Fast Monte Carlo estimation of patient and detector scatter and crosstalk contamination in SPECT imaging

Monte Carlo methods offer the potential to provide an accurate correction for scatter and detector effects, especially when imaging isotopes that emit multiple photons, or dual-isotopes. Recently, a group at Utrecht University has developed some promising acceleration techniques by combining scatter maps with convolution-based forced detection. We extend their work in four ways: using the Delta Scattering technique to rapidly determine photon interaction points, simulating scatter maps for many energy bins to improve accuracy, decoupling the simulation of the object and the detector using point spread functions which include all detector effects, and including coherent scattering in addition to photoelectric effect and Compton scattering when simulating photon interactions. We store point spread functions as multi-dimensional arrays to improve speed and accuracy. A software package was developed and validated in a torso phantom experiment. Three spheres filled with ¹¹¹In were placed in the non-uniformly attenuating phantom. CT and SPECT scans were acquired and spatially registered. A good agreement between experimental and simulated profiles was found for four energy windows (50-95, 81-104, 147-195, 212-278 keV). For the two low energy windows, primary photons contribution can be modeled correctly only if the PSF includes all detector effects. A triple-energy window method overestimated scatter profiles compared to the Monte Carlo simulated scatter.