Thrombolytic therapy of acute ischemic stroke can be successful only as long as there is penumbral tissue perfused at rates between the thresholds of normal function and irreversible structural damage, respectively. To determine the proportion of tissue at risk of infarction, cerebral perfusion was studied in 12 patients with acute ischemic stroke who underwent treatment with systemic recombinant tissue plasminogen activator (0.9 mg/kg body weight according to National Institute of Neurological Disorders and Stroke protocol) within 3 hours of onset of symptoms, using [15O]-H2O positron emission tomography (PET) before or during, and repeatedly after thrombolysis. The size of the regions of critically hypoperfused gray matter were identified on the initial PET scans, and changes of perfusion in those areas were related to the clinical course (followed by the National Institutes of Health stroke scale) and to the volume of infarcted gray matter demarcated on magnetic resonance imaging 3 weeks after the stroke. Whereas the initial clinical score was unrelated to the size of the ischemic area, after 3 weeks there was a strong correlation between clinical deficit and volume size of infarcted gray matter (Spearman's rho, 0.96; P < 0.001). All patients with a severely hypoperfused (< 12 mL/100 g/min) gray matter region measuring less than 15 mL on first PET showed full morphologic and clinical recovery (n = 5), whereas those with ischemic areas larger than 20 mL developed infarction and experienced persistent neurologic deficits of varying degree. Infarct sizes, however, were smaller than expected from previous correlative PET and morphologic studies of patients with acute stroke: only 22.7% of the gray matter initially perfused at rates below the conventional threshold of critical ischemia became necrotic. Actually, the percentage of initially ischemic voxels that became reperfused at almost normal levels clearly predicted the degree of clinical improvement achieved within 3 weeks. These sequential blood flow PET studies demonstrate that critically hypoperfused tissue can be preserved by early reperfusion, perhaps related to thrombolytic therapy. The results correspond with experimental findings demonstrating the prevention of large infarcts by early reperfusion to misery perfused but viable tissue.