A Mount Sinai-led research team has advanced an AI-powered method to improve the detection of a common sleep disorder, affecting millions worldwide, with implications for diagnosing conditions like Parkinson’s and dementia.
Millions of people worldwide can now potentially receive quicker and more accurate diagnoses for a common but challenging sleep disorder, thanks to a groundbreaking advancement by a team of researchers led by Mount Sinai. This pioneering effort has enhanced an AI-powered algorithm capable of analyzing video recordings of clinical sleep tests. The development, published in the Annals of Neurology, promises to revolutionize the identification of REM sleep behavior disorder (RBD).
RBD is a condition that involves abnormal movements or the physical acting out of dreams during the rapid eye movement (REM) phase of sleep. Isolated RBD, affecting over 1 million people in the United States alone, often serves as an early indicator of severe neurological conditions, such as Parkinson’s disease or dementia.
However, its diagnosis has been notoriously difficult, requiring the subjective review of sleep studies by health care professionals.
Previous studies hinted that advanced 3D cameras might be necessary to capture the subtleties of RBD movements, especially with bedding obscuring the view.
However, this groundbreaking Mount Sinai-led effort, in collaboration with the Medical University of Innsbruck and contributions from the Swiss Federal Technology Institute of Lausanne, has shattered this notion.
By utilizing conventional 2D cameras available in clinical sleep labs, the researchers developed an automated machine learning method capable of interpreting standard video-polysomnographic data.
“This automated approach could be integrated into clinical workflow during the interpretation of sleep tests to enhance and facilitate diagnosis, and avoid missed diagnoses,” lead author Emmanuel During, M.D., an associate professor of neurology and medicine at the Icahn School of Medicine at Mount Sinai, said in a news release.
He emphasized that the method could aid in tailoring treatment plans based on the severity of movements observed during sleep tests, ultimately allowing for more personalized patient care.
The study involved analyzing sleep recordings from approximately 80 patients diagnosed with RBD and around 90 controls, who either had a different sleep disorder or no sleep disruptions. Using AI to detect pixel motion between video frames, the team evaluated various factors, such as the rate and magnitude of movements, achieving an unprecedented accuracy rate of 92%.
The implementation of this technology in clinical settings could facilitate a more efficient diagnostic process, significantly reducing the risk of misdiagnosis and allowing earlier intervention for conditions closely associated with RBD.
The potential impact of these findings extends beyond individual diagnoses. As the health care industry continues to integrate AI and machine learning technologies, improvements in diagnostic accuracy and personalizing treatments could become commonplace.