Dear all
Actually I would like to perform automated reconstruction for my T1 image and then obtain anatomical labels after cortical parcellation. I have done fully automated reconstruction using recon-all. But for my case, the white and grey matter segmentation appears to be weird. The green line seems laying on the pial surface rather than on the white surface, and the red line seems including some non-brain structures (including small portion of CSF and skull). Thus, the resultant cortical thickness appears to be very thin after the segmentation. I just doubt if anything was going wrong during the procedure. Or need any preprocessing to enhance the segmentation quality before performing the reconstruction? If you have had similar situation as mine, please let me know how to tackle the problem.
Attached please find the images for better illustration.
fs_image_brain.tiff - use brain.mgz as aux volume fs_image_norm.tiff - use norm.mgz as aux volume fs_image_brainmask.tiff - use brainmask.mgz as aux volume and load segmentation file aseg.mgz
Thanks Angel
eek, that looks pretty bad. Can you tell us what the acquisition parameters are? On Thu, 8 Oct 2009, Angel Wong wrote:
Dear all
Actually I would like to perform automated reconstruction for my T1 image and then obtain anatomical labels after cortical parcellation. I have done fully automated reconstruction using recon-all. But for my case, the white and grey matter segmentation appears to be weird. The green line seems laying on the pial surface rather than on the white surface, and the red line seems including some non-brain structures (including small portion of CSF and skull). Thus, the resultant cortical thickness appears to be very thin after the segmentation. I just doubt if anything was going wrong during the procedure. Or need any preprocessing to enhance the segmentation quality before performing the reconstruction? If you have had similar situation as mine, please let me know how to tackle the problem.
Attached please find the images for better illustration.
fs_image_brain.tiff - use brain.mgz as aux volume fs_image_norm.tiff - use norm.mgz as aux volume fs_image_brainmask.tiff - use brainmask.mgz as aux volume and load segmentation file aseg.mgz
Thanks Angel
Hi Bruce
Attached please find the parameters for acquiring the T1 image.
Thank you for your help.
Angel
2009/10/8 Bruce Fischl fischl@nmr.mgh.harvard.edu
eek, that looks pretty bad. Can you tell us what the acquisition parameters are?
On Thu, 8 Oct 2009, Angel Wong wrote:
Dear all
Actually I would like to perform automated reconstruction for my T1 image and then obtain anatomical labels after cortical parcellation. I have done fully automated reconstruction using recon-all. But for my case, the white and grey matter segmentation appears to be weird. The green line seems laying on the pial surface rather than on the white surface, and the red line seems including some non-brain structures (including small portion of CSF and skull). Thus, the resultant cortical thickness appears to be very thin after the segmentation. I just doubt if anything was going wrong during the procedure. Or need any preprocessing to enhance the segmentation quality before performing the reconstruction? If you have had similar situation as mine, please let me know how to tackle the problem.
Attached please find the images for better illustration.
fs_image_brain.tiff - use brain.mgz as aux volume fs_image_norm.tiff - use norm.mgz as aux volume fs_image_brainmask.tiff - use brainmask.mgz as aux volume and load segmentation file aseg.mgz
Thanks Angel
I see. Your TR is quite short and your TE relatively long, so this is probably pretty poor CNR between gray and white. If you send me the orig.mgz I'll take a look and see if it's fixable On Thu, 8 Oct 2009, Angel Wong wrote:
Hi Bruce
Attached please find the parameters for acquiring the T1 image.
Thank you for your help.
Angel
2009/10/8 Bruce Fischl fischl@nmr.mgh.harvard.edu
eek, that looks pretty bad. Can you tell us what the acquisition parameters are?
On Thu, 8 Oct 2009, Angel Wong wrote:
Dear all
Actually I would like to perform automated reconstruction for my T1 image and then obtain anatomical labels after cortical parcellation. I have done fully automated reconstruction using recon-all. But for my case, the white and grey matter segmentation appears to be weird. The green line seems laying on the pial surface rather than on the white surface, and the red line seems including some non-brain structures (including small portion of CSF and skull). Thus, the resultant cortical thickness appears to be very thin after the segmentation. I just doubt if anything was going wrong during the procedure. Or need any preprocessing to enhance the segmentation quality before performing the reconstruction? If you have had similar situation as mine, please let me know how to tackle the problem.
Attached please find the images for better illustration.
fs_image_brain.tiff - use brain.mgz as aux volume fs_image_norm.tiff - use norm.mgz as aux volume fs_image_brainmask.tiff - use brainmask.mgz as aux volume and load segmentation file aseg.mgz
Thanks Angel
Thanks a lot with your help. This is because we want to perform quite a number of functional runs on the same subject at the same session. So we tried to shorten the T1 scan time to minimize the motion artifacts come from the subject.
Attached please find the orig.mgz for your information.
