Quantification limits of iterative PET reconstruction algorithms and improved estimation of kinetic constants

Abstract

Quantification of tracer kinetics is often accomplished from time-activity curves of a region of interest of dynamic PET images. The choice of reconstruction method may affect the timeactivity curves and hence the estimated kinetic parameters. Several studies have shown that statistical-iterative methods, due to non-negativity constrains, may exhibit a quantification bias in low activity regions and thus these methods, in spite of the better image quality they provide, are seldom used to estimate kinetic constants. By means of realistic dynamic simulations, we have investigated the quantitative properties of statistical-iterative (OS EM, both 2D and 3D) and FBP reconstruction methods and the accuracy of the kinetic parameters derived from images reconstructed with each algorithm. We focus on the procedure to fit kinetic constants to data. Our results show that, with appropriate measures to account for quantification bias, iterative reconstructions may be suited to derive kinetic constants from dynamic PET acquisitions.