Predicting Alzheimer progression using EEG-based digital twins

Early and accurate prognosis of Alzheimer’s disease (AD) remains limited by current biomarkers, which are often invasive, require expensive equipment, or fail to predict individual trajectories of cognitive decline. In this study, we evaluated the Digital Alzheimer’s Disease Diagnosis (DADD) model, a digital twin framework that extracts neurodegeneration-related digital biomarkers from non-invasive recordings such as electroencephalography (EEG), mapping early functional neural alterations to latent parameters linked to structural AD mechanisms. DADD was applied to a cohort of 459 participants with Subjective Cognitive Decline (SCD) or Mild Cognitive Impairment (MCI), evaluating its performance against established AD biomarkers like APOE genotype, magnetic resonance imaging (MRI) and cerebrospinal fluid (CSF) markers in predicting prognostic outcomes over a 10-year follow-up. Digital biomarkers demonstrated strong prognostic value, predicting participants who converted at follow-up, significantly outperforming standard EEG metrics. They also robustly forecast both risk and timing of clinical conversion, matching the performance of established AD biomarkers and even outperforming them in SCD participants. Moreover, combining established AD biomarkers with digital biomarkers led to a significant increase in prognostic accuracy. DADD represents a robust clinical tool, enabling scalable, affordable, non-invasive AD detection and prognostic stratification across diverse clinical settings.