You could correct this problem using FSL's FAST and see if your Caret
segmentation improves. For example:
bet2 original original_no_skull
fast --nopve -O 0 -I 4 -b -o original_no_skull original_no_skull
fslmaths original -div original_no_skull_bias orginal_bias_corrected
To average images, use fslmaths image1 -add image2 -add imageN -div N
average_output
I am unclear on how increasing the scan length or number would reduce signal
inhomogeneity, as this results from the position of the subject in the coil
(signal is strongest in the center of the coil, and reduces as you move away
from the center). If you do more averages or a longer scan, you will still
have this effect, though your SNR may improve.
Peace,
Matt.
-----Original Message-----
From: FSL - FMRIB's Software Library [mailto:[log in to unmask]] On Behalf
Of shih-Wei Wu
Sent: Tuesday, April 07, 2009 11:57 PM
To: [log in to unmask]
Subject: [FSL] utility to average different images
Dear all,
I encountered a segmentation problem of my T1-weighted image using caret.
After asking
around, the initial diagnosis is that there is quite some inhomogeneity in
my image. In
particular, the brightness of my occipital part looks higher than the
frontal.
One recommendation is to do a 15-min T1 (currently it's set to 7min).
Another
recommendation is to do multiple T1-weighted scans, each about 3 1/2 min and
then
average them after registration. My first question is, which might be a
better option?
The second question is, if I want to average, say 3 T1-weighted image to
create an
averaged T1-weighted image, what is the utility in FSL that would allow me
to do so?
Thanks very much
shihwei
|