Encrypted login | home

Program Information

Universally Reprogrammable Automated Planning

no image available
H Wang

H Wang1*, L Xing2 , (1) Electrical Engineering, Stanford University (2) Stanford University School of Medicine, Stanford, CA

Presentations

SU-E-T-339 Sunday 3:00PM - 6:00PM Room: Exhibit Hall

Purpose: Automating treatment planning has recently become a subject of intense research. The current limitation is there is no way to get around the black box nature of planning. One way is know the components inside the black box, but this idealistic approach is rarely practical. The purpose of this work is to develop a generic framework that is capable of automating any existing black box planning environment without the need to know the details of the implementation inside.

Methods: We used a record, playback, and validation mechanism to fully automate the treatment planning procedure. Varian Eclipse Treatment Planning and Coutouring were used for VMAT planning. Different actions were pre-recorded and played back with adjustable parameters calculated and inputted on the fly. The decision on which parameters to tune and how much to make the adjustment is based on previous results produced and saved.

Results: We have tested our interface with four different head and neck VMAT patients. The system successfully ran and the results were in strong agreement with clinical requirements. Each iteration consists of tuning, decision making based on previous runs, optimization, and dose calculation. For each patient 20 iterations were sufficient and each iteration took on average 34.6 minutes. The DVH was exported after the dose calculation and the priority for all the organs were adjusted in the optimizer for each adjustment iteration.

Conclusion: We have developed a generic framework that allows researchers to fully automate the treatment planning process. We have showed a proof-of-concept using Varian Eclipse. The automation does not require the user to know the details of the implementation of the treatment planning environment. It approaches treatment planning in a similar way as a manual planner would, by trying different adjustment strategies on the fly and deciding subsequently what the best action is.



Contact Email: