[ITK-users] Is image direction taken into account during registration?

[vincent ngai]

Hi all, apologies if this has been asked, but a cursory search of image
direction and registration did not turn up what i was looking for.

Specifically I am trying to find out whether image direction is required
for proper registration.

>From what I understand image direction comes from the DICOM directional
cosines. So say for sagittal images, image origin correspond to the the top
left pixel centre in world space, and the x index ideal direction is [LPS
+ve Y]( 0, 1, 0 )  and y index direction is [LPS -ve Z] ( 0, 0, -1 ). The
direction cosines comes from the table tilt of the scanner and the
scanner's sampling space remains fixed and cartesian as the LPS coordinate
system, and actual orientation of the patient's head may deviate from the
table's directions.

If my understanding of the above is correct(?) Then should the images to be
used for registration have their direction cosines replaced with the given
view's( sagittal, coronal etc ) ideal directions? I have read its dangerous
to change the image direction, but in this case I'm preserving the view's
orientation and ensuring that the registration process is comparing 2
similarly oriented spaces with differently oriented samples (i.e head tilt,
multimodal ).

What I am trying to do is based on the different DICOM scans e.g sagittal,
coronal, bring both images into the common LPS coordinate system, then
perform the registration in that space. And it would seem that preserving
the original direction cosines doesn't make any sense, since they do not
infer the actual scanner's sampling space.


Thank you.


-- 
Regards,
Vincent
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://www.itk.org/pipermail/insight-users/attachments/20140505/0f77ba74/attachment.html>