Comparison of the SSRB, MSRB, and FORE methods with the 3DRP algorithm using data from a high resolution PET scanner

Abstract

The authors used the high-resolution prototype PET camera developed at MD Anderson Cancer Center (MDAPET) to scan a cylindrical uniform phantom, a cylindrical phantom with four hot small lesions in a warm background, and the Hoffman brain phantom. The MDAPET camera is a multiring scanner having no septa that allows for three-dimensional (3D) data acquisition. The acquired data were used to test several methods of rebinning 3D-projection data into two-dimensional (2D) sinograms. For this study, first, the 3D data were reconstructed using the 3D-reprojection (3DRP) algorithm; then images were used as a reference for testing the rebinning methods. The methods the authors tested were single-slice rebinning (SSRB), multi-slice rebinning (MSRB), and Fourier rebinning (FORE). After the data were rebinned, the 2D filtered-backprojection (2DFB) or 2D ordered-subsets expectation-maximization (OSEM) algorithm was used for image reconstruction. Rebinning the 3D data and using a 2D reconstruction technique resulted in noise reduction when compared with 3DRP; however, it reduced the contrast recovery and also showed some degradation in spatial resolution. The results also indicated that for the authors' prototype camera, the SSRB followed by 2DFB, which significantly reduces reconstruction time, provides a reasonable alternative to 3DRP.