Registering the Brain/MINDS Marmoset Atlas to New Data Using ANTs

Introduction

This tutorial will teach the user how to use Advanced Normalization Tools (ANTs) to do image registation in 3D using data from Brain/MINDS data portal and how to apply the transforms/inverse transfroms from image registration.

This is an important step if the user want to map the Brain/MINDS marmoset brain atlas to their own data.

ANTs Installation

The latest release of ANTs can be found on the official github page here.

This tutorial uses v.2.3.5(Ecphorella).

Please follow the guide to compile and install ANTs for Linux/MacOS or Windows.

Download Sample Data

In this tutorial we will use the data sets from the data portal atlas download page.

We will use BMA 2017 Ex Vivo v1.1.0 and BMA 2019 Ex Vivo v1.0.0.

Image Registration using ANTs

The files we need for this tutorial are:

bma-1-mri.nii.gz (brain 1) and bma-1-region_seg.nii.gz (label 1) from BMA 2017

sp2_avg_mri_exvivo_t2wi_v1.0.0.nii.gz (brain 2) and sp2_label_512_v1.0.0.nii.gz (label 2) from BMA 2019.

Copy and paste above files into the same folder for easy processing.

You can view Nifti files using tools such as ITK-SNAP or 3D Slicer.

Here, we will demostrate how the map label 1 from BMA 2019 onto the brain 2 from BMA 2017 and vice versa.

  • Register brain 1 to brain 2:

Run the following command:

antsRegistrationSyNQuick.sh -d 3 -f sp2_avg_mri_exvivo_t2wi_v1.0.0.nii.gz -m bma-1-mri.nii.gz  -o brain1_to_brain2_

Lets go over each arguments in the above command:

-d 3: This is the dimension, since we are working with 3D volumes, the dimension is 3.

-f sp2_avg_mri_exvivo_t2wi_v1.0.0.nii.gz: This is the fixed image or the image we want to register to, in our case, this will be brain 2.

-m bma-1-mri.nii.gz: This is the moving image or the image we want to register, in our case, this will be brain 1.

-o brain1_to_brain2_: This is the output prefix, in this example, the output files will all have the prefix ”brain1_to_brain2″

  • Optional argument:

-n 8: This is the number of thread to use during registration to speed things up, it is recommended to add this to the end of the above command.

The registration process might take up to 20 minutes or more depend on how many threads you are using and how much your CPU can handle.

  • Outputs:

After the registration command, there should be 5 new files in the same folder:

brain1_to_brain2_0GenericAffine.mat: This is the output for affine transform.

brain1_to_brain2_1InverseWarp.nii.gz: This is the output for inverse displacement field.

brain1_to_brain2_1Warp.nii.gz: This is the output for displacement field.

brain1_to_brain2_InverseWarped.nii.gz: This is the result of brain2 registered to brain1.

brain1_to_brain2_Warped.nii.gz: This is the result of brain1 registered to brain2.

As you can see from the above outputs, running the registration command will give you the necessary transforms to register brain2 to brain1 and the inverse transforms.

Next we will demonstrate how to use the transform files to perform direct and inverse transforms.

  • Run the following command to map label1 to brain2:

antsApplyTransforms -d 3 -i bma-1-region_seg.nii.gz -r sp2_avg_mri_exvivo_t2wi_v1.0.0.nii.gz -o label1_to_brain2.nii.gz -t brain1_to_brain2_1Warp.nii.gz -t brain1_to_brain2_0GenericAffine.mat -n NearestNeighbor

Lets go over each arguments in the above command:

-d 3: This is the dimension, since we are working with 3D volumes, the dimension is 3.

-i bma-1-region_seg.nii.gz: This is the volume we want to register, in this case, label1.

-r sp2_avg_mri_exvivo_t2wi_v1.0.0.nii.gz: This is the reference image we are registering to, in this case, brain2.

-o label1_to_brain2.nii.gz: This is the output file, we name it label1_to_brain2.nii.gz.

-t brain1_to_brain2_1Warp.nii.gz: The transforms are applied from right to left, so this is the 2nd transform to be applied, in our case, we will apply the displacement field 2nd.

-t brain1_to_brain2_0GenericAffine.mat: This is the first transform to be applied, here we will apply the affine transforms.

-n NearestNeighbor: This is the interpolation method to be used, which is not necessary in most cases, here, since we are applying transforms to a label map, we will use NearestNeighbor.

 
ss1

You can examine the output using 3D slicer, by loading both label1_to_brain2.nii.gz and sp2_avg_mri_exvivo_t2wi_v1.0.0.nii.gz, the transformed label1 should match the shape of brain2 as shown in the image above.

  • To apply the inverse transform, run the following command to map label2 to brain1:

antsApplyTransforms -d 3 -i sp2_label_512_v1.0.0.nii.gz -r bma-1-mri.nii.gz -o label2_to_brain1.nii.gz -t [brain1_to_brain2_0GenericAffine.mat, 1] -t brain1_to_brain2_1InverseWarp.nii.gz -n NearestNeighbor

Lets go over each arguments in the above command:

–d 3: This is the dimension, since we are working with 3D volumes, the dimension is 3.

-i sp2_label_512_v1.0.0.nii.gz: This is the volume we want to register, in this case, label1.

-r bma-1-mri.nii.gz-o label2_to_brain1.nii.gz: This is the reference image we are registering to, in this case, brain1.

-t [brain1_to_brain2_0GenericAffine.mat, 1]: The transforms are applied from right to left, so this is the 2nd transform to be applied, in our case, we will apply the inverse of the affine transforms 2nd.

-t brain1_to_brain2_1InverseWarp.nii.gz: This is the first transform to be applied, here we will apply the inverse displacement field.

-n NearestNeighbor: This is the interpolation method to be used, which is not necessary in most cases, here, since we are applying transforms to a label map, we will use NearestNeighbor.

 
ss2

You can examine the output using 3D slicer, by loading both label2_to_brain1.nii.gz and bma-1-mri.nii.gz, the transformed label2 should match the shape of brain1 as shown in the image above.

If you want to compare the result with the label map before registration, you can do it by loading sp2_label_512_v1.0.0.nii.gz and bma-1-mri.nii.gz.

  • Apply multiple transformations:

Imagine a situation where we have three images A, B and C, where we have registered A to B and C to B and have the transformations.

Now we have enough information to transform A to the space of C without doing registration of A to C.

To do this, we will use :

antsApplyTransforms -d 2 -i A -r C -o AtoC -t [C_to_B_0GenericAffine.mat, 1] -t C_to_B_1InverseWarp.nii.gz -t A_to_B_1Warp.nii.gz -t A_to_B_0GenericAffine.mat

since the transforms are applied from right to left, here, we want to apply transforms in the order of A to B first then the inverse of C to B.

By following this rule, you can apply multiple transformations in order to transform images into different spaces.

Last updated 11-30-2022