Based on the short literature research I did that's visible in my other comment, I came to some rough ideas on how to make a good sleep tracker. If there are people who want to work on it, feel free to reach out. Creating this as a side project might be something I was going to do anyway to learn about data science and some simple hardware hacking. It's currently 2nd on the list (only beaten by the 1st thing: prepping for interviews).
What I think that technically needs to happen at minimum is:
1. Movement
2. HRV
3. EEG (brainwaves)
That will be the minimum viable product that may become as good as polysomnography and bring sleep lab tech to your home. Dreem claims to be doing this, but it isn't clear whether their claims are true.
Companies like Zeo produced presumably good results, yet they are now bankrupt. It's quite infuriating to see. So there's a business model (or simply an organizational problem at Zeo at the time) and I'm not clear what it is. The tech is already here! For example, Somté PSG shows that it can be done, and I'm simply thinking to myself: this could be more user friendly, couldn't it? [1]
Marketing-wise, I think the following needs to happen:
1. Get multiple unrelated universities to test your device against polysomnography and iterate against it.
2. Once it's within the 90% agreement range, then even make a YouTube video about such tests. Keep testing it against the golden standard, because if you are as good as the golden standard, then you can really state you track sleep.
3. Also test different PSGs with each other, because while PSG is the golden standard, no study told to what extent they actually agree on sleep metrics. There probably are studies on it but they didn't test this (or comment on it even, I believe) when wearables such as a Fitbit were studied as well.
4. Make ridiculous cool claims like everyone is already doing. Really go for the full appeal to authority, like Dreem is doing (nothing new).
I have a watch-like thingy that tracks my pulse and give interesting data about sleep patterns but every time I sleep with it I am self conscious about it. It emits light to track the pulse and it's annoying to have something at my wrist.
I can't imagine having a headset and sleeping as I normally sleep. The results wouldn't be fully representative.
I've noticed I'd become used to it. I've gotten used to a sleeping mask and earplugs in a similar way (they really help me).
I've noticed I don't have issues wearing something around my wrist, but I get that that's a factor. I'm also noticing that I don't get anxious about the data, because it's all a bit guesswork anyway, so I guess I'd want to see an aggregate form of data if it gets more precise, so that it still feels a bit like guesswork, unless there's something wrong in a big way.
Also, as appealing as the device sounds, I'm terrified of giving this kind of personal data away to a tech company:
>Dreem works as a long-term sleep expert and one-on-one coach: it monitors your sleep habits and sleep quality for a week, asks you for a lot of more details about your life (day and night), and downloads your activity data from other personal health apps
The overly suave marketing on the dreem site makes me wary as well. Between personal data abuses and shady marketing practices, how is anyone supposed to trust modern tech anymore?
Hi, I am the author of the mentioned article. It is not an ad :). In fact, I am Czech publisher of Matthew Walker's book "Why We Sleep" (mentioned in the article) and a bit lifehacker too – blogging in Czech-only until recently. I just wanted to check if my suspictions about the most hyped sleep-trackers are legit but waited for something better to compare them with. I believe Dreem can be used for this task, but, sure, it is not perfect – some people report worse EEG signal quality and thus less effective stimulations. Dreem removed weekly reports in last app update and there is nothing like Oura Cloud for "data mining" in Dreem, which is really a miss! I finally bought another Dreem2 for my mother, so I have to translate everything from English to Czech for her but I want to try to help her with her terrible sleep habits. The article will probably continue one day with my longitudial results (tested on me and her too). Ask whatever you want, I will try to answer.
Best, Tomáš
The normal data export only includes summary data for each night in csv:
Type (sleep, nap, relax)
Start Time
Stop Time
Sleep Duration
Sleep Onset Duration
Light Sleep Duration
Deep Sleep Duration
REM Duration
Wake After Sleep Onset
Duration
Number of awakenings
Position Changes
Mean Heart Rate
Mean Respiration CPM
Number of Stimulations
Not sure if there's any way to get the raw sensor data.
