Hi FreeSurfers,
After a months of QC'ing FreeSurfer results, my team and I are now working toward end stage analysis of our cortical thickness data. All along, we have been aiming to make improvements in the wm.mgz volume so that we can assure that the white matter and gray matter surfaces are as accurate as possible. This was under the assumption that if the white matter and gray matter surfaces are accurate, then the voxel-wise representation of the white and gray matter in the aparc+aseg.mgz file would be accurate. However, in almost all of our cases, the aparc+aseg.mgz segmentation looks very bad. The bad aparc+aseg.mgz representation of what seem to be accurate white and gray matter segmentations in the wm.mgz file is leading us to believe that the cortical correspondences created after template mapping are wrong.
Does anyone else have an explanation for such a discrepancy? Is this a common problem and if so, is there any way to remedy this situation?
Thanks,
Ryan
Hi Ryan,
what version are you using? And when you say "bad" what exactly do you mean? Can you send some snapshots? If you're doing a thickness study the aparc+aseg is irrelevant - just the white and pial surfaces matter (and the spherical registration of course)
cheers Bruce On Tue, 24 Mar 2009, Ryan Scotton wrote:
Hi FreeSurfers,
After a months of QC'ing FreeSurfer results, my team and I are now working toward end stage analysis of our cortical thickness data. All along, we have been aiming to make improvements in the wm.mgz volume so that we can assure that the white matter and gray matter surfaces are as accurate as possible. This was under the assumption that if the white matter and gray matter surfaces are accurate, then the voxel-wise representation of the white and gray matter in the aparc+aseg.mgz file would be accurate. However, in almost all of our cases, the aparc+aseg.mgz segmentation looks very bad. The bad aparc+aseg.mgz representation of what seem to be accurate white and gray matter segmentations in the wm.mgz file is leading us to believe that the cortical correspondences created after template mapping are wrong.
Does anyone else have an explanation for such a discrepancy? Is this a common problem and if so, is there any way to remedy this situation?
Thanks,
Ryan
Hello again Bruce,
We used FreeSurfer version 3.0.4, so an older version. I've attached 9 screen shots...the first 6 are from the same subject, and the last three (the jpgs with _2 at the end of them) were included just to give a better idea of the problems we're seeing in the aparc+aseg.mgz volume, despite good surfaces in the wm.mgz volume (screenshots of the wm.mgz were not included for the 2nd volume...they are equally accurate as the first case I sent). You'll notice that there are many GM areas that are not segmented and others which have a very fuzzy/noisy boundary. Do you think that there is some kind of error occuring in the spherical registration step? If so, how can one QC this step? Or is it some other issue entirely?
Thanks for your help,
Ryan
On Tue, Mar 24, 2009 at 5:01 PM, Bruce Fischl fischl@nmr.mgh.harvard.eduwrote:
Hi Ryan,
what version are you using? And when you say "bad" what exactly do you mean? Can you send some snapshots? If you're doing a thickness study the aparc+aseg is irrelevant - just the white and pial surfaces matter (and the spherical registration of course)
cheers Bruce
On Tue, 24 Mar 2009, Ryan Scotton wrote:
Hi FreeSurfers,
After a months of QC'ing FreeSurfer results, my team and I are now working toward end stage analysis of our cortical thickness data. All along, we have been aiming to make improvements in the wm.mgz volume so that we can assure that the white matter and gray matter surfaces are as accurate as possible. This was under the assumption that if the white matter and gray matter surfaces are accurate, then the voxel-wise representation of the white and gray matter in the aparc+aseg.mgz file would be accurate. However, in almost all of our cases, the aparc+aseg.mgz segmentation looks very bad. The bad aparc+aseg.mgz representation of what seem to be accurate white and gray matter segmentations in the wm.mgz file is leading us to believe that the cortical correspondences created after template mapping are wrong.
Does anyone else have an explanation for such a discrepancy? Is this a common problem and if so, is there any way to remedy this situation?
Thanks,
Ryan
For whatever reason, my email from this morning doesn't seem to have gone through. Hopefully it will work this time...
Ryan
---------- Forwarded message ---------- From: Ryan Scotton ryan.scotton@gmail.com Date: Thu, Mar 26, 2009 at 10:25 AM Subject: Re: [Freesurfer] Incorrect correspondence in the aparc+aseg.mgz volume? To: Bruce Fischl fischl@nmr.mgh.harvard.edu Cc: freesurfer@nmr.mgh.harvard.edu
Hello again Bruce,
We used FreeSurfer version 3.0.4, so an older version. I've attached 9 screen shots...the first 6 are from the same subject, and the last three (the jpgs with _2 at the end of them) were included just to give a better idea of the problems we're seeing in the aparc+aseg.mgz volume, despite good surfaces in the wm.mgz volume (screenshots of the wm.mgz were not included for the 2nd volume...they are equally accurate as the first case I sent). You'll notice that there are many GM areas that are not segmented and others which have a very fuzzy/noisy boundary. Do you think that there is some kind of error occuring in the spherical registration step? If so, how can one QC this step? Or is it some other issue entirely?
