ESTRO 2020 Abstract book

S264 ESTRO 2020

requires regular repeat imaging. To comprehensively evaluate the overall treatment dose, accurate dose accumulation is required. This is achieved by deforming the dose via the displacement vector field (DVF) obtained through deformable image registration (DIR) between repeated and planning images. Unfortunately, different DIR algorithms provide different DVFs, thus affecting the evaluated dose accumulation. These differences are subject to both geometrical errors and unquantifiable uncertainties due to the lack of ground truth. In this study, we investigate differences in the accumulated proton dose in the presence of inter-fractional anatomical changes for NSCLC patients when using 6 different clinically used DIR algorithms. Material and Methods IMPT treatment plans (PTV dose - 66 Gy-RBE), were created for 7 NSCLC patients and the doses recalculated on 9 repeat-CTs acquired at different times during the treatment course. All CTs were acquired with visually guided deep-inspiration breath-hold to limit intra- fractional motion. Recalculated doses were then warped back and accumulated on the original planning CTs using different DIR algorithms (Plastimatch-Demon, Plastimatch-Bspline, Raystation-Morpheus, Raystation- Anaconda, Velocity and Mirada). Results The initial plan quality was degraded in the presence of inter-fractional anatomical changes. On average over all fractions, the PTV V95 was reduced by 14% (between -1.5% and -40.5% for single fractions) due to anatomical changes only (dose evaluated directly on the repeat-CTs). The calculated PTV V95 after warping these dose distributions back to the planning CT in average agreed well with the dose evaluated directly on the CT of the day (average PTV V95 difference 2%.). In some cases however, these differences reached 19%. Finally, for all cases, accumulated doses differed substantially from the initial plan (figure 1 and 2), with the spread of case specific, accumulated PTV V95 values as a function of DIR algorithm being 9% (range 1-26% depending on the patient).

Conclusion In the first systematic investigation of the sensitivity of PGI-based treatment verification in clinical prostate- cancer treatments, its capability to detect strong anatomical changes has been clearly demonstrated. In clinical PGI application the sensitivity is a bit smaller than for idealized PGI simulations, still severe changes were detected for all cases. The next step is to establish a reliable automated interpretation of PGI data. In a first trial, we established a two-parametric prediction model with already encouraging results. OC-0444 Impact of deformable image registration on inter-fractional variations in lung cancer proton therapy L. Nenoff 1,2 , C.O. Ribeiro 3 , M. Matter 1,2 , L. Hafner 1,2 , A.C. Knopf 3 , J.A. Langendijk 3 , M. Walser 1 , M. Josipovic 4 , G.F. Persson 4,5,6 , D.C. Weber 1,7,8 , A.J. Lomax 2 , F. Albertini 1 , Y. Zhang 1 1 Paul Scherrer Institute, Center for Proton Therapy, Villigen PSI, Switzerland ; 2 ETH Zurich, Department of Physics, Zurich, Switzerland ; 3 University of Groningen, University Medical Center Groningen- Department of Radiation Oncology, Groningen, The Netherlands ; 4 Rigshospitalet Copenhagen University Hospital, Department of Oncology, Copenhagen, Denmark ; 5 University of Copenhagen, Faculty of Medical Sciences- Department of Clinical Medicine, Copenhagen, Denmark ; 6 Herlev-Gentofte Hospital Copenhagen University Hospital, Department of Oncology, Copenhagen, Denmark ; 7 University Hospital Bern, Department of Radiation Oncology, Bern, Switzerland ; 8 University Hospital Zurich, Department of Radiation Oncology, Zurich, Switzerland Purpose or Objective Non-small cell lung cancer (NSCLC) patients can particularly benefit from high dose gradients, which are achievable by intensity modulated proton therapy (IMPT). However, the high probability of anatomical changes during treatment, causing severe dose degradation,

Conclusion