Using ANTs landmarkmatch to register the colin27 to a study template and it looks strange - help!

[arianayoum]

Hello, I am trying to register the colin27 MNI brain to an older-adult study template. I would like to use specific markers to anchor this registration so I am using the landmarkmatch function. However, the end result looks strange with a little protruding bump on the registered brain that does not exist on the fixed nor moving brains. Attached is a picture of the registered brain with the protrusion. Any help is welcome - thank you!

[cookpa]

That script was written for hippocampus matching and uses tools that haven't been updated in a long time. Could you try with antsRegistration? Start with antsRegistrationSyN.sh, it will register the images and print a command you can modify. Then you can add a point set metric (put the image metric first).

As always, code and data enables us to give the most detailed advice.

[arianayoum]

Hi, I tried with the antsRegistrationSyN but there is still an abnormal protrusion.
I know you said it hasn't been updated in a while, but is there a way to edit the landmarkmatch function so that it reduces the amount of warping/deformation?

Attached are:

1. Warped - the output of the antsRegistrationSyN, of the registered colin27 brain to the older adult template brain
2. colin_skillstripped - the colin27 brain that's been skull-stripped with BET
3. ANTS70-74 - the older adult template brain

Warped.nii.gz
colin_skullstripped.nii.gz
ANTS70-74Years3T_brain_bias_corrected.nii.gz

[cookpa]

You can smooth the deformation field by replacing the smoothing term Gauss[ 3,0 ] to something like Gauss[ 3,1 ]

[cookpa]

The fixed image is cropped very tightly, which can cause boundary instability in registration. By padding the image

ImageMath 3 fixed_pad.nii.gz PadImage ANTS70-74Years3T_brain_bias_corrected.nii.gz 20

I was able to get better results with antsRegistrationSyN.sh.

[arianayoum]

Hello, thank you so much! It works well :) I didn't realize the cropping would affect it - makes sense, and I really appreciate your help!

[ntustison]

Also, related---general landmark matching for a variety of transforms is available in ANTsPy and ANTsR. Here's a couple of self-contained examples.