Room: Davidson Ballroom A
Purpose: To evaluate the effect of knowledge-based planning (KBP) on plan complexity and delivery accuracy of VMAT treatments for prostate cancer.
Methods: A database of 124 VMAT plans for prostate cancer was used to develop an in-house KBP model such that in-field overlap volume histograms could be used to predict bladder and rectum dose-volumes. The model was used to retrospectively guide a re-plan of 31 prostate patients who had previously been treated with a manually optimized plan. Reference plans were constructed to approximate each patient's clinical plan in order to deliver all plans on the same treatment machine. For each patient, both the KBP plan and reference plan were delivered to a commercial two-dimensional diode array. The agreement between the measured dose and the expected dose (as calculated by the planning system) was assessed using the gamma metric. Average gamma passing rates for the two sets of plans were compared using a two-sided paired t-test. Differences in planned monitor units (MUs) were also examined.
Results: KBP re-plans showed a significant increase in monitor units (p < 0.001), requiring 194.7 total MUs more than reference plans, on average. In terms of delivery accuracy, KBP plans showed significantly lower and more variable gamma pass rates than reference plans at 3%/3mm (98.3Â±1.6 vs. 99.4Â±0.6; p < 0.001) and 2%/2mm (91.5Â±4.6 vs. 93.9Â±2.2; p = 0.017). Also, the increase in KBP MUs was correlated with a decrease in KBP delivery accuracy at 3%/3mm (R = -0.59; p < 0.001) and 2%/2mm (R = -0.56; p < 0.001) compared to reference plans.
Conclusion: While KBP techniques have been shown to improve plan quality, this work suggests they can also increase plan complexity and reduce delivery accuracy. Future work is required to determine whether this results from beam model errors, machine limitations, or a combination thereof.