suggestions for BCI addressing dystonic cerebral palsy ?
marto
London
Just a bit of background. I'm totally new to OpenBCI but it looks really exciting. I'm a developer so links to code samples / tutorials will be great.
What I want to used it for?
My son, who is 10, has dystonic cerebral palsy and currently communicates with an AAC device running a software called Grid 3. Basically, it is a grid of cells that have different word and he selects the one he wants to say and the computer reads out what he has selected. He uses 2 manual buttons to move the cell selection and select. It is a pretty basic interface. The issue comes because he has dystonia which makes it very difficult to move his hand without engaging most muscles in his body. It means that he can only use his communication device for short periods of time and with assistance from an adult. We are currently looking for alternative ways for him to use his communication device.
Is it possible to use the Open BCI hardware to detect 2 different thought patterns and replace the switches we currently use with an Open BCI headset.
Which started kit should we use to get started? Do we need all the 16 electrodes? Could we start with 8? Are there any tutorials where people have trained themselves to move a virtual object or even perform different actions.
I watched a presentation where eye blinks were used but I don't think my son could reliably do these.
Thank you in advance.
Comments
Marto, hi.
You mention, "...he has dystonia which makes it very difficult to move his hand without engaging most muscles in his body..."
This would seem to imply that moving other parts of his body: feet, arms, head (up, down, left right) -- also have the same kind of issue with "engaging most muscles in his body", is that true? My initial thought was that it might be possible to attach EMG electromyography muscle sensors to other areas of his body where he has better control. But your comment above would probably rule out that possibility.
Take a look at this website, a research company in Holland (NL Netherlands), is almost ready to release their very impressive cVEP (code based Visual Evoked Potentials) device. The way this works is that your son would watch the computer screen, using his 'attention' to focus on particular letters or buttons. The BCI then picks up this attention signal and presses that button.
https://www.mindaffect.nl/
They are coordinating with the OpenBCI team in Brooklyn, to make this device available to OpenBCI users. It can even work with the lower priced Ganglion board.
I'm not sure what the pricing or time-frame will be. But I'm sure they are targeting their technology, to just your type of situation. The demos above are of their keyboard / typing application. But I've seen other companies in this arena developing icon based simpler interfaces as well. Such as g.tec's IntendiX,
http://www.biosignal.at/Intendix/
My hunch is that the MindAffect Speller system will be many times more affordable than the g.tec solution. Due to the nature of the type of BCI used. MindAffect uses cVEP technology, g.tec Speller instead uses a more difficult and involved tech called ERP / P300, Event Related Potentials. Many advances have been made in cVEP systems in recent years. This will bring costs down.
The sections listed on the IntendiX page have three choices, one is the Speller (which uses ERP), the other is listed as,
intendiX®SOCI — Screen Overlay Control Interface
This indeed is SSVEP, not ERP.
Best regards, William
I'd suggest using the contact and newsletter signup links on the MindAffect website, so that you are informed of their release plans. You can also say that you would be using OpenBCI.
Their Fall 2019 newsletter,
https://mailchi.mp/dc1224b3bf97/newsletter-august-438327
Another type of VEP BCI, is called SSVEP, Steady State Visual Evoked Potentials. In this case the object(s) on the screen are each flashing at a fixed frequency. This is a simpler concept than cVEP, but also limits the operator to fewer selection buttons. But in the case of your "Grid 3" device with only 2 buttons, it should be able to emulate that. You can read more on this search query,
https://www.google.com/search?as_q=ssvep+widget&as_sitesearch=openbci.com
https://www.google.com/search?as_q=ssvep&as_sitesearch=openbci.com
There is also an SSVEP widget under development and testing, that runs inside of the OpenBCI GUI. Richard @retiutut knows the status. Many other posts / threads on SSVEP are at the above search results.
https://github.com/OpenBCI/OpenBCI_GUI/blob/master/OpenBCI_GUI/W_SSVEP.pde
My hunch is that SSVEP might be something you initially experiment with, until the MindAffect system is available. However once the cVEP system is running, it should give expanded possibilities for typed language interactions.
We are waiting to have a neuroscientist review the SSVEP widget algorithm in the OpenBCI GUI. The Networking widget sends out a normalized output from 0 to 1 for up to four SSVEPs.
I would like to add a feature to the SSVEP widget to allow it to open in a separate window that can be made fullscreen on a second monitor.
This is Chip Audette's EEGHacker SSVEP post from 2014. There are many more SSVEP related threads on the forum here, at the above search links.
http://eeghacker.blogspot.com/2014/06/controlling-hex-bug-with-my-brain-waves.html
To generate the precise timing flashes needed, Chip actually used pre-recorded video files that he generated, played back by the operating system video player. The reason for this is that he found he could get more precise flash timing pulses with the movie file, rather than trying to generate flash pulses programmatically from his program. I'm not clear if he was using Processing or Matlab. Many of his file extensions on his Github are '.m' which is a Matlab script.
https://github.com/chipaudette/EEGHacker/tree/master/Data
@wjcroft and @retiutut thank you so much for the info. I've subscribed to the MindAffect newsletter. What they have looks very promising and I will contact them to see if they can help.
I've been watching openbci videos on youtube and also saw that you can use other muscles so I've been watching Sasha (my son) and asking him whether he can voluntarily move any other muscle without engaging most of his body. I think it might work with his cheek. I guess it will be a matter of experimenting and also once he sees something works it will motivate him to work on it.
SSVEP sounds exciting as well. Currently he is working with 2 buttons so the possibility of having 4 inputs is exciting and could make communication much faster for him.
About the hardware. I was looking at the ultra cortex headset but I don't think it will work because Sasha sits in a wheelchair and has headrest behind him which will make it really uncomfortable to wear the hat. I'm thinking of getting the following:
Do you guys think that will be enough to get started? Really appreciate the help
Marto, hi.
Sounds good. Note that the headband used by MindAffect is not the same as the one sold by OpenBCI, although there may be similarities. Generally stronger EEG signals are obtained with electrodes that have some form of saline electrolyte-skin interface involved: gel, paste, or saline pads. My guess is that the MindAffect headband is using some type of electrolyte. The OpenBCI headset uses dry electrodes, which work, but may not work as well with MindAffect because of the lower signal levels.
re: Muscle Sensors / Myoware. These basically have the function of integrating the EMG signal over time. So thus they make detection somewhat easier. But EMG can also be received directly. Cyton does a better job of direct EMG receiving, because of the higher sample rate and not using compression. Ganglion compresses the data stream and so loses some of the EMG resolution. I believe the Myoware was added to the shop more to address Ganglion users. With Cyton you could receive as many as 8 EMG channels, or a mix of EEG and EMG channels. It appears that MindAffect only requires 4 channels, but possibly could work differently / possibly better with a few more channels.
re: Sasha's control of his 'cheek' muscles. Also consider his jaw directions: open, closed, left, right, etc.
Regards,
Looking at the MindAffect Ganglion Youtube video again, at the wiring. They might only be using TWO EEG channels, not 4. Although as far as the headset they seem to have at least 5 wires coming from the headset. Two of those are likely Reference and Ground. For SSVEP you typically attach the main electrodes to the occipital lobe, at say locations O1 and O2.
Thanks for confirming wjcroft. I'm off to the store to get the hardware. I've got 2 weeks off around Christmas so I will be able to do some more research and hopefully start playing around with the headset. I'll post here about any progress and I'm sure I'll have more questions.