Multi
Voxel Pattern Classification Toolbox
The
Multi
Voxel
Pattern
Classification toolbox (
MVPC toolbox) is a MATLAB toolbox to
facilitate multi-voxel pattenr classification analysis on fMRI data.
MVPC toolbox was originally developed by Satoshi Hirose and Isao Nambu.
What can we do with this
toolbox ?
You can perform below analyses with this toolbox.
1) Preprocessings including
•Conversion from DICOM files to NIFTI files,
•Usual preprocessings
for fMRI data (slice timing correction,
co-registration, normalization etc.)
•GLM univariate analysis
2) Voxel and volume selection for decoding.
3) fMRI decoding by using various algorithms.
4) Leave-one-session-out and leave-two-session-out cross validation for
performance evaluation and adjustment of (a) hyper-parameter(s).
5) save results in graphs.
Algorithms
· Sparse Logistic Regression (SLR; Yamashita et al., 2008)
· iSLR (Hirose et al., submitted)
· L1-norm reguralized sparse logistic regresssion
· Elastic Net (Zou and Hastie, 2004)
· Support Vector Machine
· SVM with Recursive Feature Elimination (Rakotomamonjy, 2003)
External Toolboxes
Machine learning
implementation: We rely on the following toolboxes for algorithm
implementaion. Please download and install them at their webpage linked
below.
·
Sparse
Logistic Regression Toolbox for SLR
·
GLMnet in
MATLAB for L1-norm reguralized sparse logistic regresssion and
Elastic Net
·
LibSVM for
SVM
Preprocessing, univariate analysis,
ROI definition and extraction: We rely on SPM5, and its
extensions, Volumes Toolbox and Anatomy Toolbox for the processes.
·
SPM5
·
Volumes Toolbox
·
Anatomy
Toolbox
Download
· MVPC toolbox (version 1.0; 36KB)
Download !!
· Tutorial (version 1; 662MB)
Download !!
Copyright
MVPC toolbox is free but copyrighted software, distributed under the
terms of the GNU General Public Licence. Further details on copyleft
can be found at http://www.gnu.org/copyleft/.
Feedback and bug report
Any feedback and bug reports are welcome. Please mail to satoshi.hirose
[at] nict.go.jp (please replace the [at] with the '@' symbol).
Future Update
· Weight-based functional mapping
· Decoding from Beta-map (or other statistical values)
· GUI for setting parameters
· Preprocessing with recent version of SPM
· Readme for parallel processing
· Readme for clsfy
· Smaller version of Tutorial