Burst Suppression EEG in Neonatal Convulsions

Abstract

Convulsions represent a characteristic signal of neurological disease in the newborn period. Single-channel EEG is a convenient tool for continuous evaluation of neonatal convulsions and gives valuable prognostic information on neurological recovery. Among various abnormal EEG waveforms during convulsions, burst suppression (BS) pattern is distinctive and usually indicates an urgent state that therapeutic interventions should be performed appropriately and instantly. Four temporal variables (i.e. burst suppression ratio, burst frequency, burst amplitude, and suppression amplitude) were selected as BS features in this paper to describe the convulsive EEG signals from 47 full term neonates. Subjects were divided into mild convulsive group (22 neonates) and serious convulsive group (25 neonates) according to their standard clinical diagnoses. Wilcoxon rank sum test shows that both the burst suppression ratio and burst amplitude are significantly different between mild and serious groups (p << 0.01) while burst frequency and suppression amplitude exhibit similar values in two groups (p = 0.76 and p = 0.46). According to linear discriminant analysis, nearest neighbor rule (NNR) is used to perform pattern classification in the plane supported by the first two discriminant BS features (i.e. burst suppression ratio and burst amplitude). The leave-one-out recognition rate of NNR is 0.96. Results in the present paper indicate that burst suppression ratio and burst amplitude are effective discriminant BS features which help to guarantee a rigorous evaluation of neonatal convulsions.