Fast and accurate iterative reconstruction using a volume integral model with look-up tables for pinhole SPECT.

The purpose of this study was to propose an algorithm that efficiently computes detection probabilities—an essential factor in determining the relationship between voxels and detector bins in iterative reconstruction using the maximum-likelihood expectation–maximization (ML-EM) method—and accelerates the reconstruction process by tabulating this information. In the proposed method, each voxel was subdivided into sub-voxels and projected onto the detector to approximate the detection probabilities. Furthermore, by precomputing these data and using them as a look-up table (LUT), the method achieved computational acceleration while reducing memory usage. In this study, the proposed method was applied to three-dimensional pinhole SPECT imaging, and its effectiveness was evaluated using an ideal infinitesimal pinhole and a pinhole collimator with a 2-mm-diameter aperture. LUTs were also implemented for the distance-driven (DD) method and Siddon’s method, in which detection probabilities are computed on the fly, and comparative evaluations were performed. Normalized root mean squared error (NRMSE) and normalized mean absolute error (NMAE) were used as metrics, and the methods were assessed in terms of the computation time required for table generation, the reconstruction time, and the storage size of the LUTs. The results demonstrated that the proposed method achieves higher reconstruction accuracy compared with the DD and Siddon’s methods.