| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|
|
|
|||||||
|
|||||||||
State of the Art |
1 From the Morgan H. Russell Department of Radiology, Division of Neuroradiology, the Johns Hopkins Hospital, 600 N Wolfe St, Baltimore, MD 21287. Received April 16, 1998; revision requested June 24; final revision received January 13, 1999; accepted March 16. Address reprint requests to N.J.B. (e-mail: nbeauch@welchlink.welch.jhu.edu).
Abstract
Until recently, there was no efficacious treatment for acute cerebral ischemia. As a result, the role of neuroimaging and the radiologist was peripheral in the diagnosis and management of this disease. The demonstration of efficacy using thrombolysis has redefined this role, with the success of intervention becoming increasingly dependent on timely imaging and accurate interpretation. The potential benefits of intervention have only begun to be realized. In this State-of-the-Art review of imaging of acute stroke, the role of imaging in the current and future management of stroke is presented. The role of computed tomography is emphasized in that it is currently the most utilized technique, and its value has been demonstrated in prospective clinical trials. Magnetic resonance techniques are equally emphasized in that they have the potential to provide a single modality evaluation of tissue viability and vessel patency in an increasingly rapid evaluation.
Index terms: Brain, infarction, 10.781 • Brain, ischemia, 10.781 • Cerebral blood vessels, CT, 174.1211 • Cerebral blood vessels, flow dynamics, 174.76 • Cerebral blood vessels, MR, 174.12141, 174.12142, 174.12145 • State-of-art reviews
Stroke is an injury to the central nervous system that is characteristically abrupt in onset and due to a vascular insult. The term is reflective of damage to the brain secondary to ischemia or hemorrhage. It is the number three cause of mortality and the number one cause of disability in adults in the United States. Strokes are ischemic approximately 80% of the time, and until recently, there was no available beneficial intervention (1,2). In 1995, the published results of the National Institute for Neurological Diseases and Stroke (NINDS) recombinant tissue plasminogen activator (rt-PA) trial represented the first demonstration of efficacious treatment for acute cerebral ischemia (3). This has redefined the role of the radiologist and neuroimaging from peripheral to central in the management of acute cerebral ischemia. This State of the Art describes the central role of the radiologist and neuroimaging in the diagnosis and subsequent management of acute cerebral ischemia.
HISTORY OF STROKE IMAGING
The requirements of stroke imaging are reflective of available treatment options. Before the demonstration of an efficacious intervention, management was largely supportive. Of greatest importance was not the detection of manifest ischemic changes, but the detection of processes such as infection or tumor for which interventions were available. Once processes requiring emergent intervention such as infection and tumor were excluded, supportive measures were instituted, including the prevention of additional thrombus formation with use of anticoagulation. It was well accepted that in the absence of detecting the aforementioned processes, deficits lasting less than 24 hours were due to transient ischemic attacks. Neurologic deficits persisting beyond 24 hours were attributed to infarction. Imaging confirmation of manifest infarction was not needed.
The demonstrated efficacy of the intravenously administered thrombolytic, alteplase ([rt-PA] Activase; Genentech, South San Francisco, Calif), has redefined the requirements for neuroimaging. Optimal management now requires delineation of infarction prior to intervention. This requirement is best understood in the context of the role of thrombolytics for cerebral ischemia and the pathophysiology of cerebral ischemia.
THROMBOLYSIS
Prompt reestablishment of flow with lysis of an occlusive thrombus is the goal of thrombolysis. This is in contrast to anticoagulation, in which new clot formation is minimized. These agents convert the proenzyme plasminogen to the active enzyme in thrombin lysis, plasmin. Thrombolysis can be performed by using streptokinase, urokinase, or rt-PA. rt-PA is the most specific of the three agents. It is fibrin specific, activating only thrombin-bound plasminogen. For this reason, it has the most favorable therapeutic index of the three thrombolytic agents.
After thrombolysis was initially proposed in 1958, its acceptability was limited by the inability to differentiate cerebral ischemia from primary intraparenchymal hematoma, the latter a contraindication to thrombolysis (4). Enthusiasm was renewed in the 1990s because of the wide availability of computed tomographic (CT) scanners and the demonstration of efficacy of thrombolytics in the treatment of myocardial ischemia (5,6). This resulted in the performance of numerous clinical trials (3,7–10). Of these, two intravenous rt-PA clinical trials have been of particular importance in defining the current approach to treating cerebral ischemia—the NINDS trials (3) and the European Cooperative Acute Stroke Study (ECASS) (7).
