Abstracts

Seizures occur more often in the newborn period than at any other time during childhood. The EEG can play a vital role in providing a sensitive indication of cerebral function during periods when the brain is at risk. The general hypothesis of this proposal is that there are measurable changes in brain electrical activity that precede and accompany seizures. Further, these changes can be quantitatively measured to characterize and differentiate normal from epileptic conditions, and detect seizures in newborns.
Linear and nonlinear univariate and bivariate measures were studied in: (1) full-term newborns with seizures (n=3), and (2) newborns without seizures (n=3). Newborns were matched for conceptional age, 5-minute Apgar score, weight, neurological condition, AED dose/ serum level. EEG from an 11-electrode montage based on the 10-20 international system was recorded (12-bit A/D conversion, 256 Hz sampling, 0.1-70 Hz signal bandwidth). EEGs were analyzed with the following meaures: signal energy, approximate entropy (ApEn), cross-ApEn, coherence, mutual information, and short-term maximum Lyapunov exponent (STLmax). Measures were applied to 0.5-1 hour, 11-channel bipolar EEG recordings iteratively at 4-second epochs (sampling frequency 256 Hz). EEGs of physiologic and seizure intervals were divided into non-overlapping 4-second segments (1024 points).
(1) Teager energy is greatly increased during seizure periods. Teager energy remains constant in normal newborns. (2) Newborns with seizures show greater ApEn values during interictal periods. ApEn values reach their greatest value during seizures. (3) Newborns with seizures show greater cross-ApEn values compared to newborns without seizures. (4) Coherence values were observed to significantly change at seizure times for frequencies of 10-35 Hz. (5) No differences in mutual information values are observed in newborns with and without seizures. (6) STLmax values are lower during the entire recording period in newborns with seizures. STLmax reach their lowest value during seizure periods, and are persistently higher in newborns without seizures.
Results suggest that measurable differences in brain electrical activity exist between normal newborns and those prone to seizures. Further, through linear and nonlinear time series EEG analyses, seizures can be detected. The long-term goal of this research will be to develop a real-time automated monitoring system capable of detecting newborn seizures, as well as differentiating normal from epileptic infants. To accomplish this goal, it will be necessary to establish in a larger cohort of newborns the normal range of measures that distinguish the category of newborn with seizures from the category of newborn without seizures by defining the optimal confidence interval as a function of conceptional age.
[Supported by: NIH, Childrens Miracle Network, US Vet Adm]