Hi David, please post to the list ... answers below
On 11/13/2017 01:48 PM, David Beeler wrote:
Hi Doug, really helpful thanks! So if I want all the motion corrected functional data across runs to be aligned with a downsampled version of my anatomical, is it better to run preproc-sess with -per-session or should I do -per-run and then map each functional to anatomical without resampling using: mri_vol2vol --reg register.dof6.lta --mov fmcpr.nii.gz --fstarg \ --o fmcpr.new.vox2ras.nii.gz --no-resample and then run selxavg3-sess on the fmcpr.new.vox2ras.nii.gz in each run folder? I've heard motion correction with -per-run is better but I don't understand why... can you explain?
Oh, if you want to do it that way, then you would not use the --no-resample (that just changes the geometry in the header). per-run registers and resamples to the middle time point, meaning that there is generally less interpolation. You should add --fstarg orig.mgz This will have it sample into the 1mm orig space. This will create a very largefile. If you want to downsample the anat, you can run mri_convert --downsample 2 2 2 orig.mgz orig.ds2.mgz and then --fstarg orig.ds2.mgz
Further, I don't get how selxavg3-sess chooses its input. If I first do preproc-sess with -per-run, do my transformation to downsampled anatomical space (as described above), and then run selxavg3-sess with the -no-preproc flag, what volume will selxavg3-sess use by default? I assume it will search the run folders for fmcpr.nii.gz (because using f.nii.gz would defeat the purpose of doing motion correction in the first place), but is there a flag I could use to tell selxavg3-sess to use my transformed fmcpr.new.vox2ras.nii.gz as input instead? Should I just rename fmcpr.new.vox2ras.nii.gz to fmcpr.nii.gz and then selxavg3-sess will take it automatically?
When you create your analysis, you should put -funcstem fmcpr.new.vox2ras
Thanks again! David PS is this the correct way to respond to threads on this mailing list so that the thread is maintained? On 11/8/17 5:57 PM, David Beeler wrote:
Hi freesurfer people, I'm trying to understand preprocessing a little better, especially registration of functional data to anatomical space. I am currently running everything in the volume, doing motion correction per-run and then pooling all the runs in the session together for the first level analysis.As far as I can tell, running preproc-sess with -per-run takes the rawfunctional data (f.nii.gz), pulls out the middle timepoint and calls it template.nii.gz, bbregisters func to anat and creates a register.dof6.lta file in each of the functional run folders, transforms the aparc+aseg from the subject's mri folder and binarizes it into a brainmask (brain.nii.gz), and runs mc-afni using f.nii.gz as the input vol and template.nii.gz as the template vol to create the motion corrected volume (fmcpr.nii.gz). Is this correct?yes
Then when I run selxavg3-sess it somehow mri_converts an intermediary motion corrected volume from the mc-afni command above (eg <run#>/tmp.mc-afni2.20097/outvol.nii.gz) into an fmc.nii.gz file in each run folder. Are these fmc.nii.gz's supposed to be coregistered between runs? (i.e. if the subject moved a bunch between two of the runs, would those two volumes be aligned at all?). How do these files work?Not with -per-run. per-run was designed with the idea that you would beresampling to some common space (eg, surface or mni305 or cvs), and then running sexlavg3-sess on that data. If you want all the fmc.nii.gz to be in the same space, then use -per-session.
At the end I have an analysis directory with my betas, sigmaps, etc. What functional volume is everything registered to? Is the transform between this functional volume and anatomical space the register.dof6.lta file located in the session directory? Thanks for bearing with me! -David _______________________________________________ Freesurfer mailing list Freesurfer@nmr.mgh.harvard.edu <mailto:Freesurfer@nmr.mgh.harvard.edu> https://mail.nmr.mgh.harvard.edu/mailman/listinfo/freesurfer