Abstract—A graphical and analytical description of epileptic seizures based on amplitude modulation and frequency modulation components of intracranial EEG (iEEG) is proposed. This representation allows the characterization of seizures and their different stages from the iEEG by means of triangles whose vertexes are the centroids (c m ) of the signal during preictal, ictal and postictal periods. The centroid is the point defined by the average values of instantaneous amplitude and frequency, a i and f i respectively. Data were obtained from 8 patients with recurrent epilepsy, 170 records were processed, 62 of which were seizures and 108 interictal signals. Results show that the centroids of the ictal periods are located in a region of the space a i - f i distant from the centroids corresponding to interictal and postictal periods. This original representation of epileptic seizures can facilitate the visualization of stage- transitions and discrimination between the different stages of the iEEG signal. An additional advantage of the method is that the information contained in the signal is synthesized significantly. I. INTRODUCTION ANY signal processing techniques have been used in epilepsy research. The main goals of these studies have included: understanding the origin of seizures, their temporal evolution, searching new therapies to treat the disease and also defining surgical strategies to follow in patients with intractable epilepsy. The amplitude of EEG signals has been considered in the extraction of signal characteristics in time domain. Gotman used the average wave amplitude in one epoch, in a system for seizure detection based on the analysis of multiple characteristics of the EEG [1]. The detection of signal peaks for diagnostic has been treated, among others by Pang [2]. The spectral characteristics of EEG signals also have been widely explored, examples of some characteristics in the frequency domain are the dominant frequency [1], the power in frequency bands studied in the detection of seizure onset [3], and the coherence used to study the synchronization between diverse areas of the brain before seizures [3], [4], [5]. The amplitude and frequency or phase of the EEG, have been used together to study the spatio-temporal dynamics before seizures, with emphasis in synchronization. The Manuscript received April 2, 2007. M. A. Díaz is with the Department of Electronic and Circuits, Simón Bolívar University, Caracas, Venezuela (phone: +58-212-9064014; fax: +58-212-9063631; e-mail: mdiaz@ usb.ve). J. C. Viola, is with the Department of Electronic and Circuits, Simón Bolívar University, Caracas, Venezuela (e-mail: jcviola@ usb.ve). R. Esteller is with Neuropace, Mt View, California, USA (e-mail: resteller@neuropace.com). results obtained contribute to understanding epileptogenic networks and the mechanisms of seizures generation [6]. In the work of De Clercq [7], the Hilbert transform was used to consider the a i and f i in the EEG from patients with mesial temporal lobe epilepsy. The Hilbert transform was also used to measure phase synchronization and signal rhythmicity. Another method that considers the a i and f i of a signal is the one proposed by Maragos [8], [9] based on Teager energy operator (TEO) also known as nonlinear energy operator [10]. Maragos showed the utility of the method analyzing the resonances of modeled speech using an AM- FM signal [9]. The extraction of frequency and amplitude modulation components formed the basis of a hypernasal speech detection system [11], [12]. TEO potentiality as a useful feature in the detection of epileptic seizures was determined by Esteller [3]. Investigation in [13], [14] demonstrate that is possible to decompose the iEEG signal in an amplitude modulation component (a i ) and a frequency modulation component (f i ). It has been reported that when Hilbert Transform is applied to real signals, negative values of f i can be obtained [14], [15]. This can be avoided using TEO to estimate the f i . Additionally the approaches based on TEO have a smaller computational cost and faster adaptation [16]. An exhaustive comparison between Hilbert Transform and TEO methods, to estimate a i and f i features is presented in [14], [16]. Typically, it is observed that the EEG signal has a tendency to an inverse relation between the amplitude and the frequency, namely when the frequency of the signal increases its amplitude decreases and vice versa. Particularly, during an epileptic seizure the signal’s amplitude increases, so it is likely that the f i of the signal will be lower in this period than during the preictal, postictal, or interictal stages. In this work the characteristics of a i and f i in IEEG signals from patients with recurrent epilepsy are compared as an attempt to characterize the signal according to the stage of the epileptic episode. This analysis of the signal could be useful to improve the guidelines for interpretation of the iEEG signal. The understanding of the meaning of instantaneous frequency is controversial, because it depends among other factors on the nature of signals. The methods most used to determine the f i are defined for monocomponent signals. For a multicomponent signal such as the iEEG, the estimated a i Analysis of Instantaneous Amplitude and Frequency of Intracranial EEG signal to Characterize Epileptic Seizure Stages M. A. Díaz, Student Member, IEEE, J. C. Viola, and R. Esteller, Member, IEEE M Proceedings of the 29th Annual International Conference of the IEEE EMBS Cité Internationale, Lyon, France August 23-26, 2007. ThD02.3 1-4244-0788-5/07/$20.00 ©2007 IEEE 1290