Dr. Kmiecik is a fourth-year Radiology Resident at the University of California, San Francisco. He received his MD in 1996 and his PhD in biophysics in 1997, both at the University of Illinois at Urbana-Champaign. He will begin a Neuroradiology fellowship at UCSF after completing his residency.

Brain ischemia is a clinical emergency, and immediate neuroimaging is necessary to confirm stroke and to exclude alternative diagnoses. Magnetic resonance (MR) imaging offers considerable diagnostic and prognostic information, although practical difficulties have limited its role to this point. This review includes a brief summary of the clinical issues of acute cerebral infarction, a description of common MR methods and findings, and the use of MR to direct therapy and to predict clinical outcome.

Brain ischemia is a clinical emergency because neurons die rapidly when deprived of oxygen and metabolic substrates. Although the use of thrombolytic agents has recently become available for the treatment of stroke, this potentially curative therapy must be administered expeditiously to be effective. Moreover, since thrombolytics increase the incidence and severity of intracranial hemorrhage (ICH), this complication must be excluded before these agents are administered. Therefore, immediate neuro-imaging is necessary to confirm the diagnosis of acute cerebral ischemia and to exclude alternative diagnoses for stroke-like clinical syndromes. While computed tomography (CT) remains the standard for imaging patients with acute stroke, magnetic resonance (MR) imaging offers considerably more diagnostic and prognostic information. At this point, however, practical difficulties have limited the role of MR. Recent extensive research into the benefits and practicality of MR in the evaluation of stroke has indicated significant potential for improving stroke management and patient outcome.

This review summarizes the clinical issues of cerebral infarction and describes common MR methods for imaging the condition. It discusses the significance of MR findings and their use for selecting appropriate therapies and predicting clinical outcome. It also outlines proposed protocols for the application of MR in acute stroke.

The clinical impact of stroke

The potential benefit of developing techniques for the rapid diagnosis and treatment of stroke is clear--stroke is the third leading cause of death in the United States and results in an annual cost of $30 billion. ¹ Within 6 months after a stroke, 20% to 30% of patients die, 20% to 30% are moderately to severely disabled, and very few recover without disability. ²

There has been considerable effort recently to elucidate the pathophysiology of stroke and to develop treatments aimed at effectively improving functional outcome. Several major trials have investigated the benefit of early pharmacologic intervention with thrombolytic agents, and recombinant tissue plasminogen activator (rtPA) has been approved by the Food and Drug Administration for the treatment of acute stroke. ^3-8 Because intravenous rtPA thrombolysis has been proven to be beneficial only when administered within 3 hours of the onset of symptoms, rapid clinical diagnosis and imaging confirmation is vital.

The integrity of collateral flow to the ischemic region may be a critical factor that determines the outcome of an ischemic episode. Given this and other variables that influence the outcome of patients with stroke, it is necessary to stratify patients according to the likelihood of successful intervention; MR can accomplish this rapidly and effectively. The unique diagnostic and prognostic information quickly obtainable by new MR imaging techniques promises to have a substantial impact on stroke management.

Brain attack: The pathophysiology of stroke

Distinct pathophysiologic stages of cerebral ischemia can be identified, and each one can be investigated with MR techniques. The cascade begins with a blood flow abnormality, which may be focal, from embolism or in situ thrombosis; or from global hypoperfusion. Autoregulation of cerebral blood flow may compensate for a focal insult in several ways. It may increase cerebrovascular resistance to counteract decreased cerebral perfusion pressure or recruit collateral blood flow from the circle of Willis or from leptomeningeal vessels. It may also compensate by increasing oxygen extraction efficiency. If there are insufficient avenues of collateral flow, the affected tissue can switch from oxidative to glycolytic metabolism. This may lead to lactate accumulation and acidosis. Hypoperfusion below a critical level (approximately 10 to 20 mL/min/100 g tissue), however, causes electrical and cellular dysfunction, with release of cytotoxic levels of glutamate and failure of the Na ⁺ /K ⁺ ATPase. ⁹ Pump failure allows accumulation of sodium, calcium, and water within the cell, a process known as cytotoxic edema. Finally, after approximately 6 hours of ischemia, cellular structural integrity fails, causing the breakdown of the blood-brain barrier and subsequently vasogenic edema.

