note: see Buckner40 page for notes on building fsaverage
We have a new average subject that we are recommending that people use instead of "average7" on subjects reconstructed with the new stream. The new recon uses a different curvature pattern as the surface registration target, and this is not compatible with previous reconstructions. For prior reconstructions, continue to use average7. The new average subject can also replace the "talairach" subject as well. The name of the "fsaverage", and it can be found in the distribution itself:
- $FREESURFER_HOME/subjects/fsaverage
To set up a link in your SUBJECTS_DIR, do the following:
cd $SUBJECTS_DIR ln -s $FREESURFER_HOME/subjects/fsaverage fsaverage
There is a README file in $FREESURFER_HOME/subjects/fsaverage that describes how this subject was created.
IMPORTANT:
Differences between average7 and fsaverage, and FIX_VERTEX_AREA with impacts on mri_surfcluster and mri_glmfit.
As some people have noted, the total surface area of average7 was much less than that of a typical subject (by about 40%). This caused the areas of clusters as reported by mri_surfcluster to be too small. HOWEVER, a bug in the routine that computed the vertex area caused the clusters to be 50% too large. So, the net effect was that the clusters were about the right size :).
Stay with me.
The surface area of the new average subject (fsaverage) is that of a typical subject, SO, it is IMPORTANT that it be used in conjunction with the fixed vertex area.
This affects only two programs: mri_surfcluster and mri_glmfit. After today (March 6, 2006), if you use average7 with mri_surfcluster or mri_glmfit, they will give the following error:
- ERROR: you have selected subject average7. It is recommended that you use the fsaverage subject in $FREESURFER_HOME/subjects. If you really want to use average7, re-run this program with --really-use-average7 as the first argument.
If you chose to do so, then it will print out:
- INFO: you have selected subject average7 (and REALLY want to use it) instead of fsaverage. So I'm going to turn off fixing of vertex area to maintain compatibility with the pre-stable3 release.
The results should be the same as when you ran it prior to the stable3 release.
If you have used average7 along with Monte Carlo or Permutation simulation, inferences based on those simulations should still be correct. The only problem will be that the cluster area might be a little off.
If you have used a non-average subject as your common surface space or an average subject that you created prior to March 1, 2006, then the cluster sizes will appear to drop by a third the next time your run mri_glmfit or mri_surfcluster. Again, any mri_glmfit simulation results will still be valid. For backwards compatibility, you can add --no-fix-vertex area to either mri_surfcluster or mri_glmfit.
The new versions of mri_surfcluster or mri_glmfit will include the following line in their summary/logfile output:
The value will either be 1 or 0 depending upon whether it was fixed or not fixed.