Nine radiation-treated brain tumor patients were studied by positron emission tomography (PET) in an attempt to differentiate tumor recurrence from radiation necrosis. Rubidium-82 was used to define the region of absent or disturbed blood-brain barrier and [18F]2-fluoro-2-deoxy-D-glucose ([18F]FDG) was used to evaluate the metabolic state of the brain. By comparing glucose utilization in the pathologic region with utilization in the adjacent tissue, a diagnosis of recurrent tumor (increased [18F]FDG accumulation) or necrosis (decreased FDG accumulation) was made. In the seven patients who underwent surgery the PET diagnosis was confirmed by histologic examination of resected tissue. The two patients who did not undergo surgery have had a clinical course consistent with the PET diagnosis of necrosis. Dynamic 82Rb imaging showed that the rate of 82Rb accumulation was greater in tumor than in normal brain. However, this finding alone did not differentiate tumor from necrosis, as some necrotic tissue also showed high rates of 82Rb accumulation, and washout kinetics were similarly nonspecific. The differentiation of radiation necrosis from tumor recurrence is reliably achieved by [18F]FDG PET examination and is aided by information obtained from a 82Rb PET study done immediately prior to the [18F]FDG injection.