Applications of diffusion–perfusion magnetic resonance imaging in acute ischemic stroke

The basics of DWI and PI.

DWI is derived from measurements of the random movement (Brownian motion) of water molecule protons. A value known as the apparent diffusion coefficient (ADC) is determined by diffusion weighting of the imaging sequence, and this value is dependent on a number of variables including time, orientation of the imaging plane, tissue being imaged (i.e., ADC values differ in normal white and gray matter), and the energy state of the imaged tissue.7 Signal intensity on a gray scale is directly related to ADC values on DWI; brain tissue with low ADC values appears relatively hyperintense, whereas regions with higher ADC values appear hypointense. In both animals and humans, ADC values decline very rapidly after the onset of ischemia, leading to the appearance of hyperintense regions in these ischemic zones on DWI. The ADC declines are related to reductions in cerebral blood flow that cause a failure of high energy metabolism and the development of cytotoxic edema.8,9 It must be recognized that ADC declines are not specific for focal brain ischemia, as similar ADC declines occur with global ischemia, hypoglycemia, spreading waves of cortical depolarization, and status epilepticus.7 These conditions share a commonality of abrupt water shifts from the extracellular to the intracellular space and this is a potential mechanism for the observed ADC decline. A reduction of intracellular water motion or an increase of intracellular viscosity are other possible explanations for the ADC decline observed in brain ischemia.10 Hyperintense signal changes on DWI must be interpreted in relation to the clinical setting to avoid potential diagnostic confusion.

To acquire DWI studies in stroke patients, an ultrafast technique such as echoplanar imaging is preferable because it essentially eliminates movement artifact and drastically reduces the imaging time.11 DWI provides clinicians the opportunity to determine the location of the ischemic lesion in most stroke patients early after presentation.12 As with standard MRI studies, patients with metal fragments or implants cannot be imaged and excluding this possibility in acutely ill, poorly communicative patients without accompanying family members may be difficult in some cases. Additionally, monitoring acutely ill patients is somewhat more challenging within the MRI unit than with CT and requires MRI compatible equipment such as infusion pumps and ventilators.

With PI, information about the perfusion status of the microcirculation is available.13 There are currently three methods available to perform PI: contrast bolus tracking, blood oxygen level, and arterial spin tagging.14,15 The bolus contrast approach is the most widely employed method in stroke patients and animal stroke models. To acquire a bolus contrast PI study, the patient is injected with a contrast agent such as gadopentate dimeglumine (gadolinium) that has a paramagnetic susceptibility effect. A gradient-echo ultrafast imaging protocol is employed, and as the contrast agent transits the microvasculature, signal intensity declines. Multiple repetitive images are acquired every 1 to 2 seconds and as the contrast agent exits the microvasculature the signal intensity reverts to normal. A signal wash-out curve is derived and the area under the curve represents an estimate of cerebral blood volume.16,17 Given an arterial input function, which can be obtained from an artery identified in the lower brain slices, and the fact that the gadolinum concentration is proportional to the changes in R2* (1/T2*) at low (clinically applicable) gadolinum dosages (up to 3 mm/kg), then the relative cerebral blood flow can be derived, and from the central volume theorem, the relative mean transit time mapped.18,19 This provides an estimate of cerebral blood flow because the absolute arterial input function is not known. Bolus contrast PI measurements of cerebral blood volume show some correlation with positron emission computerized tomography results, but further studies of this relationship are needed.20

Applying DWI and PI to focal brain ischemia.

