Better Sleep Now, Better Cognition Later? Predicting Cognitive Function Using A Machine Learning-Based Sleep EEG...

Area of Science:

• Neuroscience
• Biomarkers
• Machine Learning

Background:

• Electrocortical activity measured by electroencephalograms (EEG) is linked to cognitive function and dementia risk.
• Machine learning can condense complex EEG data into integrated scores, such as the Brain Health Score (BHS).
• The predictive value of BHS for future neuropsychological (NP) performance is currently unknown.

Purpose of the Study:

• To investigate whether the Brain Health Score (BHS) predicts future neuropsychological (NP) performance.
• To assess the association between mid-to-late life BHS and cognitive function assessed over a decade later.

Main Methods:

• The study included 426 participants from the Framingham Heart Study (FHS) with BHS values from in-home polysomnography.
• Participants underwent digital clock drawing (dCDT) and NP testing an average of 12.6 years after BHS assessment.
• Linear regression models were used to estimate associations between BHS and cognitive scores, adjusting for covariates.

Main Results:

• Each 1-SD higher BHS was associated with significantly better performance in dCDT, memory, language, and executive function.
• Specifically, higher BHS correlated with higher scores in dCDT (β=0.16), memory (β=0.13), language (β=0.13), and executive function (β=0.10).
• These associations remained significant after adjusting for multiple demographic and health factors.

Conclusions:

• A higher Brain Health Score (BHS) in mid-to-late life is associated with better cognitive performance over a decade later.
• These findings suggest that EEG-derived, data-driven scores like BHS may serve as valuable biomarkers for future cognitive health.
• The study supports the utility of BHS in predicting long-term brain health outcomes.