We are running FS v.6.0. on a supercomputer operating on Linux. We have two sets of data - one from 1.5T Aera Siemens, second from 3.0T Achieva TX Philips. The T1 images on both scanners have comparable CNR, with an isotropic resolution of 1mm3 each. The recon-all finished with no errors, but in approximately half of the Philips data there is a significant (up to 40%) difference between the brain mask volume and eTIV (no such problem on Siemens). We inspected the Talairach registration (as is pointed that eTIV is calculated upon it) in the mismatched data with:

tkregister2 --mgz --s <subject> --fstal --surf orig

and in the corrupted data the brain mask and the WM-GM borders align in terms of placement but are of a wrong scale (proportionate to the "brainmask-to-eTIV" difference).

What may be the reason and how to cope with that? We cannot normalize the volume data to eTIV (as it is not proper), what value should we use instead?

Attached you will find example tkregister2 output screenshots.

Best regards,

Agnieszka Sabisz