The Language of Sleep: New AI Tool Predicts Over 100 Diseases Years Before Symptoms Appear

February 9, 2026

PALO ALTO, CA — Researchers at Stanford Medicine have developed a groundbreaking artificial intelligence model capable of predicting the risk of more than 130 diseases—including various cancers, heart disease, and dementia—by analyzing a single night of sleep data. The model, named SleepFM, marks a significant shift in preventative medicine, suggesting that the “hidden language” of our physiological signals during rest may hold the keys to diagnosing chronic illnesses years before overt symptoms emerge.

Beyond the Snooze: A Foundation Model for Human Physiology

For decades, sleep studies, or polysomnograms (PSG), have been used primarily as diagnostic tools for immediate issues like obstructive sleep apnea or restless leg syndrome. However, a team of researchers led by Dr. Emmanuel Mignot and Dr. James Zou at Stanford University argues that these eight-hour recordings contain a wealth of untapped biological data.

“We record an amazing number of signals when we study sleep,” says Dr. Mignot, a professor of sleep medicine and co-senior author of the study. “It’s a kind of general physiology that we study for eight hours in a subject who’s completely captive. It’s very data-rich.”

While humans and standard software are limited in their ability to process the sheer complexity of these multi-channel recordings, SleepFM uses AI to bridge the gap. Unlike previous AI tools designed for a single task, SleepFM is a foundation model. Much like the technology behind Large Language Models (LLMs), it was trained on nearly 600,000 hours of data from roughly 65,000 individuals to understand the baseline patterns of human biology.

How SleepFM “Learns” Your Body

The researchers trained the model by breaking polysomnography recordings into five-second segments. These recordings include:

• EEG: Brain activity and sleep stages

• ECG: Heart rate and rhythms

• EMG: Muscle activity and movement

• Respiratory Data: Airflow and oxygen levels

To ensure the model truly understood the relationship between these systems, the team used a “blank-filling” training method. The AI would be shown several data streams (like brain waves and breathing) but have one (like heart rhythm) hidden. It then had to reconstruct the missing data based on the others.

“SleepFM is essentially learning the language of sleep,” explains co-senior author James Zou, associate professor of biomedical data science. “One of the technical advances… is to figure out how to harmonize all these different data modalities so they can come together.”

Predicting the Future: 130 Conditions Identified

The most startling revelation of the study, published in Nature Medicine, came when researchers paired sleep data with up to 25 years of electronic health records from the Stanford Sleep Medicine Center.

By scanning over 1,000 disease categories, SleepFM identified 130 conditions it could forecast with significant accuracy. The model achieved a C-index of 0.8 or higher for several major health threats. In statistical terms, a C-index of 0.8 means the model correctly ranked which individual was at higher risk 80% of the time.

Key Diseases Predicted by SleepFM:

The “Out of Sync” Signal

What is the AI actually seeing? Interestingly, the most predictive power didn’t come from any single organ, but from the mismatch between them.

“The most information we got for predicting disease was by contrasting the different channels,” says Mignot. For instance, if the brain signals indicate deep sleep but the heart rhythm suggests a state of high stress or “wakefulness,” the model flags this physiological “desynchronization” as a precursor to future illness.

Expert Perspective and Limitations

While the results are promising, independent experts urge a balanced interpretation. Dr. Sarah Holley, a sleep researcher not involved in the Stanford study, notes that while the predictive power is impressive, “We are still in the early stages of understanding the ‘why.’ An AI can find a pattern, but clinicians need to understand the biological mechanism before we can prescribe a specific intervention.”

Furthermore, the study relied on clinical polysomnography—the gold-standard test conducted in labs. Whether this level of predictive accuracy can be translated to consumer wearables like the Apple Watch or Oura Ring remains to be seen. The Stanford team is currently working on adapting SleepFM to function with more limited data from these everyday devices.

What This Means for You

For the average consumer, this research underscores that sleep is not just “down time,” but a critical window into your long-term health. While you cannot yet ask your doctor for a “SleepFM disease forecast,” the study highlights the importance of:

1. Prioritizing Sleep Hygiene: Consistency in sleep helps keep your internal systems synchronized.

2. Tracking Changes: Significant shifts in your sleep patterns (even if they don’t feel like a “disorder”) may warrant a conversation with a physician.

3. Future Screenings: In the coming years, a routine sleep study may become as common as a blood test for early disease detection.

As AI continues to decode the complex signals of our bodies, the goal is to move from “reactive” medicine—treating a disease once it appears—to “proactive” medicine, where the warning signs are caught years in advance during a simple night’s rest.

References

https://www.earth.com/news/sleep-data-ai-analysis-sleepfm-can-help-predict-disease-risk-years-in-advance/

Medical Disclaimer: This article is for informational purposes only and should not be considered medical advice. Always consult with qualified healthcare professionals before making any health-related decisions or changes to your treatment plan. The information presented here is based on current research and expert opinions, which may evolve as new evidence emerges.