Fig. 9.

Schematic view of some classical auto-planning workflows. Left: Multi criteria optimization generates a library of pareto optimal plans from a set of objectives. In the schematic, the user navigates the plans a posteriori by use of sliders for each parameter. Right: Knowledge based planning relies on a set of high-quality clinical plans. Atlas based methods finds the best matching patient in the library when introduced to a new patient, extracts the plan and adjusts it to the new patient. Model based approaches train a predictive model on the library plans. The model is used to predict parameters for new patients, which are used to generate objectives for the optimizer.