Wearable sleep trackers have become ubiquitous. Most measure sleep based on movement, although new EEG based sleep headbands are hitting the market. How accurate are they and what can they really tell you?
Wearable technology is fast becoming the new norm. The ability to track an array of physiological, sleep and activity metrics 24 hours a day, 356 days a year, offers the opportunity to combat bad habits, improve health and obtain a level of physical enlightenment beyond our own self report and guesstimation.
From a research perspective these relatively low-cost technologies offer an unprecedented window into the health and sleep behavior of millions of individuals. And although the majority of these technologies claim that they are not intended for scientific or medical purposes, the gap between them and clinical-grade devices, is rapidly closing. Improved algorithms, the launch of a Sleep Technology Council by the National Sleep Foundation and an increased availability of high resolution raw data from software platforms such as Fitbit’s Fitabase all help to connect the dots between wearable manufacturers and clinical scientists.
Sleep sensing with wearables
Wearables such as Fitbit, Jawbone (which is now going out of business) and Polar generally sense movement and use these movement patterns (actigraphs) to predict night wakings and stages of sleep. There are a growing number of studies that attempt to independently validate their accuracy against multi-parametric polysomnography (PSG) measures which include EEG, EOG (eye movements), EMG (muscle activity) and ECG (heart rhythm). Some of these include this recent evaluation of two clinical-grade actiwatches – the GT3X+ and Actiwatch Spectrum as well as an evaluations of the commercial Jawbone UP device with adults and adolescents. These studies are surprisingly small scale (13 to 25 people) and cover just one night of activity and some included very selective and odd presentation of statistics. Nonethless, the general results of these studies suggest that overall time in bed is pretty accurately estimated by these devices, but on the other hand detection of nighttime wakings was fairly inaccurate. Essentially these devices can give you a good view of gross aspects of your sleep-wake cycles. However, any further interpretation would be overreaching.
Tracking the nuances of sleep
In terms of tracking sleep, particularly for those with sleep issues, the big challenge is that there is yet no clear understanding or gold standard for what to measure against. The traditional divisions of sleep stages based on EEG have been challenged by new studies and there is yet no clarity what really constitutes a good night’s sleep. The sleeping brain displays a rich pattern of continually shifting neural dynamics whose specific function in cognitive and emotional health are yet unexplored. To truly explore and understand sleep in all its complexity, as opposed to just measuring it, you have to use EEG to record a more detailed picture of what is going inside the brain.
EEG headbands for sleep
The arrival of EEG-based headband wearables (current ones include Dreem, Sleepshepherd, and Smartsleep) on the marketplace is a first step (on what is probably still a long road) in the development of comfortable and reliable EEG systems which could be used to record the neural dynamics of sleep across the whole head to create consumer-driven large-scale datasets across different individuals and populations.
Philips Smartsleep device
But for the moment, the current aim of recently launched EEG headbands such as the Philips Smartsleep, which was unveiled at CES earlier this year, is not to just record and provide feedback on your sleep, but also to improve it. In particular their headband claims to boost your slow wave sleep with auditory tones – something closely resembling the concept of auditory closed loop stimulation demonstrated in a 2013 Neuron paper by researchers at the University of Tübingen and University of Lübeck. In this study they showed that auditory sounds presented in phase with low-frequency waves during sleep led to an enhancement of slow wave oscillatory activity, whilst auditory sounds presented out of phase with the these low-frequency waves did not deliver an equivalent enhancement.
While the number of subjects was small (N= 11) and the results largely directional in terms of functional impact (see Figure showing the difference between stimulation and sham on a memory retention test), it does suggest that this direction may have potential. As stated on the final line of their discussion “closed-loop in-phase auditory stimulation at low intensity might be a promising tool to generally ameliorate efficacy of sleep rhythms, also in pathological conditions such as insomnia” – something that seemed to have been taken on board by Philips where they claim that “Our proprietary algorithm customizes the timing and volume of tones delivered to boost your slow waves, so you can make the most of the deep sleep time you have”.
The future of sleep technology.
Sleep technologies empower consumers to take charge of their sleep behavior, monitoring how changes to their lifestyle can improve or negatively impact their sleep, and therefore their waking cognitive and emotional functioning. They also offer opportunity for the collection of large-scale research datasets, outside of a lab environment to obtain naturalistic measures of nighttime sleep. This in turn can feedback into the development of new algorithms for computing sleep quality.
However, in the fast-paced world of wearables where new models are regularly launched to market, the secrecy behind the algorithms used to compute sleep metrics and a lack of technical information about the specific sensors within the devices pose serious challenges to these scientific endeavors.