Reconstruction of Epileptic Brain Dynamics Using Data Mining Techniques

Abstract

The existence of complex chaotic, unstable, noisy and nonlinear dynamics in the brain electrical and magnetic activities requires new approaches to the study of brain dynamics. One approach is the combination of certain multichannel global reconstruction concept and data mining techniques. This approach assumes that information about the physiological state comes in the form of nonlinear time series with noise. It also involves a geometric description of the brain dynamics for the purpose of understanding massive amount of experimental data. The novelty in this chapter is in the representation of the brain dynamics by hierarchical and geometrical models. Our approach plays an important role in analyzing and integrating electromagnetic data sets, as well as in discovering properties of the Lyapunov exponents. Further, we discuss the possibility of using our approach to control the Lyapunov exponents, predict the brain characteristics, and “correct” brain dynamics. We represent the Lyapunov exponents by fiber bundle and its functional space. We compare the reconstructed dynamical system with the geometrical model. We discuss an application of this approach to the development novel algorithms for prediction and seizure control through electromagnetic feed-back.