Hi Jeff,
the pallidum is quite small and quite low-contrast, so those numbers aren't surprising between versions that are that different. How do they look visually? Does either version look more accurate? It's amazing how just a couple of ambiguous voxels here and there can result in a 6-9% difference, and it can be hard to even decide which looks better. If you are convinced that one version looks more accurate than the other let us know.
cheers Bruce
On Tue, 9 Feb 2010, Jeff Sadino wrote:
Hello Everyone,
I ran 106 subjects through fs 3.0.5 and also through fs 4.3.1 and then compared the volumes of the subcortical structures. Most of them correlated well between versions, but for the pallidum, there was a strong bias both hemispherically and between versions. In fs3, the left pallidum was 1995 and the right was 1756 for a difference of 239. In fs4, the left pallidum was 1882 and the right was 1593 for a difference of 289.
Between fs versions, this gives a 6% difference [ (1995-1882) / ((1995+1882)/2) ] for the left pallidum and a 9.7% difference [ (1756-1593) / ((1756+1593)/2) ] for the right pallidum. Between hemispheres, the difference is 19% [ (289-239) / ((289+239)/2) ].
I was wondering particularly as to the 19% variability of the pallidum measurements between fs versions. Is this much of a difference common between these two freesurfer versions, or is it unique to our study? Also, is 6% and 9.7% difference for the pallidum between hemispheres typical in freesurfer?
Attachment is plotting automated results between fs3 and fs4.
Thank you very much for your feedback! Jeff Sadino