EEG channels with low-cost PiEEG device | Hacker News Reader

EEG channels with low-cost PiEEG device

pieeg.com

78 pointsron_8744 comments

There is also the OpenEEG project[0], the boards are way cheaper (sub 100eur[1]) and also they are both open source and open hardware.

[0] - https://openeeg.sourceforge.net/doc/index.html

[1] - https://www.olimex.com/Products/EEG/OpenEEG/

is is possible measure EEG with 8 bits ADC ?

With an active electrode, it should be good enough. After all, we are talking about hobby-grade results.

The Olimex ones are 10-bit (supporting both active and passive electrodes), while it's claimed[0] that 12-bit resolution is sufficient for EEG.

However, it's worth noting that there are obvious advantages to the higher resolution.

[0] https://www.biosemi.com/faq/24bitsystem.htm

EEG in microvolts, if ADC 12 bits, then reference for ADC should be in millivolts

He mentions active probes. Active probes have amplification.

As his link makes clear, your ADC just needs enough resolution to measure the signal when the (amplified) offset voltages don't saturate it.

yes, I told about digital EEG device from this website. To clarify analog device, I need a little more time

I've been wondering about this for a while.... have you tried?

I'm not a EEG designer but I've experimenting and learning about it. Sounds like you make PiEEG? This is really cool and I appreciate the docs on the github.

After amplification, you could get the signal 0-100uV to ~0-Vanalog; so then the entire 8-bit range could ideally be used. Realistically, after filtering, due to roll-off / artifacts one could probably use 3/4ths of that space -- so ~ 192 voltage levels that can be determined by the ADC per sample. Does this sound right?

It'd probably depend on what level of control / reliability one would want from the EEG data; I imagine that you could probably cut a lot of corners if, say, you were only measuring for hemispheric coherence in a small set of frequency bands, or stages of REM sleep (and maybe ERPs like p300).

I don't know, this is in my upcoming experiments. Originally this was an attempt to build an EEG amplifier / filter circuit for an atmega328p 10-bit ADC but for my purposes I settled for a 12-bit (and possibly hardware oversampling) on an EFM32.

I was going to point to openBCI as another option, but the definitely are not a low-cost option. I don't remember spending this much on the Cyton board a few years ago.

https://shop.openbci.com/collections/frontpage

yes, thank you!

I recently discovered that you can attach an external electrode or two to a Muse headset: https://hackaday.io/project/162169-muse-eeg-headset-making-e.... This is a fairly cheap way to hack around with an EEG system that has four electrodes at the front and then an extra electrode you can place anywhere else you want. I think it would probably work pretty well for simple stuff like detecting P300s or SSVEP.

But yeah you definitely get what you pay for with EEG. In the lab I'm in, we have a $150k EGI system and a $40k cognionics system and even the cognionics system is way more glitchy/low quality compared to the more expensive systems.

Also, I wonder what happens safety/electrical noise-wise if you're plugging a raspberry pi into the wall? Generally, EEG systems are either battery-powered or use an isolation transformer. I wonder if it's best to power the pi with a battery.

Without some serious fuse/filtering, I don't think I'd feel good about plugging it into the wall. It's easy enough to get a battery pack that can provide enough juice to run this for testing sessions, so why risk it, even if the risk is small?

only from battery, thank for your message! for details write me here https://www.linkedin.com/in/ildar-rakhmatulin-262a66112/

Is someone maintaining a big list of BCIs somewhere? If you're going with extra-cranial hardware, I think the best available involves Microwave Brain Stimulation along with passive sensing methods like this. Unidirectional brain -> computer is neat but so long as the feedback is via sight or sound it's just another move-the-mouse alternative.

> , I think the best available involves [anything]

Consider how long it took to start suspecting football and other low impact head trauma of causing CTE, and to start believing in CTE at all. Personally, with a phd in neurobiology and no real data besides gut feeling, I would not be comfortable with any system that wasn't thoroughly shown to be mostly harmless over several decades at least

Whereas others take drugs made in bathtubs and cooked by idiots. The right to experiment on oneself is, for now, an advantage individuals have over corporate interests. You do you. I'm happy just seeing what kinds of stuff they make, for now, but if someone makes something cool, I'll buy the kit.

I like the idea, but $350 plus a Pi plus getting electrodes is beyond my "impulse by to try it out" price for sure.

yes, hat electrodes + plus about 300 dollars

I work at a relatively unique EEG startup, and our single-patient-use EEG headbands cost a bit more than this project.

That said, doing an EEG on yourself is a bit disappointing imo; You won't get much interesting output.

If you are at liberty to share, how many channels and are the electrodes dry or wet?

Sure :) (this is all publically available, fyi, no trade secrets here)

8 channel headband with individual gel packets for each electrode. You pop the gel packets and then you have wet electrodes!