Thanks for your help,
Ryan
On Tue, Mar 24, 2009 at 5:01 PM, Bruce Fischl fischl@nmr.mgh.harvard.eduwrote:
Hi Ryan,
what version are you using? And when you say "bad" what exactly do you mean? Can you send some snapshots? If you're doing a thickness study the aparc+aseg is irrelevant - just the white and pial surfaces matter (and the spherical registration of course)
cheers Bruce
On Tue, 24 Mar 2009, Ryan Scotton wrote:
Hi FreeSurfers,
After a months of QC'ing FreeSurfer results, my team and I are now working toward end stage analysis of our cortical thickness data. All along, we have been aiming to make improvements in the wm.mgz volume so that we can assure that the white matter and gray matter surfaces are as accurate as possible. This was under the assumption that if the white matter and gray matter surfaces are accurate, then the voxel-wise representation of the white and gray matter in the aparc+aseg.mgz file would be accurate. However, in almost all of our cases, the aparc+aseg.mgz segmentation looks very bad. The bad aparc+aseg.mgz representation of what seem to be accurate white and gray matter segmentations in the wm.mgz file is leading us to believe that the cortical correspondences created after template mapping are wrong.
Does anyone else have an explanation for such a discrepancy? Is this a common problem and if so, is there any way to remedy this situation?
Thanks,
Ryan
Hi Ryan,
I still don't quite understand what I'm looking at. Maybe you can put a problem subject somewhere we can get to it?
cheers, Bruce
On Thu, 26 Mar 2009, Ryan Scotton wrote:
For whatever reason, my email from this morning doesn't seem to have gone through. Hopefully it will work this time...
Ryan
---------- Forwarded message ---------- From: Ryan Scotton ryan.scotton@gmail.com Date: Thu, Mar 26, 2009 at 10:25 AM Subject: Re: [Freesurfer] Incorrect correspondence in the aparc+aseg.mgz volume? To: Bruce Fischl fischl@nmr.mgh.harvard.edu Cc: freesurfer@nmr.mgh.harvard.edu
Hello again Bruce,
We used FreeSurfer version 3.0.4, so an older version. I've attached 9 screen shots...the first 6 are from the same subject, and the last three (the jpgs with _2 at the end of them) were included just to give a better idea of the problems we're seeing in the aparc+aseg.mgz volume, despite good surfaces in the wm.mgz volume (screenshots of the wm.mgz were not included for the 2nd volume...they are equally accurate as the first case I sent). You'll notice that there are many GM areas that are not segmented and others which have a very fuzzy/noisy boundary. Do you think that there is some kind of error occuring in the spherical registration step? If so, how can one QC this step? Or is it some other issue entirely?
Thanks for your help,
Ryan
On Tue, Mar 24, 2009 at 5:01 PM, Bruce Fischl fischl@nmr.mgh.harvard.eduwrote:
Hi Ryan,
what version are you using? And when you say "bad" what exactly do you mean? Can you send some snapshots? If you're doing a thickness study the aparc+aseg is irrelevant - just the white and pial surfaces matter (and the spherical registration of course)
cheers Bruce
On Tue, 24 Mar 2009, Ryan Scotton wrote:
Hi FreeSurfers,
After a months of QC'ing FreeSurfer results, my team and I are now working toward end stage analysis of our cortical thickness data. All along, we have been aiming to make improvements in the wm.mgz volume so that we can assure that the white matter and gray matter surfaces are as accurate as possible. This was under the assumption that if the white matter and gray matter surfaces are accurate, then the voxel-wise representation of the white and gray matter in the aparc+aseg.mgz file would be accurate. However, in almost all of our cases, the aparc+aseg.mgz segmentation looks very bad. The bad aparc+aseg.mgz representation of what seem to be accurate white and gray matter segmentations in the wm.mgz file is leading us to believe that the cortical correspondences created after template mapping are wrong.
Does anyone else have an explanation for such a discrepancy? Is this a common problem and if so, is there any way to remedy this situation?
Thanks,
Ryan
Hi Bruce,
I've uploaded an example of a case with fairly accurate white and pial boundries when viewed on top of the wm.mgz volume, and GM areas that are not segmented and others which have a very fuzzy/noisy boundary within the aparc+aseg.mgz volume.
Also, I'm not sure what email client you're using, but since my first email failed to come through, I ended up forwarding it and sending it to you again yesterday, which is the one you received. I noticed that in gmail, it will hide the quoted text, which contained my best description of the problems we're seeing and my FS version (3.0.4). You may have already read it, I'm not sure, but just in case, it's inline with this email.
The file I uploaded is named 5008-003-02_scotton.tar.gz
Thanks so much for your help,
Ryan
On Thu, Mar 26, 2009 at 5:36 PM, Bruce Fischl fischl@nmr.mgh.harvard.eduwrote:
Hi Ryan,
I still don't quite understand what I'm looking at. Maybe you can put a problem subject somewhere we can get to it?
cheers, Bruce
On Thu, 26 Mar 2009, Ryan Scotton wrote:
For whatever reason, my email from this morning doesn't seem to have gone
through. Hopefully it will work this time...
