Ultra-fast hybrid CPU-GPU Monte Carlo simulation for scatter correction in 3D PETs

A scatter correction algorithm is very important in 3D PET reconstruction due to large scatter contribution in measurements. Currently, one of the most popular methods is single scatter simulation (SSS), which enables a fast calculation of scattering with a relatively good accuracy; however, the accuracy of SSS is dependent on the accuracy of tail fitting to find a correct scaling factor. To overcome this drawback and to improve accuracy of scatter estimation by incorporating multiple scattering contribution, this paper implements a fast Monte Carlo (MC) simulation that considers accurate photon migration and interactions due to photoelectric absorption and Compton scattering. MC simulator provides a prompt and a scatter data separately; hence, a scaling factor can be calculated by comparing a simulated prompt data with a measured data rather than using a tail fitting, which enables a more robust calculation of the scaling factor. Furthermore, we exploit a hybrid CPU-GPU algorithm using the open multi-processing (OpenMP) and CUDA, which results in 124.3 times faster than using a single CPU.