[Insight-users] Deformable Image registration for spinal cord motion during injury

[Tim Bhatnagar]

Hello all,

I am looking to quantify the motion of the spinal cord during a traumatic
injury. I have before and after 3D MR-images (I believe they are PD, but
could be T2-weighted) of the spinal cord to work with.

My current algorithm design is to perform a rigid registration of the two
image sets, take the transformed image and the final "deformed spinal cord
data" and utilize a deformable registration class. The transforms yielded by
each registration step will then be summed together to determine the total
transformation to go from 'undeformed' to 'deformed'.

The issue I think I will hit soon, is that in the 'deformed' data set, the
visual representation of the spinal cord is very different form the
'undeformed' state. There is some degree of shearing of the cord, which
yields a drastic shape change over the MR slices; additionally, the
MR-signal acquired form the deformed cord yields a qualitatively different
pixel intensity population in matching slices between the MR data sets.
Lastly, there really aren't any dependable landmarks that the registration
process can be 'helped' with.

I realize that this is not an ideal data set to use with deformable
registration techniques, but does anyone have enough experience to recommend
a certain set of registration algorithms? I know I will have to do some
'magic' with the large number of parameters there are, but even an inkling
of a starting point would be great.

I haven't included my ITK user info here as I'm not certain it affects the
theory, but I can post any relevant info if requested. Thanks so much for
your time and help.

Tim Bhatnagar

--
View this message in context: http://itk-insight-users.2283740.n2.nabble.com/Deformable-Image-registration-for-spinal-cord-motion-during-injury-tp6971185p6971185.html
Sent from the ITK Insight Users mailing list archive at Nabble.com.