0
0
January 9, 2025 — In a groundbreaking development, a research team led by Mount Sinai has refined an artificial intelligence (AI)-powered algorithm to analyze video recordings from clinical sleep tests. The advancement promises to improve the accurate diagnosis of REM sleep behavior disorder (RBD), a condition affecting over 80 million people globally. The findings were published today in the Annals of Neurology.
RBD is a sleep disorder characterized by abnormal movements or physically acting out dreams during the REM phase of sleep. It affects more than one million Americans and is often an early indicator of serious neurological conditions such as Parkinson’s disease or dementia. Diagnosing RBD is notoriously challenging due to its subtle symptoms, which are frequently mistaken for other conditions. A definitive diagnosis currently requires a video-polysomnogram conducted in specialized facilities, where subjective data interpretation can lead to missed or delayed diagnoses.
Mount Sinai researchers have revolutionized this process by developing an automated machine learning algorithm that analyzes standard 2D video recordings from overnight sleep studies. This approach eliminates the need for costly 3D cameras, previously thought necessary to capture movements obscured by blankets or sheets.
The novel AI method, which identifies detailed movement features like rate, magnitude, and velocity, achieved a groundbreaking accuracy rate of nearly 92% in detecting RBD. Dr. Emmanuel During, corresponding author and Associate Professor at Mount Sinai, emphasized the transformative potential of this technology:
“This automated approach could be integrated into clinical workflows during sleep test interpretations to enhance and facilitate diagnosis, and avoid missed diagnoses. It could also guide treatment decisions based on movement severity, helping doctors personalize care plans for individual patients.”
The Mount Sinai team expanded on previous research from the Medical University of Innsbruck in Austria by incorporating computer vision technology—an AI discipline for analyzing visual data—into their analysis. The study evaluated video data from 80 RBD patients and a control group of 90 patients with other sleep disorders or no disruptions.
Using motion detection algorithms, the AI analyzed pixel changes between video frames to quantify movements during REM sleep. The result is a cost-effective and highly accurate tool that leverages existing 2D camera setups, making it widely applicable in clinical sleep labs.
Collaborators from the Swiss Federal Technology Institute of Lausanne contributed their expertise in computer vision to enhance the algorithm’s capabilities.
The implications of this innovation extend beyond diagnosis. By quantifying movement severity, the tool could help tailor treatment plans, providing a more personalized approach to managing RBD and related conditions.
For more details, refer to the original study: Automated Detection of Isolated REM Sleep Behavior Disorder Using Computer Vision, published in Annals of Neurology (2025).
Journal Reference:
Annals of Neurology, 2025.
