Hello,
I’m trying to write a plan for a course that I have to give next semester, so accuracy is very secondary, I’m ready to make a lot of concessions to speed it up 😉
I took the model of linac in EGS++. It is defined as a stack of cones with inscribed geometries for the jaws and mirror for example, which should be very understandable for students who have very limited knowledge of MC. I just prolonged the stack to include the air at the exit of the linac and I place a XYZ geometry made of water to acquire pdds and profiles.
My problem is that when I include that XYZ geometry, the number of the regions in the simulation increase immensely (up to more than 4 000 000 regions), so much that I have the impression the whole linac is voxelised.
Is this possible? Should I use another way to model this? Place everything in a box of air for example and model the linac and the water box separately?
Thanks.
EGSnrc Google+ Archive 
I would use a CD geometry to combine the linac with the voxel phantom. Use egs_planes as the base geometry in the CD to define 3 planes, giving you two regions between the places. Place the linac in the top region and the phantom in the bottom region. I think this would be the most efficient route and shouldn’t voxelize your whole geometry (I’m not sure what happened there). There is an example of a CD geometry in HEN_HOUSE/egs++/geometry/examples.
I would recommend using egs_chamber and turning on range rejection and cross section enhancement in the phantom. However, note that if you turn variance reduction techniques on, the tracks will not be correct.
LikeLike
I ended up using a simple union. I put the linac and the phantom in a very large box of air. It seems to work fine. I have a correct number of regions now. Thanks.
LikeLike