Clinical Implementation of Direct-to-Unit Ultra-Hypofractionated Whole-Breast Irradiation Using AI-Enabled Adaptive Radiation Therapy
Abstract
Purpose
Direct-to-unit (DTU) radiotherapy, which bypasses CT simulation and prospective planning, offers an opportunity to markedly shorten the interval from consultation to treatment and expand access to timely breast cancer care. Building on prior feasibility work, this study describes the clinical deployment of an ultra-hypofractionated whole-breast workflow enabled by CBCT-guided adaptive radiation therapy (ART), AI-based segmentation, and direct CBCT dose calculation.
Methods
Ten patients (five left- and five right-sided breast cases) completed the standard pathway of simulation, contouring, and preplan generation. In parallel, a DTU workflow was configured for each case using a generic laterality-matched placeholder patient created within a separate adaptive intent (Ethos; Varian, a Siemens Healthineers company). One fraction per patient was treated with the DTU intent, while the rest were treated using the regular adaptive workflow. During both, the physician reviewed and modified AI-generated organs-at-risk (OARs) and target volumes online. For all fractions, we recorded adaptive workflow duration, Dice similarity coefficients (DSC) comparing online-adapted targets to the preplan volumes, and dose metrics for target volumes and OARs.
Results
Patient breast volumes exhibited substantial variability (mean 1383 ± 464 cc; range 735–2186 cc). The overall treatment times for the conventional and DTU workflows were comparable (24.2 ± 3.7 vs. 24.5 ± 4.0 minutes). Target volume agreement was high in both approaches, with DSC values of 0.92 ± 0.02 for conventional fractions and 0.90 ± 0.03 for DTU fractions. Across all DTU treatments, target coverage and organ-at-risk constraints were satisfied.
Conclusion
This study demonstrates successful clinical adoption of a direct-to-treatment, simulation- and preplanning-omitted workflow for ultra-hypofractionated whole-breast irradiation using AI-enabled ART. The approach integrates seamlessly into routine practice without prolonging treatment times and maintains high plan quality. Broader implementation may meaningfully accelerate the radiotherapy timeline while conserving traditional CT simulation and planning resources for more complex cases.