Physical principles of MR stroke imaging techniques

Several complementary MR techniques are useful in the evaluation of the various stages of acute stroke. A brief discussion of these methods follows.

Diffusion-weighted imaging

Diffusion-weighted imaging (DWI) derives image contrast from the random, microscopic (Brownian) motion of water molecules in tissue. The DWI pulse sequence includes magnetic field gradients before and after a 180° radiofrequency (RF) pulse, and image readout is generally accomplished with a rapid echoplanar imaging (EPI) technique. During the first gradient period, both static and moving spins acquire a phase shift. The second diffusion-weighting gradient is identical to the first, but is effectively reversed because it follows the 180° RF pulse. For stationary spins, the net phase shift after the two diffusion gradient pulses is zero. For spins moving in the direction of the diffusion gradient during the time interval between the two gradient pulses, the dephasing caused by the first diffusion gradient pulse is not reversed by the second pulse, and thus moving spins are suppressed. Diffusion-weighted images therefore show higher signal intensity for spins in which diffusion is reduced, whereas freely diffusing spins are lower in signal intensity.

By varying the direction of the diffusion gradients over repeated acquisitions, diffusion anisotropy effects are minimized by averaging. By varying the magnitude of diffusion gradients, an apparent diffusion coefficient (ADC) can be calculated for each voxel--with high ADC values indicating freely diffusible spins, and low values signifying spins with reduced diffusion. One potential pitfall of DWI is that the EPI readout in common diffusion sequences introduces inherent T2-weighting, and increased signal in DWI may be due to "T2 shine-through" rather than to true reduced diffusion. Calculating an ADC map, however, eliminates this confounding effect (Figure 1).

Ischemic and infarcted tissue is associated with restricted motion of intracellular and/or extracellular water molecules causing increased signal on diffusion-weighted images. This finding by itself is nonspecific because other pathologies, such as abscesses and certain tumors, may show reduction of water diffusion. In the proper clinical setting, however, the diagnostic accuracy of diffusion imaging is quite high, with sensitivity and specificity for acute stroke approaching 100%. ¹⁰ Normal DWI in the face of clinical symptoms suggestive of acute stroke should provoke consideration of alternative diagnoses such as transient ischemic attack, migraine, seizure, ICH, or metabolic disorder. ¹¹

Perfusion-weighted imaging

Critically decreased regional perfusion of brain tissue is the inciting physiologic abnormality leading to infarction and, therefore, visualization of this deficit with MR perfusion-weighted imaging (PWI) is highly valuable. Although there are noninvasive MR perfusion techniques that use MR spin tagging of inflowing blood ("arterial spin-labeling") and obviate the need for an exogenous contrast agent, the most common PWI technique uses a bolus injection of 0.1 to 0.2 mmol/kg gadolinium and T2* susceptibility-weighted EPI during its first pass. ¹² Images per slice can be acquired in about 50 to 100 milliseconds. The passage of paramagnetic gadolinium chelate through the image slice causes a transient and focal decrease of signal. The amount of signal loss measured is proportional to the concentration of contrast agent within the microvasculature. With the arterial input function, the following information can be calculated: relative cerebral blood flow (rCBF, volume of contrast agent to the microvasculature per unit time); mean transit time (MTT, time for bolus of contrast agent to travel through the microvasculature); and relative cerebral blood volume (rCBV, volume of contrast agent in the microvasculature during first pass). ^13,14 Time to peak (TTP) of contrast concentration in the imaging slice can be determined directly from the images. Possible pathophysiologic implications of abnormalities of these parameters will be discussed later.

