Yu et al.'s EEG framework tracks neural aging and Alzheimer's collapse

A new study from researchers including Junjie Yu and Quanying Liu introduces a mathematical framework to quantify how neural activity changes over time and with disease. The team modeled EEG signals as distributions of windowed activity patterns, rather than single time points. They used the Wasserstein distance to measure temporal stability—how similar these distributions remain across time—and intrinsic dimensionality to estimate representational complexity. The approach was applied to multi-task, lifespan, and clinical EEG datasets, revealing consistent patterns across different cognitive conditions and age groups.

The results challenge the idea that neural representations drift freely. Instead, they show constrained, condition-specific stability. A key finding is the inverse relationship: brain regions with higher intrinsic dimensionality (like posterior areas) tend to be less stable over time than frontal regions. Healthy aging is marked by increased dimensionality and decreased stability, suggesting a richer but less reproducible neural code. However, mild cognitive impairment and Alzheimer's disease show a joint collapse of both measures, pointing to a breakdown in representational structure. The authors argue this distribution-level framework could serve as a sensitive biomarker for aging-related cognitive decline, potentially making EEG a cheaper, more accessible tool for early neurodegeneration screening.