JMIR Publications

ABSTRACT

Background:

Worldwide, between 30% and 45% of adults suffer from sleep disorders, which are associated with major health problems like diabetes and cardiovascular disease. Long-term monitoring is not feasible with traditional in-lab testing due to its high cost. Wearable AI-powered solutions provide accessible, scalable, and continuous monitoring, enhancing the identification and treatment of sleep problems.

Objective:

This scoping review aimed to provide an overview of artificial intelligence (AI)-powered wearable devices used for sleep disorders.

Methods:

Seven electronic databases (Medline, PsycINFO, Embase, IEEEXplore, ACM Digital Library, Google Scholar, and Scopus) were searched for peer-reviewed literature published before March 2024. We created a list of keywords based on three domains: sleep disorders, AI, and wearable devices. The primary selection criteria were the selection of studies that utilized AI algorithms to detect or predict various sleep disorders based on data collected from wearable devices. The selection of studies was done in two steps: first, reviewing the titles and abstracts, then, full-text screening. The selection of studies and data extraction were carried out independently by two different reviewers, with differences being settled by consensus. The extracted data was synthesized using a narrative approach.

Results:

The initial search yielded 615 articles, 46 articles meeting the eligibility were included in the final analysis. The majority of the studies focused on sleep apnea disorder. Wearable AI was widely deployed for diagnosing and screening disorders, whereas none of the studies employed it for treatment. The most popular kind of wearable technology was a commercial device, albeit many various brands were utilized instead of just one large, well-known brand. Most studies used wrist-worn devices. Respiratory data was the most often used type of data for model development, followed by heart rate and body movement. The most popular algorithm was Convolutional Neural Network (CNN), followed by Random Forest and Support Vector Machines (SVM).

Conclusions:

Wearable AI technology offers promising services for sleep disorder issues. Disorders can be screened for and diagnosed with wearable AI devices. Research on wearable technology for sleep disorders other than sleep apnea is lacking. To statistically synthesize the performance and efficacy results of the investigations, more reviews are required. Technology companies should prioritize the latest technologies, such as deep learning algorithms, and invest in wearable AI for treating sleep disorders given its promise. More research is needed to validate the ML techniques on clinical data from WDs and deliver useful analytics that might function as data collection, monitoring, prediction, classification, and recommendation devices in the context of sleep disorders.