Magnetic resonance angiography

Magnetic resonance angiography (MRA) evaluates the luminal anatomy of blood vessels by visualizing the spins in flowing blood. This can be accomplished with a number of techniques, the most common of which is two-dimensional (2D) or three-dimensional (3D) time of flight (TOF). With these sequences, a selected imaging volume is repeatedly excited with a train of rapid RF pulses that saturates the signal from stationary tissue; the signal from inflowing blood, which has not received the pulse train, is bright (Figure 2). Because of the saturation effect, slowly forming blood can be poorly visualized with 3D TOF, and 2D TOF may show stepladder artifacts on maximum-intensity projections. A combination technique called multiple overlapping thin-slab acquisitions (MOTSA) improves on these limitations. Using an intravascular contrast agent can decrease acquisition time angiographic effect by reducing the saturation of intravascular spins.

Magnetic resonance spectroscopy

The magnetic resonance spectros-copy (MRS) experiment is based on principles similar to those of MR imaging. However, the data acquired are resonance peaks in which signal frequencies depend on the local magnetic microenvironment and thus differ for each compound, and peak areas are directly related to metabolite concentrations. Spectra can be obtained from spatially limited volumes of interest using one of a variety of single-volume-element (single-voxel) sequences or 2D or 3D spectroscopic imaging techniques. Proton ( ¹ H) spectroscopy is most common, but other nuclei, such as phosphorus-31 ( ³¹ P), are MRS-visible but suffer from poor signal-to-noise ratio. Metabolites frequently identified in ¹ H spectra of healthy brain tissue include cholines, creatines, and N-acetylaspartate (NAA). Decreased NAA can be seen within hours of stroke onset, presumably indicating neuronal death. Increased lactate can often be detected in ischemic brain areas, indicating glycolytic metabolism, or, when chronic, the presence of macrophages and leukocytes (Figure 3). ¹⁵ Numerous amino acids, which may be globally or focally increased in a variety of diseases, can be detected in ¹ H spectra obtained with short echo times. With ³¹ P spectroscopy, pH can be determined readily, and ischemic areas are often acutely acidotic. ¹⁶ The use of MRS in acute stroke has not yet been proven in widespread studies. However, MRS has been shown to be helpful in early reports of neonatal ischemia, which is often difficult to assess on anatomic MR imaging studies. ¹⁷

MR findings in acute stroke

The MR imaging findings in the first 24 hours of stroke are summarized in Table 1. Due to its high sensitivity for ischemia and speed in acquisition, DWI is the current workhorse for MR imaging of acute stroke. Diffusion-weighted MR images show prompt increase in the ischemic tissue signal, even within minutes of vessel occlusion. This finding is thought to reflect primarily cytotoxic edema, with shift of extracellular water into the cell, where diffusion is constrained. Additional mechanisms for reduced diffusion, such as increased tortuosity of the extracellular compartment and increased viscosity of the intracellular compartment, have also been proposed. ¹¹

The temporal evolution of the diffusion abnormality in territorial stroke has been shown to follow a reproducible time course. After an initial low ADC in the ischemic lesion, the ADC increases gradually until about a week after symptom onset, at which point it starts to pseudonormalize. Then it remains supranormal. ¹⁸ The observed pattern has been described to represent initial reduced diffusion due to cytotoxic edema, then increasing ADC secondary to subsequent membrane disruption and vasogenic edema, and, finally, elevated ADC indefinitely due to gliosis and further increase in lesion water content. ¹¹

Without early intervention, ischemic tissue showing severe early ADC decreases has been shown to proceed almost invariably to infarction, and this tissue is considered by many to represent an ischemic core. However, surrounding tissue showing more modest diffusion abnormalities may be potentially viable. ¹⁹ In fact, with early recanalization, even severe ADC reductions have been shown to be reversible in both animal models ²⁰ and patients with stroke. ²¹

