BrainSuite 2013 Statistics Toolbox (bss)

The BrainSuite statistics toolbox allows the application of advanced statistical models to surface and curve based outputs generated from BrainSuite. This enables population or group modeling of cortical or sulcal morphology. Some features of the toolbox are:

• a python interface for manipulating morphological data using R through Rpy2
• a python interface for using statsmodels with pandas for a pure python implementation
• Ability to plot graphs, charts and visualizations on surfaces (coming soon)

Download

Currently download from https://bitbucket.org/bmapdev/bss/

Requirements (short version)

• R - built as a library with –enable-R-shlib (Default on Windows and Mac OSX)
• Python 2.7
• Base Python packages - virtualenv, pip, numpy, scipy > ##### NOTE: Alternately, install Canopy python. Makes life much easier.
• Add on Python packages - pandas, statsmodels, rpy2

Requirements (long version)

To stick to a pure python execution, one could skip installing R, and Rpy2. However, in our benchmarks, Rpy2+R currently shows the best performance.

Thus, to get all the functionality of the toolbox, it is recommended to install the following software on your computer.

• Working python installation (We recommend version 2.7) -
  with the following packages:
  • virtualenv, pip, numpy, scipy
    

NOTE: Alternately, you could download Canopy express, a comprehensive python framework geared towards scientifc computing, available for free.

• R - R software for statistical computing (built as a library with –enable-R-shlib)

  This is by default on Mac OSX and Windows.

  On Linux, for e.g. in the root directory of the R source, one can do:

  sudo ./configure  --enable-R-shlib  
  sudo make  
  sudo make install  

• Rpy2 - python interface to R

• statsmodels - pure python module for statistical inference and much more

Installation for Mac OSX/Linux - Approach 1

To be followed if all the requirements above are satisfied.

It is recommended to create a virtual python environment in your BrainSuite directory.

For e.g., if your BrainSuite installation is located at /Applications/Brainsuite13a, and the unzipped bss package is located at ~/bss-0.1dev, then open the terminal and type

virtualenv --system-site-packages /Applications/Brainsuite13a/bssenv
/Applications/Brainsuite13a/bssenv/bin/python ~/bss-0.1dev/setup.py install

Installation for Mac OSX/Linux - Approach 2

To be developed.

Bootstrap everything. Start with a basic python installation

Create a virtualenv in the BrainSuite directory

Install all the packages in virtualenv

Install the bss package in virtualenv

Example for running ANOVA for model comparison

The source package includes a test directory with sample data. Assuming the unzipped source directory is located at: ~/bmapdev-bss, open the file

~/bmapdev-bss/bss/test/data/sample1/modelspec.ini

and change the respective paths for the variables

subjectdir, demographics, phenotype_attribute_matrix, atlas_surface

to correspond to the paths on your file system. Then type,

/Applications/Brainsuite13a/bssenv/bin/bss_model.py -modelspec
~/bmapdev-bss/bss/test/data/sample1/modelspec.ini -outdir ~/ -statsengine R

License

BrainSuite/MIT