[Insight-users] question on using WarpImageFilter

[John Drozd]

Hi,

I and my colleague are using Tom's diffeomorphic demons finite
difference algorithm (that I found on the Insight Journal) to warp a nifti
MRI ADNI image1 (what I cll my atlas) to another nifti MRI ADNI image2 (what
I call my subject).  Both images have float pixel types.  I am using 5
levels of 15 iterations each with a sigma of 0.8.  I notice that when the
range between the minimum Jacobian and maximum Jacobian is smaller, the
warping is more accurate.

I first affine register the atlas to the subject and get a .tfm transform
file.

I input this .tfm transform file into the diffeomorphic demons code and warp
the atlas to the subject.
This generates a deformation field which I use to warp the atlas grey scale
image to the subject grey scale image.  The warped atlas looks identical to
the subject.

My atlas has a manually traced hippocampus, and so does my subject.  These
manual tracings were done by an anatomist.

When I warp the atlas's label map of the hippocampus to the subject.  The
warped hippocampus label map does not match the traced hippocampus of the
subject.  One side looks close but the other side is off, as if the the
warped hippocampus was twisted.  We find that it is more accurate when an
interpolator is not used on warping the label map, and when an interpolator
is used when warping the grey scale image.  Am I and my colleague doing
something wrong?

Thank you for your time.

John

-- 
John Drozd
Post-Doctoral Fellow, Robarts Research Institute
The University of Western Ontario
London, ON, Canada
http://publish.uwo.ca/~jdrozd2/index.htm
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://www.itk.org/pipermail/insight-users/attachments/20110729/4ea88729/attachment.htm>