The 115th Installment
Good Sleep and Bad Sleep: Sleep Quality According to Data Analysis Science
by Huang Xuping,
Assistant Professor, Master Program of Information Systems Architecture
The stresses of modern day life are seeing many suffer from insomnia and poor sleep quality. Few, I surmise, have a developed capacity to visualize, analyze, and understand their physical condition or sleep quality, even though they may be healthy. Moreover, while there have been many research papers that have proposed solutions based on analyzing clinical data (from pathological analyses, for example), we are still a long way from the practical application, or widespread adoption among general users, of cutting edge technologies as a means to answer a societal need.
One noteworthy solution that has come to the fore in recent years are wrist-worn IoT devices that can conveniently analyze and make sense of sleep quality. In particular, products like the Apple watch, Fitbit, Veryfit, and Jawbone are distinguishing themselves as wrist-worn heart rate monitor products that are affordable and unobtrusive when worn. They also serve as unique and fashionably designed wearable accessories, equipped with tiny sensors for monitoring body temperature, acceleration, and heart rate.
In addition to a function for tracking heart rate, almost all of these products have functions for sampling blood pressure data at fixed times and for automatically measuring sleep depth (deep sleep, shallow sleep, REM sleep, and being awake) while the user is sleeping. These mechanisms measure sleep depth logically and accurately based on clinical experience observing how heart rate and body temperature drop when we are sleeping and how our wrist movement speed changes when we turn over, as well as how we enter deep sleep shortly after falling asleep. These devices compile, organize, and visualize data such as heart rate, sleep depth, and everyday activity so that even general users can see it all at a glance. Using API, these devices can even sample data at one-minute intervals. Fig. 1 shows example data visualized via a self-built program using my own heart rate and sleep stage data.
Although EKG records have been the primary means of collecting heart rate data, it is considered difficult to use EKG to measure heart rate during sleep due to the limited ability to measure respiratory rate and bodily movement. However, if a person is wearing one of these devices, the screen will show heart rate fluctuations due to sleep apnea or sudden occurrences in those with heart conditions as they sleep. Others can observe and track the sleeper's condition in real time by symbols indicating abnormalities. While measuring precision drops slightly compared to data taken via multiple sensors placed on the face, such as when measuring brain waves or eye movement, it is still informative.
Combining frequency spectrum analysis with chronological data on things like heart rate and sleep depth allows one to even calculate stress by analyzing the power spectrums of the high-frequency and low-frequency bands. And visualizing stress could prevent death from overwork. Integrating knowledge from the disparate fields of medical engineering, sociology, and informatics, researchers are shedding light on mechanisms associated with biosignal by using data analysis methods to examine new healthcare phenomena. Furthermore, AI technologies are being put to work alongside environmental sensors to enable advice giving on the best sleep environment based on each individual's physical constitution.
Users being able to intuitively understand data and stress indicators concerning heart rate fluctuation, blood changes, and sleep depth is a very valuable step forward that has led to better health awareness and lifestyles that benefit from improved sleep and health. With the rise of IoT devices and the development of data analysis technologies, one could also say that scientific progress has leaped forward in ways that will positively affect our daily lives.