The main advantages afforded by DWI–PI in comparison to standard MRI and CT are that these new MRI techniques have the following attributes: 1) they demonstrate the location and extent of the ischemic region rapidly after stroke onset; 2) they demonstrate the region of reduced microvascular perfusion very rapidly after stroke onset; and 3) they can be performed serially to evaluate the pattern of evolution of the ischemic lesion.21 In animals, DWI and PI abnormalities are apparent within minutes after the onset of focal brain ischemia.21 In many animals, the size of the perfusion deficit is larger than the diffusion abnormality early after stroke onset, but as time passes, the DWI lesion evolves to match the initial perfusion deficit. By 2 to 3 hours after stroke onset, the ischemic lesion on DWI is highly predictive of the ultimate infarct size, as demonstrated pathologically at later time points, if no therapeutic intervention is employed.22,23 In stroke patients, ischemic lesions are described as early as 105 minutes after stroke onset, but are likely to be present earlier.24 As in animals, the perfusion lesion volume is initially larger than the diffusion lesion volume early after stroke onset in the majority of stroke patients and the magnitude of the initial ADC decline is also less severe than at later time points.25 The ability to identify the ischemic lesion shortly after onset is clearly beneficial for localization and can readily distinguish small subcortical events from larger cortical strokes, a distinction not always clear from the clinical examination.26 With ultrafast MRI, this can be accomplished in the same time that is currently required for performance of a CT scan. MR angiography can also be added to image the larger extra and intracranial vessels, requiring the expenditure of several additional minutes. MR angiography, especially contrast-enhanced three-dimensional time of flight, favorably compares with standard angiography for detecting severe extracranial stenosis and intracranial occlusion but is less reliable for intracranial stenosis.27,28 A concern with using MRI for the acute evaluation of stroke patients has been the question of sensitivity for demonstrating intracranial hemorrhage. These concerns arose because of the apparent inability to adequately image acute intracerebral hemorrhage documented in early studies that used standard MRI techniques and low-field strength magnets. Susceptibility sequences may solve this problem, because the presence of intracranial blood is readily depicted by this approach and a preliminary study demonstrated the usefulness of susceptibility imaging to clearly detect acute intracerebral hemorrhage.29 The utility of this approach for the diagnosis of acute intracerebral hemorrhage is currently being validated and will likely require a direct comparison of the sensitivity of CT with susceptibility imaging on MR obtained in close temporal approximation.

Relationship of DWI–PI to clinical neurologic outcomes and infarct volume.

An important question is whether the lesions seen on early DWI or PI scans correlate with infarct volumes and clinical neurologic outcomes. Several groups have evaluated this issue and have found significant correlations between early DWI–PI volumes and clinical outcome as well as chronic infarct volumes measured on T2 imaging.

In a preliminary study, the presence of an early abnormality on DWI was found to be highly sensitive and specific for predicting a persistent neurologic deficit.30 Subsequently, DWI and PI lesion volumes, obtained within 6.5 hours of stroke onset (before the appearance of ischemic lesions on conventional T2 MRI), were found to have a good correlation with both 24-hour National Institutes of Health Stroke Scale scores (NIHSS) as well as 7-day neurologic outcomes.31 This correlation was highest for PI volumes (R = 0.96, p < 0.001). Similar correlations were seen for early DWI volumes, particularly for patients with large hemispheric lesions. Early DWI–PI volumes have also been found to be highly predictive of chronic neurologic outcome. A recent 50-patient study demonstrated good correlations between acute DWI lesion volumes and both the chronic NIHSS and the Barthel Index.32 These correlation studies of imaging and clinical scoring scales such as the NIHSS demonstrate that the early rating on the NIHSS is also highly predictive of final outcome and is useful for selecting and following patients in clinical trials.31 However, DWI–PI provides additional information concerning lesion location that may prove useful. In a recent clinical trial of the neuroprotective agent citicoline, the correlation between change in NIHSS and change in lesion volume (initial DWI to late T2) was modest, with a correlation coefficient of 0.4, indicating that only 16% of the variability in the clinical outcome could be explained by the change in lesion size.33