2009/10/8 Bruce Fischl fischl@nmr.mgh.harvard.edu
I see. Your TR is quite short and your TE relatively long, so this is probably pretty poor CNR between gray and white. If you send me the orig.mgz I'll take a look and see if it's fixable
On Thu, 8 Oct 2009, Angel Wong wrote:
Hi Bruce
Attached please find the parameters for acquiring the T1 image.
Thank you for your help.
Angel
2009/10/8 Bruce Fischl fischl@nmr.mgh.harvard.edu
eek, that looks pretty bad. Can you tell us what the acquisition
parameters are?
On Thu, 8 Oct 2009, Angel Wong wrote:
Dear all
Actually I would like to perform automated reconstruction for my T1 image and then obtain anatomical labels after cortical parcellation. I have done fully automated reconstruction using recon-all. But for my case, the white and grey matter segmentation appears to be weird. The green line seems laying on the pial surface rather than on the white surface, and the red line seems including some non-brain structures (including small portion of CSF and skull). Thus, the resultant cortical thickness appears to be very thin after the segmentation. I just doubt if anything was going wrong during the procedure. Or need any preprocessing to enhance the segmentation quality before performing the reconstruction? If you have had similar situation as mine, please let me know how to tackle the problem.
Attached please find the images for better illustration.
fs_image_brain.tiff - use brain.mgz as aux volume fs_image_norm.tiff - use norm.mgz as aux volume fs_image_brainmask.tiff - use brainmask.mgz as aux volume and load segmentation file aseg.mgz
Thanks Angel
Hi Angel,
I'm just about done running the dataset and the results actually look ok (see attached tif). I didn't intervene at all, although you would want to to get rid of some non-brain tissue that the pial surface is grabbing, but that's pretty easy. Did you minimize your TE? I don't know what your gradient set is, but usually you can get your TE down a bit more than that. You also might be better off sacrificing a bit of resolution and lengthening your TR cheers, Bruce
On Thu, 8 Oct 2009, Angel Wong wrote:
Thanks a lot with your help. This is because we want to perform quite a number of functional runs on the same subject at the same session. So we tried to shorten the T1 scan time to minimize the motion artifacts come from the subject.
Attached please find the orig.mgz for your information.
2009/10/8 Bruce Fischl fischl@nmr.mgh.harvard.edu
I see. Your TR is quite short and your TE relatively long, so this is probably pretty poor CNR between gray and white. If you send me the orig.mgz I'll take a look and see if it's fixable
On Thu, 8 Oct 2009, Angel Wong wrote:
Hi Bruce
Attached please find the parameters for acquiring the T1 image.
Thank you for your help.
Angel
2009/10/8 Bruce Fischl fischl@nmr.mgh.harvard.edu
eek, that looks pretty bad. Can you tell us what the acquisition
parameters are?
On Thu, 8 Oct 2009, Angel Wong wrote:
Dear all
Actually I would like to perform automated reconstruction for my T1 image and then obtain anatomical labels after cortical parcellation. I have done fully automated reconstruction using recon-all. But for my case, the white and grey matter segmentation appears to be weird. The green line seems laying on the pial surface rather than on the white surface, and the red line seems including some non-brain structures (including small portion of CSF and skull). Thus, the resultant cortical thickness appears to be very thin after the segmentation. I just doubt if anything was going wrong during the procedure. Or need any preprocessing to enhance the segmentation quality before performing the reconstruction? If you have had similar situation as mine, please let me know how to tackle the problem.
Attached please find the images for better illustration.
fs_image_brain.tiff - use brain.mgz as aux volume fs_image_norm.tiff - use norm.mgz as aux volume fs_image_brainmask.tiff - use brainmask.mgz as aux volume and load segmentation file aseg.mgz
Thanks Angel
Hi Bruce
Thank you very much. The result looks much better than mine. How did you do that? Please let me know so that I can modify the other datasets if I face the same situation again.
Cheers Angel
2009/10/10 Bruce Fischl fischl@nmr.mgh.harvard.edu
Hi Angel,
I'm just about done running the dataset and the results actually look ok (see attached tif). I didn't intervene at all, although you would want to to get rid of some non-brain tissue that the pial surface is grabbing, but that's pretty easy. Did you minimize your TE? I don't know what your gradient set is, but usually you can get your TE down a bit more than that. You also might be better off sacrificing a bit of resolution and lengthening your TR cheers, Bruce
On Thu, 8 Oct 2009, Angel Wong wrote:
Thanks a lot with your help. This is because we want to perform quite a
number of functional runs on the same subject at the same session. So we tried to shorten the T1 scan time to minimize the motion artifacts come from the subject.
Attached please find the orig.mgz for your information.
2009/10/8 Bruce Fischl fischl@nmr.mgh.harvard.edu
I see. Your TR is quite short and your TE relatively long, so this is
probably pretty poor CNR between gray and white. If you send me the orig.mgz I'll take a look and see if it's fixable
On Thu, 8 Oct 2009, Angel Wong wrote:
Hi Bruce
Attached please find the parameters for acquiring the T1 image.
