| The structure of human electroencephalographic (EEG) signals is investigated using computational methods with the objective of developing automatic algorithms for clinical applications. The EEG is examined globally as a stochastic process with alpha-stable increments, and a novel parameter estimation method is used to investigate its properties. The signal is also examined locally as a smooth, deterministic dynamical system, and the performance and applicability of methods for nonlinear feature estimation are evaluated. The results are compared to standard linear and spectral methods. A set of inter-cranial recordings from epileptic patients is used to study the local behavior of the EEG, and algorithms for detection and forecasting of seizures are considered. In addition, a set of sleep-EEG recordings from a study on neonatal development is examined. An algorithm for sleep-state identification in neonates is presented, and the assessment of brain maturation in neonates using the EEG is explored. |
