Fast cone-beam-based LOR reconstruction for 3-D PET

Iterative image reconstruction algorithms have been widely studied because they can provide better image quality than analytic reconstruction algorithms. In recent years, to achieve both good image quality and high computational efficiency, the line-of-response (LOR) reconstruction algorithm based on iteration had been developed. However, this algorithm requires a large amount of computation, which limits its practical application. There were several attempts to solve this computation problem by using a GPU. Since intrinsic LORs are not ordered, however, these attempts were limited to a ray-driven approach which is not suitable for parallel processing on a GPU. In this paper, we propose a fast cone-beam-based LOR reconstruction algorithm on a GPU. Experimental results show that the proposed algorithm reduces the computation time by about 31 times compared with a conventional GPU-based LOR reconstruction algorithm.