ESTRO 38 Abstract book

S988 ESTRO 38

Purpose or Objective In proton therapy treatments, physical uncertainties can cause both aiming and range errors. Current planning strategies aim at achieving robust plans by optimizing an objective function over a discrete set of scenarios. As the problem complexity scales exponentially with the number of considered uncertainty sources, a pre-selection of scenarios is necessary to limit the computational burden and keep the computation time within clinically acceptable timescales. The set of scenarios in conventional treatment planning is statistically inconsistent due to the use of unlikely scenarios (large setup error (SE) of 5 mm AND proton range error (PRE) of +/-3%) that lie outside of the planned confidence interval (CI), potentially compromising plan quality. This study aims to establish a novel treatment planning methodology that produces plans of acceptable robustness and guarantees coverage of a certain percentage of beam positions and proton ranges, defined by a priori chosen CI. The focus is to make the methodology statistically sound, fast and computationally cheap by efficiently pre-selecting relevant scenarios which are later fed to the robust optimizer. Material and Methods We consider lung treatments with motion characterized by a 4DCT binned in ten breathing phases; robustness should be achieved for all phases. SE and PRE are distributed as Gaussian distributions with standard deviations of 2 mm and 1.6%, respectively. We perform two successive steps to determine the relevant scenarios: first, 12 scenarios are selected that cover extreme positions reached by the tumor during breathing together with spatial shifts of 5 mm and image density scaling of +/-2.5% (sampling of a 4D isoprobability hypersurface). The second selection is based on the magnitude of the PRE, estimated from maps of water-equivalent path lengths, computed for scenarios randomly sampled according to the distributions defined above. 20 scenarios are selected that produce the largest PRE evaluated for each voxel inside the CTV (see Fig. 1). All treatment plans were calculated for a lung tumor case and plan optimization was performed using the robust optimizer in RayStation 6. The dose prescription was 60 Gy (minimum coverage of 57 Gy). Conventional robustness optimization with SE of 5 mm, PRE of +/-3% and inhale/exhale breathing phases, was used as a reference. Plan robustness was evaluated with the MCsquare code by recalculating the dose distribution on a set of 100 randomly sampled error scenarios.

Conclusion Serious reduction of plan calculation time is achieved using the scenario selection methodology without significant compromise in treatment plan quality. EP-1822 Evaluation of plan robustness against tumor motion for lung SBRT treatment with non-coplanar VMAT C. Garibaldi 1 , A. Bazani 2 , F. Pansini 2 , F. Emiro 2 , S. Trivellato 2 , S. Comi 2 , G. Piperno 3 , A. Ferrari 3 , B.A. Jereczek-Fossa 3,4 , M. Cremonesi 1 , F. Cattani 2 1 IEO-European Institute of Oncology IRCCS, Unit of Radiation Research, Milano, Italy ; 2 IEO-European Institute of Oncology IRCCS, Unit of Medical Physics, Milano, Italy ; 3 IEO-European Institute of Oncology IRCCS, Dept. of Radiation Oncology, Milano, Italy ; 4 University of Milan, Department of Oncology and Hemato-Oncology, Milan, Italy Purpose or Objective to investigate robust optimization for lung SBRT treatment using a non-coplanar VMAT technique, taking into account tumor motion due to respiration. Plan robustness was evaluated against variation in time spent in each breathing phase and variation of the mean tumor position. Robust plan was compared with the ITV-based plan for 3D dose and 4D dose accumulation. Material and Methods Dynamic Wave Arc (DWA) is a novel non-coplanar VMAT technique implemented on the VERO system. The fluence modulation is achieved by a synchronized moving of gantry, ring and leaves at a fixed dose rate (400 UM/min). 4D CT of 2 patients treated for lung SBRT and CIRS phantom moving with 9 different breathing patterns (6 sinusoidal with A=5-15 mm and 3 with real patient breathing) were used for robust optimization using minimax optimization method in Raystation v8A (0.2 mm dose grid, Collapsed Cone Convolution Algorithm v3.5). Robust plan was calculated on the CT end-exale (CT EXP ) and optimized on the GTV+5 mm on each respiratory phase, while ITV-based plan was calculated on the CT average intensity projection (CT ave ), obtained from all breathing phases, and optimized on ITV+5 mm with a prescription dose of 54 Gy in 3 fractions with a goal of D 95% =95% for GTV + 5 mm and ITV+5mm, respectively. Doses were calculated on each phase, deformed onto both CT ave and CT EXP and accumulated over all phases. To assess robustness against realistic variation in time spent in each breathing phase, 3 methods of 4D dose accumulation were used: equivalent weight to all phases (∑ EQUI ), more weight to expiration phases (∑ EXP ), and more weight to the inspiration phases (∑ EQUI ). To evaluate plan robustness against a baseline shift of the tumor, perturbed doses with a shift from the isocentre of ± 5 mm in CC and AP direction independently were calculated. Both 3D and 4D doses of the robust plan were compared with the conventional ITV- based approach. Results Preliminary results for two patients and phantom with a tumor motion in CC direction of 10 and 15 mm are reported. All Robust and ITV-based plans satisfied criteria

Results As shown by Table 1, an average right lung dose reduction of 3.1 Gy reduction is achieved, whilst reducing the average target dose (D 98 ) by 2.1 Gy, with the scenario selection method as compared to a conventional one. Moreover, half of the number of optimization scenarios are necessary to produce this plan. Hence, the plan calculation time was reduced by half.