Verification of the Dose Distributions with Geant4 Simulation for Proton Therapy Tsukasa Aso (Toyama College of Maritime Tech.) A.Kimura (JST CREST) S.Tanaka ( Ritsumeikan Univ.) H.Yoshida (Naruto Univ.) N.Kanematsu (NIRS) T.Sasaki (KEK) T.Akagi (HIBMC) This work is partly supported by Core Research for Evaluational Science and Technology (CREST) of the Japan Science and Technology (JST)

IEEE-NSS Rome 2004/October2 Outline Hadron therapy facility Hadron therapy facility Bragg peak characteristics is suitable for the radio-therapeutic treatment of tumors. Bragg peak characteristics is suitable for the radio-therapeutic treatment of tumors. NIRS, National Cancer Center, HIBMC, Tsukuba-U in JAPAN NIRS, National Cancer Center, HIBMC, Tsukuba-U in JAPAN Request to develop simulation tools for Request to develop simulation tools for Designing beam delivery system Designing beam delivery system Validate or Proposing a treatment planning Validate or Proposing a treatment planning These efforts are employed by the approaches, so far, These efforts are employed by the approaches, so far, Experimental measurements (Trustable but hard to do everything) Experimental measurements (Trustable but hard to do everything) Analytical calculations (Model limitation for simplicity ) Analytical calculations (Model limitation for simplicity ) Simulation Tools is possible to include Simulation Tools is possible to include Complex geometrical effect Complex geometrical effect Material variety Material variety Different Physics processes for comparison Different Physics processes for comparison However, in order to apply simulation tools for the hadron therapy, it has to reproduce the dose profiles for patient safety. However, in order to apply simulation tools for the hadron therapy, it has to reproduce the dose profiles for patient safety. => Comparison of result is nesseary. => Comparison of result is nesseary. This talk gives a comparison of simulation with the measurement of proton beams at HIBMC including the validation of the beam delivery system as well as the dose distributions. This talk gives a comparison of simulation with the measurement of proton beams at HIBMC including the validation of the beam delivery system as well as the dose distributions.

IEEE-NSS Rome 2004/October3 HIBMC: Hyogo Ion Beam Medical Center Hyogo Ion Beam Medical Center located at Harima Science Garden City, Hyogo, JAPAN. Hyogo Ion Beam Medical Center located at Harima Science Garden City, Hyogo, JAPAN. Therapeutic beam is extracted from Synchrotron Therapeutic beam is extracted from Synchrotron 150,190,230 MeV proton 150,190,230 MeV proton 250,320 MeV/u Carbon ion 250,320 MeV/u Carbon ion Five treatment rooms, including two Gantry nozzles, those are up to 3.0 m in length, and 16 cm square irradiation field. Five treatment rooms, including two Gantry nozzles, those are up to 3.0 m in length, and 16 cm square irradiation field. Spring8 Hyogo Ion Beam Medical Center Treatment Room of isocentric rotating Gantry (Only for Protons)

IEEE-NSS Rome 2004/October4 HIBMC Gantry Nozzle Configuration  HIBMC Simulation Features  Based on GEANT4  All beam line elements in HIBMC Gantry Nozzle  Configuration param. in ASCII file Easy reconfiguration w/o recomplie  Wobbling magnetic field is set for each one of primary protons  Material parameters taken from NIST database  Ionization: Low Energy extension Bethe-Bloch + SRIM2000 swicthed at 10 MeV kinetic energy.  hadronic: LHEP_PRECO_HP Pre-equilibrium decay model. Geant4 educational package Wobbling field Lead Scatter Main Monitor Secondary Monitor(SEC) Ridge Filter Flatness Monitor Block Collimator (BLC) Multi-Leaf Collimator (MLC) Water Phantom Spreading system: Wobbling magnets/scatterer =>Uniform irradiation field Modulating system: Bar ridge filter => Spread Out Bragg Peak (SOBP) Monitor system: Ionization chamber / SEC Collimator system: BLC/MLC

IEEE-NSS Rome 2004/October5 Range (Stepping) 200MeV Proton ICRU 259.6mm G4hLowEnergyIonisation w/o ChmicalFormula NuclearStoppingOff BLK 3um RED from right 500 / 100 / 50 / 10 / 5 / 1 um Replica dz=100um GTWW500/503/504/505/506/507/510 Survived Proton fraction Depth in Water (mm)

