Author: Ozgur Ates, Chin-Cheng Chen, Chia-Ho Hua, Matthew J. Krasin, Thomas E. Merchant 👨🔬
Affiliation: St. Jude Children's Research Hospital 🌍
Purpose: To validate the use of synthetic CTs generated from CBCT images for online monitoring, ensuring accurate and reliable daily plan quality assessments in adaptive proton therapy (APT).
Methods: A whole-body anthropomorphic phantom was imaged with a diagnostic CT scanner, and CBCT images were acquired for the H&N, thorax, abdomen, and pelvis using clinical scan protocols. To simulate specific anatomical variations, an independent QA software was employed to generate ground truth deformed vector fields (DVFs) by deforming CBCT images with known amounts: a 10° pitch to mimic neck flexion in the H&N, 3 cm peak-to-peak diaphragmatic motion in the thorax, 2 cm anterior weight gain and 10° spinal curvature in the abdomen, and 2 cm lateral tissue changes in the pelvis region. The synthetic CT images were generated using a clinical image registration software, which predicted the ground truth DVFs in 25 structures. An in-house water equivalent path length (WEPL) tool was also tested on a spherical phantom with 36 coplanar beams at 10° intervals to predict known discrepancies per proton spot along the beam path.
Results: DVF analysis demonstrated that 95% of the voxels (d95%) across 25 structures were individually <2 mm between the ground truth and predicted DVFs, satisfying the AAPM TG-132 criteria. The mean DVF error of all voxels within 25 structures was 1.5 mm, with a standard deviation (SD) of ±0.4 mm. A statistical analysis utilizing the Pearson Correlation Coefficient (PCC) demonstrated strong correlations between the voxel intensities of nominal and synthetic CTs with values of 0.97, 0.96, 0.98, and 0.99 for the H&N, thorax, abdomen, and pelvis, respectively. The WEPL tool predicted discrepancies accurately with 0.3 mm mean ±0.4 mm SD.
Conclusion: The combination of synthetic CT and WEPL tools in the online monitoring workflow enables reliable daily plan quality assessments for efficient APT procedures.