Hi,
I am working on artifact removal from scalp EEG. I would like to see the effect of artifact removal on BCI performances. However, since I am completely new in BCI research, I would appreciate any kind of suggestion on this particular area, especially the requirement of artifact removal online for motor-imagery based BCI applications. What types of artifacts are more disturbing in this area, and which spectral band is analyzed for BCI classification?
Any insight on this topic is much appreciated. Thank you in advance.
Regards,
Kafiul
Hi Kafiul,
this pretty much depends on what you want to do. The most essential part is that you are dealing with a *BRAIN*-computer interface, therefore any feature you want to use has to be based on brain activity, and not on peripheral muscles. One big source of contamination are eye movements (including blinks) and facial muscles (including those on the scalp). If you want to go for imagined movement paradigms, take special care that participants are actually performing the imagined movement (i.e. not moving the respective muscles).These things are classically checked for by attaching additional electrodes to the respective muscles, or by using pressure-based pads, which can be more sensitive to micro-movements (there once was a paper about this, I forgot who wrote it). But as said, it all depends on the paradigm you want to use.
The paradigm also defines the spectral band for BCI classification. Most scalp EEG studies are done on imagined movement, which analyze the beta band (around 15-25 Hz; check papers from Pfurtscheller, Wolpaw and McFarland to start with). I myself conducted covert attention studies in the alpha band (around 10 Hz). There are also different (feature selection and classification) methods that can be used here, which brings me to the next point.
There is loads to read about BCI, but it all depends on your area of expertise, e.g. some groups are working on advanced machine learning algorithms to improve BCI (check the Berlin group around Klaus-Robert Muller), others working on integration with robotics (check Jose del R. Millan), others on application in locked-in patients (e.g. Niels Birbaumer) or on innovative paradigms like auditory BCIs (e.g. Benjamin Blankertz), other use fMRI instead of EEG (e.g. Rainer Goebel), or integrate both modalities to study effects on the brain itself (e.g. Tomas Ros) incl infamous neurofeedback applications (e.g. Martijn Arns edit: and John Gruzelier), and many more subodmains. I would propose that you read the method section of some papers to get a feel for what you really want and need. There are different approaches and opinions, and there is no one else but you (and your supervisor and co-authors) to decide what is the right choice for your endeavour. Good luck!
Thank you very much Jörn M. Horschig! Yes, loads to read about BCI. Your reply is informative which triggers me to think about the problem from different angle and more maturely for sure. :)
Did you find the effect of artifact removal on overall performance? Can you suggest some related work now?
- Badges
- Science topic
- Similar topics
- Journal Impact Factor