A Quantitative Analysis of Lung Elastography Performance Using Large-Deformation CT Scans Acquired at Residual Volume (RV) and Total Lung Capacity (TLC)
B Stiehl1*, M Lauria1, L Naumann1, K Singhrao1, I Barjaktarevic2, J Goldin2, M McNitt-Gray3, D Low1, A Santhanam1, (1) Department of Radiation Oncology, University of California, Los Angeles, Los Angeles, CA, (2) Division of Pulmonary and Critical Care, University of California, Los Angeles, Los Angeles, CA, (3) Department of Radiology, University of California, Los Angeles, Los Angeles, CA
Presentations
MO-B-TRACK 6-1 (Monday, 7/26/2021) 11:30 AM - 12:30 PM [Eastern Time (GMT-4)]
Purpose: Lung elastography during normal breathing has proven to be a useful tool in characterizing regional lung function and has been shown to be an image-based biomarker for COPD patients. However, residual volume (RV) and total lung capacity (TLC) scans representative of deep breathing are often acquired for COPD patients. In this study, we systematically analyze the feasibility of performing lung elastography using large deformation breath-hold CT images acquired at RV and TLC.
Methods: A quantitative analysis of biomechanical model performance for elasticity estimation was performed using 10 patient datasets with CT scans acquired at RV and TLC, retrospectively. Deformable image registration (DIR) was performed between RV and TLC image pairs with a published registration technique (pTVreg) to obtain ground-truth deformation vector fields (DVFs). For each DVF, a well-validated inverse elasticity approach, which employs a lung biomechanical model, was performed to calculate elasticity estimations for each voxel. For validation of elastography performance, the percentage of voxels converging successfully in accordance with the ground-truth DVFs was calculated with convergence criteria of 5% and 10% of maximum observed deformation. Jacobian distributions were calculated for ground-truth DVF and displacement vectors produced by the biomechanical model during elasticity estimation and evaluated for convergence.
Results: The mean convergence between ground-truth DVF values and displacement vectors output from the biomechanical model within 5% and 10% of the maximum deformation value were measured at 67.31% and 87.62%, respectively, with mean displacement error of 2.21 mm. Jacobian convergence within 0.3 (~0.3 mm³) of the ground-truth registration value was observed to be 80.15%.
Conclusion: A high level of convergence demonstrates the feasibility of performing lung elastography using RV and TLC CT images. The ability to acquire elasticity estimation distributions for large deformation images allows for characterization of regional biomechanical phenomena observed in patients with lung diseases such as COPD.
Handouts
Keywords
Modeling, Quantitative Imaging, Simulation
Taxonomy
IM/TH- Image Analysis (Single Modality or Multi-Modality): Quantitative imaging
Contact Email