On 2/18/2021 9:52 AM, Lisanne Michelle Jenkins wrote:
External Email - Use Caution
Hi experts,
We want to combine two large datasets (of patients with different
diseases) that have been processed and manually edited using
Freesurfer 5.3 and Freesurfer 6.0.
From the Freesurfer website and previous posts, we have read that we
should re-run the Freesurfer 5.3 data that has been manually edited
(both brain masks and pial surfaces) through Freesurfer 6.0 using
recon-all and using the -continued flag, and that the manual edits
from 5.3 should carry over.
However, we have found that around one-third of subjects have skull
stripping errors after rerunning through 6.0. Has anyone else found
this? Have we missed something?
I have not noticed this. Are you saying that your edits are not carrying
over or that there are new errors in 6.0 that were not in 5.3?
A related question: We have heard it argued that if the data has been
manually edited and quality checked, the version of Freesurfer which
it is run through should not matter, therefore it is ok to combine
data from different versions – what do you think of this argument?
This is not a question of manual editing. Different versions can have
different parameters/algorithms that can create subtle but systematic
differences in thickness. If all of one group came from one version and
all of another group came from a different version, then I think you
could see differences due to version
An additional argument along those lines that we have heard is: If one
were to re-process data with a later version of Freesurfer, using
edits from an older version would introduce biases from that version –
what are your thoughts on this argument?
I doubt it.
Thank you.
Lisanne
_______________________________________________
Freesurfer mailing list
[email protected]
https://mail.nmr.mgh.harvard.edu/mailman/listinfo/freesurfer
_______________________________________________
Freesurfer mailing list
[email protected]
https://mail.nmr.mgh.harvard.edu/mailman/listinfo/freesurfer
