Physics-Informed Digital Twin of Maternal-Fetal Hemodynamics for Predictive Risk Simulation in Preeclampsia

Area of Science:

• Biomedical Engineering
• Maternal-Fetal Medicine
• Computational Physiology

Background:

• Preeclampsia (PE) is a pregnancy disorder causing endothelial dysfunction and high maternal-fetal morbidity.
• Existing biomarkers aid diagnosis, but few systems link molecular changes to hemodynamic abnormalities.
• Predictive models for PE often lack integration of physiological constraints.

Purpose of the Study:

• To develop a Physics-Informed Digital Twin (PI-DT) simulating uteroplacental hemodynamics.
• To integrate molecular markers with physical flow abnormalities for risk prediction.
• To create an interpretable model for maternal-fetal risk assessment in preeclampsia.

Main Methods:

• Developed a PI-DT using the Navier-Stokes equations for uterine blood flow.
• Integrated clinical inputs like mean arterial pressure, sFlt-1, PlGF, and hepatic function.
• Employed a Physics-Informed Neural Network to update model parameters from clinical data.

Main Results:

• Simulated uterine-artery resistance indices showed strong correlation with Doppler measurements (R2 = 0.88, p < 0.001).
• The model accurately reproduced associations between high sFlt-1/PlGF ratios, elevated MAP, and increased preterm birth and NICU admission rates.
• The PI-DT successfully linked angiogenic imbalance to measurable hemodynamic dysfunction.

Conclusions:

• The PI-DT provides a novel framework for simulating uteroplacental hemodynamics under physiological constraints.
• This model enables interpretable prediction of maternal-fetal risk by linking molecular and hemodynamic factors in preeclampsia.
• The PI-DT demonstrates potential for improving early risk stratification and management of preeclampsia.