A simplified blood sampling scheme in FDG-PET studies

Quantitative measurement of brain glucose metabolism with positron emission tomography (PET) and fluorodeoxyglucose (FDG) involves arterial blood sampling to estimate the delivery of radioactivity to the brain. Usually, for an intra-venous bolus injection of 30 sec duration, an optimal manual sampling requires more than 25 blood samples since a frequency of 1 sample every 5 sec or less is necessary to determine the peak activity in arterial plasma during the first 2 minutes after injection. In the present work, 13 standardized sampling times were shown to be sufficient to accurately define the input curve. This standardized input curve was subsequently fitted by a polynomial function for its rising part and by spectral analysis for its decreasing part. Using the measured, the standardized as well as the fitted input curves, the regional cerebral metabolic rate for glucose (rCMRGlc) was estimated in 32 cerebral regions of interest (ROIs) in 20 normal volunteers. Metabolic values were obtained by both the autoradiographic method (ARG) and the kinetic analysis of dynamic data (DYN). Comparison of rCMRGlc values obtained with the measured and the fitted input curves showed that both procedures gave consistent results, with a maximal relative error in mean rCMRGlc of about 1% in ARG and 2% in DYN studies. This input curve fitting technique which was not dependent on the peak time occurrence, allowed an accurate determination of the input curve shape from reduced sampling schemes