COMPUTATIONAL lung models accurately predicted key ventilation parameters in patients with acute respiratory distress syndrome (ARDS).
A prospective clinical pilot study involving 10 mechanically ventilated patients demonstrated strong agreement between personalized computer simulations and measured respiratory mechanics, suggesting that physics-based modeling could eventually support more individualized ventilation strategies.
Researchers evaluated 10 adults with mild, moderate, or severe ARDS, developing individualized computational lung models using computed tomography (CT) imaging and clinical respiratory measurements.
Each model simulated how the patient’s lungs responded to changes in ventilator settings, including positive end-expiratory pressure (PEEP), tidal volume, and driving pressure.
The models incorporated patient-specific lung anatomy and physiological characteristics to estimate both overall respiratory mechanics and regional ventilation distribution.
Predictions were compared with clinical measurements obtained during standardized ventilatory maneuvers.
Computational lung models demonstrated strong correlations with measured tidal volumes (r=0.925; p<0.001) and driving pressures (r=0.940; p<0.001) across evaluated PEEP levels.
The root mean square error was 53 mL for tidal volume and 1.328 mbar for driving pressure.
Predicted end-expiratory transpulmonary pressures also closely correlated with clinical measurements (r=0.971; p<0.001).
Regional ventilation predictions were assessed using electrical impedance tomography (EIT), which measures ventilation distribution across different lung regions.
At initial PEEP settings, five of seven patients with available EIT data showed strong agreement between simulated and measured ventilation patterns, with correlation coefficients exceeding 0.8.
However, predictions were less accurate for ventilation distribution between the front and back of the lungs, particularly in dorsal regions.
The findings suggest that computational lung models could eventually help clinicians anticipate physiological responses to ventilator adjustments, potentially supporting lung-protective ventilation strategies tailored to individual patients.
However, prediction errors remained clinically relevant. Tidal volume errors reached 77 mL under some conditions, while regional ventilation estimates sometimes overestimated dorsal lung aeration.
The small cohort, repeated measurements within patients, and variable regional accuracy limit the findings. Model calibration also required manual optimization, and simulations were analyzed offline rather than during real-time clinical care.
Although the results support further development of personalized ventilation modeling, larger prospective studies are needed to establish whether model-guided adjustments improve clinical outcomes.
Reference
Sablewski A et al. Predicting global and regional respiratory response in ARDS patients through individualized physics-based computational lung models: a prospective clinical pilot study. Respir Res. 2026;27(1):414.
Featured Image: Emine on Adobe Stock.
Featured Image: Emine on Adobe Stock.
Each article is made available under the terms of the Creative Commons Attribution-Non Commercial 4.0 License.
