Rationale and objectives: A statistical based iterative single-photon emission-computed tomography (SPECT) reconstruction algorithm (OSEM) modeling the depth-dependent collimator response in three dimensions has recently been introduced (OSEM3D). The aim of this study was to evaluate the axial shape fidelity of OSEM3D in comparison to OSEM, not taking this variable into account (OSEM2D).
Materials and methods: SPECT and separate spiral CT were performed in a phantom containing spheres filled with In-111. In-111-pentetreotide-SPECT and separate spiral-CT imaging were also performed in 22 patients with neuroendocrine tumors. Using window settings adapting the transversal size of the SPECT hot spots to that on CT and the 50% isocontour as boundary, the three-dimensional extensions (dx, dy, dz) of the SPECT representation of the structures under study were measured. These variables were also determined for CT. Furthermore, an index of eccentricity was calculated by averaging the ratios between dz and dx and dz and dy (IE). For isotropically imaged spheres, IE is 1.
Results: For OSEM2D, IE was significantly different from 1 in the phantom data (P < .05); this was not the case for OSEM3D and CT. This finding was accounted for by a significantly greater dz on the OSEM2D-SPECT images. In the patient data, dz was by approximately 15.5% greater for OSEM2D than for the other two modalities (P < .05).
Conclusions: The use of OSEM3D avoids deformation of hot SPECT lesions in z-direction. This may be of particular importance in SPECT/CT hybrid imaging capitalizing on the exact match of both modalities.