- Email: [email protected]
PD-0039: An ungenuine complete geometrical description of the TrueBeam linac for Monte Carlo simulation
S12
ESTRO 33, 2014
(IFS) and OSLD calibration condition on OF were also investigated. TypeA uncertainty were estimated for minor variation in cone position, detector alignment during multiple repeat measurements. Results: The OF measured at dmax (1.4 cm depth) using SFD and FC65 combination varies from 0.666 for 4 mm cone to 0.922 for 15 mm cone. The corresponding values at 5 cm depths were 0.57 and 0.813. Selection of intermediate field size of 3×3 and 2×2 cm2 reduced the OF at dmax by 0.52% and 11.45% respectively. OF measured with SFD and PPC40 combination agrees well (-0.42%) with that of SFD and FC65. However, SFD and CC13 combination increase the OF by 3.4%. The use of CC01 reduced OF up to 19.25% for 4mm cone but become comparable for cone diameter > 7.5 mm. Type-A standard uncertainty due to positioning inaccuracy of cone and SFD mis-alignment was ±0.16% and ±0.51% respectively. The reproducibility of OSLD response was independent of cone size and was consistent within a co-efficient of variance of 1.5% for non-linear and 1.7% for standard high dose calibration mode. OSLDs in non-linear calibration condition showed deviation in OF ranging from 10.6% to 10.25% for 4 to 15 mm cones at dmax as compare to corresponding values with SFD. Although similar pattern was observed at 5 cm depth, the variation was within ±2% except for 4 mm cone, where the variation was -20.1%. OF estimated using standard high dose and non-linear calibration curve agrees within ±5% except for 4 mm cone where the variation is as high as 18.5%. Conclusions: Accurate measurement of small field OF requires choice of appropriate field detector, reference detector and intermediate field size. The reproducibility of OSLD response was independent of cone size and was consistent within a co-efficient of variance of 1.7%. As compare to SFD, Al2O3:C OSLD showed comparable OF for all applicator size at 5 depths excepting 4 mm cone. However, large deviation of up to ±10.6% was observed at dmax and is currently under investigation. PD-0038 When atomic composition should not be taken into account for scoring the dose in the context of CTV to PTV safety margins E. Sterpin1 1 Université Catholique de Louvain, Center of Molecular Imaging Radiotherapy and Oncology, Brussels, Belgium Purpose/Objective: During the inverse-optimization process in IMRT treatments, the time constraints of clinical practice often lead to the use of analytical algorithms that assume the patient as made of water material with electronic density scaled according to CT information. In the short future, more realistic dose calculation algorithms (like Acuros or Monte Carlo (MC)) will likely be more involved in the inverseoptimization process. Such algorithms are able to compute dose to medium. We qualitatively show in this communication that dose to medium based optimization in IMRT may not be adequate in the specific but yet widely adopted context of CTV to PTV margins. Materials and Methods: Treatment planning for IMRT treatments typically requires the optimization of incident energy-fluence to achieve homogeneous coverage of the PTV and sparing of the organs-at-risk. Although tumor tissues typically respond to radiation like water, the PTV may also include healthy tissues that respond differently (different dose for the same local energy fluence). To achieve a uniform dose in the PTV, an IMRT optimizer based on dose to medium implicitly compensates for these heterogeneous responses. Thus, the energy fluence is not homogeneous across the PTV. Due to geometric uncertainties, the CTV moves within this heterogeneous fluence. Therefore, the CTV does not receive a uniform dose, which is the primary aim in most current radiotherapy practice. For a TomoTherapy spectrum, ICRU compact bone shows an underresponse (compared to water) of about 4% assuming charged particle equilibrium. To illustrate our argument, a cylindrical phantom geometry was defined where the PTV includes a CTV (7 cm diameter) made of water surrounded by ICRU compact bone (1.2 cm thickness). Optimization of dose assuming everything in the PTV as water and optimization in dose to medium were performed to achieve a homogeneous dose to the PTV. Dose to medium optimization compensated for the 4% under-response in compact bone, thus delivering larger energy fluence in the PTV margin. The optimizations were performed using MC and a simple analytical model (see figure). The geometry was shifted by 1.2 cm to evaluate the impact of both optimization schemes on the coverage of the CTV.
