Flora Y. Wong

Predicting the cardiorespiratory status of extremely preterm infants using signals from bedside monitors is critical for timely intervention in neonatal intensive care units (NICUs), potentially improving long-term outcomes. However, directly adapting existing time series data analysis methods to this task is challenged by the scarcity of neonatal data and the high noise levels inherent to physiological signals in this population. To address these challenges, we propose a novel hybrid neural network that integrates C urriculum L earning and a C haos-aware G ating mechanism, called CLCG. Given the long multivariate physiological sequences, we first apply a sliding-window strategy to segment the data into overlapping windows. For each window, a shared Convolutional Neural Network (CNN) fuses multi-modal signals and extracts localized representations. To increase data diversity and enhance generalization capability, we introduce a curriculum learning-based data augmentation scheme that progressively reduces the number of windows during training. This encourages the model to learn discriminative representations by exposing it to increasingly challenging scenarios with less temporal context. To further mitigate the impact of signal noise and improve temporal modeling, we extract handcrafted features, the Lyapunov exponent, from the heart rate variability (HRV) parameters. These features are incorporated into a novel chaos-aware gate mechanism within the Long Short-Term Memory (LSTM) network. This mechanism guides the gating process based on nonlinear dynamical properties, and suppresses noise-related fluctuations. Finally, a Multi-Layer Perceptron (MLP) predicts the respiratory and hypotensive status over the next five hours. We evaluate our model on two datasets: an in-house cohort of 55 extremely preterm infants from the Monash Children's Hospital, and the publicly available MC-MED dataset from emergency department patients. Our approach outperforms 6 other state-of-the-art time-series models for both datasets, achieving significantly higher F1 scores ( p <0.05). Ablation studies confirm the effectiveness of both curriculum learning and chaos-aware gating mechanism, highlighting the potential of our framework for clinical prediction in high-noise, low-data neonatal care settings.

Neonatal brain injuries, such as stroke, cause focal ischaemic lesions that often result in lifelong neurological disabilities, yet effective treatments remain limited. Early-phase therapeutic screening requires models that can reliably reproduce injury severity while minimising confounding variables, including prolonged or variable anaesthesia, surgical stress, and invasive procedures that themselves affect injury progression. Existing models of neonatal focal ischaemia often exhibit high mortality, technical complexity, and substantial variability in lesion location and volume. As a result, there is a critical need for a rapid, ethically refined, and scalable neonatal model that produces consistent cortical injury suitable for screening neuroprotective, biomaterial-based, and regenerative therapies. We established a minimally invasive photothrombotic ischaemia model in postnatal day 10 rats by administering intraperitoneal Rose Bengal (25, 40, or 60 mg/kg) and activating it with a fixed 10-minute exposure to 565-nm light through the intact scalp and skull. This incision-free protocol allowed a total procedure duration of 19 min. We characterized dose-dependent effects on infarct volume and anatomical distribution, cortical atrophy, ventricular enlargement, apoptosis (cleaved caspase-3), astrocytic and microglial reactivity (glial fibrillary acidic protein, GFAP; ionized calcium-binding adapter molecule 1, Iba1), and sensorimotor outcomes (wire hang, cylinder rearing, adhesive tape removal) at 1, 7, and 14 days after injury. Additional analyses assessed the reproducibility of lesion size across litters and explored sex-specific differences. A 25 mg/kg dose induced a reliable and well-localized motor cortex infarct with no mortality. Higher doses of Rose Bengal produced proportionally larger infarcts with greater subcortical involvement and more pronounced secondary atrophy. Across all groups, apoptotic signalling and glial reactivity remained elevated through 14 days, indicating persistent tissue injury. Sensorimotor impairments were robust at all stages, with deficits in forepaw use, endurance, and tactile response correlating with lesion volumes in the 25 mg/kg group. No significant sex differences were observed for any histological or behavioural outcomes. This refined neonatal photothrombotic model provides a reproducible, simple, scalable, and ethically optimized platform for inducing severe focal cortical injury. The model’s stable injury territory, short, standardized procedure, and consistent functional readouts fill a major gap in current research tools and provide a practical foundation for early-phase testing of neuroprotective and regenerative interventions.

