(Sunday, 7/29/2018) 3:00 PM - 6:00 PM
Room: Exhibit Hall
Purpose: To estimate the reconstruction error of a novel on-treatment CT reconstruction method, where the CT of a treatment moment is reconstructed by linearly scaling the simulation-time respiratory model based on the treatment-time breathing amplitude.
Methods: Six 4DCT datasets, including deformation vector fields (DVFs) and digitally reconstructed radiographs (DRRs), were generated using XCAT (4D Extended Cardiac-Torso XCAT Phantom) with 8 mm to 48 mm diaphragm motions at the interval of 8 mm. Based on these ground truth, thirty six combinations (6 x 6) of different simulation-to-treatment motion variations were created. For each combination, the treatment CT was reconstructed by modeling the simulation-time motion using a B-spine deformable image registration algorithm, and linearly scaling the model based on the tumor excursion detected on the DRR images. The reconstruction errors were measured in ten different spherical regions (φ=2cm) throughout the lung area.
Results: The overall estimation error (μ±σ) was 1.7 ± 1.8 mm. The largest error occurred on the diaphragms (3.1±1.6 mm) and the smallest at the apices (0.5±0.2 mm). The error was proportional to the motion magnitude at a rate of 0.6 mm error per 1.0 mm motion magnitude. A large portion of the estimation error was due to the differences in the motion models of XCAT and DIR.
Conclusion: The novel on-treatment CT reconstruction method may be clinically useful since the reconstruction error is minimal as demonstrated in this study. An improved DIR accuracy will further reduce the reconstruction error. Additional validation work is warranted with real patient images before clinical applications.
Keywords
Not Applicable / None Entered.
Taxonomy
Not Applicable / None Entered.
Contact Email
