Hi all. I have a question regarding the use of freesurfer + fs-fast preprocessing to apply the Lausanne 2008 parcellation scheme to some structural data, and then extract from those ~1000 ROIs the average time series from a 4D resting state functional run.

I'm posting what I have carried out so far, both to verify that I haven't missed anything important, but also because I have seen evidence within this and other mailing lists that others are interested in doing something similar. This thread may then answer their questions as well.

Structural and functional data is in $SUBJECTS_DIR/sub00?/mri and $FUNC_DIR/sub00?/rest, respectively
For each subject:

• I carried out recon-all -autorecon1, -autorecon2, -autorecon 3, checking data quality in between
  
• inspected ?h.aparc.annot, possibly made minor edits ( https://surfer.nmr.mgh.harvard.edu/fswiki/tksurfer_labeledit)
• used cmtk.Parcellate() with the Lausanne 2008 scheme ( https://groups.google.com/forum/#!topic/cmtk-users/k_o3s8Yy2jQ)
• This created a number of .label files and some .annot files, notably ?h.myaparcP*.annot, which have sub-parcellations of the 1-16, 17-28 and 29-36 regions indicated in the ?h.aparc.annot files (I'm not entirely clear why they couldn't just go into a single .annot file)
• Converted the .annot files to surface space using mri_aparc2aseg --subXX --annot myaparcPXXXXXX which generated myaparcPXXXXX+aseg.mgz (3 files total, as each of three sets of lh and rh annot files describe both hemispheres in a single file)
  
• ran each of the functional runs through:
  preproc-sess -s subXX -fsd rest -surface self lhrh -mni305 -fwhm 0
  

That brings me to my question. I have my functional data:
$FUNC_DIR/subXX/rest/001/fmcpr.siemens.sm0.mni305.2mm.nii.gz (I also have 2 fmcpr.siemens.sm0.self.?h.nii.gz files, but they didn't work either)

I want to pull the average timeseries data for the functional data out of the ROIs defined in myaparcP1_17+aseg.mgz (and similarly for the other two myaparcP*.mgz files)

I tried
mri_segstats \
--seg $SUBJECTS_DIR/subXX/mri/myaparcP1_16+aseg.mgz \
--i $FUNCTIONALS_DIR/subXX/rest/001/fmcpr.siemens.sm0.mni305.2mm.nii.gz \
--annot subXX lh lh.myaparcP1_16.annot \
--sum lh_1_16.sum.txt \
--avgwf lh_1_16.wav.txt

However, this gave me the following output:

sysname  Linux
hostname Sage
machine  x86_64
user     chris
Loading /home/chris/resting/anat/subXX/mri/myaparcP1_16+aseg.mgz
Loading /home/chris/resting/func/subXX/rest/001/fmcpr.siemens.sm0.mni305.2mm.nii.gz
ERROR: dimension mismatch between input volume and seg
  input 76 76 93
  seg   256 256 256

I then tried resampling my resting state data with mri_vol2vol using my func data as the --mov parameter, and my myaparc files as the --targ parameter, and used tkregister2 on the resampled volume. mri_segstats just hung when I ran it on the resampled volume; by the next morning nothing had been written to either text output file.

I can open my ?h surface files in tksurfer, and then load in the appropriate timecourse ($FUNCTIONALS_DIR/subXX/rest/001/fmcpr.siemens.sm0.self.?h.nii.gz) and when I do so, I can click on various vertices and see timecourse plots. And I can paint my surfaces with .annot files that define my ROIs.  So it seems I should be able to get these timecourse data from ROIs in surface space without making a thousand individual ROIs in volumetric space (my guess is I'd use flsmaths on my myaparc files to do this). But I am not clear if mri_segstats can be used in this way, or if indeed it is the appropriate tool.

Any assistance to get me through this last step would be appreciated.

Thanks