A fast statistical image reconstruction algorithm for TOF-PET using DRAMA (Dynamic RAMLA) strategy

We have reported a fast iterative algorithm, DRAMA (Dynamic RAMLA), for emission CT. The relaxation parameter is controlled in such a way that the contribution of noise from subsets to the image is balanced at the end of each main-iteration. DRAMA provides a rapid convergence with a good S/N ratio by avoiding the limit-cycle problem. In this paper, we present “TOF-DRAMA” developed for time-of-flight (TOF) PET with the same strategy. We first formulated the recursive formula of TOF-DRAMA. Next, we developed a simple method to calculate the relaxation parameter taking into account of the TOF resolution. In TOF-DRAMA, we generate a rotated image, h-matrix, in the TOF direction. The h-matrix is converted to “t- matrix” by taking convolution with the TOF response function, and the ratio of t-matrix to the measured TOF projection is used for the iterative updating of the image. The sizes of h- and t- matrixes were determined by making a compromise between the imaging performance and the computation time. This paper describes the results of the simulation studies on the convergence behavior, spatial resolution, statistical noise, structural error, etc. In conclusion, TOF-DRAMA is useful as a fast statistical reconstruction algorithm for TOF-PET. The use of the TOF information improves not only the S/N but also the converging speed of the algorithm. The TOF gain in the S/N ratio was fairly close to the theoretical value.