Comparison of 10 Different Magnetic Resonance Perfusion Imaging Processing Methods in Acute Ischemic Stroke

Background and Purpose— Several methods are available to assess the magnetic resonance perfusion lesion in acute ischemic stroke. We tested 10 of these to compare perfusion lesion sizes and to assess the relation to clinical scores and final infarct extent. Methods— We recruited patients with acute ischemic stroke, performed diffusion- and perfusion-weighted imaging, and recorded stroke severity at baseline, final infarct size on T2-weighted imaging at ≥1 month, and Rankin Scale score at 3 months. We calculated 10 perfusion parameters (6 of mean transit time, MTT; 3 of cerebral blood flow; 1 of cerebral blood volume; 7 relative and 3 quantitative), measured the perfusion-weighted imaging lesion and diffusion/perfusion mismatch volumes, and compared each with clinical and radiologic outcomes. Results— Among 32 patients, the median perfusion lesion volume varied from 0 to 14 882 voxels ( P <0.0001); the proportion of patients with mismatch varied from 9% to 72% ( P <0.05), depending on the perfusion parameter. Five measures of relative MTT were associated with baseline National Institutes of Health Stroke Scale score; 1 (arrival time fitted) was also associated with clinical outcome. Final infarct size was most strongly associated with MTT measures, including arrival time fitted. There was no advantage of quantitative perfusion measures and no relation between mismatch presence/absence and infarct expansion with any of the 10 perfusion measures. Conclusions— Perfusion lesion size differs markedly depending on the parameter calculated. Relative perfusion parameters performed as well as quantitative ones. Some parameters (mainly representing MTT measures) were correlated with clinical scores; others were correlated with final infarct size; and arrival time fitted was correlated with both. These findings should be validated in other datasets. A consensus is required on which perfusion measurement and processing methods should be used.