AI Can Predict Diseases From Just One Night of Sleep, Researchers Say

 

A single night of sleep may soon reveal far more about your health than you ever imagined.

Researchers have developed a new artificial intelligence (AI) model that can predict the risk of dozens of serious diseases—using data from just one night of sleep. No blood tests. No scans. Just sleep.

The findings, published by researchers at Stanford Medicine, are already being described as a potential breakthrough in preventive healthcare.

Why This Matters Right Now

Most serious diseases don’t start suddenly. Heart disease, dementia, kidney failure, and even some cancers develop quietly over many years.

The problem is timing.

Doctors usually find out about these health problems after the symptoms show up. At this point the damage has already started. This makes it harder to treat. It costs more money. The treatment is also not as good, at making people better. Doctors have a time treating these conditions when they find out about them late. These health problems are the conditions that doctors are trying to treat.

This new AI model aims to change that by spotting early warning signs—long before a person feels sick.

The AI Model Behind the Discovery

The system, called SleepFM, was developed by researchers at Stanford Medicine.

It was trained on an enormous dataset:

• Nearly 600,000 hours of sleep recordings
• Data from around 65,000 people
• Clinical-grade sleep studies known as polysomnography

Unlike fitness trackers that record limited metrics, these sleep studies capture deep physiological signals, including:

• Brain activity
• Heart rhythms
• Breathing patterns
• Muscle and eye movement

The AI doesn’t analyze these signals separately. Instead, it studies how they interact—revealing patterns that humans can’t easily detect.

What One Night of Sleep Can Reveal

According to the researchers, SleepFM can estimate the risk of more than 100 health conditions from a single night of sleep data.

These include:

• Heart attack and heart failure
• Stroke and circulation disorders
• Dementia and Parkinson’s disease
• Chronic kidney disease
• Certain cancers
• Mental health conditions
• Overall risk of early death

The model does not diagnose disease. Instead, it predicts long-term risk—offering a potential early warning system.

How Accurate Is It?

To measure performance, researchers used a standard medical metric called the C-index.

In simple terms:

• Scores closer to 1.0 indicate stronger prediction
• Scores above 0.8 are considered highly reliable

SleepFM achieved:

• Around 0.85 when predicting dementia risk
• Above 0.8 for multiple cardiovascular conditions

These results are comparable to—and in some cases better than—traditional clinical risk models.

Why Sleep Is the Key Signal

Sleep is really interesting because it shows us what our body does when it is not under a lot of stress. Our body functions, in a different way when we are sleeping. Sleep gives us an idea of how our body works when things are calm and quiet.

When we are sleeping our heart beats in a different way we breathe a little differently and our brain works differently too. These small changes can show that something is not quite right with our body like our heart, lungs and brain are not working properly. The heart rhythm, breathing and brain activity can all be. This can happen before we even notice anything is wrong, with our heart, lungs, brain and other organs.

Until now, much of this information remained hidden because humans simply couldn’t analyze such complex, high-volume data. AI can.

Does This Mean AI Will Replace Doctors?

Not at all.

The researchers emphasize that SleepFM is designed to support, not replace, medical professionals.

Doctors would still:

• Interpret results
• Decide follow-up tests
• Consider patient history and lifestyle
• Make treatment decisions

AI provides the signal. Humans provide the judgment.

Important Limitations

Despite its promise, the technology is still in the research phase.

Key limitations include:

• The need for real-world clinical validation
• Predictions indicate risk, not certainty
• Strict privacy protections will be essential

Experts caution that the model is not yet ready for routine medical use, but its potential is significant.

What This Could Mean for the Future

If validated and widely adopted, this technology could reshape healthcare.

Instead of reacting to disease, doctors could focus on prevention.

One night of sleep data could one day:

• Trigger early lifestyle interventions
• Guide preventive treatment plans
• Reduce healthcare costs
• Improve long-term outcomes

The Bottom Line

For decades, sleep has been treated as rest.

This research suggests it may also be one of the most powerful diagnostic signals we have.

With AI now capable of reading what sleep reveals, healthcare could shift from late-stage reaction to early, quiet prevention.

And all of it might begin while you’re asleep.

 

You May Also Like

Loading...