An interactive plan and model evolution method for knowledge‐based pelvic VMAT planning

PURPOSE: To test if a RapidPlan DVH estimation model and its training plans can be improved interactively through a closed‐loop evolution process. METHODS AND MATERIALS: Eighty‐one manual plans (P(0)) that were used to configure an initial rectal RapidPlan model (M(0)) were reoptimized using M(0) (closed‐loop), yielding 81 P(1) plans. The 75 improved P(1) (P(1+)) and the remaining 6 P(0) were used to configure model M(1). The 81 training plans were reoptimized again using M(1), producing 23 P(2) plans that were superior to both their P(0) and P(1) forms (P(2+)). Hence, the knowledge base of model M(2) composed of 6 P(0), 52 P(1+), and 23 P(2+). Models were tested dosimetrically on 30 VMAT validation cases (P(v)) that were not used for training, yielding P(v)(M(0)), P(v)(M(1)), and P(v)(M(2)) respectively. The 30 P(v) were also optimized by M(2_new) as trained by the library of M(2) and 30 P(v)(M(0)). RESULTS: Based on comparable target dose coverage, the first closed‐loop reoptimization significantly (P < 0.01) reduced the 81 training plans’ mean dose to femoral head, urinary bladder, and small bowel by 2.65 Gy/15.63%, 2.06 Gy/8.11%, and 1.47 Gy/6.31% respectively, which were further reduced significantly (P < 0.01) in the second closed‐loop reoptimization by 0.04 Gy/0.28%, 0.18 Gy/0.77%, 0.22 Gy/1.01% respectively. However, open‐loop VMAT validations displayed more complex and intertwined plan quality changes: mean dose to urinary bladder and small bowel decreased monotonically using M(1) (by 0.34 Gy/1.47%, 0.25 Gy/1.13%) and M(2) (by 0.36 Gy/1.56%, 0.30 Gy/1.36%) than using M(0). However, mean dose to femoral head increased by 0.81 Gy/6.64% (M(1)) and 0.91 Gy/7.46% (M(2)) than using M(0). The overfitting problem was relieved by applying model M(2_new). CONCLUSIONS: The RapidPlan model and its constituent plans can improve each other interactively through a closed‐loop evolution process. Incorporating new patients into the original training library can improve the RapidPlan model and the upcoming plans interactively.