Veštačka inteligencija samo tokom jedne noći sna prepoznaje rizik od više od 130 bolesti
Na osnovu jedne noći provedene u laboratoriji za spavanje, veštačka inteligencija može da proceni rizik za više od 130 bolesti, od Parkinsonove bolesti do raka dojke. Softver ne otkriva uzroke, već samo korelacije.
TL;DR
- A new AI model, SleepFM, can predict the risk for over 130 diseases from one night of sleep data.
- The AI analyzes signals from polysomnography, including brain waves, heart activity, and breathing.
- Predictions can be made years before the first symptoms of diseases like Parkinson's, dementia, and cancer appear.
- SleepFM is trained on approximately 585,000 hours of sleep recordings from around 65,000 individuals.
- The model identifies statistical correlations, not causal relationships, between sleep patterns and future diagnoses.
- Cardiac signals are important for predicting cardiovascular diseases, while brain signals are key for neurological and psychological disorders.
- The AI's findings can serve as an early warning system, but diagnosis and treatment remain the responsibility of medical professionals.
- Data primarily comes from sleep labs in wealthier regions, leading to underrepresentation of other populations.