Key Takeaways:
- A new artificial intelligence model called SleepFM can estimate a person’s risk for around 130 diseases later in life based on one night’s sleep data.
- The model uses polysomnography data, which measures brain waves, heart activity, breathing, muscle tension, and eye and leg movements during sleep.
- SleepFM can make predictions years before the first symptoms become evident, including risks for Parkinson’s disease, dementia, heart disease, and prostate and breast cancer.
- The model is trained on nearly 600,000 hours of sleep data collected from 65,000 sleepers and can detect patterns in the data to identify potential health risks.
- SleepFM’s predictions are primarily based on data from sleep labs, which may not be representative of the general population, particularly those from less affluent areas.
Introduction to SleepFM
The SleepFM model is a groundbreaking artificial intelligence system that can estimate a person’s risk for various diseases later in life based on just one night’s sleep data. According to James Zou, a co-senior author of the SleepFM study, "Our results reveal that many conditions — including stroke, dementia, heart failure and all-cause mortality — are highly predictable from sleep data, further reinforcing the potential of sleep as a powerful biomarker for long-term health." The model uses polysomnography data, which measures brain waves, heart activity, breathing, muscle tension, and eye and leg movements during sleep. This data is collected from sleep labs, where patients are referred for sleep problems, and is used to train the SleepFM model.
Training the SleepFM Model
The SleepFM model was trained on nearly 600,000 hours of sleep data collected from 65,000 sleepers. The data was primarily collected from Stanford University’s Sleep Medicine Center in California, US. The model was first shown signals from the brain, heart, and body during normal sleep, with "normal" averages calculated statistically. Then, it was taught about the different stages of sleep as well as sleep apnea, a disorder where breathing repeatedly stops and starts during sleep. The researchers then connected the sleep data with electronic health records going back 25 years and examined how later health diagnoses correlated with the measurements from polysomnography. As Rahul Thapa, a PhD student in biomedical data science and co-lead author of the paper, notes, "In principle, an AI model can be trained for a very large number of possible predictions, provided the basic data is available."
Predicting Diseases with SleepFM
The SleepFM model can detect patterns in the data and identify potential health risks, including risks for Parkinson’s disease, dementia, heart disease, and prostate and breast cancer. The model can make predictions years before the first symptoms become evident, which could lead to early interventions and improved health outcomes. According to Sebastian Buschjäger, a German expert on machine learning, "An AI model can be trained for planning therapy but it’s humans — the doctors — who interpret the results and choose the therapy, often without knowing all the underlying causes." The model’s predictions are primarily based on data from sleep labs, which may not be representative of the general population, particularly those from less affluent areas.
Limitations and Future Directions
While the SleepFM model shows great promise, there are limitations to its current implementation. The model is trained on data from sleep labs, which may not be representative of the general population. Additionally, the model can only identify correlations between sleep data and disease risk, but cannot determine the underlying causes of the disease. As Matthias Jakobs, a computer scientist, notes, "Most AI methods do not learn causal relationships." However, the model can still provide valuable insights and support medical professionals in evaluating sleep data. Future research directions include expanding the model to include data from more diverse populations and exploring the potential for sleep diagnosis to go beyond current correlations between polysomnography and disease prediction.
Interdisciplinary Collaboration
The development of the SleepFM model highlights the importance of interdisciplinary collaboration between AI researchers, medical professionals, and sleep experts. As Buschjäger notes, "If our colleagues in sleep medicine suspect a connection, we AI specialists can incorporate this into a predictive system, and conversely, [we can] provide indications of where connections might exist." The collaboration between AI researchers and medical professionals can lead to a better understanding of the relationships between sleep and disease, and ultimately improve health outcomes. As Jakobs writes, "Models like SleepFM can record sleep stages or sleep apneas more efficiently, which allows doctors to spend more time with their patients."
https://www.dw.com/en/how-ai-can-detect-health-risks-just-from-the-way-you-sleep/a-75461969