Development of integrated models of the fetal and placental circulations to improve prediction of stillbirth

Objectives: Pregnancy disorders such as fetal growth restriction (FGR) and pre-eclampsia are leading contributors to stillbirth, but poor understanding of the underlying mechanisms limits detection of at-risk babies. We have previously developed placental computational models that link pathological changes in structure to altered haemodynamics. Here, we refine these models with patient-specific anatomical data, and expand them to incorporate the full fetal circulation.

Methods: A computational fetal circulation model was developed, including major vascular beds and an anatomically-realistic placenta, and personalised to data from one normal and two pathological pregnancies. Biopsies of perfusion-fixed placentae were imaged by in-line phase contrast synchrotron computed tomography (sCT). sCT data was processed and semi-automatically segmented using a deep learning approach. Placental models were parameterized with vascular:villous volume ratios (calculated by stereology), branch numbers, and villous volume fractions. Fetal umbilical artery (UA) Doppler waveforms were used to test the impact of placental structure on fetal haemodynamics.

Results: Sensitivity analysis of placental model showed that decreased placental branching densities (arising from increased branching angles, or decreased branching rates) had the largest haemodynamic impact, both increasing placental resistance. Parameterising the fetal model to FGR (decreasing vessel radius, increasing placental resistance 7-fold) predicted a 4.91% reduction in UA flow volume. Patient-specific models could be fitted to relevant UA Dopplers by only changing placental characteristics, with errors <2.3% in the S/D ratio and resistance index. However, reducing pulsatility index errors to <3% required also altering fetal parameters (with cardiac elastance having the greatest effect), suggesting a wider contribution of the fetal circulation to UA waveform shape.

Conclusion: These proof-of-principle computational models will be expanded with data from larger patient cohorts, with a wider view to identify haemodynamically informed imaging biomarkers predictive of stillbirth.