Comparison of ScintSim1 and Geant4 Monte Carlo simulation codes for optical photon transport in thick segmented scintillator arrays

  Introduction: Arrays of segmented scintillation crystals are useful in megavoltage x-ray imaging detectors for image-guided radiotherapy. Most previous theoretical studies on these detectors have modelled only ionizing-radiation transport. Scintillation light also affects detector performance. ScintSim1, our previously reported optical Monte Carlo code for such detectors, includes the main optical processes of interest. Here, we benchmarked ScintSim1 against the more complex and extensive Geant4 optical Monte Carlo toolkit. We aimed to assess the importance of some simplifications made in ScintSim1 and gain further insight into the more relevant optical processes Materials and Methods: We used Geant4 (version 9.6. p02) in this study. As the aim of this study was to compare the simulation of optical photons in the two codes, dependence of the simulation results on x- ray and electron interactions had to be excluded. To compare the two models correctly, they had to have the same inputs. So, in both models, an equal number of light photons were generated at identical points, obtained from previously computed depth-dose distributions resulting from X-ray and electron interactions. We first performed various tests to check proper implementation of the Geant4 code then various optical results (such as numbers and distributions of incidences, absorption and reflection) were compared. In the second part of the study, Geant4 was used for coupled simulation of ionizing radiation (x-rays and electrons) and optical photon transport in the scintillator with a slit to obtain the ionizing- radiation-only line spread function (LSF) of the energy deposited in the elements and the number of optical photons incident on an underlying screen (‘full simulation’ including ionizing radiation and optical photons) for each case. A composite LSF was then calculated and compared with the corresponding optical results obtained by using ScinSim1 Results: The differences between the optical outputs of the two codes ranged between 0.1% and 1.7%. The full-width-at-half-maximum values of the LSF for energy deposited and total (ionizing + light) normalized LSFs from the two codes differed by only 0.01 mm and 0.1 mm, respectively. Using the same computer, the run time required to launch and simulate 108 optical photon histories (without any prior ionizing radiation transport) was 16 minutes in ScintSim1 and 421 minutes in Geant4. Conclusion: These results show a close agreement between the outputs of the two models for the studied conditions and suggest that the optical processes not modelled in ScintSim1 have little effect on the results required for detector optimization in megavoltage imaging using an array of segmented scintillation crystals. The model simplifications in ScintSim1 offer substantial run-time reductions if computational time-reduction methods are not used. Nonetheless, the developed Geant4 model is a powerful tool for further detector design parameter optimization.