Extremely preterm infants born before 28 weeks gestation are at high risk for neurodevelopmental impairments. Amplitude-integrated EEG (aEEG) accompanied by raw EEG traces (aEEG-EEG) during the first days after birth could help predict outcomes in these infants. This study aimed to determine if specific qualitative and quantitative aEEG-EEG features predict cognitive, motor, and behavioral outcomes at ages 2-3 and 5-7 years in extremely preterm infants.

This retrospective cohort study analyzed aEEG-EEG recordings from the first 3 days after birth for extremely preterm infants born before 28 weeks gestation at Wilhelmina Children’s Hospital, Netherlands between 2008-2018. Infants with genetic/metabolic diseases or major malformations were excluded. Qualitative features were extracted, including background pattern, sleep-wake cycling, and seizures. Quantitative metrics were also extracted, grouped into spectral content, amplitude, connectivity, and discontinuity. Machine learning models evaluated if these early aEEG-EEG features predicted outcomes at follow-up, controlling for potential confounders like illness severity and medications.

Key findings showed background pattern was the strongest predictor. Infants with discontinuous background patterns were more likely to have cognitive, motor, and behavioral problems at follow-up. Quantitative features also had predictive value – increased discontinuity and decreased lower-frequency activity predicted worse outcomes. Sleep-wake cycling and seizures occurred too infrequently to assess predictive utility.

This study found early aEEG-EEG background patterns and quantitative metrics in extremely preterm infants provided valuable prognostic information about neurodevelopmental impairments at ages 2-7 years. Discontinuous background and increased discontinuity specifically were associated with cognitive, motor, and behavioral problems. These findings highlight the potential for automated, interpretable analysis of early aEEG-EEG features to aid risk stratification, decision-making, and intervention planning for this high-risk population. Future research should explore integrating these predictive EEG biomarkers into an automated prognostic tool to enable individualized predictions and support precision care for extremely preterm infants.