Message: Re: DICOM: adding magnetic field produces NaN Not Logged In (login)
 Next-in-Thread Next-in-Thread
 Next-in-Forum Next-in-Forum

None Re: DICOM: adding magnetic field produces NaN 

Forum: Medical Applications
Re: Question DICOM: adding magnetic field produces NaN (Asja Pfaffenberger)
Date: 03 Aug, 2011
From: Asja Pfaffenberger <Asja Pfaffenberger>

Hi again,

Finding another issue: I decided to throw away all the NaNs before they get written into the EvtMap in the simulation where i have theGeometry->SetMagField(G4ThreeVector(0.,0.,0.)); and then compare the result with a simulation where i do not set any magnetic field at all (that runs without producing NaNs), so that i see if the difference is small enough that i can ignore the NaNs.

You find plots attached of the dose in a water phantom throwing away the NaNs, and the difference when subtracting the dose simulated without any magnetic field at all. The irritating part is that areas of positive values exist, while the dose where something was thrown away should always be smaller than where nothing is thrown away. - And this is obviously systematic in the buil-up region, not noise. The penumbras differ, too. The resolution you see in the plots is 0.5mmx0.5mmx3mm (512x512x70) voxels, and the radiation field 2cmx2cm, number of particles 10^9. Although it may be argued whether or not such a difference is clinically relevant, i thought that switching on the magnetic field transportation should not lead to so big effects..? Need to test what happens in more complex CT geometries before trusting the magnetic field results? Is this known? Is this effect amplified when B>0T?

The difference in integral dose is less than 0.5%, so that throwing away the NaNs seems to be basically possible.

All the best,



 Add Message Add Message
to: "Re: DICOM: adding magnetic field produces NaN"

 Subscribe Subscribe

This site runs SLAC HyperNews version 1.11-slac-98, derived from the original HyperNews

[ Geant 4 Home | Geant 4 HyperNews | Search | Request New Forum | Feedback ]