Generating a data stream for inter-voxel connectivity

[jsilberman]

I'd like to be able to do the following, all in real time.

1. Acquire 19 channel EEG data.

2. Apply a LORETA algorithm (preferably eLORETA, but sLORETA or LORETA would also work) to these data.

3. Extract data reflecting current densities in two specific voxels of interest.

4. Compute a measure of connectivity (coherence would work) between these two voxels of interest.

Does anybody have recommendations regarding how I can use OpenViBE to create a software tool that will do this? Where should I start??

Any advice would be greatly appreciated.

Jordan

[yrenard]

Dear jsilberman,

thank you for your interest in OpenViBE and welcome on this forum.

The first step in order to achieve your goal is to generate a matrix that implements the xLORETA source localization. We don't provide this tool so you will have to generate the matrix yourself. If you are familiar with loretakey, then you can get this matrix from loretakey and convert it to an OpenViBE usable matrix thanks to Marco Congedo's OpenViBE tools.

Once you have this matrix available, then you will have to :

• filter the signal in whatever frequency band you want
• spatial filter the signal with the previously computed matrix
• channel select the channels that corresponds to the voxels you are interested in

The is nothing to compute connectivity or coherence right now (if you want to contribute such functionality, please join on the IRC channel so we can have a discussion).

The best you could do right now is compute the variance (using the Univariate Statistics box) / covariance (using a mix of Simple DSP and Epoch Average boxes) and see what you can get from this.

I hope this answers your question ; have fun with OpenViBE
Yann

[mcongedo]

As a complement to Yann's directions, remember that the sLORETA or eLORETA transfer matrix in the LORETA-Key space will output the (x, y, z) current triplet for each voxel (x1, y1, z1, x2, y2, z2, ..., x2349, y2349, z2349), for the 2349 voxels embedded in that head model. So, before you do anything else you will have to derive the current density as the sum of the squares of the x, y, and z component for each voxel.

Best,
Marco