Using the rPatlak plot and dynamic FDG-PET to generate parametric images of relative local cerebral metabolic rate of glucose

Local cerebral metabolic rate of glucose (LCMRGlc) is an important index for the description of neural function. Dynamic ¹⁸F-fluoro-2-deoxy-D-glucose (FDG) positron emission tomography (PET) has been used for quantitative imaging of LCMRGlc in humans, but is seldom used routinely because of the difficulty in obtaining the input function noninvasively. A reference tissue-based Patlak plot model (rPatlak) was proposed to generate parametric images of LCMRGlc in a quantitative dynamic FDG-PET study without requiring blood sampling. Dynamic emission scans (4×0. 5, 4×2 and 10×5 min) were acquired simultaneously with an IV bolus injection of 155 MBq of FDG. Arterial blood samples were collected during the scans via a catheter placed in the radial artery. Simulation data were also generated using the same scan sequence. The last ten scan data sets were used in a graphical analysis using the Patlak plot. The ratio of LCMRGlc estimated from the original Patlak (oPatlak, using plasma input) was used as the gold standard, and the standardized uptake value ratio (SUVR) was also calculated for comparison. Eight different tissues including white matter, gray matter, and whole brain were chosen as reference tissues for evaluation. Regardless of the reference region used, the slopes in the linear regression between oPatlak and rPatlak were closer to unity than the regression slopes between oPatlak and SUVR. The intercepts for the former were also closer to 0 than those for the latter case. The squared correlation coefficients were close to 1. 0 for both cases. This showed that the results of rPatlak were in good agreement with those of oPatlak, however, SUVR exhibited more deviation. The simulation study also showed that the relative variance and bias for rPatlak were less than those for SUVR. The images obtained with rPatlak were very similar to those obtained with oPatlak, while there were differences in the relative spatial distribution between the images of SUVR and oPatlak. This study validates that the rPatlak method is better than the SUVR method and is a good approximation to the oPatlak method. The new method is suitable for generating LCMRGlc parametric images noninvasively.