Seizures are common in newborn infants with hypoxic-ischemic encephalopathy and are highly associated with adverse neurodevelopmental outcomes. The impact of seizure activity on the developing brain and the most effective way to manage these seizures remain surprisingly poorly understood, particularly in the era of therapeutic hypothermia. Critically, the extent to which seizures exacerbate brain injury or merely reflect the underlying evolution of injury is unclear. Current anticonvulsants, such as phenobarbital and phenytoin have poor efficacy and preclinical studies suggest that most anticonvulsants are associated with adverse effects on the developing brain. Levetiracetam seems to have less potential neurotoxic effects than other anticonvulsants but may not be more effective. Given that therapeutic hypothermia itself has significant anticonvulsant effects, randomized controlled trials of anticonvulsants combined with therapeutic hypothermia, are required to properly determine the safety and efficacy of these drugs. Small clinical studies suggest that prophylactic phenobarbital administration may improve neurodevelopmental outcomes compared to delayed administration; however, larger high-quality studies are required to confirm this. In conclusion, there is a distinct lack of high-quality evidence for whether and to what extent neonatal seizures exacerbate brain damage after hypoxia-ischemia and how best to manage them in the era of therapeutic hypothermia.

Hypoxic-ischemic encephalopathy is brain injury resulting from the loss of oxygen and blood supply around the time of birth. It is associated with a high risk of death or disability. The only approved treatment is therapeutic hypothermia. Therapeutic hypothermia has consistently been shown to significantly reduce the risk of death and disability in infants with hypoxic-ischemic encephalopathy. However, approximately 29% of infants treated with therapeutic hypothermia still develop disability. Recent preclinical and clinical studies have shown that there is still persistent neuroinflammation even after treating with therapeutic hypothermia, which may contribute to the deficits seen in infants despite treatment. This suggests that potentially targeting this persistent neuroinflammation would have an additive benefit in addition to therapeutic hypothermia. A potential additive treatment is Exendin-4, which is a glucagon-like peptide 1 receptor agonist. Preclinical data from various in vitro and in vivo disease models have shown that Exendin-4 has anti-inflammatory, mitochondrial protective, anti-apoptotic, anti-oxidative and neurotrophic effects. Although preclinical studies of the effect of Exendin-4 in perinatal hypoxic-ischemic brain injury are limited, a seminal study in neonatal mice showed that Exendin-4 had promising neuroprotective effects. Further studies on Exendin-4 neuroprotection for perinatal hypoxic-ischemic brain injury, including in large animal translational models are warranted to better understand its safety, window of opportunity and effectiveness as an adjunct with therapeutic hypothermia.

This study introduces transformer-based deep learning models as superior alternatives to convolutional neural network (CNN) architectures for seizure detection after hypoxia–ischemia (HI) in electroencephalography (EEG) recordings in the developing brain. We further demonstrate that these models excel in classifying of varied and subtle morphologies of electrographic seizures observed in the EEG of fetal sheep after HI. The study explores model training combinations from four cohorts of Romney/Suffolk fetal sheep (HI-normothermia-term (n = 7), HI-hypothermia-term (n = 14), sham-normothermia-term (n = 5), and HI-normothermia-preterm (n = 14)), totalling 31,015 EEG segments from 17,300 h of recordings. We evaluated multiple Transformer architectures using both 2D wavelet-scalograms (WS) and 1D raw EEG signals through leave-one-out and k-fold cross-validation, comparing models’ performance with or without sham-normothermia-term data and contrasting them with CNN-based results from our previous work. The Data-Efficient image Transformer (DeiT) emerged as the top-performing model, achieving 99.60 % accuracy (AUC = 0.992) when fed with WS, surpassing the 2D WS CNN result (98.94 % accuracy, AUC = 0.967) from our prior study. Notably, the Wav2Vec2 transformer model, fed with raw 1D EEG segments, demonstrated comparable performance with 99.60 % accuracy (AUC = 0.992), outperforming the 1D CNN result (98.43 % accuracy, AUC = 0.961) from the previous study. More importantly, the transformers can adeptly distinguish between seizures in both the preterm and term brain, and under the influence of treatment with 99.92 % accuracy (AUC = 0.997). Our findings highlight the superior ability of transformer models to capture long-range dependencies within EEG data, enhancing seizure detection without the added complexity of generating 2D WS images. Transformer models show promise for generalized automated seizure detection, irrespective of the perinatal brain maturation stage and the influence of hypothermia, offering potential improvements for seizure detection after HI in newborns.

