HOW OLD IS YOUR BRAIN? A FOUNDATIONAL MODEL FOR BRAIN AGING AND COGNITION DECLINE

Brain aging is a complex process in which cognitive functions and homeostatic activity decline unevenly across domains, making it difficult to establish a comprehensive view (Campbell et al., 2024; James & Burgess, 2025). Disentangling the relationships between neural features, chronological age, and cognitive function remains a fundamental challenge, as cognitive reserve mechanisms can mask age-related decline (James & Burgess, 2025). There is a pressing need for non-invasive, functional biomarkers for early detection of neurodegenerative conditions, where early intervention is effective (Simmatis et al., 2025). We propose a foundational model trained on EEG signals across diverse ages and cognitive tasks. Given resting-state or task-based EEG, such a model predicts brain age and cognitive scores simultaneously, identifying which cognitive domains have declined, the trajectory of aging, and which brain regions drive these changes. To this end, we employed statistical feature extraction, individual supervised models, and ensemble methods. We examined correlations with and without controlling for age. Training utilised 126 subjects from open-source datasets with validation done using an external hospital dataset. Preliminary findings show Peak Alpha Frequency (C-PAF, N-PAF) maintains significant age-controlled correlations with cognitive performance (NART: ρ = −0.420, p = 0.001; QMCI: r = 0.308, p < 0.05). ElasticNet achieved MAE = 5.78 years, while SVR yielded MAE = 4.33 years on external hospital data. Feature and channel importance analyses reveal which brain regions and spectral signatures best predict aging, offering mechanistic insights into cognitive decline and establishing a foundation for accessible cognitive health monitoring across diverse populations.