The model-free toolbox for short is a simple to use Matlab toolbox for performing model-free analysis of multivariate data sets. Its graphical user interface enables users to easily try out various model-free algorithms, together with additional pre- and postprocessing algorithms and reliability analyses. The design of the toolbox is modular, so it can be easily extended to include your algorithm of choice. It plugs into SPM and can be used to perform model-free analysis of biomedical image time series such as fMRI or PET data sets, but can also be used stand-alone.
The toolbox has been developed by the Computational Intelligence and Machine Learning Group at the Institute of Biophysics, University of Regensburg, Germany. The main contributors are Ingo R. Keck, Peter Gruber and Fabian J. Theis. The software is provided as free software subject to the GNU General Public License. Associated member of the mfBox development-team is Elmar Lang, who greatly helped during planning, development, testing and of course with funding. It also contains contributions by Jean-François Cardoso and Juha Karvanen.
This toolbox is a product of our BMBF-funded projects 'ModKog' and 'Brain Plasticity', where we analyze functional magnetic resonance imaging (fMRI) scans using information-theoretic methods. The toolbox provides, among others, the two novel algorithms mdSOBI and stJADE and stSOBI developed within the 'modkog' project, specifically fit to the structure of biomedical image time series. For more details, please check out Fabian's page on biomedical applications.
If it is to be run from within SPM, just extract the toolbox into the folder 'toolbox'. You should have a subfolder mf_box. The program can be launched from the toolbox dropdown box in SPM. For usage outside of SPM but for fMRI analysis, just add SPM to your path and start the toolbox by starting 'spm_mf_box' within the mfBox-main folder.