Reliability of estimating stochastic lung tissue heterogeneity from pulmonary impedance spectra: a forward-inverse modeling study.

Heterogeneity of regional lung mechanics is an important determinant of the work of breathing and may be a risk factor for ventilator associated lung injury. The ability to accurately assess heterogeneity may have important implications for monitoring disease progression and optimizing ventilator settings. Inverse modeling approaches, when applied to dynamic pulmonary impedance data (Z(L)), are thought to be sensitive to the detection of mechanical heterogeneity with the ability to characterize global lung function using a minimal number of free parameters. However, the reliability and bias associated with such model-based estimates of heterogeneity are unknown. We simulated Z(L) spectra from healthy, emphysematous, and acutely injured lungs using a computer-generated anatomic canine structure with asymmetric Horsfield branching and various predefined distributions of stochastic lung tissue heterogeneity. Various inverse models with distinct topologies incorporating linear resistive and inertial airways with parallel tissue viscoelasticity were then fitted to these Z(L) spectra and evaluated in terms of their quality of fit as well as the accuracy and reliability of their respective model parameters. While all model topologies detected appropriate changes in tissue heterogeneity, only a topology consisting of lumped airway properties with distributed tissue properties yielded accurate estimates of both mean lung tissue stiffness and the spread of regional elastances. These data demonstrate that inverse modeling approaches applied to noninvasive measures of Z(L) may provide reliable and accurate assessments of lung tissue heterogeneity as well as insight into distributed lung mechanical properties.