I can't figure out exactly where it's all going wrong.
When you begin, image A3 (T1 weighted scan with same slice prescription as EPI)
should be in exact register with images A2 (functional EPI scans). You can
check this using the <Check Reg> button. If they are not in register, then make
sure they are. You may need to delete any .mat files associated with the images
(if you do this, it means that the realignment parameters for A2 have been
lost, so you may need to realign them again). Once you have all the images (A3
and A2) in the same orientation, then you can start to register them with A1.
If you change the origins or orientations in A3 using <Display>, ensure that
you also select images A2. This applies the same spatial transformations to
all images (indirectly via changing their .mat files). If you use <Check Reg>,
then A3 and A2 should all still be in register.
When you coregister, possibly using mutual information (which works best with
the updated MI coregistration software, which will need recompiling if you are
not using Linux and Matlab 6, or Solaris), it is easiest if the source image
is A1, and the target is A3. This will change the .mat file for A1. The
alternative would be to have source:A3, target:A1 and others:A2. This has
more potential for problems though, as if it goes wrong, then the orientation
of A3 and A2 will be changed, rather than just the orientation of A1.
After this step, all the images should be in register (check using <Check Reg>).
You now need to spatially normalise. To do this, you would estimate a spatial
transformation by matching A1 to the T1 template. Because all the images should
be in register, applying the same warp to them all should bring them all into
alignment with the template.
I really hope this helps.
Best regards,
-John
> Essentially, it's:
> A. First collect your data:
> 1. Collect a whole brain T1 weighted scan.
> 2. Collect your functional EPI scans.
> 3. Collect a T1 weighted scan with the same slice prescription as your
> functional EPIs. I'll call this one coplanar. Use a 2D sequence like FSE.
>
> B. Now co-register it like this:
> 1. Motion correct your EPI's (ie realign in SPM speak). Create a mean EPI.
> 2. Co-register your coplanar to the whole brain T1 and carry the EPI's
> along. This gets the EPIs very close to where they should be. Use mutual
> information registration if you've chopped off a lot of data. 3. Co-
> register your mean EPI image to the whole brain using mutual information.
> Because of step 2, your starting conditions are very good and MI
> coregistration will get you the rest of the way.
> *******************************
>
> I'm having trouble at step B2. I've been trying to get it to work with
> some frontal lobe (anterior to ac-oblique along the ac-pc line) functional
> images and I'm having the following problems. I was hoping you could
> advise. Here's the problem.
>
> Corregistration of in-plane and functional images to SPGR fails (images are
> not aligned). I can go into display and set the origin for in-plane and
> functionals it corregisters fine but then when I normalize, the head is
> tipped upThere is a little script I found called "reorient.m" which
> reslices after I reorient in DISPLAY. Then normalization on these
> corregistered images works fine. If I run this "reorient.m" after I've
> motion corrected (without reslicing) will I be overwriting the motion
> correction .mat files (realign in spm-speak)? If I corregister before
> motion correction, will I just be adding a constant and then I could just
> motion correct after that or would that screw up motion correction
> (realignment)?
--
Dr John Ashburner.
Functional Imaging Lab., 12 Queen Square, London WC1N 3BG, UK.
tel: +44 (0)20 78337491 or +44 (0)20 78373611 x4381
fax: +44 (0)20 78131420 http://www.fil.ion.ucl.ac.uk/~john
|