Digital Twins of Mechanically Ventilated Preterm Neonates with Respiratory Distress Syndrome
Background: Mechanical ventilation is life-saving for preterm infants with respiratory distress syndrome but can also contribute to lung injury and long-term morbidity. Protective ventilation strategies are recommended, yet implementation in neonatal intensive care units remains inconsistent, and infants continue to be exposed to injurious ventilator settings. Objective: To develop and validate a cohort of neonatal digital twins, based on mechanistic models of cardiopulmonary physiology calibrated to individual patient data, as a tool for simulating and optimising protective ventilation strategies. Methods: A high-fidelity computational simulator of human cardiopulmonary physiology was adapted to neonatal-specific parameters, including lung compliance, dead space, pulmonary vascular resistance, oxygen consumption, and fetal haemoglobin oxygen affinity. Digital twins were generated using data at 65 time points from 11 preterm neonates receiving volume-controlled ventilation. Model parameters were calibrated to minimise the error between simulated and observed PaO2, PaCO2, and peak inspiratory pressure (PIP). Results: Digital twins reproduced measured data with mean absolute percentage errors of 3.9% (PaO2), 3.0% (PaCO2), and 5.8% (PIP) across the cohort. Predictions for uncalibrated variables (pHa, SaO2, mean and minimum airway pressure) also showed high accuracy, with errors <5%. Strong correlations and narrow limits of agreement were observed across all patients and time points. Conclusions: This study demonstrates, for the first time, the feasibility of creating fully mechanistic digital twins of mechanically ventilated neonates with RDS. The twins accurately captured patient-specific gas exchange and respiratory mechanics, supporting their potential as a platform for conducting virtual clinical trials and for the design of individualized, lung-protective ventilation strategies.