UC Irvine

Hypoxic-ischemic encephalopathy (HIE) is a leading cause of death and long-term neurodevelopmental disability in neonates globally. Independent of etiology, high seizure burden is a significant risk factor for mortality in neonates. While prior EEG studies have shown that measures such as amplitude, power, entropy, and long-range temporal correlations can differentiate HIE severity grade, it is unknown whether these same measures can differentiate neonates who develop seizures in the first few days of life from those who do not. Understanding seizure risk in this population is critical for developing a treatment plan. Therefore, we investigated computational EEG markers of seizures in a retrospective cohort of 96 neonates with HIE who underwent therapeutic hypothermia (TH) at Children's Hospital of Orange County. Continuous scalp EEG was recorded throughout the cooling and rewarming period, and six categories of EEG metrics were computed across the 6-90 hour postnatal window: amplitude, spectral edge frequency (SEF), power spectral density (PSD), Shannon and permutation entropy, detrended fluctuation analysis (DFA) scaling exponent and intercept, and cross-correlation connectivity. Metrics were compared between neonates with (SZ+) and without (SZ-) seizures using the Wilcoxon rank-sum test, and across neonatal encephalopathy (NE) severity grades (mild, moderate, severe) using the Kruskal-Wallis test followed by pairwise comparisons, with false discovery rate (FDR) correction applied in both cases.Broadband amplitude and PSD in the delta, theta, alpha, beta, and broadband frequency bands were lower in the SZ+ group across most of the 84-hour study window; Shannon entropy in the delta, theta and alpha bands was also lower in SZ+ neonates, but this difference emerged later. Permutation entropy in the delta band was higher in SZ+ neonates and also emerged later in the timeframe. The DFA scaling exponent in the delta, theta, and alpha bands was higher in SZ+ neonates at most time points, and cross-correlation connectivity was higher in the SZ+ group during the majority of the recording. SEF, and DFA intercept did not differentiate the groups consistently across the timeframe. Most of these same metrics, with the exception of mean connectivity, also differentiated the severe NE grade from the mild and moderate grades, with no consistent differences found between mild and moderate grades. To our knowledge, this is the first study to show that seizure status in neonates with HIE can be quantitatively assessed from EEG activity during TH. These findings indicate that background EEG activity is altered in neonates who develop seizures, and this change can be detected early, during TH, using computational EEG metrics. This work identifies candidate biomarkers that could support earlier identification of seizure risk and inform clinical monitoring and intervention strategies for this vulnerable population.