Routine sleep studies generate far more data than clinicians typically use to make treatment decisions. Now, an artificial intelligence model has been developed to extract long-term health risk information from that same data, according to News-Medical.
Polysomnography, the standard overnight sleep study, records brain activity, breathing patterns, oxygen levels, heart rate, and body movements. It is most commonly used to diagnose conditions like sleep apnea. The AI model in this research was designed to look beyond those immediate diagnostic signals and identify patterns connected to patients' future health.
The model was trained on data from large numbers of patients who had undergone sleep studies and were then followed over time. By comparing the sleep data to what happened to those patients medically in the years afterward, the AI learned to recognize patterns that predict elevated risk for conditions such as cardiovascular disease and other serious long-term health problems.
One of the central advantages researchers cited is that the data required already exists. Patients who have had sleep studies do not need additional tests for the model to assess their long-term risk. The AI can work with recordings that were already collected for other purposes.
Sleep medicine researchers have long observed that poor sleep quality and breathing disruptions during sleep are associated with a range of serious health outcomes. The challenge has been translating that general association into specific, actionable risk information for individual patients. A model that can produce individual risk assessments from existing data would move that research from the population level to the clinical level.
The model's ability to identify high-risk patients earlier could give clinicians a window to intervene before those conditions develop or worsen. Early intervention in cardiovascular disease, for example, can significantly alter a patient's trajectory.
The research represents a broader trend in medicine toward finding additional value in data that is already being collected. Sleep studies are expensive and time-consuming for patients. Getting more clinical information from a test a patient has already completed fits with efforts to reduce redundant testing while improving preventive care.
Researchers indicated that further validation of the model across different patient populations would be a necessary step before it could be widely adopted in clinical practice.
