TY - JOUR T1 - Optimized mesh generation for OSEM SPECT reconstruction JF - Journal of Nuclear Medicine JO - J Nucl Med SP - 407 LP - 407 VL - 50 IS - supplement 2 AU - Bo Yu AU - Yao Lu AU - Levon Vogelsang AU - Andrzej Krol AU - Yuesheng Xu AU - David Feiglin Y1 - 2009/05/01 UR - http://jnm.snmjournals.org/content/50/supplement_2/407.abstract N2 - 407 Objectives To generate triangular mesh that is optimized for the ordered subsets expectation maximization (OSEM) SPECT reconstruction in the mesh domain. Methods We developed a new method for optimized triangular mesh generation that consists of the following steps: (1) Assuming that all boundaries are known, nodes are uniformly distributed on all boundaries with minimum nearest-neighbor distance smaller than pixel size; (2) Maximum approximation error (MAE) of polygonal approximation using these nodes is calculated for each boundary and global greedy algorithm is used to add additional nodes such that approximation error is smaller than the smallest MAE among all boundaries; (3) An initial mesh is created by Delaunay triangulation; (4) Size and error functions are defined for every node and are used to eliminate “bad” mesh elements. The meshes were obtained for a test object and a thorax phantom. Synthetic SPECT projection sets were created for these two objects. Mesh-based OSEM reconstructions using our mesh and Engwirda-Persson (E-P) mesh were performed. Reconstructed image quality was investigated. Results Number of nodes in our mesh (2779) was significantly lower than in E-P mesh (3791) for the thorax phantom. OSEM reconstructions using our mesh produced superior images, as compared with reconstructions using E-P mesh, including: increased Contrast-to-Noise-Ratio, lower noise, lower bias, lower normalized mean squared error and higher uniformity. Conclusions OSEM mesh-domain reconstructed image quality can be significantly improved by application of our optimized mesh generation method. ER -