Does energy non-linearity affect noise estimates from Monte Carlo simulations of X-ray imaging detectors?

Publicly available Monte Carlo simulation packages for light transport in scintillator-based x-ray imaging detectors utilize a linear yield model. However, scintillators are known to be non-linear in the lower end of the energy range. Is this assumption reasonable for x-ray imaging simulations? We modified a freely available Monte Carlo package for modeling x-ray scintillators (MANTIS) to compare a linear versus a nonlinear optical yield model. We report simulations in the diagnostic x-ray energy range for a CsI:Tl. To determine the effect of the light yield model on imaging performance, we calculated the distribution of signal outputs characterized by the information or Swank factor. We find that the choice of yield model plays a significant role in the outcome statistics increasing the intensity of lower energy peaks favored by the non-linear model. This observation is confirmed by the results for the Swank factor with larger values for the non-linear models for an increase with respect to the linear model results of 42 and 26% for a high-resolution and a high-light-output model respectively. Our findings indicate that the assumption of linear light yield in Monte Carlo simulations of imaging detectors might introduce a significant bias in the estimates of noise as expressed by the Swank factor. More research is needed to implement realistic nonlinear yield models, to calculate the effect on realistic simulations including x-ray spectra of interest, and to experimentally validate these models.