Stroke Wars: Episode IV CT Strikes Back
=======================================

* Robert D. Zimmerman

At the end of *Episode III*, upstart MR had seemingly vanquished CT as the imaging procedure of choice for detection of hyperacute infarction. Diffusion-weighted (DW) imaging has proved to be more sensitive than CT for detection of infarction and the combination of fluid-attenuated inversion recovery (FLAIR) and gradient echo sequences were shown to be superior to CT for detection of hemorrhagic transformation of infarction as well as other types of intracranial hemorrhage, including subarachnoid hemorrhage. But, as we begin *Episode IV*, a new threat has arisen. CT has returned with new technological weapons. The combination of multirow scanners and advanced easy-to-use 3D workstations has led to the development of the “Anti-Death Star” (well, okay, it is really a “multi-modal CT stroke protocol,” but I think Anti-Death Star sounds more dramatic). Nonenhanced CT followed by CT angiography (CTA) and CT perfusion (CTP) can be performed within 5 minutes and provides information on the status of the extra- and intracranial circulation, the amount of brain that has been infracted, and the tissue at risk. This technique has swept through academic practices formerly in the MR camp and spread to many clinical practices. Tragically, in some centers, MR advocates and CT promoters, formerly trusted colleagues, now war for patients. (I exaggerate. No one really wars; we just scowl at each other at faculty meetings and social events.)

From my little parody, it is apparent where my sympathies lie, but it is important to realize that both imaging techniques can provide information that would have seemed impossible to obtain a few years ago. Each technique currently has a role, depending on the clinical circumstance and local constraints (type and availability of technology, experience of staff, etc.). The future is a different story. I firmly believe that MR, with its ability to look at many aspects of the pathophysiological features of infarction rather than one parameter (perfusion), will provide information that CT will not be able to match.

Let me begin with the issue that colors this debate. A commonly stated argument against the use of MR imaging is that it is too difficult to perform these examinations in the setting of hyperacute infarction. It is claimed that patients are too unstable or that it is difficult to get the patients on an MR unit quickly enough in light frequency of use of MR imaging and length of examination. It is time that we stop using these problems as excuses and do the work necessary to overcome them. We are 4 years into the new millennium. We should be by using the most powerful tool at our disposal to deal with one of the most common and debilitating diseases we encounter. Patient monitoring has become easier with development of MR-compatible equipment and short-bore magnets. MR imaging in stroke patients can be performed within 15 minutes. New-generation magnets are easier to site, and many institutions have multiple MR systems, some deployed near acute care services. MR systems are more than money-making (or consuming) machines to be fed by healthy outpatients with damaged knees or headaches (or both). The sooner we devote ourselves to overcoming the logistical and technical problems inherent in performing MR imaging in the setting of hyperacute infarction, the sooner we will be able to take advantage of the potential of MR imaging.

## Patients Who Are Not Candidates for Thrombolysis

In individuals presenting 6 hours after symptom onset or those with medical conditions that preclude use of thrombolytics, the goal of imaging is the detection of the presence and extent of infarction and the detection or exclusion of other causes of acute neurologic dysfunction, including intracranial hemorrhage, tumors, infection, posterior reversible encephalopathy syndrome, and venous thrombosis. MR imaging is clearly superior to noncontrast CT for the identification of acute infarction and the neurologic disorders that may mimic infarction. It should be the procedure of choice unless contraindicated (eg, if the patient has a pacemaker or other device incompatible with MR imaging).

