An analysis of knowledge-based planning for stereotactic body radiation therapy of the spine

Purpose Planning for spine stereotactic body radiation therapy (SBRT) is time consuming, and differences in planner experience and technique result in discrepancies in plan quality between facilities. Here, knowledge-based planning is analyzed to determine if it may be effective in improving the quality and efficiency of spine SBRT planning. Materials and methods Thirty-eight spine SBRT cases were collected from the University of Michigan database and inverse planned to deliver 3 10-Gy fractions to the planning target volume (PTV). These plans were used to train a knowledge-based model (model A) using RapidPlan (Varian Medical Systems). The model was evaluated for outliers and validated in 10 independent cases. Each of these cases was manually planned to compare the quality of the model-generated plans with the manual plans. To further test the robustness of the software, 2 additional models (models B and C) were created with intentional outliers resulting from inconsistent contouring. Results Using models A, B, and C, all 10 generated plans met all dose objectives for modeled organs at risk (OARs) (spinal cord, cord planning risk volume, and esophagus) without user intervention. The target coverage and OAR dose sparing was improved or equivalent to manual planning by an expert dosimetrist; however, manually created plans typically required 1 to 1.5 hours to produce and model-generated plans required only 10 to 15 minutes with minimal human intervention to meet all dose objectives. Conclusions The clinical quality of plans produced by RapidPlan were found to improve on or be similar to the manually created plans in terms of normal tissue objectives and PTV dose coverage and could be produced in a fraction of the time. RapidPlan is a robust technique that can improve planning efficiency in spine SBRT while maintaining or potentially improving plan quality and standardization across planners and centers.