Pituitary Segmentation

Atlas-based segmentation of the pituitary in MR images.

New Features

• Now supports multi-atlas segmentation. Check out multi_atlas_example.py for more details.
• Also provides a command-line tool for automatic pituitary segmentation without dependency on Python or ANTsPy.
• Python tools
  • For segmentation using a single atlas, use main.py.
  • For an example of how to leverage multiple atlases to give more robust segmentation, see multi_atlas_example.py.
• Bash tool (recommended)
  • For the command-line tool written bash that supprots multiple atlases, use pituitary_segmentation.sh.

Bash tool

Dependencies

• ANTs

Check out ANTs' official repository for instructions on how to install ANTs.

Usage

Atlas-based pituitary segmentation using ANTs.

Usage: ./pituitary_segmentation.sh <input> <output> [-t transform] [-c cutoff] [-m threads] [-n] [-h]

Options:
  <input>         Input image filename.
  <output>        Output image filename.
  -t transform    Type of transform to use in registration. Default: Affine. Currently supported: Affine, SyN, SyNQuick.
  -c cutoff       Cutoff value for the mask. Default: 5.
  -m threads      Number of threads to use. Default: 1. Increase this value to speed up the registration process.
  -n              Apply N4 bias correction to the input image.
  -h              Display this help message.

Python tool

Dependencies

• ANTsPy

To install ANTsPy, run:

pip install antspyx

Usage

python main.py --in <input_image> --out <output_image>

Options

• --transform: Specify the type of transform to use for ANTs registration. Default is Affine. For other options, see ANTsPy documentation.
• --n4: Optionally apply N4 bias correction to the input image.

Acknowledgements

• MNI152 2009a Nonlinear Asymmetric atlas taken from here.
• Pituitary mask annotated by expert.