Leveraging gEUD objectives in the development and clinical implementation of a RapidPlan knowledge-based planning model for lung SBRT in Eclipse.

Abstract

To provide a detailed and reproducible description of the development and clinical implementation of a RapidPlan knowledge‑based planning (KBP) model for lung stereotactic body radiotherapy (SBRT) in Eclipse v15.6, including training dataset composition, model analytics, objective template design, and prospective validation. A RapidPlan model was constructed using 161 coplanar VMAT SBRT plans calculated with Acuros XB (2.0 mm grid) and 6 MV FFF beams. Structure matching, outlier identification, and DVH prediction behavior were assessed using Varian Model Analytics report. A unified optimization template was assembled with explicit target priorities, a PTV gEUD objective, and gradient‑focused constraints addressing near‑target fall‑off and chest‑wall interfaces. The model was prospectively validated on 21 patients by comparing KBP plans with previously approved clinical plans. Model Analytics indicated stable DVH prediction bands after refinement; 4 plans were removed for atypical geometric or dosimetric behavior. The finalized template produced clinically acceptable plans for both peripheral and central tumors across the prescription range represented in the training set. In validation, most plans met target and OAR criteria on first optimization; where needed, minor adjustments (typically <10 minutes) resolved remaining issues. This technical note provides a complete, practical workflow for constructing a lung SBRT RapidPlan model in Eclipse v15.6. The reported training selection, analytics‑based refinement, and optimization template parameters enable other centers to reproduce or adapt the model for standardized SBRT planning without outcome comparisons.