Adaptive tracking of human ECoG network dynamics

Extracting and modeling the low-dimensional dynamics of multi-site electrocorticogram (ECoG) network activity is important in studying brain functions and dysfunctions and for developing translational neurotechnologies. Dynamic latent state models can be used to describe the ECoG network dynamics with low-dimensional latent states. But so far, non-stationarity of ECoG network dynamics has largely not been addressed in these latent state models. Such non-stationarity can happen due to a change in brain state or recording instability over time. A critical question is whether adaptive tracking of ECoG network dynamics can lead to further dimensionality reduction and more parsimonious and precise modeling. This question is largely unaddressed.

We investigate this question by employing an adaptive linear state-space model for ECoG network activity constructed from ECoG power feature time-series over tens of hours from 10 human subjects with epilepsy. We study how adaptive modeling affects the prediction and dimensionality reduction for ECoG network dynamics compared with prior non-adaptive models, which do not track non-stationarity.

Across the 10 subjects, adaptive modeling significantly improved the prediction of ECoG network dynamics compared with non-adaptive modeling, especially for lower latent state dimensions. Also, compared with non-adaptive modeling, adaptive modeling allowed for additional dimensionality reduction without degrading prediction performance. Finally, these results suggested that ECoG network dynamics over our recording periods exhibit non-stationarity, which can be tracked with adaptive modeling.

These results have important implications for studying low-dimensional neural representations using ECoG, and for developing future adaptive neurotechnologies for more precise decoding and modulation of brain states in neurological and neuropsychiatric disorders.