Dear Richard, Marko,
perhaps this is useful to you: when you obtained the seg_(inv)_sn.mat
files, you can use the function spm_get_orig_coord5 to transform a set
of x,y,z points back and forth between native space and MNI. That's how
I normalize native space DTI-fiber-tracts. See the embedded help for
more detail on syntax.
Good luck,
Bas
Marko Wilke schreef:
> Hi Richard,
>
>> Bear with me - I am a novice....
>
> We all either are or were :)
>
>> I realise there is a code for reversing the linear affine components
>> (although I have yet to track it down) but I am also wondering
>> whether there is now a way in which the nonlinear component can be
>> reversed.
>
> Yes, by segmenting the image in SPM5 (quite counterintuitive, I know).
> During this procedure, a "spatial normalization" mat file will be
> generated, as well as the inverse of that. This will do what you want,
> as described by John on July 31:
>
>> If you used the Segment button in SPM5, then there will also be a
>> *_seg_inv_sn.mat file (as well as a *_seg_sn.mat). You can use this
>> to do a "Write Normalised". If you set the bb to nan(2,3) and the
>> voxel size to nan(1,3), then the dimensions of the inverse normalised
>> image should be (about) the same as those of the image from which the
>> parameters were derived from.
>>
>> If you didn't use the Segment button, then the Deformations utility
>> of SPM5 can be used for taking a *_sn.mat file, inverting it and
>> writing an inverse normalised image. The SPM5 manual should give you
>> a few clues.
>>
>> Best regards,
>> -John
>
> Just one more thing:
>
>> and then adjust the position manually, however, our concern is that
>> the linear components are least effective in our area of interest
>> (which is the temporal pole) and we infact need the non-linear
>> component.
>
> The overall scaling that is occurring in the image is quite dominated
> by the linear scaling. If you look at John's paper from 1997
> ("Incorporating prior knowledge...") you will see that by the affine
> component alone, the image is scaled by about 1.4. Regional or local
> effects of the non-linear normalization interfere, and I agree that a
> combined approach is better. However, if you look at images that are
> normalized using an affine-only approach, you will be surprised how
> well they are registered to the template already (at least I was :)
>
> Best,
> Marko
--
--------------------------------------------------
Dr. S.F.W. Neggers
Division of Brain Research
Rudolf Magnus Institute for Neuroscience
Utrecht University Medical Center
Visiting : Heidelberglaan 100, 3584 CX Utrecht
Room B.01.1.03
Mail : Huispost B.01.206, P.O. Box 85500
3508 GA Utrecht, the Netherlands
Tel : +31 (0)88 7559609
Fax : +31 (0)88 7555443
E-mail : [log in to unmask]
Web : http://www.fmri.nl/people/bas.html
--------------------------------------------------