Ryan
---------- Forwarded message ---------- From: Ryan Scotton ryan.scotton@gmail.com Date: Thu, Mar 26, 2009 at 10:25 AM Subject: Re: [Freesurfer] Incorrect correspondence in the aparc+aseg.mgz volume? To: Bruce Fischl fischl@nmr.mgh.harvard.edu Cc: freesurfer@nmr.mgh.harvard.edu
Hello again Bruce,
We used FreeSurfer version 3.0.4, so an older version. I've attached 9 screen shots...the first 6 are from the same subject, and the last three (the jpgs with _2 at the end of them) were included just to give a better idea of the problems we're seeing in the aparc+aseg.mgz volume, despite good surfaces in the wm.mgz volume (screenshots of the wm.mgz were not included for the 2nd volume...they are equally accurate as the first case I sent). You'll notice that there are many GM areas that are not segmented and others which have a very fuzzy/noisy boundary. Do you think that there is some kind of error occuring in the spherical registration step? If so, how can one QC this step? Or is it some other issue entirely?
Thanks for your help,
Ryan
On Tue, Mar 24, 2009 at 5:01 PM, Bruce Fischl <fischl@nmr.mgh.harvard.edu
wrote:
Hi Ryan,
what version are you using? And when you say "bad" what exactly do you mean? Can you send some snapshots? If you're doing a thickness study the aparc+aseg is irrelevant - just the white and pial surfaces matter (and the spherical registration of course)
cheers Bruce
On Tue, 24 Mar 2009, Ryan Scotton wrote:
Hi FreeSurfers,
After a months of QC'ing FreeSurfer results, my team and I are now working toward end stage analysis of our cortical thickness data. All along, we have been aiming to make improvements in the wm.mgz volume so that we can assure that the white matter and gray matter surfaces are as accurate as possible. This was under the assumption that if the white matter and gray matter surfaces are accurate, then the voxel-wise representation of the white and gray matter in the aparc+aseg.mgz file would be accurate. However, in almost all of our cases, the aparc+aseg.mgz segmentation looks very bad. The bad aparc+aseg.mgz representation of what seem to be accurate white and gray matter segmentations in the wm.mgz file is leading us to believe that the cortical correspondences created after template mapping are wrong.
Does anyone else have an explanation for such a discrepancy? Is this a common problem and if so, is there any way to remedy this situation?
Thanks,
Ryan
ok, we'll take a look On Fri, 27 Mar 2009, Ryan Scotton wrote:
Hi Bruce,
I've uploaded an example of a case with fairly accurate white and pial boundries when viewed on top of the wm.mgz volume, and GM areas that are not segmented and others which have a very fuzzy/noisy boundary within the aparc+aseg.mgz volume.
Also, I'm not sure what email client you're using, but since my first email failed to come through, I ended up forwarding it and sending it to you again yesterday, which is the one you received. I noticed that in gmail, it will hide the quoted text, which contained my best description of the problems we're seeing and my FS version (3.0.4). You may have already read it, I'm not sure, but just in case, it's inline with this email.
The file I uploaded is named 5008-003-02_scotton.tar.gz
Thanks so much for your help,
Ryan
On Thu, Mar 26, 2009 at 5:36 PM, Bruce Fischl fischl@nmr.mgh.harvard.eduwrote:
Hi Ryan,
I still don't quite understand what I'm looking at. Maybe you can put a problem subject somewhere we can get to it?
cheers, Bruce
On Thu, 26 Mar 2009, Ryan Scotton wrote:
For whatever reason, my email from this morning doesn't seem to have gone
through. Hopefully it will work this time...
Ryan
---------- Forwarded message ---------- From: Ryan Scotton ryan.scotton@gmail.com Date: Thu, Mar 26, 2009 at 10:25 AM Subject: Re: [Freesurfer] Incorrect correspondence in the aparc+aseg.mgz volume? To: Bruce Fischl fischl@nmr.mgh.harvard.edu Cc: freesurfer@nmr.mgh.harvard.edu
Hello again Bruce,
We used FreeSurfer version 3.0.4, so an older version. I've attached 9 screen shots...the first 6 are from the same subject, and the last three (the jpgs with _2 at the end of them) were included just to give a better idea of the problems we're seeing in the aparc+aseg.mgz volume, despite good surfaces in the wm.mgz volume (screenshots of the wm.mgz were not included for the 2nd volume...they are equally accurate as the first case I sent). You'll notice that there are many GM areas that are not segmented and others which have a very fuzzy/noisy boundary. Do you think that there is some kind of error occuring in the spherical registration step? If so, how can one QC this step? Or is it some other issue entirely?
Thanks for your help,
Ryan
On Tue, Mar 24, 2009 at 5:01 PM, Bruce Fischl <fischl@nmr.mgh.harvard.edu
wrote:
Hi Ryan,
what version are you using? And when you say "bad" what exactly do you mean? Can you send some snapshots? If you're doing a thickness study the aparc+aseg is irrelevant - just the white and pial surfaces matter (and the spherical registration of course)
cheers Bruce
On Tue, 24 Mar 2009, Ryan Scotton wrote:
Hi FreeSurfers,
After a months of QC'ing FreeSurfer results, my team and I are now working toward end stage analysis of our cortical thickness data. All along, we have been aiming to make improvements in the wm.mgz volume so that we can assure that the white matter and gray matter surfaces are as accurate as possible. This was under the assumption that if the white matter and gray matter surfaces are accurate, then the voxel-wise representation of the white and gray matter in the aparc+aseg.mgz file would be accurate. However, in almost all of our cases, the aparc+aseg.mgz segmentation looks very bad. The bad aparc+aseg.mgz representation of what seem to be accurate white and gray matter segmentations in the wm.mgz file is leading us to believe that the cortical correspondences created after template mapping are wrong.
