Dear Erik-Jan,
> However, registration-step (1) (EPI -> high res) fails. I am using a Siemens
> Vision scanner, with an MPRAGE sequence, and use this to register:
> flirt -in EPI -ref MPRAGE -omat epi2mprage.mat
> It does produce a matrix without an error message, but the registration is
> wrong. The two images are already in the same axial orientation.
>
> Can anyone give a hint why this would fail? Step (2) does work.
You have definitely understood the principles correctly.
There are any number of reasons why registration can fail.
Large differences in FOV, different amounts of non-brain matter,
bias field corruption ...
Without seeing the images there is very little I can do to diagnose
the problem.
Note that non-brain matter can and should be removed with BET prior to
registration. Do you do this? If not, then see if this fixes the problem.
Otherwise, put the images somewhere that I can download and I'll have a look
at the problem myself.
> Because I wanted to continue, I tried a different solution: SPM comes with an
> EPI template. I simply registered the EPI images to this template (this does
> work fine, using flirt -in EPI -ref EPI_TEMPLATE -omat epi2template.mat).
>
> Can anyone tell me whether this method is just as valid as the other scheme?
This is valid, though probably not as accurate as when the other one works well.
The reason being that both the EPI and the EPI template have low spatial
resolution (are fuzzy) and so cannot be as precise with the alignment.
So, I recommend trying to fix the first method initially.
Meanwhile you can still use the EPI template, although it will a bit inaccurate.
All the best,
Mark
|