O20 Detection of rapid changes in epileptic EEG

Neurophysiol Clin (1990) 20, S, 7s 9s 7S

© Elsevier , P a r i s

Communications orales Oral communications

SESSION 3

ModUles exp6rimentaux et informatisds des signaux dectrophysiologiques spontan6s et 6voqu6s

Experimental and computerized models of spontaneous and evoked electrophysiological signals

019

CORTICAL TOPOGRAPHICAL ANALYSIS AND SOURCE LOCALIZATION OF EPILEPTIFORH DISCHARGES

R. D. Sidman, D.B.Smith, L.Lee, D.J.Major,

The University of Southwestern Louisiana, Lafayette (USA) and the Oregon Comprehensive Epilepsy Program, Portland (USA)

We present a method fo r s imu la t ing p o t e n t i a l f i e l d s on the cortical surface. This cortical ~maging ~echnique (CIT) is validated by applying it Eo the potentials evoked by right median nerve stimulation where actual cortical maps are available for comparison.

CIT is then applied to several cases of epileptiform discharges where the sites of the discharges have been con- firmed. This technique appears to distinguish features of the EEG that are not apparent in standard scalp topograph- ical maps and may make it easier to localize the sources of discharges without recourse to invssive recording tech- niques.

020

D E T E C T I O N O F RAPID C H A N G E S I N E P I L E P T I C E E G

Gath, I.*, Feuerstein, C.**, Pham, D.T.*, Bonnet, C.** Vergnes, M.* * *. • Laboratoire LMC/IMAG, UJF, Grenoble. ** Laboratoire de Physiology section Neurophysiology, INSERM U318 CHU, Grenoble * * * Cent ~p ~

Neurochimie, CNRS, Strasbourg.

Power spect rum estimation is a useful tool in the quant i ta t ive analysis of EEG recorded during epileptic seizures. However, the "epileptic" EEG sig- nal recorded either in experimental animals or in humans might show marked nonstat ionari ty, and hence, est imation of the power spectrum using conven- tional techniques such as the F F T results in bad estimates, or in "smearing" of the changes due to the averaging process inherent in the FFT technique. A different approach to the est imation of the power spectrum and phase and coherence spectra is based on parametr ic modelling of the multichannel EEG. The AR model est imator is required to be efficient in case of short record lengths (2 sec. or less). Continuous recordings of EEG were carried out in a genetically epilepsy-prone rat , and in a normal rat. Recording was made from left and right frontal and occipital leads by implanted electrodes. Da ta acquisition was carried out on IBM P C / A T using a sampling rate of 128 Hz. In addition, several depth recordings (right and left amygda la and hippocam- pus) from patients having generalized seizures with focal onset were digitized. Signal segments including both normal EEG and ictal activity were chosen for analysis. Power spectrum estimation and coherence and phase spectra were calculated using multichannel (vectorial) autoregressive modeling, through a process of adaptive segmentation. Thus, estimates of the spectra of quasi- s ta t ionary segments of the signal were obtained. The genetically-prone epilep- tic ra t had abundance of short epochs of multi-spike and spike and wave ac- tivity, a l ternat ing with normal background activity. 3-D plots of the power spectra and contour maps demonst ra ted tha t the pathological activity could be quanti tat ively distinguished from periods of normal background activity by spectral peak at around 9-10 Hz, with increased coherence between the different leads. The instantaneous spectral peaks corresponding to the human seizure were found to decrease gradually in frequency by 2-3 Hz , and together with the coherence spectra gave quanti tat ive information related to the dy- namic changes which occur prior and during the seizure. It is argued tha t power spectrum estimation together with estimation of coherence and phase spectra, using parametr ic multichanne] modeling of the EEG, could be an aid to the quanti tat ive analysis of EEG during epileptic seizures.