search

OpenGATE / Gate

Official public repository of Gate
http://www.opengatecollaboration.org
GNU Lesser General Public License v3.0
236 stars 264 forks source link

Greater number of delayed coincidences than random event coincidences in output ROOT files #435

Open robbietuk opened 3 years ago

robbietuk commented 3 years ago

I recently posted this issue on the users list but had no joy in diagnosing the issue. I have been simulating PET acquisitions using GATE on an XCAT phantom for reconstruction using STIR, which is also used to unlist the ROOT files into STIR compatable sinograms. In my experiments there is no radioactive decay or deadtime modeling.

I have been trying to debug an issue regarding a significant discrepancy between the number of event in the delayed coincidence window and the number of "random" events in the coinciudence window. I unlist the delayed coincidence ROOT file and find 9.15008e+07 events. I unlist the coincidence ROOT file, but set a filter to only process events if (eventID1 != eventID2), then I get 8.23657e+07 events, which is approximately 10% fewer. As I understand the theory, there should not be this signiciant variation.

As far as I can tell, my GATE macros are set correctly (see below) and there are no difference between the coincidence processing and delayed coincidence processing besides the offset. I apologies if there is a mistake that is unnoticed.

My coincidence and delayed outputs can be found here. My digitiser macro can be found here.

robbietuk commented 3 years ago

I have an update of this issue. I simulated a single voxel source with high activity (1e+08 in a {4, 4, 4} mm voxel) and no attenuation, which I think is air. The activity was simulated high enough that there was a presence of random coincidence events in both the coincidence window and delayed window.

The GATE simulation was run for approximately 2 hours (GATE time) and then I counted the number of different type of events as before to realise the following contributions to the ROOT file.

Event Types Coincidences Delayed
Non-random & unscattered events 2.42 e+9 0
Random & unscattered events 1.50 e+8 2.81 e+8
Non-random & scattered events 1.89 e+8 0
Random & scattered events 3.54 e+5 5.53 e+5

As we can see, the delayed window contains approximately double the number of events as the coincidence window, w.r.t. random events (from different eventIDs), and there are no contributions due to non-random events in the delayed coincidence data.

sparzych commented 3 years ago

Hi, I recently made a similar check between the number of random coincidences by setting a eventID1 != eventID2 condition and by using delayed time window in the digitizer (as you have). In my case I simulated a J-PET detector based on plastic scintillators and as a source I used a cylinder filled with water and activity. While I do not see such big discrepancy in the results, there is a ~1.5% increase in the delayed coincidences. If I understood correctly in your simulation you used only a source without any phantom. However, in the table you separate scattered events from unscattered. Here the scatters are then detector scatters?

robbietuk commented 3 years ago

If I understood correctly in your simulation you used only a source without any phantom.

Correct

However, in the table you separate scattered events from unscattered. Here the scatters are then detector scatters?

I believe so but I am not a GATE expert am not 100% sure if GATE assumes a vacuum or "Air" by default, where scatter may occur.