This so-called "ischemic penumbra" of potentially viable tissue, which surrounds a core of tissue doomed to infarct, is seen in a majority (65% to 95%) of patients with stroke who were imaged within 24 hours of symptom onset. ^22-24 This concept has been the subject of numerous recent studies comparing perfusion and diffusion abnormalities in acute stroke. ^25-29

Within the core of the lesion defined by a severe diffusion abnormality, perfusion imaging generally shows increased relative MTT and significantly decreased rCBV and rCBF. ¹⁴ Very often, the volume of the perfusion abnormalities on PWI is larger than the diffusion abnormality on DWI. While this lesion volume mismatch between DWI and PWI may represent peripheral ischemic, it could very well be viable tissue, and, as such, has the potential of being saved with early therapeutic intervention and recanalization. In untreated patients, the presence of a perfusion-diffusion mismatch in initial imaging has been shown to correlate with subsequent lesion growth. ³⁰ In contrast, if the DWI and PWI lesion volumes are similar, there is generally no significant increase of lesion size. ¹¹ Increased rCBV in the penumbra region ^13,14 may represent reactive vasodilation, while prolonged MTT and TTP may represent indirect blood flow via collateral pathways (Figure 4). ⁹

Patient selection for thrombolytic therapy

Potentially, the identification of a DWI-PWI mismatch can be exploited to select a subset of patients with stroke who would most likely benefit from thrombolysis (Figure 5), possibly even after the currently accepted time window for rtPA has expired. The benefit of expanding the window for thrombolytic therapy is substantial; the most common reason for intravenous rtPA ineligibility is patient presentation beyond 3 hours from symptom onset. ²⁹ Indeed, studies have indicated that the penumbra region may remain viable up to, and even beyond, 48 hours. The viability may also vary among individuals due to differences in location of ischemic insult and the adequacy of collateral flow. ^31,32 Conversely, MR DWI/PWI imaging may identify those patients with matched lesions for whom thrombolytic therapy is unlikely to be effective, sparing them the attendant risks. Verification of the existence of potentially viable tissue could also be useful in selecting patients who may benefit from other therapeutic measures, such as angioplasty, or promising neuroprotective agents, such as dichloroacetate. ³³

MR angiography and MRS evaluation may also prove helpful in selecting patients for whom immediate treatment is indicated. Although patients with proximal middle cerebral artery (MCA) branch occlusions often recanalize with thrombolysis, ²⁹ it has been shown that patients with distal internal carotid artery (ICA) occlusions are less likely to benefit from thrombolytic intervention. ³⁴ Additionally, normal MRA in the setting of acute territorial stroke suggests spontaneous recanalization, in which case thrombolytic therapy would be superfluous and potentially dangerous. In a study evaluating the ability of ¹ H MRS to predict infarct growth, expansion of the initial diffusion abnormality was seen in patients in whom peri-infarct lactate/choline ratios were increased. ³⁵ This finding suggests that elevated lactate may indicate tissue at risk to infarct, for which thrombolytic therapy may be indicated.

Prediction of functional outcome

Whether or not thrombolytic therapy is administered, it is important to obtain timely information about the likely outcome of the stroke patient. There is a strong positive correlation of neurologic deficit with volume of the initial DWI lesion, as well as with the severity of the initial ADC abnormality. ³⁶ Lesion volume at early DWI was combined into a single indicator using two clinical measures of stroke severity (the National Institutes of Health stroke scale score, a measure of neurologic deficit, and the Barthel Index, an assessment of the ability to perform 10 specific activities of daily life) and then calculated. The resultant measure correlated with final functional outcome better than any of the individual components. ³⁷

Several groups have reported that mismatches of volume between the early DWI lesion and the lesions on rCBV and MTT maps correlate with growth in DWI lesion volume on follow-up studies. ²⁵ This finding suggests that although cerebral vascular autoregulation may preserve viability in the penumbra region early in the process, compensation is tenuous, and eventually will fail and lead to infarct. Also, although DWI and PWI lesion volumes at early imaging correlate individually with clinical outcome, the size of the mismatch between DWI and rCBV was best at predicting lesion growth and functional outcome.

