Installation

cvu is hosted on pypi (the python packages index) and github.

Dependencies

cvu has the following dependencies:

• numpy/scipy
• wxpython
• traits
• traitsui (and pyface, enable)
• matplotlib
• mayavi (and tvtk)
• chaco
• bctpy
• mne-python
• nibabel (for locating subcortical structures)

Other than bctpy and mne-python, these packages should largely be available from most scientific python distributions including canopy/epd, anaconda, and python(x,y), though note that python(x,y) is only available for windows. Note that the free version of canopy does not include mayavi, and you may have to use canopy's package manager to login and manually install it (or alternately, build mayavi yourself). However the academic version of canopy, available for free to academics with a .edu email address, does include mayavi.

Previously, I recommended the scientific python distribution canopy. The reason I recommended canopy is because anaconda does not ship with wxpython and I originally wrote for the wx python. However I have done more work getting cvu to work with the Qt backend and fixing some of the minor issues, such as with file dialog window which is clunky. So now I recommend to use either one.

If you want to use the Qt backend on canopy, you may need to set the environment variable ETS_TOOLKIT='qt4'. On anaconda, it is necessary to install some packages using conda, including mayavi and pyqt4/pyside.

bctpy, nibabel and mne-python are more specialized tools that do not come with typical scientific python distributions and you may have to install them yourself. mne and bctpy can be easily installed by using easy-install or pip (pip install mne; pip install bctpy; 'pip install nibabel').

nibabel is only needed for the definition of ROIs from subcortical structures. If your data does not use subcortical ROIs, you do not need nibabel.

Distribution

You can download cvu with tools such as pip and easy-install. These tools will install the package into your python installation. Equivalently, you can download the source (which is in pure python) and run the setup script (i.e. python setup.py install) On UNIX systems, this will create a scripts directory which will be added to your $PATH. You can execute the command cvu in a terminal. Note that if you do not want to install the package to your python installation (perhaps because your python installation is shared and you do not have permissions), as long as you have the requisite libraries in your python distribution cvu will work just fine when distributed as a "dumb" installation (because it is pure python) where the launcher is executed directly from your home directory or somewhere else.

On windows, the launcher script is instead added to the scripts directory of your python installation, which is considerably less useful. I am looking into providing a Windows installer to facilitate installation on Windows.

Troubleshooting

If on OS X, you receive a message about "cannot find the screen, use pythonw", then in spite of my efforts to fix this, you can do one of several things:

• Download the development version from github.
• Navigate to the cvu directory in your python installation, and edit the launcher script to use pythonw instead of python
• Navigate to the cvu directory in your python installation, and invoke cvu by calling pythonw main.py. Automated scripting and command line arguments will not work.