Abstract
In order to improve counting rate and reduce patients’ dose for diagnosis, the new generation of PET scanners retracted septa during scanning and collected LOR with large acceptance angle. For such kinds of LORs, 3D image reconstruction algorithm is necessary. Except for time-consuming 3D reconstruction algorithm, alternative choice is rebinning algorithm, which converted 3D data set to a parallel stack of 2D data set. Consequently 2D data set could be implement by 2D image reconstruction algorithm, and the time-saving method could be applied clinically in practice. There are three normal rebinning methods, including single slice rebinning (SSRB), multi-slice rebinning (MSRB), and Fourier rebinning (FORE). In this thesis, we mentioned Orthogonal projection MSRB (ortho-MSRB) to improve the problem of blurring belong the z direction in MSRB. We made a stack of sinogram table by using single slice rebinning (SSRB) with difference one sinogram to estimate the source distribution belong the LOR. Based on the sinogram table, we can rebin 3D data set in the z direction accurately. To understand the performance of Ortho-MSRB, we simulated several point sources and line sources in 3D space by SimSET, which released by Washington University, and treated the 3D data respectively by using the four rebinning methods. As the result, the Ortho-MSRB has good performance for the resolution of z direction.