Dosimetry Simulations with the UF-B Series Phantoms using the PENTRAN-MP Code System

Transcription

1 Dosimetry Simulations with the UF-B Series Phantoms using the PENTRAN-MP Code System A. Al-Basheer, M. Ghita, G. Sjoden, W. Bolch, C. Lee, and the ALRADS Group Computational Medical Physics Team Nuclear & Radiological Engineering University of Florida

2 Overview Dosimetry Simulations with the UF-B Series Phantoms Structure of Calculations PENTRAN-MP Code System UF Computational Phantoms Results and Discussion

3 PENTRAN-MP Code System Pre-processing GHOST-3D and DXS (3-D General Collapsing Code determines an effective phantom material distribution, DXS yields sources distributions ) PENMSH-XP (prepares mesh, source, and material distributions) CEPXS (prepares multi-groups Cross-section libraries ) Transport Calculation (Parallel Environment Neutral-particle TRANsport) Postprocessing 3D-Dose (calculate total 3D-dose distributions for all energy Group) 3DI (extracts data via a 3-D linear interpolation, and compares with reference data)

4 CT Images Simulation Methodology for Dose Computations CEPXS High Resolution Voxelized Human Phantom GHOST Voxelized Human Phantom For Transport Simulations Cross sections MCNP5 benchmarking PENMSH-XP PENTRAN PENDATA Flux distribution Built-in cross section Library (ENDF/B-VI) DXS X-Ray source generator Dose distribution Optimized Clinical Techniques

5 GHOST-3D 3-D General Collapsing Code By A. Al-Basheer & M. Ghita X-Y-Z voxel collapsing using two methodologies: Density Averaging Methodology Cubical collapsing Preset equivalent material distribution Density averaging based on ICRP46 output generation for MCNP and PENMSHxp Dominant Material Methodology Cubical collapsing Conservation of original material distribution Dominant material collapsing output generation for MCNP and PENMSHxp

6 PENMSH-xp (Yi & Haghighat) A Cartesian-based 3-D mesh generator 3-D Cartesian-based mesh generator prepares material distribution source distribution Methodology: a physical model has to be partitioned into slices along an axis (e.g., the z- axis) each z-level has to be partitioned into 2-D coarse meshes. each x-y coarse-mesh is partitioned into grids (fine meshes) of equal size Block adaptive meshing: different coarse meshes may have different grid densities.

7 The UF Computational Phantoms Series B UF Pediatric Series A phantoms developed from CT images of live patients. The UF Series B pediatric phantoms represent an extension of the UF Series A pediatric phantoms within which patient-specific in body dimensions and internal organ masses remained as viewed in the original CT images.

8 Model of Study: 11 Year Old Male Phantom This model considered up to 73 materials Matrix size [398,242,252] Voxel resolution= * * 0.6 cc Volumetric source ( cc) over the left side of the phantom chest 8 energy groups (10-90 Kev) 189,600 fine meshes divided uniformly to 30 coarse meshes unbiased quadrature sets were tested using S24, S32, and S42 Still need to investigate biased (splitting) quadrature sets

9 11 Year Old Male Phantom Cross-Sectional view UF-Series B phantom (11 years old male) with ( voxels); the corresponding PENTRAN downsampled models with ( voxels) and 66 materials, and another with ( voxels) and 72 materials

10 11 Year Old Male Phantom Cross-Sectional view UF_11yr voxel PENTRAN model - whole body

11 SELECTED IMAGE SLICES THE UF SERIES B PHANTOMS AS A PENTRAN INPUT and MATERIAL variation CT Image slice 27M voxel phantom 4-materials 189k voxel phantom 4-materials 900k voxel phantom 72-materials 900k voxel phantom

12 SELECTED IMAGE SLICES THE UF SERIES B PHANTOMS AS A PENTRAN INPUT and MATERIAL variation CT Image slice, 27M voxel phantom 4-materials 189K voxel phantom 4-materials 900K voxel phantom 72-materials 900K voxel phantom

13 SELECTED IMAGE SLICES THE UF SERIES B PHANTOMS AS A PENTRAN INPUT and MATERIAL variation Pentran discretized spatial-mesh slice, 4-materials 900k voxel Phantom CT Image slice, 27M voxel phantom

14 90 KeV source 3-D scalar flux distributions for four energy groups with S42 quadrature G1 (80-90) Kev G3 (50-60) Kev G5 (30-40) Kev G8 (80-90) Kev X-ray 3-D Group 1, 3, 5, and 8 scalar flux distribution computed by PENTRAN with the cepxs cross section library; an S42 angular quadrature (1848 directions) with P4 scattering anisotropy.

