Implementation and comparison of KBPRapidPlan models in breast cancer radiotherapy

Breast cancer is the most common type of cancer among women globally and in Brazil, where it is also the leading cause of cancer-related mortality. Radiotherapy plays a crucial role in treatment, utilizing linear accelerators to deliver X-ray beams. With technological advancements, Volumetric Modulated Arc Therapy (VMAT) has emerged, providing more precise dose conformity to the target and greater protection to adjacent organs. The effectiveness of this technique is further enhanced by using Knowledge-Based Planning (KBP) algorithms, such as RapidPlan, which employ Dose-Volume Histogram (DVH) estimation models to reduce planning time and variability, ensuring safer and more effective treatment plans. This study aims to develop and validate RapidPlan models to optimize the treatment of left breast cancer. It evaluates the efficacy of RapidPlan models with different database sizes and seeks to implement the most efficient model in clinical practice. Various models were developed and validated with databases containing 30, 50, 100, 150, 200, 300, and 400 breast treatment plans (designated as RP30, RP50, RP100, RP150, RP200, RP300, and RP400), as well as a specific model with 150 plans performed on the Halcyon linear accelerator (HCY150). Thirty patients not included in the models were analyzed, totaling 240 plans optimized with RapidPlan. The comparison between the plans generated by the models, without human interference, was conducted in pairs, all normalized to V95%-D95%. The model with the largest database (RP400) showed a significant reduction in low doses to the left lung in the V150cGy, V350cGy, and V780cGy metrics, without compromising the plan quality compared to previously approved plans and those generated by models with fewer data. Additionally, no significant differences were observed between the specific model for the Halcyon accelerator (HCY150) and the mixed model (RP150), indicating that a more comprehensive database is more beneficial than specialization for a single type of equipment. This study recommends creating models with at least 150 left breast plans to ensure optimal performance. Models with fewer than 150 plans are still viable but may require additional adjustments by experienced planners. The most efficient model, RP400, demonstrated excellent optimization capabilities, potentially matching or surpassing plans created by experienced planners. The RP400 model was approved for clinical use in the radiotherapy units of the Oncoclínicas Group and provides a solid foundation for the implementation of new RapidPlan models in various anatomical sites, contributing to the improvement of VMAT treatment quality and efficiency. (author)