Advanced analyses of physiological signals and their role in Neonatal Intensive Care

Abstract

Preterm infants admitted to the neonatal intensive care unit (NICU) face an array of life-threatening diseases requiring procedures such as resuscitation and invasive monitoring, and other risks related to exposure to the hospital environment, all of which may have lifelong implications. This thesis examined a range of applications for advanced signal analyses in the NICU, from identifying of physiological patterns associated with neonatal outcomes, to evaluating the impact of certain treatments on physiological variability. Firstly, the thesis examined the potential to identify infants at risk of developing intraventricular haemorrhage, often interrelated with factors leading to preterm birth, mechanical ventilation, hypoxia and prolonged apnoeas. This thesis then characterised the cardiovascular impact of caffeine therapy which is often administered to prevent and treat apnoea of prematurity, finding greater pulse pressure variability and enhanced responsiveness of the autonomic nervous system. Cerebral autoregulation maintains cerebral blood flow despite fluctuations in arterial blood pressure and is an important consideration for preterm infants who are especially vulnerable to brain injury. Using various time and frequency domain correlation techniques, the thesis found acute changes in cerebral autoregulation of preterm infants following caffeine therapy. Nutrition in early life may also affect neurodevelopment and morbidity in later life. This thesis developed models for identifying malnutrition risk using anthropometry and near-infrared interactance features. This thesis has presented a range of ways in which advanced analyses including time series analysis, feature selection and model development can be applied to neonatal intensive care. There is a clear role for such analyses in early detection of clinical outcomes, characterising the effects of relevant treatments or pathologies and identifying infants at risk of later morbidity.