Nonlinear spatio-temporal analysis for neural activity characterization in MEG data

Nonlinear spatio-temporal analysis has been performed on neural activity recorded using 148-channel whole-head magnetoencephalography. The analysis consists of two phases: artifact removal and nonlinear feature evaluation. In the first phase, known artifacts, produced by cardiac and eye movement, and unknown artifacts have been isolated from intrinsic neural activity by using two adaptive filters in a cascade configuration. In the second one, phase space reconstruction of multivariate Magnetoencephalography measurements have been performed by using both temporal and spatial embedding. Indices of nonlinear dynamics have been defined and evaluated showing invariant features both in time and space.