Serial DWI imaging during the first week after stroke onset frequently reveals progressive enlargement of the DWI lesion over several days.34 It has been suggested that early reperfusion (as demonstrated by rapid resolution of the PI deficit) may prevent expansion of the DWI lesion and potentially reduce the ultimate clinical deficit. Alternatively, patients who experience late reperfusion, or who do not reperfuse (persistence of the initial PI deficit), may be at increased risk for developing larger infarct volumes and more severe clinical deficits. This hypothesis suggests that early DWI–PI imaging may provide a rough estimate of the ischemic penumbra. Some preliminary data are available to address this issue. In one recent study, patients who were scanned within 6 hours of stroke onset and had initial PI volumes that were larger than the early DWI lesion had ultimate infarct volumes that were considerably larger than the early DWI lesion. In contrast, patients who did not have a large PI lesion on early imaging typically did not demonstrate significant enlargement of the DWI lesion.25 Another recent study demonstrated that patients imaged within several hours of stroke onset typically had larger perfusion than diffusion lesions. Over the course of the next several days, the perfusion lesions typically reduced in size, whereas the diffusion lesions increased in volume by about 30%.31 These studies suggest that a PI volume larger than an associated DWI volume early after stroke onset, a so-called “diffusion–perfusion mismatch,” may in part identify the existence of an ischemic penumbra. It is likely that the region of the diffusion–perfusion mismatch will not precisely identify the ischemic penumbra for several reasons. 1) Not all of the hypoperfused zone on PI has a sufficient cerebral blood flow decline to induce infarction.7 2) In animals, some DWI abnormalities can be reversed and these regions ultimately may not be infarcted at postmortem, implying that not all DWI changes define irreversible ischemic injury.35 3) It is likely that MRI measurements beyond ADC and perfusion abnormalities will be needed to more precisely characterize the potentially salvageable ischemic penumbra from irreversibly injured ischemic tissue.36 What remains to be determined is whether either attenuating expansion or reducing acute DWI volumes with acute stroke interventions will correlate with clinical benefits. If this is the case, then the ability of an investigational agent to prevent the expansion of early DWI volumes (which potentially can be assessed in relatively small numbers of patients) may be predictive of the ultimate therapeutic benefits of that investigational agent.

Are DWI–PI studies clinically useful?

Although at an early stage of evaluation, DWI–PI imaging is evolving as a tool for clinical assessment of stroke patients. DWI and PI studies can be rapidly obtained and repeated several times, if necessary, during the acute phase of ischemic stroke. Early studies demonstrate that DWI appears to be superior to conventional MRI for detecting ischemic lesions, not only in the acute time period (within 6 hours of stroke onset), but also in patients who are imaged between 6 and 48 hours after stroke onset.37,38 Early identification of the clinically significant lesion in patients presenting with acute stroke by the use of DWI and PI is useful in: 1) clarifying whether the acute lesion is in the anterior or posterior circulation (figure 2), 2) differentiating small deep infarcts from larger cortical or subcortical infarcts, and 3) delineating the pattern of the ischemic changes.

• Download figure
• Open in new tab
• Download powerpoint

Figure 2. This patient presented with right arm weakness and unsteady gait and was considered clinically to have a posterior circulation stroke. The T2 image is normal at 24 hours; however, a small lesion is present in the left hemisphere on diffusion-weighted imaging (DWI) (arrowhead). This lesion is demonstrated to be acute because it is dark on the apparent diffusion coefficient (ADC) map (arrowhead).

A fundamental question that influences early diagnostic and therapeutic decisions for patients who present with acute brain ischemia is whether the lesion localizes to the anterior or posterior circulation. For patients with anterior circulation events, imaging of the carotid bifurcation ipsilateral to the event becomes essential because of the well-established utility of carotid endarterectomy for appropriate patients with symptomatic significant carotid stenosis. If the acute lesion is identified to be in the posterior circulation, treatment is more problematic. Specific interventions, such as vertebral angioplasty and anticoagulation, are being evaluated for specific subgroups with high-grade posterior circulation lesions.

Early differentiation of small deep infarcts from larger cortical or cortical/subcortical lesions might influence both diagnostic and therapeutic management. Small, deep infarcts are most frequently caused by occlusion of small penetrating arteries. Large vessel stenoses or cardiac embolism are uncommon causes for small, deep infarcts.39 Detailed cardiac or carotid imaging has a low yield in patients with small, deep infarcts. Therefore, early identification of a typical small, deep (lacunar) infarct may limit the diagnostic evaluation. Alternatively, early identification of the symptomatic lesion as a cortical or cortical/subcortical stroke increases the probability that the etiology is related to large artery atherosclerosis or cardiac embolism, and the diagnostic evaluation can be tailored appropriately.

