Program Information
Harnessing Prior Clinical Knowledge to Automate Radiation Therapy Treatment Planning
Y Yang*, H Liu , P Dong , L Xing , Stanford Univ School of Medicine, Stanford, CA
Presentations
WE-RAM2-GePD-JT-1 (Wednesday, August 2, 2017) 10:00 AM - 10:30 AM Room: Joint Imaging-Therapy ePoster Theater
Purpose: To develop a clinically practical autonomous planning strategy with effective use of prior knowledge for IMRT/VMAT treatment planning.
Methods: For a given patient, an assemble of prior treatment reference plans with similar anatomy is chosen automatically with pre-defined geometric criteria. A plan is first generated using a commercial treatment planning system (TPS) with its planning parameters initialized based on the reference plans. Script/API/Microsoft Visual Studio Coded UI programming is implemented to interact with the TPS. A plan evaluator with consideration of clinical outcome data and organ dose-volume status is developed to assess the TPS-generated plan. A new set of TPS optimization parameters, such as the weighting factors and dose/volume constraints, are generated based on the evaluation results and are fed back to the TPS to drive the search toward an improved solution as measured by the plan evaluator. The procedure proceeds in an iterative fashion until an acceptable tradeoff among target dose coverage/homogeneity and sparing of critical structures. A prostate case and a head-and-neck case is used to demonstrate the proposed approach.
Results: It is found that the iterative interaction between our evaluation software tool and the commercial TPS can successfully drive the optimization toward a solution consistent with our prior knowledge. Our study also indicates that the proposed strategy can dynamically balance the competing objectives and automatically balance the tumor coverage and sparing of the OARs with adequate priority, primarily because of the use of prior knowledge in our modeling. The treatment plans generated using the proposed approach compare favorably with the manually generated plans for both prostate and head-and-neck cases.
Conclusion: Clinical inverse planning process can be automated effectively with the guidance of prior knowledge and automated optimization parameter update. The approach has the potential to significantly improve the radiation therapy workflow and patient care.
Funding Support, Disclosures, and Conflict of Interest: This work is partially supported by NIH (5R01CA176553) and Varian Medical Systems
Contact Email: