200 likes | 367 Views
Outline. Intro: voxel phantom + MCNPX Building geometry Source sampling (ext, int) Scoring (average, distribution) Examples: Klara, Godwin Summary. Voxel Phantom. Anatomical Realism. before. now. Millions of boxes assigned to different organs and media. MCNPX.
E N D
Outline • Intro: voxel phantom + MCNPX • Building geometry • Source sampling (ext, int) • Scoring (average, distribution) • Examples: Klara, Godwin • Summary
Voxel Phantom Anatomical Realism before now Millions of boxes assigned to different organs and media
MCNPX • Monte Carlo general purpose code • MCNP-4C3 + LAHET + Fluka89 + CEM2k + ... • 34 particles • Mix-and-match (tables vs models) • Fast lattice initialization new • Fast scoring in lattice new • And more…
Voxel Phantom Specification • Voxel dimensions X-Y-Z • 3D array of organ IDs ~ 10 MB in binary • Organ ID => Material • Material definitions elements, density
MCNPX Repetitive Geometry Traditional cells/tallies: 10e6? Repeated Structures voxel 3D lattice
x x y y z z STEP 1: Constructing 3D Lattice • Define voxel surfaces • Define lattice 1 rpp 0 0.2 0 0.20 0.8 Block 2 999 0 -1lat=1 u=999 &fill=0:254 0:131 0:219 & Block 1 Indices: Followed by array of universes (organ IDs)
STEP 2: Defining Basic Universes • Universe = Organ • Define a list of universes1 10 -0.95 7 u= 1$ universe #1 5 6 -1.05 7 u= 5$ universe #5 Block 1 Mapping: mat + rho => universe Plane 7 pz -1000e2 Block 2
x y z STEP 3Filling Phantom Box with Lattice • Define surfaces of phantom box2rpp 0 510 26.4 0 176 • Fill the phantom box cell with lattice 100 0 -2 fill=999 Block 2 Block 1
Miscellaneous • Geometry plot! • Compressor of voxel data bin 12 MB => ASCII 3 MB • READ card + NOECHO option read phantom.geom noechoread phantom.mat noecho • Suppress huge tables in production runs print -128 -140 saves RAM
Source Sampling • External Sourcesdef par=p erg=3 sur=2.1 x=51 y=d2 z=d3 vec=-1 0 0 dir=1si2 0 26.7$ samples Y from 0 to 26.7 cmsp2 0 1 $ samples Y uniformlysi3 0 176 $ samples Z from 0 to 176 cmsp3 0 1 $ uniformly • Internal Source(efficiency!)sdef par=p erg=0.1 cel=d5 x=d1 y=d2 z=d3 eff=1e-4si5 L (11<999<100) $ all voxels of uni 11 in lat999 in cell 100 sp5 1 $ uniform sampling# si1 sp1 si2 sp2 si3sp3 $ three distr-s for coord-s sampling0 00 0 0 0 $ coord-s start at low voxel border0.2 1 0.21 0.8 1
MCNPX Scoring in Lattice • Recent versions speedup for scoring in latticeby orders of magnitude in full mesh! • Unified syntax (source sampling and scoring)*f18:p,e (1<999<100) $ id=1 Adipose tissue, head • Newest version 2.5.0: F8 tallies in all voxels*f8:p,e (999<999[0:254 0:131 0:219]<100) talnp $ tally no printprdmp 2j 1 $ get data via MCTAL file
Adjusted Voxel Phantoms: Klara, Godwin • Based on CT data: woman and man • Adjusted to ICRP 89 Reference Man (V, m, h)
Godwin (2.085 × 2.085 × 8) mm 255 × 132 × 220 7.4e6 voxels 2e6 non-void 85 organs
mcplib02 Fixed XS
1 source photon No transport: NOTRN Time = 14 min
1-MeV Neutron Beam TECplot MCNPX plot
Summary • Geometry objects used • Fast initialization: < 1 min • Acceptable speed: 1-2 h per 10e6 src photons • RAM used: 500 MB (scoring, tables!) • Good agreement with EGSnrc • Get fresh version of MCNPX • MCNP-5