104 Automated Analysis of Long-Term Continuous Multichannel Eeg in Premature Infants

Abstract

Background: Continuous long-term multichannel EEG monitoring of the premature infant is feasible and could be a useful tool in early detection of pathological processes. This could help in early prediction of long-term outcome. To date interpretation and quantification of EEG have been based on visually edited recordings after rejection of artifacts. This is a cumbersome and time consuming process, specially in long-term monitoring. Thus a method for automated analysis is greatly needed. Aim: To develop an algorithm for automated analysis of the EEG background activity of the continuous EEG from day one to three. Method: 42 healthy infants with GA< 31 wk were included, monitored (NicoletOne monitor) continuously for 3 days soon after birth. 8 EEG electrodes were applied. The data were analyzed by visually removing artifacts (pruned), and by an algorithm removing the highest 5, 10, 15 or 20% of the total absolute bandpower (tABP). Results are depicted in fig 1. (δ-tABP) There were good correlations between the pruned and the 4 different algorithms used, the 5% removal showing the closest relationship with 95 % match of the removed data. Conclusion: This study shows a good relationship between pruned and automated analyses of the EEG recording.