ESTRO 37 Abstract book

ESTRO 37

S532

PO-0966 Translation of a certified MR-only synthetic CT solution for prostate from photon to proton therapy C. Kurz 1,2,3,4 , M. Maspero 3,4 , G. Landry 2 , C. Belka 1,5 , K. Parodi 2 , P.R. Seevinck 4 , B.W. Raaymakers 3,4 , C.A.T. Van den Berg 3,4 1 University Hospital- LMU Munich, Department of Radiation Oncology, Munich, Germany 2 Faculty of Physics- Ludwig-Maximilians-Universität München, Department of Medical Physics, Munich, Germany 3 Universitair Medisch Centrum Utrecht, Department of Radiation Oncology, Utrecht, The Netherlands 4 Universitair Medisch Centrum Utrecht, Center for Image Sciences, Utrecht, The Netherlands 5 DKTK, German Cancer Consortium, Munich, Germany Purpose or Objective MRI plays an increasingly important role in radiotherapy (RT). It allows for improved target and OAR delineation as well as pre-treatment and online image-guidance in adaptive RT scenarios. Still, electron density or, in the case of proton RT, stopping power ratio (SPR) information for dose calculation is typically derived from CT imaging data. An MR-only RT workflow could reduce clinical workload, radiation exposure and uncertainties from CT- to-MRI registration, but requires the generation of so- called synthetic CTs (sCT) from MRI data for dose calculation. In this work, we investigate for the first time whether a sCT generation algorithm certified for photon RT can be used for accurate dose calculation in proton therapy for prostate cancer. Material and Methods Ten prostate cancer patients with rigidly aligned CT and 3T MRI data acquired in RT position within less than 2.5 hours were included in this study. A bulk-assignment method with a constrained bone shape model was used for sCT generation from a T1w gradient echo MRI scan with Dixon reconstruction [1]. As addition to the photon solution, an algorithm for detecting internal air on only the MRI data was implemented. Still, in order to minimize the impact of inter-scan differences in sCT and CT, internal air cavities were copied from CT to sCT during data evaluation (Figure 1). Body contours of both data- sets were matched by cropping and filling operations. Opposing field IMPT treatment plans optimized on CT were recalculated on sCT for comparison in terms of gamma-index analysis and clinically relevant target and OAR DVH parameters using the same structure set on CT and sCT. The proton range in beam’s eye view (BEV) was

compared utilizing single field uniform dose (SFUD) plans

Results Optimization of the assigned synthetic Hounsfield Unit (sHU) values of soft and cortical bone, using data of two exemplary patients, was crucial to yield satisfying proton range agreement on CT and sCT. Accurate localization of internal air was also critical. After adapting the bone sHU and copying internal air from CT to sCT, a mean difference in SPR of 0.1±0.3% (1σ) over all patients was found comparing CT and sCT adp in the region receiving at least 10% of the prescribed dose. The mean proton range difference for the SFUD plans was 0.1mm. More than 96% of all analyzed dose profiles in BEV had a range agreement better than 3mm (Figure 2). The average pass- rate for a (2%,2mm) gamma criterion with 10% dose threshold was 98±1%. Mean differences in all considered DVH parameters were below 1Gy (1.5%).

Conclusion Results suggest that MR-only proton dose calculation of comparable accuracy to CT is feasible using a fully automatic bulk-assignment sCT generation algorithm originally designed for photon RT. Adaptation of the assigned bone sHU values and accurate localization of internal air was required. Acknowledgments DFG-MAP, Deutsche Krebshilfe, ZonMw IMDI. Ethical protocol 15-444/C. [1] Helle et al., 2014, Proc. Int. Soc. Magn. Reson. Med.