A feasibility study of a risk-based stochastic optimization approach for radiation treatment planning under setup uncertainty. (September 2019)

• Record Type: Journal Article
• Authors: Khabazian, Azin; Zaghian, Maryam; Lim, Gino J.
• Is Part Of: Computers & industrial engineering. Volume 135(2019)
• Page Start: 67
• Page End: 78
• Publication Date: 2019-09
• Subjects: Chance-constrained programming -- Radiation treatment planning -- Risk-based approach -- Setup uncertainty -- Stochastic optimization; Engineering -- Data processing -- Periodicals; Industrial engineering -- Periodicals; 620.00285
• DOI: 10.1016/j.cie.2019.05.031
• Languages: English
• ISSNs: 0360-8352

Abstract

Highlights: A risk-based CCP model for radiation therapy planning. It is flexible to accommodate the planner's risk profile. It can meet patient specific treatment goals. It is less conservative than a robust optimization approach. Planner defined radiation treatment planning approach under uncertainty. Abstract: This article introduces a planner-driven flexible stochastic decision making model to develop radiation treatment plans for cancer patients under patient-setup uncertainty. The clinical goal is to deliver the prescribed amount of radiation dose to the target tissue(s) while sparing the organs nearby. However, it is difficult to achieve the goal because organs are often closely located in the body. Therefore, some tissues may receive a higher radiation dose than desired. To minimize such violations and allow to make a trade-off between tumor coverage and healthy tissue sparing, we present a chance constrained programming (CCP) optimization method. A planner can use the CCP approach to specify how much clinical violation can be allowed for a specific patient. Assuming that the uncertain dose distribution follows a known (or estimated) probability distribution function, the CCP model was tested using five clinical cases. The resulting treatment plans were compared with the plans generated by the conventional robust worst-case optimization method using dose-volume histograms. Our results support the CCP approach over the robust optimization method in terms of healthy tissues sparing and the clinical target dose requirements. Overall, the risk-based CCP model is not only flexible to accommodate the planner's risk profile and to meet patient specific treatment goals, but has potential to compromise for overly-conservative treatment plans generated by robust optimization methods.