32 bit limit is something I'm starting to worry about. With resolutions of functional data ever increasing, we're getting to the point where it's likely we'll need to address more than 2 gigs of memory in the next year or two.
Are there any plans to address this? I'm not sure what you can do with FS-FAST, since it depends on Matlab and I don't know if there are 64-bit versions of matlab available (especially for affordable platforms like AMD64 and the G5).
<aside> I'll insert my plug for moving FS-FAST to python here, as python natively works on 64 bit platforms. </aside>
But with Freesurfer, how hard will it be to build a 64 bit version? If you need a testing machine, I'm working from a dual G5 that would be more than happy to build Freesurfer :-)
Nick
On Thu, Apr 15, 2004 at 01:51:47PM -0400, Nick Knouf wrote:
This 2gig, 32 bit limit is something I'm starting to worry about. With resolutions of functional data ever increasing, we're getting to the point where it's likely we'll need to address more than 2 gigs of memory in the next year or two.
Are there any plans to address this?
freesurfer-wise, i am currently working on a 64-bit opteron port.
--vicka
Hi Nick,
unfortunately, os/x doesn't allow you to address more than 4G either. On the bright side though, we have a couple of opteron machines around that we are actively trying to port to, so I suspect we'll have a 64 bit version sometime in the next couple of months. We'll let people know when it's ready (you're welcome to be a beta site).
Bruce
On Thu, 15 Apr 2004, Nick Knouf wrote:
On Apr 15, 2004, at 11:16 AM, Yasunari Tosa wrote:
Currently, "recon-all" script can take the option of "-cm" (conform to min voxel size). This may end up creating more slices than 256, but freesurfer should be able to handle as long as there are enough memory left. Make sure that you have 2 Gbyte of memory and 2 GByte of swap space. Here, we are approaching the 32 bit memory address limit.
{snip}
This 2gig, 32 bit limit is something I'm starting to worry about. With resolutions of functional data ever increasing, we're getting to the point where it's likely we'll need to address more than 2 gigs of memory in the next year or two.
Are there any plans to address this? I'm not sure what you can do with FS-FAST, since it depends on Matlab and I don't know if there are 64-bit versions of matlab available (especially for affordable platforms like AMD64 and the G5).
<aside> I'll insert my plug for moving FS-FAST to python here, as python natively works on 64 bit platforms. </aside>
But with Freesurfer, how hard will it be to build a 64 bit version? If you need a testing machine, I'm working from a dual G5 that would be more than happy to build Freesurfer :-)
Nick
Freesurfer mailing list Freesurfer@nmr.mgh.harvard.edu http://mail.nmr.mgh.harvard.edu/mailman/listinfo/freesurfer
Hello "freesurfers", I have a very basic question... Following the tutorial about freesurfer on the FSL home page and also the freesurfer tutorial I try to create the talairach.xfm trasformation file using mritotal. But apparently mritotal don't like the COR format files.
I did several trials but always the same result:
'cannot open COR-001'
and from mincinfo I got:
'is not a netcdf file'
...
Someone can help me?
Thank you in advance
Francesco Musso
FWIW, it looks like Mac OSX Tiger will be able to address 64 bits of virtual memory. Finally, eh. Any plans to utilize this in a future release of FS?
http://www.apple.com/macosx/tiger/64bit.html
Ray
On Apr 15, 2004, at 9:38 PM, Bruce Fischl wrote:
Hi Nick,
unfortunately, os/x doesn't allow you to address more than 4G either. On the bright side though, we have a couple of opteron machines around that we are actively trying to port to, so I suspect we'll have a 64 bit version sometime in the next couple of months. We'll let people know when it's ready (you're welcome to be a beta site).
Bruce
On Thu, 15 Apr 2004, Nick Knouf wrote:
On Apr 15, 2004, at 11:16 AM, Yasunari Tosa wrote:
Currently, "recon-all" script can take the option of "-cm" (conform to min voxel size). This may end up creating more slices than 256, but freesurfer should be able to handle as long as there are enough memory left. Make sure that you have 2 Gbyte of memory and 2 GByte of swap space. Here, we are approaching the 32 bit memory address limit.
{snip}
This 2gig, 32 bit limit is something I'm starting to worry about. With resolutions of functional data ever increasing, we're getting to the point where it's likely we'll need to address more than 2 gigs of memory in the next year or two.
Are there any plans to address this? I'm not sure what you can do with FS-FAST, since it depends on Matlab and I don't know if there are 64-bit versions of matlab available (especially for affordable platforms like AMD64 and the G5).
