ANTs — Advanced Normalization Tools

Overview

ANTs (Advanced Normalization Tools) is an open-source software suite for image registration and segmentation, developed at the University of Pennsylvania. In the DTI pipeline, ANTs is used for two key tasks:

1. Brain extraction (skull stripping) of T1 structural images — using a template-based approach that is more accurate than simpler methods
2. Tissue segmentation — classifying voxels into CSF, gray matter, and white matter (used for ICV calculation)

ANTs is not used in every pipeline stage, but the brain extraction it produces in Step 2 is a critical input for later registration steps.

GitHub: https://github.com/ANTsX/ANTs

Installation

Option 1: Pre-built Binaries (Recommended)

The easiest installation method — no compiling required.

1. Go to the ANTs Releases page
2. Download the archive for your platform (e.g., ants-2.5.1-centos7-X64-gcc.zip for Linux)
3. Extract to a permanent location:

# Download (check releases page for latest version)
wget https://github.com/ANTsX/ANTs/releases/download/v2.5.1/ants-2.5.1-centos7-X64-gcc.zip

# Extract
unzip ants-2.5.1-centos7-X64-gcc.zip -d /opt/

# Add to your shell profile
echo 'export ANTSPATH="/opt/ants-2.5.1/bin"' >> ~/.bashrc
echo 'export PATH="$ANTSPATH:$PATH"' >> ~/.bashrc
source ~/.bashrc

Option 2: conda

conda install -c aramislab ants
# or
conda install -c conda-forge ants

conda packages may lag behind official releases. Verify that antsBrainExtraction.sh is included — some minimal packages omit the shell scripts.

Option 3: Build from Source

Building ANTs from source gives you the latest features but takes 1–2 hours and requires significant RAM (~8 GB).

# Prerequisites
sudo apt-get install cmake git build-essential zlib1g-dev

# Clone
git clone https://github.com/ANTsX/ANTs.git
cd ANTs
mkdir build && cd build

# Configure and build
cmake -DCMAKE_INSTALL_PREFIX=/opt/ANTs \
      -DBUILD_TESTING=OFF \
      -DCMAKE_BUILD_TYPE=Release \
      ..
make -j$(nproc)
make install

# Add to PATH
echo 'export ANTSPATH="/opt/ANTs/bin"' >> ~/.bashrc
echo 'export PATH="$ANTSPATH:$PATH"' >> ~/.bashrc
source ~/.bashrc

• The build uses a lot of RAM. On systems with less than 16 GB, reduce parallelism: make -j2 instead of make -j$(nproc)
• If cmake fails, check your cmake version: ANTs requires cmake ≥ 3.16. Update with pip install cmake or from cmake.org
• BUILD_TESTING=OFF skips the test suite, cutting build time significantly

Verify Installation

# Check that key commands are available
which antsBrainExtraction.sh
# Expected: /opt/ANTs/bin/antsBrainExtraction.sh (or your install path)

which Atropos
# Expected: /opt/ANTs/bin/Atropos

# Quick version check
antsRegistration --version

Brain Extraction Templates

ANTs brain extraction (antsBrainExtraction.sh) requires a template and a template probability mask. These tell the algorithm what a "typical brain" looks like so it can identify brain vs. non-brain tissue in your images.

Downloading Templates

Download templates from the official ANTs figshare repository: https://figshare.com/articles/dataset/ANTs_ANTsR_Brain_Templates/915436

Which Template Should I Use?

Template	Population	Best For
OASIS	Adults (18–96 years)	General adult neuroimaging studies
NKI	Adolescents and adults	Studies spanning teenage to adult
MNI/ICBM	Adults	Standard reference, widely used
NIH Pediatric	Children (4–18 years)	Pediatric studies — brain shape differs significantly from adults
Study-specific	Your participants	If you have 20+ subjects, consider building a study-specific template

Using an adult template on pediatric data (or vice versa) will produce poor brain extractions. If your participants are children, elderly, or a clinical population with atypical brain anatomy, choose a template that matches your population.

Template File Structure

After downloading, you should have two files per template:

OASIS/
  T_template0.nii.gz                  # The template brain image
  T_template0_BrainCerebellumProbabilityMask.nii.gz  # Brain probability mask

These are passed to antsBrainExtraction.sh with the -e (template) and -m (mask) flags. See Step 2: Skull Stripping for the full command.

Key Commands for DTI

antsBrainExtraction.sh

Used in Step 2: Skull Stripping to remove non-brain tissue from T1 structural images.

antsBrainExtraction.sh \
    -d 3 \
    -a "$t1_image" \
    -e "$template" \
    -m "$template_mask" \
    -o "$output_prefix"

Flag	Meaning
-d 3	3-dimensional image
-a	Input anatomical image (your T1)
-e	Template image
-m	Template brain probability mask
-o	Output prefix (produces *Brain.nii.gz and *BrainMask.nii.gz)

Atropos

Used in ICV Calculation for tissue segmentation (CSF, gray matter, white matter).

Atropos -d 3 \
    -a "$brain_image" \
    -i KMeans[3] \
    -o "$output_prefix" \
    -x "$brain_mask"

Common Issues

Issue	Cause	Solution
antsBrainExtraction.sh: command not found	ANTs scripts not on PATH	Add $ANTSPATH to PATH, and check that the Scripts/ directory is also on PATH
Brain extraction runs for hours	Normal — ANTs registration is thorough	Expect 20–45 min per subject. Use nohup or tmux
Poor brain extraction	Wrong template for your population	Try a different template; check that the T1 is not corrupted
Build fails at cmake	cmake too old	Update cmake: pip install cmake --upgrade
ITK: ERROR during build	Out of memory	Reduce build parallelism: make -j2

References

• Avants BB, Tustison NJ, Song G, Cook PA, Klein A, Gee JC (2011). A reproducible evaluation of ANTs similarity metric performance in brain image registration. NeuroImage, 54(3), 2033-2044.
• Tustison NJ, et al. (2014). Large-scale evaluation of ANTs and FreeSurfer cortical thickness measurements. NeuroImage, 99, 166-179.
• ANTs GitHub Wiki — Official documentation and examples