Characterizing correlation changes of complex pattern transitions: The case of epileptic activity

We analyze the time-dependent spectrum of eigenvalues of the correlation matrix for multivariate EEG data at the transition to epileptic seizures. By a mechanism of level repulsion between states at both edges of the spectrum of the correlation matrix, relevant information about quantitative correlation changes is reflected in the largest and smallest eigenvalues and corresponding eigenvectors. By the application of measures from random matrix theory we provide evidence that statistically relevant information can be obtained both at the upper and the lower end of the spectrum. In addition, information about spatial characteristics of correlation changes can be extracted.