Earlier, you would come to know about a disease only when your condition became very serious. But with advances in science, it is now possible to find out what diseases you have, how serious they are, and what diseases you may develop in the future. Researchers have created an artificial intelligence model that can predict a person’s future risk of 130 diseases based on sleep data. This model is called SleepFM. It has been created by researchers from many institutions including America’s Stanford University. It was trained on approximately 600,000 hours of sleep data collected from 65,000 people. The results of this research have been published in the medical journal Nature Medicine. Let us understand how this device works and how it can predict what diseases you may get in the future.
How does SleepFM work?
Initially, this AI system was tested to identify common sleep problems, such as tracking different sleep stages or determining the severity of sleep apnea. Then, the sleep data was combined with the patients’ medical records to see which diseases they might be at risk for in the future. Researchers said that this model was able to predict 130 diseases out of more than 1,000 diseases present in health records with great accuracy.
Important signs of health are hidden in sleep
According to Emmanuel Mignot, professor of sleep medicine at Stanford University, “During sleep, many signals from the body are recorded. The body’s normal activities are studied in such detail for eight hours that the data becomes extremely rich.”
What kind of data is collected?
Polysomnography was used for sleep analysis, which is considered the most reliable method for studying sleep. Using sensors, it records a number of signals, including:
brain activity
heartbeat
breathing pattern
eye movement
Muscle Activity SleepFM analyzes all these data streams together, and understands their relationships.
A new way to train AI
The team adopted a technique called ‘leave-one-out’ contrastive learning to train the AI. In this approach, one type of data is intentionally hidden, and the AI is challenged to infer the missing information based on the rest of the signal. This improves the understanding and accuracy of the model.
What diseases does it best identify? Research has found that this AI is particularly good at predicting a number of diseases, including:
cancer
Pregnancy related problems
diseases related to heart and blood flow
mental health problems
In many cases, its C-index score was greater than 0.8, indicating good predictive accuracy. According to researchers, the diseases for which SleepFM can predict risk from just one night of sleep data include:
dementia
heart attack
heart failure
chronic kidney disease
stroke
atrial fibrillation
Apart from this, this model also proved to be effective in predicting the risk of diseases like Parkinson’s and developmental problems in children. Overall, this research shows that your sleep is not just a way to relieve fatigue, but it is also a mirror of your future health.