A Multicenter Diagnostic Study on the Utility of AI-assisted Continue EEG Diagnosis of Neonatal Seizures

The occurrence of neonatal seizures may be the first, and perhaps the only, clinical sign of a central nervous system disorder in the newborn infant. The incidence of neonatal seizures is variable based on gestational age. The etiology of seizures may indicate the presence of a potentially treatable etiology and should prompt an immediate evaluation to determine the cause and to initiate etiology-specific therapy. Importantly, the earlier treatment of seizures positively affects the infant's long-term neurological development. However, even when continue electroencephalogram (cEEG) monitoring is available, the availability of on-site expertise to interpret cEEG signals is limited and in practice, the diagnosis is still based only on clinical signs. The previous study indicated that the reliable seizure detection was as little as 10% of seizure events. Therefore, an early automated seizure detection tool has been developed based on machine learning. The lack of an automated seizure detection tool has been validated in the external neonatal seizures cohort. The evidence on the utility of the automated seizure detection tool remains uncertain. This is a prospective, continuous double-blind designed diagnostic accuracy study. The study aims to validate the accuracy of the artificial intelligence (AI)-assisted cEEG diagnostic tool comparing the manually cEEG interpretation as the golden standard of neonatal seizure in neonatal intensive care units. Analysis of sensitivity and specificity is to evaluate the accuracy of AI-assisted cEEG diagnostic tool.