Hello.
I've asked around a few places without luck but maybe someone here has
an idea.
I am trying to convert the warp image from epi_reg (run with a fieldmap)
to something useable with ITK based software (e.g. ANTS).
Theoretically, the 4D warp image (*_warp.nii.gz) created by epi_reg
should be convertible to 3D vector format with c3d. The c3d_affine_tool
works well converting fsl .mat affine transform files to itk transform
files, but this is my first time using it for warp images.
If I naively use c3d to convert the 4D warp image and apply it to the
source image, the resulting brain is mostly out of the field of view and
clearly in the wrong spot. If I remove the --postmat from the
convertwarp step near the end of epi_reg the result is better but not
correct. In this case, converting the affine part ( ${vout}.mat ) to itk
and applying that produces a brain in the correct spot, but it looks
like the warp is not applied at all. For example, I can change the
--shiftdir in convertwarp from y- to y and the resulting image is
identical. Neither image is the same as the warped image from epi_reg;
they're in the same spot but not actually warped/stretched.
So, I guess my questions are:
1. Has anyone gotten the 4D warp to ITK warp conversion to work somehow?
2. the *_fieldmaprads2epi_shift image appears to be 3D. The following line
$FSLDIR/bin/convertwarp -r ${vrefhead} -s ${vout}_fieldmaprads2epi_shift
--postmat=${vout}.mat -o ${vout}_warp --shiftdir=${fdir} --relout
produces a 4D *_warp image with or without including the --postmat portion.
What is contained in the shift image (how is the shift represented) and
what changes are made to it in convertwarp?
Any ideas anyone has would be appreciated.
Thanks
j
--
Jason F. Smith, Ph.D. _.--~~~.~-._
Affective and Translational Neuroscience Lab .( ./ ).
Department of Psychology ( { ).
University of Maryland College Park (' / _ ./ )
[log in to unmask] (' .~~~' / .}
Rm 1147J BPS building ( .~~ )
College Park MD, 20742-4411 '~~( _,,'{{}}'
PH) 301.314.1566 '..:~~'' {{{}}}
|