<aside> I'll insert my plug for moving FS-FAST to python here, as python natively works on 64 bit platforms. </aside>
But with Freesurfer, how hard will it be to build a 64 bit version? If you need a testing machine, I'm working from a dual G5 that would be more than happy to build Freesurfer :-)
Nick
Freesurfer mailing list Freesurfer@nmr.mgh.harvard.edu http://mail.nmr.mgh.harvard.edu/mailman/listinfo/freesurfer
Freesurfer mailing list Freesurfer@nmr.mgh.harvard.edu http://mail.nmr.mgh.harvard.edu/mailman/listinfo/freesurfer
Hi Ray,
I had heard a rumor to this effect. Apple has agreed to loan us a G5 for the port in a month or two, so we will hopefully have a 64 bit os/x version by early fall (and a 64 bit opteron version before that)
cheers, Bruce
On Mon, 28 Jun 2004, Ray Fix wrote:
FWIW, it looks like Mac OSX Tiger will be able to address 64 bits of virtual memory. Finally, eh. Any plans to utilize this in a future release of FS?
http://www.apple.com/macosx/tiger/64bit.html
Ray
On Apr 15, 2004, at 9:38 PM, Bruce Fischl wrote:
Hi Nick,
unfortunately, os/x doesn't allow you to address more than 4G either. On the bright side though, we have a couple of opteron machines around that we are actively trying to port to, so I suspect we'll have a 64 bit version sometime in the next couple of months. We'll let people know when it's ready (you're welcome to be a beta site).
Bruce
On Thu, 15 Apr 2004, Nick Knouf wrote:
On Apr 15, 2004, at 11:16 AM, Yasunari Tosa wrote:
Currently, "recon-all" script can take the option of "-cm" (conform to min voxel size). This may end up creating more slices than 256, but freesurfer should be able to handle as long as there are enough memory left. Make sure that you have 2 Gbyte of memory and 2 GByte of swap space. Here, we are approaching the 32 bit memory address limit.
{snip}
This 2gig, 32 bit limit is something I'm starting to worry about. With resolutions of functional data ever increasing, we're getting to the point where it's likely we'll need to address more than 2 gigs of memory in the next year or two.
Are there any plans to address this? I'm not sure what you can do with FS-FAST, since it depends on Matlab and I don't know if there are 64-bit versions of matlab available (especially for affordable platforms like AMD64 and the G5).
<aside> I'll insert my plug for moving FS-FAST to python here, as python natively works on 64 bit platforms. </aside>
But with Freesurfer, how hard will it be to build a 64 bit version? If you need a testing machine, I'm working from a dual G5 that would be more than happy to build Freesurfer :-)
Nick
Freesurfer mailing list Freesurfer@nmr.mgh.harvard.edu http://mail.nmr.mgh.harvard.edu/mailman/listinfo/freesurfer
Freesurfer mailing list Freesurfer@nmr.mgh.harvard.edu http://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
i'm currently doing two other 64-bit ports (ppc64 and x86_64). so we are certainly working on that. we'll see what it actually entails once we have a copy of tiger to play with :)
--vicka
On Mon, Jun 28, 2004 at 08:38:14PM -0400, Ray Fix wrote:
FWIW, it looks like Mac OSX Tiger will be able to address 64 bits of virtual memory. Finally, eh. Any plans to utilize this in a future release of FS?
http://www.apple.com/macosx/tiger/64bit.html
Ray
On Apr 15, 2004, at 9:38 PM, Bruce Fischl wrote:
Hi Nick,
unfortunately, os/x doesn't allow you to address more than 4G either. On the bright side though, we have a couple of opteron machines around that we are actively trying to port to, so I suspect we'll have a 64 bit version sometime in the next couple of months. We'll let people know when it's ready (you're welcome to be a beta site).
Bruce
On Thu, 15 Apr 2004, Nick Knouf wrote:
On Apr 15, 2004, at 11:16 AM, Yasunari Tosa wrote:
Currently, "recon-all" script can take the option of "-cm" (conform to min voxel size). This may end up creating more slices than 256, but freesurfer should be able to handle as long as there are enough memory left. Make sure that you have 2 Gbyte of memory and 2 GByte of swap space. Here, we are approaching the 32 bit memory address limit.
{snip}
This 2gig, 32 bit limit is something I'm starting to worry about. With resolutions of functional data ever increasing, we're getting to the point where it's likely we'll need to address more than 2 gigs of memory in the next year or two.
Are there any plans to address this? I'm not sure what you can do with FS-FAST, since it depends on Matlab and I don't know if there are 64-bit versions of matlab available (especially for affordable platforms like AMD64 and the G5).
<aside> I'll insert my plug for moving FS-FAST to python here, as python natively works on 64 bit platforms. </aside>
But with Freesurfer, how hard will it be to build a 64 bit version? If you need a testing machine, I'm working from a dual G5 that would be more than happy to build Freesurfer :-)
Nick
Freesurfer mailing list Freesurfer@nmr.mgh.harvard.edu http://mail.nmr.mgh.harvard.edu/mailman/listinfo/freesurfer
Freesurfer mailing list Freesurfer@nmr.mgh.harvard.edu http://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 -
Denis -
F.Musso -
Geoffrey Aguirre -
Nick Knouf -
Ray Fix -
Vicka Corey -
Yasunari Tosa