6 - Détection temps-échelle d'évènements paroxystiques intercritiques en électroencéphalogramme

The electroencephalographic signals recorded from surface and depth electrodes are, in epilepsy, the basic elements for the spatiotemporal analysis ofparoxystic processes . These signals can be characterized in the interictal periods by fast discharges that must be detected. These non predictable transients are superposed to a background activity which is itself a complex mixture of signals . This paper describes a detector based on Wavelet Transform whose aim is to detect theses events. A two stage structure is proposed : the first level carries out the detection of all the transients (including the artifacts) ; the second level achieves the discrimination between the events of interest and the artifactual waveforms . The decision criteria associated to the stages of the detector are reported and their adaptive fitting is emphasized . The resulting performances are evaluated on simulated and real data and they are compared to another detection scheme which has been recently proposed .