Thank you for your help.
Angel
2009/10/8 Bruce Fischl fischl@nmr.mgh.harvard.edu
eek, that looks pretty bad. Can you tell us what the acquisition
parameters are?
On Thu, 8 Oct 2009, Angel Wong wrote:
Dear all
Actually I would like to perform automated reconstruction for my T1 image and then obtain anatomical labels after cortical parcellation. I have done fully automated reconstruction using recon-all. But for my case, the white and grey matter segmentation appears to be weird. The green line seems laying on the pial surface rather than on the white surface, and the red line seems including some non-brain structures (including small portion of CSF and skull). Thus, the resultant cortical thickness appears to be very thin after the segmentation. I just doubt if anything was going wrong during the procedure. Or need any preprocessing to enhance the segmentation quality before performing the reconstruction? If you have had similar situation as mine, please let me know how to tackle the problem.
Attached please find the images for better illustration.
fs_image_brain.tiff - use brain.mgz as aux volume fs_image_norm.tiff - use norm.mgz as aux volume fs_image_brainmask.tiff - use brainmask.mgz as aux volume and load segmentation file aseg.mgz
Thanks Angel
Freesurfer mailing list Freesurfer@nmr.mgh.harvard.edu https://mail.nmr.mgh.harvard.edu/mailman/listinfo/freesurfer
I didn't do anything but run recon-all with the latest version. Remind me what version you were running? On Sat, 10 Oct 2009, Angel Wong wrote:
Hi Bruce
Thank you very much. The result looks much better than mine. How did you do that? Please let me know so that I can modify the other datasets if I face the same situation again.
Cheers Angel
2009/10/10 Bruce Fischl fischl@nmr.mgh.harvard.edu
Hi Angel,
I'm just about done running the dataset and the results actually look ok (see attached tif). I didn't intervene at all, although you would want to to get rid of some non-brain tissue that the pial surface is grabbing, but that's pretty easy. Did you minimize your TE? I don't know what your gradient set is, but usually you can get your TE down a bit more than that. You also might be better off sacrificing a bit of resolution and lengthening your TR cheers, Bruce
On Thu, 8 Oct 2009, Angel Wong wrote:
Thanks a lot with your help. This is because we want to perform quite a
number of functional runs on the same subject at the same session. So we tried to shorten the T1 scan time to minimize the motion artifacts come from the subject.
Attached please find the orig.mgz for your information.
2009/10/8 Bruce Fischl fischl@nmr.mgh.harvard.edu
I see. Your TR is quite short and your TE relatively long, so this is
probably pretty poor CNR between gray and white. If you send me the orig.mgz I'll take a look and see if it's fixable
On Thu, 8 Oct 2009, Angel Wong wrote:
Hi Bruce
Attached please find the parameters for acquiring the T1 image.
Thank you for your help.
Angel
2009/10/8 Bruce Fischl fischl@nmr.mgh.harvard.edu
eek, that looks pretty bad. Can you tell us what the acquisition
parameters are?
On Thu, 8 Oct 2009, Angel Wong wrote:
Dear all
> > > Actually I would like to perform automated reconstruction for my T1 > image > and then obtain anatomical labels after cortical parcellation. I have > done > fully automated reconstruction using recon-all. But for my case, the > white > and grey matter segmentation appears to be weird. The green line seems > laying on the pial surface rather than on the white surface, and the > red > line seems including some non-brain structures (including small > portion > of > CSF and skull). Thus, the resultant cortical thickness appears to be > very > thin after the segmentation. I just doubt if anything was going wrong > during > the procedure. Or need any preprocessing to enhance the segmentation > quality > before performing the reconstruction? If you have had similar > situation > as > mine, please let me know how to tackle the problem. > > Attached please find the images for better illustration. > > fs_image_brain.tiff - use brain.mgz as aux volume > fs_image_norm.tiff - use norm.mgz as aux volume > fs_image_brainmask.tiff - use brainmask.mgz as aux volume and load > segmentation file aseg.mgz > > Thanks > Angel > > > >
Freesurfer mailing list Freesurfer@nmr.mgh.harvard.edu https://mail.nmr.mgh.harvard.edu/mailman/listinfo/freesurfer
My version is 4.3.0. I will download the latest version and try again.
Thank you.
2009/10/10 Bruce Fischl fischl@nmr.mgh.harvard.edu
I didn't do anything but run recon-all with the latest version. Remind me what version you were running?
On Sat, 10 Oct 2009, Angel Wong wrote:
Hi Bruce
Thank you very much. The result looks much better than mine. How did you do that? Please let me know so that I can modify the other datasets if I face the same situation again.
