Simultaneous Localization of Multiple Oscillations Using Dynamic Solutions to the Inverse Problem in E/MEG Source Analysis

One critical challenge encountered in non-invasive electrophysiological studies is inferring the spatiotemporal dynamics of cortical sources from measured electro/magneto-encephalography (E/MEG) recordings. In this paper, we describe a dynamic inverse solution utilizing state-space oscillator models to enable simultaneous source localization of multiple oscillations under distributed source modeling. Our solution framework incorporates symmetric spatial dependencies with sparse priors and performs efficient state inference and parameter learning via the expectation-maximization algorithm. We show promising simulation results of successful source localization of oscillatory activity that improve over past non-oscillating solutions. We also demonstrate the added ability to simultaneously localize both alpha and slow rhythms from the same recording in both simulated and real resting-state EEG data. This method adds to a growing body of time-domain modeling methods to characterize neural oscillations from electrophysiological data and paves the way for analyzing cortical neural activity with improved precision beyond traditional static inverse solutions.