Since both critical hypoperfusion and benign oligemia may create the perfusion abnormality, correlations of perfusion abnormalities with functional outcome may be improved if thresholds of severity are used to define the spatial extent of the lesion. For example, the mismatch volume between the DWI abnormality and a PWI abnormality defined by a mild prolongation of MTT or TTP does not correlate as strongly to final infarct volume as it does when defined by a more pronounced delay. ²⁶ Pixel-by-pixel analyses, more robust than simple volumetric thresholding, are also being investigated to account for the occasional appearance of PWI lesion heterogeneity. ²⁷

Metabolite concentrations as measured by MRS have also been shown to correlate with final infarct size and with the degree of neurologic impairment. Although the concentrations of both NAA and lactate within the ischemic core at initial imaging have been shown to be independently linked to functional outcome, several studies suggest that there is a stronger correlation between lactate and functional outcome. ^38,39 Lactate levels in the chronic state continued to correlate with functional status, P <0.01. ³⁵

Intravascular enhancement due to slow flow is an early sign of ischemia. However, the initial area of intravascular enhancement does not seem to predict final infarct size, even if the area of initial intravascular enhancement is larger than that of the initial DWI abnormality. ⁴⁰ Nonetheless, a tendency toward poor clinical condition in patients showing the hyperintense vessel sign within the full MCA territory on fluid-attenuated inversion recovery imaging has been described. ⁴¹ Additionally, the absence of MCA flow on MRA in patients with DWI-PWI mismatch may serve as an independent predictor of DWI lesion expansion and worse clinical outcome. ²²

Evaluation of future risk for stroke

Stratification of relative risk of future ischemic events in patients after initial stroke, or in patients without prior stroke but with flow-limiting carotid disease, would be beneficial in directing therapy. A multifunctional MR imaging study (MR imaging, MRA, MRS) evaluated 115 patients with angiographically confirmed ICA occlusion presenting with transient or mild ischemic symptoms and with 4-year follow-up. A low ratio of NAA/choline in the white matter of the ipsilateral centrum semiovale was associated with a yearly risk of recurrent ischemic event of nearly 4 times that of patients with normal ratios. This finding suggests a possible increased benefit of external-internal carotid bypass in this subset of patients. ⁴²

In studies in the United States, Japan, and Europe, MR evidence of clinically silent prior infarcts was found in approximately 25% of thousands of elderly volunteers without a history of stroke, and was associated with hypertension. ^43-45 Annual incidence of stroke was 1.5 to 13 times higher in patients with silent infarcts than in those without. This association remained even after adjusting for age, sex, and other (multiple) risk factors for stroke. These findings may eventually lead to MR screening in certain high-risk patient populations.

Investigating stroke etiologies

Because advanced MR imaging techniques are fast and repeatable, they can be very effective in elucidating the pathophysiologic mechanisms of secondary stroke. In subarachnoid hemorrhage (SAH), acute or delayed ischemia secondary to vasospasm is an important cause of morbidity and mortality. ⁴⁶ The natural history of ischemia secondary to vasospasm differs from that caused by occlusive disease, because vasospasm represents a protracted low-flow state with eventual reperfusion. In an animal model of acute SAH in which the circle of Willis of anesthetized rats was perforated during repeated DWI acquisitions, there was rapid decline in ADC in the ipsilateral somatosensory cortex in the distribution of the distal MCA. This finding suggests acute vasospasm as the likely etiology, with the time course and spatial extent of the ADC abnormalities related to the severity of SAH. ⁴⁷

In serial DWI studies of patients with SAH, all patients with vasospasm suggested by transcranial Doppler ultrasound had associated DWI abnormalities, despite the fact that some patients were asymptomatic. However, DWI was normal in patients with SAH without vasospasm. ⁴⁸ In some patients with vasospasm, the associated ADC decreases were reversible.