Cheers Angel
2009/10/10 Bruce Fischl fischl@nmr.mgh.harvard.edu
Hi Angel,
I'm just about done running the dataset and the results actually look ok (see attached tif). I didn't intervene at all, although you would want to to get rid of some non-brain tissue that the pial surface is grabbing, but that's pretty easy. Did you minimize your TE? I don't know what your gradient set is, but usually you can get your TE down a bit more than that. You also might be better off sacrificing a bit of resolution and lengthening your TR cheers, Bruce
On Thu, 8 Oct 2009, Angel Wong wrote:
Thanks a lot with your help. This is because we want to perform quite a
number of functional runs on the same subject at the same session. So we tried to shorten the T1 scan time to minimize the motion artifacts come from the subject.
Attached please find the orig.mgz for your information.
2009/10/8 Bruce Fischl fischl@nmr.mgh.harvard.edu
I see. Your TR is quite short and your TE relatively long, so this is
probably pretty poor CNR between gray and white. If you send me the orig.mgz I'll take a look and see if it's fixable
On Thu, 8 Oct 2009, Angel Wong wrote:
Hi Bruce
Attached please find the parameters for acquiring the T1 image.
Thank you for your help.
Angel
2009/10/8 Bruce Fischl fischl@nmr.mgh.harvard.edu
eek, that looks pretty bad. Can you tell us what the acquisition
parameters > are? > > On Thu, 8 Oct 2009, Angel Wong wrote: > > Dear all > > > >> >> Actually I would like to perform automated reconstruction for my T1 >> image >> and then obtain anatomical labels after cortical parcellation. I >> have >> done >> fully automated reconstruction using recon-all. But for my case, the >> white >> and grey matter segmentation appears to be weird. The green line >> seems >> laying on the pial surface rather than on the white surface, and the >> red >> line seems including some non-brain structures (including small >> portion >> of >> CSF and skull). Thus, the resultant cortical thickness appears to be >> very >> thin after the segmentation. I just doubt if anything was going >> wrong >> during >> the procedure. Or need any preprocessing to enhance the segmentation >> quality >> before performing the reconstruction? If you have had similar >> situation >> as >> mine, please let me know how to tackle the problem. >> >> Attached please find the images for better illustration. >> >> fs_image_brain.tiff - use brain.mgz as aux volume >> fs_image_norm.tiff - use norm.mgz as aux volume >> fs_image_brainmask.tiff - use brainmask.mgz as aux volume and load >> segmentation file aseg.mgz >> >> Thanks >> Angel >> >> >> >> >> _______________________________________________
Freesurfer mailing list Freesurfer@nmr.mgh.harvard.edu https://mail.nmr.mgh.harvard.edu/mailman/listinfo/freesurfer
Angel,
I noticed you have problems with some remaining dura in your images. you might want to try a new utility that is being tested for inclusion in a future release. Its from a group in Singapore, and does a good job removing dura. Its called:
mri_gcut
and can be downloaded for the linux and mac platforms from here:
ftp://surfer.nmr.mgh.harvard.edu/pub/dist/freesurfer/misc/
copy it to your $FREESURFER_HOME/bin
usage (with reference to paper) is found in:
mri_gcut --help
and intended usage is:
cd subjid/mri mri_gcut -110 -mult brainmask.mgz T1.mgz brainmask.mgz
then run:
recon-all -s subjid -autorecon2 -autorecon3
but of course inspect the brainmask.mgz to see if actually helped.
(bruce, can you point me at her image? i can run it locally)
Nick
On Sat, 2009-10-10 at 21:46 +0800, Angel Wong wrote:
My version is 4.3.0. I will download the latest version and try again.
Thank you.
2009/10/10 Bruce Fischl fischl@nmr.mgh.harvard.edu I didn't do anything but run recon-all with the latest version. Remind me what version you were running?
