BACKGROUND: The ability to quickly and efficiently identify the ischemic penumbra in the acute stroke clinical setting is an important goal for stroke researchers and clinicians. Early and accurate identification of potentially salvageable versus irreversibly infarcted brain tissue may enable selection of the most appropriate candidates for early stroke therapies and identify patients who may still benefit from late recanalization or neuroprotective treatment. Recent advances in magnetic resonance imaging of the ischemic penumbra have been driven by serial MRI studies characterizing the natural evolution of cerebral infarction as well as the brain's response to reperfusion. SUMMARY OF COMMENT: Based on these studies, various models for imaging the penumbra with MRI have been proposed, including the pioneering diffusion-perfusion mismatch model and later multivariate approaches. Each model has its own unique advantages and disadvantages. CONCLUSIONS: There now are sufficient data to support paradigm shifts in a variety of central tenets regarding MRI and the ischemic penumbra. These include the insights that diffusion-perfusion mismatch does not optimally define the penumbra; that early diffusion lesions are in part reversible and often include both irreversibly infarcted tissue and penumbra; that the visible zone of perfusion abnormality overestimates the penumbra by including regions of benign oligemia; that MRI is a very practical method for acute stroke imaging; and that therapeutic salvage of the ischemic penumbra has been demonstrated in humans using diffusion-perfusion MRI.