The NINDS trial was the first trial to demonstrate efficacy in the use of the thrombolytics for the treatment of cerebral ischemia (3). The protocol was based on a 3-hour treatment window from the time of ictus to the administration of rt-PA. A CT scan was obtained, and the detection of intracerebral thrombosis was an exclusion criterion. Although no short-term benefits were demonstrated, patients in the treatment group achieved improved stroke scale ratings as compared with those in the placebo group at 3 months. This resulted in approval by the U.S. Food and Drug Administration of rt-PA in June of 1996 and the widespread implementation of acute stroke teams across North America.
The ECASS trial allowed a longer treatment window of 6 hours (7). This time-window prolongation increased the likelihood of manifest infarction at the time of reestablishing flow and thus the risk of reperfusion hemorrhage. In an attempt to avoid reperfusion hemorrhage, a CT scan was obtained to exclude patients with extensive infarction. Patients were excluded if on-site interpretation detected infarction in greater than 33% of the middle cerebral artery (MCA) distribution. Unfortunately, the ECASS group failed to demonstrate efficacy in the intention to treat group. Notably, in addition to the on-site interpretation, CT scans were later reviewed by the ECASS CT review panel. The panel determined that "misinterpretation" by the on-site readers resulted in inappropriate entrance into the trial of 52 individuals (7). By excluding this subgroup of protocol violators, efficacy was demonstrable.
The need for subgroup analysis led to skepticism about the efficacy of thrombolysis beyond 3 hours, which is reflected in the aforementioned approval of a 3-hour treatment window and the relative lack of enthusiasm in the European Community toward implementing thrombolysis. Despite this, the design and outcome of the ECASS trial has had a major effect on defining the optimal approach to stroke intervention in that it has redefined the role for neuroimaging in acute stroke. In this study, CT was used to detect the presence of infarction as opposed to its prior role of excluding other processes such as hemorrhage, tumor, or infection. The ECASS trial also "suggested," based on subgroup analysis, that with neuroimaging it is possible to extend the treatment window beyond 3 hours. This is of great relevance in that on the basis of estimates from the International Stroke Trial, only 4% of acute stroke patients presenting to the hospital are able to arrive within 3 hours (11). Extending the treatment window to 6 hours would make intervention available to approximately 16% of acute stroke patients (11,12).
Finally, this trial demonstrates that "in practice" the early CT changes of infarction are subtle. For example, in 11% of the patients in the ECASS trial, initial interpretation did not detect the presence of infarcts seen on post hoc review. Thus, although enabling of efficacy when the treatment window is less than 3 hours, CT is not the optimal technique for delineating the presence or absence of infarction in the acute period. This "insensitivity" becomes even more important in considering extension of the treatment window beyond 3 hours when the likelihood of infarcted tissue being present and risks of posttreatment hemorrhage are greater.
NEWLY DEFINED ROLE OF CT
CT detection of extended infarction is a recommended exclusion criterion based on the American Heart Association Thrombolysis Practice Advisory Guidelines (13). It is now incumbent on all radiologists to be familiar with the CT evaluation of acute stroke. The sensitivity of CT in the detection of the early changes of infarction is related to the underlying principles of image acquisition. CT images are obtained by projection of a high-kilovoltage collimated beam through the brain. Beam attenuation is due to absorption proportional to the linear attenuation coefficient of the materials through which it passes. The relatively high-kilovoltage x rays used in CT result in linear attenuation primarily due to tissue density. Differentiation of adjacent tissues such as gray matter and white matter depends on perceivable differences in electron density; detection of a pathologic condition requires a perceivable change in electron density.
Fluid redistribution characterizes cerebral ischemia and imparts the changes in electron density detectable with CT. Maintenance of a normal cell volume involves the double-Donnan-equilibrium between electrolytes and fluid exchange between the extra- and intracellular spaces (14,15). This requires a continuous supply of energy to the membrane pumps necessary in maintaining normal intra- and extracellular electrolyte concentration gradients and fluid distribution. With the interruption of blood flow, these energy needs are no longer met, resulting in cytotoxic edema, which is defined as cellular injury with influx of fluid in the intracellular space without an increase in vascular permeability (16).
Neurons are the most sensitive cell to ischemia. These cells are located in gray matter. Initially more dense than the white matter, gray matter becomes increasingly less dense with an increase in water content. The normal gray matter–white matter differentiation that characterizes the normal CT scan progressively declines and results in three of the CT signs of acute cerebral infarction: (a) blurring of the clarity of the internal capsule, (b) loss of distinctness of the insular ribbon cortex, and (c) loss of differentiation between the cortical gray matter and the subjacent white matter (Fig 1) (17,18). In addition to the attenuation changes, morphologic changes occur due to the accumulation of intracellular fluid causing swelling of the cortical gyri. This results in effacement of the spaces demarcated by the gyral infoldings (sulci), known as "sulcal effacement" (19). This is demonstrated prior to the appearance of hypoattenuation. Although these findings become prominent over time, detectable alterations in attenuation are not present at the time of ictus (20).
|
|
|
|
Administration of contrast material may be helpful in delineating infarcts in the subacute period when there may be obscuration of the infarct by "fogging" (24). Fogging was described by Becker et al (25) in 1979 as the phenomenon of an area of previously hypoattenuating infarction evolving to a state of isoattenuation between days 14 and 21 after infarction; this occurrence was believed to be secondary to small petechial hemorrhages or infiltration of the infarcted tissue with macrophages (25). Although difficult to see on a nonenhanced study, the area of infarction intensely enhances at the periphery.
