Stroke is a leading cause of death, along with cancer and coronary heart disease, and is the most common cause of physical disabilities in adults. Conventional therapies for ischemic stroke include thrombolytic therapy, prevention of inappropriate coagulation and thrombosis, and surgery to repair vascular abnormalities. Over 10 years have passed since the US Food and Drug Administration approved intravenous tissue plasminogen activator (tPA) for use in acute stroke patients. Most major stroke trials have failed during the last 2 decades, including the following:

1) Large clinical trials of secondary prevention (enrolling up to tens of thousands of patients in each study), such as the Warfarin-Aspirin Symptomatic Intracranial Disease Study,1 the Management of Atherothrombosis with Clopidogrel in High-Risk Patients with Recent Transient Ischemic Attack or Ischemic Stroke, and Clopidogrel for High Atherothrombotic Risk and Ischemic Stabilization, Management and Avoidance studies.2

2) Recent randomized trials of new thrombolysis agents, such as the Desmoteplase In Acute Ischemic Stroke (DIAS)-II and Abciximab in Emergent Stroke Treatment-II trials.3

3) Stroke Therapy Academic Industry Roundtable criteria-guided Neuroprotection Trial (NXY-059).4

These results indicate the need for measuring patient heterogeneity in stroke (i.e., individualization of the patient).5 Unlike coronary heart disease, strokes are caused by numerous etiologies, including large-artery atherosclerosis, cardioembolism, and lacunar stroke. Neuroimaging techniques now allow changes to be observed in patients from the acute to the recovery phase. The role of MRI in stroke evaluation and treatment is discussed herein.

In a prospective, multicenter study, pretreatment MRI could be used to differentiate between subgroups of stroke patients likely to benefit from reperfusion therapies given 3-6 h after stroke onset.27 The study also provided a simple but valuable MRI categorization strategy with profiles that are strongly associated with clinical outcomes after reperfusion therapy (i.e., mismatched vs. matched vs. small vs. malignant). Using 30-day fluid-attenuated inversion recovery (FLAIR) images, the Diffusion-weighted Imaging Evaluation For Understanding Stroke Evolution (DEFUSE) trial group reported a significant association between recanalization and reduced infarct growth in patients with mismatch who were treated with tPA within 3-6 hours after stroke onset.28 Patients who exhibited a target mismatch pattern had a favorable clinical response to recanalization therapy (Fig. 1). However, diffusion-perfusion mismatch is only one contributor to tissue fate, even when recanalization is complete.29

Two variables characterize the extent of hypoperfusion: the volume of hypoperfused tissue and the intensity or severity of hypoperfusion within these regions.27 Most studies have focused on the extent of hypoperfusion. The hemodynamic effects of the collateral circulation are important in maintaining perfusion in the penumbral regions.30 Using pretreatment angiographic and MRI data in acute stroke patients, we have shown that pretreatment collaterals may influence the severity of ischemic injury over the hypoperfused region.29 Patients with good collaterals had larger areas of mildly hypoperfused tissue than those with poor collaterals,29 and infarct growth within the penumbral zone was smaller when collaterals were better, irrespective of the degree of recanalization.29,31

The collateral supply can be visualized using a dedicated MRI method (subtracting the image of the first movement map).32 We found that, compared to the large cortical DWI pattern, the deep-infarcts pattern exhibited less severe hypoperfusion related to good collateral flows.33 In addition, the perfusion status may largely depend on the stroke subtype; stroke patients with large intracranial atherosclerosis had different mismatch profiles, which were related to better collaterals, compared to other subtypes.34