The COVID-19 pandemic drastically changed face-to-face interactions, as face masks cover facial zones and affect face perception. Preterm birth is associated with altered face perception, which is important in socio-cognitive development. Use of face masks in neonatal units could further impede face perception development. We assessed the impact of face masks on face perception, in preterm (23–30 weeks gestation at birth, n = 24) and term infants (n = 24) at 6 months’ post-term age, using functional near-infrared spectroscopy (fNIRS). The infants viewed alternating 5-s trials with images of full or masked happy faces, interspersed with 9-15 s baseline trials. Changes in oxy- and deoxy-haemoglobin concentration (ΔHbO and ΔHbR, µM), were measured using multichannel fNIRS covering inferior-frontal, temporo-parietal, and lateral occipital regions. Term infants showed higher ΔHbO than preterm infants for full faces in the right inferior-frontal region. The ΔHbO was higher for full than masked faces in the right temporo-parietal region for term infants, but not preterm infants. These findings reveal lower brain activation in preterm compared to term infants in the inferior-frontal region which is important for face and emotion processing. Preterm infants also show no differential brain response between full vs masked faces, reflecting altered face perception at 6 months’ post-term age.

Within the first 28 days after birth, more than 1 in every 2500 newborns will suffer a stroke. The weekly-adjusted risk of stroke for a term-born infant is threefold higher than for a male smoker aged 50 to 59 years with hypertension and diabetes. Neonatal stroke has significant clinical and socio-economic consequences, leading to cerebral palsy, epilepsy, and a range of motor, sensory, and cognitive impairments. Currently, there is no treatment for the brain damage caused by neonatal stroke. In this review, we outline the differences in the complex interplay of inflammation, excitotoxicity, oxidative stress, and cell death after stroke between adults and neonates, which limits the direct transfer of knowledge between studies for understanding injury. We comprehensively document what is known about the pathophysiology of neonatal stroke and critically evaluate current therapeutic strategies, emphasising the urgent need for innovative treatments tailored to suit the neonatal brain. This analysis reveals that treatment with an injectable hydrogel scaffold, a three-dimensional, water-swollen polymer network, may be an innovative, viable approach to improve outcomes for infants suffering from the most severe forms of brain injury arising from neonatal stroke.

Background: Kangaroo mother care (KMC) is WHO-recommended for low-birth-weight infants, yet its impact on autonomic cardiovascular control in preterm foetal growth-restricted (FGR) infants remains unclear. We hypothesised that KMC would promote autonomic cardiovascular control, benefiting preterm FGR infants with reduced baseline autonomic function compared to appropriate for gestational age (AGA) infants. Methods: Autonomic control was assessed via heart rate variability (HRV) in low frequency (LF) and high frequency (HF) bands using spectral analysis. Preterm FGR (n = 22) and AGA (n = 20) infants were assessed for 30-min before and 60-min during KMC. Comparisons were made between FGR and AGA infants; and between infants with baseline HRV below and above median. Results: Overall, no significant HRV changes were observed during KMC for FGR or AGA infants compared to baselines. Infants with low baseline HRV LF showed increased HRV during KMC (p = 0.02 and 0.05 for the entire group and FGR group, respectively). This effect was absent in the AGA group regardless of baseline HRV. Infants with high baseline HRV had decreased HRV during KMC. Conclusions: Infants with low baseline HRV, suggesting reduced autonomic control, are more likely to benefit from KMC with increased HRV. Further, this effect is stronger in FGR than AGA infants. Impact: Kangaroo mother care (KMC) is WHO-recommended for low-birth-weight infants, yet its impact on autonomic cardiovascular control in preterm foetal growth-restricted (FGR) infants is unclear. Preterm infants with low baseline heart rate variability (HRV) are more likely to benefit from KMC and increase their HRV suggesting improved autonomic control. This effect is stronger in preterm FGR infants than those with appropriate growth for age.

Summary: This observational study investigated the effects of sleep position and sleep state on short apneas and periodic breathing in hospitalized preterm infants longitudinally, in relation to postmenstrual age. Preterm infants (25–31 weeks gestation, n = 29) were studied fortnightly after birth until discharge, in prone and supine positions, and in quiet sleep and active sleep. The percentage of time spent in each sleep state (percentage of time in quiet sleep and percentage of time in active sleep), percentage of total sleep time spent in short apneas and periodic breathing, respectively, the percentage of falls from baseline in heart rate, arterial oxygen saturation and cerebral tissue oxygenation index during short apneas and periodic breathing, and the associated percentage of total sleep time with systemic (arterial oxygen saturation < 90%) and cerebral hypoxia (cerebral tissue oxygenation index < 55%) were analysed using a linear mixed model. Results showed that the prone position decreased (improved) the percentage of falls from baseline in arterial oxygen saturation during both short apneas and periodic breathing, decreased the proportion of infants with periodic breathing and the periodic breathing‐associated percentage of total sleep time with cerebral hypoxia. The percentage of time in quiet sleep was higher in the prone position. Quiet sleep decreased the percentage of total sleep time spent in short apneas, the short apneas‐associated percentage of falls from baseline in heart rate, arterial oxygen saturation, and proportion of infants with systemic hypoxia. Quiet sleep also decreased the proportion of infants with periodic breathing and percentage of total sleep time with cerebral hypoxia. The effects of sleep position and sleep state were not related to postmenstrual age. In summary, when sleep state is controlled for, the prone sleeping position has some benefits during both short apneas and periodic breathing. Quiet sleep improves cardiorespiratory stability and is increased in the prone position at the expense of active sleep, which is critical for brain maturation. This evidence should be considered in positioning preterm infants.