Conversely, despite histologic breakdown of the blood-brain barrier beginning on the day of infarction, contrast enhancement is not generally seen during the first 3 days. Further, there is a theoretical concern about intravenous contrast material causing further parenchymal injury in the presence of a disrupted blood-brain barrier (26). Finally, intravenous contrast material has not been demonstrated to increase the detectability of ischemic changes within the first 24 hours (27). Therefore, it is not recommended in the acute evaluation of stroke.
In summary, nonenhanced head CT has a clearly defined role in the current management of acute stroke (Table). A CT scan for the exclusion of hemorrhage is a listed requirement in the package insert for alteplase. Also, in the practice guidelines for the imaging of acute stroke, the Stroke Council of the American Heart Association recommends an initial CT scan (13). It has enabled the implementation of the first efficacious treatment of ischemia. CT is appealing as an imaging study in that it is widely available, can be performed quickly and safely on critically ill patients, and is relatively inexpensive. Further, its utility has been proved in large-scale clinical trials.
|
The current approach to evaluation and treatment of cerebral ischemia with thrombolytics demands further optimization. An analysis of rt-PA thrombolysis estimated that on the basis of the current approach to thrombolysis, there is a 12% chance of an improved outcome at 3 months (28). However, there is also a concomitant 3% increased risk of intracranial hemorrhage. Reperfusion into areas of infarction is associated with an increased risk of hemorrhage (28). As previously noted, in the ECASS study, trained investigators did not detect these "subtle but present changes" in approximately 11% of the patients. von Kummer et al (29) demonstrated that in the first 6 hours after onset of ischemia, 31% of CT scans may be misinterpreted (Fig 3). Although CT may show findings of infarction as early as 3–6 hours after ictus, up to 60% of CT scans are normal in the first few hours after ischemic insult. Thus, despite some reports of high sensitivity in the acute period, it is still accepted that its overall sensitivity is relatively poor (30,31).
|
|
Another 10%–20% had arterial dissections, occlusions sites not readily susceptible to thrombolysis, or minor branch occlusions that without intervention have a favorable natural outcome (33). The lack of sensitivity and specificity in the diagnosis of stroke limits the attainable benefit. Therefore, although CT has proved to be of tremendous benefit, other modalities such as magnetic resonance (MR) imaging may provide an opportunity to further improvements in patient care (12).
CONVENTIONAL MR IMAGING FOR INCREASED SENSITIVITY IN DETECTION OF ACUTE STROKE
Earlier detection of areas of infarction is possible with MR imaging. MR imaging involves placing a patient in an applied external magnetic field. Because of the precessional properties of certain nuclei with an odd molecular mass or odd charge, the most abundant of which is hydrogen, the nuclei will act as small magnets that align with the magnetic field. With the use of applied radio-frequency pulses and receiver coils, it is possible to generate and detect different signals from tissue based on the number of protons per unit volume (ie, spin density) of different substances. MR imaging takes advantage of the fact that water protons have different biophysical relaxation properties, described by times T1 and T2, depending on the microscopic environment of the water protons. Acute cerebral ischemia is characterized by changes in electrolyte and water balance that result in alterations in this environment. These differences can be exploited to increase contrast, and as a result MR imaging is extremely sensitive in the detection of acute infarction.
Acute infarcts are more visible on MR images than on CT scans, with over 80% of MR images positive in the 1st day compared to 60% of CT scans (19). MR imaging is particularly superior in the detection of stroke in the posterior fossa where CT is limited due to beam-hardening artifact from the adjacent skull base. Lacunar infarcts and small cortical strokes are also seen with higher conspicuity (19).
The earliest MR changes are loss of normal intravascular flow voids, morphologic swelling of the gray matter, and increased signal intensity on the T2-weighted and intermediated–weighted (3,000/30 [repetition time msec/echo time msec]) images [Editor's note: The convention of Radiology is to term this sequence intermediate weighted although it has also been termed proton-density weighted]. There is normally a loss of intraarterial signal with standard spin-echo (SE) sequences, referred to as a "flow void," due to precessional phase changes induced by blood flow. In low- or no-flow states, there is a loss of the "flow void" such that signal intensity is demonstrated in the involved vessel (19). Similar to the "hyperdense MCA sign" in CT, absence of flow in involved vessels can be seen immediately after occlusion.
