Dosimetric Study of Adaptive Proton Pencil-Beam Scanning Treatment Plans Using CBCT-Based Synthetic CTs Generated From Commercial Image Information Systems

Abstract

<h3>Purpose/Objective(s)</h3> Evaluate the accuracy of using CBCT-based synthetic CTs to trigger adaptive planning in pencil-beam scanning proton therapy. The CBCT-based synthetic CTs can decrease the number of necessary verification CT scans during treatment and trigger earlier adaptive planning for more effective treatment. <h3>Materials/Methods</h3> A retrospective study of 20 patients across 3 different sites: 10 head and neck, 5 breast, and 5 lung cases were studied. MIM Maestro 7.0.4 and Velocity 4.1 were used to generate synthetic CTs from CBCTs. The verification plans calculated on both the synthetic CT (synCT) and the verification CT (verCT) performed on the same day were compared in a treatment planning system. Dose metrics from each synCT plan generated are subtracted from the corresponding dose metrics from the same day verCT plan. Outliers with greater than 10% deviation on D95% were excluded from the analysis. <h3>Results</h3> The CTV D95% is less than 2% different between the synCT and verCT plans, and 75% of Velocity patients and 55% of MIM patients falls within 95% confidence levels when comparing target D95%. The average difference from verCT D95 was 1.6% for Velocity and -0.88% for MIM. Serial OAR Dmax in synCT plans average deviation from verCT was less than 1Gy on average. Parallel OAR volumetric and mean dose in the synCT plans were consistent with the verCT plans with deviations on average less than 1% by volume and 0.5Gy mean. Maximum doses have significant variations ranged from -7.7Gy to +4.4Gy for the target and -11.76Gy to +9.83Gy for OARs. The D2% of treatment volume in synCT plans showed was consistent with less than 1% deviation from the verCT plans. <h3>Conclusion</h3> Both systems can automatically generate synthetic CTs from CBCTs. The dosimetric similarities of synCT plans compared with verCT plans were higher for treatment sites with less heterogeneities such as breast patients when compared with head and neck patients. Outliers can happen when there are significant anatomy changes which result in inaccurate synthetic CTs due to inadequate image deformation or HU corrections. Users should pay extra attention to tissue discrepancies while evaluating the images. The Dmax for critical structures can vary greatly between synCT and verCT, so a small volumetric analysis is recommended. Both systems can be used to reduce verification CT scans while a patient is on treatment and can improve the adaptive plan workflow. Synthetic CT generation can be a more effective way to discover significant dosimetric differences caused by changes in patient anatomy, and therefore a more efficient path to adaptive treatment.