EEGs obtained from 55 newborns, recorded at the Montreal, Sydney, and Texas Children’s Hospitals, were reviewed by 3 types of automatic analysis of sequential epochs aimed at detecting rhythmic paroxysmal discharges, repetitive spike patterns, arrhythmic runs of spikes, and low frequency discharges, and the methods are reported from the Montreal Neurological Institute, and the Montreal Children’s Hospital, Canada. Initial evaluation detected 71% of seizures and 78% of seizure clusters, and the false detection rate was 1.7/hour of recording. 
COMMENT. The EEG is particularly important in the recognition of seizures in the newborn, since clinical observation alone may not be diagnostic. An algorithm that extracts rhythmic features from the EEG by spectral analysis may identify paroxysmal patterns indicative of seizure activity, but false detections may be a concern.
In a second publication, the authors evaluated their automatic EEG method of neonatal seizure detection, using recordings from a new set of 54 patients. The average seizure detection rate in the 3 institutions providing recordings was 69%, and the average false detection rate was 2.3/hour. Fluctuations in the false detection rates, ranging from a low of 1 to a high of 4/h, were a reflection of the technical quality and level of supervision of recordings. An experienced electroencephalographer must review “seizure” detections in conjunction with clinical observations, so that false or artifactual patterns may be excluded.