Does anyone else have an explanation for such a discrepancy? Is this a common problem and if so, is there any way to remedy this situation?
Thanks,
Ryan
Ryan,
Be sure to have a look at ribbon.mgz, as that is a volume file created using the white and pial surfaces (from mris_volmask), and would be the most accurate of the volume files in regards to gm segmentation.
Nick
On Fri, 2009-03-27 at 10:53 -0400, Ryan Scotton wrote:
Hi Bruce,
I've uploaded an example of a case with fairly accurate white and pial boundries when viewed on top of the wm.mgz volume, and GM areas that are not segmented and others which have a very fuzzy/noisy boundary within the aparc+aseg.mgz volume.
Also, I'm not sure what email client you're using, but since my first email failed to come through, I ended up forwarding it and sending it to you again yesterday, which is the one you received. I noticed that in gmail, it will hide the quoted text, which contained my best description of the problems we're seeing and my FS version (3.0.4). You may have already read it, I'm not sure, but just in case, it's inline with this email.
The file I uploaded is named 5008-003-02_scotton.tar.gz
Thanks so much for your help,
Ryan
On Thu, Mar 26, 2009 at 5:36 PM, Bruce Fischl fischl@nmr.mgh.harvard.edu wrote: Hi Ryan,
I still don't quite understand what I'm looking at. Maybe you can put a problem subject somewhere we can get to it? cheers, Bruce On Thu, 26 Mar 2009, Ryan Scotton wrote: For whatever reason, my email from this morning doesn't seem to have gone through. Hopefully it will work this time... Ryan ---------- Forwarded message ---------- From: Ryan Scotton <ryan.scotton@gmail.com> Date: Thu, Mar 26, 2009 at 10:25 AM Subject: Re: [Freesurfer] Incorrect correspondence in the aparc+aseg.mgz volume? To: Bruce Fischl <fischl@nmr.mgh.harvard.edu> Cc: freesurfer@nmr.mgh.harvard.edu Hello again Bruce, We used FreeSurfer version 3.0.4, so an older version. I've attached 9 screen shots...the first 6 are from the same subject, and the last three (the jpgs with _2 at the end of them) were included just to give a better idea of the problems we're seeing in the aparc +aseg.mgz volume, despite good surfaces in the wm.mgz volume (screenshots of the wm.mgz were not included for the 2nd volume...they are equally accurate as the first case I sent). You'll notice that there are many GM areas that are not segmented and others which have a very fuzzy/noisy boundary. Do you think that there is some kind of error occuring in the spherical registration step? If so, how can one QC this step? Or is it some other issue entirely? Thanks for your help, Ryan On Tue, Mar 24, 2009 at 5:01 PM, Bruce Fischl <fischl@nmr.mgh.harvard.edu>wrote: Hi Ryan, what version are you using? And when you say "bad" what exactly do you mean? Can you send some snapshots? If you're doing a thickness study the aparc+aseg is irrelevant - just the white and pial surfaces matter (and the spherical registration of course) cheers Bruce On Tue, 24 Mar 2009, Ryan Scotton wrote: Hi FreeSurfers, After a months of QC'ing FreeSurfer results, my team and I are now working toward end stage analysis of our cortical thickness data. All along, we have been aiming to make improvements in the wm.mgz volume so that we can assure that the white matter and gray matter surfaces are as accurate as possible. This was under the assumption that if the white matter and gray matter surfaces are accurate, then the voxel-wise representation of the white and gray matter in the aparc +aseg.mgz file would be accurate. However, in almost all of our cases, the aparc+aseg.mgz segmentation looks very bad. The bad aparc+aseg.mgz representation of what seem to be accurate white and gray matter segmentations in the wm.mgz file is leading us to believe that the cortical correspondences created after template mapping are wrong. Does anyone else have an explanation for such a discrepancy? Is this a common problem and if so, is there any way to remedy this situation? Thanks, Ryan
Freesurfer mailing list Freesurfer@nmr.mgh.harvard.edu https://mail.nmr.mgh.harvard.edu/mailman/listinfo/freesurfer
Hi Ryan,
I think a lot of the spottyness you are seeing is a bug in FS3 that was fixed. Try using the new FS to just regenerate the aparc+aseg and see if it looks better.
Bruce On Fri, 27 Mar 2009, Nick Schmansky wrote:
Ryan,
Be sure to have a look at ribbon.mgz, as that is a volume file created using the white and pial surfaces (from mris_volmask), and would be the most accurate of the volume files in regards to gm segmentation.
