Can MRI at term equivalent age predict neurodevelopmental outcomes in preterm born infants?

Student thesis: Doctoral ThesisDoctor of Philosophy


More than 15 million infants are born prematurely worldwide each year and the incidence is rising.1 Preterm birth represents the single largest cause of mortality in under five year olds globally1 and whilst progress in perinatal medicine has resulted in improved survival, the rates of major morbidity are unchanged and endemic amongst the survivors.2

The introduction of antenatal steroids and routine surfactant administration in these infants, coupled with cessation of post-natal steroids, has coincided with a reduction in major lesions such as periventricular leukomalacia (PVL) and haemorrhagic parenchymal infarction (HPI) and a resultant reduction in cerebral palsy rates. 2–5 Despite these advances, 25% of extremely preterm-born infants still experience moderate to severe neurodevelopmental impairment and up to 50% have milder deficits including cognitive, fine motor and visuo-perceptual problems, evident by early childhood.6,7

Cranial ultrasound has long been the stalwart of preterm brain imaging and is readily able to identify the aforementioned major lesions however these are now only seen in 10-14% of preterm infants which proffers the question: why do so many preterm-born infants have poor neurodevelopmental outcomes? The emergence of neonatal magnetic resonance imaging (MRI) identified several new entities that may, in part, explain some of the neurodevelopmental impairments seen in these infants and also identified imaging biomarkers able to aid in prognostication, thus allowing targeting of early rehabilitative interventions. In addition, advanced MRI techniques have offered surrogate markers of the underlying pathological process and demonstrated differences in regional brain volumes compared to term-born infants. However most of these advanced techniques currently remain time-consuming, confining them for use as research tools.

Several studies have attempted to form tools that offer better prediction of motor and cognitive outcomes in preterm-born infants using standard MRI sequences and whilst they succeeded in improving prognostication, the result was scoring systems, ill-suited for busy clinical radiological practice.8,9 This thesis sought to utilise standard structural MRI sequences in order to better understand what is ‘normal’ for the preterm brain at term-equivalent age and define simple tools that would be immediately applicable to clinicians reporting preterm neonatal brain MRI and aid them in offering more accurate prognostic information.

Chapter 5 examined the frequency of acquired lesions in a large, unselected cohort of very preterm born infants (<33 weeks gestation) undergoing routine MRI at term-equivalent age (TEA) and related these lesions to neurodevelopmental outcomes at 20 months. We found lesions in this population were common, occurring in 76% of infants however normal neurodevelopment was seen in 56% of the cohort. We concluded that a normal MRI in a preterm-born infant imaged at term-equivalent age is specific but poorly sensitive for a normal motor and cognitive outcome. Lower socioeconomic status and longer periods of ventilation were universally associated with poorer outcomes.

Chapter 6 sought to determine whether anthropomorphic measures of the brain were different between term and preterm born infants at TEA and understand how and when the growth trajectory of the preterm brain deviates from that of a healthy fetus or term neonate. We found few measures were correlated with gestational age but all correlated with post menstrual age and postulated that this reflected the accelerated growth phase known to occur post term. This was confirmed by the regression slopes comparing fetal and term born controls showing a steeper slope in the post natal period. Measures confirmed the scaphocephalic preterm head shape and the suggestion of a dose-dependent association between prematurity and biparietal diameter with major lesions demonstrating little effect on biometry. Smaller corpus callosal length, vermis height and pontine diameter were associated with adverse motor outcomes which we postulated reflected Wallerian degeneration in white matter tracts.

Chapter 7 built on the findings from chapter 5 to define ‘normal’ anthropomorphic measures of the brain for the preterm infant imaged at TEA. Overall, we found measures falling below the fifth centile to be poorly sensitive but highly specific at predicting outcomes at 20 months and performed better at predicting motor than cognitive development. These findings reflect the wider paradigm of trying to predict later cognitive outcomes in the neonatal period; a domain that is so heavily influenced by other factors.

Chapter 8 used both quantitative and qualitative measures to examine myelination within the posterior limb of the internal capsule (PLIC). Previous studies have demonstrated strong associations between abnormal appearances of the myelin within the PLIC and poor motor outcomes in preterm infants with HPI and term infants with hypoxic-ischaemic injury.10,11 We found that even in the absence of major lesions, the myelin within the PLIC was significantly thinner in the preterm compared to term-born infants and that this translated to poorer motor outcomes. Where major lesions were present, quantitative measures of the myelin offered additional prognostic information on cognitive outcomes. Qualitative assessment of the myelin within the PLIC predicted all infants with high levels of functional motor impairment at 20 months and continued to offer improved prognostic information beyond the presence of major lesions.

Chapter 9 concludes that MRI in very preterm infants imaged at term equivalent age is useful in contributing towards prediction of neurodevelopmental, particularly motor, outcomes however it is probably unrealistic to expect structural imaging at such an early age and from a single time-point, to predict longer term prognosis. This is particularly true of cognitive and language outcomes where extraneous factors such as socioeconomic status and environment are all contributory. This chapter also discusses possible directions of future work, including models incorporating all the facets of this thesis, in order to offer individualised prognosis as well as the use of advanced myelin-specific techniques.

Date of Award1 Jun 2022
Original languageEnglish
Awarding Institution
  • King's College London
SupervisorMary Rutherford (Supervisor) & David Edwards (Supervisor)

Cite this