Summary: Researchers have created a machine-learning model that estimates a person’s “brain age” from their sleep EEG. The model identifies subtle brain-wave features that predict future dementia risk, suggesting sleep recordings could become a noninvasive early-detection tool for cognitive decline years before clinical symptoms appear.
The study shows that when a person’s EEG-derived brain age is older than their chronological age, their risk of developing dementia increases substantially. Conversely, EEG signatures indicating a younger brain age are linked to lower dementia risk. These associations were discovered by analyzing fine-scale sleep microstructure rather than conventional sleep measures, highlighting the potential of sleep EEG and machine learning for dementia risk screening.
Key Facts & Statistics
- The “Brain Age” Gap: Each 10-year increase in estimated brain age relative to actual age corresponded to a nearly 40% higher risk of dementia.
- Protective Youthfulness: Participants whose brain waves appeared younger than their chronological age faced a substantially lower risk of dementia.
- Large-Scale Data: The analysis pooled overnight EEG recordings from about 7,100 participants (ages 40–94) across five community cohorts, with follow-up up to 17 years.
- Invisible Patterns: Traditional sleep metrics such as total sleep time or sleep stages showed no clear link to dementia; the signal came from 13 microstructural EEG features identified by the machine-learning algorithm.
- Critical EEG Markers:
- Delta Waves: Linked to deep, restorative sleep.
- Sleep Spindles: Short bursts of activity important for memory consolidation.
- Kurtosis: Sudden large spikes in EEG activity that were, unexpectedly, associated with a lower dementia risk.
Source: UCSF
Overview: A collaborative study led by UC San Francisco and Beth Israel Deaconess Medical Center used interpretable machine learning to derive a sleep EEG brain age index (BAI). The model integrates 13 microstructural EEG features from central channels in overnight polysomnography and estimates the difference between EEG-based brain age and chronological age. That difference—BAI—was then tested for association with incident dementia in community-dwelling adults.
Across the pooled cohorts, researchers observed that each 10-year increase in sleep EEG–based brain age compared with chronological age was associated with a 39% higher hazard of developing dementia after adjusting for demographic and lifestyle factors. Associations remained robust after adjusting for comorbidities, sleep-disordered breathing measures, and genetic risk (APOE ε4), and were consistent across sex and age groups.
The cohorts included in this individual participant data meta-analysis were the Multi-Ethnic Study of Atherosclerosis (MESA), the Atherosclerosis Risk in Communities (ARIC) study, the Framingham Heart Study–Offspring Study (FHS-OS), the Osteoporotic Fractures in Men Study (MrOS), and the Study of Osteoporotic Fractures (SOF). Participants (n ≈ 7,105) were free of dementia at baseline, ranged in age from 40 to 94, and were followed for between 3.5 and 17 years. During follow-up, roughly 1,000 participants developed dementia or probable dementia.
Why sleep EEG? Sleep is a time when the brain performs essential maintenance, including clearance of toxic proteins linked to Alzheimer’s disease. Fine-grained EEG microstructure captures aspects of these restorative processes that traditional sleep stage metrics do not. As the authors note, broad sleep measures fail to represent the complex, multidimensional physiology of sleep that machine learning can extract from EEG.
Brain-Wave Patterns and Cognitive Health
Several of the EEG microfeatures that drive the brain age estimate are well-known to relate to memory and brain health. Delta waves reflect deep sleep and restorative processes; sleep spindles support memory consolidation; and unexpectedly, higher kurtosis—indicating brief large-amplitude transients—was associated with lower dementia risk in this analysis. These microstructural markers provided predictive value above and beyond typical sleep measures.
Potential for Early Detection and Applications
Because sleep EEG can be recorded noninvasively, the sleep EEG brain age index may be deployable as a digital biomarker for early dementia risk screening in community settings, including through wearable or home-based devices as technologies advance. The model’s interpretable design also helps highlight which EEG features matter most for risk prediction.
The researchers emphasize that an “older” brain age is not a current diagnosis of dementia but a predictive signal indicating accelerated brain aging. Lifestyle and health interventions that improve sleep quality—such as weight management and exercise that reduce sleep apnea—may help preserve healthier brain-wave patterns, although there is no single remedy to reverse brain aging.
Authors and Funding
The model was developed by Haoqi Sun, Robert J. Thomas, M. Brandon Westover, and colleagues, with senior authors including Yue Leng. Funding sources included multiple National Institutes of Health grants, the National Institute on Aging, the National Science Foundation, the National Health and Medical Research Council, and the American Academy of Sleep Medicine.
Key Questions Answered:
A: Not necessarily. An older EEG-derived brain age indicates accelerated electrical aging of the brain during sleep and suggests higher future risk, but it is not a clinical diagnosis of dementia.
A: Sleep facilitates critical restorative processes, including clearance of amyloid-beta and other metabolic waste. Disrupted or altered sleep EEG microstructure may signal failures in these processes that contribute to long-term cognitive decline.
A: Quantity alone is not enough; the quality and microstructure of sleep matter. Addressing health factors like body weight and exercise that reduce sleep-disordered breathing and improve sleep quality may help preserve healthier EEG patterns, but there is no simple, guaranteed intervention to reverse brain aging.
Editorial Notes:
- This article was edited by a Neuroscience News editor.
- The journal paper was reviewed in full.
- Additional context was added by staff.
About this sleep and dementia research news
Author: Suzanne Leigh
Source: UCSF
Contact: Suzanne Leigh – UCSF
Image: The image is credited to Neuroscience News
Original Research: Open access. “Machine Learning–Based Sleep Electroencephalographic Brain Age Index and Dementia Risk: An Individual Participant Data Meta-Analysis” by Haoqi Sun and colleagues. Published in JAMA Network Open. DOI: 10.1001/jamanetworkopen.2026.1521
Abstract
Machine Learning–Based Sleep Electroencephalographic Brain Age Index and Dementia Risk: An Individual Participant Data Meta-Analysis
Importance
Sleep EEG microstructures are closely tied to cognition and change with age. Their multidimensional nature makes conventional interpretation difficult. The EEG brain age index (BAI) uses machine learning to quantify the gap between sleep EEG–predicted brain age and chronological age.
Objective
To evaluate the association between sleep EEG–based BAI and incident dementia in community populations.
Data Sources and Study Selection
This IPD meta-analysis pooled overnight polysomnography data from five longitudinal cohorts (MESA, ARIC, FHS-OS, MrOS, SOF). Adults without dementia at baseline who underwent polysomnography were included.
Methods and Outcomes
Interpretable machine learning computed BAI from central-channel sleep EEG features. Fine-Gray competing-risk models assessed BAI’s association with incident dementia in each cohort; estimates were pooled by random-effects meta-analysis. The primary outcome was incident dementia or probable dementia, accounting for death as a competing risk.
Results
The analysis included 7,105 participants across cohorts with follow-up ranging from 3.6 to 16.9 years in different studies. Each 10-year increase in BAI was associated with a 39% higher risk of incident dementia after covariate adjustment (HR, 1.39; 95% CI, 1.21–1.59; P < .001). Associations remained significant after controlling for comorbidities, apnea-hypopnea index, and APOE ε4.
Conclusions and Relevance
A higher sleep EEG–based brain age index was consistently associated with greater risk of incident dementia. These findings support further evaluation of the BAI as a noninvasive digital biomarker for early detection of dementia in community settings.