ESTRO 38 Abstract book

S997 ESTRO 38

Nucleare INFN, Trento Institute for Fundamental Physics and Applications TIFPA, Trento, Italy Purpose or Objective Protontherapy treatments with pencil beam scanning (PBS) offer the highest flexibility when proton fields are optimized simultaneously (i.e. with multi-field optimization, MFO). Due to the high in-field dose gradients obtained with MFO, this might come at the price of reduced robustness. Robust optimization techniques have been proposed to handle this issue, but they require extensive computational power. Therefore, they are hardly compatible with the clinical application of Monte Carlo (MC) dose calculation algorithms, which have been demonstrated to significantly improve the accuracy of proton treatment plans. The purpose of this work is to implement a robust MFO technique compatible with the application of MC algorithm and to evaluate its robustness to both physical and biological uncertainties. Material and Methods 9 patients (3 brain, 5 head-and-neck, 1 spine) underwent proton treatment generated by a novel robust MFO technique. A hybrid (hMFO) approach was implemented, planning dose coverage on a PTV compensating for setup errors, whereas range calibration uncertainties are incorporated in PTV robust optimization process as CT calibration uncertainties (i.e.: three scenarios are optimized, corresponding to nominal and ±3.5% uniform mass density scaling). A commercial MC dose calculation engine was adopted. hMFO was compared with single-field optimization (SFO), both by robustness analysis (considering both CT calibration uncertainties and isocenter shifts due to setup errors, for a total of 16 perturbed scenarios) on CTV and organs at risk (OARs) and by assessing in the nominal plans the potential impact of variable relative biological effectiveness (RBE). Results Nominal hMFO plans were superior compared to SFO in terms of target coverage (p=0.004), without difference for OARs sparing (p=0.280). The improvement in target coverage obtained with hMFO is preserved in worst-case scenarios (p=0.004), confirming that hMFO is as robust as SFO to physical uncertainties in terms of target coverage. This is summarized in the boxplots of Figure 1, showing the difference between prescription and actual CTV doses for the patients considered in the study. Similarly, the difference between OARs planning and prescription dose for the different scenarios is shown in Figure 2. On OARs, physical (i.e. worst-case scenario) and biological (i.e. variable RBE) uncertainties resulted into significant (p<0.01) dose increase for both hMFO and SFO (by 3-7 Gy), but without significant difference between these two hMFO allows improving plan quality compared to SFO, without affecting robustness to setup, range and RBE uncertainties. We show that hMFO is compatible with the application of MC-based robust optimization with minimal impact on computation time. Our data also indicate that uncertainties due to variable RBE, even though generally neglected in clinical practice, might be comparable with those resulting from physical uncertainties. Figure 1 techniques. Conclusion

Figure 2

EP-1838 Dosimetrical comparison of brain SRS treatment plans using IRIS and InCise 2 MLC of the CyberKnife M. Vekas 1 , G. Stelczer 1 , L. Janvary 1 , T. Major 1 , C. Polgar 1 1 National Institute of Oncology, Center of Radiotherapy, Budapest, Hungary Purpose or Objective A Cyberknife M6 (CK) system was installed at our institution in 2018 January. The aim was to compare the dosimetrical properties of the treatment plans and treatment times with IRIS and MLC collimator. Material and Methods Ten patients with single brain tumours were selected for this study. For each patient three treatment plans were made. Optimal treatment plans were created with MLC and IRIS. Due to the big differences in treatment time between these two plans, one additional IRIS plan was made with time reduction (IRIS TR), to get as short treatment time as possible for a clinically still acceptable plan. The dose prescription of each IRIS plan was set in order to have the same target coverage as the MLC plan. MLC plans were compared to IRIS and IRIS TR plans pairwise with respect to PTV coverage (V 100% ), treatment time, ICRU 83 homogeneity index (HI), conformity index (nCI), monitor units (MU), D 0,04ccm , D 0,5ccm for brainstem, D 0,2ccm , D 0,04ccm for optic pathway and maximum dose of a shell (one voxel thick ring ROI around the target at 10 mm distance). Wilcoxon matched pair test of Statistica 7 software was used for statistical analysis. Results The average volume of PTVs was 13,4 cm 3 . There were no differences between the target coverages of the three plans. However, average MUs were significantly lower for the MLC plans compared to IRIS (p=0,005) or to IRIS TR (p=0,005) plans, 7908 MU, 23109 MU and 19467 MU respectively. The mean calculated treatment time was significantly less with MLC (18 mins) compared to IRIS (27 mins, p=0,008) or IRIS TR (23 mins, p=0,008). There was a trend for average HI where the IRIS plans had better values (0,15) than MLC (0,18), while IRIS TR (0,19) values were similar. IRIS plans resulted better nCI values (1,27) than