Skip to content

A solver of EEG/MEG inverse problem using a multivariate auto-regressive model on the source space

Notifications You must be signed in to change notification settings

BBELAOUCHA/iSDR_p

Repository files navigation

iSDR_p (iterative source and dynamics reconstruction)

A solver of EEG/MEG inverse problem using a multivariate auto-regressive model (MVAR) on the source space

picture

iSDR_p is a C++/Python package for solving the EEG/MEG inverse problem using structural/functional prior on the causality between brain regions/sources. iSDR_p solve the following functional:

to obtain the brain activity with an initial MAR model A_i's. Then, brain regions/sources interactions are obtained by optimizing the following cost function:

Where:

 * A_i: i=1,..,p are the matrices of the MVAR model of order p.

 * M_t: EEG or/and MEG measurements at time t.

 * G: Lead field matrix which project brain activity into sensor space.

 * J_{t-i}: Brain activity (distributed source model with fixed position) at time t-i.

 * alpha ]0,100[: percentage of the maximum alpha_max (reg parameter which results in zero brain activity).

Requirements

1-MATIO > 1.5.2 2-HDF5 3-FLENS 4-MKL intel 5-CMake

Cite

(1) Brahim Belaoucha, Théodore Papadopoulo. Large brain effective network from EEG/MEG data and dMR information. PRNI 2017 – 7th International Workshop on Pattern Recognition in NeuroImaging, Jun 2017, Toronto, Canada.

(2) Brahim Belaoucha, Mouloud Kachouane, Théodore Papadopoulo. Multivariate Autoregressive Model Constrained by Anatomical Connectivity to Reconstruct Focal Sources. 2016 38th Annual International Conference of the IEEE Engineering in Medicine and Biology Society (EMBC), Aug 2016, Orlando, United States. 2016.

Author

Belaoucha Brahim Theodore Papadopoulo

Installation

Assuming your are in the source directory mkdir build cmake -DMKL_ROOT_DIR=YOUR_MKL_INSTALL_DIR -Dmatio_DIR=YOUR_MATIO_INSTALL_DIR -DFLENS_INCLUDE_DIR=YOUR_FLENS_INSTALL_DIR .. make

Details

More details about the package input parameters can be obtained by:

iSDR_p --help (-h).

The first argument of ./iSDR_p is a .mat file containing the following:

* M: measurements.

* G: Lead field matrix.

* GA: lead field matrix times initial values of A_i's, i=1,..,p i.e. GA = [GA1,..,GAp].

* SC: structural connectivity matrix (symmetric).

* n_c: size of sensor space.

* n_t: number of measurements samples.

* n_s: size of source space.

* m_p: order of MVAR.

Example

An example of iSDR can be seeing in the examples folder.

Test

To test the iSDR_p package, you can run the test module:

test_MxNE_iSDR

To use it in Python

Now, it is possible to use iSDR in Python to reconstruct the brain activity. Go to ./python_wrapper for more details on how to use the package on python