The pattern seen on DWI may suggest a particular stroke mechanism, such as proximal source of embolism or a hemodynamic large artery stenosis. DWI imaging can detect small, acute lesions in multiple vascular distributions that are not detected on T2-weighted imaging (figure 3).12 This pattern is suggestive of a proximal source of embolism (such as the heart). A borderzone (watershed) type of infarct pattern can be detected more readily at early time points with DWI imaging than conventional T2 imaging in some patients.7 A hemodynamic cause for these infarcts may be suggested. However, data are lacking at this time regarding the positive and negative predictive valve of such patterns for specific cardiac and vascular pathology. Caution should be used in inferring pathogenetic mechanisms from DWI patterns until more extensive information is available. Furthermore, the effect of these diagnostic approaches on improving outcome remains unknown.

• Download figure
• Open in new tab
• Download powerpoint

Figure 3. T2 images (T2), diffusion-weighted images (DWI), and apparent diffusion coefficient (ADC) maps are shown at 24 hours after stroke onset. All 6 images were obtained at the same time point. The top row displays images at the level of the midbrain. The images in the bottom row are from a higher cortical level. A vague lesion in the right parietal lobe is poorly seen on the T2 image. This lesion is easily seen on DWI (arrow) and is confirmed to be an acute lesion because it is dark on the ADC map (arrow). Two additional acute lesions are seen in the left hemisphere, one in the left temporal lobe (large arrowhead) and the other in the parietal lobe (small arrowhead) on the DWI and ADC maps. These lesions were not clearly visible on the T2 images. A cardiac source for emboli was detected on transesophageal echocardiogram.

Utility of DWI–PI for assessing acute therapy.

The ability to rapidly detect the location and extent of the ischemic focus on DWI, as well as the region of abnormal perfusion on PI, suggests that these two MRI techniques may be useful in monitoring acute therapy.12 Many animal studies demonstrated that the response to therapy of focal ischemic brain injury can be monitored by DWI.40,41 The demonstration of in vivo salvage of ischemic regions by DWI was then confirmed by delayed pathologic assessment. Occasionally, surprising results are observed when therapeutic effects are monitored by DWI studies. For example, two groups observed that therapy with a glycine antagonist initiated 30 minutes after permanent MCA occlusion in rats had no effect on early ischemic lesion volumes.40,42 However, several hours later, the ischemic lesion volume began to shrink, and a greater than 40% reduction of ischemic lesion volume was observed on delayed, postmortem evaluation. If only the traditional histologic method of assessing therapeutic response had been performed, the delayed initiation of the treatment effect would not have been known. A recent study used a combination of DWI and PI to evaluate and compare intra-arterial therapy with ProUrokinase (Abbott Laboratories) with IV ProUrokinase therapy and a placebo group in a rat embolic stroke model.43 PI demonstrated that in both ProUrokinase groups significant shrinkage of the volume of hypoperfused ischemic brain occurred. On DWI, this reperfusion effect was accompanied by a significant reduction in the evolution of the ischemic lesion volume over time. A similar DWI–PI study in an embolic stroke model by another group documented the beneficial effects of t-PA on abnormal cerebral blood flow and the evolution of the ischemic lesion.44 These brief examples exemplify how DWI and PI can be useful in the preclinical assessment of potential new acute ischemic stroke therapies. Clinical application of DWI and PI in human stroke trials is ongoing.

Future directions.

We anticipate that many potential benefits for the clinician and drug trialist will likely be provided by DWI–PI. Advising the patient and his or her family about prognosis is likely to be influenced by early subtyping of the stroke, as patients with small lesions generally have a more favorable prognosis. The previously described ability of DWI–PI to identify the size and location of the acute ischemic lesion rapidly after symptom onset provides potentially important diagnostic information early after stroke onset. However, the impact of this new information on patient outcome remains to be determined. Cost–benefit analyses will need to be performed to determine if the diagnostic information provided saves money by reducing the performance of other expensive diagnostic tests and hospital length of stay. Proof that DWI–PI can guide therapeutic decisions is currently lacking, but may be available soon, as ongoing clinical trials employing these new MRI techniques are completed. The cost of DWI–PI is uncertain in many locales because reimbursement rates by third-party payers remain unestablished. Hopefully, charges will not exceed the usual fee for a standard MRI study and replacing a standard MRI with a DWI–PI will be neutral regarding cost.

