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

The choice of lung protective ventilation settings for mechanical ventilation has a considerable impact on patient outcome, yet identifying optimal ventilatory settings for individual patients remains highly challenging due to the inherent inter- and intra-patient pathophysiological variability. In this validation study, we demonstrate that physics-based computational lung models tailored to individual patients can resolve this variability, allowing us to predict the otherwise unknown local state of the pathologically affected lung during mechanical ventilation. For seven ARDS patients undergoing invasive mechanical ventilation, physics-based, patient-specific lung models were created using chest CT scans and ventilatory data. By numerically resolving the interaction of the pathological lung with the airway pressure and flow imparted by the ventilator, we predict the time-dependent and heterogeneous local state of the lung for each patient and compare it against the regional ventilation obtained from bedside monitoring using Electrical Impedance Tomography. Excellent agreement between numerical simulations and experimental data was obtained, with the model-predicted anteroposterior ventilation profile achieving a Pearson correlation of 96% with the clinical reference data. Even when considering the regional ventilation within the entire transverse chest cross-section and across the entire dynamic ventilation range, an average correlation of more than 81% and an average root mean square error of less than 15% were achieved. The results of this first systematic validation study demonstrate the ability of computational models to provide clinically relevant information and thereby open the door for a truly patient-specific choice of ventilator settings on the basis of both individual anatomy and pathophysiology.

💡 Research Summary

**
This paper presents the first systematic validation of patient‑specific, physics‑based computational lung models for mechanically ventilated ARDS patients. The authors enrolled seven ARDS patients who received routine chest CT scans and pressure‑controlled ventilation with recorded airway pressure, flow, and esophageal pressure waveforms. Using a deep‑learning segmentation pipeline, the CT data were processed to extract individualized anatomical representations of the lungs, lobes, and airways. These geometries were then mapped onto a reduced‑dimensional biomechanical model that captures nonlinear elastic and visco‑elastic behavior of lung tissue and airway resistance. The model is driven by the recorded ventilator waveforms, solving for time‑dependent airflow distribution and tissue deformation throughout the respiratory cycle.

A novel aspect of the work is the coupling of the biomechanical simulation with an electrodynamic model to generate synthetic Electrical Impedance Tomography (EIT) data. The regional ventilation predicted by the biomechanical model is translated into local changes in electrical conductivity, reflecting the inverse relationship between air content and tissue conductivity. The same 32‑electrode belt configuration used clinically is reproduced in the simulation, and the resulting voltage patterns are reconstructed with the GREIT algorithm to produce virtual EIT images.

The virtual EIT maps were directly compared with bedside EIT recordings, which serve as the reference standard. Validation metrics included Pearson correlation coefficients and root‑mean‑square error (RMSE) across three spatial scales: (1) anteroposterior ventilation profiles, (2) the full transverse cross‑section (32 × 32 pixels), and (3) the dynamic range over the entire breathing cycle. The model achieved a Pearson correlation of 0.96 for the anteroposterior profile, and an average correlation exceeding 0.81 with RMSE below 15 % for the full cross‑sectional analysis. Computational runtime for the entire pipeline (segmentation, biomechanical simulation, electrodynamic calculation, and EIT reconstruction) was on the order of 5–7 minutes, demonstrating feasibility for clinical workflow integration.

The authors discuss the clinical significance of these findings: the ability to predict heterogeneous regional ventilation and strain provides insight into the spatial distribution of ventilator‑induced lung injury (VILI), which cannot be inferred from global parameters such as tidal volume or airway pressure alone. By accurately reproducing bedside EIT, the model offers a trustworthy tool for personalized ventilator setting optimization, potentially reducing VILI risk.

Limitations acknowledged include the small cohort size, reliance on low‑resolution 32‑electrode EIT, the use of a single static CT (which may not capture rapid recruitment/derecruitment dynamics), and the simplified linear mapping between air content and electrical conductivity that neglects contributions from blood flow and inflammation.

Future work is outlined to incorporate multi‑time‑point imaging, integrate additional physiological sensors (e.g., ultrasound, blood gas analysis), and develop a closed‑loop system where the model informs real‑time ventilator adjustments.

In conclusion, the study demonstrates that a physics‑based, patient‑specific lung model, built from routinely available clinical data, can reliably predict regional ventilation as validated against bedside EIT. This establishes a solid foundation for model‑informed, precision ventilation strategies in the intensive care unit.