P35-T An automated EEG grading system for the early neonatal brain injury severity prediction post-perinatal asphyxia

Abstract

Background Perinatal asphyxia can lead to hypoxic-ischemic encephalopathy (HIE). The prognosis is improved by therapeutic hypothermia, indicated for moderate and severe HIE. Reliable evaluation of cerebral injury is necessary in the first hours of life. The objective was to develop a quantitative EEG-based automatic grading system for neonatal HIE. Material and methods Neonatal EEGs were recorded in full term infants in the first 6 h of life after perinatal asphyxia. The severity of HIE was determined by the visual conventional EEG grades (French classification), assessed by two neurophysiologists blinded to clinical data. Six EEG quantitative features were selected based on their correlation scores with the 3 visual grades. Thereafter, the 6 selected features were analysed using Discriminant Factorial Analysis (DFA) to predict the severity grade and the long-term outcome. Results 90 EEGs were analysed between 2013 and 2017. The EEG quantitative features measuring the discontinuity and the amplitude of the signal were able to discriminate the 3 visual grades. The DFA results showed an accuracy of 86.7% for predicting EEG grades and 79.8% for predicting outcome at one year. Conclusion The proposed automated system using DFA was effective for grading initial EEG and predicting long-term outcome early after perinatal asphyxia. This system is based on simple quantitative features already proposed in marketed programs and could be easily used in clinical routine by unexperienced users. It may facilitate the evaluation of HIE’s severity within 6 h after birth and then be useful to determine whether therapeutic hypothermia has to be initiated.