Classification of Epileptic EEG Data Using Multidimensional Scaling

A methodology is proposed and developed for epileptic seizures prediction through multifeatures extracted from EEG, and submitted to space reduction. Concepts from energy, frequency-time, and nonlinear dynamics are used to obtain the set of 14 features. Multidimensional scaling is used for space reduction from high dimensional to three dimensional space, in the VISRED platform. Results show the potentiality of this methodology. A computational system with an algorithms base and a data base, under development, is briefly sketched, in order to allow to face the high variability of biological systems.