Results: For both analytic and MC computations, compensating for the under-response of the bone material during the optimization phase (dose to medium optimization) leads to an unnecessary overdosage of the CTV for the shifted CTV. Meanwhile, optimization in dose assuming everything as water leads to a uniform coverage of the CTV Conclusions: In the context of IMRT using CTV to PTV safety margins, optimization in dose to medium may be inadequate considering the objective of uniform coverage of the CTV in the presence of geometric errors. Thus, computing the dose assuming everything as water with correct electron density should be preferred. This issue can be circumvented with robust optimization options, which is not widely available at the present time. PD-0039 An ungenuine complete geometrical description of the TrueBeam linac for Monte Carlo simulation L. Brualla1, M. Rodriguez2, L. Cozzi3, A. Fogliata3, W. Sauerwein1, J. Sempau2 1 Universitätsklinikum Essen, Strahlenklinik, Essen, Germany 2 Universitat Politècnica de Catalunya, Institut de Tècniques Energètiques, Barcelona, Spain 3 Oncology Institute of Southern Switzerland, Medical Physics Unit, Bellinzona, Switzerland Purpose/Objective: TrueBeam is the most recent linac marketed by Varian. The geometrical details corresponding to the patientindependent part of the linac head are not available to TrueBeam users. Instead, to allow Monte Carlo (MC) simulation of these machines, Varian offers a set of phase-space files (PSF) to selected costumers. These PSFs can be used as particle sources for simulations of the patient-dependent part of the head: jaws and multi-leaf collimators. MC simulation of a linac based on third-party PSFs imposes limitations: (i) it is impossible to adapt the initial beam parameters so to match experimental dosimetry for a specific machine; (ii) the energy deposited in the ion chamber, which is part of the undisclosed geometry, cannot be determined and, thus, only relative dosimetry can be done accurately; (iii) the finite size of the PSF imposes a lower bound on the statistical uncertainty; (iv) simulations downstream the patient-independent part are conditioned by possible inaccuracies of the code used for generating the PSFs. The purpose of this work is to provide an ungenuine geometrical description of the undisclosed part of the TrueBeam linac operating in Free Flattening Filter (FFF) mode that can be used for MC simulation and thus
ESTRO 33, 2014 to overcome the aforementioned limitations. We name this ungenuine geometry as FakeBeam so to make clear that it is a fake geometry that must not be assumed as the actual geometry of the cited linac. Materials and Methods: Experimental profiles measured in water for several reference fields ranging from 3x3 up to 40x40 cm2 were obtained from both the Varian Golden Beam Data Set (ion chamber) and from inhouse measurements (diode). Based on these data, an ansatz geometry for the patient-independent part of the head was conceived. MC simulations, with PENELOPE, were used to test and iteratively adapt the geometry until close agreement between experiments and MC results was reached. Additionally, MC simulations with the PSFs provided by Varian were also done for comparison. The agreement between dose profiles was assessed using gamma analysis with the criteria of 1%/1mm.
S13 Materials and Methods: A Monte Carlo code based on Geant4 was developed to simulate the acquisition of IDDs with a BPC. The selected energies were in the clinical range (26 energies between 100 and 226 MeV with a constant energy spread of 0.67 MeV). The cross-sectional area of the beam was energy-dependent and it had a gaussian intensity profile with a σ ranging from 5.1 mm for the lowest energy to 2.6 mm for the highest. These values were measured in air at the isocenter plane of our facility with a 2D scintillation-based detector. The code was validated in dose, measuring beam profiles (acquired with a compact cylindrical chamber) for each energy listed in Table at different depths in water and comparing shapes and ranges of the IDD curves. In the Table are reported R90 of several IDD measured and simulated.
Results: The comparison of the measured and calculated lateral profiles at 5 cm depth for the most demanding case, a 40x40 cm field at 6 MV, is presented in figure 1. The gamma analysis shows that 98% of the data points obtained with the proposed geometry pass the test. For the Varian's PSF only 55% of points pass the test. Similar or better results are obtained for different profiles at both 6 and 10 MV.
