EEG functional connectivity metrics wPLI and wSMI account for d distinct types of brain functional interactions

Abstract

AbstractFunctional connectivity (FC) estimation methods are extensively used in neuroimaging to measure brain inter-regional interactions. The weighted Phase Lag Index (wPLI) and the weighted Symbolic Mutual Information (wSMI) represent relatively robust exemplars of spectral (wPLI) and information-theoretic (wSMI) connectivity measures that recently gained increased popularity due to their relative immunity to volume conduction. wPLI and wSMI are posited to have different sensitivity to linear and nonlinear relationships between neural sources, but their performance has never been directly compared. Here, using simulated high-density (hd-)EEG data, we evaluated the accuracy of these two metrics for detecting distinct types of regional interdependencies characterised by different combinations of linear and nonlinear components. Our results demonstrate that while wPLI performs generally better at detecting functional couplings presenting a mixture of linear and nonlinear interdependencies, only wSMI is able to detect exclusively nonlinear interaction dynamics. To evaluate the potential impact of these differences on real experimental data, we computed wPLI and wSMI connectivity in hd-EEG recordings of 12 healthy adults obtained in wakefulness and deep (N3-)sleep. While both wPLI and wSMI revealed a relative decrease in alpha-connectivity during sleep relative to wakefulness, only wSMI identified a relative increase in theta-connectivity, while wPLI detected an increase in delta-connectivity, likely reflecting the occurrence of traveling slow waves. Overall, our findings indicate that wPLI and wSMI provide distinct but complementary information about functional brain connectivity, and that their combined use could advance our knowledge of neural interactions underlying different behavioural states.