Tracking Dynamic Functional Connectivity Using Time-Varying Multivariate Autoregressive Models: an EEG Study of
Abstract:
Brain functions are the results of complex neural activations and dynamic interactions across distinct cortical regions. Dynamic functional connectivity (dFC) represents a powerful tool for studying these transient interactions and permits to extract connectivity even at very high temporal resolutions. One example is given by time-varying multivariate autoregressive (tvMVAR) models, which track dFC instantaneously. These methods rely on different parameters that need to be carefully tuned to achieve an accurate estimation of connectivity. In this study we analyse dFC during a sensorimotor control task using electroencephalography (EEG) data. The aims of the present work are: i) to investigate the performance of tvMVAR and the effect of its parameters through a realistic simulation study, and ii) to explore the dFC patterns in real EEG data from healthy subjects executing a visually guided pointing task. TvMVAR performance is influenced by parameter selection, thus simulations are needed to better determine their values. On real data, dFC analysis showed task-related modulations of the network, specifically highlighting the dominant role of the contralateral motor cortex as source of information flow toward the ipsilateral motor cortex, premotor cortex, and posterior parietal cortex.Clinical Relevance- A better understanding of network modulation over time during cognitive and motor processes could bring benefits in neurorehabilitation developing strategies for recovering lost motor functions.
