ESTRO 2021 Abstract Book

S23

ESTRO 2021

equally important as ensuring target dose in the CTV. At our center, clinical robustness evaluations are based on the consensus within the Dutch Proton Therapy (DUPROTON) group. CTV and serial OARs doses are prescribed to and evaluated on voxel-wise minimum/maximum (VWmin/VWmax) doses, based on 28 evaluation scenarios. We assessed the DUPROTON robustness evaluation protocol in terms of CTV and serial OARs dose in a clinical cohort of neuro-oncological patients. Polynomial chaos expansion (PCE) is applied to benchmark clinically relevant DVH parameters and evaluate population-based metrics. Materials and Methods A clinical cohort of 26 neuro-oncological patients treated at our center was used, including cases in which robustness was clinically sacrificed. Prescribed doses (D Pres ) were 45 Gy(RBE) (1 case), 50.4 Gy(RBE) (15 cases), 54 Gy(RBE) (2 cases) and 59.4 Gy(RBE) (7 cases) in 1.8 Gy(RBE) fractions; and 60 Gy(RBE) (1 case) in 2 Gy(RBE) fractions. PCE was applied to provide an accurate and fast, patient- and plan-specific model of the voxel-dose dependence on treatment uncertainties, derived from known geometrical and stopping-power prediction error distributions. After its validation, PCE was used to evaluate 100,000 complete fractionated treatments per patient to determine population-based statistics of relevant dosimetric parameters, e.g., D 98%, CTV and D 0.03cc for serial OARs such as brainstem surface, brainstem-core, and the optical system. We analyzed correlations with clinically VWmin/max DVH parameters. CTV results were grouped according to patients for whom adequate coverage could be achieved and for whom the CTV had to be underdosed due to OAR constraints. Results Comparisons between accurate patient-specific PCE-based and voxel-wise clinical evaluation metrics for the D 0.03cc, OARs and the D 98%, CTV /D Pres are shown in Figure 1 and 2. In Figure 1, all the OAR points are located below the identity line (dashed, black), on average there is 6.4% room relative to widely used (EPTN) OAR constraints in 98% (R 2 = 0.98) of the sampled treatments (dashed, red). From Figure 2, a 2.7% higher CTV dose than prescribed is realized on average in 98% (R 2 = 0.90) of the sampled treatments with the clinical robustness protocol, indicating a possible reduction of setup robustness (SR). Moreover, treatments in which coverage was sacrificed showed a larger variation between the PCE-based and voxel-wise evaluation due to the harder trade-off between the CTV and OAR sparing.