Nick
On Fri, 2009-03-27 at 10:53 -0400, Ryan Scotton wrote:
Hi Bruce,
I've uploaded an example of a case with fairly accurate white and pial boundries when viewed on top of the wm.mgz volume, and GM areas that are not segmented and others which have a very fuzzy/noisy boundary within the aparc+aseg.mgz volume.
Also, I'm not sure what email client you're using, but since my first email failed to come through, I ended up forwarding it and sending it to you again yesterday, which is the one you received. I noticed that in gmail, it will hide the quoted text, which contained my best description of the problems we're seeing and my FS version (3.0.4). You may have already read it, I'm not sure, but just in case, it's inline with this email.
The file I uploaded is named 5008-003-02_scotton.tar.gz
Thanks so much for your help,
Ryan
On Thu, Mar 26, 2009 at 5:36 PM, Bruce Fischl fischl@nmr.mgh.harvard.edu wrote: Hi Ryan,
I still don't quite understand what I'm looking at. Maybe you can put a problem subject somewhere we can get to it? cheers, Bruce On Thu, 26 Mar 2009, Ryan Scotton wrote: For whatever reason, my email from this morning doesn't seem to have gone through. Hopefully it will work this time... Ryan ---------- Forwarded message ---------- From: Ryan Scotton <ryan.scotton@gmail.com> Date: Thu, Mar 26, 2009 at 10:25 AM Subject: Re: [Freesurfer] Incorrect correspondence in the aparc+aseg.mgz volume? To: Bruce Fischl <fischl@nmr.mgh.harvard.edu> Cc: freesurfer@nmr.mgh.harvard.edu Hello again Bruce, We used FreeSurfer version 3.0.4, so an older version. I've attached 9 screen shots...the first 6 are from the same subject, and the last three (the jpgs with _2 at the end of them) were included just to give a better idea of the problems we're seeing in the aparc +aseg.mgz volume, despite good surfaces in the wm.mgz volume (screenshots of the wm.mgz were not included for the 2nd volume...they are equally accurate as the first case I sent). You'll notice that there are many GM areas that are not segmented and others which have a very fuzzy/noisy boundary. Do you think that there is some kind of error occuring in the spherical registration step? If so, how can one QC this step? Or is it some other issue entirely? Thanks for your help, Ryan On Tue, Mar 24, 2009 at 5:01 PM, Bruce Fischl <fischl@nmr.mgh.harvard.edu>wrote: Hi Ryan, what version are you using? And when you say "bad" what exactly do you mean? Can you send some snapshots? If you're doing a thickness study the aparc+aseg is irrelevant - just the white and pial surfaces matter (and the spherical registration of course) cheers Bruce On Tue, 24 Mar 2009, Ryan Scotton wrote: Hi FreeSurfers, After a months of QC'ing FreeSurfer results, my team and I are now working toward end stage analysis of our cortical thickness data. All along, we have been aiming to make improvements in the wm.mgz volume so that we can assure that the white matter and gray matter surfaces are as accurate as possible. This was under the assumption that if the white matter and gray matter surfaces are accurate, then the voxel-wise representation of the white and gray matter in the aparc +aseg.mgz file would be accurate. However, in almost all of our cases, the aparc+aseg.mgz segmentation looks very bad. The bad aparc+aseg.mgz representation of what seem to be accurate white and gray matter segmentations in the wm.mgz file is leading us to believe that the cortical correspondences created after template mapping are wrong. Does anyone else have an explanation for such a discrepancy? Is this a common problem and if so, is there any way to remedy this situation? Thanks, Ryan
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
Hi guys,
Does this mean that the measurements are probably solid, but the aparc+aseg.mgz volume inaccurately represents these measurements, or will I need to recompute the last two steps (aparc2aseg and wmparc) in order to attain accurate cortical thickness values? It doesn't like that way, since the thickness files are generated earlier in the recon stream. I ask this because we are trying to get the stats as fast as possible without having to rerun all of our cases (200+) over again.
Ryan
---------- Forwarded message ---------- From: Bruce Fischl fischl@nmr.mgh.harvard.edu Date: Fri, Mar 27, 2009 at 2:47 PM Subject: Re: Fwd: [Freesurfer] Incorrect correspondence in the aparc+aseg.mgz volume? To: Nick Schmansky nicks@nmr.mgh.harvard.edu Cc: Ryan Scotton ryan.scotton@gmail.com, Freesurfer Mailing List < freesurfer@nmr.mgh.harvard.edu>
Hi Ryan,
I think a lot of the spottyness you are seeing is a bug in FS3 that was fixed. Try using the new FS to just regenerate the aparc+aseg and see if it looks better.
Bruce
On Fri, 27 Mar 2009, Nick Schmansky wrote:
Ryan,
Be sure to have a look at ribbon.mgz, as that is a volume file created using the white and pial surfaces (from mris_volmask), and would be the most accurate of the volume files in regards to gm segmentation.
Nick
On Fri, 2009-03-27 at 10:53 -0400, Ryan Scotton wrote:
Hi Bruce,
I've uploaded an example of a case with fairly accurate white and pial boundries when viewed on top of the wm.mgz volume, and GM areas that are not segmented and others which have a very fuzzy/noisy boundary within the aparc+aseg.mgz volume.
