ESTRO 38 Abstract book

S1149 ESTRO 38

asymmetric and dependent on the noise level in dose distributions. When setting a noisy dose as reference (MCref) in the gamma test, the passing rate (PR) is underestimated, while setting a noisy dose as evaluated dose (MCeval) results in an overestimated PR. MC algorithms nowadays can be fast, but not enough to avoid substantial noise. This work proposes a method to correct the gamma test result when applied to MC doses, while keeping a tractable computation time. Material and Methods The method was derived thanks to the similarities observed between multiple patients. It consists of extrapolating the PR corresponding to a low noise MC dose from several PR computed for various high noise levels. We avoid this way a too costly MC computation. The fitting function has 7 parameters. This model being quite flexible, the optimization is performed with 50 random initial conditions uniformly distributed in a given interval, and the best solution is kept. This method was applied to 5 PBS proton plans: prostate, lung, liver, brain, H&N. For each of them, using the fast code MCsquare, we computed 21 MC doses with different uncertainty levels: 7 for the fitting (high noise) and 14 for the validation (low noise). Gamma tests were performed between the MC and the TPS dose maps, for each available noise level and 4 different criteria. We then predicted the PR corresponding to the lowest MC uncertainty and compared our results with the actual PR previously calculated. Points used for the fitting corresponded to maximum 2E7 or 4E7 simulated protons in the MC calculation. Results Only the results for MCref case are presented here, but MCeval case is similar. As the method is not deterministic, we applied it 60 times and report here the results as mean values with corresponding standard deviations (SD). Figure 1 illustrates the method for a brain case, showing the fit and the error in each point for two gamma criteria. Table 1 gives the results for all patients. The gain obtained with our method is defined as the difference between two errors: the error made by the classical gamma index for a MC dose calculated with uncertainty S, and the error made by our method when using a smallest fitting uncertainty S. We are thus comparing the methods at equal MC computation time. We see here that the error on the PR at the lowest uncertainty is most of the times between 1% and 3%, with a SD about 1%. The gain is between 0.5% and 12%.

and summed using Fiji. RLI data were corrected for perspective distortion with a home-made program written using Matlab2018b. A perpendicular single beam (Source- Surface Distance = 80cm) was set with different field sizes. Two circular fields with diameter equal to 10 and 40 mm were tested. Measured planar dose distributions and dose profiles were compared against radiochromic films (GC), including FWHM estimate. Measurement repeatability was tested on both beam by moving and repositioning the CMOS for three times. Results The image correction procedure was confirmed to be able to reduce the perspective distortion at a negligible level and, at the same time, was reproducible. The agreement of the measured dose profiles and beam sizes is fairly good as shown in figure 2; light scattering due to the buildup slab slightly affected the flat region of the profile. The measured FWHM of the 40 mm beam obtained with RLI and CG were respectively equal to 42.4 mm and 42.7 mm.

Figure 2 Comparison of normalized RLI and GC beams profiles. The FWHM of the 10 mm beam were respectively equal to 10.9 mm and 11.3 mm: repeatability tests confirmed RLI values for FWHM within 0.3 mm. The RLI signal was detected in dimmed condition, showing the RLI dose measurements possible even when the lights in the treatment room cannot be totally removed. Conclusion A novel RLI approach based on a thin scintillator screen and a CMOS detector was tested for CK dosimetry, in particular for periodic constancy checks, also allowing real time dose information. Preliminary results show a good agreement with GC dose measurements; the set-up of the measurement (positioning of the detector, fixation, PMMA slab optimization, light collection parameters) may likely be further improved and future investigation is warranted. EP-2082 An adapted use of the gamma index method for Monte Carlo dose distributions comparison M. Cohilis 1 , S. Edmond 1,2 , L. John A. 1 , S. Kevin 1 1 Université Catholique de Louvain- Institute of Experimental & Clinical Research, Molecular Imaging- Radiotherapy and Oncology, Brussels, Belgium ; 2 Katholieke Universiteit Leuven, Department of Oncology, Leuven, Belgium Purpose or Objective In IMRT/IMPT, QA is essential. The quality of the treatment can be verified by comparing the TPS dose to measurements or to Monte Carlo (MC) simulations, which are slower and noisy due to their stochasticity, but more accurate. The most common method of dose comparison is the gamma test. However, this method is known to be