ESTRO 2024 - Abstract Book

S4588

Physics - Optimisation, algorithms and applications for ion beam treatment planning

ESTR0 2024

probabilistic outcomes and, subsequently, assess its influence on the dose delivered to the healthy brain and hippocampi.

Material/Methods:

The clinical treatment plans of 20 neuro-oncological cases, prescribed (D pres ) to 50.4 GyRBE (15/20), 54 GyRBE (2/20), or 59.4 GyRBE (3/20) to be delivered in 1.8 GyRBE fractions were used for analysis. All treatment plans fulfilled the robustness evaluation criteria of 3 mm setup/3% range robustness settings. The clinical treatment plans were scaled to meet the prescription: VWmin-D 98%,CTV =0.94D pres , based on the PTV-based photon plan calibration in [1]. Polynomial chaos expansion (PCE) was applied to simulate 100,000 complete fractionated treatments, allowing for the derivation of statistically accurate probability distributions for clinically relevant DVH parameters. Stopping-power prediction errors of 1.2%±1.0% (1SD) and systematic and random setup errors of 0.92 mm (1SD) and 1 mm (1SD) respectively were used, consistent with the 3 mm margin from Van Herk’s recipe. To investigate the interpatient variation of the VWmin approach, we analyzed the population median and full ranges of (i) the absolute differences and the ratios between the VWmin-D 98%,CTV and the 10 th percentile of the PCE-derived D 98%,CTV distribution; (ii) the dose corresponding to the 10 th percentile of the PCE-derived D 98%,CTV distribution and (iii) the CTV volume in which 95% of D pres is achieved at a 90% probability. The percentile and probability choices are based on Van Herk's criteria. Hereafter, (iv) the probability of achieving 95% of D pres in the median volume derived under (iii) was determined per patient. Furthermore, (v) we analyzed what would happen if scaling based on the PCE-derived D 98%,CTV instead of the VWmin D 98%,CTV was used. The ratios in (i) were used to determine patient-specific scaling factors that ensure that the 10 th percentile of the PCE-D 98%,CTV =0.95D pres and, subsequently, the impact on the mean dose to healthy brain and hippocampi was investigated.

Results:

The 10 th percentile of the PCE- D

98%,CTV distribution exceeded the VWmin- D 98%,CTV for all patients with a median

absolute difference (i) of 1.36 (range: 0.85-1.91) GyRBE and a median ratio of 0.974 (range: 0.966-0.984).

The 10 th percentiles of the PCE-D 98%,CTV distributions (ii) ranged from 0.956D pres to 0.973D pres (median: 0.965D pres ) and the CTV volume receiving 0.95D pres (iii) ranged from 98.9% to 99.9% (median: 99.7%) (Figure 1). The probability that the population median volume of 99.7% received 0.95D pres (iv) ranged from 0 to 99.6% for the individual patients (Figure 2).

Rescaling the plans (v) based on the ratios obtained in (i) resulted in a median reduction of the D mean, brain-CTV from 6.4 GyRBE to 6.2 GyRBE and of the D mean , hippocampi from 13.2 GyRBE to 12.8 GyRBE.