Purpose: Knowledge-based planning (KBP) tools, which incorporates prior treatment planning experience, has the potential to improve the quality and consistency of treatment plans. In this work, we evaluated the performance of a proton-specific KBP model in the generation of robustly optimized IMPT (RO-IMPT) plans for treatment of head and neck (HN) patients.
Methods: Sixty head and neck cancer patients previously treated with volumetric modulated arc therapy (VMAT) were selected and replanned with RO-IMPT. Plans were robustly optimized using ±3 mm setup uncertainty (in cardinal directions) along with ±3% proton range uncertainty. Fifty IMPT plans were selected to build the HN proton KBP model library. The remaining ten patients were used for model validation. The model quality was assessed using model generated plots such as DVH plots, regression and residual plots. The comparison of the dosimetric indexes for target and OARs between KBP generated and expert plans were performed by two-sided paired t-test, p<0.05 indicating significance.
Results: Compared with expert plans, the nominal KBP plans delivered significantly lower doses to almost all OARs (-4.9Gy ± 2.74Gy for larynx Dmean, -6.16Gy ± 3.87Gy for constrictor Dmean, -6.84Gy ± 4.65Gy for spinal cord Dmax, -7.06Gy ± 4.66Gy for left cochlea Dmean) except the mandible and the oral cavity which showed no statistically different significance. When including all uncertainty scenarios, the target coverage was better in the KBP plans (4.59% ± 7.58% for CTV70 V100, 4.07% ± 5.84% for CTV56 V100).
Conclusion: This work demonstrated that the proton-specific KBP model was able generate plans with similar target coverage and much better OAR sparing as compared to the ones generated by the expert for HN patients. The HN KBP plans were also more robust and homogenous. Use of proton KBP model for generation of IMPT plans has a potential to improve treatment planning efficiency.
Funding Support, Disclosures, and Conflict of Interest: This work was supported by a research grant by Varian Medical Systems