Mike,
The 2008-03-26 GCA atlas was built with the code which is more robust in the presence of background noise, and also code to catch some missing priors. The results of Dice coefficient testing, against manually segmented subjects, indicated that this atlas is identical to the 2007-08-08 atlas (actually, just a hair better, but not significantly). So it is ok to mix data processed under 4.0.5 and 4.0.2 (although in general, we recommend processing a subject set with the same version of freesurfer).
Nick
On Thu, 2008-07-10 at 09:15 -0500, Michael Harms wrote:
Hello, Comparing the recon-all script for FS v4.0.2 against FS v4.0.5, I noticed that somewhere between those versions the "default" GCA files changed from ones dated 2007-08-08 to 2008-03-26.
Is this related to the "new feature" of FS v4.0.3 described on the Release Notes as "talairach_with_skull.lta generation is now robust in presence of increased background noise levels." ?
What are the implications (if any) of this change in the GCA for the intermingling of data processed under early version 4 releases (4.0.2 or earlier) with later version 4 releases (4.0.3 or later)? Is the difference in the GCA minor enough that it would be reasonable to intermix data processed under 4.0.5 and 4.0.2, without needing to update the v4.0.2 data? e.g., Does the difference have any relevant impact on the subcortical segmentation?
thanks, Mike H.