Researchers, guided by Mount Sinai, have amplified an artificial intelligence (AI) mechanism’s abilities to scrutinize video footage from clinical sleep examinations. As a result, they have improved the precision in diagnosing a globally widespread sleep disorder that affects over 80 million individuals. The research results were published in the Annals of Neurology journal on January 9.
The sleep disorder in question is REM Sleep Behavior Disorder (RBD). It is a condition that provokes abnormal movements or physical enactment of dreams during the phase of sleep characterized by rapid eye movement (REM). When RBD manifests in otherwise healthy adults, it’s referred to as “isolated” RBD. This condition affects more than a million individuals in the United States and is often an early indicator of Parkinson’s disease or dementia in nearly all cases.
Diagnosing RBD is quite challenging as its symptoms can be overlooked or mistaken for other illnesses. A confirmed diagnosis necessitates a sleep study, a video-polysomnogram, conducted by a health professional in a facility equipped with sleep-monitoring technology. However, the data collected in these studies are subjective and can be challenging to interpret universally due to the many complex variables involved, such as sleep stages and muscle activity levels. Although the video data is systematically recorded during a sleep test, it is often disregarded and discarded post-interpretation.
Earlier research suggested that high-end 3D cameras might be required to detect movements during sleep as bedsheets or blankets could obscure the activity. However, this study is the first to present the development of an automated machine learning method that assesses video recordings collected with a standard 2D camera during overnight sleep tests. The method also introduces additional “classifiers” or characteristics of movements, achieving an impressive accuracy rate of nearly 92 percent in detecting RBD.
“This automated approach could be integrated into clinical workflow during the interpretation of sleep tests to enhance and facilitate diagnosis, and avoid missed diagnoses,” said Dr. Emmanuel During, Associate Professor of Neurology (Movement Disorders), and Medicine (Pulmonary, Critical Care, and Sleep Medicine) at the Icahn School of Medicine at Mount Sinai. He added that the method could also guide treatment decisions based on the severity of the movements recorded during sleep tests, ultimately assisting doctors in tailoring patient-specific care plans.
The Mount Sinai group replicated and expanded an automated machine learning analysis proposal for movements during sleep studies developed by researchers at the Medical University of Innsbruck in Austria. Using computer vision, an AI field that enables computers to interpret visual data, they used standard 2D cameras found in sleep labs to monitor patients’ sleep overnight. Analysis of video recordings from about 80 RBD patients and a control group of approximately 90 patients without RBD (either having another sleep disorder or no sleep disruption) was carried out. An automated algorithm calculated the motion of pixels between consecutive frames in a video, enabling the detection of movements during REM sleep. The researchers extracted and analyzed five features of short movements, achieving the highest accuracy to date, at 92 percent.
Researchers from the Swiss Federal Technology Institute of Lausanne (École Polytechnique Fédérale de Lausanne) in Lausanne, Switzerland also contributed to the study by sharing their expertise in computer vision.
Comments are closed for this post.