Analysis of task-related MEG functional brain networks using dynamic mode decomposition.

Objective. Functional connectivity networks explain the different brain states during the diverse motor, cognitive, and sensory functions. Extracting connectivity network configurations and their temporal evolution is crucial for understanding brain function during diverse behavioral tasks. Approach. In this study, we introduce the use of dynamic mode decomposition (DMD) to extract the dynamics of brain networks. We compared DMD with principal component analysis (PCA) using real magnetoencephalography data during motor and memory tasks. Main results. The framework generates dominant connectivity brain networks and their time dynamics during simple tasks, such as button press and left-hand movement, as well as more complex tasks, such as picture naming and memory tasks. Our findings show that the proposed methodology with both the PCA-based and DMD-based approaches extracts similar dominant connectivity networks and their corresponding temporal dynamics. Significance. We believe that the proposed methodology with both the PCA and the DMD approaches has a very high potential for deciphering the spatiotemporal dynamics of electrophysiological brain network states during tasks.