Accelerated Iterative Algorithm For Image Reconstruction

Iterative algorithms such as algebraic reconstruction technique (ART) and expectationmaximum (EM) algorithm are rapidly becoming the standard for image reconstruction forsimplicity,efficiency and performance. However, the major drawback with the iterative imagereconstruction is the computation, that is , the convergence of the reconstructed image is veryslow. This is why the iterative method had never been used in clinical studies. To acceleratethe iterative image reconstruction, this paper introduce the ordered subset (OS) method,subset sequence method and statistically regulated method, and we call the (OS) methodapplied to EM the OS-EM, subset sequence method applied to EM subset sequence EM andstatistically regulated method applied to EM statistically regulated EM. We analyzeconvergence situation of OSEM when the order of the projection direction is various andsubset level is different and rightly introduce subset sequence method when OSEM may notconvergence at the situation of high subsets level and high noise level, and provide themethod of selecting the sequence of subsets. We modify the original test statistic ofstatistically regulated EM. and enlarge its using bound. Simulated and experimental resultsshow that these methods can provide high quality of the reconstruction image after a smalliterative and when the projection data is noisy the second and the third methods show theirsuperiority.