Skip to Main Content
We have evaluated the 3D Ordered Subset Expectation Maximization (OSEM) algorithm for reconstruction of the projection data from a high-resolution 3D PET scanner. For this study, we used the inter-update Metz filtered OSEM (IMF-OSEM) algorithm, which has been developed by PARAPET project. The IMF-OSEM is an implementation of the OSEM algorithm with some additional capabilities such as inter-update filtering and random permutation of the subsets in each iteration. The projection data were acquired with the high-resolution PET camera developed at MD Anderson Cancer Center (MDAPET). This prototype camera, which is a multiring scanner with no septa, has a transaxial resolution of 2.8 mm that allows a better evaluation of the algorithm. We scanned three phantoms: a cylindrical uniform phantom, a cylindrical phantom with four small lesions, and the Hoffman brain phantom. The evaluation of the OSEM algorithm was performed by computing the noise level of the reconstructed images of the uniform phantom and by studying the contrast recovery for the hot lesions in warm background and also by visual inspection of images especially for the Hoffman brain phantom. In addition, the effects of post filtering and filtering during the reconstruction process have been evaluated. We observed that for the high statistics data, a good compromise between contrast recovery and noise level was achieved between 20 to 40 iterations for plain OSEM algorithm. By visually inspecting the images of Hoffman brain phantom and hot lesions, we observed that plain-OSEM algorithm, especially when followed by post-filtering, could also reasonably reproduce the phantom's structure. We also found that inter-update filtering has the potential to reach a noise level and contrast comparable to those from plain-OSEM at a smaller iteration number; however, it also has a higher tendency to develop noise artifacts.