A team of researchers led by Mount Sinai has significantly advanced an artificial intelligence (AI)-powered algorithmdesigned to analyze video recordings from clinical sleep tests, enhancing the accuracy of diagnosing REM Sleep Behavior Disorder (RBD)—a common sleep disorder affecting over 80 million people worldwide. This breakthrough, published in the journal Annals of Neurology on January 9, promises to improve diagnostic precision and aid early detection of Parkinson’s disease and dementia, conditions often heralded by RBD.

RBD is characterized by abnormal movements or the acting out of dreams during the REM phase of sleep. When it occurs in otherwise healthy individuals, it is referred to as “isolated RBD,” which affects more than one million people in the United States alone. Nearly all cases of isolated RBD are early indicators of neurodegenerative conditions such as Parkinson’s disease or dementia.

Diagnosing RBD has long been a challenge for clinicians due to the disorder’s subtle symptoms, which can go unnoticed or be misdiagnosed as other conditions. A definitive diagnosis requires a video-polysomnogram (v-PSG), a type of sleep study that records video alongside other physiological data. However, interpreting the video data is complex because it involves multiple variables such as sleep stages, muscle activity, and movement patterns. The data is often subjective, and video recordings, which could provide crucial insight into the disorder, are frequently overlooked or discarded after the test.

Previous studies in this area have suggested that specialized 3D cameras might be necessary to detect movements during sleep, as blankets and sheets often obscure them. However, the new study conducted by Mount Sinai researchers is the first to introduce an automated machine-learning method that analyzes video collected by standard 2D cameras—which are already commonly used in clinical sleep studies. This method identifies additional movement features, or “classifiers,” improving the accuracy of RBD detection to nearly 92%.

Dr. Emmanuel During, the corresponding author and Associate Professor of Neurology at Mount Sinai’s Icahn School of Medicine, emphasized that this automated approach could revolutionize the way sleep tests are interpreted in clinical settings. “This method could be integrated into the clinical workflow during the interpretation of sleep tests to enhance diagnosis and reduce missed cases,” he said.

In addition to improving diagnostic accuracy, the AI algorithm could play a role in treatment decisions by assessing the severity of movements during sleep. This would allow doctors to tailor care plans based on the individual severity of RBD symptoms, leading to more personalized patient care.

Building on a proposal for automated machine-learning analysis developed by researchers at the Medical University of Innsbruck in Austria, the Mount Sinai team used computer vision—a field of AI that enables computers to analyze visual data, such as images and videos. The researchers utilized routine 2D cameras, which are already standard in most sleep labs, to monitor patients’ movements during sleep.

The team analyzed data from 80 RBD patients and 90 control subjects without RBD (either healthy or suffering from other sleep disorders). The AI algorithm tracked the motion of pixels between consecutive video frames, which allowed it to detect movements during REM sleep. By analyzing five key movement features—ratemagnitudevelocityimmobility ratio, and movement ratio—the researchers achieved 92% accuracy in detecting RBD, setting a new benchmark in sleep study analysis.

The innovative approach developed by Mount Sinai’s research team not only improves the diagnosis of REM sleep behavior disorder but also marks a significant leap in the use of AI for sleep medicine. As AI technology continues to evolve, its potential to enhance the diagnosis and treatment of a wide range of sleep disorders remains promising, offering hope for millions of patients worldwide.

This breakthrough represents a major step forward in the effort to diagnose early and treat more effectively complex sleep disorders like RBD, with the ultimate goal of improving outcomes for individuals at risk of developing debilitating neurodegenerative diseases like Parkinson’s and dementia.

By Impact Lab