Abstracts

RATIONALE: Nonlinear analysis of electroencephalographic (EEG) signals has been shown to detect dynamical changes before seizures onset. The similarity method has been used to measure preictal changes of EEG signals from selected patients suffering from well-localized intractable focal epilepsies, originating from the mesial temporal lobe (Le Van Quyen et al., The Lancet 2001, 357 : 183-88) or from the neocortex (Navarro et al., Brain 2002, 125 : 640-655).
The efficiency of this method on unselected patients undergoing intracranial investigations needs to be evaluated. The nature of the dynamical changes detected prior to seizures needs to be defined.
METHODS: All patients (n=7) undergoing an intracranial EEG investigation for medically intractable partial epilepsy at the Montreal Neurological Institute between December 2001 and April 2002 were included.
A total of 51 seizures were studied, according to the following criteria : electro-clinical or electrographic seizures, at least 1 hour of EEG recording before each seizure and between two successive seizures.
All EEG channels were analysed by the similarity nonlinear method, that compares the signal dynamics of successive windows with that of a 5 min long [ssquote]reference[ssquote] period, taken at the beginning of the recording.
The presence of a preictal change was defined by a deviation of the similarity index of more than 5 standard deviations above that of the [soquote]reference[scquote] period. These changes should persist until the seizure, in at least 3 channels and for more than 1 min.
In order to explain the nature of these dynamical changes, EEG recordings were then visually inspected and video recordings were reviewed.
RESULTS: (i) Preictal changes were detected in a total of 31 / 51 seizures.
Among these 31 seizures, visually detectable changes of the EEG may account for the preictal changes in 24 seizures, corresponding to moderate modifications of the background activity before 2 seizures, changes of sleep stages before 7 seizures, changes of the epileptic interictal activity before 13 seizures, and concomitant artifacts before 2 seizures. No visually detectable change of the EEG could be related to the preictal changes before 7 seizures.
(ii) In 14 seizures, changes of the similarity index were found, but did not meet our criterion. Among these 14 seizures, changes were found before 9 seizures, but they included fewer than 3 channels. In 5 other seizures, the preictal changes could not be distinguished from post-ictal changes due to a previous seizure.
(iii) No preictal change was detected in 6 seizures.
(iiii) Preictal changes were found in channels adjacent to the epileptogenic focus, as well as at a distance.
CONCLUSIONS: The similarity nonlinear method was able to detect preictal changes in 60 % of the analyzed seizures, in an unselected population of patients with various localizations and extensions of the epileptogenic area.
Those changes could be related to visually undetectable dynamical modifications or to varied obvious modifications of the EEG. In the latter, preictal changes were sometimes difficult to distinguish from physiological changes.
[Supported by: Canadian Institutes of Health Research grant FRN 10189
Fondation pour la Recherche Médicale]