In a retrospective study of posttraumatic cerebral infarction, which complicates 2% to 10% of head-trauma cases, focal mass effect or brain herniation was the most common cause. ⁴⁹ The investigators noted, however, that because the angiography was not performed in these cases, the impact of vasospasm could not be evaluated. Also, outcomes were worse in the patients with posttraumatic SAH than in those without.

Feasibility of MR imaging for acute stroke

The critical time constraints for identifying patients with ischemic stroke who could benefit from therapy demand that the imaging protocol be both accurate and efficient. Additionally, the test must also rule out hemorrhage to prevent the administration of thrombolytics or anticoagulants in these patients. Ideally, an imaging protocol for acute stroke must exclude ICH, verify the presence of ischemic disease, and take no more than 20 minutes to perform.

If MR is to replace CT as the standard of care for imaging acute stroke, its superiority must be definitively established. In a direct comparison between MR and CT performed within 6 hours of onset in patients with clinically suspected acute hemispheric stroke, DWI/MRA was found to be superior to CT in diagnosing acute infarction, revealing occluded MCAs, and disclosing subtle hemorrhage. Additionally, MR was found to be more accurate than CT in revealing infarction in >33% of the MCA territory, whereas intravenous rtPA is relatively contraindicated. ⁵⁰

The identification of acute ICH has been the domain of CT, although the sensitivity of MR for acute hemorrhage has been questioned. The sensitivity of MR imaging for hyperacute ICH was investigated in a study of 5 patients with hyperacute-to-acute ICH. Susceptibility-weighted EPI identified ICH in all 5 patients as early as 23 minutes after symptom onset. ⁵¹ Similar sensitivity has been reported by other investigators. ^52,53

An additional concern that conventionally favors CT over MR concerns efficiency as it relates to the time between patient arrival and imaging completion. However, the results of a 2-year study of this issue showed that MR was competitive with CT in this regard, at least in larger institutions with stroke centers. ⁵⁴ The availability of MR imaging on an emergent basis in smaller facilities may be more limited. Considering the duration of the MR examination, Sunshine and colleagues ²⁸ have detailed a protocol for the emergency MR evaluation of acute stroke that is performed routinely in <15 minutes. The sequences include sagittal T1 scout, axial T2 turbo gradient and spin echo (GRASE), axial EPI DWI, and gadolinium-bolus PWI, which is routinely performed in <15 minutes. Difusion-weighted and perfusion-weighted images are postprocessed automatically, with ADC and TTP maps immediately available at the console.

Other considerations for the application of MR to acute stroke are body movements in the acutely ill patient (who is often uncooperative), patient monitoring that requires specialized MR-compatible equipment, and the unique requirement to screen for contraindications to the MR examination, such as metal fragments and MR-incompatible implants. More research is needed to determine the practical significance of these issues.

Future directions

Early reports of the use of novel MR techniques in the diagnosis and treatment of various brain pathologies indicate the increasing use of this modality for evaluating and managing stroke. Preliminary research with a MR contrast agent, which was created by conjugating iron oxide particles with a protein known to bind to membrane phospholipids of apoptotic cells, has demonstrated the ability of an MR technique, used in both in vitro and tumor models, to detect cells destined to die. This type of research may lead to a specific early MR indicator of brain tissue certain to infarct in stroke. ⁵⁵ Significant research is also under way on the use of MR in guiding neurointerventional procedures directly, such as the coil embolization of aneurysms. ⁵⁶ In stroke, such a system could provide real-time monitoring of the patency of occluded vessels, as well as territorial perfusion during thrombolysis or angioplasty.

Conclusion

By providing superior anatomic and physiometabolic information, MR imaging clearly offers substantial benefit for rapidly and accurately evaluating the extent, severity, etiology, and chronicity of cerebral ischemia. These capabilities promise to improve the clinical management and outcome for patients with strokes. These capabilities are compelling motives for increasing the use of MR imaging in acute stroke.