EEG Single-channel minimum electrode montage

mark.o.smark.o.s
edited February 2018 in Other Platforms
Hello,

I am trying to re-design the Cyton, for a specific application which only requires a single differential channel of EEG.
I want to design a board with as little components and as small a footprint as possible.
My main issue is, I hope to only use 2 electrodes to record a differential channel of EEG (for example C3-C4) and I'm not sure if that is possible. 

To test this I used the openBCI Cyton with an arbitrary waveform generator (AWG) outputting +/-50uV EEG signals.
I connected the output of the AWG to 1P. I connected the ground of the AWG to 1N and to Bias (essentially shorting 1N and Bias). 
In the openBCI software, I included Chan1 in Bias and turned off SRB.
It worked perfectly and I got a correlation of 0.99 between the openBCI data and the original AWG data. 

I modified Fig. 72 from the ADS1299 datasheet to show how I intend to setup the channels. 
image
Is this possible? If so why hasn't everyone been doing it? I have some theoretical reasons why it might not work.
Firstly, if I short 1N with Bias, then the - input of the differential amplifier will no longer have the common-mode signals, as they will be inverted and summed with 1N, and cancelled. But 1P will still have them, so I'm effectively ruining the common-mode rejection?
Secondly, ground is usually on the ear as there is no EEG source signals on the ear lobe, only the interference signals (mains, EMI, etc.), which we want to be cancelled out.
Whereas, for example if the signals from C3 and C4 are summed, inverted and injected back into the patient through C4, then interference signals will be cancelled out, but also potentially the EEG signal sources close to the electrode at C4?

Has anybody experimented with such minimalist electrode montages? Is it possible to short ref and bias? 
If not, then is it possible to place 1P at C3 and then separate 1N and bias electrodes very close together at C4? Will I still record the equivalent of C3-C4, without diminishing or corrupting the signal? 

Thanks

Mark

Comments

  • wjcroftwjcroft Mount Shasta, CA
    Mark, hi.

    Joel @biomurph may want to comment on the circuit implications. But I'll add something regarding Ground / Bias placement on the head.

    On clinical grade EEG caps, such as the Electro-cap or the g.tec GAMMAcap, the Ground / Bias is placed on the midline. In the case of Electro-cap at AFz.


    image

    For those EEG systems that use Ground to inject reverse phase mains noise (common mode), called DRL, or driven ground, driven right leg -- the midline placement has a better chance of phase canceling on both hemispheres. So Ground is NOT always placed on an ear lobe or mastoid bone.

    William

  • mark.o.smark.o.s
    edited February 2018
    Hi William,

    Thanks for that. You're right, so that negates my second point mostly. 
    I still wonder, if only 1P and 1N are included in the bias, and they are only 3cm apart (roughly distance between C3-C4), then surely in addition to the offset, 50Hz, etc., there is EEG signals common between both which will be removed by the DRL?

    Having thought a bit more about the first point, maybe if I connect Bias to both the 1P and 1N pins, then the common-mode signals will be cancelled on both + and - channels.
    I have made up a rough schematic of what I am hoping to design.
    Any pointers on where I'm going wrong would be great. I will use a unipolar 5V supply.
    Regarding the RC filter between 1P and 1N, in the ADS1299 datasheet, it says:
    "the filter is advised to be formed by using
    a differential capacitor CFIlt that shunts the inputs rather than individual RC filters whose capacitors shunt to
    ground. The differential capacitor configuration significantly improves common-mode rejection because this
    approach removes dependence on component mismatch."
    This is not done on the openBCI Cyton. Any reason for this? I choose R and C values to give a Fc of 6.7kHz as used in the ADS1299EEG-FE Eval Board. Although they still connect to Agnd as seen in the smaller image below.
    image
    image
  • biomurphbiomurph Brooklyn, NY
    @mark.o.s
    Interesting idea.
    In your schematic, you are connecting the 8N and 8P inputs together with the BIAS, so you will likely not se any signal, since they are tied together.
    Also, you are connecting the inputs together with the capacitor C4. This will likely also cause correlation between the input signals?

    The Cyton does use an RC filter on the inputs. In our schematic, check out R6 and C34 for example. In your design, you would want to have a capacitor on each of the inputs coupled to ground, like the image you included.

    The differential inputs 8N and 8P will be measured against each other, and the difference becomes the signal that channel 8 outputs. If you have these electrodes placed close together on the scalp, then they will likely show minimal signal, since they will share much of the same signal.
    Instead, use 8N and 7N, and also use the SRB2 with 8P and 7P internally connected to the SRB2 pin. Have the SRB2 electrode connected to a part of the head that is away from the 8N and 7N electrodes, and ideally at a place that will not have any potentials riding in it (mastoid, earlobe).
    The SRB2 electrode is the signal that the 8N and 7N will be measured against. You also need the BIAS connected to the subject. I think you can put this almost anywhere, like William says, but don't connect it directly to the electrode inputs.

    It can be useful to use a 'linked ear' reference, if you are measuring EEG from both sides of the head. That will 'balance' the reference between the two hemispheres. In this case two ear clips are connected to the SRB2 pin.
  • mark.o.smark.o.s
    edited May 2018
    @biomurph
    Thank you for your answer.

    As for the differential cap not being shunted to ground, that seems to be the recommended  configuration. It is explained in detail here by a TI employee on the TI forum.

    As for the 2 electrode configuration, the below design was suggested to me. I still don't fully understand how it will work, but it does mean that 8N and 8P are no longer shorted.
    Any help or comments on the design would be very much appreciated.

    Thanks!


    https://e2e.ti.com/resized-image/__size/700x0/__key/communityserver-components-multipleuploadfilemanager/079a2162_2D00_01b9_2D00_402d_2D00_a907_2D00_741417b5e96c-347099-complete/Schematic-ADS1299-2-electrode-V5.png
  • biomurphbiomurph Brooklyn, NY
    @mark.o.s

    Your image did not resolve....
  • mark.o.smark.o.s

    Uploaded below!
    Thanks for the help
    image
  • biomurphbiomurph Brooklyn, NY
    Well, you are coupling both of the inputs with the biasout... I am not sure if that's a good idea. You would want to connect the biasout to the body biasout is the DRL, or driven ground. by connecting it to the inputs, I would imagine everything would be flat.
  • mark.o.smark.o.s
    Think the idea is that the capacitors at + and - will remove the DC component. Then if BiasOut is set to mid-supply and applied to both the inputs through a high impedance resistor, the EEG signals should be driven to the middle of the ADC voltage range.
    That was my understanding from what I was told on the TI forum, but I may be wrong.
    Still not sure what signal BiasInv is inverting. Is it inverting the AC common-mode signals only or is the DC bias which is common between between both inputs and AC common-mode?
    Thanks for your help.
Sign In or Register to comment.

© Vanilla Keystone Theme 2025