Chapter 4 - Pattern adapted wavelet-based seizure prediction

Abstract

Electroencephalographic (EEG) signals are analyzed based on event-related potentials (ERPs) and event-related oscillations (EROs). The wavelet transform is the best suitable method to analyze the temporal and spatial characteristics of EROs at the same instant. Fundamental analysis of EEG is followed by artifact removal, feature selection, feature extraction, and classification. EEG signals from an epileptic patient are divided into five periods or stages: (i) the nonseizure period, when no epileptic syndrome is visible; (ii) the ictal period, which is the actual seizure period, with a typical duration of 1 to 3 minutes; (iii) the preictal period, with a duration of 30 to 60 minutes before the ictal period; (iv) the postictal period, with a duration of 30–60 minutes after the ictal period; and (v) the interictal period, which is the period between the postictal period and the preictal period of the immediate next ictal. Prediction and detection of seizures by analyzing ictal, preictal, and interictal periods could help to identify subsequent seizures, provide better treatments, and improve safety. The objective of this chapter to analyze EEG signals for identification of subsequent changes during the preseizure, seizure, and normal states. This identification of the preseizure state could lead to early detection of the occurrence of seizures to avoid accidents that are observed with epileptic patients.