Background: Intrauterine inflammation is considered a major cause of brain injury in preterm infants, leading to long-term neurodevelopmental deficits. A potential contributor to this brain injury is dysregulation of neurovascular coupling. We have shown that intrauterine inflammation induced by intra-amniotic lipopolysaccharide (LPS) in preterm lambs, and postnatal dopamine administration, disrupts neurovascular coupling and the functional cerebral haemodynamic responses, potentially leading to impaired brain development. In this study, we aimed to characterise the structural changes of the neurovascular unit following intrauterine LPS exposure and postnatal dopamine administration in the brain of preterm lambs using cellular and molecular analyses. Methods: At 119–120 days of gestation (term = 147 days), LPS was administered into the amniotic sac in pregnant ewes. At 126-7 days of gestation, the LPS-exposed lambs were delivered, ventilated and given either a continuous intravenous infusion of dopamine at 10 µg/kg/min or isovolumetric vehicle solution for 90 min (LPS, n = 6; LPS_DA, n = 6). Control preterm lambs not exposed to LPS were also administered vehicle or dopamine (CTL, n = 9; CTL_DA, n = 7). Post-mortem brain tissue was collected 3–4 h after birth for immunohistochemistry and RT-qPCR analysis of components of the neurovascular unit. Results: LPS exposure increased vascular leakage in the presence of increased vascular density and remodelling with increased astrocyte “end feet” vessel coverage, together with downregulated mRNA levels of the tight junction proteins Claudin-1 and Occludin. Dopamine administration decreased vessel density and size, decreased endothelial glucose transporter, reduced neuronal dendritic coverage, increased cell proliferation within vessel walls, and increased pericyte vascular coverage particularly within the cortical and deep grey matter. Dopamine also downregulated VEGFA and Occludin tight junction mRNA, and upregulated dopamine receptor DRD1 and oxidative protein (NOX1, SOD3) mRNA levels. Dopamine administration following LPS exposure did not exacerbate any effects induced by LPS. Conclusion: LPS exposure and dopamine administration independently alters the neurovascular unit in the preterm brain. Alterations to the neurovascular unit may predispose the developing brain to further injury.

Background: Cord-clamping strategies may modify blood pressure (BP) and cerebral tissue oxygen saturation (rStO ₂ ) immediately after birth. Methods: We conducted a sub-study nested within the Baby-Directed Umbilical Cord-Clamping trial. Infants ≥32 ⁺⁰ weeks’ gestation assessed as requiring resuscitation were randomly allocated to either physiologically-based cord clamping (PBCC), where resuscitation commenced prior to umbilical cord clamping, or standard care where cord clamping occurred early (ECC). In this single-site sub-study, we obtained additional measurements of pre-ductal BP and rStO ₂ . In a separate observational arm, non-randomised vigorous infants received 2 minutes of deferred cord clamping (DCC) and contributed data for reference percentiles. Results: Among 161 included infants, n=55 were randomly allocated to PBCC (n= 30) or ECC (n= 25). The mean (SD) BP at 3-4 minutes after birth (primary outcome) in the PBCC group was 64 (10) mmHg compared to 62 (10) mmHg in the ECC group, mean difference 2 mmHg (95% confidence interval −3 – 8 mmHg, p= 0.42). BP and rStO ₂ were similar across both randomised arms and the observational arm (n=106). Conclusion: We found no difference in BP or rStO ₂ with the different cord clamping strategies. We report reference ranges for BP and rStO ₂ for late-preterm and full-term infants receiving DCC. Impact: Among late-preterm and full-term infants receiving resuscitation at birth, blood pressure (BP, at 3-4 minutes and 6-7 minutes) and cerebral tissue oxygen saturation (rStO ₂ ) are not influenced by timing of cord clamping in relation to establishment of ventilation. Infants in this study did not require advanced resuscitation, where different cord clamping strategies may yet influence BP and rStO ₂ . The reference ranges for BP and rStO ₂ represent the first, to our knowledge, for vigorous late-preterm and full-term infants receiving deferred cord clamping. rStO ₂ >90% (∼90 ^th percentile) may be used to define cerebral hyperoxia, for instance when studying oxygen supplementation after birth.