Room: Exhibit Hall | Forum 7
Purpose: To develop an automation solution for generating clinically acceptable plans using a commercial treatment planning system (TPS).
Methods: Unmanned planning trials on the Elekta Monaco TPS were performed with updated constraints, which were managed by the template that controls all input parameters for optimization. A software tool called Robot Framework was used to interface with Monaco to allow repeated plan trial launch. The resulting plan was evaluated in order to identify the plan quality indices that failed to meet the acceptance criteria in the prescription. The results of evaluation, along with certain optimization parameters in Monaco, such as desired and achieved values of organ-at-risk constraints that conflicted with target coverage, and their adaptively adjusted relative weightings, were used as the input to a template modifier to derive the new constraints for the next trial. The template modifier, developed in Python, mimicked the thought process of an experienced planner. Acceptable plans, as with all prescribed constraints satisfied, were retrievable for further evaluation by clinicians.
Results: The feasibility of our automated planning approach was validated by testing cases. For a prostate VMAT case, automatically plans were very similar to the clinical plan judged by common dose-volume histogram criteria. Acceptable plans could be generate by less than 30 trials for fluence optimization only, and the quality of final plans were weakly dependent of the parameters in the starter template and the choices of the step sizes for changing the dosimetric constraint values.
Conclusion: A planning automation solution facilitated by the template in Monaco has been developed. It will potentially replace the lengthy process of planner-TPS interactions. Plannerâ€™s experience to find suitable but case-specific constraints is no longer needed. The central piece of the solution is a template modifier that is being further developed for improving the robustness of our automation approach.