Also, I'm not sure what email client you're using, but since my first email failed to come through, I ended up forwarding it and sending it to you again yesterday, which is the one you received. I noticed that in gmail, it will hide the quoted text, which contained my best description of the problems we're seeing and my FS version (3.0.4). You may have already read it, I'm not sure, but just in case, it's inline with this email.
The file I uploaded is named 5008-003-02_scotton.tar.gz
Thanks so much for your help,
Ryan
On Thu, Mar 26, 2009 at 5:36 PM, Bruce Fischl fischl@nmr.mgh.harvard.edu wrote: Hi Ryan,
I still don't quite understand what I'm looking at. Maybe you can put a problem subject somewhere we can get to it? cheers, Bruce On Thu, 26 Mar 2009, Ryan Scotton wrote: For whatever reason, my email from this morning doesn't seem to have gone through. Hopefully it will work this time... Ryan ---------- Forwarded message ---------- From: Ryan Scotton <ryan.scotton@gmail.com> Date: Thu, Mar 26, 2009 at 10:25 AM Subject: Re: [Freesurfer] Incorrect correspondence in the aparc+aseg.mgz volume? To: Bruce Fischl <fischl@nmr.mgh.harvard.edu> Cc: freesurfer@nmr.mgh.harvard.edu Hello again Bruce, We used FreeSurfer version 3.0.4, so an older version. I've attached 9 screen shots...the first 6 are from the same subject, and the last three (the jpgs with _2 at the end of them) were included just to give a better idea of the problems we're seeing in the aparc +aseg.mgz volume, despite good surfaces in the wm.mgz volume (screenshots of the wm.mgz were not included for the 2nd volume...they are equally accurate as the first case I sent). You'll notice that there are many GM areas that are not segmented and others which have a very fuzzy/noisy boundary. Do you think that there is some kind of error occuring in the spherical registration step? If so, how can one QC this step? Or is it some other issue entirely? Thanks for your help, Ryan On Tue, Mar 24, 2009 at 5:01 PM, Bruce Fischl <fischl@nmr.mgh.harvard.edu>wrote: Hi Ryan, what version are you using? And when you say "bad" what exactly do you mean? Can you send some snapshots? If you're doing a thickness study the aparc+aseg is irrelevant - just the white and pial surfaces matter (and the spherical registration of course) cheers Bruce On Tue, 24 Mar 2009, Ryan Scotton wrote: Hi FreeSurfers, After a months of QC'ing FreeSurfer results, my team and I are now working toward end stage analysis of our cortical thickness data. All along, we have been aiming to make improvements in the wm.mgz volume so that we can assure that the white matter and gray matter surfaces are as accurate as possible. This was under the assumption that if the white matter and gray matter surfaces are accurate, then the voxel-wise representation of the white and gray matter in the aparc +aseg.mgz file would be accurate. However, in almost all of our cases, the aparc+aseg.mgz segmentation looks very bad. The bad aparc+aseg.mgz representation of what seem to be accurate white and gray matter segmentations in the wm.mgz file is leading us to believe that the cortical correspondences created after template mapping are wrong. Does anyone else have an explanation for such a discrepancy? Is this a common problem and if so, is there any way to remedy this situation? Thanks, Ryan
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
yes, the thickness is independent of these. I think the only issue are the spots in the aparc+aseg, which I think we've fixed. You should be able to just rerun the very last (Fast) steps to regenerate them.
cheers, Bruce
On Fri, 27 Mar 2009, Ryan Scotton wrote:
Hi guys,
Does this mean that the measurements are probably solid, but the aparc+aseg.mgz volume inaccurately represents these measurements, or will I need to recompute the last two steps (aparc2aseg and wmparc) in order to attain accurate cortical thickness values? It doesn't like that way, since the thickness files are generated earlier in the recon stream. I ask this because we are trying to get the stats as fast as possible without having to rerun all of our cases (200+) over again.
Ryan
---------- Forwarded message ---------- From: Bruce Fischl fischl@nmr.mgh.harvard.edu Date: Fri, Mar 27, 2009 at 2:47 PM Subject: Re: Fwd: [Freesurfer] Incorrect correspondence in the aparc+aseg.mgz volume? To: Nick Schmansky nicks@nmr.mgh.harvard.edu Cc: Ryan Scotton ryan.scotton@gmail.com, Freesurfer Mailing List < freesurfer@nmr.mgh.harvard.edu>
Hi Ryan,
I think a lot of the spottyness you are seeing is a bug in FS3 that was fixed. Try using the new FS to just regenerate the aparc+aseg and see if it looks better.
Bruce
On Fri, 27 Mar 2009, Nick Schmansky wrote:
Ryan,
Be sure to have a look at ribbon.mgz, as that is a volume file created using the white and pial surfaces (from mris_volmask), and would be the most accurate of the volume files in regards to gm segmentation.
Nick
On Fri, 2009-03-27 at 10:53 -0400, Ryan Scotton wrote:
Hi Bruce,
I've uploaded an example of a case with fairly accurate white and pial boundries when viewed on top of the wm.mgz volume, and GM areas that are not segmented and others which have a very fuzzy/noisy boundary within the aparc+aseg.mgz volume.