15 11 year old male phantom using 4 distinct materials Equivalent Geometry The MCNP5 model was defined to be equivalent to that used in PENTRAN model Equivalent Source spectra The Source was defined in a consistent manner with the various multi-group libraries considered Equivalent Volumetric meshtally (F4) tallies Tallies were equivalent to the discretized Sn volumes UF_11yr voxel MCNP model - whole body

16 PENTRAN/CEPXS Vs MCNP5 Flux through various parts of the model along Y axis, Model(189k voxel, 4-materials and 8 energy groups) Fig 1 Fig 2 Fig 3 Fig 4 Fig 1,2,3& 4: deterministic results using S42 angular quadrature (1848 directions) with P4 CEPXS showed an acceptable agreement with Monte Carlo

17 PENTRAN/CEPXS Vs MCNP5 Flux through various parts of the model along Y axis, Model(189k voxel, 4-materials and 8 energy groups) Fig 5 Fig 6 Fig 7 Fig 8 Fig 5,6,7& 8: deterministic results using S42 angular quadrature (1848 directions) with P4 CEPXS showed an acceptable agreement with Monte Carlo

18 PENTRAN/CEPXS Vs MCNP5 Flux through various parts of the model along Y axis, Model(189k voxel, 4-materials and 8 energy groups) Y=4.35cm X=13.0cm sigma-src<5% Y=16.6cm X=13.0cm G1-G6 sigma-src<6% sigma-5%-15% 5.00E E E E+05 Flux 4.00E E E E E E E E E E E E E E E+02 Z(cm) MC(G1) Sn42(G1)p4 MC(G3) Sn42(G3)p4 MC(G5) Sn42(G5)p4 MC(G7) Sn42(G7)p4 Flux 6.01E E E E E E E E E E E E E+02 Z(cm) MC(G1) Sn(G1) MC(G3) Sn(G5) MC(G5) Sn(G3) MC(G7) Sn(G7) Fig 1 Fig 2 Fig 1, Fig2: deterministic vs Monte Carlo results along the Z axis at Y=4.35 cm, X= 13.0 cm and at Y=16.6 cm, X= 13.0 cm both results showed acceptable agreement.

19 Simulation Methodology for Dose Computation UF_11YR Voxel Model UF Phantom As a PENTRAN Input Phantom Dose Distribution

20 SELECTED IMAGE SLICES OF THE UF SERIES B PHANTOMS COMPARED TO PENTRAN/CEPXS INPUT AND THE CORRESPONDING DOSE DISTRIBUTION CT Image slice 27M voxel phantom Corresponding dose distribution 4materials 900K voxel phantom

21 SELECTED IMAGE SLICES OF THE UF SERIES B PHANTOMS COMPARED TO PENTRAN/CEPXS INPUT AND THE CORRESPONDING DOSE DISTRIBUTION CT Image slice 27M voxel phantom Corresponding dose distribution 72-materials 900K voxel phantom

22 SELECTED IMAGE SLICES OF THE UF SERIES B PHANTOMS COMPARED TO PENTRAN/CEPXS INPUT AND THE CORRESPONDING DOSE DISTRIBUTION CT Image slice 27M voxel phantom Corresponding dose distribution 4materials 900K voxel phantom

23 Individual Organ dose is readily obtainable via post processing

24 Dose Volume Histograms (DVH) Left lung (organ in the radiation field). Liver (organ is partially in the radiation field) Right lung (organ is out side the radiation field where MC didn t converge).

25 Parallel PENTRAN Decomposition for Sn, Model (189k voxel and 8 energy groups) quadrature level (directions ratio*) Processors # Decomposition A E S S24 (1.0) ,21,24, S32 (1.74) ,21,24, S42 (2.96) ,21,24, Cumulative time** and memory requirements for four quadrature sets Phantom matrix size [79,48,50] Iterations # for (G1,G2,G3&G4) Parallel Cumulative time for MC Processors # 16 Cumulative Problem Time (hr) 17.6 Cumulative Problem Time (hr) Nondedicated Time ratio to S We estimated that more than 2000 hr, for tallies 70 cm away from the source, is needed to converge to less than 5% * Directions and time ratio are referred to the S12 quadrature set. ** Non dedicated running time Time ratio*

26 Monte Carlo Vs. Deterministic Term Deterministic Monte Carlo Geometry Discrete/ Exact Exact Energy treatment- cross section Discrete Exact Direction Discrete/ Truncated series Exact Input preparation simplified* simple Computer memory required Large Small/large Computer time Small Large Numerical issues Convergence Statistical uncertainty Amount of information Large Limited Parallel computing Complex Trivial Comparison between Monte Carlo methods and Deterministic techniques

27 The SN approach is an effective alternative to the Monte Carlo method for Accurately solve the transport equation over an entire human phantom Use in optimizing current radiation therapy treatment plans. A major advantage of the SN method combined with NURBS Computational Phantoms it can be sculpted to represent the patient avoiding Findings whole-body images of the patient (too costly) detailed organ segmentation of the partial body CT images for treatment planning Collisional kerma approximation Can is accurate for low beam energies dose calculations For therapeutic beams (high energies) require a different approach employing better representation of electron transport