Robustness of individualized inferences from longitudinal resting state dynamics

Abstract

Tracking how individual human brains change over extended timescales is crucial in scenarios ranging from healthy aging to stroke recovery. Tracking these neuroplastic changes with resting state (RS) activity is a promising but poorly understood possibility. It remains unresolved whether a persons RS activity over time can be reliably decoded to distinguish neurophysiological changes from confounding differences in cognitive state during rest. Here, we assessed whether this confounding can be minimized by tracking the configuration of an individuals RS activity that is shaped by their distinctive neurophysiology rather than cognitive state. Using EEG, individual RS activity was acquired over five consecutive days along with activity in tasks that were devised to simulate the confounding effects of inter-day cognitive variation. As inter-individual differences are shaped by neurophysiological differences, the inter-individual differences in RS activity on one day were analyzed (using machine learning) to identify a distinctive configuration in each individuals RS activity. Using this configuration as a classifier-rule, an individual could be re-identified with high accuracy from 2-second samples of the instantaneous oscillatory power acquired on a different day both from RS and confounded-RS. Importantly, the high accuracy of cross-day classification was achieved only with classifiers that combined information from multiple frequency bands at channels across the scalp (with a concentration at characteristic fronto-central and occipital zones). These findings support the suitability of longitudinal RS to support robust individualized inferences about neurophysiological change in health and disease.