The penumbra area is defined as a region of hypoperfusion but with some remaining metabolic activity. Thus, metabolic conditions such as the oxygen extraction fraction (OEF) and the cerebral metabolic rate of oxygen also influence tissue fate after recanalization. Metabolic conditions in the ischemic penumbra may be predicted more accurately than simple diffusion-perfusion mismatch (which frequently overestimates the final lesion). They can enable visualization of the area of stage II hemodynamic failure (increased OEF).35 Although PET measurement of cerebral blood volume (CBV), CBF, OEF, and regional cerebral metabolic rate of oxygen can identify stage II hemodynamic failure, its clinical use is limited, especially in acute stroke settings. MRI can detect misery perfusion (increased brain OEF in an area with reduced CBF but preserved oxygen metabolism).36,37 T2*-weighted gradient-echo (GRE) imaging can also assess brain tissue viability.36,38-40 GRE imaging is extremely sensitive to magnetic field inhomogeneities, since acute decreases in GRE imaging occur in animal models and stroke patients due to blood deoxygenation and increased OEF.36,41,42 The paramagnetic effect of deoxyhemoglobin produces blood-oxygen-level-dependent (BOLD) contrast.43 OEF values from MRI and PET are generally consistent.39 By applying either quantitative BOLD imaging36,39,44 or directly demonstrating intravascular deoxygenation changes on T2*-weighted GRE imaging,40,45 increased OEF can be measured as an increase in deoxyhemoglobin. These measures offer an estimation of the oxygen utilization and provide additional information concerning the metabolic state of the threatened brain.

Fig. 2 shows a case with prominent hypointense leptomeningeal vessels. Although mismatch areas showed an extensive, severe delay in time-dependent PWI [time to peak after deconvolution (T_max) delayed by 8 s or more], suggesting an ischemic core, increased CBV and GRE-hypointense vessels within the hypoperfused area suggested the presence of viable tissue.36,39 CBV was increased in these areas, reflecting an efficient collateral blood flow soon after arterial occlusion, which may indicate the ability of ischemic tissue to compensate for a decreased CBF.40 Hypointense vessels were correlated with a larger perfusion defect and a larger perfusion delay, but an increased CBV.45 These hypointense areas on GRE imaging became infarcted during the subacute and chronic phases.36,46 Brain areas with hypointense signaling on GRE imaging may be targets for acute therapy to improve compromised cerebral circulation.36 However, further validation of GRE imaging for assessing brain tissue viability is necessary, and it is limited by susceptibility artifacts.38

No parameters among CBV, CBF, mean transit time (MTT, first movement of the signal intensity curve), time to peak (TTP), and time to maximum tissue residue function (T_max, time to peak after deconvolution), is superior in predicting tissue fate after recanalization. The size of a perfusion lesion differs markedly depending on which of the ten PWI postprocessing methods are used.47 The time-domain perfusion parameters (e.g., MTT, TTP, and T_max) are being used increasingly in clinical practice, but may have considerable drawbacks in certain situations. For example, PWI may yield inconsistent data from time-dependent and nondependent domains; patients with severe delays in perfusion may a have preserved CBV (Fig. 2). Prominent dispersion and CBV preservation or augmentation might sustain the ischemic regions. A multiparameter approach has been suggested to help define the PWI abnormalities, by combining data from relative CBF, regional CBV, and time-domain maps.48 In certain clinical settings of acute ischemic stroke with prominent collaterals at the time of PWI acquisition (such as populations in which intracranial occlusive disease is prevalent), time-domain PWI parameters may overestimate the perfusion severity and extent and should be interpreted with caution; a non-time-domain PWI parameter may be needed.48

Patients who do not exhibit mismatch patterns are unlikely to have a good clinical response to recanalization therapy, because it has been shown that in the presence of reperfusion, an increasing mismatch ratio was associated with a higher response rate (Fig. 3).49 The DEFUSE trialists exhibited the current mismatch ratio of 1.2 to 1.8-2.6, although this may reduce the number of patients with mismatch (eligible for reperfusion therapy) by 15-25%.49 Overestimation of the penumbra area may have contributed to the failure of the recent reported DIAS-II trial to document benefits of recanalization therapy in mismatch patients.50 Thus, reperfusion in patients with a mismatch ratio below these cutoffs may have no beneficial effects, and may in fact be harmful. Further studies are needed to determine the optimal definition of a target mismatch pattern, as well as the mismatch ratio, PWI parameter, and each threshold.26,50

