Hi Mel,
when manually checking the talairach transform, make sure you are checking the one that is actually used for the eTIV computation (that changed over the years and versions).
http://www.freesurfer.net/fswiki/eTIV
so I think currently it is the talairach.xfm
you can use lta_diff (maybe you need lta_convert before) with dist 5 or 7 to find how much scaling is contained in each transform. The difference in scaling should account for the eTIV differences. 30% is pretty severe and you should be able to see that much scaling differences in the transforms.
Best, Martin
On 04/04/2017 10:42 AM, Melanie Ganz wrote:
Dear list,
we are currently working with some longitudinal data where we would like to measure volume changes over time and compare the cross-sectional and longitudinal stream on.
When we run the cross-sectional stream and just look at changes of eTIV within subjects over a period of 6 months to a year we get 3-4 subjects out of 110 that have eTIV changes of 20-30%. I've double checked the tailarach transforms using the talairach_afd tools and also manually checked the transforms with tkregister. Everything seems to be ok.
I know when using the longitudinal stream we can use the single eTIV value for volume correction, but for comparison in case of the cross-sectional stream we want to use the single time point values of eTIV.
Any ideas what else I can check or is this simply noise transferred from the linear estimation of eTIV?
Cheers,
Mel