The way the first step of the registration works is that it does an 18
parameter fit of the two images to their respective templates.
Image1 is fit using a matrix generated by R1*S, where R1 is a rigid body
transformation (3 rotations and 3 translations) and S is a transformation
parameterised by 3 shears and 3 zooms (including a negative zoom for L-R
flipping).
Image2 is also simultaneously registered by R2*S, where S is the same as in
the first transformation and R2 is a different set of rotations and
translations.
R1*S and R2*S are full 12 parameter affine transformations that register the
images to the templates, but they are also used to register the individual
images together by a rigid body transformation. This can be thought of as an
affine mapping from one image to the templates and back to the other image:
(R1*S)*inv(R2*S)
= R1*S*inv(S)*inv(R2)
= R1*inv(R2)
.... which is a rigid body transformation that does not include any zooms,
shears or flips. After the registration, the images remain unflipped and
therefore will still be L=R.
Best regards,
-John
> I'd like to check my understanding of some orientation issues:
>
> I have
> 1) an spgr (3D anatomical T1) and
> 2) a functional image
> which I wish to coregister to one another.
>
> Both are in radiological (L=R) orientation.
>
> What I expect to happen (based on the spm archive materials)is that the
> coregistration process will first normalize each image to the appropriate
> template (thus flipping them into neurological orientation). Then it will
> register them to each other.
>
> So I will end up with the images in neurological orientation (L=L) and
> registered to each other.
>
> I have not changed any spm defaults, so the normalization procedure should
> expect images in radiological orientation.
--
Dr John Ashburner.
Wellcome Department of Cognitive Neurology.
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
mail: [log in to unmask]
