Sociological research will tell you that sleep management in families entails "doing gender" - or how men and women experience snoring in unequal ways which affect the quality of relationships. The COVID-19 pandemic may have disrupted circadian rhythms; there is greater stress, worry and fear. Lack of a structured schedule with work-from-home lifestyles is gradually changing the way we sleep and snore.
While snoring is common and harmless, it is often seen as a symptom of obstructive sleep apnoea (OSA), which can cause a person’s breathing to stop and start while they sleep, wake up a lot or snore loudly. OSA has raised the risk for diabetes, obesity, hypertension, stroke and heart conditions. Sleep aid apps are proliferating in the market and so are the ones associated with snoring. Gone are the days when breakfast table conversations revolved around threats to surreptitiously record the loudest snoring family member. Now one can place a mobile phone next to their bedside pillow to record the intensity of one's wheezing and study their snorescore in the morning. Something called the SnoreLab is already available for iOS and android users. Researchers at the University of Sheffield are developing a new iOS app, SoundSleep, to monitor breathing while people sleep and help discover the causes, factors and solutions to snoring and sleep disorders using artificial intelligence. Dr Ning Ma, research fellow at the University of Sheffield’s department of computer science, has more.
Q\ When was SoundSleep first conceived and how did it set out to be different from the many sleep aid apps that already exist?
A\ The use of acoustic analysis in healthcare is one of our research focuses at Sheffield that sets out to use deep learning methods to better understand and treat sleep-disordered breathing. The machine learning algorithm in the SoundSleep App is the output of a Knowledge Transfer Partnerships project between Passion for Life and University of Sheffield five years ago. The algorithm was subsequently fine-tuned during a recent Innovate UK project, which just finished this year. We believe the acoustic analysis of breathing sounds during sleep can offer an inexpensive and unobtrusive way for monitoring sleep-disordered breathing. A key part of this technology is to look at how to differentiate between snoring sounds and background noise, and leverage state-of-the-art AI to identify sleep disorders from sounds. The machine learning algorithm has been designed to work in bedroom environments and can detect snores while ignoring other sounds in the bedroom.
Q\ Tell us about your work with the AI technology in your app which records and tracks snoring levels and provides nightly reports? How are these nightly reports created?
A\ We understand privacy is one of major concerns many users have for many sleep monitoring apps. The AI technology that records and tracks snoring has a small footprint in power consumption and the analysis is done entirely on the smartphone. We designed it in such a way so that it does not stream any audio away to a remote server and thus protects users’ privacy. The technology continuously records sound when the monitoring is switched on, and once enough acoustic evidence is buffered, it will start to analyse the buffer and try to identify snore events among many acoustic events in the buffer. This process is repeated throughout the night. The amount of snoring and the snoring levels can be computed from the identified snore events and are displayed in a sleep report ready for the user in the morning. One of the advantages of such an unobtrusive sleep monitoring approach is that it allows continuous monitoring across multiple nights. We believe this will be particularly useful for users to monitor their sleep over a long period of time, and the snoring trend will help users identify any issues such as sleep apnoea at an early stage, allowing early intervention.
Q\ What are the other ways in which AI is being harnessed to promote better sleep health?
A\ We have utilised AI to detect and screen for obstructive sleep apnoea from entirely sound. Our next plan is to look at AI tools that will enable efficient sleep disorder assessment, by integrating acoustics and small, low-cost sensors that can be placed unobtrusively on the body in a unified approach. The aim is to implement this software on a mobile device, offering a smartphone-based solution for home monitoring and treatment of sleep-disordered breathing.