The morphologic changes at MR imaging relate to cellular swelling. These have been shown to appear before the T2 signal changes, although these changes are very subtle. More easily detected are the signal intensity changes that result from an increase in intracellular fluid. These can be detected on both the T2-weighted images or on the intermediated-weighted images with the T1-weighted signal intensity changes the least sensitive for detection of stroke. The changes on the intermediate-weighted changes are often more conspicuous than those on the T2-weighted images, particularly in strokes involving the cortical gray matter (19). The earliest signal intensity changes usually involve the gray matter, with the white matter typically appearing normal in the first 24 hours.
Intermediate-weighted images are also superior for detection of lesions near a ventricle or subarachnoid space. This is because the presence of cerebrospinal fluid (CSF) can make it difficult to identify the lesions on the T2-weighted images given that both CSF and infarcted tissue may have similar signal intensities. CSF is isointense to gray matter, and thus areas of abnormal hyperintensity can be detected. Similarly, intermediate-weighted images enable differentiation of lacunar infarcts from perivascular spaces: the former hyperintense and the latter isointense to CSF.
Intravenous contrast material can provide additional information. Vascular enhancement can be seen in cortical infarcts up to 75% of the time (34). This is believed to be due to slow flow in the region of the infarction. Meningeal enhancement can also be seen acutely in approximately 33% of patients, possibly due to meningeal inflammation (34). Both types of enhancement resolve toward the end of the 1st week. This can potentially delineate the acuteness of a lesion in patients demonstrating more that one area of infarction. In general, however, it is not necessary to administer contrast material to assess the enhancement characteristics of an ischemic lesion. As will be discussed subsequently, the acuteness of a lesion can be adequately determined by using MR diffusion imaging. Administration of contrast material is typically reserved for those instances in which MR perfusion imaging is performed.
New MR techniques are now available to generate images with T2 weighting in less time. A fast SE sequence enables the acquisition of T2-weighted images in a fraction of the time required to obtain similar conventional SE images(roughly four to 16 times faster). This is made possible by acquiring multiple lines in k space per repetition time, TR (number of lines acquired equal to the echo train length). In conventional SE imaging, only one line of k space is acquired per repetition time. Whereas the time of acquisition of an SE image is number of signals acquired (NSA) x TR x the number of phase encoding steps (Nphase), time of acquisition in fast SE imaging is (NSA x TR x Nphase)/ETL, where ETL is the echo train length. For example, with an echo train length of 8, a fast SE evaluation of equal resolution could be performed as much as eight times faster. This is of particular value in evaluating patients with acute stroke for which prompt diagnosis and institution of therapy portends the greatest likelihood of tissue salvage. Other benefits include reduced flow artifacts due to even echo rephasing, a result of the symmetric sequence design within a repetition time cycle (35). This may increase the conspicuity of posterior fossa lesions where flow-related artifacts can be particularly problematic (36). Further, the ability to rapidly acquire the images enables timely performance of high-spatial-resolution T2-weighted imaging and limits patient motion artifact. Although the information obtained is similar with equivalent detectability of large lesions, the detection of small lesions is decreased (37). This may be due to the image blurring that occurs in the phase direction due to unequal echo delays in the echo train (38). Further, true proton-density–weighted images are not as reliably obtained, limiting the ability to differentiate perivascular spaces from lacunar infarcts.
Lesion conspicuity can be further optimized by using an inversion-recovery sequence in which an initial 180° radio-frequency pulse is followed at an inversion time by a 90° radio-frequency pulse. The latter converts the partially recovered longitudinal magnetization to transverse magnetization, at which point it is detected (39). For a given T1 relaxation time, an inversion time can be used that will result in zero transverse magnetization. A fluid-attenuated inversion-recovery (FLAIR) sequence using an inversion time approximating the null point of CSF (2,200 msec) combined with a long echo time to obtain T2 weighting suppresses the signal from CSF while maintaining the hyperintensity associated with pathologic processes (40). This leads to reduced artifacts due to CSF flow and volume averaging and results in excellent conspicuity of both small and large lesions (41). FLAIR is superior to fast SE T2-weighted and intermediate-weighted images in the evaluation of cerebrovascular disease, with lacunar and cortical infarcts more conspicuous (Fig 4) (42).
|
|
|
MR Exclusion of Intraparenchymal Hemorrhage
As previously mentioned, exclusion of intraparenchymal hemorrhage is central to the evaluation of acute stroke. Conventional SE techniques are highly sensitive for the detection of subacute and chronic hemorrhage. Deoxyhemoglobin, methemoglobin, and hemosiderin are the breakdown products of hemoglobin (45). These substances are paramagnetic and with retraction, into a fibrin clot, they cause a marked alteration in relaxation properties (46). In the hyperacute stages of hemorrhage in which there has been minimal conversion of oxyhemoglobin to breakdown products and no mature clot formation, conventional MR sequences are less sensitive. This has caused reluctance to rely on MR as the primary imaging modality in acute stroke due to concerns of false-negative results.
