Treatment Plan Source on opengate_core

[MartiDvrk]

New TP Source implementation:

Python side:

• TreatmentPlanPBSource inherits from SourceBase
• User must provide:
  • beamline object, containing beam model and geometry parameters (e.g stearing magnets distance)
  • treatment plan filepath (.txt or .dcm). Alternatively, the user can input a beam_data_dict, dictionary with plan info, generated via 'spots_info_from_txt', or manually generated
  • number of primaries 'n'
  • particle type
  • source position and rotation with respect to the mother volume
  • beam_nr -> number of the beam to simulate (indexing from 1)
• Additional options:
  • flat_generation (default False) -> if True, each spot gets assigned the same number of primaries (flat probability density function PDF) and a weight proportional to the spot's msw. Otherwise (default), weight is 1 for each spot, and the number of primaries per spot is sampled from a PDF that weights each spot according to its msw.
  • sorted_spot_generation (default True) -> if True, each spot is exhausted before moving to the next one. Otherwise, at each event a different spot is sampled from the PDF.
  • gantry rotation axis (default 'z')
• User_info filled automatically at initialization:
  • PDF -> vector with a probability for each spot, defined accordingly to its msw (unless flat generation)
  • one vector for each spot feature, including: position, rotation, phase space parameters (X and Y direction separately), weight, energy, energy sigma.

Cpp side:

• GateTreatmentPlanPBSource inherits from GateVSource
• Implementation partly based on the Gate 9 GateSourceTPSPencilBeam
• Main idea:
  • a single particle source (fSPS) of type GateSingleParticleSourcePencilBeam is created only once, at initialization
  • in PrepareNextTime, a spot is selected. Either the next spot that still has primaries to generate (sorted) or a spot sampled from the PDF.
  • NOTE: sampling of from the PDF relies on the CLHEP::RandGeneral function (as in Gate 9)
  • if the spot is different from the one selected at the previous Time, the parameters of the fSPS are updated with the ones of the current spot
  • in GeneratePrimaries, the configured fSPS generates a vertex. The weight of the event is assigned accordingly to the one of the current spot
• Note on PrepareNextTime:
  • the simulation time is updated as the user had set an activity, with the difference that there is no random factor affecting the lenght of the time increment.
  • the number of particles is converted into time, by considering the following expression for the next time: next_time = t + 1/(Ntot*Bq).
  • the advantage of this approach, compare to simply setting a max number of particles, is that in case of multiple runs, the primaries are spread across the run_timing_intervals proportionally to the lenght of the interval, instead of being exhausted in the first run.
  • NOTE: somehow I currently simulate one particle less per run, still need to figure out why

Tests:

• I adapted the tp source tests (test059) to use the new source instead of the TreatmentPlanSource class, which was not a real source but rather an object responsible to initialize several PencilBeamSources.
• The tests seem to pass.
• More test needed (?)

I leave this PR as a draft, we can discuss together what to do with it :)

[dsarrut]

Hi @MartiDvrk should we merge it?

[MartiDvrk]

Hi @MartiDvrk should we merge it?

Hi David! yes, it would be great if you could merge it :)