dHCP Structural Pipeline

The dHCP structural pipeline performs structural analysis of neonatal brain MRI images (T1 and T2) and consists of:

• cortical and sub-cortical volume segmentation
• cortical surface extraction (white matter and pial surface)
• cortical surface inflation and
• projection to sphere

It is described in detail in:

A. Makropoulos, E. C. Robinson et al. "The Developing Human Connectome Project: a Minimal Processing Pipeline for Neonatal Cortical Surface Reconstruction" link

Developers

Antonios Makropoulos: main author, developer of the structural pipeline, and segmentation software. more

Andreas Schuh: contributor, developer of the cortical surface extraction, and surface inflation software. more

Robert Wright: contributor, development of the spherical projection software.

License

The dHCP structural pipeline is distributed under the terms outlined in LICENSE.txt.

Running the pipeline from dockerhub

You can run the pipeline in a docker container. This will work on any version of any platform and is simple to set up. First, install docker:

https://docs.docker.com/engine/installation/

Next, you need to make a directory to hold the images you want to analyze and the results from the pipeline. For example:

$ mkdir data
$ cp T1w.nii.gz data
$ cp T2w.nii.gz data

The T1 image is optional. You can use any names for the images and the directory, though you'll obviously have to modify the next command slightly.

Get the latest version of the pipeline from dockerhub like this:

$ docker pull biomedia/dhcp-structural-pipeline:latest 

And finally, execute the pipeline like this:

$ docker run --rm -t \
    -u $(id -u):$(id -g) \
    -v $PWD/data:/data \
    biomedia/dhcp-structural-pipeline:latest subject1 session1 44 \
            -T1 data/T1w.nii.gz -T2 data/T2w.nii.gz -t 8

Substituting subject and session codes, and the post-menstrual age at scan, see below.

Once the command completes, you should find the output images in your data folder.

The dhcp-pipeline.sh script has the following arguments:

./dhcp-pipeline.sh <subject_ID> <session_ID> <scan_age> -T2 <T2_image> \
    [-T1 <T1_image>] [-t <num_threads>]

where:

Examples:

./dhcp-pipeline.sh subject1 session1 44 -T2 subject1-T2.nii.gz -T1 subject1-T1.nii.gz -t 8
./dhcp-pipeline.sh subject2 session1 36 -T2 subject2-T2.nii.gz -T1 subject2-T1.nii.gz 
./dhcp-pipeline.sh subject3 session4 28 -T2 subject3-T2.nii.gz 

The output of the pipeline is the following directories:

• sourcedata: folder containing the source images (T1,T2) of the processed subjects
• derivatives: folder containing the output of the pipeline processing

Measurements and reporting for the dHCP Structural Pipeline can be computed using:

https://github.com/amakropoulos/structural-pipeline-measures

Rebuild the docker image

In the top directory of dhcp-structural-pipeline, use git to switch to the branch you want to build, and enter:

$ docker pull ubuntu:xenial
$ docker build -t biomedia/dhcp-structural-pipeline:latest .
$ docker push biomedia/dhcp-structural-pipeline:latest

Install natively

If you want to work on the code of the pipeline, it will be more convenient to install natively to your machine. Only read on if you need to do a native install.

FSL

The dHCP structural pipeline uses FSL. You'll need to read their install pages.

Packages

The dHCP structural requires installation of the following packages.

macOS (tested on version 10.9.5)

This is easiest with homebrew. Install that first, then:

$ brew update
$ brew install gcc5 git cmake unzip tbb boost expat cartr/qt4/qt
$ sudo easy_install pip
$ pip install contextlib2

Ubuntu (tested on version 16.04)

$ sudo apt -y update
$ sudo apt -y install g++-5 git cmake unzip bc python python-contextlib2 
$ sudo apt -y install libtbb-dev libboost-dev zlib1g-dev libxt-dev 
$ sudo apt -y install libexpat1-dev libgstreamer1.0-dev libqt4-dev

Debian GNU (tested on version 8)

$ sudo apt -y update
$ sudo apt -y install git cmake unzip bc python python-contextlib2 
$ sudo apt -y install libtbb-dev libboost-dev zlib1g-dev libxt-dev libexpat1-dev 
$ sudo apt -y install libgstreamer1.0-dev libqt4-d
$ # g++-5 is not in the default packages of Debian
$ # install with the following commands:
$ echo "deb http://ftp.us.debian.org/debian unstable main contrib non-free" | sudo tee -a /etc/apt/sources.list
$ sudo apt-get -y update
$ sudo apt-get -y install g++-5

CENTOS (tested on version 7)

$ sudo yum -y update
$ sudo yum -y install git cmake unzip bc python tbb-devel boost-devel qt-devel zlib-devel libXt-devel expat-devel gstreamer1-devel 
$ sudo yum -y install epel-release
$ sudo yum -y install python-contextlib2
$ # g++-5 is not in the default packages of CENTOS, install with the following commands:
$ sudo yum -y install centos-release-scl
$ sudo yum -y install "devtoolset-4-gcc*"
$ # then activate it at the terminal before running the installation script
$ scl enable devtoolset-4 bash

Red Hat Enterprise Linux (tested on version 7.3)

$ sudo yum -y update
$ sudo yum -y install it cmake unzip bc python tbb-devel boost-devel qt-devel zlib-devel libXt-devel expat-devel gstreamer1-devel
$ # the epel-release-latest-7.noarch.rpm is for version 7 of RHEL, this needs to be adjusted for the user's OS version
$ curl -o epel.rpm https://dl.fedoraproject.org/pub/epel/epel-release-latest-7.noarch.rpm
$ sudo yum -y install epel.rpm
$ sudo yum -y install python-contextlib2
$ # g++-5 is not in the default packages of RHEL, install with the following commands:
$ sudo yum-config-manager --enable rhel-server-rhscl-7-rpms
$ sudo yum -y install devtoolset-4-gcc*
$ # then activate it at the terminal before running the installation script
$ scl enable devtoolset-4 bash

Installation

$ ./setup.sh [-j <num_cores>] 

where num_cores the number of CPU cores used to compile the pipeline software.

The setup script installs the following software packages.

The '-h' argument can be specified to provide more setup options:

$ ./setup.sh -h

Once the installation is successfully completed, if desired, the different commands/tools built (workbench, MIRTK and pipeline commands) can be included in the shell PATH by running:

$ . parameters/path.sh