What is the story about?
Researchers have developed an artificial intelligence model that can predict one’s risk of developing over a hundred different health conditions using sleep data.
Named 'SleepFM', the model was developed by researchers, including those from the US’ Stanford University, and trained on nearly six lakh hours of sleep data, collected from 65,000 participants.
The AI system, described in a paper in the journal Nature Medicine, was initially tested on standard tasks involving sleep analysis, such as tracking different stages of sleep or diagnosing severity of sleep apnea.
The model was then used to predict the future onset of disease by analysing sleep data, with health record data sourced from a sleep clinic.
More than 1,000 disease categories in the health records were looked at and 130 could be predicted with reasonable accuracy using a patient’s sleep data, the researchers said.
"We record an amazing number of signals when we study sleep. It’s a kind of general physiology that we study for eight hours in a subject who’s completely captive. It’s very data rich,” senior author Emmanual Mignot, professor in sleep medicine at Stanford University’s department of psychiatry and behavioural sciences, said.
Polysomnography — considered the gold standard in sleep studies — is a common means of collecting sleep data that uses sensors to record brain activity, heart function, respiratory signals and eye movements, among other aspects.
The AI model was able to incorporate multiple streams of data — such as electroencephalography (electrical activity of brain), electrocardiography, electromyography (electrical activity of muscles), pulse reading and breathing airflow — and glean how they relate to each other, the researchers said.
The team developed a new technique for training the AI, called 'leave-one-out' contrastive learning. This essentially hides one modality or stream of data and challenges the model to reconstruct the missing piece based on the other signals.
The AI system’s predictions were found to be particularly strong for cancer, pregnancy complications, circulatory conditions and mental disorders, achieving a ’C-index’ score higher than 0.8.
C-index, or concordance index, is a common measure of an AI’s predictive performance — specifically, its ability to predict which of any two individuals in a group will experience an event first, the researchers said.
"From one night of sleep, SleepFM accurately predicts 130 conditions with a C-Index of at least 0.75, including all-cause mortality (C-Index, 0.84), dementia (0.85), myocardial infarction (0.81), heart failure (0.80), chronic kidney disease (0.79), stroke (0.78) and atrial fibrillation (0.78)," the authors wrote.
"Although performance varies across categories, SleepFM demonstrates strong results in several areas, including neoplasms (tumours), pregnancy complications, circulatory conditions and mental disorders," they said.
Strong performance of the AI model was also seen in predicting risk of Parkinson’s disease — in which sleep conditions are considered among the early indicators — and developmental delays and disorders.
Named 'SleepFM', the model was developed by researchers, including those from the US’ Stanford University, and trained on nearly six lakh hours of sleep data, collected from 65,000 participants.
The AI system, described in a paper in the journal Nature Medicine, was initially tested on standard tasks involving sleep analysis, such as tracking different stages of sleep or diagnosing severity of sleep apnea.
The model was then used to predict the future onset of disease by analysing sleep data, with health record data sourced from a sleep clinic.
More than 1,000 disease categories in the health records were looked at and 130 could be predicted with reasonable accuracy using a patient’s sleep data, the researchers said.
"We record an amazing number of signals when we study sleep. It’s a kind of general physiology that we study for eight hours in a subject who’s completely captive. It’s very data rich,” senior author Emmanual Mignot, professor in sleep medicine at Stanford University’s department of psychiatry and behavioural sciences, said.
Polysomnography — considered the gold standard in sleep studies — is a common means of collecting sleep data that uses sensors to record brain activity, heart function, respiratory signals and eye movements, among other aspects.
The AI model was able to incorporate multiple streams of data — such as electroencephalography (electrical activity of brain), electrocardiography, electromyography (electrical activity of muscles), pulse reading and breathing airflow — and glean how they relate to each other, the researchers said.
The team developed a new technique for training the AI, called 'leave-one-out' contrastive learning. This essentially hides one modality or stream of data and challenges the model to reconstruct the missing piece based on the other signals.
The AI system’s predictions were found to be particularly strong for cancer, pregnancy complications, circulatory conditions and mental disorders, achieving a ’C-index’ score higher than 0.8.
C-index, or concordance index, is a common measure of an AI’s predictive performance — specifically, its ability to predict which of any two individuals in a group will experience an event first, the researchers said.
"From one night of sleep, SleepFM accurately predicts 130 conditions with a C-Index of at least 0.75, including all-cause mortality (C-Index, 0.84), dementia (0.85), myocardial infarction (0.81), heart failure (0.80), chronic kidney disease (0.79), stroke (0.78) and atrial fibrillation (0.78)," the authors wrote.
"Although performance varies across categories, SleepFM demonstrates strong results in several areas, including neoplasms (tumours), pregnancy complications, circulatory conditions and mental disorders," they said.
Strong performance of the AI model was also seen in predicting risk of Parkinson’s disease — in which sleep conditions are considered among the early indicators — and developmental delays and disorders.
