A Fast List-Mode Reconstruction Algorithm with Dedicated Correction for Random Coincidences for the Clear-PEM System

A fast list-mode MLEM reconstruction algorithm including low-noise random correction for the Clear-PEM imaging system is presented. In this approach, the sensitivity matrix is estimated based on planar phantom acquisitions, rather than tracing along lines of response (LORs). The algorithm is implemented by distributing the calculation of the acquired list-mode events into different threads yielding significantly increased computational speed. A list-mode random correction approach is presented in which reconstructed data are corrected according to a smooth correction image estimated from delayed coincidence data. Monte Carlo simulations based on GATE were performed to validate the proposed approach. A Derenzo phantom study shows that the 1.2-mm hot rods can be resolved using the presented algorithm. A breast-torso phantom study as well as a study based on acquired patient data demonstrate its feasibility in a clinical environment. This method reaches a significant higher calculation efficiency in which the total reconstruction time is less than 30 seconds for a typical clinical study. Results show that bias from random coincidences is largely suppressed. The presented algorithm achieves a fast and low-noise reconstructions without compromising intrinsic system resolution and sensitivity.