Author: Zhenzhen Dai, Anthony J. Doemer, Ryan Hall, Kenneth Levin, Bing Luo, Benjamin Movsas, Karen C. Snyder, Kundan S Thind, Eleanor Walker 👨🔬
Affiliation: Henry Ford Health, HFHS 🌍
Purpose: To investigate the feasibility of a predictive tool for efficient allocation of hypofractionated whole-breast irradiation patients between Varian Truebeam and Ethos systems.
Methods: A fully automated, knowledge-based automated planning tool integrated via the Eclipse (Varian, Palo Alto) scripting interface was developed in C# to simultaneously generate intensity‐modulated tangential plans for Truebeam and Ethos. The tool calculates posterior breast separation to inform Truebeam energy selection. For larger separations, Truebeam plans use two 6 MV intensity‐modulated tangent fields and two 15/18 MV static tangents, with the ratio determined empirically Ethos plans are limited to 6 MV-FFF. Fluences are then generated via a machine‐learning DVH predictive model trained on an independent cohort, followed by post‐processing for skin flash and fluence smoothing. A cohort of 28 breast cancer patients treated with 26 Gy in 5 fractions (Fast Forward trial) were replanned with the tool, producing automated Truebeam (autoTB) and Ethos (autoEthos) plans. Plans were renormalized to identical coverage (D95% = 95% Rx), and clinical goal compliance was assessed per Fast Forward constraints.
Results: autoTB plans met clinical goals in 86% of patients (24/28), while autoEthos achieved 25% compliance (7/28). Analysis revealed posterior breast separation as a key factor influencing success. autoEthos plans failing dose homogeneity criteria (n=17) had an average separation of 24.7 cm, exceeding the 23.4 cm observed in successful plans. Notably, 76% (17/21) of failing autoEthos plans met all criteria with autoTB. These findings suggest that incorporating separation as a criterion for treatment system selection can enhance automated planning effectiveness.
Conclusion: This study demonstrates the feasibility of a novel auto-planning tool for predictive patient allocation in hypofractionated whole-breast radiotherapy, using posterior breast separation as a key criterion. Further validation is warranted in larger cohorts to refine these criteria and identify additional easily measurable patient factors.