looking for OpenViBE users in North America for assistance developing an EEG neurofeedback scenario

[KJeffrey]

Community:

We are first-time users of OpenViBE and are having a few difficulties with implementing a neurofeedback scenario. Reduced to essentials, our protocol is as follows:

Stage 1:
a. Present a randomly selected picture (without replacement) from two categories for 1 second with a 1 second blank screen following. There are 150 pictures in each category, 300 pictures total per run. One category is designed to induce a stage of agitation, and the other category has neutral pictures.
b. Collect continuous EEG and train a classifier unique to each individual participant to differentiate between the two categories based solely on the EEG. It is presumed that each category will have somewhat different EEG signatures that will be picked up by the classifier and differentiated. Previous work we are trying to replicate demonstrated a 70% rate of correct classification, and we think we can obtain 90% classification with our modifications.

Stage 2:
Present the same set of pictures (in different random order) and perform the above classification after each presentation to feed back to the participant in a visual display. The participant's task is to suppress the difference in their EEG caused by the agitation-inducing pictures, and thus make their EEG to those pictures look more like the EEG to the neutral pictures.

In stage 2, our processing pipeline consists of doing this for 1 second of 64 channel EEG at 500 fps after each picture is displayed
a. bandpass and notch filter
b. remove eyeblinks
c. average reference
d. continuous wavelet transform
e. feature extraction (select subset of significant features identified in the classifier training)
f. classify and modify feedback to participant

The above steps have to be competed in the 1 second between the finish of each picture and the display of the next. Does anyone out there have experience with doing something like this? Are we being too ambitious? The work we are trying to replicate claimed to do this in 2017 and earlier using a NeuroScan 64-channel EEG system and Matlab based BCI tools.

We have had trouble testing this protocol. It has proven difficult to pre-record data and try to run it through the scenario and review the results. You may review our problems in our previous threads by searching for my last name Eriksen.

We may be willing to pay for assistance if anyone thinks they would be able to help us in some way. Please email me if you have any comments or suggestions.

We are really hoping for help from someone in North America so the time difference is not so onerous.

-Jeff Eriksen
jeriksen@downeurobiology.org
Dow Neurobiology Laboratory
Legacy Research Institute
Portland, OR
USA

[KJeffrey]

previous support threads

OpenViBE threads.png (28.32 KiB) Viewed 15575 times