Implementation of scatter corrected list-mode OP-EM reconstruction algorithm and a dual (histogram/list-mode) reconstruction scheme for dynamic PET imaging

Abstract

We describe an implementation of ordinary Poisson list-mode expectation maximization (OP-LMEM) algorithm for the high resolution research tomography (HRRT) with a scatter correction method based on the single scatter simulation (SSS) technique, and a random correction method based on the variance-reduced delayed-coincidence technique. A hybrid EM algorithm has also been implemented in list-mode reconstruction (H-LMEM) using delayed-coincidence event subtraction technique with the same scatter correction as in OP-LMEM. The reconstructed images of a dynamic scanning sequence of a contrast phantom have been compared with those reconstructed using the histogram-mode reconstruction, in particular, 3D ordinary Poisson ordered subset expectation maximization (3D-OP) with the same variance-reduced random and estimated scatter. The transaxial and axial profile analyses have shown excellent agreement between histogram and both list-mode reconstructions. Likewise the preliminary contrast and noise analyses have shown a close agreement between histogram-mode and both list-mode reconstructions. Based on these results, a dual reconstruction scheme can now be applied to dynamic imaging in positron emission tomography (PET) with scatter correction; i.e. histogram-mode reconstruction (3D-OP) can be applied to frames with a large number of counts, and list-mode reconstruction (OP-LMEM) will subsequently be used for low statistics frames, as an effort to obtain efficient and quantitatively accurate reconstructions applicable to state-of-the-art dynamic PET imaging using the HRRT.