Room: Exhibit Hall
Purpose: To determine if a knowledge-based planning model (KBPM) built from VMAT plans on Varian Linacs can generate step-and-shoot IMRT plans for Siemens Linacs.
Methods: We created two KBPMs using clinical prostate cancer cases with the Varian RapidPlan module in Eclipse Treatment Planing System(v.13.6). The first VMAT-KBPM was trained from 30 clinical plans with 2-full-arcs on TrueBeam Linacs with Millenium 120 MLCs. The second IMRT-KBPM was trained from 24 clinical 9-beam step-and-shoot IMRT plans on a Siemens Artiste Linac with Siemens 160 MLC (0.5cm leaf width). Both models were iteratively adjusted until they match or exceed the training plans in quality and then employed to automatically generate step-and-shoot IMRT plans for the Artiste for another 20 prostate cases. All resulting plan pairs from both models were normalized to have 100% prescription dose (46Gy) covering 95% of the PTV (prostate + Lymph nodes). The total MUs and DVH parameters of PTV and OARs including bladder, rectum, femurs and penile bulb were compared using paired sample t-test.
Results: The IMRT-KBPM based plans outperformed the VMAT-KBPM based on conformity-index (CI, 1.092Â±0.07 vs. 1.103Â±0.07, p<0.05) and heterogeneity-index (HI, 0.14Â±0.02 vs. 0.15Â±0.02, p<0.05) of the PTV, and use fewer MUs (1118Â±225 vs 1659Â±997, p<0.05). The VMAT-KBPM plans had a lower Dmean (16.49Â±349 vs. 19.60Â±4.14 in Gy, p<0.05) and D50% (13.13Â±4.02 vs. 17.82Â±5.07 in cGy, p<0.05) for the rectum, but IMRT-KBPM created plans with significantly lower (p<0.05) femur doses evaluated by D50%. Both models performed similarly on the Dmean of the penile bulb, D50% and Dmean of the bladder.
Conclusion: To generate step-and-shoot IMRT plans on Artiste, the IMRT-KBPM model based on Artiste IMRT plans performed better than VMAT-KBPM based on Varian RapidArc plans in most aspects. The application of KBPM may be sensitive to the MLC-type and require models for specific MLC and techniques.