One-Minute Verification and Fully Autonomous VMAT TBI Planning In Raystation for Pediatric Patients
Abstract
Purpose
VMAT total body irradiation (TBI) planning is complex and time-intensive, involving multiple isocenters, junction management, and iterative hotspot control. Many automation approaches still require repeated user interaction, limiting efficiency gains. This work presents a fully autonomous VMAT TBI planning workflow requiring brief upfront verification followed by unattended optimization and plan completion.
Methods
An end-to-end VMAT TBI planning framework was developed using the RayStation (2024A) scripting API. User interaction is limited to two early verification steps: (1) review of automatically generated multi-isocenter geometry, and (2) confirmation of automatically created beams and parameters, including jaw limits (TrueBeam M120-MLC), robustness settings, junction structures, derived target sub-volumes, and initial optimization objectives. After verification (≤1 minute from launch), the workflow runs unattended, performing sequential optimization, hotspot control, objective refinement, and result saving. The system was evaluated on 12 simulated pediatric cases derived from craniospinal irradiation datasets extending from head to mid-thigh (ages 4–14 years). Mean PTV length was 66 ± 6 cm (volume 13,306 ± 3,493 cc). Mean lung and kidney volumes were 839 ± 355 cc and 131 ± 33 cc, respectively. All plans used eight VMAT arcs with a prescription of 200 cGy × 1 fraction.
Results
Following ≤1 minute of total user verification, all plans completed autonomously. Target coverage was consistent, with PTV D90 = 200.3 ± 0.2 cGy. Hotspots were controlled, with PTV D0.03cc = 231.5 ± 1.8 cGy. Mean lung dose of 112.5 ± 0.5 cGy (56.3 ± 0.2% Rx) and mean kidney dose of 145.9 ± 0.5 cGy (73.0 ± 0.2% Rx). Total MU was 2204.7 ± 126.8 (all isocenters). Hands-on planning effort was reduced from hours to brief upfront verification.
Conclusion
This fully autonomous VMAT TBI planning workflow enables a “verify once, run unattended” paradigm, delivering consistent plan quality while substantially reducing planning time and inter-planner variability.