Individualised risk prediction of fetal infection and development in pregnancies that deliver preterm using advanced MRI techniques and machine learning

Research Question Can fetal MRI be utilised to establish an individualised, aetiology-driven prediction of infection and adverse neonatal outcomes associated with spontaneous preterm birth (sPTB) Background Very preterm birth (Aims and Objectives In a cohort of pregnancies that subsequently deliver very preterm to utilise MRI to: assess placental, fetal thymus, lung and regional brain macro and microstructure and perfusion, and relationships with proven infection. utilise machine learning to develop predictive tools, identifying: fetal infection. fetuses at highest risk of adverse outcome. Methods 100 women with uncomplicated pregnancies will be recruited and 75 at high risk of sPTB (16-32 weeks gestation) including those with: risk factors for sPTB (previous premature delivery or cervical surgery) and a high risk of delivery, determined by clinical predictive algorithms (https://quipp.org) or preterm premature rupture of membranes or exposed membranes through premature cervical shortening. Fetal MRI will be undertaken including T2, T2* (reflecting perfusion) and diffusion (reflecting microstructure) imaging of the fetal brain, thorax and placenta. Post-processing pipelines will generate thymus, placenta, lung and regional brain volumes and apparent diffusion co-efficients, fractional anisotropy values and T2* maps of the same regions. Imaging will be repeated every two weeks (maximum three) in the high-risk cohort if undelivered. Postnatal confirmation of infection will be ascertained from placental histology. Clinical outcome parameters including mortality, abnormalities on brain imaging, bronchopulmonary dysplasia or retinopathy of prematurity will be recorded to obtain a postnatal composite of adverse outcome. Additional datasets will be obtained from pregnancies imaged in other research studies (>650 datasets). Results of antenatal imaging findings and postnatal outcome measures will be used to generate predictive models for adverse outcomes. Models will be tested with datasets from collaborators (with >10% of the numbers for model development) prior to larger multicentre trials. Timelines for delivery Month 1-36 patient recruitment, MRI phenotyping of preterm and control pregnancies. 36-48 data analysis, predictive-model development, model testing, dissemination of findings, planning of RCTs. Impact and Dissemination Outputs will be disseminated through a study website, peer-reviewed journals, international conferences and through social media. The preterm birth PPI group and national pregnancy charities will also disseminate findings. Development of an effective prediction imaging tool for adverse fetal outcomes following sPTB would be followed by RCTs to evaluate if outcomes of pregnancies identified 'at risk' are improved by management strategies such as timing of delivery, or antenatal therapies.