Installation#

Pip#

CMRsim can be installed using the python package manager pip. The package can be found in the registry of the repository hosted on the ETH-Gitlab instance

Versioning: Stable release versions are denoted with the ususal major.minor notation. Additionally, development versions are constantly updated on solved issued and can be installed using the major.minor.devX notation

Installation in command line:

pip install cmrsim --extra-index-url https://__token__:<your_personal_token>@gitlab.ethz.ch/api/v4/projects/23798/packages/pypi/simple

Requirements#

CMRsim does rely on functionality of 3 non-standard python stack packages:

  1. TensorFlow: Pretty much all computationally heavy calculations.

  2. Pyvista: For processing and visualizing graphs (e.g. 3D Velocity fields).

  3. CMRseq: For defining MR-sequences used for Fourier-Encoding calculation and Bloch-simulation.

Further (already indirectly included) requirements are: -h5py~=2.10.0 -scipy>=1.4.1 -tqdm

Docker#

For each major release a set of docker images are compiled and published in the container registry associated with repository. Within the docker images, all requirements are installed for simulation deployment.

Furthermore, there are three image-tags available. To use the pull the docker image use (e.g. for jupyter tag):

docker pull registry.ethz.ch/jweine/cmrsim/jupyter:latest

CPU#

Based on the Tensorflow:…-cpu:latest docker image. Containing all cmrsim dependencies and cmrsim

GPU#

Based on the Tensorflow:…-gpu:latest docker image. Containing all cmrsim dependencies and cmrsim

Jupyter#

Based on the gpu:latest tag, but also includes an installation of jupyter-lab for interactive simulation / rendering.