OCMDemo.jl

This is demo code for the paper

Preiswerk, Frank, et al. "Hybrid MRI‐Ultrasound acquisitions, and scannerless real‐time imaging." Magnetic Resonance in Medicine (2016).

How to use

If you have already installed Julia version 0.5 or higher, you can skip the first step.

Install Julia

You can downoaad a version of Julia for your operating system from the Julia website, or use your system's package manager. For example, on MacOS with Homebrew installed, you can type

brew cask install julia

If installation succeeds, you will be able to start Julia (i.e. the Read/Evaluate/Print/Loop, or REPL) by typing julia on your command line.

Install the package

In the REPL, install this package using Julia's built in package manager by typing

Pkg.clone("https://github.com/fpreiswerk/OCMDemo")

This will automatically install the package and all its dependencies. It might take a while, especially if you have freshly installed Julia.

Download the sample data

Hybrid OCM-MRI data of three subjects (A, B and H from the paper) were made available. You can download the data by calling the download_data.jl script from the REPL,

include(joinpath(Pkg.dir("OCMDemo"),"examples","download_data.jl"))

The data is ~400MB, so this might take a while.

Playing around

When the sample data is downloaded, you can take a look at the example script run.jl. It is located in the same folder as above, so typing

joinpath(Pkg.dir("OCMDemo"),"examples","run.jl")

in the REPL will reveal its location. Open it and play around, e.g. using

edit(joinpath(Pkg.dir("OCMDemo"),"examples","run.jl"))

All experiments are defined in XML files: Three experiment definitions are available, A1.xml, B2.xml and H2.xml, respectively, corresponding to acquisitions used in the paper. You can run any of them by changing the value of the experiment_file variable.

Run the program using the following command,

include(joinpath(Pkg.dir("OCMDemo"),"examples","run.jl"))

or directly from the shell using

julia /path/to/run.jl

The script will produce an m-mode visualization, similar to Figure 5 in the paper. The result will be saved as an image, and should automatically open using your system's default image viewer.

Optimization and parallel processing

The code is optimized for speed, essentially through in-place processing using pre-allocated buffers and Julia's broadcast function, as well as @fastmath and @simd. Using Julias's built in support for parallelism through @parallel however caused more overhead, at least on the Julia versions I've tried. Still, you can play around with it by inserting

OCMDemo.init_workers()
@everywhere using OCMDemo # load module on all workers

at the beginning of run.jl (after using OCMDemo). This will parallelize the reconstruction of each plane (here 2).

Cleaning up

To uninstall the package, as well as delete the downloaded sample data, simply type

Pkg.rm("OCMDemo")

Problems or questions

Please don't hesitate to contact me via email if you are having problems using the code or if you have questions or comments.

Credits

This work was performed with the following co-authors:

• Matthew Toews, École de technologie supérieure, Montreal
• Cheng-Chieh Cheng, Brigham and Women's Hospital, Harvard Medical School, Boston
• Jr-yuan George Chiou, righam and Women's Hospital, Harvard Medical School, Boston
• Chang-Sheng Mei, Soochow University, Taipei
• Lena F. Schaefer, Brigham and Women's Hospital, Harvard Medical School, Boston
• W. Scott Hoge, Brigham and Women's Hospital, Harvard Medical School, Boston
• Benjamin M. Schwartz, Google Inc., New York
• Lawrence P. Panych, Brigham and Women's Hospital, Harvard Medical School, Boston
• Bruno Madore, Brigham and Women's Hospital, Harvard Medical School, Boston