Hi Diederick,
the ICV is computed from the linear talairach.xfm transform (including skull). There are all kinds of other talairach transforms produced by recon all, so make sure you are looking at the talairach.xfm (not ...lta or ..._with_skull.lta).
You might want to read: http://surfer.nmr.mgh.harvard.edu/fswiki/eTIV
It is very likely that the clipping on your outlier image (missing cerebellum etc). produce an incorrect talairach.xfm and thus outlier ICV.
Best, Martin
On Tue, 2010-12-21 at 17:32 +0100, Diederick Stoffers wrote:
Hi Martin,
OK, I didn't expect different scanners/fieldstrengths to have that big an effect on ICV. Given that there is a systematic bias, one would expect a very high correlation between ICV at timepoint one and two. I inspected a scatterplot and this was indeed the case, except for a single outlier. The Tailairach registration looks fine for both timepoints in this outlier. However, the scan at one timepoint was badly planned, part of the cerebellum and brainstem are missing. Could this explain the low ICV value in this subject? If not, what other things should I check?
Cheers,
Diederick
On 20 dec 2010, at 15:26, Martin Reuter wrote:
Hi Diederick,
In a longitudinal study you need to ensure identical acquisition and processing, else you'll introduce a systematic bias. Some of my recent analyses indicate that even updating the software on the scanner can bias your results. Hardware changes are worse.
Best Martin
On Dec 20, 2010, at 6:48 AM, Diederick Stoffers d.stoffers@gmail.com wrote:
Hi all,
I have a dataset with AD patients that were scanned twice, once at 1.5T and once at 3T at an interval of a few years. The ICV values are lower for almost all subjects at timepoint two (FS 5.0). Isn't ICV in the later FS versions supposed to be independent of brain volume as it is based on a scaling factor derived from the Tailairach transform of the skull? Many thanks!
Cheers,
Diederick _______________________________________________ Freesurfer mailing list Freesurfer@nmr.mgh.harvard.edu https://mail.nmr.mgh.harvard.edu/mailman/listinfo/freesurfer