The Science behind CALM. – Sleep Analysis

The science of sleep analysis with CALM.

In part one of the science behind CALM. I want to talk about sleep. We spend a third of our life in slumber, and yet we are unconscious and seldom aware of its quality and effects. Indeed, there is a recent boom of interest in sleep quality, and how to improve it. The latest wearables are able to detect weather you are sleeping or not, with very good accuracy (around 95%), but falls short when analyzing sleep-stage, with accuracy of around 60% for ring type devices [1] and even lower for wrist worn devices [2] , as they are usually based on movement, and sometimes temperature and heart rate.

CALM. detects sleep-stage using movement, and the ECG signal from your heart. ECG is the electrical signals that are making your heart move, measured directly from the heart, as opposed to heart rates monitored at the wrist or finger, which are measuring the rush of blood through your veins as a result of your heart pumping. This means that we can get millisecond accuracy on the activity of your heart, and also interesting (and useful!) noise from your lungs. From the ECG, we can derive heart rate very accurately, as well as your breathing rhythm with reasonable accuracy using a well known technique called ECG Derived Respiration (or EDR) [3].

The heart rate data is so accurate that we are able to take advantage of and analyze the correlation between your heart beat timings and your autonomous nervous system which controls your unconscious body functions, such as heart beat, breathing, and even sexual arousal. The autonomous nervous system acts very differently when we asleep, which is why CALM. is able to use that as additional information to analyze your sleep-stage and quality. We use Machine Learning and data sets from clinical sources and open data to create the sleep-stage classifier used in CALM. and we are currently able to achieve around 80% accuracy, and hope to bring that up past 90% soon.

As for EDR, we use that to see how well you are breathing during the night. If you are known to snore at night, chances are your breathing is not optimal. When we are asleep, the muscle in our tongue relaxes, and (among other things) causes snoring. Unless someone is awake next to you listening to you snore, it’s hard to analyze the breathing quality when you are asleep, even though it is critical to your body’s recovery. With CALM. this is made possible by getting the approximation of your breathing rhythm from the ECG signal; if you are not breathing smoothly, the rhythm will have some irregularities. Breathing quality adds another dimension to the sleep quality analysis that is simply impossible with wrist or finger worn activity trackers.

Even with its advanced analytics, CALM. is not a medical device and should not be used to diagnose sleep disorders. If you suspect something may be wrong with your sleeping quality, consult a doctor for advice, and get a professional examination.

In the next post, we will look at the CALMness score, a metric we came up with for a simple representation of your mental state of relaxation, focus, concentration, and stress.

[1] Zambotti et al. (2017), The Sleep of the Ring: Comparison of the ŌURA Sleep Tracker Against Polysomnography
[2] Mantua et al. (2016), Reliability of Sleep Measures from Four Personal Health Monitoring Devices Compared to Research-Based Actigraphy and Polysomnography
[3] Moody et al. (1985), Derivation of Respiratory Signals from Multi-lead ECGs