Hi,
in our population of MCI patients, unfortunately parts of the cerebellum were not completely scanned. I ran the data throught the whole FS-4.5.0 pipeline and I see that the eTIV normalized volumes of the segmented structues have very large variance compared to the eTIV normalized group of independent age matched subjects with the full brain coveage. I assume that this is possibly due to the suboptimal registration to the template, which is used to compute eTIV. In fact, all the subjects with incomplete cerebellum failed to pass the quality check of the registration to Talairach, but on the first glance the registration seemed fine to me so I just used -notal-check flag to complete recon-all -all.
I know I could go back, check more carefully and correct the Talairach transform manually, but I hesitate to go this way. Rather I thougt if it would be possible/feasible to remove the cerebellum from each subject AND the template and possibly achieve better registration from these images.
Thanks in advance for opinions,
Martin