Reconstruction of linear kinetic parameters directly from projection PET data

Dynamic Positron Emission Tomography (PET) data provide functional information. Usually, this is measured in the form of pharmacokinetic parameters derived from the temporal response of each region. Recent trends have shown that when pharmacokinetic parameters are estimated directly from the projection data, they are less affected by noise. This work investigates an existing parametric maximum likelihood expectation maximization algorithm applied to [18F]DOPA data using reference-tissue input function. The study reveals how direct reconstruction of pharmacokinetic parameters from the measured data can be performed optimally. It explains how to optimize the speed of the standard iterative algorithm and it compares the results with the existing FBP method. The improvement of the quality of the parametric images preserving quantification suggests the usefulness of direct estimation of the kinetic parameters. This algorithm is freely available within the open-source library STIR 2.1.