Low-cost 16 EEG channels cap (hat) for PiEEG | Hacker News Reader

Low-cost 16 EEG channels cap (hat) for PiEEG

pieeg.com

37 pointsMarat_197521 comments

It's $390 and the actual buy link is https://www.elecrow.com/cap-eeg-kit-16-channels.html.

This is just for the cap. You would also need 2x of the $350 pi eeg hats and of course the pi itself.

soon will be available pieeg-16 for 390$ https://pieeg.com/news/pieeg-16-is-availabe-in-the-market/

This is pretty cool, and the price (390USD) isn't too bad. Might be useful for self-monitoring of my sleep apnea treatment (all sleep tracking that isn't an EEG is essentially unreliable).

Almost tempted to buy one, but unclear where they're shipping from.

SpO2 monitoring is actually quite reliable.

It really isn't. Most sensors suck. The reliable ones (i.e. the ones used in sleep labs) are uncomfortably tight (because they need to stay on to give useful readings). Most sensors stop being reliable as soon as any movement is involved. Even the really expensive ones in rings aren't all that.

And either way, SpO2 monitoring is not very useful for sleep apnea monitoring. The problem is not oxygen desaturation but rather arousals[1], which cannot be detected by looking at an SpO2 chart. Some people, even when sleeping with CPAP, have hundreds of arousals per night, wrecking their sleep quality, while their oxygen saturation remains perfectly fine.

[1] If sleep apnea is not treated at all, desaturation may be more prevalent, but even then arousals are a bigger problem unless saturation goes way down (below 85% or so).

Anybody has experience and is it actually useful?

Heard a lot of "it captures mostly face muscle contractions" and "capturing a brain signal with it is like listening to a whisper being spoken in a grand canyon, while being miles away"

> it captures mostly face muscle contractions

Maybe it is useful for people who suffer from bruxism.

Listening to a whisper in the grand canyon is possible with modern ML ;)

As a neuroscientist focused on intracortical recordings, these can be useful for certain clinical diagnostics but they’re quite limited in terms of signal quality. They signals are analogous to listening to an orchestra from outside the building. You may be able to tell that a song is playing but you lose the vast majority of the detail, and what we can learn from this is somewhat limited. Noise from facial muscles is another huge issue if you’re focused on the brain. Some interface devices focus on these signals instead for silent speech decoding etc. that can work, but the applications are still somewhat niche.

It is a lot like that.

I only played with these briefly in biomedical engineering, we experimented with signal analysis algorithms while monitoring engineering students on varying levels of caffeine (it was hard to find volunteers for the control group) and doing various tasks. You could correlate EEG measurements to tell whether someone was focusing on a reaction time test and exactly when they reacted, whether they were at a stressful or relaxed part of a movie, estimate their WPM level while typing (calibrated to individual users), and similar stuff, which was at least interesting. I think one group did a sleep study and correlated EEG readings to heart rate and EMG data, looking for sleep phases and REM state activity levels, which is a common use case too.

Whether it's useful or not is a completely separate question, but those are the kinds of things that you can measure.

Some people (eg. Youtuber Perrikaryal) are trying various things like mapping "concentrate really hard on pushing this cube" to moving a video game character on a screen, which is astonishingly impressive to me, whose last exposure to the field is over a decade old. But even at the bleeding edge, it's nowhere near the point of "I'm imagining a shape, now 3D print it" or anything like that. Just as with AI, the hype in this sector is high because people imagine that it will eventually do that and they want to be the first ones there, but there's a lot of work between then and now if it's even possible at all.

Putting in a plug for Open BCI

https://shop.openbci.com/collections/openbci-products

which has been making EEG equipment for DIY use for years, and has some scientific validity claims.

Including their https://shop.openbci.com/products/ultracortex-mark-iv for the same price as the article's

No affiliation, just a fan.

Emotiv did 24 channels for <500 if I recall and that was a decade ago. The raw data become accessible from a hack. Wonder what happened to them? Terrible customer service

What makes these so expensive in general? Even $390 seems insanely expensive for what (to this layperson) seems like a bunch of wires on a skullcap.

you have to amplify electrical signals from the scalp that are in the order of microvolts, which requires having low impedance between the electrode and scalp, proper amplification circuitry, and the ability to filter out lots of environmental electrical noise. I am a student in a lab where we have a EGI system that costs in the ballpark of ~$100k (with $10-20k in nets that need to be replaced every couple of years), and it is even a world of difference between that system and a $30k mobile EEG system that we recently acquired.

Isn’t that amplification and filtering part of whatever the “hat” would plug into? I.e not part of the product offering anyway?

The wires are pretty complicated - trying to pull out some very low signal voltages and high noise levels, reading through a high-impedance skin layer with exotic silver chloride electrodes. It's not like wiring a dishwasher.

Plus, there's a very big obstacle with the required low volume, high complexity assembly. This isn't a big enough market to be worth automating, it involves a seamstress and a technician with minutes of hands-on time assembling it.

I'm an industrial controls engineer, we just wrote a quote and did some concepting to build a piece of equipment that automates wiring harness construction for a major office furniture - think cubicle dividers with embedded lights and outlets. It involves multicolored single-stranded wiring harnesses pulled/pushed through flexible metal conduit, with wires picked up by a robot, stripped, terminated with a crimped-on insulated spade terminal, and tested. Project cost to lights-out build something that would take an electrician with hand tools 5 minutes to accomplish (and would require several electricians, working without fault, for 2 shifts a day, without quitting due to boredom and RSI) is on the order of $350k.

It is a bunch of wires on a skullcap, and that's what it costs to make it.

FWIW, the OpenBCI equivalent:

https://shop.openbci.com/products/openbci-eeg-electrocap

is $500.

The electrodes themself seem to cost $45 for 30 on that site? Is the rest just some low impedance wire? I appreciate economies of scale, etc, but they advertise these as “open” solutions, but is it some knowledge gatekeeping that stops this being truly DIY?

You are welcome DIY it yourself with a sewing machine, soldering iron, and some hours of manual skilled labor.

But sewing skills are becoming less common, that labor and equipment has a cost, and the market has pretty well settled at these prices.

If you started a company buying those electrodes for $25/16 (or much less in bulk) and selling these super-simple skullcaps at $50 each, undercutting the market by 10x, I'm sure demand would be high and people would flock to your door.

Heck, I'll build you an automated wiring harness machine for on the order of $500,000 to churn out rainbow ribbon cables with molded snap electrodes on one side and a 16-pin IDC Dupont parallel connector on the other end if you've got the cash and can allocate $0.50 per unit towards that automation equipment over the first million projected sales. I've also done automated fabric cutting and sewing (in upholstery and automotive sectors, not clothing/medical). Sewing is much more complicated and expensive than wiring, and typically involves significant amounts of hands-on time.

The absence of anyone doing that suggests that it's not just gatekeeping and price fixing that sets prices where they are.

The modern clothing industry that churns out T-shirts for a few bucks and cheap blue jeans for $20 is working miracles, you should not use that as a bar with which to set expectations.

> You are welcome DIY it yourself with a sewing machine, soldering iron, and some hours of manual skilled labor.

Thats exactly what I would expect to do, but it seems instead of buying the electrodes and handling the wiring as part of a DIY project, these "open" solutions are just selling their own 10x-more-expensive-than-the-parts kits, and the knowledge to make it yourself is treated as some lost-to-the-ages arcane magic

EDIT: ok, thats not fair of me - i just found it on their website at: https://docs.openbci.com/

soon will be available pieeg-16 for 390$ https://pieeg.com/news/pieeg-16-is-availabe-in-the-market/

Reply on news.ycombinator.com