Also, I'm not sure what email client you're using, but since my first email failed to come through, I ended up forwarding it and sending it to you again yesterday, which is the one you received. I noticed that in gmail, it will hide the quoted text, which contained my best description of the problems we're seeing and my FS version (3.0.4). You may have already read it, I'm not sure, but just in case, it's inline with this email.
The file I uploaded is named 5008-003-02_scotton.tar.gz
Thanks so much for your help,
Ryan
On Thu, Mar 26, 2009 at 5:36 PM, Bruce Fischl fischl@nmr.mgh.harvard.edu wrote: Hi Ryan,
I still don't quite understand what I'm looking at. Maybe you can put a problem subject somewhere we can get to it? cheers, Bruce On Thu, 26 Mar 2009, Ryan Scotton wrote: For whatever reason, my email from this morning doesn't seem to have gone through. Hopefully it will work this time... Ryan ---------- Forwarded message ---------- From: Ryan Scotton <ryan.scotton@gmail.com> Date: Thu, Mar 26, 2009 at 10:25 AM Subject: Re: [Freesurfer] Incorrect correspondence in the aparc+aseg.mgz volume? To: Bruce Fischl <fischl@nmr.mgh.harvard.edu> Cc: freesurfer@nmr.mgh.harvard.edu Hello again Bruce, We used FreeSurfer version 3.0.4, so an older version. I've attached 9 screen shots...the first 6 are from the same subject, and the last three (the jpgs with _2 at the end of them) were included just to give a better idea of the problems we're seeing in the aparc +aseg.mgz volume, despite good surfaces in the wm.mgz volume (screenshots of the wm.mgz were not included for the 2nd volume...they are equally accurate as the first case I sent). You'll notice that there are many GM areas that are not segmented and others which have a very fuzzy/noisy boundary. Do you think that there is some kind of error occuring in the spherical registration step? If so, how can one QC this step? Or is it some other issue entirely? Thanks for your help, Ryan On Tue, Mar 24, 2009 at 5:01 PM, Bruce Fischl <fischl@nmr.mgh.harvard.edu>wrote: Hi Ryan, what version are you using? And when you say "bad" what exactly do you mean? Can you send some snapshots? If you're doing a thickness study the aparc+aseg is irrelevant - just the white and pial surfaces matter (and the spherical registration of course) cheers Bruce On Tue, 24 Mar 2009, Ryan Scotton wrote: Hi FreeSurfers, After a months of QC'ing FreeSurfer results, my team and I are now working toward end stage analysis of our cortical thickness data. All along, we have been aiming to make improvements in the wm.mgz volume so that we can assure that the white matter and gray matter surfaces are as accurate as possible. This was under the assumption that if the white matter and gray matter surfaces are accurate, then the voxel-wise representation of the white and gray matter in the aparc +aseg.mgz file would be accurate. However, in almost all of our cases, the aparc+aseg.mgz segmentation looks very bad. The bad aparc+aseg.mgz representation of what seem to be accurate white and gray matter segmentations in the wm.mgz file is leading us to believe that the cortical correspondences created after template mapping are wrong. Does anyone else have an explanation for such a discrepancy? Is this a common problem and if so, is there any way to remedy this situation? Thanks, Ryan
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
Thanks for your help, Nick and Bruce.
I'm trying to regenerate this volume in order to verify that this is, in fact, an older bug that has been resolved in FS4. I'm assuming that by the last (fast) steps, you meant aparc2aseg and wmparc, so I used this command line: recon-all -s 2368-004-01 -aparc2aseg -wmparc > aparc_fix.txt &
However, I keep getting this error (I included a little bit before the error just to give an idea of where it was at in the aparc2aseg process):
Loading rh annotations from /Autism/autism/longitudinal/2368-004/2368-004-01/FreeSurfer/2368-004-01/label/rh.aparc.annot Have color table for lh white annotation Have color table for rh white annotation Loading ribbon segmentation from /Autism/autism/longitudinal/2368-004/2368-004-01/FreeSurfer/2368-004-01/mri/ribbon.mgz ERROR: cannot find /Autism/autism/longitudinal/2368-004/2368-004-01/FreeSurfer/2368-004-01/mri/ribbon.mgz ERROR: loading /Autism/autism/longitudinal/2368-004/2368-004-01/FreeSurfer/2368-004-01/mri/ribbon.mgz Linux gaea.ia.unc.edu 2.6.18-92.1.22.el5 #1 SMP Tue Dec 16 11:57:43 EST 2008 x86_64 x86_64 x86_64 GNU/Linux
recon-all exited with ERRORS at Mon Mar 30 10:31:11 EDT 2009
Any idea of what's going on here? I find it perplexing that it's searching for a ribbon.mgz rather than an ?h.ribbon.mgz, since I found no evidence on the wiki of a file just called ribbon.mgz....