Hemorrhagic transformation (HT)-which is a much-feared complication of recanalization therapy for acute ischemic stroke-and hemorrhagic stroke can be visualized with high sensitivity and specificity. GRE imaging accentuates the paramagnetic properties of blood products such as deoxyhemoglobin, intracellular methemoglobin, and hemosiderin, and can detect hemorrhage and intravascular clots. A prospective multicenter trial has shown that GRE imaging is as accurate as CT at detecting acute hemorrhage in patients with acute stroke.51

Fig. 4 shows a case with HT, and demonstrates the relationship between the severity of tissue damage and the development of subsequent HT. A recent multicenter study found that the risk of symptomatic intracranial hemorrhage after thrombolytic therapy increased with increased DWI lesion volume.52 The DEFUSE trial defined this malignant profile as extensive severe hypoperfusion with a large DWI lesion volume.27 In that study, early reperfusion was associated with fatal intracranial hemorrhage in patients with a malignant profile.

There have been numerous efforts to predict HT using MRI. Diffusion-perfusion characterization of the ischemic territory may help identify patients at increased risk for HT after recanalization therapy: DWI lesion volume,52 apparent diffusion coefficient (ADC),53 and degree of hypoperfusion.27,54 Additional MRI parameters may also reflect an increased risk for HT after thrombolysis: leukoaraiosis,55 cerebral microbleeds, early parenchymal enhancement,56 and early cerebrospinal fluid hyperintensity.57

BBB permeability dysfunction often precedes HT. Gadolinium (Gd) contrast agents are routinely used to detect BBB opening in patients with strokes or tumors, although routine Gd enhancement does not detect modest changes in BBB permeability. Suboptimal delivery of the contrast agent to the affected region due to a lowered CBF may result in these subtle changes in BBB permeability being missed.58 Permeability assessment methods have shown permeability derangement in patients with ischemic stroke or malignant brain tumors,59,60 and recently with ischemic stroke that could not be demonstrated by simple postcontrast spin-echo imaging alone.61,62 We have applied a novel MRI permeability assessment technique that employs perfusion scan source data to assess BBB dysfunction in acute ischemic stroke and evaluate the association between permeability changes and subsequent HT after recanalization therapy.62 In this study, some degree of HT occurred in 12 of 32 patients. Permeability image abnormalities at baseline were present in 7 of 12 patients with HT and in none of 20 patients without HT on follow-up images. These preliminary data suggest that permeability images derived from pretreatment perfusion MRI data identify patients at risk of HT with high specificity. MRI guidance of interventional decision-making may be improved by supplementing standard penumbral imaging (indicating the potential to benefit from recanalization) with the use of slope imaging (which identifies individuals with BBB disruption who are at high risk of harmful treatment effects due to HT). Excluding patients at high risk of HT may improve the safety profile and risk: benefit ratio.

Permeability derangement is a dynamic process associated with ischemic stroke pathophysiology and recanalization therapy. Although disruption of the BBB is a necessary (albeit not sufficient) condition for intracerebral hemorrhage, it can be benign in certain conditions.63,64 Even if BBB leakage is present, the patterns and destiny of permeability derangement vary. Moreover, predictors of permeability derangement may be different pre- and posttreatment.64

Finally, the clinical features and lesion site are also important; location may be more important than the volume of salvageable tissue. Patients with a similar infarct volume may show variable severity of neurologic deficits; some patients may have lesions in critical regions such as the corticospinal area, versus less critical areas such as the association cortex (e.g., the case presented in Fig. 3). The latter patients might not be candidates for potentially harmful recanalization therapy. Dr. Lev indicated the importance of "location-weight" scoring over simple volumetric data in penumbra areas.65 Locating the mismatch before treatment may help to predict the potential benefits of reperfusion.66

Multimodal MRI can identify the age and vascular territories involved in infarcts (Fig. 5). In most centers, the diagnosis of ischemic stroke subtype is performed during the first few days or weeks after hospitalization. Multimodal MRI, including DWI and magnetic resonance angiography (MRA), performed within a few hours of arrival at the hospital allows the rapid and accurate identification of early ischemic subtypes.67,68 In particular, a cardioembolism can easily be established when patients exhibit acute lesions in multiple vascular territories, enabling the physician to consider early anti-coagulant use.67

