Human airway material characterization via inverse finite element analysis and neural network surrogate

This study combined planar biaxial mechanical testing, inverse finite element analysis, and a neural-network surrogate to characterize the anisotropic material properties of human central airway tissues. Tracheal, left primary bronchial, and right primary bronchial smooth muscle specimens were obtained from seven transplant-eligible donor lungs, producing 46 samples for mechanical characterization. Equibiaxial tensile testing revealed strongly nonlinear, bilinear, and anisotropic behavior, with substantial stiffening at higher strains. Experimental force-displacement data were used to calibrate a Holzapfel-Gasser-Ogden constitutive model containing parameters describing the isotropic matrix response, collagen-fiber contribution, nonlinear fiber recruitment, and fiber dispersion. Traditional inverse finite element analysis was first performed on nine reference specimens, after which 2,000 finite element simulations were used to train a feedforward neural-network surrogate. The trained network reproduced previously unseen finite element responses with approximately 0.9% error and reduced material-parameter calibration for all 46 samples from an estimated several weeks of conventional inverse analysis to approximately six minutes. Mean calibrated HGO parameters were similar among the trachea and primary bronchi, with no statistically significant regional differences. The fitted fiber-dispersion parameter remained low across regions, supporting strong preferential fiber alignment in the axial airway direction. Direct analysis of the experimental stress-strain curves also showed marked nonlinear stiffening, with human tracheal initial moduli of approximately 22.3 ± 7.4 kPa axially and 13.2 ± 6.5 kPa circumferentially, increasing to ultimate moduli of approximately 592.3 ± 542.1 kPa axially and 284.1 ± 355.3 kPa circumferentially. Overall, the study provides experimentally informed constitutive properties for human tracheal and bronchial tissues and an efficient computational framework intended to improve airway biomechanical models, patient-specific stent simulations, and intervention planning.