Estimation of time-varying causal connectivity on EEG signals with the use of adaptive autoregressive parameters

In this paper, we address the problem of time-varying causal connectivity estimators on Electro-encephalographic (EEG) signals by means of Directed Transfer Function (DTF). The DTF method reveals causal information flows between brain areas, while direct DTF (dDTF) is able to distinguish and estimate only direct flows. Since neuro-physiological signals such as EEG and event related potentials (ERP) can be nonstationary, their temporal dynamics cannot be satisfactorily represented. Time-varying dDTF can be estimated using Kalman Filter for adaptive calculation of multivariate autoregressive coefficients. This approach can reveal transient causal relations and model time-dependent flow patterns. This approach was applied to simulated signals and the results indicated that time-varying dDTF can provide efficient estimates of connectivity patterns.