Estimation of dynamic neural activity including informative priors into a Kalman filter based approach

The EEG recordings contain dynamic information inherent to its nature, therefore, the accurate estimation of neural activity is highly dependent on the inclusion of such information in the inverse problem solution. The present study proposes the inclusion of informative priors into a Kalman filter based solution, aimed to include the different dynamics present on the data. This is achieved by decomposing a space-time-frequency, here after s-f-t, representation of the data to extract different dynamics contained in the EEG signals. Attained results using physiological-based simulations, show that including more informative s-f-t priors along with a temporal-based solution, the reconstruction of neural activity can be improved, in the present study, we achieved an average localization error of 4 mm, compared to 47 mm using the baseline approach.