Back-up plan.

Think residency expansion and relaxed supervision requirements are making life hard for the next generation of radiation oncologists? AI-based treatment planning algorithms are coming for dosimetrists. If you’re familiar with Varian’s new Ethos treatment platform, it utilizes knowledge-based planning (KBP) to generate an optimal treatment plan based on similar prior plans. In this study, rad oncs were presented with over 150 pairs of human-generated plans and knowledge-based plans. The former had been generated during clinical practice, and the latter were generated for the purposes of the study. In over half the cases, KBP plans were preferred over human plans and overall were deemed non-inferior. Surprisingly, lung plans and head and neck plans had the highest success rates (>60%) for KBP while prostate plans had the lowest (<50%). However, if you’ve ever used pareto planning algorithms, you’ll recognize how the magnitude of possible solutions for a head and neck plan greatly exceed the practical abilities of a single treatment planner. TBL: Knowledge-based planning may be just as good as human-generated plans, and may be even better when optimizing is more complex. | Cornell, Int J Radiat Oncol Biol Phys 2019