Conventional SE imaging techniques have a lower sensitivity for the detection of blood products due to an applied 180° radio-frequency phase refocusing pulse. Fast SE techniques have an even lower sensitivity. However, gradient-recalled-echo (GRE) imaging employs a reversal of the readout gradient, without a refocusing radio-frequency pulse, resulting in high sensitivity to magnetic field inhomogeneity induced by the presence of the paramagnetic breakdown products of blood. As a result, GRE images are particularly sensitive in the detection of acute intraparenchymal hemorrhage (45) (Fig 5). When conventional SE sequences are supplemented with GRE hemosiderin-sensitive sequences, the earlier stages of hemorrhage can be diagnosed equally well with MR imaging and CT (45,47). The criteria for the presence of hemorrhage on MR images are increased signal intensity on the T1-weighted images, decreased signal intensity on the T2-weighted images, and/or decreased signal intensity on the GRE images when the signal intensity is compared to that of white matter. In the hyperacute stage, the central area of hemorrhage will appear somewhat similar to CSF, but there is often a subtle halo of hypointensity on GRE images.
|
|
|
Assessing Large-Vessel Patency
As discussed earlier, approximately 20% of patients presenting with symptoms of acute cerebral ischemia do not have an angiographically demonstrable vascular occlusion. An additional 10%–20% have small-vessel occlusions or dissections for which thrombolysis is unlikely to be of benefit. Thus, in addition to demonstrating areas of infarction, it is equally important to demonstrate that there is a vascular lesion that is amenable to thrombolysis.
MR angiography enables the noninvasive evaluation of vascular patency. Images can be generated on which flow within the vessel is increased in signal intensity (bright blood) or on which the lumen is depicted as decreased in signal intensity (black blood); the former is the more commonly used technique. Time of flight is the most frequently used bright-blood MR angiographic technique. Key issues in MR angiography are optimizing intravascular signal intensity and maximizing suppression of stationary spins. Time of flight applies a GRE sequence, enabling the use of a shorter echo time and repetition time than those of SE imaging, important in minimizing signal loss due to intravascular intravoxel phase dispersion and optimizing suppression of stationary tissue, respectively. For evaluating the intracranial circulation, a volumetric (three-dimensional) acquisition is optimal, enabling thinner voxels, which limits intravoxel phase dispersion and the associated artifactual signal loss in the region of stenoses. In time-of-flight MR angiography, background suppression is particularly important. Magnetization transfer and manipulation of echo time such that fat and water are out of phase yield improved background suppression. Additionally, increasing the flip angle in the direction of flow minimizes saturation effects, which optimizes intravascular signal recovery in the more distal vessels. Loss of phase coherence limits usefulness in small peripheral vessels by preventing generation of an intravascular signal. However, an image similar to a conventional arteriogram can be reconstructed from the intravascular signals that is sensitive to large vessel occlusion or narrowing in the internal carotid, vertebral, basilar, and first and second segments of the anterior, middle, and posterior cerebral arteries. Although the value of MR angiography has yet to be shown in a large-scale clinical trial, the need for differentiating which patients have lesions amenable to thrombolysis seems clear given the potential risks of treatment (Fig 6).
|
|
|
TREATABLE TISSUE BEYOND 3 HOURS: THE PENUMBRA
Compared to other tissues such as muscle, the brain has little capacity to store energy. In zero-flow states (complete ischemia), the available energy will maintain cell viability for approximately 2–3 minutes. More often, ischemia is incomplete, with the area of greatest compromise located centrally with a large area of surrounding tissue at risk but remaining viable. The central region in which cell death will occur generally corresponds to a region in which energy supply is insufficient to maintain membrane pumps leading to loss of the membrane potential. The surrounding at-risk area can have detectably abnormal evoked potentials but has not undergone cell depolarization. This area is referred to as the penumbra (50,51).
The progression of acute stroke is a common occurrence. The mechanisms for progression of infarct are multiple, including cerebral edema, hemorrhagic transformation, poor collateral flow, low blood pressure, or increased blood glucose levels. Animal and human studies (19,52) have documented both reversal of ischemia and progression of ischemic tissue to infarction over periods as long as 36 hours. In a study by Castillo et al (52), 34% of patients demonstrated such progression with a drop in their Canadian Stroke Scale score of 1 or more over 48 hours. Bryan et al (19) demonstrated that at least one-third of MR imaging studies showed that stroke areas increased in size after 24 hours. This suggests that there is salvageable tissue beyond the currently accepted 3-hour therapeutic window.
