Adding translation step improves ANTs registration in some datasets

[MarkNelson86]

Unless all MRI images in a multimodal analysis are pretty well aligned to start, we've noticed that the addition of a simple translation step to align images by their geometric center improves registration with ANTs. It's a simple add-on and should only help registration. Here's an example:

# [fixedImage,movingImage,initializationFeature]
        centeralign="[${T1nativepro_brain},${dwi_b0},0]"                                                                        

# AFFINE: dwi to T1w
    antsRegistrationSyN.sh -d 3 -f "$T1nativepro_brain" -m "$dwi_b0"  \ 
             -o "$str_dwi_affine" -t a -n "$threads" -p d  \ 
             -i ${centeralign}

# Apply T1w to dwi affine transformation
    antsApplyTransforms -d 3 -i "$T1nativepro" -r "$dwi_b0"  \
             -t ["$mat_dwi_affine",1] -o "$T1nativepro_in_dwi" -v -u int

This addition also helped improve our T1w-func registration.

[rcruces]

Thank you for your comment. This enhancement is included in our recently released version v0.2.3 which is already on our dockerhub:

docker pull micalab/micapipe:v0.2.3