Python

The scientific python ecosystem is also available on Aalto Linux workstations, including the anaconda (Python 3) and anaconda2 (Python2) modules providing the Anaconda python distribution. For a more indepth description see the generic python page under scientific computing docs.

The “neuroimaging” environment

On the Aalto Linux workstations there exists a conda environment under the anaconda3 module called “neuroimaging” which contains an extensive collection of Python packages for the analysis of neuroimaging data, such as fMRI, EEG and MEG.

To use it:

$ ml purge
$ ml anaconda3
$ source activate neuroimaging

A more up-to-date yet under development version of the environment can be activated with:

$ ml purge
$ ml anaconda3
$ source activate neuroimaging2

To see the full list of packages what are installed in the environment, use:

$ conda list

Some highlights include:

  • Basic scientific stack
    • numpy
    • scipy
    • matplotlib
    • pandas
    • statsmodels
  • fMRI:
    • nibabel
    • nilearn
    • nitime
    • pysurfer
  • EEG/MEG:
    • mne
    • pysurfer
  • Machine learning:
    • scikit-learn
    • tensorflow
    • pytorch
  • R:
    • rpy2 (bridge between Python and R)
    • tidyverse

Finally, if you get binaries from the wrong environment (check with which BINARYNAME) you may need to update the mappings with:

$ rehash

MNE Analyze

Note: this was tested only for NBE workstations. If you wish to run mne_analyze from your workstation you should follow this procedure. Open a new terminal and make sure you have the bash shell (echo $SHELL, if you do not have it, just type bash) and then:

$ module load mne
$ source /work/modules/Ubuntu/14.04/amd64/common/mne/MNE-2.7.4-3434-Linux-x86_64/bin/mne_setup_sh
$ export SUBJECTS_DIR=PATHTOSUBJECTFOLDER
$ export SUBJECT=SUBJECTID
$ mne_analyze

Please note that the path of the “source” command might change with most up to date versions of the tool. Please note that the “PATHTOSUBJECTFOLDER” and “SUBJECTID” are specific to the data you have. Please refer to MNE documentation for more help on these.

Mayavi

If you experience problems with the 3D visualizations that use Mayavi (for example MNE-Python’s brain plots), you can try forcing the graphics backend to Qt5:

  • For the Spyder IDE, set Tools -> Preferences -> Ipython console -> Graphics -> Backend: Qt5
  • For the ipython consoles, append c.InteractiveShellApp.matplotlib = 'qt5' to the ipython_config.py and ipython_kernel_config.py configuration files. By default, these can be found in ~/.ipython/profile/default/.
  • In Jupyter notebooks, execute the magic command %matplotlib qt5 at the beginning of your notebook.

Installation of additional packages

The “neuroimaging” environment aims to provide everything you need for the analysis of neuroimaging data. If you feel a package is missing that may be useful for others as well, contact Marijn van Vliet. To quickly install a package in your home folder, use pip install <package-name> --user.

Notes for admins

The trick to getting Mayavi to play nicely with a modern Python environment is to install it from Git:

$ ml purge
$ ml anaconda3
$ source activate neuroimaging
$ pip install git+https://github.com/enthought/traits.git@a7a83182048c08923953e302658b51b68c802132
$ pip install git+https://github.com/enthought/pyface.git@13a064de48adda3c880350545717d8cf8929afad
$ pip install git+https://github.com/enthought/traitsui.git@ee8ef0a34dfc1db18a8e2c0301cc18d96b7a3e2f
$ pip install git+https://github.com/enthought/mayavi.git