antsRegistration V.S. antsMotionCorr

[zw1989]

Hi,

I have tried to use antsRegistration to do motion estimation and correction as the code shows below. The epi1.nii is the 4D EPI images and the epi1avg4D.nii is the replication of the average of the 4D EPI along time. The results are good and the run time is 214 seconds.

antsRegistration --dimensionality 4 --float 0 \
--output [run4D_,run4D_Warped.nii.gz,run4D_inverswarped.nii.gz] \
--interpolation Linear \
--initial-moving-transform [epi1avg4D.nii, epi1.nii, 0] \
--winsorize-image-intensities [0.005,0.995] \
--use-histogram-matching 0 \
--transform Rigid[0.1] \
  --metric MI[epi1avg4D.nii, epi1.nii,1,32,Regular,0.25] \
  --convergence [1000x500x250x100,1e-6,10] \
  --shrink-factors 8x4x2x1 \
  --smoothing-sigmas 3x2x1x0vox \
--transform Affine[0.1] \
  --metric MI[epi1avg4D.nii, epi1.nii,1,32,Regular,0.25] \
  --convergence [1000x500x250x100,1e-6,10] \
  --shrink-factors 8x4x2x1 \
  --smoothing-sigmas 3x2x1x0vox \

Then I find there is another function antsMotionCorr that can do the motion correction, too (http://stnava.github.io/fMRIANTs/). I then tested the following code whose parameter setting is quite similar to that in antsRegistration. The results are good, too, but the run time is about 1010 seconds.

antsMotionCorr --dimensionality 3 \
--output [run4D_,run4D_.nii.gz,run4D_avg.nii.gz] \
--interpolation Linear \
--use-histogram-matching 0 \
--transform Rigid[0.1] \
  --metric MI[epi1avg.nii, epi1.nii,1,32,Regular,0.25] \
  --shrinkFactors 8x4x2x1 \
  --smoothingSigmas 3x2x1x0 \
  --u 1 -e 1 -i 100x70x50x20 -n 3 \
--transform Affine[0.1] \
  --metric MI[epi1avg.nii, epi1.nii,1,32,Regular,0.25] \
  --shrinkFactors 8x4x2x1 \
  --smoothingSigmas 3x2x1x0 \
  --u 1 -e 1 -i 100x70x50x20 -n 3 \

At first I thought antsMotionCorr is a wrap of antsRegistration, but it is actually different. My questions is that what's the main difference between antsMotionCorr and antsRegistration? Why does it take so much longer when the registration parameters are the same? Which one is recommended to use?

In addition, in this post http://stnava.github.io/fMRIANTs/, the deformable registration is done by using antsRegistration, but why not use antsMotionCorr which can also have the Syn transform?