Hi Ryan
I'm not sure we have much useful advice to give you as it depends a lot
on your subject population. Basically we make the decision of whether or
not a given dataset is good enough quality to extract biologically
meaningful results from. If you are only interested in a bit of the brain
then maybe you don't care about artifacts in other regions. Alternatively
if you have tons of subject maybe you don't mind getting rid of low quality
data. You do have to be careful that you aren't biasing your sampling
though.
cheers
Bruce
On Tue, 5 Mar 2019, Ryan Wales wrote:
External Email - Use Caution
Hi,
Although there are many ways to correct for pial, WM, and segmentation defects,
it seems common in
the literature to forego any edits and simply either accept or reject a
subject's FreeSurfer output
in order to avoid subjectivity and potential over-editing. Another group
rejected subjects if their
defects spanned 6 or more slices, for example.
Are there any guide lines on how much editing is too much editing? Should it
just be edited enough
so it looks tolerable by eye? Maybe by a few different raters?
The defects vary in their severity, so it's difficult to know if a subject
should be considered
passable or requiring edits based on a minor defect.
My current plan is to conduct my analysis once with excluding subjects with
poor segmentation and
then again, including those subjects' edited outputs. Do you agree with this
logic?
Thanks for your advice,
Ryan
--
Ryan Wales
Graduate Student
Cognition and Motor Control Neuroscience Laboratory
Integrative Neuroscience
Psychology Department
Stony Brook University
E-mail: ryan.wa...@stonybrook.edu
_______________________________________________
Freesurfer mailing list
Freesurfer@nmr.mgh.harvard.edu
https://mail.nmr.mgh.harvard.edu/mailman/listinfo/freesurfer