On Sat, 10 Oct 2009, Angel Wong wrote: Hi Bruce Thank you very much. The result looks much better than mine. How did you do that? Please let me know so that I can modify the other datasets if I face the same situation again. Cheers Angel 2009/10/10 Bruce Fischl <fischl@nmr.mgh.harvard.edu> Hi Angel, I'm just about done running the dataset and the results actually look ok (see attached tif). I didn't intervene at all, although you would want to to get rid of some non-brain tissue that the pial surface is grabbing, but that's pretty easy. Did you minimize your TE? I don't know what your gradient set is, but usually you can get your TE down a bit more than that. You also might be better off sacrificing a bit of resolution and lengthening your TR cheers, Bruce On Thu, 8 Oct 2009, Angel Wong wrote: Thanks a lot with your help. This is because we want to perform quite a number of functional runs on the same subject at the same session. So we tried to shorten the T1 scan time to minimize the motion artifacts come from the subject. Attached please find the orig.mgz for your information. 2009/10/8 Bruce Fischl <fischl@nmr.mgh.harvard.edu> I see. Your TR is quite short and your TE relatively long, so this is probably pretty poor CNR between gray and white. If you send me the orig.mgz I'll take a look and see if it's fixable On Thu, 8 Oct 2009, Angel Wong wrote: Hi Bruce Attached please find the parameters for acquiring the T1 image. Thank you for your help. Angel 2009/10/8 Bruce Fischl <fischl@nmr.mgh.harvard.edu> eek, that looks pretty bad. Can you tell us what the acquisition parameters are? On Thu, 8 Oct 2009, Angel Wong wrote: Dear all Actually I would like to perform automated reconstruction for my T1 image and then obtain anatomical labels after cortical parcellation. I have done fully automated reconstruction using recon-all. But for my case, the white and grey matter segmentation appears to be weird. The green line seems laying on the pial surface rather than on the white surface, and the red line seems including some non-brain structures (including small portion of CSF and skull). Thus, the resultant cortical thickness appears to be very thin after the segmentation. I just doubt if anything was going wrong during the procedure. Or need any preprocessing to enhance the segmentation quality before performing the reconstruction? If you have had similar situation as mine, please let me know how to tackle the problem. Attached please find the images for better illustration. fs_image_brain.tiff - use brain.mgz as aux volume fs_image_norm.tiff - use norm.mgz as aux volume fs_image_brainmask.tiff - use brainmask.mgz as aux volume and load segmentation file aseg.mgz Thanks Angel _______________________________________________ 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 Nick
Thanks for your suggestion. I will look at it in detail later.
Angel
2009/10/10 Nick Schmansky nicks@nmr.mgh.harvard.edu
Angel,
I noticed you have problems with some remaining dura in your images. you might want to try a new utility that is being tested for inclusion in a future release. Its from a group in Singapore, and does a good job removing dura. Its called:
mri_gcut
and can be downloaded for the linux and mac platforms from here:
ftp://surfer.nmr.mgh.harvard.edu/pub/dist/freesurfer/misc/
copy it to your $FREESURFER_HOME/bin
usage (with reference to paper) is found in:
mri_gcut --help
and intended usage is:
cd subjid/mri mri_gcut -110 -mult brainmask.mgz T1.mgz brainmask.mgz
then run:
recon-all -s subjid -autorecon2 -autorecon3
but of course inspect the brainmask.mgz to see if actually helped.
(bruce, can you point me at her image? i can run it locally)
Nick
On Sat, 2009-10-10 at 21:46 +0800, Angel Wong wrote:
My version is 4.3.0. I will download the latest version and try again.
Thank you.
2009/10/10 Bruce Fischl fischl@nmr.mgh.harvard.edu I didn't do anything but run recon-all with the latest version. Remind me what version you were running?
On Sat, 10 Oct 2009, Angel Wong wrote: Hi Bruce Thank you very much. The result looks much better than mine. How did you do that? Please let me know so that I can modify the other datasets if I face the same situation again. Cheers Angel 2009/10/10 Bruce Fischl <fischl@nmr.mgh.harvard.edu> Hi Angel, I'm just about done running the dataset and the results actually look ok (see attached tif). I didn't intervene at all, although you would want to to get rid of some non-brain tissue that the pial surface is grabbing, but that's pretty easy. Did you minimize your TE? I don't know what your gradient set is, but usually you can get your TE down a bit more than that. You also might be better off sacrificing a bit of resolution and lengthening your TR cheers, Bruce On Thu, 8 Oct 2009, Angel Wong wrote: Thanks a lot with your help. This is because we want to perform quite a number of functional runs on the same subject at the same session. So we tried to shorten the T1 scan time to minimize the motion artifacts come from the subject. Attached please find the orig.mgz for your information. 2009/10/8 Bruce Fischl <fischl@nmr.mgh.harvard.edu> I see. Your TR is quite short and your TE relatively long, so this is probably pretty poor CNR between gray and white. If you send me the orig.mgz I'll take a look and see if it's fixable On Thu, 8 Oct 2009, Angel Wong wrote: Hi Bruce Attached please find the parameters for acquiring the T1 image. Thank you for your help. Angel 2009/10/8 Bruce Fischl <fischl@nmr.mgh.harvard.edu>
eek, that looks pretty bad. Can you tell us what the acquisition parameters are? On Thu, 8 Oct 2009, Angel Wong wrote: Dear all ActuallyI would like to perform automated reconstruction for my T1
image and then obtainanatomical labels after cortical parcellation. I have
done fully automatedreconstruction using recon-all. But for my case, the
white and grey mattersegmentation appears to be weird. The green line seems
laying on the pial surfacerather than on the white surface, and the
red line seems includingsome non-brain structures (including small
portion of CSF and skull).Thus, the resultant cortical thickness appears to be
very thin after thesegmentation. I just doubt if anything was going wrong
during theprocedure. Or need any preprocessing to enhance the segmentation
quality beforeperforming the reconstruction? If you have had similar
situation as mine, please let me know how to tackle the problem. Attachedplease find the images for better illustration.