Conclusions: The proposed geometry allows the accurate MC simulation of a TrueBeam linac in FFF mode for 6 and 10 MV beams. The main advantages of our approach is that users can fine tune initial beam parameters to reproduce a particular machine and that the number of histories that can be simulated, and therefore the statistical uncertainty that can be reached, is theoretically unlimited. Full details on the FakeBeam geometry will be given and made freely available. Additionally, the geometry will be coded in the PRIMO (www.primoproject.net) system for its simulation. PD-0040 Monte Carlo simulation tool for commissioning measurements corrections of a proton pencil beam scanning (PBS) mode F. Fracchiolla1, S. Lorentini2, M. Schwarz3 1 Agenzia Provinciale per la Protonterapia (ATREP) and "Sapienza" University of Rome, Post Graduate School of Medical Physics, Ruvo di Puglia, Italy 2 Agenzia Provinciale per la Protonterapia (ATREP), Medical Physics, Trento, Italy 3 Agenzia Provinciale per la Protonterapia (ATREP) and Azienda Provinciale per i Servizi Sanitari (APSS), Medical Physics, Trento, Italy Purpose/Objective: Validation of a Monte Carlo tool to evaluate the impact of the Bragg Peak Chamber (BPC) volume and positioning errors in the measurement of proton Integral Depth Doses (IDD).
We simulated 0.5 mm-thick scoring planes at various depths in water with no detailed modeling of the chamber geometry. A comparison between the signal in the area covered by the BPC and the signal scored in the whole plane was carried out. An evaluation of the maximum acceptable shift, due to misalignments, between the scanning direction of the BPC and the beam axis was carried out. Results: The validation of Monte Carlo shows a very good agreement between measurements and simulations. The local signal lost by the BPC grows with the energy.It is less than 1% for low energy and reaches maximum values of almost 7% for the highest. Losses depends on the analyzed region of the IDD. They are low in the entrance and in the peak region and reach maximum values in the distal part of the plateau region, proximal to the peak, where secondary protons are scattered with large angles due to nuclear interactions. In Figure is shown the estimation of signal lost by the BPC(expressed in local percentage) for eight energies. No appreciable errors due to misalignments were introduced as long as they are less than 15.0 mm.
ESTRO 33, 2014
(IFS) and OSLD calibration condition on OF were also investigated. TypeA uncertainty were estimated for minor variation in cone position, detector alignment during multiple repeat measurements. Results: The OF measured at dmax (1.4 cm depth) using SFD and FC65 combination varies from 0.666 for 4 mm cone to 0.922 for 15 mm cone. The corresponding values at 5 cm depths were 0.57 and 0.813. Selection of intermediate field size of 3×3 and 2×2 cm2 reduced the OF at dmax by 0.52% and 11.45% respectively. OF measured with SFD and PPC40 combination agrees well (-0.42%) with that of SFD and FC65. However, SFD and CC13 combination increase the OF by 3.4%. The use of CC01 reduced OF up to 19.25% for 4mm cone but become comparable for cone diameter > 7.5 mm. Type-A standard uncertainty due to positioning inaccuracy of cone and SFD mis-alignment was ±0.16% and ±0.51% respectively. The reproducibility of OSLD response was independent of cone size and was consistent within a co-efficient of variance of 1.5% for non-linear and 1.7% for standard high dose calibration mode. OSLDs in non-linear calibration condition showed deviation in OF ranging from 10.6% to 10.25% for 4 to 15 mm cones at dmax as compare to corresponding values with SFD. Although similar pattern was observed at 5 cm depth, the variation was within ±2% except for 4 mm cone, where the variation was -20.1%. OF estimated using standard high dose and non-linear calibration curve agrees within ±5% except for 4 mm cone where the variation is as high as 18.5%. Conclusions: Accurate measurement of small field OF requires choice of appropriate field detector, reference detector and intermediate field size. The reproducibility of OSLD response was independent of cone size and was consistent within a co-efficient of variance of 1.7%. As compare to SFD, Al2O3:C OSLD showed comparable OF for all applicator size at 5 depths excepting 4 mm cone. However, large deviation of up to ±10.6% was observed at dmax and is currently under investigation. PD-0038 When atomic composition should not be taken into account for scoring the dose in the context of CTV to PTV safety margins E. Sterpin1 1 Université Catholique de Louvain, Center of Molecular Imaging Radiotherapy and Oncology, Brussels, Belgium Purpose/Objective: During the inverse-optimization process in IMRT treatments, the time constraints of clinical practice often lead to the use of analytical algorithms that assume the patient as made of water material with electronic density scaled according to CT information. In the short future, more realistic dose calculation algorithms (like Acuros or Monte Carlo (MC)) will likely be more involved in the inverseoptimization process. Such algorithms