Multiple Automatically Generated Templates brain segmentation algorithm

---

Looking for the MAGeT Morphology pipeline? Use the morpho-labelmask branch instead.

---

Given a set of labelled MR images (atlases) and unlabelled images (subjects), MAGeT produces a segmentation for each subject using a multi-atlas voting procedure based on a template library made up of images from the subject set.

Here is a schematic comparing 'traditional' multi-atlas segmentation, and MAGeT brain segmentation:

The major difference between algorithms is that, in MAGeT brain, segmentations from each atlas (typically manually delineated) are propogated via image registration to a subset of the subject images (known as the 'template library') before being propogated to each subject image and fused. It is our hypothesis that by propogating labels to a template library, we are able to make use of the neuroanatomical variability of the subjects in order to 'fine tune' each individual subject's segmentation.

To cite MAGeTbrain in publications, please use:

Pipitone J, Park MT, Winterburn J, et al. Multi-atlas segmentation of the whole hippocampus and subfields using multiple automatically generated templates. Neuroimage. 2014;

M Mallar Chakravarty, Patrick Steadman, Matthijs C van Eede, Rebecca D Calcott, Victoria Gu, Philip Shaw, Armin Raznahan, D Louis Collins, and Jason P Lerch. Performing label-fusion-based segmentation using multiple automatically generated templates. Hum Brain Mapp, 34(10):2635–54, October 2013. (doi:10.1002/hbm.22092)

Dependencies

In order to use MAGeTBrain, you will need to install:

• Python 3, we recommend you install the anaconda python 3 distribution
• minc-toolkit-v2
• pyminc
• minc-stuffs
• for cluster integration qbatch
• gnu-parallel

or

Just use the pipeline on Compute Canada's Niagara Cluster.

Getting started, for the impatient:

git clone https://github.com/CobraLab/MAGeTbrain.git
source MAGeTbrain/bin/activate
mb init segmentations
cd segmentations
mb import atlas1_t1.mnc atlas1_labels.mnc
mb import atlas2_t1.mnc atlas2_labels.mnc
... 
mb check
mb run

Quick start

0. Check out a copy of this repository somewhere handy,

    git clone https://github.com/CobraLab/MAGeTbrain.git
   

1. Add the MAGeTbrain/bin folder to your path. This can be done easily by running, source MAGeTbrain/bin/activate (revert PATH by typing deactivate).

2. Create a new folder for your project and,

    mb init
   

3. Copy/link your atlases, templates, subjects into input/atlases, input/templates, and input/subjects, respectively. As per always, MR images go in brains.

   Atlas labels are expected to have the same name as their MR image but with a postfix of _labels.mnc. Use the import command to simplify the renaming process: mb import /path/to/image_file.mnc /path/to/label_file.mnc

   This copies the image and label files into the proper folder and rename the label file accordingly. Use the mb import -l to use symbolic links rather than copying files.

4. MAGeT Brain is operated using the mb command. In order to run the entire pipeline, simply run:

    mb run
   

   This uses the default settings to execute the necessary commands. In particular, it assumes you have access to a qbatch-enabled batch queuing system, and uses the utility qbatch to submit commands in batches. If you are running this on Niagara, you should have no problems.

   If you can also run MAGeT Brain using the GNU parallel on a single machine, like so:

    mb run -q parallel
   

Currently, only majority voting label fusion is enabled. For a more complete implementation, check the 'master' branch.

MAGeT Brain Commands

mb init

mb init [<folder>]

The init command is used to start a new project. It creates a folder structure to hold MAGeT brain inputs and outputs. If no folder name is supplied then the current folder is used. init creates the following folders:

project-folder/
    input/
        atlases/
            brains/
            labels/
        templates/
            brains/
        subjects/
            brains/
    output/

mb import

mb import <image.mnc> <labels.mnc> [<mask.mnc>]

The given image and corresponding labels are imported into the input/atlases folder, with the proper naming convention.

mb check

Checks the input configuration and reports any anomalies and suggestions to correct them. If this command doesn't return any messages then you should be set to run MAGeT brain.

mb status

Reports on the progress of MAGeT brain and shows the number of remaining tasks.

mb run

Starts the MAGeT brain pipeline.

Tutorials

For an in-depth guideline, follow the MAGeTbrain and the MAGeTmorph tutorials.

---

http://tinysong.com/y9lO