In the dim sleep lab in Mount Sinai, the machine quietly records the mind, breathing patterns and the heart rate of the sleeping patients. Now, newly developed AI systems can process this data in an unprecedented way, rather than traditional fragmented approaches to overall sleep.
Researchers at the Icahn School of Medicine have launched an AI tool that has processed an astonishing 1,011,192 hours of human sleep, making it one of the largest sleep analytical studies ever. The results of the study are published in the journal sleep March 13.
The model, known as the “basic transformer of sleep” (PFTSLEEP), represents a significant deviation from conventional sleep analysis methods. Traditional approaches often involve human experts manually scoring short segments of sleep data or using AI models limited to analyzing short intervals.
“This is a step forward in AI-assisted sleep analysis and interpretation,” said Benjamin Fox, first author and doctoral candidate at the Icahn School of Medicine at Mount Sinai. “By leveraging AI in this way, we can learn relevant clinical features directly from sleep study signal data and use them for sleep scores and use other clinical applications in the future, such as detecting sleep apnea or assessing health risks associated with sleep quality.”
The system is built on the same transformer architecture for language models such as Chatgpt, analyzing brain waves, muscle activity, heart rate, and breathing patterns throughout the night. This comprehensive approach allows it to identify patterns and relationships that may be missed when checking only the isolated 30-second segment.
The technology arrives when sleep disorders affect nearly one-third of Americans, and many cases are not diagnosed due to the complex and time-consuming nature of sleep analysis. Sleep experts spend hours reviewing multiple illustrations (capturing records of multiple physiological signals during sleep) to diagnose conditions such as sleep apnea or insomnia.
Dr. Ankit Parekh, co-training author and assistant professor of medicine at Icahn Medical School, sees a wider application. “Our findings show that AI can change the way we learn and understand sleep,” he explained. “Our next goal is to refine technologies for clinical applications, such as more efficiently identifying health risks associated with sleep.”
The reason for setting this model is its training method. The researchers not only rely on human-labeled data, but instead adopt “self-learning”, allowing AI to independently identify relevant patterns. This approach helps the system learn directly from physiological signals, potentially revealing subtle relationships that human experts may ignore.
The team trained PFTSLEEP using three major sleep research databases: Sleep Heart Health Study, Wisconsin Sleep Cohort and Men’s Study of Osteoporosis Fractures. They then verified their performance against independent datasets, achieving an impressive accuracy score. The model’s Kappa score (a measure of inter-evaluator reliability) ranged from 0.59 to 0.81 in different test sets, comparable to expert human analysis.
Traditionally, sleep medicine has struggled with the problem of standardization. Different sleep centers and individual experts may interpret the same sleep data in different ways, resulting in inconsistent diagnosis. Systems like PftSleep can help solve this problem by providing a more unified analysis method.
“By analyzing the entire sleep with greater consistency, we can gain a deeper understanding of sleep health and its connection to overall well-being,” noted Dr. Girish N. Nadkarni, collaborator and chair of the Windreich Department of Artificial Intelligence and Human Health.
Dr. Nadkarni, who also serves as director of the Hasso Plattner Institute for Digital Health, stressed that this AI-driven approach could revolutionize sleep research, although he warned that the technology is designed to enhance rather than replace clinical expertise.
The meaning goes beyond sleep disorders. There is growing evidence linking sleep quality to cardiovascular health, cognitive function, and immune system performance. More effective sleep analyses can help researchers better understand these connections and potentially identify early warning signs for health problems.
For the average person struggling with sleep problems, tools like PFTSleep may eventually lead to more accessible sleep assessments. Multiple current terms usually require overnight stays in a professional facility, which makes it inconvenient and expensive. AI systems that can effectively analyze sleep data may support home-based alternatives in the future.
Researchers are already seeking to extend the functionality of the model. In addition to classifying sleep stages, they also hope to develop systems that can detect specific sleep disorders and even predict health outcomes based on how they sleep.
As AI continues to penetrate healthcare, this advance in sleep analytics shows how emerging technologies can cope with long-standing clinical challenges. Whether this leads to a better diagnosis, more personalized treatments, or just a better night’s sleep remains to be seen, but for the moment, AI is learning to observe you while you’re sleeping.
If you find this piece useful, consider supporting our work with a one-time or monthly donation. Your contribution allows us to continue to bring you accurate, thought-provoking scientific and medical news that you can trust. Independent reporting requires time, effort, and resources, and your support makes it possible for us to continue exploring stories that are important to you. Together, we can ensure that important discoveries and developments attract those who need them the most.