The infarct pattern on DWI is correlated with the pathogenic mechanisms underlying the stroke and may predict stroke recurrence and outcome.69 Small acute lesions in multiple vascular beds on DWI provide insight into the stroke mechanism by predicting the proximal source of the embolism.70,71 In addition, the ADC may be useful for estimating the lesion age and distinguishing acute from subacute DWI lesions.72 Acute ischemic lesions can be divided into hyperacute lesions (low ADC and DWI-positive) and subacute lesions (normalized ADC). Chronic lesions can be differentiated from acute lesions by normalization of ADC and DWI. The presence of multiple DWI lesions of varying ages suggests active early recurrences over time and portends a higher early risk of future ischemic events.73

Multimodal MRI may also indicate stroke mechanisms. Silent cerebral microbleeds that are visible on T2*-weighted imaging may be seen as corticomeningeal microbleeds in patients with infective endocarditis,74 enabling the provision of appropriate antimicrobial therapy and avoiding antithrombotic treatment, which could cause increased mortality. In contrast, it has been shown that patients with nonbacterial thrombotic endocarditis exhibit multiple disseminated infarctions of varying size, with at least one medium or large lesion on DWI.75 Large thrombi that are fragile (i.e., lacking an inflammatory reaction and with little cellular organization) could underlie this DWI pattern, distinguishing it from infective endocarditis or cardioembolic stroke.75 Finally, patent foramen ovale is a potential risk factor for ischemic stroke and shows different infarct patterns on DWI. Patients with patent foramen ovale showed embolic infarcts, especially multiple lesions in the posterior circulation,76 which may reflect the finding of higher blood flow in the posterior circulation than the anterior circulation during the Valsalva maneuver.77

Small-vessel disease resulting from hypertension is the most common cause of stroke, and has characteristic clinical features and a good prognosis. DWI allows the detection of silent infarctions at different sites from the symptomatic, small, deep infarction, and concomitant small lesions outside the striatocapsular distribution could be identified. We have reported that proximal middle cerebral artery lesions are a common cause of small deep infarcts,78 and that patients with parental arterial disease (by branch atherosclerosis) are more likely to have recurrent strokes and a poor long-term prognosis.79 These results emphasize the importance of performing vascular studies in intracranial vessels, as indicated in Fig. 6.

The traditional criterion of an infarct size of <15 mm has recently been challenged.80,81 The infarct size in patients with symptomatic small arterial occlusions varied from 3.1 to 38.7 mm.81 A new stroke classification (Scandinavian Stroke Scale-Trial of Org 10172 in Acute Stroke Treatment)82 that used an infarct size criterion of 20 mm rather than 15 mm was designed for the diagnosis of lacunar stroke. The large variation of infarct size may be related to the branching patterns of perforating arteries (Fig. 6); Cho et al.83 recently showed an atypical branch pattern using high-resolution MRA.

It has been shown that recurrent intracranial hemorrhage is more frequent following lacunar vs. nonlacunar infarction.84 Another study has shown that the prevalence of microbleeds was higher in lacunar stroke (62%) than in other infarct subtypes (21-30%),85 which may be attributable to the increased risk of subsequent intracranial hemorrhage. These results suggest the importance of performing GRE imaging to avoid aggressive antithrombotic therapy and possible bleeding complications in patients with multiple microbleeds.

Atherosclerosis is an inflammatory disease. Coronary and carotid histology show that inflammatory cells such as macrophages mediate the development and progression of atherosclerosis. In addition, inflammatory markers such as C-reactive protein are associated with the progression of atherosclerosis and stroke recurrence. Vulnerable plaques typically have a substantial lipid core and a thin fibrous cap. Thrombotic coronary artery occlusion usually follows rupture of an unstable atherosclerotic plaque; the at-risk or vulnerable asymptomatic atherosclerotic coronary artery plaque is not associated with high-grade stenosis. Carotid atherosclerosis has many identical clinical and pathological features to coronary atherosclerosis.

