Dear experts,

We were comparing recon-all output of Freesurfer 5.3 and 6.0 in our database, and found that eTIV values can increase by up to 40% going from 5.3 to 6.0. However, closer inspection revealed that it was not due to the version per se, but how we call recon-all. We had computed one version with recon-all -all, and the other with recon-all -autorecon1, thinking that eTIV output from these two were identical. Some testing in two versions with two flags showed that in both versions, using -all flag automatically add

--atlas 3T18yoSchwartzReactN32_as_orig

in the talairach_avi command, while -autorecon1 does not. The talairach_avi.log indicates that the former set the target to 3T18yoSchwartzReactN32_as_orig, while the latter use the 711-2C_as_mni_average_305.

My first question is, is this some kind of bug or is it intentional?

We thought that the difference in the templates caused the change in the estimates, but it looks like that's not the only answer:
We still get very different eTIV values from -autorecon1 option in FS5.3 and 6.0, even though the target images are the same. The results from -all option are more comparable.

Here is the eTIV from one subject we tested:

-FS5, autorecon1: 1318267

-FS6, autorecon1: 1561456

-FS5, -all: 1662005

-FS6, -all:1678336

Here I attached the shortened recon-all log files from two versions of Freesurfer, with with -all or -autorecon1, as well as eTIV output from the four variations we tested in one subject.
Could you help us resolve why we get the varying estimates?

Thank you!