fs_image_brain.tiff - use brain.mgz as aux volume
fs_image_norm.tiff - use norm.mgz as aux volume
fs_image_brainmask.tiff - use brainmask.mgz as aux volume and load
segmentation file aseg.mgz
Thanks Angel _______________________________________________ Freesurfer mailing list Freesurfer@nmr.mgh.harvard.eduhttps://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 Bruce
I have downloaded the latest version (v4.5.0) and tried again. But unfortunately, I cannot reproduce the results same as yours. The new generated results are similar to the results generated by the older version. The command I used is
recon-all -i <directory that contains COR files> -s <subjectid> -sd <subjects_dir> -all
and I viewed the images using
tkmedit <subjectid> brainmask.mgz lh.white -aux-surface rh.white
I am wondering why there is a big difference between your results and mine. Is there any pre-setting needed before running the recon-all or additional arguments needed during the processing?
Cheers Angel
2009/10/11 Angel Wong angel.delver@gmail.com
Hi Nick
Thanks for your suggestion. I will look at it in detail later.
Angel
2009/10/10 Nick Schmansky nicks@nmr.mgh.harvard.edu
Angel,
I noticed you have problems with some remaining dura in your images. you might want to try a new utility that is being tested for inclusion in a future release. Its from a group in Singapore, and does a good job removing dura. Its called:
mri_gcut
and can be downloaded for the linux and mac platforms from here:
ftp://surfer.nmr.mgh.harvard.edu/pub/dist/freesurfer/misc/
copy it to your $FREESURFER_HOME/bin
usage (with reference to paper) is found in:
mri_gcut --help
and intended usage is:
cd subjid/mri mri_gcut -110 -mult brainmask.mgz T1.mgz brainmask.mgz
then run:
recon-all -s subjid -autorecon2 -autorecon3
but of course inspect the brainmask.mgz to see if actually helped.
(bruce, can you point me at her image? i can run it locally)
Nick
On Sat, 2009-10-10 at 21:46 +0800, Angel Wong wrote:
My version is 4.3.0. I will download the latest version and try again.
Thank you.
2009/10/10 Bruce Fischl fischl@nmr.mgh.harvard.edu I didn't do anything but run recon-all with the latest version. Remind me what version you were running?
On Sat, 10 Oct 2009, Angel Wong wrote: Hi Bruce Thank you very much. The result looks much better than mine. How did you do that? Please let me know so that I can modify the other datasets if I face the same situation again. Cheers Angel 2009/10/10 Bruce Fischl <fischl@nmr.mgh.harvard.edu> Hi Angel, I'm just about done running the dataset and the results actually look ok (see attached tif). I didn't intervene at all, although you would want to to get rid of some non-brain tissue that the pial surface is grabbing, but that's pretty easy. Did you minimize your TE? I don't know what your gradient set is, but usually you can get your TE down a bit more than that. You also might be better off sacrificing a bit of resolution and lengthening your TR cheers, Bruce On Thu, 8 Oct 2009, Angel Wong wrote: Thanks a lot with your help. This is because we want to perform quite a number of functional runs on the same subject at the same session. So we tried to shorten the T1 scan time to minimize the motion artifacts come from the subject. Attached please find the orig.mgz for your information. 2009/10/8 Bruce Fischl <fischl@nmr.mgh.harvard.edu> I see. Your TR is quite short and your TE relatively long, so this is probably pretty poor CNR between gray and white. If you send me the orig.mgz I'll take a look and see if it's fixable On Thu, 8 Oct 2009, Angel Wong wrote: Hi Bruce Attached please find the parameters for acquiring the T1 image. Thank you for your help. Angel 2009/10/8 Bruce Fischl <fischl@nmr.mgh.harvard.edu>
eek, that looks pretty bad. Can you tell us what the acquisition parameters are? On Thu, 8 Oct 2009, Angel Wong wrote: Dear all ActuallyI would like to perform automated reconstruction for my T1
image and then obtainanatomical labels after cortical parcellation. I have
done fullyautomated reconstruction using recon-all. But for my case, the
white and grey mattersegmentation appears to be weird. The green line seems
laying on the pial surfacerather than on the white surface, and the
red line seemsincluding some non-brain structures (including small
portion of CSF and skull).Thus, the resultant cortical thickness appears to be
very thin after thesegmentation. I just doubt if anything was going wrong
during theprocedure. Or need any preprocessing to enhance the segmentation
quality beforeperforming the reconstruction? If you have had similar
situation
as mine, please let me know how to tackle the problem. Attachedplease find the images for better illustration.
fs_image_brain.tiff - use brain.mgz as aux volume
fs_image_norm.tiff - use norm.mgz as aux volume
fs_image_brainmask.tiff - use brainmask.mgz as aux volume and load
segmentation file aseg.mgz
Thanks Angel _______________________________________________ Freesurfer mailing list Freesurfer@nmr.mgh.harvard.eduhttps://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
hmmm, no idea. Can you send me the recon-all.log file for that subject? On Tue, 13 Oct 2009, Angel Wong wrote:
Hi Bruce
I have downloaded the latest version (v4.5.0) and tried again. But unfortunately, I cannot reproduce the results same as yours. The new generated results are similar to the results generated by the older version. The command I used is
recon-all -i <directory that contains COR files> -s <subjectid> -sd <subjects_dir> -all
and I viewed the images using
tkmedit <subjectid> brainmask.mgz lh.white -aux-surface rh.white
I am wondering why there is a big difference between your results and mine. Is there any pre-setting needed before running the recon-all or additional arguments needed during the processing?
