Online PET Reconstruction From List-Mode Data

This work proposes an alternative to ordered subsets to improve the convergence speed of list-mode expectationmaximization image reconstruction algorithms. Instead of subdividing the input data into non-overlapping subsets, the stream of measured coincidence events is immediately processed online. The reconstruction algorithm maintains a sliding window covering the events that contribute to the current image estimate. The image is seamlessly updated by adding a new contribution for the next event read from the list-mode and possibly subtracting old contributions from a batch of events that leave the window. This incremental event-by-event estimation method can reconstruct a dynamic image sequence in real-time during data acquisition. If the reconstructed object is static, the width of the sliding window can be expanded during the reconstruction process to balance between early estimation and global convergence behaviors. Encouraging results are shown on image reconstructions from a simulated static phantom and from a clinical dataset of a dynamic cardiac perfusion study.