IEEE-NSS Rome 2004/October6 Proton Range in Water -Default- ICRU mm Depth in water (mm) Number of survived Protons (%) 200 MeV proton G4hIonisation G4hLowEIonisation w/ NuclearStopping Range in default settings is shorter than ICRU value. MeanExcitationEnergy = eV

IEEE-NSS Rome 2004/October7 Proton Range in Water - ExcitationEnergy - defaultCutValue = 3um RED G4hIonisation BLK G4hLowEnergyIonisation w/o ChemicalFormula(“H_2O”) NuclearStoppingOff GTWW500/ MeV proton ICRU mm w/ MeanExcitationEnergy = 75eV (ICRU)

IEEE-NSS Rome 2004/October8 Range of Protons in water (mm) First of all, we have examined the range of proton in water. Our simulation uses G4hLowEnergyIonisation with low energy electro-magnetic processes. In the low energy hadrons ionization process, the energy loss function below the kinetic energy of 2 MeV is changed according to the setting of “ ChemicalFormula ”. If ChemicalFormula is set to “ H_2O ”, the energy loss is derived from the fit function of the ICRU, while it is not set, the energy loss is calculated from the sum of electronic stopping power of the elements in the material. Proton rate ChemicalFormula= “” ChemicalFormula= “ H_2O ” 200MeV protons ICRU 259.6mm If we dose not set the ChemicalFormula, simulated range is consistent with ICRU value. But with the setting of ChemicalFormula= “ H_2O ”, the simulated range become shorter.

IEEE-NSS Rome 2004/October9 Range of in water Eloss (MeV/mm) KinE (MeV) 2MeV G4hBetheBlochModelG4hICRU49p (Chem-NoChem)/NoChem KinE (MeV) ChemicalFormula= “” ChemicalFormula= “ H_2O ” For smooth connection at 2 MeV, the BetheBloch value is multiplied by the factor (ParamB). We changed the energy loss calculation at the region from 0.8 to 2 MeV to use the sum of electronic stopping power in G4hParamterisedLossModel. G4 used the correction factor “ paramB ” in the G4hLowEnergyIonisation Class. But the correction factor becomes large in Water ， so that the range become shorter than the measured value. ParamB = (Eloss_LowE)/(Eloss_Bethe-Bloch) Our Tentative Patch

IEEE-NSS Rome 2004/October10 Range of proton in Water 150MeV 157.8mm 190MeV 237.8mm 230MeV 329.5mm NIST PSTAR program calculation based on continuous-slowing-down approximation, with ICRU Report49(1993) ICRU 150MeV mm 190MeV mm 230MeV mm After the modification of the energy loss calculation, the G4hLowEnergyIonisation can reproduce the range of proton in water given by the ICRU Report. Geant4

IEEE-NSS Rome 2004/October11 Connection Factor The paramB Problem had been reported to Geant4 Bugzilla. After bug report, we got a comment to use SRIM2000 Model rather than default.

IEEE-NSS Rome 2004/October12 Ranges of proton are obtained in the simulation by switching off all processes except for ionization process. As a reference, the PSTAR program (National Institute of Standards and Technology, NIST) was used for comparison. The ranges obtained from Geant4 simulation are good agreement with reference values. The agreement is better than 0.1%, 0.3%, and 0.2% for water, lead, and aluminum, respectively. Material Validation: Range of protons Lines : PSTAR NIST Red symbols: Simulation

IEEE-NSS Rome 2004/October13 Beam line elements tuning/validation Wobbling radius Wobbling radius Standard Deviation ~ 14mm Wobbling trace w/o scatter 190 MeV Beam dispersion by scatter Beam dispersion by scatter Wobbling radius ~ 99 mm Beam dispersion w/o wobbling The magnetic fields of a pair of wobbling magnets are adjusted to fit a radius of circular trace in the measurement. Then a set of wobbling fields is randomly changed for each one of primary protons in the simulation. The primary beam dispersion at the nozzle entrance is derived from the comparison with measurements and included in the simulation, where parallel beam and Gaussian shape intensity are assumed. -1.5%-3% Thickness in measurement 1.6mm 150MeV 2.5mm 190MeV 3.5mm 230MeV

