Monte Carlo simulation based scatter correction in 3D list-mode image reconstruction

Abstract

We developed a Monte Carlo simulation based scatter correction method for 3D list-mode image reconstruction, and tested the method with Monte Carlo simulations. First, an emission image without scatter correction was reconstructed using MLEM. A transmission image was generated with the CT image. Then, based on the emission and transmission images, GATE was used to simulate coincidence events with their line-of responses (LORs) grouped according to their spatial positions (e.g. interaction positions or detector modules). The scatter ratio (scatter vs total coincidences) in each LOR group was calculated and stored in a scatter table. Finally, the scatter table was applied in a new image reconstruction to correct the scatter on the basis of each LOR group. The method was implemented in a simulated brain-size PET, with 300×300×100 mm3 FOV, 2×2×30 mm3 LYSO crystals, and 5 mm depth-of-interaction (DOI) resolution. Images of a 150×150×80 mm3 PMMA phantom inserted with three different radioisotope distributions were studied, including a point source array, a hot rod matrix, and a uniform source. We used detector module as the criteria to group LORs. With scatter correction, image resolution was almost the same as measured by point sources at different FOV positions; hot-rod sources showed visually improved image quality with reduced background noise; image SNR of the uniform source was not impacted. This method has been successfully implemented in the brain-size PET with improved image quality. It can be potentially applied to other list-mode 3D PET systems, with considering the accuracy and variation of scatter ratio in LOR grouping.