Targeted multi-criteria optimisation in IMRT planning supplemented by knowledge based model creation

Intensity-modulated radiation therapy (IMRT) planning is an inherently multi-criteria task. A multi-criteria workflow (MCW) typically passes the following steps: create an optimisation model with multiple criteria, approximate the Pareto frontier, and visualise the generated plans to the decision-maker (DM) for inspection. This interactive plan selection and manipulation allow to create better treatment plans as judged by physicians. However, once an optimisation model is specified, optimisation objectives cannot be modified any more. Thus this fixed model implies that a planner has to guess an appropriate model to begin with. Only after Pareto frontier approximation is calculated, the planner can assess the goodness of the model by exploring the trade-offs. The shortcoming of a MCW becomes apparent when the proposed model fails to generate expected trade-offs and the planner is thus forced to refine the model and repeat the calculations. To circumvent this drawback in the MCW, we propose a local multi-criteria workflow (L-MCW) designed and implemented in a collaboration between Fraunhofer ITWM and Varian Medical Systems. L-MCW enables local exploration around an initial, promising plan. The initial plan is automatically inferred by a knowledge-based algorithm (RapidPlan™). The decision-maker can thus evaluate trade-offs in the most interesting region surrounding the initial plan. Clinical results of the combination of knowledge-based planning and L-MCW with a cohort of Prostate and stereotactic ablative radiotherapy (SABR) Lung cases demonstrate substantially reduced planning time and improved organ-at-risk sparing compared to manual planning. The L-MCW provides an intuitive and flexible mechanism to adapt knowledge-based-planning models to similar, but not identical clinical situations and allows the practitioner to quickly determine and realise the most beneficial trade-offs in a treatment plan.