ADVANCED MR IMAGING TECHNIQUES FOR BETTER DEFINITION OF STROKE
The imaging requirements for extending the treatment window are delineation of areas of infarction and risk of infarction, evaluation of large-vessel patency, and exclusion of nonischemic causes of symptom onset. Conventional CT combined with CT angiography and CT perfusion imaging has promise due to its accessibility and ability to demonstrate areas of hemorrhage, large-vessel occlusions, and regions of hypoperfusion (52a). However, it remains somewhat limited by its current low sensitivity for delineating irreversibly injured tissue in the hyperacute period. Numerous other imaging techniques can contribute in assessing individual parameters, but few are both comprehensive and expeditious. These include xenon-enhanced CT, positron emission tomography (PET), single photon emission CT, and transcranial Doppler ultrasonography. For example, whereas PET provides important quantitative information related to blood flow and oxygen extraction, current limitations in anatomic delineation (and accessibility) have hampered its routine use in accessing stroke patients (53). Conversely, MR imaging generally has higher spatial resolution than other tomographic techniques (similar to that of CT), enabling anatomic, physiologic, and metabolic analysis in a single examination. Therefore, rather than presenting the wide array of available imaging techniques, in reviewing the state of the art for acute stroke imaging, emphasis is focused on the potential of advanced MR imaging techniques to provide the information relevant to the treatment of cerebral ischemia in a time-efficient and comprehensive fashion. Three newer techniques hold great promise for increasing the sensitivity of detecting infarction and demarcating the area at risk and are reviewed in detail. These techniques include MR diffusion imaging, MR perfusion imaging, and proton MR spectroscopic imaging.
MR Diffusion Imaging
The effects of molecular self-diffusion on the amplitude of SE MR signals was recognized early in the development of nuclear MR spectroscopy (54), and the equations and pulsed field gradient methods published by Stejskal and Tanner (55) in 1965 form the basis of modern diffusion-weighted MR imaging (56). In 1990, Moseley et al (57) demonstrated in an animal stroke model that the apparent diffusion coefficient (ADC) decreased by approximately 30%–50% within 30 minutes after onset of focal ischemia. This change occurs while other MR image types such as T2- or intermediate-weighted images remain normal (Fig 7) (57,58). It is now well demonstrated that the majority of the diffusion changes coincide with the membrane depolarization due to Na+,K+-ATPase pump failure (59–61), and this can be measured instantaneously by using MR imaging (58,62). This coincides with water redistribution between intra- and interstitial space upon energy failure, although the exact physical mechanism for the changes in diffusion-weighted imaging is still a point of discussion (63–67). Although many mechanisms have been proposed, it is generally believed that the most likely cause of the decrease in ADC values is the redistribution of water from the interstitial to the (diffusion-restricted) intracellular space, as the energy-dependent Na+,K+-ATPase pumps fail (ie, the initiation of cytotoxic edema) (63,65,68–70). Of note is that there is also ample proof that the explanation based on a high interstitial and low intracellular diffusion constant is somewhat of an oversimplification. Further experimental data are needed to support the interstitial diffusion values that underlie this assumption.
|
|
|
Therefore, there have been attempts to associate diffusion-weighted imaging with other imaging parameters (such as T2), which, when used in combination, may better stage stroke progression (58,82). This was first performed in a permanent MCA occlusion model in the rat; after the initial ADC drop, by 12 hours, as vasogenic edema begins to accumulate, the T2 begins to increase (this stage of stroke is particularly conspicuous on diffusion-weighted images, which, because of limited gradient strength on clinical imagers, are both diffusion- and T2-weighted in most cases) (83). As ischemia continues, the ADC begins to return to normal (12–24 hours in this model) and eventually becomes elevated (24–48 hours) as necrosis occurs. Various combinations of ADC and T2 ("signatures") were identified, which corresponded to the different histopathologic conditions of evolving infarction. The same behavior has also been observed in human stroke, although the time scale appears to be different, with several groups finding a slower recovery and increase in ADC (after several days) compared to that in the animal studies (84,85). However, this has been somewhat controversial and may be partly technique dependent.
