ESTRO 2020 Abstract Book

S91 ESTRO 2020

a generalized horse-shoe shaped oropharhyngeal disease surrounded by two parotids and a spinal cord as avoidance structures. The optimized beams were then copied onto the uncorrected scan, OMAR corrected scan and AMPP corrected scan. The dose distributions were recalculated without reoptimizing the proton beams in order to analyze dose distribution differences. Results Noticeable dose distribution differences were observed in the plans using OMAR corrected and uncorrected scans. Figure 1 shows the same axial slice on the phantom, centered inside the artifact filled region, for all the scan sets. There is under dosing in the posterior region near the spinal cord in the OMAR and uncorrected scans. In contrast, the dose distributions after AMPP show a similar dose distribution to the baseline scan without artifacts. AMPP also showed comparable target coverage compared to the baseline (96.8%) scan with 97.5% of target covered with prescription dose. The plan on the AMPP corrected images also showed a larger number of hotspots inside the target which could be the reason for the higher percentage of target coverage. Alternatively, the uncorrected and the OMAR corrected scans presented lower target coverages, with 84.7% and 77.9%, respectively.

dose distribution to the artifact-free baseline scan. The proton plans copied onto the OMAR and the uncorrected scans presented an underdosing of the target in the posterior region, further represented by the 77.9% and 84.7% of prescribed target dose coverage respectively. In comparison, the plans on the baseline and AMPP presented comparable prescribed target coverages. PD-0185 Development of a Monte-Carlo head model for a fast online validation of 1.5 T MR-linac plans M. Nachbar 1 , O. Dohm 2 , M. Friedlein 3 , J. Winter 1 , D. Mönnich 1 , D. Zips 2 , D. Thorwarth 3 1 University Hospital Tübingen, Section for Biomedical Physics- Department of Radiation Oncology, Tuebingen, Germany ; 2 University Hospital Tübingen, Department of Radiation Oncology, Tuebingen, Germany ; 3 University Hospital Tübingen, Section for Biomedical Physics, Tuebingen, Germany Purpose or Objective MR-linac (MRL) systems are a promising technology in which online plan adaptation during treatment is required. For a fast control of new plan characteristics, we developed an automatic verification routine using a Monte- Carlo (MC) head model for the 1.5 T MRL (Unity, Elekta AB, Stockholm, Sweden) which neglects the influence of the magnetic field. In this work we evaluate the physical differences and validate the clinical usability of the proposed method by comparison against experimental plan verification. Material and Methods An independent MC head model for the 1.5 T MRL was developed in our in-house treatment planning software Hyperion. It was designed based on measurements of the MRL and evaluated based on PDD, square fields and output factors (OF) without the magnetic field effect (B = 0 T). In the automatic online verification, Monaco (Elekta AB, Stockholm, Sweden) dose distributions (3 mm grid, MC- uncertainty 1%) are compared with a global γ-analysis (6mm/3%) to the recalculated dose (3 mm grid, MC- uncertainty 5 %). The γ-values are calculated evaluating all voxel doses >40% D max . To benchmark this fast online procedure against experimental plan QA, we compare it for n=100 IMRT plans against measurements evaluated with a local γ-criterion of 3mm/3% for 6 different tumor entities (table 1). Additionally, interpolation of the grid to 1mm is performed to compare dose distributions and measurements with a global γ-criterion of 3mm/3%. Results The developed MRL head model (B = 0 T) shows reasonably good agreement with the measured MRL beam data (B = 1.5 T) with mean absolute differences of 1.6% (inplane), 3.1 % (crossplane) and 2.2 % (PDD) in a 10 x 10 cm 2 field. Figure 1 shows OFs and PDD, cross- and inplane profiles of an exemplary 10 x 10 cm 2 field in comparison to the TPS (B = 1.5 T). Differences are visible due to the magnetic field effects (shorter build-up region, shifted profile in crossplane direction). The analysis shows a good mean agreement over all plans between the TPS and the measurement (98.67 %), the original voxel grid (97.34 %) and the interpolated grid (95.58 %). The mean calculation time above all entities was 01:27 min. Additional entity specific analyses are depicted in table 1. Based on a two-sided wilcoxon rank test only the measured data in mamma, H&N and rectal cancer was significantly different (p < 0.01) compared to the MC-based recalculations. Conclusion In this work we propose a fast method for an online verification of treatment plans, showing good agreement between measurement and calculation for entities treated with multiple beam angles and little air-tissue interfaces.

Conclusion Proton dose distributions were evaluated using different metal artifact reduction techniques. The in-house AMPP algorithm outperformed the uncorrected and OMAR corrected scans showing a treatment plan with a similar