IEEE-NSS Rome 2004/October14 Beam line validation: Irradiation field Proton flux at an isocenter Proton flux at an isocenter Uniform lateral irradiation field at an isocenter is obtained by the combination of a pair of wobbler magnets and scatterer. In this case, the beam energy of 190 MeV with 99 mm wobbling radius and 2.5 mm lead scatter are selected. Edges of Multi-Leaf Collimator Simulation shows that 15 cm square field with +-2% uniformity is obtained, which follows the requirement in the measurement. 16cm diameter uniform filed

IEEE-NSS Rome 2004/October15 Dose verification(1): Bragg peak In order to compare two distributions, the following procedure is performed. (1)Depth-dose distributions are normalized at the peak position. (2) The Simulated distribution is fitted by the measured distributions with a displacement (a) and normalization (b) fitting parameters as f(z) = b D(z+a). a = mm b = a = mm b = a = mm b = Difference is about 2~3% The depth-dose distributions agree each other better than 4% accuracy. Dot: Meas. Solid: G4 150MeV 190MeV 230MeV

IEEE-NSS Rome 2004/October16 Range modulation: Bar ridge filter Bar ridge filter is used as the range-modulating system in order to obtain Spread Out Bragg Peak (SOBP). The ridge filter made of aluminum has 24 bars with the pitch of 5mm. It is processed within an accuracy of 30 um by the micro fabrication technique. The ridge filters used to produce SOBP at Gantry Nozzle were designed and built for 150 MeV and 190 MeV beams, respectively. The height of ridges is about 4cm and 6cm for 9cm width and 12cm width SOBP, respectively. Proton Energy w/o Ridge Filter w/ Ridge Filter for 9cm SOBP

17 Dose Verification(2): SOBP Normalization and fitting procedure are performed at the same manner with Bragg peak comparison. a= b= a= b= a= b= a= b= The shape of the SOBP is similar to the measurements. For 12cm width SOBP, fan beam effect is clearly reproduced. There are some discrepancy, but those are less than 4 % at the maximum. The small bump in the measurement is thought to be a fan beam effect which is never seen at the analytical calculation with the parallel beam approximation Dot: Measured Solid: G4 150MeV 190MeV SOBP 9cm SOBP 12cm

IEEE-NSS Rome 2004/October18Summary The GEANT4 based simulation for proton therapy has been developed. The GEANT4 based simulation for proton therapy has been developed. The obtained dose distributions are agree well better than 4 % difference with the measured value at HIBMC. The obtained dose distributions are agree well better than 4 % difference with the measured value at HIBMC. The results are derived from only the simple assumption of beam spot size at nozzle entrance. The use of realistic beam profile will improve the simulation. The results are derived from only the simple assumption of beam spot size at nozzle entrance. The use of realistic beam profile will improve the simulation. Further development is processed. Further development is processed. DICOM interface has already available. DICOM interface has already available. The plastic phantom DICOM data will be simulated and compared with the treatment system predictions. The plastic phantom DICOM data will be simulated and compared with the treatment system predictions.

IEEE-NSS Rome 2004/October19 Discussion Displacement of Bragg peak position Displacement of Bragg peak position Measurement systematic error, Measurement systematic error, Not sure, but ~500um, by the comparison of w/ wobbling and w/o wobbling. Simulation shows about 500um shift due to wobbling, but measured data is not. Not sure, but ~500um, by the comparison of w/ wobbling and w/o wobbling. Simulation shows about 500um shift due to wobbling, but measured data is not. Beam energy ambiguity, Beam energy ambiguity, Not sure. But peak shifts roughly 1mm/0.5MeV. Energy resolution of the synchrotron is thought to be 0.1%. If the beam energy shift about 0.3~0.5% higher, it is consistent. Not sure. But peak shifts roughly 1mm/0.5MeV. Energy resolution of the synchrotron is thought to be 0.1%. If the beam energy shift about 0.3~0.5% higher, it is consistent.

IEEE-NSS Rome 2004/October20 Ionization (Process connection) Ionization loss is scaled by ionloss*=(1.0+paramB*highEnergy/lowEdgeEnergy)

IEEE-NSS Rome 2004/October21 w/o Wobbling a= b= a= b= a= b=