Abstract: A major drawback of the iterative image reconstruction algorithms is their high computational cost. To decrease computation time, an Ordered Subsets (OS) method was proposed by Hudson et al. Because the OS algorithm uses a fixed number of subsets for each iteration, the quality of images depends upon the number of subsets. In this paper, we develop a new algorithm called the sequence subsets SART, which decrease the number of subsets after each iteration. Thus we not only accelerate convergence of the reconstructed image but also recover various frequency components of the reconstructed image in early iterative steps. Experimental results show that the proposed sequence subsets SART can provide higher quality reconstructed images that are very insensitive to noise at a small number of iterations.