Purpose: Lung tissue elasticity distributions that are estimated from 5DCT are indicative of function and severity of lung disease progression. However, changes in patient breathing pattern during 5DCT acquisition could introduce perturbations in the estimated lung tissue elasticity. A systematic analysis of the lung tissue elasticity consistency is needed for validating its usage in a clinical workflow.
Methods: We quantitatively analyzed the consistency of the estimated lung elasticity using 10 lung cancer patients, for whom 5DCT data sets were retrospectively analyzed. For each subject, two 5DCT models were generated with a pre-specified finite time interval between the acquired set of images (12 free breathing CTs) in order to represent potential drifts in the patient breathing. Then, 5% and 85% end-exhalation CT datasets were generated from each of the 5DCT models to represent lung deformation during normal breathing. Using a well validated elastography process, we estimated the lung tissue elasticity using the deformation vector fields (DVF) that mapped the 5% to the 85% images. A finite element biomechanical lung model was employed for this purpose. The boundary constraints for the biomechanical model were obtained from the lung boundary DVFs. Quantitative analyses were performed on the resulting elasticity distributions to verify the model consistency. Specifically, histograms of 1 kPa bin width were generated and differences in voxel count were compared across 5DCT models.
Results: The estimated tissue elasticity matched the underlying lung substructureâ€™s elasticity from literature. In addition, the comparison of estimated elasticity distributions between models had an average weighted percent difference voxel count of 0.860% and 1.31%, for right and left lung, respectively.
Conclusion: These results suggest consistency in our lung biomechanical model and elasticity estimation process when using 5DCT models generated from free breathing scans.
Funding Support, Disclosures, and Conflict of Interest: This research was funded by National Institute of Health grant R56-HL139767.