O-Ring Platform Automated Planning for Lattice SBRT across Diverse Anatomical Sites: Strengths and Limitations
Abstract
Purpose
The Intelligent Optimization Engine (IOE) on Varian Ethos has demonstrated efficient Lattice SBRT planning for lung tumors. However, generalizability of IOE performance across diverse anatomical sites and tumor sizes remains unexplored. This study evaluates IOE-generated Lattice plans across multiple anatomical regions to establish the scope and limitations of automated planning compared to clinically delivered C-arm linac plans.
Methods
Eleven previously treated tumors spanning thoracic, abdominal, pelvic, and extremity sites (volume: 2039.6 ± 2043.6 cm3) were retrospectively replanned using standardized RT Intent with IOE on Ethos platform. Original plans were created in Eclipse on C-arm linacs with Millennium-120 and HD-120 MLCs. Dosimetric evaluation followed LITE SABR M1 protocol criteria: PTV_Peak mean dose, PTV_Avoid mean dose, dose ratio (DR), conformity index (CI), PTV V95%, DR_1.5cm, and monitor units (MUs).
Results
For targets ≤24 cm in superior-inferior dimension, IOE produced plans meeting LITE SABR M1 metrics across all anatomical sites. PTV_Peak mean dose exceeded 95% with improved dose conformality (CI: 1.0) compared to Eclipse (p = 0.04). Targets exceeding 28 cm required multi-isocenter planning, which failed to achieve characteristic lattice dose distributions. Plan quality degraded for large volumes (PTV>3500 cm3), with PTV_Peak mean dose and PTV V95% falling below 90% and DR below 3. IOE demonstrated reduced MUs in 63% of cases (two-tailed t-test, p = 0.01). Four full arcs with widely spaced collimator angles consistently achieved comparable plan quality, while C-arm linacs with HD-120 MLCs demonstrated an advantage in cases involving non-coplanar beam geometries.
Conclusion
Ethos IOE generates clinically acceptable Lattice SBRT plans across diverse anatomical sites for targets under 24 cm, reducing MUs and planning complexity. It provides a robust, site-independent automated planning solution for appropriately sized targets, with some performance limitations in large volumes.