EEG Analysis with Neuronal Cell Models: From Detection towards Prediction of Epileptic Seizures.
Abstract number :
1.124
Submission category :
Year :
2001
Submission ID :
124
Source :
www.aesnet.org
Presentation date :
12/1/2001 12:00:00 AM
Published date :
Dec 1, 2001, 06:00 AM
Authors :
K. Schindler, MD, PhD, Neurology, Inselspital, Bern, Switzerland; R. Wiest, MD, Neurology, Inselspital, Bern, Switzerland; M. Kollar, REEGT, Neurology, Inselspital, Bern, Switzerland; F. Donati, MD, Neurology, Inselspital, Bern, Switzerland
RATIONALE: We have recenctly developed a method for real-time and automatic detection of epileptic seizures based on simulated neuronal cell models, which are used for EEG analysis (1). The objective of the study presented here is to test if our method can be modified to detect slow changes of excitability in the preictal state and thereby provide a new measure to anticipate epileptic seizures.
METHODS: The core of our seizure detection method are 2 simulated leaky integrate and fire units (LIFU), which are classical simple neuronal cell models. The LIFUs are connected to a signal preprocessing stage that detects those parts of 8 foramen ovale EEG signals (4 EEG signals on each side) with slopes larger than a preset threshold [italic]H[sub]th[/sub][/italic] and marks them with unit pulses. The LIFUs increase their spiking rates the higher the frequency and the synchrony of the impinging pulse trains. To detect EEG signals with large slopes as occur at the beginning of epileptic seizures, [italic]H[sub]th[/sub][/italic] has typically to be set in a high range of 2500-6000[mu]V/s. Here we set [italic]H[sub]th[/sub][/italic] to low values of 400-500[mu]V/s and test if our method can detect slow changes of excitability during the preictal state.
RESULTS: We analysed 3 long-term foramen ovale EEGs (24[plusminus]1h recording time) of patients with uni- or bilateral drug resistant temporal lobe epilepsy. 14 seizures were recorded. All the seizures were preceded by a non-monotonic increase of the spiking rates [italic]SR[/italic] of the 2 LIFUs. Before 12 seizures the increase of [italic]SR[/italic] could be monitored by computing the moving average [italic]SR[sub]av[/sub][/italic] for a sliding time window of 30 minutes duration. The minimum of [italic]SR[sub]av[/sub][/italic] in each inter-/preictal interval occurred 166[plusminus]99min before seizure onset. For the remaining 2 seizures the minimum of [italic]SR[sub]av[/sub][/italic] coincided with seizure onset. However, a local minimum of [italic]SR[sub]av[/sub][/italic] could be demonstrated 8 and 16min before seizure onset when computing the moving average [italic]SR[sub]av[/sub][/italic] with a time window of 10min duration.
CONCLUSIONS: We conclude that simulated neuronal cell models can be used to detect slow changes of excitability during the preictal state by monitoring the moving average of their spiking rates [italic]SR[/italic]. We suggest that the gradual increase of [italic]SR[/italic] that occurs before seizure onset could serve as a new measure to anticipate epileptic seizures. This opens the interesting possibility of using the same small network of simulated neuronal cell models operated in two different modes (with [italic]H[sub]th[/sub][/italic] set to low and high values respectively) for seizure prediction and seizure detection.
[italic]References[/italic]
(1) Schindler K, Wiest R, Kollar M and Donati F. Using simulated neuronal cell models for detection of epileptic seizures in foramen ovale and scalp EEG. [italic]Clin Neurophysiol[/italic] 2001, in press
Support: This work was partially supported by the Swiss National Science Foundation, SNF No. 31-55298.98