In silico Patient-Specific Quality Assurance for Rapidarc Dynamic Breast Treatments
Abstract
Purpose
Measurement-based patient-specific QA is often impractical in adaptive workflows, motivating the use of complexity metrics. For tumor-site-specific class solutions [NCS 35], consistent beam characteristics allow the definition of targeted complexity thresholds to assess plan consistency. Building on prior results, this study performs a longitudinal beam-level analysis of complexity metric variation for a RapidArc Dynamic (RAD) breast class solution.
Methods
52 RAD breast plans were retrospectively analyzed: 20 plans (10 commissioning validated with IC/film and 10 clinical validated with EPID) formed the training set, and 32 clinical plans served as the validation set. Prescriptions consisted of 15x2.67Gy to the whole breast with a 3.2Gy SIB using 6X-FFF energy. Plans included one RAD sub-arc with two static ports and an avoidance sector to limit lung dose, and a second VMAT sub-arc with fixed collimator angle (10°) for boost delivery. Statistical process control (SPC) limits [TG-218] were derived per beam from the training set for 4 metrics: MU per dose per fraction (MU/dose/fx), leaf sequence variability (LSV), aperture area variability (AAV), and equivalent square field size (EqFS). To accommodate non-normal distributions, robust control limits were defined using the median and a robust variability estimate based on the Median Absolute Deviation. Validation-set plans with fewer than 6/8 metrics within expected variation were considered outliers.
Results
Expected variations were narrow and beam-specific. For the RAD beam, MU/dose/fx, LSV, AAV and EqFS ranges were [167.825-222.952], [0.828-0.917], [0.267-0.444], and [32.248-55.296], respectively. Corresponding VMAT beam ranges were [71.583-125.673], [0.572-0.835], [0.053-0.404], and [11.177-28.581]. In the validation cohort, 93.8% of plans met the ≥6/8 metric criterion.
Conclusion
Beam-specific, complexity-based SPC limits captured systematic differences between RAD and VMAT beams in a RAD breast class solution, enabling targeted, beam-level QA thresholds. This approach supports efficient screening of class solution compliance. Further research will explore expected complexity variation in adaptive settings.