are able to compute dose to medium. We qualitatively show in this communication that dose to medium based optimization in IMRT may not be adequate in the specific but yet widely adopted context of CTV to PTV margins. Materials and Methods: Treatment planning for IMRT treatments typically requires the optimization of incident energy-fluence to achieve homogeneous coverage of the PTV and sparing of the organs-at-risk. Although tumor tissues typically respond to radiation like water, the PTV may also include healthy tissues that respond differently (different dose for the same local energy fluence). To achieve a uniform dose in the PTV, an IMRT optimizer based on dose to medium implicitly compensates for these heterogeneous responses. Thus, the energy fluence is not homogeneous across the PTV. Due to geometric uncertainties, the CTV moves within this heterogeneous fluence. Therefore, the CTV does not receive a uniform dose, which is the primary aim in most current radiotherapy practice. For a TomoTherapy spectrum, ICRU compact bone shows an underresponse (compared to water) of about 4% assuming charged particle equilibrium. To illustrate our argument, a cylindrical phantom geometry was defined where the PTV includes a CTV (7 cm diameter) made of water surrounded by ICRU compact bone (1.2 cm thickness). Optimization of dose assuming everything in the PTV as water and optimization in dose to medium were performed to achieve a homogeneous dose to the PTV. Dose to medium optimization compensated for the 4% under-response in compact bone, thus delivering larger energy fluence in the PTV margin. The optimizations were performed using MC and a simple analytical model (see figure). The geometry was shifted by 1.2 cm to evaluate the impact of both optimization schemes on the coverage of the CTV.
Results: For both analytic and MC computations, compensating for the under-response of the bone material during the optimization phase (dose to medium optimization) leads to an unnecessary overdosage of the CTV for the shifted CTV. Meanwhile, optimization in dose assuming everything as water leads to a uniform coverage of the CTV Conclusions: In the context of IMRT using CTV to PTV safety margins, optimization in dose to medium may be inadequate considering the objective of uniform coverage of the CTV in the presence of geometric errors. Thus, computing the dose assuming everything as water with correct electron density should be preferred. This issue can be circumvented with robust optimization options, which is not widely available at the present time. PD-0039 An ungenuine complete geometrical description of the TrueBeam linac for Monte Carlo simulation L. Brualla1, M. Rodriguez2, L. Cozzi3, A. Fogliata3, W. Sauerwein1, J. Sempau2 1 Universitätsklinikum Essen, Strahlenklinik, Essen, Germany 2 Universitat Politècnica de Catalunya, Institut de Tècniques Energètiques, Barcelona, Spain 3 Oncology Institute of Southern Switzerland, Medical Physics Unit, Bellinzona, Switzerland Purpose/Objective: TrueBeam is the most recent linac marketed by Varian. The geometrical details corresponding to the patientindependent part of the linac head are not available to TrueBeam users. Instead, to allow Monte Carlo (MC) simulation of these machines, Varian offers a set of phase-space files (PSF) to selected costumers. These PSFs can be used as particle sources for simulations of the patient-dependent part of the head: jaws and multi-leaf collimators. MC simulation of a linac based on third-party PSFs imposes limitations: (i) it is impossible to adapt the initial beam parameters so to match experimental dosimetry for a specific machine; (ii) the energy deposited in the ion chamber, which is part of the undisclosed geometry, cannot be determined and, thus, only relative dosimetry can be done accurately; (iii) the finite size of the PSF imposes a lower bound on the statistical uncertainty; (iv) simulations downstream the patient-independent part are conditioned by possible inaccuracies of the code used for generating the PSFs. The purpose of this work is to provide an ungenuine geometrical description of the undisclosed part of the TrueBeam linac operating in Free Flattening Filter (FFF) mode that can be used for MC simulation and thus
ESTRO 33, 2014 to overcome the aforementioned limitations. We name this ungenuine geometry as FakeBeam so to make clear that it is a fake geometry that must not be assumed as the actual geometry of the cited linac. Materials and Methods: Experimental profiles measured in water for several reference fields ranging from 3x3 up to 40x40 cm2 were obtained from both the Varian Golden Beam Data Set (ion chamber) and from inhouse measurements (diode). Based on these data, an ansatz geometry for the patient-independent part of the head was conceived. MC simulations, with PENELOPE, were used to test and iteratively adapt the geometry until close agreement between experiments and MC results was reached. Additionally, MC simulations with the PSFs provided by Varian were also done for comparison. The agreement between dose profiles was assessed using gamma analysis with the criteria of 1%/1mm.
