[Gate-users] Re: benchmark
Churchill, JF (Jonathan)
J.F.Churchill at rl.ac.uk
Wed Jun 6 19:35:23 CEST 2007
Hello Everyone
I have been installing the Gate 3.1.1 release on a Linux RHEL4 AMD
x86_64 platform and I see a mismatch on the SPECT (and PET) benchmarks
to the values quoted in the 3.0.0 manual (see results below). I notice
that a number of other users on this list have the same concern (Suhaili
Zakaria, Antigoni Divoli, Daniel Woodsworth and recently Maria
Holstensson).
Any suggestions as what can be done to resolve the discrepancy ?
Any thoughts or suggestions would be very welcome.
Kind regards
Jonathan
--- Details follow --------
Installation details:
Gcc 3.4.6
CLHEP 2.0.2.3
GEANT4 8.1.p02 (With G4EMLOW4.2 data)
Root 5.12
SPECT Benchmark results
Some results are very close but some are quite a way off.
For comparison to Maria's values on a SUSE system:
Percentage of unscattered photons
My average value: 35.66% , Benchmark value: 32.90% , Out by 8%
Percentage of photons whose last scattered event occured in the
collimator:
My average value: 0.35% , Benchmark value: 2.10% , Out by 83%
Percentage of photons whose last scattered event occured in the
crystal:
My average value: 6.42% , Benchmark value: 8.50% , Out by 24%
PET Benchmark Results
Same pattern as SPECT where some results are close and some are a way
off. Below is my output
>From the benchmarkPET.C root analysis. Also the gif produced does not
match the benchmark expected
output very well on the axial scatter fraction plot.
==> Expected total number of decays during the acquisition is
3.68126e+07 +/- 6067.34
There are 3.68151e+07 recorded decays
There are 313755 true unscattered coincidences
There are 23542 random coincidences
There are 368884 scattered coincidences
==> there are 706181 coincidences (true, scattered, and random)
==> global scatter fraction = 0.540379
==> absolute sensitivity = 0.852246 %
Measured O-15 life-time = 121.853 [s]
Nominal O-15 life-time = 122.24 [s]
==> difference = -0.316621 %
Gamma acolinearity FWHM = 0.606136 degree (expected: 0.58)
More information about the Gate-users
mailing list