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.wales@stonybrook.edu