Plaque vulnerability is identified by MRI86 in three ways (Fig. 7): 1) multicontrast MRI, 2) commercially available nonspecific contrast agents, and 3) molecular imaging probes (for details see the review by Briley-Saebo et al.86). Multi-contrast MRI characterizes the key structures (e.g., the lipid core, fibrous cap, and intraplaque hemorrhage) using signal variation after pulse sequences (with T1-, T2-, and protondensity-weighted images being the most commonly used) to identify plaque composition. MRI-based tissue quantification is accurate and reproducible; when compared with carotid endarterectomy specimens, in vivo multicontrast MRI could distinguish advanced lesions from earlier atherosclerotic plaques.87,88 Commercially available nonspecific contrast agents (i.e., Gd-diethylenetriamine penta-acetic acid) can reveal plaque structures that are indicative of vulnerability (e.g., necrotic lipid cores and fibrous caps) and allow the assessment of plaque neovascularization.86 Contrast agents that characterize thrombi are under development.89,90

Molecular imaging probes targeted to biochemical and/or cellular targets (such as macrophages, matrix metalloproteinase) can indicate plaque vulnerability.86 Ultrasmall superparamagnetic particles of iron oxide (USPIO) are the best studied in stroke patients.91,92 USPIO are nanoparticles of iron oxide that can be taken up by macrophages, thus decreasing the MRI signal. Macrophage accumulation in the carotid plaque can be visualized using GRE imaging (as a signal decrease in part of the vessel wall after USPIO administration).91 Both histological analysis and MRI of symptomatic patients who have undergone carotid endarterectomy have shown that macrophage accumulation (shown by USPIO) was more prevalent in the ruptured and rupture-prone lesions than in the stable lesions.91 Areas of reduced signal intensity were observed in 24 out of 27 patients (89%) with symptomatic carotid stenosis.92

Thrombogenicity is related to hemorheology (e.g., high shear stress, oscillatory shear stress, and local stasis) as well as local (i.e., plaque structure) or systemic (e.g., inflammation) conditions.93-97 Clinical studies of intra-aneurysmal hemodynamics have shown that specific flow patterns may be related to risk of aneurysm rupture.98 Similarly, the greatest atherosclerotic plaque accumulation typically occurs on the outer wall of the proximal segment of the sinus of the internal carotid artery, in the region with the lowest wall shear stress.99 MRI can measure wall shear stress vectors in normal subjects or in patients with carotid atherosclerosis.100,101 The shapes and locations of low-shear-stress wall surfaces differ between individuals,100,102 which may explain why strokes recur in some patients and not in others. Although the degree of stenosis is the key factor to consider before intervention, patients with the same degree of stenosis on the carotid bifurcation or intracranial vessels do not have the same rate of progression of stenosis or stroke recurrence. Measurement of the role of flow pattern (i.e., wall shear stress and oscillation) on the development and progression of atherosclerosis using angiographic or noninvasive imaging merits further study.

The evaluation of changes in motor function can be achieved with functional MRI, diffusion tensor imaging, and molecular MRI. This research is outside the scope of the present review.

The American Stroke Association/American Heart Association Stroke Council has issued the statement that multimodal CT and MRI data may improve the diagnosis of ischemic stroke (class I, level A).17 Using MRI techniques to understand individual case pathophysiologies will allow the future development of rational stroke therapies that are tailored to the specifics of each case. Measuring salvageable tissue and permeability derangements on MRI may help select patients for recanalization therapy.104 MRI can also guide decision-making in stroke intervention, such as stenting and carotid endarterectomy, by providing information on plaque characteristics and the rheological aspects of atherosclerotic stenosis.

We are living in an era in which stroke physicians have difficulty keeping up with developments in imaging techniques. Multimodal MRI has significant potential for improving treatments and outcomes in stroke, as well as improving decision-making algorithms and estimates of the NNT or the number needed to harm.