Regarding Spatial filter and Voxel Visualizer

[ChangeWind]

I am trynig to figure out how to apply this combination of boxes to my EEG data of 64 channels. I have looked into the Voxel-Display.xml Scenario.

Question1:
What configuration I need to define to use it for 64 channels?

The example provided by Openvibe has 60channels.
The Voxel Visualizer box only need the Voxels Filename, consisting of 2394 voxels. Which I believe is Head Mesh specific and I do not need to change it.

The critical configuration is in the Spatial Filter.
Of the 3 parameters, the No of Input and output channels will be 64 each (am I right?).

Question2:
The question now is how do I define the Spatial filter coefficients?

I have read the doc on Spatial filter, the example is for Laplace value determination. I am not sure what I need to do to provide the coefficients for 64 channels to obtain an inverse solution needed for VoxelDisplay.

Question3: How are the values in the "Voxel-Display-settings-file.txt" provided by OpenVIBE calculated for 60 channels.

Could the OpenVIBE developer provide some hints and tips?

Thanks

CW.

[yrenard]

Dear ChangeWind,

Question1:
What configuration I need to define to use it for 64 channels?

The example provided by Openvibe has 60channels.
The Voxel Visualizer box only need the Voxels Filename, consisting of 2394 voxels. Which I believe is Head Mesh specific and I do not need to change it.

The critical configuration is in the Spatial Filter.
Of the 3 parameters, the No of Input and output channels will be 64 each (am I right?).

You are right, it should be 64.

Question2:
The question now is how do I define the Spatial filter coefficients?

I have read the doc on Spatial filter, the example is for Laplace value determination. I am not sure what I need to do to provide the coefficients for 64 channels to obtain an inverse solution needed for VoxelDisplay.

The coefficients are computed with an inverse model (e.g. sLORETA, see http://www.uzh.ch/keyinst/loreta.htm).

Question3: How are the values in the "Voxel-Display-settings-file.txt" provided by OpenVIBE calculated for 60 channels.

Could the OpenVIBE developer provide some hints and tips?

We do not provide any tool to compute these coefficients. Either you have the possibility to compute them yourself using an inverse model such as sLORETA, either you use the exact 60 channels we use in the sample scenario.

Hope this helps,
Best regards,
Yann

[ChangeWind]

Hi Yrenard,
After some days of studying sLoreata as you have suggested. I have the following feedbacks:

@VoxelDisplay is one of the features that makes OpenVIBE Very unique. It will help user to have more access to the box if the following information is provided.

A) The computation of the spatial filter coeficient "specific" for the VoxelDisplay depends on

(i) leadfield matrix which in turn depends on the grid of 2394 voxels used in VoxelDisplay. Can I assume that this grid is "Fixed" for the VoxelDisplay or one could load a different set of grids compatible in coordinate scale to the one used in current VoxelDisplay?

(ii) Hence, one needs to input the grid information to SLoreata to compute the coefficient. The "standard" way to use SLoreta is to provide sensor info to generate the transformation matrix using the sLoreta's internal voxel grid

(iii) I am not sure if the sLoreta grid is compatible with the OpenVIBE's 2394 grid. If not, the next question is could one use the SLoreta's grid (see A(i) ) in VoxelDisplay.

(iv)Therefore it is not going to be straight forward to hack SLoreta to generate the coefficient used in OpenVIBE spatial filter.

(B)Alternative One:
Restrict oneself to use only the 60 Channels provided.
Could the team provide the Labels of the 60 channels used in the example. So the user can perhap select those channels out of 64 that match in (x,y,z) to the 60 channels for feeding into OpenVIBE.

(C)Alternative two.
The team supprise us with a Christmas gift:
(i) For each standard set of channel configuration based on 10-20 system, provide the precomputated coefficients for 32, 64 and perhaps 128 channels and make them available in the Share folder so we could use VoxelDisplay other than the provided 60 channels.

These are wish list for future development of OpenVIBE.

[yrenard]

Dear ChangeWind,

thank you very much for your feedbacks about the VoxelDisplay. I'm glad that you find it a usefull feature !

I will try to answer your questions as precisely as possible...

A) (i) The spatial voxel localization are contained in the file ${Path_Root}/share/openvibe-plugins/simple-visualisation/voxeldisplay/voxel_set_2394.txt which is a parameter of the Voxel Display box (as shown in the sample scenario). So no, this is not fixed position for the Voxel Display, it can be whatever you want as soon as you have the preceeding spatial filter that computes the correct values for those positions.

A) (ii) You have the grid.

A) (iii) You probably can. Check the matrix file format and write your own with whatever voxel positions you prefer.

B) I did not do this tutorial myself, I will try to find this back for you. Without answer by the end of next week, please bump me up.

C) This could be a nice feature request, please fill in a new entry in the bug tracker about it. Unfortunately, no christmas gift for this year except the upcoming 0.5.0 release. We can't be on all the aspects at once. It surely would be a nice feature but we won't do it in the short term. I hope you understand.

Thank you again for your feedbacks, we need them !

Yann

[yrenard]

Dear Changewind,

I am definitely late to answer your request. However, here are the information about the file we used for the voxel display box tutorial. It is the first minute of the record of subject 1 in the IIIa dataset of bci competition III. The full file can be downloaded from their website as soon as you have a login available. More information can be found on the appropriate page. If you don't have a login yet, feel free to create one from the dedicated page

According to their description, the channels are placed as described in the attached picture. I hope this helps. We'll see if we can release a matrix for standard location.

Best regards,
Yann