Intrahepatic fluorine-18-fluorodeoxyglucose kinetics measured by least squares nonlinear computer modelling and Gjedde–Patlak–Rutland graphical analysis

Abstract

We aimed to use simple physiological equations to show similarities and differences in blood fluorine-18-fluorodeoxyglucose (18F-FDG) clearance in the liver (Ki) and distribution volume (V0) in the liver, respectively, generated from nonlinear least squares computer modelling and Gjedde–Patlak–Rutland graphical analysis of dynamic 18F-FDG PET. We show theoretically that when, as is usually the case, vascular fraction (fraction of liver occupied by blood, Vb) is included as a parameter in modelling but ignored in graphical analysis, the ratio of Ki values, respectively, generated by graphical anlaysis (GKi) and by modelling (MKi) is equal to 1−Vb. This theoretical prediction was then confirmed from dynamic PET data acquired in a clinical population of patients undergoing routine 18F-FDG PET/computed tomography and from a review of the literature in which it can be seen that GKi/MKi ranges from 0.47 to 0.98. When Vb is not included as a parameter in modelling, GKi is theoretically equal to MKi and to MV0·k3, and GV0 is equal to MV0. There are several attractions to normalising Ki to V0 with respect to liver. Thus, first, there is no need to correct imaging for photon attenuation. Second, it makes no difference whether uptake constant is expressed as blood or plasma clearance. Third, it circumvents the effects of hepatic fat, which, because it accumulates negligible 18F-FDG, physically dilutes the 18F-FDG signal and reduces both uptake constant and distribution volume.