Although diffusion-weighted imaging can help identify stroke location, therefore, it is questionable if it alone can differentiate reversible from irreversible ischemic injury. However, the ability to positively diagnose stroke from an imaging perspective is extremely valuable for initiating stroke therapy (86). Diffusion-weighted imaging has also been shown to enable differentiation of acute from nonacute infarcts (87), to help predict outcome (88,89), and to facilitate correlation with final infarct size, although final infarcts were larger than abnormalities at diffusion-weighted imaging seen at 6, 12, or 24 hours (78). It was suggested that enlargement of the area of infarction was likely to occur if the area of perfusion deficit was larger than the area of diffusion abnormality (90). We and others also found that acute perfusion abnormalities tend to be larger than final infarct size (88,91,92). While the exact time scale for ADC reversal is still debated, the ability of diffusion-weighted imaging to distinguish an acute from a chronic lesion is of great clinical interest (93). In addition, it was found possible to predict which strokes would improve versus those that would not on the basis of the diffusion-weighted image: In one report of 40 patients, only four had no detectable abnormality on diffusion-weighted images; in all four of these patients, the neurologic deficit subsequently resolved (90). However, there have been isolated reports of patients with negative diffusion-weighted imaging studies who subsequently developed infarction (94).
For all of these reasons, diffusion-weighted imaging is becoming an essential stroke imaging modality, but diffusion-weighted imaging of the human brain also poses numerous technical challenges. Perhaps the most difficult problem is the elimination of motion artifacts on the heavily diffusion-weighted images. Large phase errors can occur during application of the diffusion gradients as the result of macroscopic sample motion. Various different approaches have been tried to address this problem. One approach is to image as rapidly as possible in order to minimize potential motion. If the image can be recorded in a single shot, for instance by using echo-planar imaging, every line of k space has the same phase error, so no image artifacts are present after Fourier transformation and magnitude calculation (90). Other rapid imaging techniques, such as gradient- and spin-echo (GRASE) or fast low-angle shot (FLASH), may also be used (95,96). An alternative approach is to measure, and correct for, the phase error on each line of k space, by recording a second, non–phase-encoded echo (called a "navigator" echo) (97). This approach allows diffusion imaging to be performed with conventional SE sequences, which are generally less demanding on MR instrumentation and have better signal-to-noise ratio and fewer susceptibility artifacts than those of single-shot imaging sequences. However, adequate correction with navigator echoes is only possible for relatively simple, small amplitude motions, and this approach is less favorable than fast imaging methods, particularly for noncompliant patients, and/or when multiple b values and gradient directions are required. As a result, echo-planar diffusion is the most practical and commonly used method for stroke imaging (79).
Another technical issue regards the directional dependence of the ADC. While the ADC in liquids is isotropic, in biological samples the motion of water molecules is partially restricted, resulting in different diffusivities in different directions (anisotropy). Mathematically, diffusion therefore has to be described as a tensor, and the signal attenuation in diffusion-weighted imaging will depend on the direction and amplitude of the applied gradients, which is proportional to the proton gyromagnetic ratio, the gradient length, and the diffusion time, with the time between the (start of) two gradients in the diffusion experiment, and x, y, and z are the orthogonal directions defined by the scanner magnetic field gradients.
In the brain, ADC anisotropy is particularly evident in white matter, due to the greater diffusivity of water parallel (as opposed to perpendicular) to the axonal fibers. This property can potentially be exploited to noninvasively map brain fiber tracts and directions (70), and to study the processes of myelination (98) and demyelinating diseases (99). However, in the context of stroke imaging, diffusion anisotropy is largely an added complication: If only a single diffusion-gradient direction is used (as in the case of most early studies of human stroke), hyperintensity is often observed in the densely myelinated corticospinal tracts, corpus callosum, or other white matter regions, which may be difficult to distinguish from reductions in ADC due to ischemia. This important problem can be avoided by using certain combinations of these ADCs that are not orientation-dependent, such as the so-called trace of the diffusion tensor (100). Trace imaging is a more reliable method for measuring changes in ADCs and may avoid interpretation errors due to the intrinsic anisotropy of the brain (81,100).
With appropriate pulse sequence design, the orientation-independent trace or average value of the diffusion constant, Dav, can be measured, which provides nearly uniform diffusion-weighted images in normal brain (90,100). Normal Dav values are approximately 3.2 x 10-5 cm2/sec for CSF and 0.8 x 10-5 cm2/sec for human brain parenchyma (101). Gray matter may have a slightly higher (by approximately 10%) Dav than that of white matter, although it is difficult to measure cortical gray matter regions of interest that are completely free from CSF contamination. Trace images can be constructed either by measuring Dxx, Dyy, and Dzz in separate experiments, or by using sequences that give isotropic diffusion-weighting in a single shot (102). Each approach has its own advantages and disadvantages: The single-shot trace methods are fast and are not subject to errors due to head motion (or different eddy-current artifacts) between different sequences. However, they are more demanding on gradient strength, and with typical currently available gradient sets, maximum b values are rather low using acceptable echo times (eg, on our current GE Medical Systems unit [Milwaukee, Wis] with "echo-speed" gradients, Gmax = 2.2 G/cm, with an echo time approximately 126 msec, bmax approximately 550 sec/mm2) (103,104). Ideally, b values in the range of 1,000–1,500 sec/mm2 have been shown to give optimal contrast for stroke diffusion-weighted imaging studies (105). Individual measurements of Dxx, Dyy, and Dzz, as well as applying gradient combinations in the xy, xz, and yz directions, allow the full diffusion tensor to be determined, and to calculate indexes of diffusion anisotropy, which may also change in ischemia (106). Higher b values can be obtained for better contrast, but at the expense of increased imaging time compared to that of the single-shot trace methods. A minor advantage of the single-shot trace sequence, but nevertheless an important one in the clinical environment, is that less offline processing is required to view orientation-independent images. Newer imagers with "scalable" gradient driver hardware can produce gradient strengths of up to 4 G/cm, which should enable single-shot imaging with adequate b-value weighting (104).
