Hi everyone,
setting up the AC-PC and reorienting images is a recurrent issue... so
we came up with some automatic solution.
Check this:
https://dl.dropboxusercontent.com/u/15856584/SPM_autoreorient/spm_auto_reorient.m
and its help to use it.
This can be included in the batching system (of SPM8) by overwriting &
copying these 2 functions
https://dl.dropboxusercontent.com/u/15856584/SPM_autoreorient/spm_cfg.m
https://dl.dropboxusercontent.com/u/15856584/SPM_autoreorient/spm_cfg_autoreorient.m
in your SPM/config subdirectory.
Obviously NO guarantee whatsoever that this will work 100% of the times!
And I have NOT checked when it breaks down in term of rotation angle or
displacement. All I can claim is that it does a good job with our data
(young and old healthy subjects + AD/PD patients)...
HTH,
Chris
PS: go the Red Devils ! :-)
===================================================
Christophe Phillips, Ir, PhD
FRS-FNRS. Research Associate
Adjunct assistant professor in applied sciences
Cyclotron Research Centre, B30
University of Liege, Sart Tilman
4000 Liege, Belgium
Tel: +32 4 366 2316 (secr.)
+32 4 366 2366
Fax: +32 4 366 2946
email: [log in to unmask]
web: http://www.cyclotron.ulg.ac.be
===================================================
Le 3/07/2014 19:33, Helmut Nebl a écrit :
> To add to the confusion :o)
>
> 1) The "default" approch (as described in the SPM manual) is to coregister the structure = source image onto the EPI mean = reference image. In that case "other images" is left empty. As only the structure is reoriented this is fast.
> 2) Alternatively, you can coregister the EPI mean = source onto the structure, but then you have to select the EPI series as "other images". In that case all the EPI volumes including the mean have to be reoriented, which is slower than 1).
>
> If you want to minimize the risk of problems during automatic preprocessing you would try to manually reorient the whole data set into MNI space before the automatic segmentation/normalisation step. As the resolution of the EPIs is poor you typically load the structure, set origin near AC, reorient according to AC-PC, and apply these parameters to the whole series. If you now start with the preprocessing as described in 1), the structure is jiggled onto the mean and thus no longer corresponds to the exact, manually adjusted orientation. In general (motion within a limited range of just a few mm) the coregistered structure is still quite close (close enough and especially, closer than without the manual adjustment) to the template, so this displacement should not matter. From a theoretical perspective it is nicer of course if the manually defined origin remains fixed. Thus you might prefer 2) Another solution is to start with the automatic preprocessing including realignment and coregistration and then manually reorient the whole data set into MNI space based on the stucture. This combines the advantages of 2) (origin and orientation is set to exact the same region in any subject) and 1) (a little faster, as you only reorient the structure during coregistration).
>
> In practice, it should be irrelevant which way to go as long as motion within the EPI run and motion between EPI run and structure is small. If motion between structure and EPI run is rather large you might want to manually reorient the EPI series and the structure separately anyway ;-)
>
|