Systematic Review of Diffusion and Perfusion Imaging in Acute Ischemic Stroke

Abstract

Background and Purpose —Recent advances in neuroimaging have raised hopes of early and accurate identification of ischemic brain and the discrimination of dead from salvageable tissue. We sought to determine whether the data published so far are enough to establish the roles of these techniques in everyday clinical practice. Methods —A systematic review of studies of MR diffusion-weighted imaging (DWI), perfusion imaging (PI), or a combination of the two, in human stroke, excluding abstracts and case reports. One reviewer extracted information on the size of each study, its main purpose, methodological details, and results. Results —We identified 47 studies of DWI, 18 studies of MR PI alone or in combination with another advanced imaging modality, and 19 studies of DWI and PI together. Although high proportions of the studies were prospective and gave good details of the imaging sequences used, the majority gave very limited details on patient selection and clinical characteristics or blinded imaging assessment. Pathophysiological changes were inferred from DWI/PI patterns that were not supported by other data. Conclusions —Despite considerable enthusiasm for and promise of these techniques, there is not sufficient information available in these studies to enable us to draw firm conclusions about the sensitivity and specificity of these techniques for identification of either ischemic lesions not visible by other means or salvageable tissue. Future studies should include larger numbers of carefully described patients, assess the contribution of DWI over and above other imaging, obtain follow-up at an appropriate time interval to determine accurate clinical and neuroradiological outcomes, and assess DWI/PI abnormality with reperfusion in randomized treatment trials. Investigators should also be encouraged to combine their individual patient data in meta-analyses to obtain a more robust assessment of the value of DWI and PI from larger sample sizes.