Abstract
Accurate and rapid characterization of lung mechanics remains a major challenge in the management of respiratory disease. Physics-informed poroelastic finite-element (FE) models can capture detailed tissue–airflow interactions, but their substantial computational expense precludes real-time clinical deployment. Conversely, lumped-parameter compartment models are routinely used at the bedside for fast assessment of disease severity, yet they provide limited mechanistic insight and no spatial resolution. In this work, using porcine-specific data, we integrate multi-fidelity poroelastic FE simulations with machine learning to enable fast, uncertainty-aware estimation of respiratory system compliance and resistance.