[의학신문·일간보사=정광성 기자] A domestic professor team is attracting attention by developing an AI-based wearable device model that can screen children for ADHD and sleep disorders early.
Korea University Anam Hospital, Department of Psychiatry, Professor Cheol-Hyun Cho<사진>The team announced on the 28th that it has identified the possibility of an AI model that can screen children’s attention deficit hyperactivity disorder and sleep disorder early through a wearable device in collaboration with ‘Luman Lab’, a digital healthcare company for infants and toddlers.
According to the professor’s team, early diagnosis of attention deficit hyperactivity disorder (ADHD) and sleep disorders in children is very important for children’s mental health and growth development.
However, it is difficult to screen early in everyday life, and the diagnosis method through existing interviews and questionnaires has limitations, so the need for more convenient and objective early screening technology in daily life is emerging.
Accordingly, the professor’s team utilized the children’s wearable data and ADHD and sleep disorder diagnosis results accumulated through research on adolescent brain and cognitive development conducted in the United States.
Specifically, 21 days of wearable data of 5725 children, such as heart rate, number of steps, sleep time, sleep stages, napping, and calories burned, were analyzed based on circadian rhythm, and 12,348 data for ADHD diagnosis model, sleep disorder diagnosis We used 39,160 data points for the model.
As a result of the study, the ADHD diagnostic model showed AUC (closer to 1, higher performance), which evaluates the performance of the model, with a sensitivity of 0.798, a sensitivity of 0.756, and a specificity of 0.716. The specificity was 0.632.
Both models showed a level of performance that enabled early screening using digital phenotypes in daily life, and the professor’s team explained that this provided the basis for early detection and early treatment of ADHD and sleep disorders in children through wearable data.
Professor Chul-Hyun Cho said, “As it is a machine learning diagnosis model using digital phenotypes obtained in everyday life, it will be easy, objective, and early screening and intervention will be possible.” Familiarity with and utilization of digital devices is increasing, which will lead to therapeutic effects when linked with personalized digital treatment services in the future.”
Meanwhile, this study was published in the JAMA Network Open (IF=13.37), an academic journal of the American Medical Association.