Abstract Book

ESTRO 37

S505

vertex and two lateral fields) intensity-modulated proton therapy plans prescribing doses of 54-59.4 Gy(RBE) to the PTVs were generated in the Varian Eclipse treatment planning system. The FLUKA Monte Carlo code was used to recalculate dose and obtain dose-averaged LET (LET d ) from the treatment plans. RBE-weighted dose distributions were calculated and compared for the brainstem and its substructures using the Carabe (CAR), McNamara (MCN) and Wedenberg (WED) models, as well as LET-weighted dose (LWD) (physical dose weighted purely on LET) and RBE=1.1 dose (RBE 1.1 ). RBE-weighted isodose lines were calculated for the different models and compared for isodose levels ranging from 5-60 Gy(RBE). Results The WED model typically calculated somewhat higher biological doses compared to CAR and MCN. The LWD and RBE 1.1 dose were overall lower than the doses derived using the WED, CAR and MCN models for all three patients (Pons D Median [Gy(RBE)] for the glioma patient: WED=45.5, MCN=43.9, CAR=44.6, LWD=36.0, RBE 1.1 =33.0). The shape of the different isodose curves were similar for all models, however, the RBE 1.1 and LWD isodose volumes were smaller compared to CAR, WED and MCN (Figure 1/Table 1). The doses derived with the variable RBE models were similar where the WED model in most cases reported the largest dose volumes within the substructures. Meanwhile, the RBE 1.1 dose volumes were typically the smallest (Midbrain V 50Gy(RBE) [%] for the ependymoma patient: WED=51, RBE 1.1 =41) (Table 1).

Conclusion A model that combines deep learning and radiomics can outperform the individual methods, indicating that there is complementary information in deep learning and radiomics features. Bibliography 1. D’Souza, G. et al. N. Engl. J. Med. 356, 1944–1956 (2007). 2. Ang, K. K. et al. N. Engl. J. Med. 363, 24–35 (2010). 3. Lassen, P. et al. Radiother. Oncol. 113, 310–316 (2014). PO-0933 Biological dose to brainstem substructures in scanning proton therapy of paediatric brain tumours L.F. Fjæra 1 , D.J. Indelicato 2 , K.S. Ytre-Hauge 1 , Z. Li 2 , L.P. Muren 3 , Y. Lassen-Ramshad 4 , S. Flampouri 2 , O. Dahl 5 , C.H. Stokkevåg 5 1 University of Bergen, Department of Physics and Technology, Bergen, Norway 2 University of Florida, Department of Radiation Oncology, Jacksonville, USA 3 Aarhus University/Aarhus University Hospital, Department of Medical Physics, Aarhus, Denmark 4 Aarhus University Hospital, Department of Oncology, Aarhus, Denmark 5 Haukeland University Hospital, Department of Oncology and Medical Physics, Bergen, Norway Purpose or Objective When irradiating paediatric posterior fossa tumours with protons, the dose to the brainstem is of great concern. It is recognised that the substructures of the brainstem may potentially have different radiosensitivities, particularly the transverse pontal fibers located in the anterior part of the pons. Proton therapy clinics are currently calculating treatment plans based on a constant relative biological effectiveness (RBE) of 1.1. Since the RBE varies with quantities such as the linear energy transfer (LET), the dose calculated using variable RBE models can be different, making additional evaluation of these doses beneficial. The main objective of this study was therefore to investigate how the choice of RBE models will influence biological isodoses to the substructures of the brainstem. Material and Methods The brainstem, brainstem surface (outer 3 mm) and core (brainstem cropped by 3 mm), midbrain, pons (including posterior and anterior pons) and medulla oblongata were delineated on CT/MR images for three paediatric patients previously treated with proton therapy. The diagnoses included craniopharyngioma, ependymoma and low-grade glioma in vicinity of the brainstem. Three field (one