Various approaches estimating local cerebral glucose utilization by positron emission tomography of labeled deoxyglucose are compared. Autoradiographic methods that predict the glucose utilization rate from a single scan are unreliable in pathologic tissue because of abnormal values of the model rate constants. A normalization procedure using the ratio of measured tissue activity to activity calculated with standard rate constants is proposed to readjust the values of the rate constants. Reliable estimates of metabolic rates can be obtained from dynamic recordings of tracer uptake. In the graphic approach, metabolic rate can be derived from the slope of a segment of a transformed uptake curve, which becomes linear at 15-20 min after intravenous tracer injection, with an accuracy comparable with that in complete dynamic studies. However, by recording and analyzing full-length uptake curves, in addition to metabolic rate, the model rate constants can be determined regionally. The physiological significance of those parameters is demonstrated in crossed cerebellar deactivation in 30 patients with supratentorial infarcts. Mild hypometabolism both within the ischemic lesion and in the morphologically intact cerebellum is accompanied by a reduction of the phosphorylation rate only. Severe metabolic depression, by contrast, affects both cerebellar transport and phosphorylation processes, whereas in the cerebrum, only the rate constant k1 is significantly correlated with the degree of metabolic disturbance.