Perinatal brain injury secondary to hypoxia-ischemia and/or infection/inflammation remains a major cause of disability. Therapeutic hypothermia significantly improves outcomes, but in randomized controlled trials nearly half of infants still died or survived with disability, showing that additional interventions are needed. There is growing evidence that brain injury spreads over time from injured to previously uninjured regions of the brain. At least in part, this spread is related to opening of connexin hemichannels and pannexin channels, both of which are large conductance membrane channels found in many brain cells. Opening of these membrane channels releases adenosine triphosphate (ATP), and other neuroactive molecules, into the extracellular space. ATP has an important role in normal signaling, but pathologically can trigger the assembly of the multi-protein inflammasome complex. The inflammasome complex promotes activation of inflammatory caspases, and release of inflammatory cytokines. Overall, the connexin hemichannel appears to play a primary role in propagation of injury and chronic disease, and connexin hemichannel blockade has been shown to be neuroprotective in multiple animal models. Thus, there is potential for some blockers of connexin or pannexin channels to be developed into targeted interventions that could be used in conjunction with or separate to therapeutic hypothermia.

Therapeutic hypothermia significantly improves outcomes after neonatal hypoxic-ischemic (HI) encephalopathy but is only partially protective. There is evidence that cortical inhibitory interneuron circuits are particularly vulnerable to HI and that loss of interneurons may be an important contributor to long-term neurological dysfunction in these infants. In the present study, we examined the hypothesis that the duration of hypothermia has differential effects on interneuron survival after HI. Near-term fetal sheep received sham ischemia or cerebral ischemia for 30 min, followed by cerebral hypothermia from 3 h after ischemia end and continued up to 48 h, 72 h, or 120 h recovery. Sheep were euthanized after 7 days for histology. Hypothermia up to 48 h recovery resulted in moderate neuroprotection of glutamate decarboxylase (GAD)⁺ and parvalbumin⁺ interneurons but did not improve survival of calbindin⁺ cells. Hypothermia up to 72 h recovery was associated with significantly increased survival of all three interneuron phenotypes compared with sham controls. By contrast, while hypothermia up to 120 h recovery did not further improve (or impair) GAD⁺ or parvalbumin⁺ neuronal survival compared with hypothermia up to 72 h, it was associated with decreased survival of calbindin⁺ interneurons. Finally, protection of parvalbumin⁺ and GAD⁺ interneurons, but not calbindin⁺ interneurons, with hypothermia was associated with improved recovery of electroencephalographic (EEG) power and frequency by day 7 after HI. The present study demonstrates differential effects of increasing the duration of hypothermia on interneuron survival after HI in near-term fetal sheep. These findings may contribute to the apparent preclinical and clinical lack of benefit of very prolonged hypothermia.

Brain maturity and many clinical treatments such as therapeutic hypothermia (TH) can significantly influence the morphology of neonatal EEG seizures after hypoxia–ischemia (HI), and so there is a need for generalized automatic seizure identification. This study validates efficacy of advanced deep-learning pattern classifiers based on a convolutional neural network (CNN) for seizure detection after HI in fetal sheep and determines the effects of maturation and brain cooling on their accuracy. The cohorts included HI–normothermia term (n = 7), HI–hypothermia term (n = 14), sham–normothermia term (n = 5), and HI–normothermia preterm (n = 14) groups, with a total of >17,300 h of recordings. Algorithms were trained and tested using leave-one-out cross-validation and k-fold cross-validation approaches. The accuracy of the term-trained seizure detectors was consistently excellent for HI–normothermia preterm data (accuracy = 99.5%, area under curve (AUC) = 99.2%). Conversely, when the HI–normothermia preterm data were used in training, the performance on HI–normothermia term and HI–hypothermia term data fell (accuracy = 98.6%, AUC = 96.5% and accuracy = 96.9%, AUC = 89.6%, respectively). Findings suggest that HI–normothermia preterm seizures do not contain all the spectral features seen at term. Nevertheless, an average 5-fold cross-validated accuracy of 99.7% (AUC = 99.4%) was achieved from all seizure detectors. This significant advancement highlights the reliability of the proposed deep-learning algorithms in identifying clinically translatable post-HI stereotypic seizures in 256Hz recordings, regardless of maturity and with minimal impact from hypothermia.