Cheers Angel
2009/10/11 Angel Wong angel.delver@gmail.com
Hi Nick
Thanks for your suggestion. I will look at it in detail later.
Angel
2009/10/10 Nick Schmansky nicks@nmr.mgh.harvard.edu
Angel,
I noticed you have problems with some remaining dura in your images. you might want to try a new utility that is being tested for inclusion in a future release. Its from a group in Singapore, and does a good job removing dura. Its called:
mri_gcut
and can be downloaded for the linux and mac platforms from here:
ftp://surfer.nmr.mgh.harvard.edu/pub/dist/freesurfer/misc/
copy it to your $FREESURFER_HOME/bin
usage (with reference to paper) is found in:
mri_gcut --help
and intended usage is:
cd subjid/mri mri_gcut -110 -mult brainmask.mgz T1.mgz brainmask.mgz
then run:
recon-all -s subjid -autorecon2 -autorecon3
but of course inspect the brainmask.mgz to see if actually helped.
(bruce, can you point me at her image? i can run it locally)
Nick
On Sat, 2009-10-10 at 21:46 +0800, Angel Wong wrote:
My version is 4.3.0. I will download the latest version and try again.
Thank you.
2009/10/10 Bruce Fischl fischl@nmr.mgh.harvard.edu I didn't do anything but run recon-all with the latest version. Remind me what version you were running?
On Sat, 10 Oct 2009, Angel Wong wrote: Hi Bruce Thank you very much. The result looks much better than mine. How did you do that? Please let me know so that I can modify the other datasets if I face the same situation again. Cheers Angel 2009/10/10 Bruce Fischl <fischl@nmr.mgh.harvard.edu> Hi Angel, I'm just about done running the dataset and the results actually look ok (see attached tif). I didn't intervene at all, although you would want to to get rid of some non-brain tissue that the pial surface is grabbing, but that's pretty easy. Did you minimize your TE? I don't know what your gradient set is, but usually you can get your TE down a bit more than that. You also might be better off sacrificing a bit of resolution and lengthening your TR cheers, Bruce On Thu, 8 Oct 2009, Angel Wong wrote: Thanks a lot with your help. This is because we want to perform quite a number of functional runs on the same subject at the same session. So we tried to shorten the T1 scan time to minimize the motion artifacts come from the subject. Attached please find the orig.mgz for your information. 2009/10/8 Bruce Fischl <fischl@nmr.mgh.harvard.edu> I see. Your TR is quite short and your TE relatively long, so this is probably pretty poor CNR between gray and white. If you send me the orig.mgz I'll take a look and see if it's fixable On Thu, 8 Oct 2009, Angel Wong wrote: Hi Bruce Attached please find the parameters for acquiring the T1 image. Thank you for your help. Angel 2009/10/8 Bruce Fischl <fischl@nmr.mgh.harvard.edu>
eek, that looks pretty bad. Can you tell us what the acquisition parameters are? On Thu, 8 Oct 2009, Angel Wong wrote: Dear all ActuallyI would like to perform automated reconstruction for my T1
image and then obtainanatomical labels after cortical parcellation. I have
done fullyautomated reconstruction using recon-all. But for my case, the
white and grey mattersegmentation appears to be weird. The green line seems
laying on the pial surfacerather than on the white surface, and the
red line seemsincluding some non-brain structures (including small
portion of CSF and skull).Thus, the resultant cortical thickness appears to be
very thin after thesegmentation. I just doubt if anything was going wrong
during theprocedure. Or need any preprocessing to enhance the segmentation
quality beforeperforming the reconstruction? If you have had similar
situation
as mine, please let me know how to tackle the problem. Attachedplease find the images for better illustration.
fs_image_brain.tiff - use brain.mgz as aux volume
fs_image_norm.tiff - use norm.mgz as aux volume
fs_image_brainmask.tiff - use brainmask.mgz as aux volume and load
segmentation file aseg.mgz
Thanks Angel _______________________________________________ Freesurfer mailing list Freesurfer@nmr.mgh.harvard.eduhttps://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 Bruce
Actually, the program is still running (running the last few stages), so the log file is not complete. But I can send you the log file which was generated by the older version (v4.3.0). I think these two log files might be similar as the results generated by these two version look similar too.
Once my recon-all program finishes, I may send you the log file generated by the latest version.
