Using dynamic [18F]fluorodeoxyglucose (FDG) positron emission tomography with a high-resolution, seven-slice positron camera, the kinetic constants of the original three-compartment model of Sokoloff and co-workers (1977) were determined in 43 distinct topographic brain regions of seven healthy male volunteers aged 28-38 years. Regional averages of the cerebral metabolic rate for glucose ( CMRglu ) were calculated both from individually fitted rate constants ( CMRglukinetic ) and from activity maps recorded 30-40 min after FDG injection, employing a four-parameter operational equation with standard rate constants from the literature ( CMRgluautoradiographic ). Metabolic rates and kinetic constants varied significantly among regions and subjects, but not between hemispheres. k1 ranged between 0.0485 +/- 0.00778 min-1 in the oval center and 0.0990 +/- 0.01347 min-1 in the primary visual cortex. k2 ranged from 0.1198 +/- 0.01533 min-1 in the temporal white matter to 0.1472 +/- 0.01817 min-1 in the cerebellar dentate nucleus. k3 was lowest (0.0386 +/- 0.01482 min-1) in temporal white matter and highest (0.0823 +/- 0.02552 min-1) in the caudate nucleus. Maximum likelihood cluster analysis revealed four homogeneous groups of brain regions according to their respective kinetic constants: (1) white matter and mixed brainstem structures; (2) cerebellar gray matter and hippocampal formations; (3) basal ganglia and frontolateral and primary visual cortex; and (4) other cerebral cortex and thalamus. Across the entire brain, k1 and k2 were positively correlated (r = 0.79); k1 and k3 showed some correlation (r = 0.59); but no significant linear association was found between k2 and k3. A strong correlation with CMRglu could be demonstrated for k1 (r = 0.88) and k3 (r = 0.90), but k2 was loosely correlated (r = 0.56). CMRglu kinetic ranged from 17.0 +/- 2.45 mumol/100 g/min in the occipital white matter to 41.1 +/- 5.62 mumol/100 g/min in the frontolateral cortex. In most regions the mean values of CMRglu kinetic did not differ significantly from CMRglu autoradiographic. With few exceptions, however, within-region variance was significantly less for CMRglu kinetic than for CMRglu autoradiographic, suggesting greater individual reliability of results obtained by the kinetic approach.