LATENT-TRAJECTORY MODELING OF LONGITUDINAL EEG ALTERATIONS IN PRODROMAL Α-SYNUCLEINOPATHIES

Idiopathic REM sleep behavior disorder (iRBD) represents a prodromal stage of alpha-synucleinopathies, with a high risk of progression to Parkinson’s disease or Dementia with Lewy Bodies over a prolonged preclinical phase. Early neurodegeneration is associated with subtle, distributed alterations in cortical network dynamics, which can be non-invasively assessed using high-density EEG. However, EEG-derived biomarkers are often high-dimensional, heterogeneous across subjects, and non-monotonic over time, limiting their usefulness for longitudinal disease modeling.
Here, we propose a latent-trajectory framework that integrates multivariate EEG criticality metrics into low-dimensional representations designed to capture longitudinal structure. Resting-state high-density EEG was collected from multicentric cohorts in Italy and South Korea, including healthy controls (N=65) and iRBD patients (N=95) with heterogeneous follow-up and conversion trajectories. EEG metrics included detrended fluctuation analysis, bistability index, weighted phase lag index, and orthogonalized amplitude correlations, capturing complementary aspects of cortical network dynamics. These measures were harmonized across recording sites and frequency bands to enable consistent comparisons across cohorts. Across the cohort, EEG-derived metrics showed associations with striatal dopamine uptake, while correlations with clinical scores were weaker, indicating that EEG may capture early neurodegenerative changes that are not yet evident clinically. Latent components captured global variability at baseline, with some showing weak correlations with striatal dopamine and subtle group-related differences, highlighting early, distributed EEG alterations. Overall, these findings demonstrate that early EEG alterations in iRBD are complex and heterogeneous, emphasizing the value of trajectory-based, multivariate approaches for capturing prodromal neurodegenerative changes that may be missed by single-feature biomarkers.