How Does CBCT Image Quality Impact on Deformably Mapped Targets and Accumulated Dose Distributions?

Purpose/Objective(s): We perform quantitative analysis of differences in deformable image registration (DIR) and deformable dose accumulation (DDA) computed on CBCT datasets reconstructed using the standard (Feldkamp-Davis-Kress: FDK_CBCT) and a novel iterative (iterative_CBCT) CBCT reconstruction algorithms. Materials/Methods: Nine head/neck (H&N) and 10 prostate cancer patients were randomly selected for retrospective analysis. Both FDK_CBCT and iterative_CBCT images were reconstructed for 608 fractions of treatment. Planning CT images were deformably registered to CBCT image sets using an intensity-based, B-spline algorithm (Elastix ITK/VTK) with optimized parameters. After dose distributions were computed for each fraction on both deformed CT image sets, they were mapped to the planning CT to evaluate the deformed dose. Results based on FDK_CBCT and iterative_CBCT images were compared before and after the dose mapping operation, respectively. Dose constraints for targets and critical structures were evaluated using DDA based on different CBCT images. Dice similarity coefficient (DSC), mass-center distance, and Hausdorff distance analysis of the prostate gland deformed from the planning CT to CBCT were compared against manually delineated contours. Differences due to CBCT images reconstructed with standard and iterative modes were evaluated. Results: Average daily dose differences and normalized DDA differences were ≤1 cGy. Comparing with results before dose mapping operation, maximum daily point dose differences increased from 0.81 ± 0.54 Gy to 1.19 ± 0.48 Gy for H&N patients, and increased from 0.22 ± 0.06 Gy to 1.33 ± 0.38 Gy for prostate patients after the dose mapping operation. Maximum normalized DDA differences were up to 0.43 Gy (0.29 ± 0.09 Gy) and 0.80 Gy (0.42 ± 0.19 Gy) for H&N and prostate patients, respectively. Differences in target minimum doses were up to 0.82 Gy (0.47 ± 0.32 Gy) and 8.31 Gy (-0.62 ± 4.60 Gy) and differences in critical structure doses were 0.10 ± 0.55 Gy and 0.70 ± 1.49 Gy, for H&N and prostate patients, respectively. It is noteworthy that the largest differences occurred in regions with the highest dose gradients, as expected. Comparison between mapped prostate gland and manually defined contours showed that for iterative_CBCT (relative to standard FDK_CBCT), DSC increased by 0.18 ± 0.11 (from mean 0.57 to 0.75), mass-center distances decreased by 4.4 ± 3.3 mm, and Hausdorff distances decreased by 6.1 ± 5.1 mm. Conclusion: Better image quality afforded with iterative_CBCT enhances accuracy of DIR between planning CT and CBCT. This will improve accuracy of mapped volumes of interest, deformed dose distribution, and DDA. iCBCT has bigger effect on image quality in regions like the prostate because the increased patient scatter in such areas has a larger impact on image quality degradation versus sites like the head/neck