"CERN - PET"
This page compiles material related to simulation and reconstruction studies performed at
The following idealized axial 'brain PET' geometry (pdf file) is used.
The active area is a cylindrical arrangement of 14 identical modules
diameter of the inner circumscribed circle : 320 mm
diameter of the inner circumscribed circle: 380 mm
Crystal size: 3.2 x 3.2 mm, L = 100 mm
The crystal are arranged without any dead space
The modules are arranged without any dead space
Important: For FP6 related simulations the following more realistic axial 'brain PET' geometry (pdf file) shall be used.
The Erlangen study has the goal to apply the ITEM reconstruction algorithm, developed by this group and so far applied to Compton image reconstruction, to the problem of PET. As starting point of the reconstruction, a file with simulated gamma detection point pairs are used. This data file has been produced (CJ) using the above described axial geometry. The crystals have the dimensions (x,y) 3.2 x 3.2 mm2, however their length in z is infinite. The polar emission angle has been chosen such, that the z-coordinates are limited to values of approximately -60 < z < 60 mm.
1.) | 7 x 5 point sources.
The simulation includes positron diffusion and gamma non-collinearity, however not Compton scattering. The crystal material is not defined, however a gamma absorption length of 20 mm (similar to YAP:Ce) is assumed. All gammas have an energy of 511 keV ± 7.5% (FWHM). The z- resolution is assumed to be 5.9 mm (FWHM). This file (pdf file) contains plots showing the source configuration and the detector hits, before and after taking into account the detector granularity/resolution. The gamma detection points in the scintillators are listed in this data file (ascii, compressed with gzip). The format is: x1 y1 z1 E1 x2 y2 z2 E2 (all dimensions in mm, energies in MeV). It contains the coordinates of 1M gamma pairs. |
The study consists of two phases. (1) Adaptation of the existing simulation software EIDOLON, developed at HUG, to the the above 'axial' geometry, and (2) reconstruction of simulated data with conventional 2D or 3D statistical methods.
Crystal configuration (see sketch)
comments | |
Crystal size: 3.2 x 3.2 mm, pitch: 4 mm, L = 100 mm | (0.8 mm gaps between crystals) |
Crystal type: YAP:Ce | foresee LSO for later |
Light absorption length la: 75 mm | |
Geometrical factor kg:
1.2
|
path enhancement due to bouncing
|
Light Yield | |
N: = 875 · E / 511 keV |
detected photoelectrons per side for a deposited gamma energy E, la = infinite. Take E from simulation |
NL = N: exp(-kg z / la) | left photo detector |
NR = N: exp(-kg (L-z)/ la) | right photo detector |
Take z from the simulation,
calculate NL and NR and then ‘smear’ NL
and NR according to Poisson statistics
|
|
Determination of the detection point in 3D | |
xd, yd: use center xc, yc of hit crystal
|
|
zd: 0.5 (L + kg la log(NR/NL)) |
use smeared values of NL and NR |
Determination of the deposited energy | |
Ed = 511 keV ( NL exp(kg z / la) + NR exp(kg (L-z)/ la) ) / (2 N:) |
The Poisson fluctuation of NL and NR propagate automatically to zd and Ed. No further smearing is required.
Last updated by CJ: 21/02/03