I am currently developing a FLUKA model of a SPECT scanner for radiation protection purposes. A key component of the system involves the collimators mounted in front of the detector heads. For realism, I have been referring to the specifications of an LEHRS (Low-Energy High-Resolution-Sensitivity) collimator:
Hole diameter: 1.43 mm
Septal thickness: 0.13 mm
Hole length: 32 mm
Given that the physical collimator spans over half a metre in both width and height, it’s evident that modelling such detailed geometry poses a considerable challenge, not only in terms of computation time, as highlighted in this paper, but also in setting up the geometry itself.
To address this, I initially simulated a small section of the collimator using a standard Nuclear Medicine phantom, recording the ratio of incident to transmitted fluence. I then replaced the detailed collimator with a homogenous lead block and ran multiple simulations with varying densities to match the same transmission ratio. I found that a density of ρ = 0.55 g/cm³ yielded comparable attenuation. However, this simplified model surprisingly required more simulation time to achieve acceptable statistics, somewhat contradicting the findings of the aforementioned paper.
Should I pursue modelling the full collimator geometry, despite the significant setup effort and large input files? I have two collimators (one per detector head), so I would have to define the regions separately for both.
Or should I use the simplified “bulk” collimator, even if it increases run-time and may compromise statistical efficiency?
If the detailed geometry is the preferred approach, is there a more efficient method of constructing it in FLUKA—perhaps a more automated or parametric way—rather than defining it plane by plane?
I would greatly appreciate any insights or recommendations from the community.
Concerning the choice of implementation—whether to use a detailed or simplified collimator geometry—I can certainly offer some advice, though keep in mind that this somewhat goes beyond the typical scope of the forum, as it touches on modelling strategy decisions which depend heavily on your specific application and constraints.
In the particular case of SPECT, since the source of radiation is within the patient, one could consider using a simplified model of the collimator and appropriate biasing. For radiation protection purposes, I guess the majority of the dose in the SPECT room—and outside of it—will primarily result from photons escaping the patient rather than those passing through the collimator material. In this context, a simplified collimator could be a solution for first estimates. Evidently, this simplified approach would not be suitable if the objectives of the simulation were the design of the detector system.
In any case, if you wish to pursue a more detailed modelling, you could explore the LATTICE capabilities of FLUKA: you can define a small portion of the collimator (for instance as you’ve already done which looks still manageable) and create replicas of it. The use of lattices is an advanced geometry topic: you can have a look at the Advanced Geometry lecture and of course at the LATTICE manual page for more details
I hope this helps and clarifies your options.
Feel free to reach out if further questions arise.
Best,
Davide
PS: I had a look at your file, and it seems there’s a geometry error: the PHANTOM and BLKBODY regions are overlapping
Thank you for your helpful response, and apologies for the delayed reply.
Your explanation was excellent. As you rightly pointed out, it’s not particularly useful to model the entire collimator in detail if the primary radiation escaping into (and out of) the room is scatter from the patient, or in my case, the phantom. I’ll adopt this simplified approach in my setup going forward.
Regarding the LATTICE capability, I’ve been aware of it for some time but never explored it properly. Now that I understand its potential, especially as I need to repeat the same procedure for a different collimator, I’ll make use of it. While I’ll still only simulate a small portion of the collimator, I plan to create an even smaller base unit and use LATTICE to replicate it to complete the geometry.
Thank you again. I really appreciate your guidance!