(Tuesday, 7/16/2019) 7:30 AM - 9:30 AM
Room: 221AB
Purpose: Hounsfield Unit ventilation (HU-V) methods approximate voxel volume changes from the measured variations in HU between corresponding tissue locations, as identified by deformable image registration (DIR), within an inhale/exhale CT pair. HU-V assumes volume changes are caused solely by air content changes, or equivalently, that apparent mass is conserved. In addition to lung segmentation and DIR, HU-V methods require 1) preprocessing high-intensity vessel segmentation to prevent DIR errors from corrupting measured HU variations, and 2) a heuristically applied post-processing smoothing to counteract CT noise. We introduce the novel Mass Conserving Volume Change (MCVC) method for robust ventilation.
Methods: MCVC estimates a series of subregional volume changes using the integrated mass conservation model and HU-measured density means. Numerical uncertainty is quantified as the standard error of the sample density mean, which allows for acquiring estimates with controllable levels of uncertainty. A ventilation image is recovered from the subregional information, without requiring vessel segmentation or post-processing, by solving a constrained least squares problem. MCVC and HU-V reproducibility was assessed with respect to DIR accuracy using ten 4DCT data-sets from www.dir-lab.com. Results were also compared to the robust transformation-based Integrated Jacobian Formulation (IJF) ventilation method.
Results: Test cases were registered twice with two DIRs. Across all cases: Mean landmark errors for the two DIRs were 1.14 (0.25) and 0.90 (0.16) millimeters. Average Pearson correlation between HU-V images computed from the two DIRs was 0.94 (0.03), while for MCVC it was 1.00 (0.00). Average Pearson correlation between HU-V and IJF was 0.56 (0.11), while it was 0.81 (0.14) between MCVC and IJF
Conclusion: Our results indicate that MCVC is a more robust method for calculating ventilation than traditional HU-V. MCVC is more consistent with the robust IJF method, which suggests that incorporating robustness leads to more consistent results across DIRs and ventilation algorithms.
Keywords
Ventilation/perfusion, CT, Image Processing
Taxonomy
IM- CT: Quantitative imaging/analysis
Contact Email
