Dementia Prediction

Machine learning-based sleep test may help diagnose and predict dementia in older adults

brain eeg

Brain Age Index estimates the difference between a person’s chronological age and the biological age of their brain based on measurements of electrical activity during sleep. Image: chaikom/Getty Images.

Harvard Medical School investigators based at Massachusetts General Hospital and Beth Israel Deaconess Medical Center have developed a machine learning-based sleep test that shows potential as a biomarker associated with unhealthy brain aging, including processes leading to dementia.

The findings, published in JAMA Network Open on Sept. 28, may help clinicians identify patients who have dementia or are at risk of developing it.

Get more HMS news here

The research team recently created the Brain Age Index (BAI), a machine learning model that estimates the difference between a person’s chronological age and the biological age of their brain. The model analyzes electroencephalogram measurements of the electrical activity of an individual’s brain during sleep. A higher BAI signifies deviation from normal brain aging, which could reflect the presence and severity of dementia.

“The model computes the difference between a person’s chronological age and how old their brain activity during sleep ‘looks’ to provide an indication of whether a person's brain is aging faster than is normal,” said senior author M. Brandon Westover, HMS associate professor of neurology and director of data sciences at the Mass General McCance Center for Brain Health.

Sleep EEG tests are increasingly accessible in non-sleep laboratory environments, using inexpensive technologies such as headbands and dry EEG electrodes, the authors said.

“This is an important advance, because before now it has only been possible to measure brain age using brain imaging with magnetic resonance imaging, which is much more expensive, not easy to repeat, and impossible to measure at home,” said study first author Elissa Ye, a member of Westover’s laboratory.

To test whether high BAI values obtained through EEG measurements may be indicative of dementia, the researchers computed values for 5,144 sleep tests in 88 individuals with dementia, 44 with mild cognitive impairment, 1,075 with cognitive symptoms but no diagnosis of impairment and 2,336 without dementia.

BAI values rose across the groups as cognitive impairment increased, and patients with dementia had an average value of about four years older than those without dementia. BAI values also correlated with neuropsychiatric scores from standard cognitive assessments conducted by clinicians before or after the sleep study.

“Because [it is] quite feasible to obtain multiple nights of EEG, even at home, we expect that measuring BAI will one day become a routine part of primary care, as important as measuring blood pressure,” said co–senior author Alice Lam, HMS assistant professor of neurology at Mass General. “BAI has potential as a screening tool for the presence of underlying neurodegenerative disease and monitoring of disease progression.”

Adapted from a Mass General news release.