Greetings,<br><br>I am trying to determine how bad a talairach.xfm registration has to be to warrant a manual edit.  I have found that even minor edits can have noticable (9%) effects on major subcortical volumes (such as thalamus or hippocampus).  Very few registrations seem to have completely failed, but most seem like they could be made better by a tweak here or there.  At what point would I just be adding randomness by manually editing?<br>
<br>About what percent of a given dataset would you expect to manually edit?<br><br>Much appreciated,<br>-Eric Cunningham<br>