Example: matlab

This page shows a simple example on how to containerize your matlab script for this challenge. The source code can also be found on github: hjkuijf/wmhchallenge.

The following matlab script simply thresholds the orig/FLAIR image at gray value 800. To load/save the images, it uses the “Tools for NIfTI and ANALYZE image“. This is placed next to the script in a separate folder, which we add to the matlab path during runtime.

function example()
inputDir = '/input';
outputDir = '/output';

% Add subfolders to the path
currentFolder = fileparts(which(mfilename)); 
addpath(genpath(currentFolder));

nii = load_untouch_nii([inputDir '/orig/FLAIR.nii.gz']);

nii.img(nii.img <  800) = 0;
nii.img(nii.img >= 800) = 1;

save_untouch_nii(nii, [outputDir '/result.nii.gz']);

Since we cannot run the matlab GUI inside a Docker, we need to create a standalone application from this matlab script. The instructions on the matlab website are quite clear, but some small details:

  • Select “Runtime downloaded from web”, since we will pre-install the runtime in the Docker container. Hence it does not need to be included.
  • Be sure to include all “Files required for your application”.

After compilation, you only need the output in the for_redistribution_files_only directory.

Note: for this example, I used MATLAB R2016a running on CentOS 7 64bit.

The Dockerfile should use the correct CentOS version, install the matlab runtime, and add our compiled script. Hence the Dockerfile looks like:

FROM centos:latest
MAINTAINER hjkuijf

RUN yum update -y
RUN yum install wget unzip libXext libXt-devel libXmu -y
RUN mkdir /mcr-install && cd /mcr-install &&  \
    wget -nv http://www.mathworks.com/supportfiles/downloads/R2016a/deployment_files/R2016a/installers/glnxa64/MCR_R2016a_glnxa64_installer.zip && \
    unzip MCR_R2016a_glnxa64_installer.zip && \
    ./install -mode silent -agreeToLicense yes
RUN rm -Rf /mcr-install
ENV LD_LIBRARY_PATH=$LD_LIBRARY_PATH:/usr/local/MATLAB/MATLAB_Runtime/v901/runtime/glnxa64:/usr/local/MATLAB/MATLAB_Runtime/v901/bin/glnxa64:/usr/local/MATLAB/MATLAB_Runtime/v901/sys/os/glnxa64:/usr/local/MATLAB/MATLAB_Runtime/v901/sys/java/jre/glnxa64/jre/lib/amd64/native_threads:/usr/local/MATLAB/MATLAB_Runtime/v901/sys/java/jre/glnxa64/jre/lib/amd64/server:/usr/local/MATLAB/MATLAB_Runtime/v901/sys/java/jre/glnxa64/jre/lib/amd64
ENV XAPPLRESDIR=/usr/local/MATLAB/MATLAB_Runtime/v901/X11/app-defaults
ENV MCR_CACHE_VERBOSE=true
ENV MCR_CACHE_ROOT=/tmp

ADD matlab/example/for_redistribution_files_only /wmhseg_example
RUN ["chmod", "+x", "/wmhseg_example/example"]

The middle block of code downloads and installs the matlab runtime and correctly updates the environment variables. The final ADD command copies our compiled code into the container at the location /wmhseg_example. Finally we chmod the executable, so we can run it.

With the following command, we build a Docker container from our Dockerfile and the compiled matlab code:

docker build -f Dockerfile -t wmhchallenge/[TEAM-NAME] .

Once your container is ready, we can run it with the following command:

docker run -dit -v [TEST-ORIG]:/input/orig:ro -v [TEST-PRE]:/input/pre:ro -v /output wmhchallenge/[TEAM-NAME]

The -v options map the input folder into the container at /input, read-only. The last -v creates an output directory.

This command outputs the Container ID, which you can also look up with:

docker ps

Next, we will execute the example Python script:

docker exec [CONTAINER-ID] python /wmhseg_example/example.py

Since this script is quite small, it doesn’t take long to finish. Next we copy the output from the container to our local machine:

docker cp [CONTAINER-ID]:/output [RESULT-LOCATION]

Finally, we shut down the running container. This also removes the created /output folder and any other changes made to the container.

docker rm -v [CONTAINER-ID]