Interactive visual guidance for automated stereotactic radiosurgery treatment planning


The growing technology industry has led to the steady enhancement of expert systems, often at the cost of increased complexity for the systems' end users. Efforts to improve the prescriptive elements of systems, however, often prove unsuccessful, since the nature of complex and high-dimensional decision problems is difficult to capture precisely by models and algorithms. To rectify this deficiency, complementary softwares may be used to accept decision-making input from users. In this paper, we introduce a graphical interface-based multi-criteria decision support system for designing radiation therapy treatment plans. While many automated strategies for treatment plan generation exist in the literature, they often require a large amount of iteration and a priori decision-making in practice, so much of the planning is done manually. Our interface, morDiRECT (the Medical Operations Research Laboratory’s Display for Ranking and Evaluating Customized Treatments) uses the variability associated with the planning parameters to generate diverse plan sets automatically, creating a comprehensive and visible decision space for users. We demonstrate morDiRECT’s generation process, built-in analytical tooling and graphical display using four clinical case studies. In three cases, we find plans that fully dominate the benchmark forward plans, as well as additional plans that possess potentially desirable tradeoffs for all cases. Our results demonstrate that with relatively little upfront effort, users can pre-generate and choose from a diverse set of clinically acceptable plans, leading to reliable treatments for head-and-neck patients.

Expert Systems with Applications
Dionne M. Aleman, PhD, PEng
Dionne M. Aleman, PhD, PEng
Professor of Industrial Engineering