Ryan
On Fri, Mar 27, 2009 at 4:49 PM, Bruce Fischl fischl@nmr.mgh.harvard.eduwrote:
yes, the thickness is independent of these. I think the only issue are the spots in the aparc+aseg, which I think we've fixed. You should be able to just rerun the very last (Fast) steps to regenerate them.
cheers, Bruce
On Fri, 27 Mar 2009, Ryan Scotton wrote:
Hi guys,
Does this mean that the measurements are probably solid, but the aparc+aseg.mgz volume inaccurately represents these measurements, or will I need to recompute the last two steps (aparc2aseg and wmparc) in order to attain accurate cortical thickness values? It doesn't like that way, since the thickness files are generated earlier in the recon stream. I ask this because we are trying to get the stats as fast as possible without having to rerun all of our cases (200+) over again.
Ryan
---------- Forwarded message ---------- From: Bruce Fischl fischl@nmr.mgh.harvard.edu Date: Fri, Mar 27, 2009 at 2:47 PM Subject: Re: Fwd: [Freesurfer] Incorrect correspondence in the aparc+aseg.mgz volume? To: Nick Schmansky nicks@nmr.mgh.harvard.edu Cc: Ryan Scotton ryan.scotton@gmail.com, Freesurfer Mailing List < freesurfer@nmr.mgh.harvard.edu>
Hi Ryan,
I think a lot of the spottyness you are seeing is a bug in FS3 that was fixed. Try using the new FS to just regenerate the aparc+aseg and see if it looks better.
Bruce
On Fri, 27 Mar 2009, Nick Schmansky wrote:
Ryan,
Be sure to have a look at ribbon.mgz, as that is a volume file created using the white and pial surfaces (from mris_volmask), and would be the most accurate of the volume files in regards to gm segmentation.
Nick
On Fri, 2009-03-27 at 10:53 -0400, Ryan Scotton wrote:
Hi Bruce,
I've uploaded an example of a case with fairly accurate white and pial boundries when viewed on top of the wm.mgz volume, and GM areas that are not segmented and others which have a very fuzzy/noisy boundary within the aparc+aseg.mgz volume.
Also, I'm not sure what email client you're using, but since my first email failed to come through, I ended up forwarding it and sending it to you again yesterday, which is the one you received. I noticed that in gmail, it will hide the quoted text, which contained my best description of the problems we're seeing and my FS version (3.0.4). You may have already read it, I'm not sure, but just in case, it's inline with this email.
The file I uploaded is named 5008-003-02_scotton.tar.gz
Thanks so much for your help,
Ryan
On Thu, Mar 26, 2009 at 5:36 PM, Bruce Fischl fischl@nmr.mgh.harvard.edu wrote: Hi Ryan,
I still don't quite understand what I'm looking at. Maybe you can put a problem subject somewhere we can get to it? cheers, Bruce On Thu, 26 Mar 2009, Ryan Scotton wrote: For whatever reason, my email from this morning doesn't seem to have gone through. Hopefully it will work this time... Ryan ---------- Forwarded message ---------- From: Ryan Scotton <ryan.scotton@gmail.com> Date: Thu, Mar 26, 2009 at 10:25 AM Subject: Re: [Freesurfer] Incorrect correspondence in the aparc+aseg.mgz volume? To: Bruce Fischl <fischl@nmr.mgh.harvard.edu> Cc: freesurfer@nmr.mgh.harvard.edu Hello again Bruce, We used FreeSurfer version 3.0.4, so an older version. I've attached 9 screen shots...the first 6 are from the same subject, and the last three (the jpgs with _2 at the end of them) were included just to give a better idea of the problems we're seeing in the aparc +aseg.mgz volume, despite good surfaces in the wm.mgz volume (screenshots of the wm.mgz were not included for the 2nd volume...they are equally accurate as the first case I sent). You'll notice that there are many GM areas that are not segmented and others which have a very fuzzy/noisy boundary. Do you think that there is some kind of error occuring in the spherical registration step? If so, how can one QC this step? Or is it some other issue entirely? Thanks for your help, Ryan On Tue, Mar 24, 2009 at 5:01 PM, Bruce Fischl <fischl@nmr.mgh.harvard.edu>wrote: Hi Ryan, what version are you using? And when you say "bad" what exactly do you mean? Can you send some snapshots? If you're doing a thickness study the aparc+aseg is irrelevant - just the white and pial surfaces matter (and the spherical registration of course) cheers Bruce On Tue, 24 Mar 2009, Ryan Scotton wrote: Hi FreeSurfers, After a months of QC'ing FreeSurfer results, my team and I are now working toward end stage analysis of our cortical thickness data. All along, we have been aiming to make improvements in the wm.mgz volume so that we can assure that the white matter and gray matter surfaces are as accurate as possible. This was under the assumption that if the white matter and gray matter surfaces are accurate, then the voxel-wise representation of the white and gray matter in the aparc +aseg.mgz file would be accurate. However, in almost all of our cases, the aparc+aseg.mgz segmentation looks very bad. The bad aparc+aseg.mgz representation of what seem to be accurate white and gray matter segmentations in the wm.mgz file is leading us to believe that the cortical correspondences created after template mapping are wrong. Does anyone else have an explanation for such a discrepancy? Is this a common problem and if so, is there any way to remedy this situation? Thanks, Ryan
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
freesurfer@nmr.mgh.harvard.edu
-
Bruce Fischl -
Nick Schmansky -
Ryan Scotton