Efficient point clouds based scatter correction for fully 3D PET

摘要

A novel way to perform scatter correction for Positron Emission Tomography by using point clouds representation is presented. The scatter correction method is based on single scatter simulation. The strategy of scatter point sampling during scatter calculation will affect the accuracy of scatter correction directly and the number of sampled scatter points determines the computation complexity. Compared with regular voxel based scatter point sampling, the new method, which has been greatly reduced in size, adopts content-adaptive point clouds structure. The point clouds method can capture both anatomical and functional information, and cover the critical regions with limited number of points. Reduced sampled points significantly bring down the computation cost. Experiments with Monte Carlo simulations prove that such process can reduce the time consumption while at the same time maintain the reconstruction accuracy at the same level.