The model, called SleepFM, was developed by Stanford Medicine<\/a> researchers and collaborators. <\/p>\n It was trained on nearly 600,000 hours of polysomnography data from about 65,000 people, using the kind of overnight sleep<\/a> study that tracks the brain, heart, breathing, movement, and more.<\/p>\n Mining data with SleepFM<\/p>\n Polysomnography is often treated as a clinical tool: you do the study, score sleep stages, look for sleep apnea, and move on. The team argues that\u2019s only a small slice of what these recordings contain.<\/p>\n \u201cWe record an amazing number of signals when we study sleep,\u201d said co-senior author Emmanuel Mignot, a professor of sleep medicine at Stanford.\u00a0<\/p>\n \u201cIt\u2019s a kind of general physiology that we study for eight hours in a subject who\u2019s completely captive. It\u2019s very data rich.\u201d<\/p>\n The problem, until recently, was that humans and standard software could only digest so much of that complexity. <\/p>\n AI changes that equation, at least in principle, by learning patterns across thousands of nights and multiple body systems at once.<\/p>\n Sleep studies with AI <\/p>\n Medical AI has been booming in fields like radiology and cardiology. Sleep has lagged behind, even though it sits at the intersection of brain function, metabolism, breathing, and cardiovascular health<\/a>.<\/p>\n Study co-senior author James Zou is an associate professor of biomedical data science.<\/p>\n \u201cFrom an AI perspective, sleep is relatively understudied. There\u2019s a lot of other AI work that\u2019s looking at pathology or cardiology, but relatively little looking at sleep, despite sleep being such an important part of life,\u201d said Zou.<\/p>\n That gap shaped the team\u2019s approach. Instead of building a model for a single task, they developed a foundation model designed to learn broad patterns first and adapt to specific predictions later.<\/p>\n SleepFM and the language of sleep<\/p>\n SleepFM was trained like a large language model, but instead of words, it learned from tiny slices of physiology. <\/p>\n The polysomnography recordings were chopped into five second segments, so the model could process long nights as sequences and learn what normally follows what. \u201cSleepFM is essentially learning the language of sleep,\u201d Zou said.<\/p>\n The model pulled in multiple channels at once, including signals such as electroencephalography for brain activity<\/a>, electrocardiography for heart rhythms, electromyography for muscle activity, plus pulse and airflow data.\u00a0<\/p>\n Training a model for reliability <\/p>\n The goal wasn\u2019t just to read each channel. It was to understand how the channels relate to one another.<\/p>\n To do that, the researchers created a training method designed to make the model fill in blanks. One stream of data would be hidden, and the model would have to reconstruct it from the others.<\/p>\n \u201cOne of the technical advances that we made in this work is to figure out how to harmonize all these different data modalities so they can come together to learn the same language,\u201d Zou said.<\/p>\n After training, the team fine-tuned SleepFM for familiar sleep medicine tasks. They tested whether it could classify sleep stages and assess sleep apnea<\/a> severity, among other standard measures.<\/p>\n On those benchmarks, the system performed as well as or better than leading models already used in the field. <\/p>\n That step mattered because it suggested the model wasn\u2019t just learning noise. It could do the basics reliably before being asked to do something more ambitious.<\/p>\n Sleep data and disease risk<\/p>\n Then came the real swing: forecasting future disease from one night\u2019s sleep. To do this, the researchers paired sleep data with long-term medical outcomes, using decades of patient records from a major sleep clinic.<\/p>\n The Stanford Sleep Medicine Center was founded in 1970 by the late William Dement. For this project, the largest dataset came from about 35,000 patients aged 2 to 96 whose polysomnography tests were recorded between 1999 and 2024.\u00a0<\/p>\n The team matched those sleep studies to electronic health records, giving up to 25 years of follow-up data for some individuals.<\/p>\n SleepFM scanned more than 1,000 disease categories and identified 130 that it could predict with reasonable accuracy using sleep data alone.