Probabilistic initiation and termination for MEG multiple dipole localization using sequential Monte Carlo methods

The paper considers an electromagnetic inverse problem of localizing dipolar neural current sources on brain cortex using magnetoencephalography (MEG) or electroencephalography (EEG) data. We aim to localize the unknown and time-varying number of dipolar current sources using data from multiple MEG coil sensors. In this work, we model the problem in a Bayesian framework, we propose a linear prior detection method as well as a probabilistic approach for target number estimation, and target state initiation/termination. We then use a sequential Monte Carlo (SMC) algorithm to numerically estimate location and moment of the dipolar current sources. We apply the algorithm in both simulated and measured data. Results show that the proposed approach is able to estimate and localize the unknown and time-varying number of dipoles in simulated data with reasonable tracking accuracy and efficiency.