Hi Krish, Bruce, et al.
I just wanted to follow up on the messages below.
We have the problem that when we re-run mris_volmask using a high-res volume (after recon-all and then after re-sampling the aseg file), we get anomalous segmentation: it looks like the surfaces are inside out, as you can see by the side by side comparison. The standard resolution ribbon looks good, but the high resolution ribbon looks inverted. We do not understand why and cannot seem to fix it. The new ribbon is properly aligned to the anatomy, but is mostly unlabeled, except right along the tissue boundaries.
I include a link to the T1 file and the whole subject directory in case anyone is able to take a look at, or just suggest a fix. Of course I understand if no one has time for that!
t1: http://white.stanford.edu/~winawer/brain/2011/t1.nii.gz (33 MB) whole subject dir: http://white.stanford.edu/~winawer/brain/ fs_subject.zip (325 MB)
Many thanks.
Jon
PS: The code used to create the images is here:
# give the subject a name subj=homer
# recon all using hires t1 (0.7 mm isotropic voxels) recon-all -i ~/t1.nii.gz -all -subjid $subj -all
# define the working directory d=$SUBJECTS_DIR/$subj/mri
# back up original ribbon and aseg files mkdir $d/1mm; mv $d/*ribbon* $d/aseg.mgz -t $d/1mm
## Create Hi-Res Ribbon##################### # reslice the aseg file to the size and resolution of rawavg mri_convert $d/1mm/aseg.mgz -o $d/aseg.mgz --reslice_like $d/ rawavg.mgz --resample_type nearest
# re-create surface at resolution of rawavg mris_volmask --label_left_white 2 --label_left_ribbon 3 -- label_right_white 41 --label_right_ribbon 42 --save_ribbon -- save_distance $subj #####################################
# back up high-res files mkdir $d/hires; cp $d/*ribbon* $d/aseg.mgz $d/hires/
# Compare the two ribbon files. Original (1x1x1 mm) looks good: tkmedit $subj 1mm/aseg.mgz -segmentation 1mm/ribbon.mgz
# HiRes looks bad: tkmedit $subj hires/aseg.mgz -segmentation hires/ribbon.mgz
On Apr 14, 2011, at 2:11 PM, Bruce Fischl wrote:
Hi Jon,
I'll leave this to Krish, as he's the local expert on mris_volmask, but it definitely could. Do you want to upload an example of a subject that failed for us to try out?
Bruce
On Thu, 14 Apr 2011, Jonathan Winawer wrote:
Hmm. At least when I use tkmedit, the surfaces seem to be correctly registered to both the high resolution volume (rawavg.mgz) and the 1x1x1 volume (T1.mgz). So it does not look like a coordinate problem, at least to my un-expert eyes. But maybe tkmedit does a different transform of the surfaces than mris_volmask?
In any case, the failure of mris_volmask to produce a good ribbon file follows a very characteristic pattern that looks more like a topology confusion to me, as you can see in the attached png. I say this because voxels along the gray-white border are labeled white (with some gaps) and voxels along the pial surface are labeled as gray. Also, some of the regions which are unlabeled holes in the correct (original) ribbon become labeled islands in the new, high-res ribbon. So it looks like the surfaces are in the right place, but they are somehow not filled in correctly. This precise pattern happens with every high-resolution brain I use when I try to re-run mris_volmask. In contrast when I rerun mris_volmask on bert, it works perfectly, though in this case the rawavg.mgz is already registered to T1.mgz and has the same resolution.
Could a coordinate problem cause this?
Sorry for all the questions. I seem to be stuck though. Many thanks again for your quick responses!
Jon
The information in this e-mail is intended only for the person to whom it is addressed. If you believe this e-mail was sent to you in error and the e-mail contains patient information, please contact the Partners Compliance HelpLine at http://www.partners.org/complianceline . If the e-mail was sent to you in error but does not contain patient information, please contact the sender and properly dispose of the e-mail.