S13 Materials and Methods: A Monte Carlo code based on Geant4 was developed to simulate the acquisition of IDDs with a BPC. The selected energies were in the clinical range (26 energies between 100 and 226 MeV with a constant energy spread of 0.67 MeV). The cross-sectional area of the beam was energy-dependent and it had a gaussian intensity profile with a σ ranging from 5.1 mm for the lowest energy to 2.6 mm for the highest. These values were measured in air at the isocenter plane of our facility with a 2D scintillation-based detector. The code was validated in dose, measuring beam profiles (acquired with a compact cylindrical chamber) for each energy listed in Table at different depths in water and comparing shapes and ranges of the IDD curves. In the Table are reported R90 of several IDD measured and simulated.
Results: The comparison of the measured and calculated lateral profiles at 5 cm depth for the most demanding case, a 40x40 cm field at 6 MV, is presented in figure 1. The gamma analysis shows that 98% of the data points obtained with the proposed geometry pass the test. For the Varian's PSF only 55% of points pass the test. Similar or better results are obtained for different profiles at both 6 and 10 MV.
Conclusions: The proposed geometry allows the accurate MC simulation of a TrueBeam linac in FFF mode for 6 and 10 MV beams. The main advantages of our approach is that users can fine tune initial beam parameters to reproduce a particular machine and that the number of histories that can be simulated, and therefore the statistical uncertainty that can be reached, is theoretically unlimited. Full details on the FakeBeam geometry will be given and made freely available. Additionally, the geometry will be coded in the PRIMO (www.primoproject.net) system for its simulation. PD-0040 Monte Carlo simulation tool for commissioning measurements corrections of a proton pencil beam scanning (PBS) mode F. Fracchiolla1, S. Lorentini2, M. Schwarz3 1 Agenzia Provinciale per la Protonterapia (ATREP) and "Sapienza" University of Rome, Post Graduate School of Medical Physics, Ruvo di Puglia, Italy 2 Agenzia Provinciale per la Protonterapia (ATREP), Medical Physics, Trento, Italy 3 Agenzia Provinciale per la Protonterapia (ATREP) and Azienda Provinciale per i Servizi Sanitari (APSS), Medical Physics, Trento, Italy Purpose/Objective: Validation of a Monte Carlo tool to evaluate the impact of the Bragg Peak Chamber (BPC) volume and positioning errors in the measurement of proton Integral Depth Doses (IDD).
We simulated 0.5 mm-thick scoring planes at various depths in water with no detailed modeling of the chamber geometry. A comparison between the signal in the area covered by the BPC and the signal scored in the whole plane was carried out. An evaluation of the maximum acceptable shift, due to misalignments, between the scanning direction of the BPC and the beam axis was carried out. Results: The validation of Monte Carlo shows a very good agreement between measurements and simulations. The local signal lost by the BPC grows with the energy.It is less than 1% for low energy and reaches maximum values of almost 7% for the highest. Losses depends on the analyzed region of the IDD. They are low in the entrance and in the peak region and reach maximum values in the distal part of the plateau region, proximal to the peak, where secondary protons are scattered with large angles due to nuclear interactions. In Figure is shown the estimation of signal lost by the BPC(expressed in local percentage) for eight energies. No appreciable errors due to misalignments were introduced as long as they are less than 15.0 mm.