The company is called Ceribell; Happy to answer more questions if you have them. We do not sell to individuals, but if you're in the research space, I'm happy to get in touch.

sounds great,, can you give a website for your project ?

Thanks for replying! It looks very intriguing. I have only worked with caps that have electrodes covering the whole sculp (32 channels). I have no research funding at the moment, but thanks for offering!

This seems as if it could be cool, but the board is $350 with a BoM of about 50 bucks. Considering the current nascent state of things, this doesn't really feel like a project on a successful path.

why 50 ?

The Elecrow page says that the analog front end used on the board is a ADS1299. The eight channel version of it is 60€+VAT on Digikey. The rest of the parts is most likely very cheap.

yes, but also ICs for Voltage! I use with the low noise. Also price for PCB board and for soldering and in my case also included software) But anyway I will try make price lower in the near future

The BoM is of course only a small part of the cost and there is lots of work time and prototypes that need to be paid off by selling finished devices. I‘m sorry if that sounded like I thought that the price might be too high. Thanks for your efforts!

you are very polite, thank you very much

I have tremendous respect for your efforts and goals. I should have reiterated that in my original comment. If you need to price these boards at $350 as a measure to ensure the success of your project, and you have people willing to participate at this stage of development at that price, more power to you. But your git repo looks like you need people to help and contribute, so I might suggest that it’s a little early to be amortizing your R&D costs onto the prototype hardware pricing. The projects I’ve seen successfully transit his stage usually arrange group buys and the like to get it into the hands of potential contributors as quickly and affordably as possible. This is also unlikely to be the last revision you want to make to the hardware.

What are good papers to get started on BCI in tandem with the recent advancements in Large Language Models?

Our preprint does exactly that :). https://arxiv.org/abs/2403.05583

Forgot to say, 8 channels now and 16 in the few months

Anyone recommend particular electrodes for this or have a good comparison with alternatives?

dry electrodes Ag/AgCl (silver-silver chloride)

Is this how the Neuralink device works?

NeuraLink is using a specialized robot to surgically insert very thin electrodes into the brain.

This device here uses good old wet or dry EEG electrodes that are touching the outside of the head (e.g. on an EEG cap).

As a bad analogy, I've heard the difference between EEG and implants described as pointing microphones at the outside of a building vs having them inside the room.

Well at least you can tell if there’s a party going on in room 106.

Well, EEG is more like undirected microphones that record the voices of several thousand people at the same time outside of a building, whereas implants are more like directed microphones inside the building that only record the voices of a few people...

It's worth noting that if each electrode is thin enough to be placed in extracellular medium getting signals from several neurons around it, you can do some fairly simple signal processing to separate the signals of individual neurons. This is called spike sorting [0] (a "spike" is an impulse transmitted by a single neuron).

So, continuing with the microphone analogy, with the right setup, recording a few voices is almost as good as recording each voice separately. It's not perfect because sometimes people talk over each other, and people distant from the mic may have muffed voices, and stuff like that. But generally speaking, we can roughly understand each individual people in a recording, because each people speak in a different way (and just like that, neurons spike in a different way).

Here's some academic code that does this [1] (it has accompanying articles [2] [3]) (Random article found in Google, no affiliation).

The first thing it does is removing the low frequency component of the signal (the LFP [4], that doesn't originate from any specific neuron), then it stores each individual spike waveform as a time series (a vector of N dimensions, where each dimension is the value of the signal at a moment in time), and then just does some k-means clustering to separate spikes in clusters, then assign each cluster to an individual neuron (by hypothesizing that each neuron likes to spike in a specific way which depends on its characteristics). This is all very vanilla / undergraduate stuff, and it's been done for more than 10 years [5] or perhaps even 20 years or more.

Trouble is, with EEG all you have is essentially an average of the LFP over a large area; you don't get single-neuron spikes. So EEG fundamentally can't listen to each neuron individually. However it isn't right to simply throw away LFP information, and you can still do stuff with it.

[0] http://www.scholarpedia.org/article/Spike_sorting

[1] https://github.com/akcarsten/spike_sorting

[2] https://towardsdatascience.com/using-signal-processing-to-ex...

[3] https://towardsdatascience.com/whos-talking-using-k-means-cl...

[4] https://en.wikipedia.org/wiki/Local_field_potential

[5] https://www.ncbi.nlm.nih.gov/pmc/articles/PMC3657693/

Sounds great! thank you, you seem to be good at this

Not really, I just took a couple of undergraduate courses in neuroscience, but I didn't major in it or anything. For a time I thought about doing research in this area, but I got sick of how the lab animals were treated. Take a look at this

https://en.wikipedia.org/wiki/Remote_control_animal

How does the neuralink get an advantage though, do neuron spike signals contain any information

of course not

Reply on news.ycombinator.com