Thanks again for your help.
Angel
2009/10/13 Bruce Fischl fischl@nmr.mgh.harvard.edu
hmmm, no idea. Can you send me the recon-all.log file for that subject?
On Tue, 13 Oct 2009, Angel Wong wrote:
Hi Bruce
I have downloaded the latest version (v4.5.0) and tried again. But unfortunately, I cannot reproduce the results same as yours. The new generated results are similar to the results generated by the older version. The command I used is
recon-all -i <directory that contains COR files> -s <subjectid> -sd <subjects_dir> -all
and I viewed the images using
tkmedit <subjectid> brainmask.mgz lh.white -aux-surface rh.white
I am wondering why there is a big difference between your results and mine. Is there any pre-setting needed before running the recon-all or additional arguments needed during the processing?
Cheers Angel
2009/10/11 Angel Wong angel.delver@gmail.com
Hi Nick
Thanks for your suggestion. I will look at it in detail later.
Angel
2009/10/10 Nick Schmansky nicks@nmr.mgh.harvard.edu
Angel,
I noticed you have problems with some remaining dura in your images. you might want to try a new utility that is being tested for inclusion in a future release. Its from a group in Singapore, and does a good job removing dura. Its called:
mri_gcut
and can be downloaded for the linux and mac platforms from here:
ftp://surfer.nmr.mgh.harvard.edu/pub/dist/freesurfer/misc/
copy it to your $FREESURFER_HOME/bin
usage (with reference to paper) is found in:
mri_gcut --help
and intended usage is:
cd subjid/mri mri_gcut -110 -mult brainmask.mgz T1.mgz brainmask.mgz
then run:
recon-all -s subjid -autorecon2 -autorecon3
but of course inspect the brainmask.mgz to see if actually helped.
(bruce, can you point me at her image? i can run it locally)
Nick
On Sat, 2009-10-10 at 21:46 +0800, Angel Wong wrote:
My version is 4.3.0. I will download the latest version and try again.
Thank you.
2009/10/10 Bruce Fischl fischl@nmr.mgh.harvard.edu I didn't do anything but run recon-all with the latest version. Remind me what version you were running?
On Sat, 10 Oct 2009, Angel Wong wrote: Hi Bruce Thank you very much. The result looks much better than mine. How did you do that? Please let me know so that I can modify the other datasets if I face the same situation again. Cheers Angel 2009/10/10 Bruce Fischl <fischl@nmr.mgh.harvard.edu> Hi Angel, I'm just about done running the dataset and the results actually look ok (see attached tif). I didn't intervene at all, although you would want to to get rid of some non-brain tissue that the pial surface is grabbing, but that's pretty easy. Did you minimize your TE? I don't know what your gradient set is, but usually you can get your TE down a bit more than that. You also might be better off sacrificing a bit of resolution and lengthening your TR cheers, Bruce On Thu, 8 Oct 2009, Angel Wong wrote: Thanks a lot with your help. This is because we want to perform quite a number of functional runs on the same subject at the same session. So we tried to shorten the T1 scan time to minimize the motion artifacts come from the subject. Attached please find the orig.mgz for your information. 2009/10/8 Bruce Fischl <fischl@nmr.mgh.harvard.edu> I see. Your TR is quite short and your TE relatively long, so this is probably pretty poor CNR between gray and white. If you send me the orig.mgz I'll take a look and see if it's fixable On Thu, 8 Oct 2009, Angel Wong wrote: Hi Bruce Attached please find the parameters for acquiring the T1 image. Thank you for your help. Angel 2009/10/8 Bruce Fischl <fischl@nmr.mgh.harvard.edu>
eek, that looks pretty bad. Can you tell us what the acquisition parameters are? On Thu, 8 Oct 2009, Angel Wong wrote: Dear all ActuallyI would like to perform automated reconstruction for my T1
image and then obtainanatomical labels after cortical parcellation. I have
done fullyautomated reconstruction using recon-all. But for my case, the
white and grey mattersegmentation appears to be weird. The green line seems
laying on the pial surfacerather than on the white surface, and the
red line seemsincluding some non-brain structures (including small
portion of CSF and skull).Thus, the resultant cortical thickness appears to be
very thin after thesegmentation. I just doubt if anything was going wrong
during theprocedure. Or need any preprocessing to enhance the segmentation
quality beforeperforming the reconstruction? If you have had similar
situation
as mine, please let me know how to tackle the problem. Attachedplease find the images for better illustration.
fs_image_brain.tiff - use brain.mgz as aux volume
fs_image_norm.tiff - use norm.mgz as aux volume
fs_image_brainmask.tiff - use brainmask.mgz as aux volume and load
segmentation file aseg.mgz
Thanks Angel _______________________________________________ Freesurfer mailing list Freesurfer@nmr.mgh.harvard.eduhttps://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
-
Angel Wong -
Bruce Fischl -
Nick Schmansky