External Email - Use Caution
Alright thanks, mri_coreg looks a little better than bbregister. But similar to bbregister, using 6 dof results in a brain that is a little too small and using 12 dof has both scaling and translation issues.
What would be the way to use CVS? I tried using the m3z from mri_cvs_register as described in my first message, but got those weird artifacts (http://web.mit.edu/dsbeeler/www/images/m3z-mni.png).
DB
------------------------------------------------------------------------
Douglas N. Greve wrote:
Use mri_coreg instead of bbregister. Make sure to use 12 DOF. You could use CVS as well since that is non-linear
On 05/07/2018 04:11 PM, David Beeler wrote:
Hi Doug, Ok so if I am understanding correctly, SPM or FSL will be better at registering functional data to CVS or MNI space for group analyses than freesurfer will be because they have the capability of doing nonlinear registrations. If that's the case I may give SPM a shot and compare the results. But if I were to use freesurfer to do this registration, is there something better than the following commands?: bbregister --s MNI152_FS --mov funcVol --reg func2mni.lta --bold
mri_vol2vol --mov funcVol --reg func2mni.lta --o funcVolMNI --fstarg --no-resample --nearest Adding --12 to bbregister and using the same mri_vol2vol as above actually makes the transformation much worse, is there something I'm missing? I will use SPM if I have to, but I feel like there is probably something smarter I could be doing with freesurfer. And for clarification, is the --12 flag in bbregister making the registration affine (ie accounting for translation, rotation, scaling, and sheers)? Thanks again for your help, much appreciated! DB
Douglas N. Greve wrote: If that is really what you want to do, you should use a tool appropriate for this type of analysis (eg, SPM and FSL have non-linear intersubject registration)
On 05/04/2018 01:13 PM, David Beeler wrote:
Hi Doug,
funcVol is an individual functional resting state scan. We are doing some clustering analyses and trying to visually compare the clustering across subjects (I would be a bit sketched out trying to quantify the overlap between subjects due to individual differences in anatomy, etc, but just looking at it might give us a clue on what is going on). We do all our analyses in the individual subject's native space, but it's hard to see what the similarities / differences are.
If not bbregister, what would be the correct way to align all subjects to a common space?
Thanks, DB
Douglas N. Greve wrote: What is funcVol? An individual functional? If so, then use --12 (12 dof to account for scaling). In general, we don't recommend doing cross subject registration with BBR as it is really not appropriate for that kind of thing.
On 05/03/2018 01:30 PM, David Beeler wrote:
Hi,
I'm trying to register and transform some raw functional data (108x108x72x175, 2mm iso voxels) to MNI space, while keeping the low res dimensions.
I have tried: bbregister --s MNI152_FS --mov funcVol --reg func2mni.lta --init-fsl --bold mri_vol2vol --mov funcVol --reg func2mni.lta --o funcVolMNI --fstarg --no-resample --nearest mri_vol2vol --mov funcVol --reg func2mni.lta --o funcVolMNI --fstarg --no-resample
And this looks pretty good, but not perfect... the functional data is oriented correctly but the brain seems to be a centimeter or two smaller in multiple dimensions, see:
http://web.mit.edu/dsbeeler/www/images/bbregister-mni.png
I'm not too surprised since the MNI brain is blurry and oddly round, but presumably there is a way to do this transformation more accurately. I've run mri_cvs_register for this subject using the --mni flag:
mri_cvs_register --openmp 8 --mni --mov $subjID --outdir $mniDir
And for other applications I use this m3z file to transform volumes between MNI orig.mgz and individual subject orig.mgz, but is there a way to use it to transform functional data to MNI space while keeping the functional dimensions?
I could do:
mri_vol2vol --mov funcVol --targ $SUBJECTS_DIR/$subjID/mri/orig.mgz --reg register.dof6.lta --nearest --o template-in-anat.nii.gz
mri_vol2vol --noDefM3zPath --targ $SUBJECTS_DIR/$subjID/mri/orig.mgz --mov template-in-anat.nii.gz --m3z $mniDir/final_CVSmorph_tocvs_avg35_inMNI152.m3z --o template-in-MNI-m3z.nii.gz --nearest
But not only is this unideal because it requires upsampling the functional data to 256x256x256, it also gets these funky wavy distortions:
http://web.mit.edu/dsbeeler/www/images/m3z-mni.png
Any thoughts are appreciated, thanks!
-DB
Freesurfer mailing list Freesurfer@nmr.mgh.harvard.edu https://mail.nmr.mgh.harvard.edu/mailman/listinfo/freesurfer
Those CVS images look about right to me for CVS, so I don't think you've done anything wrong. You could do the reg in two parts: (1) run mni152reg to register the anatomical to the mni152, and (2) run bbregister (6dof) to register the fMRI to the anatomical. Then concatenate them into one registration and apply that to the fMRI to bring it into the 152. This would all still be affine.
