Author: Shahed Badiyan, Chien-Yi Liao, Mu-Han Lin, David D.M. Parsons, Justin D. Visak, Brien Timothy Washington, Kenneth Westover, Yuanyuan Zhang 👨🔬
Affiliation: Department of Radiation Oncology, UT Southwestern Medical Center, UT Southwestern Medical Center, Medical Artificial Intelligence and Automation (MAIA) Lab & Department of Radiation Oncology, UT Southwestern Medical Center, Department of Radiation Oncology, UT Southwestern Medical Center, Dallas, TX 🌍
Purpose: Adaptive radiotherapy (ART) programs are resource-intensive due to their technical complexities, requiring highly skilled planners. Leveraging integrated automated treatment planning system (TPS) technology is critical to reducing planner workload. We propose a method to generate high-quality x-ray-based adaptive lung stereotactic ablative radiotherapy (SAbR) plans using a novel high-fidelity mode and an integrated automatic intensity-modulated radiation therapy (IMRT) beam geometry optimizer (BGO).
Methods: Fifteen patients with early-stage NSCLC or lung metastases previously treated with 40–60Gy in five fractions using Ethos1.1 TPS were selected. Plans were reoptimized in Ethos2.0 using a simplified planning strategy with high-fidelity mode and automatic IMRT beam selection. For five patients with suboptimal default automatic 7-field IMRT beam geometry (default-auto7B), plans were further enhanced by leveraging the BGO. Field optimization was achieved by generating an interim treatment intent with 10 fields (enhanced-auto10B) through temporarily increasing the fractional dose. Default-auto7B and enhanced-auto10B plans were compared to clinical counterparts using RTOG-based metrics, maximum organ-at-risk (OAR) doses, and total monitor units (MUs).
Results: Enhanced-auto10B plans demonstrated improved intermediate dose fall-off compared to default-auto7B plans and more closely resembled human-selected field geometry. Enhanced-auto10B achieved 4.8%±2.5% improvement in D2cm (p = 0.02) and a reduction in the gradient index by 0.97±0.1 (p = n.s.) compared to default-auto7B. The conformity index improved from 1.1±0.1 (p = n.s.) to 1.06±1.1 (p = n.s.) for enhanced-auto10B versus default-auto7B, with clinical plans achieving 1.01±0.4. No statistically significant differences were observed in maximum OAR doses across all plans. Total MUs were statistically similar between default-auto7B (3221±629) and enhanced-auto10B (3410±686) but were significantly lower than clinical plans (6285± 812). The proposed routine added approximately 15 minutes to the offline workflow.
Conclusion: Enhancing the default BGO algorithm in Ethos2.0 improves lung SAbR reference plan quality, addresses limitations of default settings, and maintains workflow efficiency for patients undergoing lung SAbR.