ML reconstruction of dynamic PET images from projections and Clist

We present methods for maximum likelihood reconstruction of parametric dynamic images directly from observed positron emission tomography data, avoiding reconstruction of intermediate time frames. We show how to use an Expectation Maximization algorithm in this framework and derive a new optimality condition iteration algorithm. We also extend our approach to the reconstruction directly from a coincidence list (Clist) in continuous time. A Clist is a list of coincidence events (pairs of simultaneously detected photons) that are serially stored in the order of their registration. Reconstruction from a Clist is more accurate and computationally efficient as compared to the use of intermediate projection frames.