ESTRO 38 Abstract book

S299 ESTRO 38

uncertainty include motion, range and prediction of relative biological effectiveness (RBE). We present a platform to assess the combined impact of these uncertainties on dose distributions and DVHs in a variance- based sensitivity analysis (SA). Material and Methods Since the statistical approach of the variance-based SA requires a large number of RBE-weighted dose (RWD) calculations (10 5 -10 6 ), a fast, GPU-accelerated modeling of patient and range shifts was implemented, based on the pencil beam algorithm in an ion therapy extension of the research treatment planning system CERR. It was combined with the repair-misrepair-fixation (RMF) model for fast RBE calculation. In each calculation the input parameters (motion, range shifts, the biological reference parameter α/β for X-rays and RMF model parameters) are sampled independently within their assumed normal distributions (range: 3%+1mm, biological parameters: 10%, motion: 1mm in all 3 dimensions). The parameters are ranked by statistical formalisms according to their impact on the uncertainty of the RWD in every voxel, resulting in relative, normalized sensitivity indices (“S=0: no impact, S=1: only influential input”). Results are visualized in sensitivity maps and DVHs. Results The complete SA calculation including 5·10 5 RWD calculations was performed in <2 hours for a meningioma proton plan (dose/fx: 1.8 Gy (RBE)). The largest local uncertainty (0.6 Gy (RBE)) was discovered towards and after the distal fall-off of the spread out Bragg peak (SOBP) (fig. 1), where it is dominated by range uncertainties. In the lateral direction the overall uncertainty is governed by motion, while biological modeling is the most relevant contribution to the smaller uncertainty (0.15 Gy (RBE)) in the center of the SOBP. Consequently, OARs downstream of the target are affected primarily by range uncertainties and OARs lateral to the beam are affected more by motion. The uncertainty of the CTV D95% is affected by mostly by range and biology (fig. 2).

Fig.2: DVHs and sensitivities of selected quantiles. In the pie charts, the sensitivities of the input parameters are combined to three groups: motion (blue), range (yellow) and biology (turquoise). Conclusion A comprehensive sensitivity analysis framework for uncertainties in particle therapy was implemented. It is a powerful and flexible tool to assess the combined impact and interplay of motion, range and biological uncertainties, with possible implications for PTV definition and robust planning. Acknowledgment DFG-KA4346/1-1 OC-0570 Dosimetric study to guide preclinical trials in proton minibeam radiotherapy G. Consuelo 1 , L. De Marzi 2 , Y. Prezado 3 1 Centre national de la recherche scientifique, Imagerie et Modélisation en Neurobiologie et Cancérologie, Orsay Ville, France; 2 Institut Curie, Centre de Protonthérapie d’Orsay, Orsay Ville, France ; 3 Imagerie et Modélisation en Neurobiologie et Cancérologie, Imagerie et Modélisation en Neurobiologie et Cancérologie, Orsay Ville, France Purpose or Objective Proton minibeam radiation therapy (pMBRT) is a novel concept that combines the benefits of protons for therapy with the remarkable normal tissue preservation observed when irradiated with submillimetric spatially fractionated beams [1]. Thanks to multiple Coulomb scattering, the tumor may receive a homogeneous dose distribution, while normal tissues in the beam path benefit from the spatial fractionation of the dose. This promising technique [2] has already been implemented at a clinical center (Institut Curie-Proton therapy center of Orsay, ICPO) by means of a first prototype of a multislit collimator [3]. The goal of this work was to develop a set of dosimetric tools to be able to guide reliably preclinical studies. Material and Methods The complete ICPO beamline and pMBRT irradiations setup were modelled using GATEv7.0 simulations. A clinically relevant energy (100 MeV) was used. For minibeam generation the brass multi-slits collimator used in the experiments was modelled [3]. Dose distributions were recorded in a water phantom and voxelized rat CT images (7 week male Fischer rats), whose whole brains were irradiated at ICPO. One part of the animals was implanted RG2 glioma tumors intracranially. This study includes a control group (tumor-bearing rats, non-irradiated) and a group of tumor-bearing rats that received pMBRT (70 Gy peak dose in one fraction) with very heterogeneous dose distributions. Results The agreement between the MC and experimental data provided us with a benchmark. We generate a virtual source in the nozzle exit in good agreement with the measurements. Starting from our modelling, we are able to perform preclinical simulations with a significant reduction of the computational time. Hence, we have

Fig.1: Impact of uncertainties in motion, biological modeling and range on the overall uncertainty (local empirical standard deviation) on a meningioma treatment plan consisting of one IMPT field. The structures shown are the CTV (red), brain stem (green) and optic nerves (blue).