Hi, I had a couple of questions regarding cortical/subcortical segmentation-
1. In our cortical segmentation we keep getting what look to be inaccurate segmentation of the rostral and caudal acc. It looks some of what should be labeled as one of those two areas is getting labelled corpus callosum or "unlabelled subcortical" instead, sometimes fairly large chunks. Does this indicate any problem earlier in the processing stream, and is there anything that can be done to prevent this? If not I saw that both tksurfer and tkmedit can be used to edit the segmentation, what would be the easiest way to do this?
2. When registering a freesurfer anatomical to Feat output (using reg-feat2anat) the registration seems to be somewhat inaccurate. Registering brainmask.mgz to our example_func using flirt has given us a better registration so far. Since reg-feat2anat seems to use flirt in its registration I was wondering what the difference might be resulting from. Does reg-feat2anat use T1.mgz as it's reference scan, and if so is there any way to have it use brainmask instead, as that seems to give better results?
Thanks, Jeff
Hi Jeff,
With regards to the cortical parcellation and the mislabelling of the anterior cingulate cortex (rostral anterior cingulate in the parcellation atlas), we are aware of this as a problem and are in the process of fixing it as we speak. If you were interested in editing the parcellation labels, my suggestion would be to first convert the *.aparc.annot file into individual labels (i.e. *h.rostralanteriorcingulate.label, *h.corpuscallosum.label) and edit the individual labels using tksurfer. You can view the parcellation labels with the curvature visible (Under View-->Label Style-->choose Outline instead of Filled) to differentiate cortical areas from non-cortical areas (i.e. corpus callosum, unknown).
best,
Rahul
On Tue, 20 Jun 2006, Jeffrey Spielberg wrote:
Hi, I had a couple of questions regarding cortical/subcortical segmentation-
- In our cortical segmentation we keep getting what look to be inaccurate
segmentation of the rostral and caudal acc. It looks some of what should be labeled as one of those two areas is getting labelled corpus callosum or "unlabelled subcortical" instead, sometimes fairly large chunks. Does this indicate any problem earlier in the processing stream, and is there anything that can be done to prevent this? If not I saw that both tksurfer and tkmedit can be used to edit the segmentation, what would be the easiest way to do this?
- When registering a freesurfer anatomical to Feat output (using
reg-feat2anat) the registration seems to be somewhat inaccurate. Registering brainmask.mgz to our example_func using flirt has given us a better registration so far. Since reg-feat2anat seems to use flirt in its registration I was wondering what the difference might be resulting from. Does reg-feat2anat use T1.mgz as it's reference scan, and if so is there any way to have it use brainmask instead, as that seems to give better results?
Thanks, Jeff
Hi, thanks, would the fix for this problem be released as something that I then would use to rerun mris_ca_label (and mris_anatomical_stats and mri_aprac2aseg) and get better results, or would I have to rerun a whole bunch of earlier steps (so I would want to hold off with processing for now)? Jeff
On 6/20/06, Rahul Desikan rahul@nmr.mgh.harvard.edu wrote:
Hi Jeff,
With regards to the cortical parcellation and the mislabelling of the anterior cingulate cortex (rostral anterior cingulate in the parcellation atlas), we are aware of this as a problem and are in the process of fixing it as we speak. If you were interested in editing the parcellation labels, my suggestion would be to first convert the *.aparc.annot file into individual labels (i.e. *h.rostralanteriorcingulate.label, *h.corpuscallosum.label) and edit the individual labels using tksurfer. You can view the parcellation labels with the curvature visible (Under View-->Label Style-->choose Outline instead of Filled) to differentiate cortical areas from non-cortical areas (i.e. corpus callosum, unknown).
best,
Rahul
On Tue, 20 Jun 2006, Jeffrey Spielberg wrote:
Hi, I had a couple of questions regarding cortical/subcortical
segmentation-
- In our cortical segmentation we keep getting what look to be
inaccurate
segmentation of the rostral and caudal acc. It looks some of what
should be
labeled as one of those two areas is getting labelled corpus callosum or "unlabelled subcortical" instead, sometimes fairly large chunks. Does
this
indicate any problem earlier in the processing stream, and is there
anything
that can be done to prevent this? If not I saw that both tksurfer and tkmedit can be used to edit the segmentation, what would be the easiest
way
to do this?