For stroke therapy trials, the ability to reliably determine stroke subtype with DWI–PI will impact upon trial design and implementation. For example, white matter ischemic injury appears to have different pathophysiologic mechanisms than cortical or gray matter injury, so it is likely that some classes of neuroprotective agents will be ineffective or less effective in white matter ischemia (i.e., NMDA antagonists).45 Clinically, it can be difficult to reliably distinguish small deep white matter strokes from cortical events, but DWI–PI has this capability.26 The ability of DWI–PI to evaluate therapeutic effects in animal stroke models has encouraged clinical stroke trialists to consider employing these MRI modalities for assessment of new treatments.46 Currently, the results of only one moderately sized trial of citicoline in 81 subjects are available.33 This study had a 24-hour enrollment window and was stopped early before the predetermined defined sample size was obtained. The preliminary results indicated that beneficial trends on MRI studies were not well-correlated with trends of clinical benefit. This small study employing DWI to assess stroke therapy should be viewed as preliminary. Another endpoint to consider for acute stroke therapy trials might be effects on DWI lesion growth over the initial 3 to 5 days after onset, although developing edema might confound this type of data analysis. There are currently at least four phase 2 or 3 acute stroke trials (with both neuroprotective and thrombolytic agents) that include these MRI techniques. Other larger trials are being planned. These trials should help to determine if DWI–PI is useful in acute stroke drug development.

Another potential role for DWI–PI studies is in phase 2 clinical trials. For example, if an intervention designed to reduce infarct size by preclinical assessment has no effect on ischemic lesion evolution, determined by early and late MRI studies, then it is probably less likely to achieve significant improvement effects on clinical outcome scales. Currently available data suggest that appropriately targeted stroke populations with less than 100 patients per treatment group will be needed to perform an adequately powered phase 2 study. A significant treatment effect on ischemic lesion evolution determined by MRI studies could provide added justification for a costly and time-consuming large phase 3 trial. This technique might also be used to help estimate optimal drug dosing and treatment duration in phase 2 trials. MRI-based acute stroke drug development could provide a means to evaluate potential therapies in a more cost-effective and expeditious manner, if current ongoing studies clarify the utility of imaging-supported stroke drug development. This approach, however, assumes that there is a close correspondence between reduction in lesion size and clinical outcome. Such a close relationship remains to be demonstrated.

DWI–PI might also be useful for monitoring patients who receive thrombolytic therapy. Recent PET and SPECT reports suggest that reperfusion following recombinant tissue-type plasminogen activator (rt-PA) therapy can be documented by these techniques and is correlated with clinical improvement.47,48 PI could similarly evaluate reperfusion effects with rt-PA and DWI could assess the effect of reperfusion on ischemic lesion evolution. Preliminary, small studies documented the feasibility of DWI–PI studies in patients treated acutely with rt-PA.49 The utility of DWI–PI in relationship to rt-PA therapy remains to be established by appropriately designed and performed studies.

Preliminary experimental data in animals and humans suggest that potentially salvageable ischemic tissue may be identified by DWI–PI.35,50 If this distinction can be made reliably by these MRI techniques, then future stroke therapy trials could be altered dramatically. Patients with no potentially salvageable ischemic tissue could be excluded from trials and patients with some predetermined percentage of ischemic, potentially salvageable tissue could be included. Stroke therapy trials could move from the current fixed time window protocols to studies that allow patient entry based on the status of the brain tissue. Much more work needs to be done to demonstrate convincingly that DWI–PI can distinguish potentially salvageable ischemic tissue from irreversibly injured tissue, and it is likely that successful treatment trials will be the best way to verify this concept in stroke patients.

The potential ability of DWI–PI to distinguish possibly salvageable ischemic tissue from that already irreversibly injured is a most exciting prospect for the future of acute stroke therapy. DWI–PI MRI may become the preferred imaging modality and replace CT, especially if susceptibility-weighted MRI sequences can reliably identify hemorrhagic lesions. CT-based perfusion studies and CT angiography are also showing promise and suggest that CT will be of continued utility for evaluating acute ischemic stroke in facilities unable to procure advanced MRI equipment.51,52 However, the advantages of DWI–PI over CT for depicting the parenchymal injury during the initial time period after stroke onset will be of value in choosing stroke therapies that improve patient outcome. Further research is need to establish the clinical utility of DWI–PI with certainty, but currently available data indicate that DWI–PI will prove to be a major advance in the care of patients with stroke that should be available to all.