EEG results and subject movement

Hi all,

I am looking into EEG as a possible solution for a particular problem where the subject connected to the EEG sensors will be moving (walking, running, perhaps rolling around), needless to say in my mind the sensors will have to very well secured in place on the scalp to get any degree of success.

I assume I am safe to infer from examples I have seen of an EEG's being performed that stillness is desired, but: 
  • Is stillness a requirement? 
  • Does anyone know what effect physical movement of the subject can/does have on the EEG results?
  • I assume artefacts would be created, do they clutter the results beyond repair?
  • Can anyone point me to texts online that discuss this topic as I cannot find any?
Thanks in advance, hopefully this is not a completely pointless set of questions. Also apologies if this is not quite the right place for this post.

Sam

Comments

  • wjcroftwjcroft Mount Shasta, CA
    Sam, hi.


    https://www.google.com/search?q=eeg+while+walking

    Yes, this can be done, especially with the right kind of tight fitting cap and very short electrode wires that do not move or swing around during movement. The signal processing involved however is fairly sophisticated. You might check some of the clinical research papers on those links for leads on what algorithms are employed.

    There are limits to what artifacts can be removed. Running and rolling around might be beyond what is possible. Consider that simple eye blinks with stationary sitting EEG are significant artifacts. As well as scalp motion, head and jaw muscle activations, etc.

    Regards, William

  • Thank you so much for the response Will.
    That's enough at least to justify some research!

    Regards,
    Sam
  • Hi Sam,

    I'm pretty much hardware-only, but I'll add some.  William's comments make all the sense in the world to me, with short wires, minimal wire-movement, minimal variation of forces on electrodes, etc, plus wires should be snug to the cap, and with as little as possible "loop area" as any two differential electrode signals make their way from electrodes to electronics.  Cyton & Ganglion already help a lot just by enabling head-mounting which I hope you're able to do somehow, and with battery power, and with wireless signaling.


    If you find that's not sufficient for what you need, then it's worth considering whether the motion-artifact is due to something coupling in via differential-mode, or coupling in via common-mode.  For sure there must be better descriptions on the web but I don't happen to know where, and there are descriptions that I think are much worse, so I'll describe.  If you already understand this you can skip it, but if you don't, you might want to clarify by drawing yourself some sketches or something.

    -- An example of differential-mode coupling would be this.  The electrode wires going from the electrodes to the + and - inputs of a channel of course have some separation between them, the 2 wires can't occupy the same place in space at the same time, not in the universe I'm kinda familiar with.  So there's some "enclosed area" formed by their separation, often referred to as loop area.  If there's a magnetic field that has "entered" into that loop area, then that's fine, won't cause a voltage at the A/D inputs, as long as the magnetic field isn't time-varying.  But if the magnitude of the coupled field is varying in time ... whether the field itself is varying, or the subject is moving to a different location/orientation such that the field magnitude in the loop area changes ... then a voltage will be generated.  Generally the solutions would be to reduce the field itself, and/or to pay attention to that loop-area (shorten the specific wires in question, bring them close together, even much-better is to twist them together because you get a cancelling effect, etc.).

    -- In contrast, common-mode coupling is when, instead of something different getting onto the + wire than onto the - wire, there's some mechanism that causes the same voltage to be presented to both electrodes sites (or to both electrode wires or to wherever the offending signal is coupling in).  Being common to both electrode-sites, we cleverly give it the name common mode.  An example of common mode interference is when line voltage wiring might be near, even inside a wall meters away from, the subject/patient.  The patient has some capacitance to that voltage, and let's also say has some capacitance to earth ground, so let's say there's a divider effect such that the subject is at 10 volts peak-to-peak relative to earth ground.  Ideally, the Ganglion's or Cyton's A/D ground will also be at the same 10 volts peak-to-peak relative to earth ground, but let's say that A/D ground is only "99%" of the same as the patient, so it's 9.9 volts peak-to-peak relative to earth; then it might be that there's 0.1 volt peak-to-peak between patient-average-potential and A/D ground.  Then yow, your wish is to get noise down to maybe 1's or 10's of microvolts so you can see 100-microvolt difference-signals between + and - electrode sites; but the + and - electrode sites additionally have 100,000 microvolts common-mode on them.  You might say, not a problem, because hardware or software will take the difference between + and -, and since common mode is the same on both, voila, the difference will be zero.  But the problem with common mode voltages is that as you can imagine, it takes very little asymmetry between + and -, on the way from the electrode-sites to where they're digitized signals, for a difference (a differential-mode signal) to be formed.  In our example, if there's 1% asymmetry, so to speak, the resulting differential-mode signal might be 1,000 microvolts.
    In the case of patient motion, the patient's "DC" (low-frequency) voltage to earth may change due to 'static electricity' for example, especially as you mentioned rolling-around.  At one location & time the patient may be at +30V relative to earth, and after some motion maybe +60V.  Some of the above line-voltage discussion doesn't hold water, but some does, and you can get motion-artifact that way too.

    In addition to avoiding asymmetry, hardware designers have other ways to accomplish common mode rejection.  One way is for the board electronics to include an amplifier circuit, with the output going out to the patient, and with the patient in the amplifier's feedback path.  I believe it was first used with ECG (my specialty), and the amplifier's output went to the patient's right leg, so it became known as a right leg drive (RLD) circuit.  I won't describe how it works or how to choose topology & component values, I see lots of hits googling "eeg common mode rejection active drive", I don't know how readable & by what audience, but likely lots of good info.  

    Specific to openbci,

    ** If you're using Ganglion .... it doesn't have active drive.  I remember reading that testing was done and it was found to be unnecessary.  But of course that can depend on test method & how well it matches the application.  Ganglion does benefit from having connection to the D_G signal, which I gather would-have-meant Driven_Ground, and it still is driving ground, but it will have less common mode rejection because it doesn't have the patient in a feedback loop.  If you want, there's some chance that some open bci forum people like myself could help you figure out how to modify a board to increase common mode rejection.  Among other things it would be good to know how many electrode wires you're using and whether you would plan to always use those.

    ** I'm less familiar with Cyton but am looking at a .jpg schematic I squirreled away from github several months ago, and at the spec for the ADS1299 chip.  I see that Cyton does use active drive & feedback, by way of the "BIAS" pin on the Input-Connector.  The circuit topology threw me at first, it's different than most I've used or seen, and I thought, could it be a circuit mistake? .... but I can see some reasoning behind why it basically works, not to say that it's optimal or even stable in some conditions ..... to understand better, I would want to study it carefully and discuss with maybe the Cyton board designers or others who understand RLD (but I still can't help wonder whether it could be a circuit mistake that has worked well-enough).  In any case, again if you want, I'm among the people who might be able to help you optimize the circuit and increase CMR.  Again, it would be important to know which electrode wires you're using and also in this case, whether you plan to change the ADS1299 Programmable Gain Amplifiers' (PGAs') gains or whether you can stay with constant gain.


    Depending on your background, much of the above may be gobbledy-gook, so if you think it might help and you have access to someone who has such a background, it might be good to get with them to explain it.  And/or, submit more here and maybe some of us can help.  Good luck!

         -- Bruce P.

Sign In or Register to comment.