Kalman sinogram restoration for fast and accurate PET image reconstruction

Sinogram restoration followed by filtered backprojection has been shown to be a fast alternative to direct image reconstruction method for obtaining PET images of good quality and quantitative accuracy. In previous studies, the authors used detector motion and the method of projection onto convex sets for sinogram restoration; but, the performance was observed to degrade rapidly when the noise level of PET data increased. In this paper, the authors describe a sinogram restoration method based on Kalman filtering that provides surprisingly good results with very fast computations. The new method shows greatly improved noise performance at realistic noise levels, and its computational cost is little more than that of conventional PET image reconstruction by filtered backprojection.