Patient-specific prediction of regional lung mechanics in ARDS patients with physics-based models: A validation study

Lung-protective ventilation significantly influences outcomes in ARDS patients, but identifying optimal settings remains a challenge due to pronounced inter- and intra-patient variability in lung anatomy and pathophysiology. This study demonstrates that physics-based computational lung models tailored to individual patients can predict otherwise unobservable local lung states, enabling a quantitative analysis of regional ventilation and the mechanical load experienced by lung parenchyma during ventilation. For seven mechanically ventilated ARDS patients, patient-specific computational models were generated using chest CT scan and ventilatory waveform data. By numerically resolving the lung‘s interaction with ventilator-imposed pressure and flow, we predict both the regional ventilation as well as the dynamic, spatially heterogeneous states of the lung. Model-predicted ventilation distributions were validated against clinical measurements from bedside Electrical Impedance Tomography (EIT). The predicted anteroposterior ventilation profiles exhibited excellent agreement with EIT, achieving a Pearson correlation of 96%. Across the full transverse cross-section and over the dynamic ventilation range, the models achieved an average correlation exceeding 81% and a root mean square error below 15%. This first systematic validation study indicates that computational lung models can reliably estimate patient-specific regional ventilation. These findings support the use of such models as a tool for individualized decision-making in mechanical ventilation, offering insights into both anatomical and functional lung characteristics that are not directly observable at the bedside. By leveraging detailed patient data and physical modeling, these models have the potential to inform more personalized and physiologically grounded ventilator settings, improving care in critically ill ARDS patients.