Detecting nonlinearity in multichannel epileptic EEG

Distinguishing between determinism and noise has attracted many interests in nonlinear dynamical analysis of electroencephalograph (EEG). We apply the phase-randomized Fourier transform algorithm to generate surrogate data of multichannel EEG time series to detect nonlinearity in the epileptic EEG. For this purpose, fifty EEG segments from ten patients were analyzed, and the modified Grassberger and Procaccia algorithm (GPA) with multivariate embedding was used to calculate the discriminating statistic. The results indicate that epileptic EEGs exhibit nonlinearity with high confidence level, and the application of measures from nonlinear dynamics to epileptic EEG analysis appears reasonable.