MR Perfusion Imaging
MR imaging of cerebral perfusion can be performed by injecting a bolus of gadopentetate dimeglumine (107). In normal brain, the paramagnetic contrast agent gadopentetate dimeglumine remains enclosed within the cerebral vasculature; this shortens the longitudinal relaxation time (T1) of blood, which causes a paramagnetic frequency shift within the vessel, and creates local magnetic field inhomogeneities in surrounding tissues, which shorten the transverse relaxation time constant T2* (108–113). Under certain conditions, each one of these effects is linearly proportional to the concentration of the tracer (contrast agent) within the image pixel. Therefore, by using the theory of tracer kinetics, rapid MR images acquired during the first pass of the contrast agent through the brain can be analyzed to produce images of relative regional cerebral blood volume rCBV (114). If an estimate of the arterial input function is available, the mean transit time MTT, and regional cerebral blood flow rCBF can also be estimated (rCBF = rCBV/MTT) (115,116). In lesions where the blood-brain barrier is disrupted, the latter part of the signal intensity–time curve can also be analyzed to give an index of relative blood-brain barrier permeability (117), although in most patients with acute stroke this may be expected to be small.
The theory of calculating regional cerebral blood flow from data sets of this type is complicated and too involved to be repeated here (116,118). A number of assumptions are required in order to perform the calculation, including such factors as estimation of capillary hematocrit, the assumption that the change in R2* (
R2* = -ln[S/S0]/TE) signal intensity changes follows the same dependency in the microvasculature and major vessels, and the difficulty of performing the appropriate mathematical deconvolution on rather coarsely digitized and noisy data. In some experiments, the input function may not be available, and it is debatable as to where the input function should be determined in stroke patients with occluded major vessels. For these reasons, all MR perfusion studies of human stroke to date have used relative measures of regional cerebral blood volume or other, nonquantitative perfusion-related parameters.
Theoretically, the distribution of an intravenously administered bolus leaving the heart-lung system should follow a gamma-variate function (119). Cerebral blood volume is proportional to the area under the curve; it can be seen that several simple parameters are easily measured and sensitive to changes in blood flow: peak signal intensity changes, time to peak, and width at half height (Fig 8). Each of these parameters is loosely proportional (or inversely proportional) to blood flow. Since, from PET studies, regional cerebral blood volume is known to be relatively nonspecific in stroke (120), these alternative, hemodynamic, flow-related parameters may be more sensitive for detection of abnormalities. PET has shown that cerebral blood volume may be unchanged, increased (as the result of compensatory vasodilatation), or decreased (as the result of collapse and/or occlusion of the ischemic vasculature) in cerebrovascular disease. Blood flow, however, is always reduced unless reperfusion and hyperemia have occurred. We found that arrival time (ie, time to peak) maps were much more sensitive than cerebral blood volume maps in detecting abnormalities in acute stroke, and this has also been the experience of others (92,121). In patients with occluded vessels, bolus arrival times were substantially longer than in contralateral hemispheres with patent vessels. Increased arrival times could be due to two reasons: (a) blood flow may be decreased, with increased mean transit times (at constant blood volume), or (b) the flow may be collateral or contralateral, with the result that the bolus has to traverse a longer path in order to reach the ischemic region. It is also possible that both mechanisms are involved. This delineation is benefited by assessing for large-vessel occlusion with MR angiography.
|
|
|
MR Spectroscopic Imaging
Proton MR spectra of the brain enable detection of resonances that are relevant in the evaluation of patients with cerebral ischemia: lactate at 1.33 parts per million (ppm) and N-acetylaspartate (NAA) at 2.02 ppm. Lactate is not present in sufficiently high concentrations to measure in the brain under normal conditions. NAA is normally present and is found only in neurons or axons in mature brain (134,135).
Numerous studies (136,139,141) of animal models have indicated
