Abstract
The authors studied the quality of dynamic PET images reconstructed from not a sinogram measured at each time but a set of sinograms measured during the observation. Representing a time activity curve observed at each voxel using set of basis functions, the authors reconstructed the images by estimating the coefficients for linear sum of the basis functions that maximize the likelihood. The likelihood is defined under the assumption that the measurement obey the Poisson distribution and maximized by means of an EM algorithm. In this presentation, the authors report that you can improve the image quality by simulnaneously reconstructing dynamic PET images from a set of all sinogram data.
