Emily Poon Frank Verhaegen March 6, 2006 GEANT4 for dosimetric study of an intracavitary brachytherapy applicator Emily Poon Frank Verhaegen March 6, 2006 McGill University Montreal, Canada
Ir-192 HDR Brachytherapy 3.5 mm 5 mm 1.1 mm 2 m 0.6 mm Nucletron HDR ‘classic’ model We generated a phsp file for 40 million photons reaching the capsule surface in a vacuum.
Validation of TG-43 parameters 0.5 0.6 0.7 0.8 0.9 1.0 1.1 2 4 6 8 10 12 14 radial distance (cm) radial dose function g(r) GEANT4 Williamson and Li (1995) (b) Radial dose function Anisotropy function Agreement: within 0.5% Agreement: within 2 %
Modeling of rectal applicator 2 cm made of silicone rubber 8 catheters for HDR 192Ir source allows for insertion of shielding shielding made of lead or tungsten
Applicator with balloon water or contrast medium tumor balloon with iodine solution protection for healthy tissue contrast medium for dose reduction and better localization of balloon
Plato treatment planning system CT-based dose calculations according to TG-43 assumes homogeneous water medium does not account for applicator and patient anatomy
TG-43 vs dose kernel calculations 100x100x40 voxels (10x10x10 cm3) computed using DOSXYZ because GEANT4 is too slow Lead shielding
3-D patient calculations using dose kernels 50 % 100 % 300 % no shielding lead shielding
GEANT4 simulations Low energy model Photon transport only Kerma calculations using track length estimation
Dose around the tip region no shielding tungsten shielding
Dose around the balloon no shielding tungsten shielding
Experimental validation Solid lines: GEANT4 dotted lines: EBT no shielding tungsten shielding Good agreement between GEANT4 and GafChromic EBT film measurements
conformal distributions can be attained by proper selection no shielding tungsten shielding conformal distributions can be attained by proper selection of source positions and dwell times tungsten shielding offers significant radiation protection
Ion chamber measurements Extradin A14P chamber 192Ir source variable thickness 30 x 30 x 30 cm3 Lucite phantom
GEANT4 vs ion chamber measurements DoseW/DosenoW Ion chamber GEANT4 Ion chamber: high uncertainties in partially shielded regions
Speed issues Number of voxels Time/history (ms) 125 0.815 1000 2.23 125000 120 CPU time for a 2.4 GHz processor to simulate a photon history in a 30x30x40 cm3 water phantom GEANT4 is too slow for patient calculations!
Boundary crossing problems We use track-length estimator for kerma calculations Dose dependent on photon step size When voxels are constructed as segments of a sphere, some photons cross the boundaries without stopping Errors in calculations Error is larger when θ spans a smaller angle
User code… We have an isotropic point source originating from (0,0,0) Phantom is homogeneous water Voxels are constructed as shown below: void Sphere01VoxelParameterisation::ComputeDimensions(G4Sphere& voxel,const G4int copyNo, const G4VPhysicalVolume*) const { voxel.SetInsideRadius(rInner[copyNo]); voxel.SetOuterRadius(rOuter[copyNo]); voxel.SetStartPhiAngle(0.*deg); voxel.SetDeltaPhiAngle(360.*deg); voxel.SetStartThetaAngle(87.5*deg); voxel.SetDeltaThetaAngle(5.*deg); } In this case, all photons reaching the voxel regions should have θ between 87.5º and 92.5º.
User code (cont’d) We set the voxels as “sensitive detector” (SD) As a test: in “ProcessHits” function of the user SD class, we recorded the pre-step and post-step positions of photons entering the voxels G4ThreeVector preStepPos=aStep->GetPreStepPoint()->GetPosition(); G4ThreeVector postStepPos=aStep->GetPostStepPoint()->GetPosition(); preStepPos.theta() and postStepPos.theta() should be between 87.5º and 92.5º In a test run of 1 million histories, 0.6% of the photons crossed the boundaries
Conclusions Dosimetric properties of a novel intracavitary brachytherapy applicator have been studied. GEANT4 results are in good agreement with GafChromic EBT film and ion chamber measurements. A phsp file of the 192Ir source that we generated using GEANT4 will be used in another code (to be developed) for fast Monte Carlo calculations. Speed and some boundary crossing issues need to be addressed.