- When registering a freesurfer anatomical to Feat output (using
reg-feat2anat) the registration seems to be somewhat inaccurate. Registering brainmask.mgz to our example_func using flirt has given us a better registration so far. Since reg-feat2anat seems to use flirt in
its
registration I was wondering what the difference might be resulting
from.
Does reg-feat2anat use T1.mgz as it's reference scan, and if so is there
any
way to have it use brainmask instead, as that seems to give better
results?
Thanks, Jeff
--
hopefully the former, but until we understand what's going on it's hard to say. On Wed, 21 Jun 2006, Jeffrey Spielberg wrote:
Hi, thanks, would the fix for this problem be released as something that I then would use to rerun mris_ca_label (and mris_anatomical_stats and mri_aprac2aseg) and get better results, or would I have to rerun a whole bunch of earlier steps (so I would want to hold off with processing for now)? Jeff
On 6/20/06, Rahul Desikan rahul@nmr.mgh.harvard.edu wrote:
Hi Jeff,
With regards to the cortical parcellation and the mislabelling of the anterior cingulate cortex (rostral anterior cingulate in the parcellation atlas), we are aware of this as a problem and are in the process of fixing it as we speak. If you were interested in editing the parcellation labels, my suggestion would be to first convert the *.aparc.annot file into individual labels (i.e. *h.rostralanteriorcingulate.label, *h.corpuscallosum.label) and edit the individual labels using tksurfer. You can view the parcellation labels with the curvature visible (Under View-->Label Style-->choose Outline instead of Filled) to differentiate cortical areas from non-cortical areas (i.e. corpus callosum, unknown).
best,
Rahul
On Tue, 20 Jun 2006, Jeffrey Spielberg wrote:
Hi, I had a couple of questions regarding cortical/subcortical
segmentation-
- In our cortical segmentation we keep getting what look to be
inaccurate
segmentation of the rostral and caudal acc. It looks some of what
should be
labeled as one of those two areas is getting labelled corpus callosum or "unlabelled subcortical" instead, sometimes fairly large chunks. Does
this
indicate any problem earlier in the processing stream, and is there
anything
that can be done to prevent this? If not I saw that both tksurfer and tkmedit can be used to edit the segmentation, what would be the easiest
way
to do this?
- When registering a freesurfer anatomical to Feat output (using
reg-feat2anat) the registration seems to be somewhat inaccurate. Registering brainmask.mgz to our example_func using flirt has given us a better registration so far. Since reg-feat2anat seems to use flirt in
its
registration I was wondering what the difference might be resulting
from.
Does reg-feat2anat use T1.mgz as it's reference scan, and if so is there
any
way to have it use brainmask instead, as that seems to give better
results?
Thanks, Jeff
--
I'll leave 2 for Doug. In 1. by segmentation do you mean the wm.mgz volume? Or the cortical parcellation? If the latter, which one (as we generate two by default).
cheers, Bruce
On Tue, 20 Jun 2006, Jeffrey Spielberg wrote:
Hi, I had a couple of questions regarding cortical/subcortical segmentation-
- In our cortical segmentation we keep getting what look to be inaccurate
segmentation of the rostral and caudal acc. It looks some of what should be labeled as one of those two areas is getting labelled corpus callosum or "unlabelled subcortical" instead, sometimes fairly large chunks. Does this indicate any problem earlier in the processing stream, and is there anything that can be done to prevent this? If not I saw that both tksurfer and tkmedit can be used to edit the segmentation, what would be the easiest way to do this?
- When registering a freesurfer anatomical to Feat output (using
reg-feat2anat) the registration seems to be somewhat inaccurate. Registering brainmask.mgz to our example_func using flirt has given us a better registration so far. Since reg-feat2anat seems to use flirt in its registration I was wondering what the difference might be resulting from. Does reg-feat2anat use T1.mgz as it's reference scan, and if so is there any way to have it use brainmask instead, as that seems to give better results?
Thanks, Jeff
Hi, sorry, I meant the cortical parcellation, the one that is used in aparc+aseg.mgz (I believe this uses the desikan killiany atlas?), thanks, Jeff
On 6/20/06, Bruce Fischl fischl@nmr.mgh.harvard.edu wrote:
I'll leave 2 for Doug. In 1. by segmentation do you mean the wm.mgz volume? Or the cortical parcellation? If the latter, which one (as we generate two by default).
cheers, Bruce
On Tue, 20 Jun 2006, Jeffrey Spielberg wrote:
Hi, I had a couple of questions regarding cortical/subcortical
segmentation-
- In our cortical segmentation we keep getting what look to be
inaccurate
segmentation of the rostral and caudal acc. It looks some of what
should be
labeled as one of those two areas is getting labelled corpus callosum or "unlabelled subcortical" instead, sometimes fairly large chunks. Does
this
indicate any problem earlier in the processing stream, and is there
anything
that can be done to prevent this? If not I saw that both tksurfer and tkmedit can be used to edit the segmentation, what would be the easiest
way
to do this?