There appear to be two terminals on the band which are not used by the normal user. I assume they are for flashing the firmware on bricked devices.
They don't give you the raw data, I emailed support a couple months ago and their response was:
"Unfortunately, we do not provide you with a raw data report containing your EEG data. However, we do provide it via user-friendly hypnogram via the Dreem app."
Disclaimer: I won't be able to find all my sources. I'm sorry for that.
I just looked into this 2 days ago! While I'm not the OP, this submission happens to be one of the most exciting articles I found (I should've submitted it :P). I was asking myself whether I should update my Fitbit Flex 2 to a Fitbit Versa or an Apple Watch or anything else really.
The answer is: don't upgrade, keep the Fitbit Flex 2, it's about as good as journalling your own sleep times and that's all I need really, a very rough estimate. More importantly, the newer wearable devices are not good enough at giving a reliable breakdown of your sleep that polysomnography (the gold standard would give).
With that said: the Basic Health Tracker (doesn't exist anymore, I think), Oura Ring, Philips Actiwatch Spectrum were all able to do things that trackers that the Fitbit and other wearables couldn't.
Specifically:
> Only mean values of sleep efficiency (time asleep/time in bed) from Actiwatch correlated with PSG, yet this correlation was weak. [2]
> Light sleep time differed from PSG (nREM1 + nREM2) for all devices. Measures of Deep sleep time did not differ from PSG (SWS + REM) for Basis. [2]
> From EBE analysis, ŌURA ring had a 96% sensitivity to detect sleep, and agreement of 65%, 51%, and 61%, in detecting "light sleep" (N1), "deep sleep" (N2 + N3), and REM sleep, respectively. [4]
A good literature review on it is [3].
Some facts (which can be found in [3] or its cited papers):
* All trackers overestimate sleep by about 30 minutes.
* The biggest problem for all these wearable trackers are: estimating when you're simply lying awake in bed but you're not moving (it thinks you're asleep).
* Devices like the Fitbit Flex are about as accurate as sleep journalling with regards to total sleep time.
Dreem has the following issues. It's hard to see whether their scientific papers are peer-reviewed. Yes, they produce science articles in PDF formats but not in journals. Yes, they collaborate with universities but it isn't clear how. Nothing is peer reviewed and almost nothing seems to be cited. This was echoed in a Gizmodo article [1], but I figured it out myself first.
I'm not saying they aren't legit, because another thing I have noticed is that the scientific community is a bit but not too keen on testing these devices with the golden standard (polysomnography/PSG). There is still space for a startup who is clearly capable of demonstrating this and other research institutes confirming it independently for their device.
Edit: I hope I gave enough sources to make my comment a bit credible.
What I think that technically needs to happen at minimum is:
1. Movement
2. HRV
3. EEG (brainwaves)
That will be the minimum viable product that may become as good as polysomnography and bring sleep lab tech to your home. Dreem claims to be doing this, but it isn't clear whether their claims are true.
Companies like Zeo produced presumably good results, yet they are now bankrupt. It's quite infuriating to see. So there's a business model (or simply an organizational problem at Zeo at the time) and I'm not clear what it is. The tech is already here! For example, Somté PSG shows that it can be done, and I'm simply thinking to myself: this could be more user friendly, couldn't it? [1]
Marketing-wise, I think the following needs to happen:
1. Get multiple unrelated universities to test your device against polysomnography and iterate against it.
2. Once it's within the 90% agreement range, then even make a YouTube video about such tests. Keep testing it against the golden standard, because if you are as good as the golden standard, then you can really state you track sleep.
3. Also test different PSGs with each other, because while PSG is the golden standard, no study told to what extent they actually agree on sleep metrics. There probably are studies on it but they didn't test this (or comment on it even, I believe) when wearables such as a Fitbit were studied as well.
4. Make ridiculous cool claims like everyone is already doing. Really go for the full appeal to authority, like Dreem is doing (nothing new).
[1] https://www.compumedics.com.au/products/somte-psg/