Data-driven wavelet coherence approach to assess neurovascular coupling in neonatal hypoxic-ischemic encephalopathy

Neonatal hypoxic-ischemic encephalopathy (HIE) remains a leading global cause of mortality and morbidity. Neurovascular coupling (NVC), assessed via wavelet transform coherence (WTC) between electroencephalogram and cerebral tissue oxygenation ( ${\mathrm{SctO}}_2$ ), has shown promise in identifying brain injury severity; however, statistical estimation of WTC was derived using Monte Carlo (MC) simulations with surrogate non-physiological signals.

To assess NVC using a data-driven method without MC for distinguishing the severity of HIE on the first day of life.

NVC was assessed on the first day of life with direct data in neonates diagnosed with HIE ranging from mild to severe. Neonates with moderate to severe HIE received therapeutic hypothermia (TH) at 5 h of life (TH group), whereas those with mild HIE did not based on evidence-based protocols (non-TH group). Significant time-scale ranges in WTC were identified using a cluster-based permutation test, and average NVC within these ranges was compared among groups with a linear mixed-effects model over the first 20 h of recording.

A total of 57 full-term neonates with HIE (29 non-TH and 28 TH) were included. The linear mixed-effects revealed a significant interaction between group and time within the 25- to 60-min time (0.28 to 0.67 mHz) scale ( $p<0.001$ ), indicating reduced NVC with increased encephalopathy severity. Specifically, NVC was significantly reduced in neonates in the TH group ( $p=0.0103$ ).

We demonstrate that a data-driven approach can significantly distinguish NVC patterns by HIE severity without relying on MC simulations. By enhancing robustness and bedside applicability in the early hours of life, it may support informed decisions regarding initiation of TH to improve outcomes.