Hi,
Are your images in the same initial orientation?
And is the brain extraction working successfully or not?
I assume the answer to the latter is no.
I would try using fslreorient2std on the images first,
if they are not in a similar orientation to begin with.
Then I would crop the images to remove most of
the neck, if there is a substantial amount of neck.
Then I would try bet with the defaults, and if this
does not work well then try -R or different -f values.
Once you have a good brain extraction and the
images are in a similar orientation to begin with
(less than 45 degrees different) then the linear
registration should work fine.
All the best,
Mark
On 14 May 2012, at 19:27, Alessandra Carla' wrote:
> Dear Sir,
> I tried to perform a nonlinear registration on an image downloaded by OASIS.
>
> I followed the instructions in
>
> http://www.fmrib.ox.ac.uk/fslcourse/lectures/practicals/reg/index.html
>
> I compared the "registration quality" between my registration and those
> obtained by using the data in
>
> fsl_course_data/reg
>
> making a joint histogram both between the images fnirted_old_subj and
> MNI152_T1_2mm
>
> and MNI152_T1_1mm and highres_flirted_old_subj.
>
> I observed a good registration when using the image gave in the course,
> while I obtain a scarce quality when using my image.
>
> Another, I have already bad results in the linear registration phase
> (flirted images).
>
> So I tried to modify the bet command needed to skull strip the subject
> changing
>
> bet old_subj old_brain -f 0.35
>
> with
>
> bet old_subj old_brain_032 -A -R -S -B -f 0.32 -g 0
>
> but I don't get much.
>
> Please, help me.
>
> Best regards
>
|