Inquiry Regarding Suboptimal Results in Macaque T1 Image Segmentation Using antsbrainextraction.sh

[AnastasiaWen]

Inquiry Regarding Suboptimal Results in Macaque T1 Image Segmentation Using antsbrainextraction.sh

[cookpa]

I expect it's because of antsAI downsampling the images too much. I can work on a fix, but in the interim, you could try getting an initial transform with

antsRegistrationSyNQuick.sh -f vic_ants/vic_NMT_prior.nii.gz -m vic_ants/vic_N4.nii.gz -t r -o vicToNMT_Affine

and then adding

-r vicToNMT_Affine0GenericAffine.mat

to the antsBrainExtraction.sh command.

[cookpa]

Initial affine worked for me, BTW

[ntustison]

Can you upload the full input dataset so that we can reproduce the issue on our end?

[AnastasiaWen]

Can you upload the full input dataset so that we can reproduce the issue on our end?

Thanks a bunch for your help. Unfortunately, the file size was too large to upload directly to GitHub，so I've uploaded the files to Google Drive. You can download them using the following link:
https://drive.google.com/drive/folders/1yf8d0uEasbpz0alxk70H-SM4pMIMQmqa?usp=sharing

Here's the command I used (note:name=vic):
$ANTSPATH/antsBrainExtraction.sh -d 3 -a ${name}_N4.nii.gz -e ${name}_NMT_prior.nii.gz -m ${name}_NMT_brainmask_prob_prior.nii.gz -o ${name}_ -f ${name}_NMT_brainmask_prior.nii.gz

Additionally, my goal is to perform segmentation and utilize the results to generate surfaces in FreeSurfer, following the workflow outlined below:
https://prime-re.github.io/structural/surfaces_and_flatmaps_notebook/Surfaces_and_Flatmaps.html

Thank you for your time and assistance.

[AnastasiaWen]

Thanks for your effective help! It works well :D

[YCHuang0610]

I expect it's because of antsAI downsampling the images too much. I can work on a fix, but in the interim, you could try getting an initial transform with

antsRegistrationSyNQuick.sh -f vic_ants/vic_NMT_prior.nii.gz -m vic_ants/vic_N4.nii.gz -t r -o vicToNMT_Affine

and then adding

-r vicToNMT_Affine0GenericAffine.mat

to the antsBrainExtraction.sh command.

Dear cookpa
I have met the same problem when using antsBrainExtraction.sh and antsCorticalThickness.sh to deal with monkey brains. I tried your solution and It works when using antsBrainExtraction.sh. But I didn't find the same argument (antsBrainExtraction.sh -r to do the Initial moving transform) in antsCorticalThickness.sh.
Since antsCorticalThickness.sh calls antsBrainExtraction.sh , I am wondering whether the same argument to do the Initial moving transform should be added to _antsCorticalThickness.sh pipeline as well?
Besides, when I am doing antsRegistrationSyNQuick.sh to get an initial transform, if I have to add the "-t r" argument to specifiy the "rigid" transform type? Since some T1 imgs can't be well registered only by using rigid transform.

[cookpa]

The -t r is just to do a fast initial transform. It's only a starting point for the rest of the registration, so it's OK if it's not a very good alignment. You can switch to -t a if you like.

Regarding antsCorticalThickness.sh, it doesn't support an initial transform but it will find an existing brain mask. So if you run it with -o myOutput_ and there exists myOutput_BrainExtractionMask.nii.gz, it will not re-run the brain extraction and will carry on from the next step.