Author: Maria Jose Almada, Bruno Forti, Patricia Murina, Carlos Daniel Venencia 👨🔬
Affiliation: Instituto Zunino - Fundacion Marie Curie, Dra. 🌍
Purpose: The use of VMAT for the breast requires duplication of the patient's CT, several planning structures, and patient-specific dose-volume constraints depending on the patient's anatomy, fractionation, and the complexity of the PTV. These plans can be automated using scripts and KBP. We aim to show our four years of experience with automatic breast VMAT planning.
Methods: ESAPI and Microsoft Visual Studio created two C# executable scripts in Eclipse v15.6.06 (Varian). These include the creation of planning structures, VMAT plan generation (technique, prescription, and isocenter position), optimization based on RapidPlan (PO_15606), dose calculation, and plan evaluation (automatic dosimetric data extraction from the ARIA database). RapidPlan models were created and used, considering breast laterality, fractionation (5, 16, and 20 fractions), and the complexity of the PTV (breast with or without lymph nodes).
Results: From January 2021 to January 2025, 2,400 plans were completed. The number of automatic plans in 2021 was 492; in 2022, it was 584; in 2023, it was 622; and in 2024, it was 683. VMAT planning in five fractions was introduced in March 2023. Several improvements were made after the first year, including a timer, logs, a progress bar, stage indicators, and a database. The RapidPlan models were improved, doubling the number of plans (a minimum of 60 plans per model). With manual planning in 2020, the percentage of plans that reverted from the acceptance stage to the planning stage was 9.4%. With auto planning, this percentage decreased to 4.1% in 2021 and around 2% in the following years.
Conclusion: Automatic breast VMAT plans are clinically acceptable and reduce replanning, regardless of the user’s skills. In the future, we will improve RapidPlan models, adapt Scripts to Eclipse v18 (upgrade), improve hardware (GPU), and include PSQA.