Neural representations expand with aging but collapse in mild cognitive impairment and Alzheimer's disease.
Abstract
Objective
Neural activity continuously changes over time, yet how these changes relate to cognitive states and disease remains unclear.
Methods
We modeled EEG activity as distributions of short-term activity patterns and quantified distributional changes using Wasserstein distance. Intrinsic dimensionality was estimated to characterize representational complexity and its association with these changes.
Results
Across multi-task, lifespan, and clinical EEG datasets, higher intrinsic dimensionality was consistently associated with greater distributional changes. Healthy aging showed increased dimensionality and modestly increased displacement, whereas mild cognitive impairment and Alzheimer's disease showed reduced dimensionality and displacement, indicating distinct alterations in representational geometry.
Conclusions
Cognitive aging and impairment are associated with distinct alterations in the geometry of neural representations.
Significance
A geometric perspective on neural representations provides a complementary approach for characterizing cognitive states and brain dynamics across aging and disease.