Signal Processing 83 (2003) 1275–1289 www.elsevier.com/locate/sigpro Wavelet analysis of generalized tonic-clonic epileptic seizures Osvaldo A. Rosso a ; * , Susana Blanco a , Adrian Rabinowicz b a Instituto de C alculo, Facultad de Ciencias Exactas y Naturales, Universidad de Buenos Aires, Pabell on II, Ciudad Universitaria, 1428 Ciudad de Buenos Aires, Argentina b Departament of Neurology and Epilepsy Program, Instituto de Investigaciones Neurol ogicas Ra ul Carrea (FLENI), Monta˜ neses 2325, 1429 Ciudad de Buenos Aires, Argentina Received 10 July 2000; received in revised form 13 September 2002 Abstract The analysis of generalized tonic-clonic seizures is usually dicult with quantitative EEG techniques due to muscle artifact. We applied two quantiers based on the Wavelet Transform to evaluate 20 seizures from eight consecutive patients admitted for video-EEG monitoring. We studied the relative wavelet energy and the wavelet entropy over time. In 16 20 events we found signicant decremental activity in the relative wavelet energy associated with frequency band 0.8–3:2 Hz (delta activity) at the seizure onset, indicating that the seizure is dominated by medium frequency bands 3.2–12:8 Hz (theta and alpha bands). In 19 20 events the mean wavelet entropy presents lower values during the ictal period compared to the pre-ictal period, indicating that the associated dynamic is more ordered and simple. Thus the employed tools show good accuracy for detecting changes in the system dynamic. We conclude that this behavior could be induced by frequency tuning in the neuronal activity triggered by an hypothetic epileptic focus. ? 2003 Elsevier Science B.V. All rights reserved. Keywords: Epileptic seizures; EEG; Wavelet; Signal entropy 1. Introduction Synchronous neuronal discharges create rhythmic potential uctuations, which can be recorded from the scalp through electroencephalography. The electroen- cephalogram (EEG) can be roughly dened as the mean brain electrical activity measured at dierent sites of the head. EEG patterns correlated with normal functions and diseases of the central nervous system * Corresponding author. Tel.: +54-11-4576-3375; fax: +54-11-4786-8114. E-mail addresses: rosso@ulises.ic.fcen.uba.ar, rosso@ba.net (O.A. Rosso), blanco@ulises.ic.fcen.uba.ar (S. Blanco), alr@eni. org.ar (A. Rabinowicz). are dened on an empirical basis. The clinical inter- pretation of EEG attempts to link pathological features (clinical symptomatology) with visual inspection and pattern recognition of EEG. Although this traditional analysis is quite useful, visual inspection of the EEG is subjective and hardly allows any systematization [29]. To overcome this, quantitative EEG analysis (qEEG) introduces objective measures reecting the charac- teristics of the brain activity as well as the associated dynamics. We must remark, however, that these meth- ods have not been developed to substitute traditional EEG visual analysis, but rather to complement it. The concept of ergodicity of the time series, which requires stationarity of the signal, is the common assumption in most traditional methods of signal 0165-1684/03/$ - see front matter ? 2003 Elsevier Science B.V. All rights reserved. doi:10.1016/S0165-1684(03)00054-9