View Abstract

A4681 - A Computational Model of Lung Tissue Mechanics Representing Physiological Heterogeneity in Density from Medical Imaging Data in Healthy Lungs
Author Block: L. Parisi1, H. Kumar1, A. R. Clark1, K. S. Burrowes2, E. A. Hoffman3, S. R. Hopkins4, M. H. Tawhai1; 1Auckland Bioengineering Institute, University of Auckland, Auckland, New Zealand, 2Department of Chemical and Materials Engineering, and Auckland Bioengineering Institute, University of Auckland, Auckland, New Zealand, 3Department of Radiology, Carver College of Medicine, University of Iowa, Iowa City, IA, United States, 4Department of Radiology and Department of Medicine, University of California, San Diego, La Jolla, CA, United States.
RATIONALE
Lung tissue is highly deformable, and its regional expansion determines its ability to act as an efficient gas exchanger. Lung tissue is not a uniform continuum, as there are localised regions of stiffer tissue that are more pronounced in pathology. However, most of the existing computational models of lung tissue deformation, important in predicting the relationship between pathology and local tissue stress, assume uniform tissue properties. In this study, we present a methodology to parameterize computational models of lung tissue using subject-specific tissue density derived from CT. We show that incorporating regional variation in tissue properties is essential to capture physiological heterogeneity in tissue density observed from imaging more accurately.
METHODS
CT images at functional residual capacity in 24 healthy adult subjects (11 males, 13 females; age:23.4±4.2; height:1.72±0.10m; BMI:24±3kg/m2) were analyzed from the “Human Lung Atlas” (NIH R01-HL-064368). The distribution of CT-derived tissue density was characterized by calculating the coefficient of variation (CoV), gradient in the gravitational direction, and area under the curve (representing the relationship between density and lung height in the anterior-posterior, cranio-caudal and medio-lateral directions). Lung structures were segmented, and subject-specific finite element meshes were generated. Mean lung density and stiffness were used as model inputs. We hypothesized that heterogeneity in tissue mechanics is associated with lung structure. Therefore, a spatial distribution of tissue density was applied to the computational model. Macro-scale lung tissue deformation was computed using a nonlinear hyperelastic material law. The relationship between CT- and model-based outputs of tissue density was assessed using bivariate linear regression via the Pearson’s product-moment correlation coefficient (r2). To determine whether the model provided an improved prediction of lung tissue expansion compared with previous models that typically assume uniform tissue properties, CT-based density distributions were compared with predicted density distributions.
RESULTS
Our model improved prediction of CoV (38% CT vs 38% our model, r2=0.84, p=0.03) compared to the same model but with uniform tissue properties (38% CT vs 14% uniform model, r2=0.24, p=0.20). Gradients of density were also better predicted when using heterogeneous densities (r2=0.78, p=0.0001) than when assuming uniform density (r2=0.31, p=0.12).
CONCLUSION
We show that a finite element model, parameterized to an individual, more accurately represents the underlying density distribution from medical imaging data (CT scans) than a model with the typical assumption of uniform density in the unloaded state. This is potentially important for providing clinically valuable quantification of effects due to pathophysiological tissue deformations.