Hi Mark,
Thanks for your responses. I've put my responses in CAPS.
I haven't heard about that problem of one run failing but
a re-run working. If you have a dataset where you can
replicate this (by taking a fresh copy of the original data
and doing it all again) then please upload it for us to have
a look at.
I COULDN'T REPLICATE THIS PROBLEM AS YOU HAVE SUGGESTED HERE WITH THE ORIGINAL DATA, SO I UPLOADED A SUBJECT WHO DISPLAYED THIS PROBLEM.
UPLOAD NUMBER: 701725. YOU WILL SEE THAT IN THE 313_all_fast_firstseg file THE RIGHT AMYDALA APPEARS APPROXIMATELY IN THE JAW AREA.
As for the other question, you have to look at it. If we had
an automated way of knowing how good something was, then
we'd use that to develop a better method for doing it!
INDEED, JUST THOUGHT I'D CHECK. BACK TO THE DUVERNOY ATLAS...
Thanks again and have a good weekend!
Dana
All the best,
Mark
On 29 Jan 2010, at 14:23, Dana Eldreth wrote:
> I was wondering if anyone else noticed this problem with FIRST on
> the newest
> release 4.1.5.
>
> First, I'd like to note that FSL 4.1.5 does an excellent job at
> boundary
> correction on our older brains. It's extremely impressive.
>
> 1. For some reason, this newest release occasionally does not
> accurately
> parcellate some structures correctly (e.g., hippocampus in the brain
> stem),
> but when I rerun it again it appears to fix the problem and the
> volumes are
> where they should be in the brain. The volume sizes are also
> relatively
> close to the volume sizes that I had in the earlier version of FIRST.
> Interestingly, the volumes that appear to be parcellated in the
> wrong area
> produce the same volume size as when FIRST produces the volumes in the
> correct location. Has anyone else noticed this, and should this be
> something that I should be concerned about?
>
> I also had another question associated with FIRST:
> 1. How can I tell how confident the program is that it has accurately
> identified a specific structure, or can I only tell via visual
> inspection
> using a neuroantaomical atlas?
>
> As always, thank you so much!
>
> Dana
>
|