Author: Ryan Clark, Jon Cornelisse, Kenny Guida, Anthony Magliari, Lesley Rosa, Megan Smith, Shane R Stecklein 👨🔬
Affiliation: Office of Medical Affairs, Varian, A Siemens Healthineers Company, Department of Radiation Oncology, University of Kansas Medical Center 🌍
Purpose: Post-mastectomy radiation therapy (PMRT) can pose a challenge for planners of all levels of experience. For cases where 3D conformal therapy yields undesirable results (i.e., dose heterogeneities, excessive organ at risk (OAR) dose), volumetric modulated arc therapy (VMAT) can improve plan quality. Our goal was to improve an in-house knowledge-based planning (KBP) model using dosimetric scorecard (DSC) analytics to aid in VMAT optimization for PMRT cases.
Methods: 70 unilateral PMRT cases were planned to 4256 cGy in 16 fractions with a virtual bolus VMAT technique. An original PMRT KBP model, KBP_RP, relied solely on RapidPlan statistical analysis tools, including Cook’s distance and modified z-score, to identify outlier plans. A second model, KBP_DSC, employed additional DSC analysis during the model tuning process. DSCs totaled 197 possible points based on departmental standards. A validation set of 15 PMRT cases were used to test both models.
Results: DSC analysis identified where PMRT plan scores could be improved, enabling planners to better visualize how target volumes and OARs needed to be prioritized during iterative planning to improve KBP_DSC; plans that did not exceed a threshold score of 78% or failed to meet acceptable coverage goals or OAR constraints were removed from the model. Outlier analysis led to replanning and the eventual removal of 12 and 14 plans in KBP_RP and KBP_DSC models, respectively. Plans in the KBP_DSC model yielded an average scorecard of 86.5±5.3%, as compared to 82.1±7.9% for KBP_RP (P<0.05). Scorecard analysis on the validation set showed significant improvement in plan quality for plans that employed KBP_DSC during optimization (91.9±3.4% versus 86.8±4.3%, P=0.00006).
Conclusion: DSC analysis was successfully utilized to improve KBP modeling and, subsequently, PMRT VMAT plan quality. This modeling technique ensures that only high-quality plans comprise a KBP model, thus providing planners with a robust tool to tackle difficult cases.