Audio signal assists in rapid review of neonatal electroencephalography recordings

Background: Reviewing the 'raw' electroencephalography (EEG) waveform when assessing suspected seizures on amplitude-integrated electroencephalography (aEEG) traces is important but can be time-consuming. Video and audio fast playback of long EEG recordings was previously used for review of traces from analog EEG recording systems. This study aimed to develop and apply a system for rapid visual and audio review of digital neonatal EEGs from aEEG monitoring. 

Methods: Two channels of aEEG with 'raw' EEG were monitored in newborns at risk of seizures. We developed software using MATLAB (The MathWorks Inc., USA) to generate sound and display traces from rapid EEG playback. 

Results: Recordings of 3 hours duration from eight term newborn infants were played back at 60 times the recorded speed (at 60 times, 3 hours plays in 3 minutes); seizure activity was noted to have a distinct buzz-type sound differentiable from background activity. Of 170 combined EEG seizure events in the two channels, six were missed on audio, and there were four false positive events. The sensitivity and positive predictive value were 0.96 and 0.98, respectively.

Conclusions: A computer system for rapid EEG review from neonatal aEEG recordings was developed and applied; the software generates EEG signal sound as well as simultaneously visually displaying the EEG traces. Audio generated from EEG traces can help identify seizure activity and aid background assessment.