Fifty-seventh annual meeting of the American association of physicists in medicine
TH-CD-303-01: On Characterizing 4D Lung Motion Using Coupled Multi-Physics Framework of Computational Fluid Dynamics, Linear Elasticity Estimation and 4D Lung Deformable Image Registration
Modeling human lung dynamics is critical in quantitatively representing the subject specific 4D tumor motion. However, the process is numerically complicated by fluid-structure interactions (FSI) between air and the visco-elastic lung tissue and the lack of information on the anisotropic lung elasticity.
We present a mathematical method to couple image registration and physics-based modeling approaches in order to represent a consistent volumetric lung dynamics. The Computational Fluid Dynamics (CFD) simulation assumes the lung to be a poro-elastic medium with spatially dependent elastic property. The simulation is performed on a 3D lung geometry reconstructed from four-dimensional computed tomography (4DCT) scan dataset of a human subject. The heterogeneous linear elastic Young's Modulus (YM) values are estimated from a linear elastic deformation model with the same lung geometry and 4D lung Deformable Image Registration (DIR). The deformation obtained from the CFD simulation is then coupled with the displacement obtained from the DIR by means of the Tikhonov regularization (TR) algorithm. Numerical results present an analytic inter-model consistent estimation of the volumetric lung dynamics. The fusion method was validated by comparing the optimal displacement with 5D model-generated landmarks.
A set of five 4DCT datasets was employed for this study. For the landmarks inside the lung, the optimization results in a significant reduction in error. Specifically, in the left lung, the average error for the landmarks on the surface and edges is reduced by 1.11 mm while this value is 2.96 mm for the nodes, while, in the right lung, the average displacement errors after optimization are reduced by 1.34 mm and 0.86 mm for the interior and surface nodes respectively.
These results shows that although optimization reduces the displacement error both at the surface and the interior nodes, it is more effective for the error reduction of interior nodes.