On 05/08/2018 11:45 AM, David Beeler wrote:
Alright thanks, mri_coreg looks a little better than bbregister. But similar to bbregister, using 6 dof results in a brain that is a little too small and using 12 dof has both scaling and translation issues. What would be the way to use CVS? I tried using the m3z from mri_cvs_register as described in my first message, but got those weird artifacts (http://web.mit.edu/dsbeeler/www/images/m3z-mni.png). DB
Douglas N. Greve wrote: Use mri_coreg instead of bbregister. Make sure to use 12 DOF. You could use CVS as well since that is non-linear
On 05/07/2018 04:11 PM, David Beeler wrote:
Hi Doug, Ok so if I am understanding correctly, SPM or FSL will be better at registering functional data to CVS or MNI space for group analyses than freesurfer will be because they have the capability of doing nonlinear registrations. If that's the case I may give SPM a shot and compare the results. But if I were to use freesurfer to do this registration, is there something better than the following commands?: bbregister --s MNI152_FS --mov funcVol --reg func2mni.lta --bold
mri_vol2vol --mov funcVol --reg func2mni.lta --o funcVolMNI --fstarg --no-resample --nearest Adding --12 to bbregister and using the same mri_vol2vol as above actually makes the transformation much worse, is there something I'm missing? I will use SPM if I have to, but I feel like there is probably something smarter I could be doing with freesurfer. And for clarification, is the --12 flag in bbregister making the registration affine (ie accounting for translation, rotation, scaling, and sheers)? Thanks again for your help, much appreciated! DB
Douglas N. Greve wrote: If that is really what you want to do, you should use a tool appropriate for this type of analysis (eg, SPM and FSL have non-linear intersubject registration)
On 05/04/2018 01:13 PM, David Beeler wrote:
Hi Doug,
funcVol is an individual functional resting state scan. We are doing some clustering analyses and trying to visually compare the clustering across subjects (I would be a bit sketched out trying to quantify the overlap between subjects due to individual differences in anatomy, etc, but just looking at it might give us a clue on what is going on). We do all our analyses in the individual subject's native space, but it's hard to see what the similarities / differences are.
If not bbregister, what would be the correct way to align all subjects to a common space?
Thanks, DB
Douglas N. Greve wrote: What is funcVol? An individual functional? If so, then use --12 (12 dof to account for scaling). In general, we don't recommend doing cross subject registration with BBR as it is really not appropriate for that kind of thing.
On 05/03/2018 01:30 PM, David Beeler wrote:
Hi,
I'm trying to register and transform some raw functional data (108x108x72x175, 2mm iso voxels) to MNI space, while keeping the low res dimensions.
I have tried: bbregister --s MNI152_FS --mov funcVol --reg func2mni.lta --init-fsl --bold mri_vol2vol --mov funcVol --reg func2mni.lta --o funcVolMNI --fstarg --no-resample --nearest mri_vol2vol --mov funcVol --reg func2mni.lta --o funcVolMNI --fstarg --no-resample
And this looks pretty good, but not perfect... the functional data is oriented correctly but the brain seems to be a centimeter or two smaller in multiple dimensions, see:
http://web.mit.edu/dsbeeler/www/images/bbregister-mni.png
I'm not too surprised since the MNI brain is blurry and oddly round, but presumably there is a way to do this transformation more accurately. I've run mri_cvs_register for this subject using the --mni flag:
mri_cvs_register --openmp 8 --mni --mov $subjID --outdir $mniDir
And for other applications I use this m3z file to transform volumes between MNI orig.mgz and individual subject orig.mgz, but is there a way to use it to transform functional data to MNI space while keeping the functional dimensions?
I could do:
mri_vol2vol --mov funcVol --targ $SUBJECTS_DIR/$subjID/mri/orig.mgz --reg register.dof6.lta --nearest --o template-in-anat.nii.gz
mri_vol2vol --noDefM3zPath --targ $SUBJECTS_DIR/$subjID/mri/orig.mgz --mov template-in-anat.nii.gz --m3z $mniDir/final_CVSmorph_tocvs_avg35_inMNI152.m3z --o template-in-MNI-m3z.nii.gz --nearest
But not only is this unideal because it requires upsampling the functional data to 256x256x256, it also gets these funky wavy distortions:
http://web.mit.edu/dsbeeler/www/images/m3z-mni.png
Any thoughts are appreciated, thanks!
-DB
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
-
David Beeler -
Douglas N. Greve