Predict illness
After the training phase, the researchers were able to fine-tune the model for different tasks.
First, they tested the model on standard sleep analysis tasks, such as classifying different sleep stages and diagnosing sleep apnea severity. SleepFM performed as well or better than the state-of-the-art models used today.
The researchers then tackled a more ambitious goal: predicting future disease onset from sleep data. To identify conditions that could be predicted, they had to match training polysomnography data with long-term health outcomes from the same participants. Fortunately, they had access to more than half a century of health records from a sleep clinic.
THE Stanford Sleep Medicine Center was founded in 1970 by the late William DementMD, PhD, widely considered the father of sleep medicine. The largest cohort of patients used to train SleepFM – some 35,000 patients aged 2 to 96 – had their polysomnographic data recorded at the clinic between 1999 and 2024. Researchers linked these patients’ polysomnographic data to their electronic health records, which provided up to 25 years of follow-up for some patients.
(The clinic’s polysomnographic records go back even further, but only on paper, says Mignot, who directed the sleep center from 2010 to 2019.)
SleepFM analyzed more than 1,000 disease categories in health records and found 130 that could be predicted with reasonable accuracy from a patient’s sleep data. The model’s predictions were particularly strong for cancers, pregnancy complications, circulatory problems and mental disorders, reaching a C index above 0.8.
The C-index, or concordance index, is a common measure of a model’s predictive performance, particularly its ability to predict which of two individuals in a group will experience an event first.
“For all possible pairs of individuals, the model gives a ranking of those who are most likely to experience an event – a heart attack, for example – sooner. A C-index of 0.8 means that 80% of the time the model’s prediction agrees with what actually happened,” Zou said.
SleepFM excelled in predicting Parkinson’s disease (C-index 0.89), dementia (0.85), hypertensive heart disease (0.84), heart attacks (0.81), prostate cancer (0.89), breast cancer (0.87), and death (0.84).
“We were pleasantly surprised to find that, for a fairly diverse set of conditions, the model is able to make informative predictions,” Zou said.
Less precise models, with C indices around 0.7, such as those that predict a patient’s response to different cancer treatments, have proven useful in clinical settings, he added.
Interpretation of the model
The team is working on ways to further improve SleepFM’s predictions, perhaps by adding data from wearable devices, and to understand exactly what the model is interpreting.
“It doesn’t explain this to us in English,” Zou said. “But we developed different interpretation techniques to determine what the model is looking for when it predicts a specific disease.”
The researchers note that while heart signals play a larger role in predicting heart disease and brain signals play a larger role in predicting mental health, it was the combination of all data modalities that yielded the most accurate predictions.
“Most of the information we have obtained to predict disease has been obtained by comparing the different channels,” Mignot said. Components of the body that were out of sync – a brain that appears asleep but a heart that appears awake, for example – seemed to cause problems.
Mignot and Zou are members of the Wu Tsai Neuroscience Institute.
Rahul Thapa, a doctoral student in biomedical data science, and Magnus Ruud Kjaer, a doctoral student at the Technical University of Denmark, are co-lead authors of the study. Thapa is a Knight-Hennessy scholar.
Researchers from the Technical University of Denmark, Copenhagen University Hospital – Rigshospitalet, BioSerenity, University of Copenhagen and Harvard Medical School contributed to the work.
The study received funding from the National Institutes of Health (grant R01HL161253), Knight-Hennessy Scholars, and the Chan-Zuckerberg Biohub.