TOF-PET ordered subset reconstruction using non-uniform separable quadratic surrogates algorithm

Time of fight (TOF) PET reconstruction provides statistically improved images and converges fast. However, the convergence rates of large and small regions are often different due to the measured counts for those regions are different. In particular, signal to noise ratio (SNR) for a small region is lower than that of a large region. This results in problems that can degrade the image quality during the TOF-PET reconstruction. To address this issue, the uniform convergence is necessary to improve the quality of the reconstructed image. In this paper, we propose a TOF-PET reconstruction algorithm exploiting an ordered subset non-uniform separable quadratic surrogates (OS-NU-SQS) algorithm. Although the OS-NU-SQS algorithm has been proposed for the transmission reconstruction, each step of the OS-NU-SQS can be easily extended to the emission reconstruction. In computer simulations, we confirm that the proposed method provides more accurate images and converges uniformly compared to the TOF-based conventional ordered subset expectation maximization (OSEM).