\u00a0<\/p>\n Predicting diseases years ahead<\/p>\n The strongest results were reported for cancers<\/a>, pregnancy complications, circulatory conditions, and mental disorders, with a C-index above 0.8 in those groups.<\/p>\n The C-index is a way to score how well a model ranks risk across people. It\u2019s not about certainty for one person. It\u2019s about whether the model tends to place higher-risk individuals above lower-risk ones.<\/p>\n \u201cFor all possible pairs of individuals, the model gives a ranking of who\u2019s more likely to experience an event \u2013 a heart attack, for instance \u2013 earlier. A C-index of 0.8 means that 80% of the time, the model\u2019s prediction is concordant with what actually happened,\u201d Zou said.<\/p>\n The model performed especially well for several specific outcomes, including Parkinson\u2019s disease, dementia<\/a>, hypertensive heart disease, heart attack, prostate cancer, breast cancer, and death.<\/p>\n \u201cWe were pleasantly surprised that for a pretty diverse set of conditions, the model is able to make informative predictions,\u201d Zou said.<\/p>\n What the SleepFM model \u201csees\u201d<\/p>\n Even with strong performance numbers, the obvious question remains: what exactly is SleepFM picking up? The team says they\u2019re working on interpretation tools and may also try improving predictions by adding data from wearables.<\/p>\n \u201cIt doesn\u2019t explain that to us in English,\u201d Zou said. \u201cBut we have developed different interpretation techniques to figure out what the model is looking at when it\u2019s making a specific disease prediction.\u201d<\/p>\n One pattern already stands out. The most accurate predictions didn\u2019t come from a single channel. They came from comparing channels and spotting mismatches.<\/p>\n \u201cThe most information we got for predicting disease was by contrasting the different channels,\u201d Mignot said.<\/p>\n In other words, it may be the body being out of sync that signals trouble. A brain that looks asleep while the heart looks \u201cawake,\u201d for example, could hint that something deeper is off.<\/p>\n The study is published in the journal Nature Medicine<\/a>.<\/p>\n \u2014\u2013<\/p>\n Like what you read? Subscribe to our newsletter<\/a> for engaging articles, exclusive content, and the latest updates.<\/p>\n Check us out on EarthSnap<\/a>, a free app brought to you by Eric Ralls<\/a> and Earth.com.<\/p>\n \u2014\u2013<\/p>\n","protected":false},"excerpt":{"rendered":"One bad night\u2019s sleep can result in a foggy brain for the entire day. But new research suggests…\n","protected":false},"author":2,"featured_media":274475,"comment_status":"","ping_status":"","sticky":false,"template":"","format":"standard","meta":{"footnotes":""},"categories":[20],"tags":[365,363,364,111,139,69,145],"class_list":{"0":"post-274474","1":"post","2":"type-post","3":"status-publish","4":"format-standard","5":"has-post-thumbnail","7":"category-artificial-intelligence","8":"tag-ai","9":"tag-artificial-intelligence","10":"tag-artificialintelligence","11":"tag-new-zealand","12":"tag-newzealand","13":"tag-nz","14":"tag-technology"},"_links":{"self":[{"href":"https:\/\/www.newsbeep.com\/nz\/wp-json\/wp\/v2\/posts\/274474","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/www.newsbeep.com\/nz\/wp-json\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/www.newsbeep.com\/nz\/wp-json\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/www.newsbeep.com\/nz\/wp-json\/wp\/v2\/users\/2"}],"replies":[{"embeddable":true,"href":"https:\/\/www.newsbeep.com\/nz\/wp-json\/wp\/v2\/comments?post=274474"}],"version-history":[{"count":0,"href":"https:\/\/www.newsbeep.com\/nz\/wp-json\/wp\/v2\/posts\/274474\/revisions"}],"wp:featuredmedia":[{"embeddable":true,"href":"https:\/\/www.newsbeep.com\/nz\/wp-json\/wp\/v2\/media\/274475"}],"wp:attachment":[{"href":"https:\/\/www.newsbeep.com\/nz\/wp-json\/wp\/v2\/media?parent=274474"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/www.newsbeep.com\/nz\/wp-json\/wp\/v2\/categories?post=274474"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/www.newsbeep.com\/nz\/wp-json\/wp\/v2\/tags?post=274474"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}
\n ![\"EarthSnap\"\/](\"https:\/\/www.newsbeep.com\/nz\/wp-content\/uploads\/2026\/01\/earthsnap-banner-news.webp.webp\")
\n<\/a><\/p>\n