Comment by pedalpete | Hacker News Reader

Comment by pedalpete | Hacker News Reader

I have a bit of experience in this space as we're making a sleep EEG device to improve the efficiency of deep sleep.

The hardware is one of the easier parts to build for an EEG system. Along with this device, you could also buy a very similar device from https://openbci.com (about $1000 for 12 channel, $400 for 4 channel).

The challenge in letting "anyone connect their brain to a computer" is in the electrode design and comfort/weirdness of the device.

First you've just got the dry electrodes, which can be expensive, and in many designs (Muse S) wear over time, requiring replacement headbands.

Many devices try going for the through hair electrode design, which is challenging for people with lots of hair, or curly afros (like myself).

Once you've got your electrodes figured out, then you have to figure out how to keep them in place, comfortably, over an extended period of time.

Ok, so you've got that sorted, now get that into a form factor a person will be happy to wear.

We are starting to see in-ear EEG devices built into headphones like the Emotiv Mn8 (https://www.emotiv.com/workplace-wellness-safety-and-product...).

However, even in these locations the electrodes are going to be sensitive to movement. So great it you are sitting still at work, but what are you really trying to accomplish.

So that's all the bad news. The good news is that with the low cost and experimentation that is happening in this space, I suspect we are going to learn more about the brain in the next 10 years than we have in the last 50.

It is becoming easier to watch the brain and learn more about it's functioning.

Many people ask what the benefit of a BCI device is. Unlike fitness monitors, which are giving users data, I believe the near future of BCI is in neurofeedback.

We're seeing a bunch of development from different start-ups (a bunch in Sydney, AUS - so reach out if you're local) that are looking at neurofeedback for treatment of psychological disorders.

At https://soundmind.co, we're using neurofeedback during sleep to increase the efficiency of deep sleep.

The link to the 2nd study on the search page doesn't work (links to an expired PDF link), should be https://pubmed.ncbi.nlm.nih.gov/31691831/

Thanks, I'll fix that up.

> At https://soundmind.co, we're using neurofeedback during sleep to increase the efficiency of deep sleep.

That's honestly a fascinating concept. How is it panning out so far? Do you control for participants having sleep apnea/UARS/other organic sleep pathology?

It's of interest, and are looking into a few things in that space, but we're starting out with known science where we can get to market "quickly" and help people currently in need.

There is LOTS of areas to look into once we have revenue and are looking more long-term opportunities.

I say "get to market 'quickly'" as it is still neuroscience, and as I describe in the parent, neurotech/eeg is not an easy physical thing to solve, then you've got the AI to build which is understanding the brainwave, providing the stimulation, and then adjusting the stimulation to the individual based on the resulting change in brainwave state.

How does your device overcome the barrier of the mains power frequency being different (50 or 60 Hz) in different regions? Since brain gamma waves are 32-100 Hz you would need a high attenuation notch filter at the correct frequency.

Although for your sleep application, you may only be working with delta/theta waves.

We do need to filter for 50hz/60hz noise. The EEG front-ends include what's called RLD which is used to cancel common mode noise.

But you are also correct we are currently only interested in the lower bands, mostly delta/theta/alpha, potentially beta when looking at wakefulness. We're not currently doing anything in gamma.

Biomedical engineering background here. You didn’t mention anything about the quality of the resulting signal. My experience with with EEG was that you can only get a few bits/second of info. Is this still the case?

When I was mentioning electrodes, I made the assumption signal quality was the prime target, but you're right, I didn't address that directly. Of course there are other considerations as well, but signal quality is the main one.

I'm not sure where you are getting your "few bits/second" number. EEG chips can sample thousands of times per second, and it would also depend on how many channels you are running, and what you're looking for.

So the example OpenBCI board I linked to is an 8 channel device with 8,000 samples per second and 24 bits per sample - so 15m bits/second.

Or perhaps I'm misunderstanding the question.

SoundMind product is fascinating. Is there a risk you become dependent on the device after longer term use and will have worse than pre-soundmind sleep when you have to go without it (e.g. it breaks or any other arbitrary reason)?

who knows...

neuroscientists and sleep doctors?

Nothing in the existing research suggests this to be the case.

It is important to understand, we don't help you to fall asleep. This isn't an insomnia device. We increase the efficiency of deep sleep specifically, and it is most effective on people who already have lower/less deep sleep.

I mean, what if the brain learns to expect the stimulus, and learns that it does not need to provide the stimulus on its own initiative anymore, or at least not to the same extent. Maybe it means the brain is spending that energy on other processes now that it is relieved of this task. Is that possible?

> Nothing in the existing research suggests this to be the case

Was it considered?

I'm not sure I completely understand how you're phrasing that question.

However, I will say that there is now more than a decade of research into slow-wave enhancement, and multiple studies looking at safety, and efficacy.

If dependency is a concern, it's somewhat the opposite. We're stimulating a sleep function that naturally degrades as we age, and trying to get it to maintain it's power. So you naturally lose this ability.

Note: we can't create slow waves (yet) we can only increase what is already there.

Not affiliated, but I would bet on the opposite being the case, that this device (if it works) would teach you a pattern of sleep that persists after ceasing use.

In the same way brain kindling works for seizures, my guess is sleep is a learnable skill, and once the pathway is etched it sticks around.

The device is built on known science, and the brain may learn the response, but it is in response to a stimulus. Without the stimulus, the brain does not increase increase delta power on it's own.

There will be other discoveries in the future, who knows what we'll learn, but the current body of knowledge (and there is a decent amount of it) does not suggest a learned response to be the case.

Furthermore, the sleep we stimulate is deep sleep, which naturally degrades as we age, which perhaps suggests that it isn't a learned response, or else why would we see a reduction with age.

Do you have any reference paper that details the difference between different headset and why one should chose over another for a given purpose?

We don't, but as we get closer to market, it will probably be more apparent.

I'm curious if you need test subjects, or do lab demos.

We have more "test subjects" than we have capacity, and we don't have our own sleep lab. So unless you're coming home and sleeping with me, I'm afraid not.

How cuddly are you?

thank you!!! great comment

Reply on news.ycombinator.com