Abstracts

Rationale: Recently amplitude-integrated EEG (aEEG) has been widely used as a central nerve system monitor in neonates, and it's usefulness of neonatal seizure identification are reported. On the other hands, reports of density spectral array (DSA) in NICU remain scarce, although utility of DSA has been reported for seizure detection in adult and pediatric ICU. The aim of this study is to evaluate the diagnostic accuracy of DSA and aEEG, and the adequacy of reduced channel DSA and aEEG for neonatal and pediatric seizure identification. Methods: We transformed a set of 26 neonatal and 5 pediatric eight-channel EEGs with seizures into 4 types of single-channel DSA and aEEG displays (Fp1-Fp2, C3-C4, T3-T4, and O1-O2) and 4 types of double-channel DSA and aEEG displays (Fp1-C3 and Fp2-C4, C3-O1 and C4-O2, Fp1-T3 and Fp2-T4, and T3-O1 and T4-O2). Two authors with sufficient experience investigated the series of DSA and aEEG records, blinded to the raw EEG, and marked events suspected to be seizures. Their performance was compared to seizures identified on the underlying raw EEG. The underlying conditions in neonates were hypoxic-ischemic encephalopathy in 9, neonatal epilepsy in 6 and others in 11. All pediatric patients were diagnosed with acute encephalopathy. Results: In neonatal records, the 26 EEG recordings contained 103 discrete seizures over 27 hours. In pediatric records, the 5 EEG recordings contained 172 discrete seizures over 43 hours. In neonatal patients, median sensitivity for seizure identification across all recordings was 38 % (range: 28-46 % in each record) with DSA and 60 % (range: 45-75% in each record) with aEEG. On the other hands, median sensitivity for seizure identification across all recordings was 90 % (range: 81-92% in each patient) with DSA and 91 % (range: 78-94 % in each patient) with aEEG in pediatric patients. Mean misidentification rate was 0.29 times/hour with DSA and 0.39 with aEEG. In neonatal patients, sensitivity of aEEG was higher, and misidentification rate was lower than those of DSA (p <0.001). In comparison between single and double-channel recordings, sensitivity with double channels was higher than that of single channel both on DSA and aEEG (p =0.01) without significant difference in misidentification rate. On the other hand, in pediatric records, sensitivity and misidentification rate had no differences between DSA and aEEG, and had better results than that in neonatal patients. Conclusions: Our results showed that aEEG has an advantage over DSA for seizure identification in neonates. On the other hand, in pediatric patients, both DSA and aEEG was useful for seizure identification as previous report (Stewart, C et.al. Neurology 2010). The reason of this difference might be the difference of patients back grounds. EEG activities in the pediatric patients were suppressed or slowed by the disease or sedative agents, while neonates were not deeply sedated. The difference of underlying disorders and sedative agents may result in the difference of sensitivity and accuracy between neonatal and pediatric patients.