The conventional measurement of the regional cerebral metabolic rate of glucose (rCMRGlc) with fluorodeoxyglucose (FDG) and positron emission tomography (PET) requires arterial or arterialised-venous (a-v) blood sampling at frequent intervals to obtain the plasma input function (IF). We evaluated the accuracy of rCMR-Glc measurements using population-based IFs that were calibrated with two a-v blood samples. Population-based IFs were derived from: (1) the average of a-v IFs from 26 patients (Standard IF) and (2) a published model of FDG plasma concentration (Feng IF). Values for rCMRGlc calculated from the population-based IFs were compared with values obtained with IFs derived from frequent a-v blood sampling in 20 non-diabetic and six diabetic patients. Values for rCMRGlc calculated with the different IFs were highly correlated for both patient groups (r > or = 0.992) and root mean square residuals about the regression line were less than 0.24 mg/min/100 g. The Feng IF tended to underestimate high rCMRGlc. Both population-based IFs simplify the measurement of rCMRGlc with minimal loss in accuracy and require only two a-v blood samples for calibration. The reduced blood sampling requirements markedly reduce radiation exposure to the blood sampler.