Can't extract brain from skull on T1-weighed MRI images

[ValeriePineauNoel]

Describe the problem I have a T1-weighed image sequence from a human head and I want to extract, on each image, the brain and get rid of the skull for registration. I tried this tutorial (Brain extraction in ANTs) with the Kirby brain template provided here (ANTs brain templates). I had an ERROR: command exited with nonzero status 139. From the console, I can see that the error is from the command antsApplyTransforms (0x10f91f6a5 int ants::antsApplyTransforms<double, 3u, double>(itk::SmartPointer&, unsigned int) + 7909), I'm guessing from the file antsBrainExtraction.sh.

To Reproduce I basically followed this : ANTs brain templates However, instead of putting this command

antsBrainExtraction.sh -d 3 -a t1.nii.gz -e T_template0.nii.gz \
-m T_template0_BrainCerebellumProbabilityMask.nii.gz -o output

I put this :

antsBrainExtraction.sh -d 3 -a Ax_3D_postt1_Mprage_8.nrrd -e <path>/Kirby/S_template3.nii.gz -o T1_brain_without_skull.nii.gz

System information (please complete the following information)

• OS: Mac OS 2017
• OS version: Ventura 13.4
• Type of system: Desktop
• CPU architecture: x86_64

ANTs version information

• ANTs code version: 2.4.4.post6-g28ce5ff
• ANTs installation type: Compiled from source. I think? I basically cloned the github with `git clone https://github.com/ANTsX/ANTs.git` and made a folder for my ANTs path.

Additional information

[ntustison]

At minimum, please post the entire error output. Even better, though, would be to post the data so that we can try to reproduce the error on our end.

Alternatively, I would recommend using ANTsPy/ANTsPyNet or ANTsR/ANTsRNet (brain extraction tutorial here.

[cookpa]

Yes I would need to see more context from the complete logs. They are often too long to paste but you can do

antsBrainExtraction.sh -d 3 -a Ax_3D_postt1_Mprage_8.nrrd -e <path>/Kirby/S_template3.nii.gz \
  -o T1_brain_without_skull.nii.gz 2>&1 | tee abe_command_output.txt

and then upload the text file here.

[ValeriePineauNoel]

Hi! Thanks for your quick answers. Indeed I didn't put the logs because it was quite long. Here is the command_output.txt : abe_command_output.txt

Sadly, because I am working with human brain scans, I cannot share my data for ethical reasons. However, I can tell you that my image sequence looks like this, but with a skull : CIT MRI brain.

[cookpa]

Thanks! Can you try

antsBrainExtraction.sh -d 3 -a Ax_3D_postt1_Mprage_8.nrrd -e <path>/Kirby/S_template3.nii.gz -o T1_brain_without_skull

It's possible that the extra extensions in the output are causing the problem.

[cookpa]

I just tried to reproduce and the extension mangling does not cause a problem for me (ants 2.5.1), neither does using a .nrrd reference image.

Is it possible that you are running out of memory? Is the reference image very large?

[ValeriePineauNoel]

My reference image sequence is a .nrrd and has 176 slices. The images are 16-bit and are of dimension 220 X 320 pixels. In total, the file is 24.8 MB. I didn't think the file is too big. Do you know how I could verify if I run out of memory on my MacOS?

[ValeriePineauNoel]

I had the same error after trying this : antsBrainExtraction.sh -d 3 -a Ax_3D_postt1_Mprage_8.nrrd -e <path>/Kirby/S_template3.nii.gz -o T1_brain_without_skull abe_command_output2.txt

[cookpa]

OH, I think I see it now

    Using antsBrainExtraction with the following arguments:
      image dimension         = 3
      anatomical image        = Ax_3D_postt1_Mprage_8.nrrd
      extraction template     = /Users/dcclab/Documents/Marc/shavasana_registration/T1-CIT168onpost/BrainTemplates/Kirby/S_template3.nii.gz
      extraction reg. mask    = 
      extraction prior        = 
      output prefix           = T1_brain_without_skull.nii.gz
      output image suffix     = nii.gz

You have specified the extraction template, but not the prior probability mask to go with it. Is there a brain mask or brain probability mask to go with S_template3.nii.gz? You need to have that and pass it to the script with -m

[ValeriePineauNoel]

Oh I see! But there are no probability mask with the Kirby brain template. Oasis has a file "...ProbabilityMask.nii.gz" but not the Kirby.

[ValeriePineauNoel]

I will try with NKI.

[cookpa]

Sorry about that, you can make a mask with

ThresholdImage 3 S_template3_BrainCerebellum.nii.gz S_template3_BrainCerebellumMask.nii.gz 0.01 Inf

[cookpa]

Or another template will work also. Note that -m is where you pass the brain outline, while the optional -f is for a mask to constrain registration.

Example here for NKI, the brain mask -m is in red ie (T_template_BrainCerebellumProbabilityMask.nii.gz)

[ValeriePineauNoel]

Oh ok I see! So I could really get any brain template then. I tried it with the NKI and it worked very well. Thank you both for you help! The problem came from the fact that I didn't select a probability mask. abe_command_output3.txt