Abstract
INTRODUCTION: Mind-wandering is a highly dynamic phenomenon involving frequent fluctuations in cognition. However, the dynamics of functional connectivity between brain regions during mind-wandering have not been extensively studied. METHODS: We employed an analytical approach aimed at extracting recurring network states of multilayer networks built using amplitude envelope correlation and imaginary phase-locking value of delta, theta, alpha, beta, or gamma frequency band. These networks were constructed based on electroencephalograph (EEG) data collected while participants engaged in a video-learning task with mind-wandering and focused learning conditions. Recurring multilayer network states were defined via clustering based on overlapping node closeness centrality. RESULTS: We observed similar multilayer network states across the five frequency bands. Furthermore, the transition patterns of network states were not entirely random. We also found significant differences in metrics that characterize the dynamics of multilayer network states between mind-wandering and focused learning. Finally, we designed a classification algorithm, based on a hidden Markov model using state sequences as input, that achieved a 0.888 mean area under the receiver operating characteristic curve for within-participant detection of mind-wandering. DISCUSSION: Our approach offers a novel perspective on analyzing the dynamics of EEG data and shows potential application to mind-wandering detection.