- When registering a freesurfer anatomical to Feat output (using
reg-feat2anat) the registration seems to be somewhat inaccurate. Registering brainmask.mgz to our example_func using flirt has given us a better registration so far. Since reg-feat2anat seems to use flirt in
its
registration I was wondering what the difference might be resulting
from.
Does reg-feat2anat use T1.mgz as it's reference scan, and if so is there
any
way to have it use brainmask instead, as that seems to give better
results?
Thanks, Jeff
okay, Rahul is looking into it. On Wed, 21 Jun 2006, Jeffrey Spielberg wrote:
Hi, sorry, I meant the cortical parcellation, the one that is used in aparc+aseg.mgz (I believe this uses the desikan killiany atlas?), thanks, Jeff
On 6/20/06, Bruce Fischl fischl@nmr.mgh.harvard.edu wrote:
I'll leave 2 for Doug. In 1. by segmentation do you mean the wm.mgz volume? Or the cortical parcellation? If the latter, which one (as we generate two by default).
cheers, Bruce
On Tue, 20 Jun 2006, Jeffrey Spielberg wrote:
Hi, I had a couple of questions regarding cortical/subcortical
segmentation-
- In our cortical segmentation we keep getting what look to be
inaccurate
segmentation of the rostral and caudal acc. It looks some of what
should be
labeled as one of those two areas is getting labelled corpus callosum or "unlabelled subcortical" instead, sometimes fairly large chunks. Does
this
indicate any problem earlier in the processing stream, and is there
anything
that can be done to prevent this? If not I saw that both tksurfer and tkmedit can be used to edit the segmentation, what would be the easiest
way
to do this?
- When registering a freesurfer anatomical to Feat output (using
reg-feat2anat) the registration seems to be somewhat inaccurate. Registering brainmask.mgz to our example_func using flirt has given us a better registration so far. Since reg-feat2anat seems to use flirt in
its
registration I was wondering what the difference might be resulting
from.
Does reg-feat2anat use T1.mgz as it's reference scan, and if so is there
any
way to have it use brainmask instead, as that seems to give better
results?
Thanks, Jeff
reg-feat2anat uses brain.mgz. But you'll have to give us more detail in order to evaluate the problem, "somewhat inaccurate" is not very informative. Please give cmd lines, versions, and log files. reg-feat2anat does use flirt, and it will print out the flirt cmd line. You can compare it to the one that you use. I assume you are using NIFTI?
doug
Bruce Fischl wrote:
I'll leave 2 for Doug. In 1. by segmentation do you mean the wm.mgz volume? Or the cortical parcellation? If the latter, which one (as we generate two by default).
cheers, Bruce
On Tue, 20 Jun 2006, Jeffrey Spielberg wrote:
Hi, I had a couple of questions regarding cortical/subcortical segmentation-
- In our cortical segmentation we keep getting what look to be
inaccurate segmentation of the rostral and caudal acc. It looks some of what should be labeled as one of those two areas is getting labelled corpus callosum or "unlabelled subcortical" instead, sometimes fairly large chunks. Does this indicate any problem earlier in the processing stream, and is there anything that can be done to prevent this? If not I saw that both tksurfer and tkmedit can be used to edit the segmentation, what would be the easiest way to do this?
- When registering a freesurfer anatomical to Feat output (using
reg-feat2anat) the registration seems to be somewhat inaccurate. Registering brainmask.mgz to our example_func using flirt has given us a better registration so far. Since reg-feat2anat seems to use flirt in its registration I was wondering what the difference might be resulting from. Does reg-feat2anat use T1.mgz as it's reference scan, and if so is there any way to have it use brainmask instead, as that seems to give better results?
Thanks, Jeff
Freesurfer mailing list Freesurfer@nmr.mgh.harvard.edu https://mail.nmr.mgh.harvard.edu/mailman/listinfo/freesurfer
freesurfer@nmr.mgh.harvard.edu
-
Bruce Fischl -
Doug Greve -
Jeffrey Spielberg -
Rahul Desikan