It is true that CT can depict many infarcts within the first few hours of onset. Signs of hyperacute infarction include loss of gray matter density (ie, reduced gray matter–white matter differentiation), minimal mass effect, and increased density in a major artery, most often the distal internal carotid or middle cerebral artery (the “dense vessel sign”) signifying an acute clot in the vessel. The loss of gray matter density is always ascribed—without, as far as I can tell, experimental verification—to the presence of edema. It seems to me that this assumption is incorrect or at least incomplete for a number of reasons, and I would like to offer an alternate hypothesis as to its etiology. First, the cytotoxic edema that dominates the early period of infarction is a shift of water from the extracellular to the intracellular space without a net increase in tissue water. Mild vasogenic edema occurs in the early phases of infarction but produces only a minimal increase in tissue water, and mass effect is usually absent. Minimal mass effect may be encountered near the end of the hyperacute phase (around 6 hours). If there really is an increase in water sufficient to cause hypoattenuated CT findings, this should cause obvious hyperintense T2-weighted and FLAIR findings. In a recent study of more than 600 infarcts evaluated within the first 24 hours of ictus, hyperacute infarcts produced T2-weighted and FLAIR findings that were either falsely negative (42%) or only slightly hyperintense (58%) (1). In my experience, these changes are always mild and often so subtle that they are appreciated only in retrospect. It seems improbable that this small change in water content would be more apparent on CT scans than on MR images. So what causes infarcts to be hypoattenuated on CT scans, and why does this have any bearing on this discussion? One needs to ask why normal gray matter is hyperattenuated relative to normal white matter in the first place. The most likely cause is intraluminal blood. Cerebral blood volume (CBV) in gray matter is twice that of white matter (2), and the intrinsic hyperattenuation of blood could account for the increased attenuation in gray matter. On contrast-enhanced CT scans, gray matter–white matter differentiation improves because of increased iodine in the gray matter compared with that in white matter, thus accentuating the effect of differential blood volume. The CT finding of hypoattenuated gray matter in acute infarction is most likely a direct consequence of decreased CBV in the infarcted brain and thus a crude sign of hypoperfusion rather than edema. Gray matter–white matter signal intensity differences at MR imaging are due to water content, not blood volume, so decreased CBV cannot be depicted on unenhanced MR images. If true, this hypothesis implies that CT depicts alteration with a single parameter, perfusion. Perfusion is important but surely it is not the only variable of interest in acute infarction.

How good is CT? Many studies have addressed this issue with variable results. Sensitivity has been reported to be as high as 85% in patients who are candidates for intraarterial thrombolysis and are imaged within 6 hours. This detection rate is deceptively high for several reasons. First, high rates of detection are usually seen in studies that are dominated by large middle cerebral artery–distribution infarcts, the easiest lesions to detect (3). In a large retrospective study of more than 600 patients examined at Harvard University, the detection rate for CT in the first 24 hours for all infarcts was only 40%, whereas detection with DW imaging was 97%. Several studies have documented a detection rate with DW imaging of more than 90% (1,4,5). Second, two other factors have been identified that affect detection with CT but not MR imaging; namely, the experience of the reader and clinical information (4,5). Expert readers do better than novices, and if one knew that the patient was having a stroke, it was easier to find it by using CT. This confirms what all of us know about detection of hyperacute infarction with CT. It is hard. Scans must be of high quality. Loss of gray matter density is subtle and can be difficult to detect even for the experienced reader. Gyri that curve in and out of the imaging plane may be hypoattenuated because of partial volume effects from adjacent CSF. Vessels may be dense because of calcification and may appear asymmetrically attenuated because of patient angulation or partial volume effects or both. Because of this, experienced neuroradiologists do much better than novices at detecting subtle infarcts and differentiating them from artifacts. The stroke neurologist at my hospital has an uncanny knack for correctly identifying acute infarcts on the basis of one hyperattenuated middle cerebral artery branch in the sylvian fissure and one image with a smudged gray white interface. But, of course, he is “cheating.” He has already examined the patient and cannily knows exactly where to look on the CT scan to identify the infarct he knows is present. Thus, in academic centers with experienced neuroradiologists and stroke neurologists, CT detection is good. Most patients with hyperacute infarcts, however, do not present to academic centers. In community hospitals, radiologists and neurologists interpreting CT findings often do not have this degree of expertise or clinical skill. The European thrombolysis trial (ECAS) was initially deemed to have failed, but evaluation of the data revealed that this failure was due to misinterpretation of CT findings by the treating physicians. In routine clinical practice, one should not expect to see even the modest detection rates reported in the Harvard study (1). Alternatively, detection of hyperacute infarction on MR images is easy and requires little experience with either image interpretation or clinical neurology. Medical students can identify the “light bulb” sign on a DW image after 2 days of a neurology or neuroradiology rotation. I find it ironic that, after my talk promoting emergent MR imaging, experienced radiologists often say that they do not like to use MR imaging in these circumstances because it is complex, whereas CT is simple. CT is simple to perform, but MR findings are simple to interpret.

If MR imaging is to be used in these patients, how is it to be performed? We use different protocols in different clinical situations. In emergent circumstances, we use a fast protocol consisting of sagittal T1-weighted and axial DW, gradient echo, FLAIR, and MR angiography (MRA) sequences. The sequences are performed in this order to ensure that the most critical information is obtained if the examination must be prematurely terminated. The sequences are optimized for shortest acquisition time allowable for diagnostic-quality images. The sagittal T1-weighted image acts as the localizer image and allows for the detection of herniation and hyperintensity secondary to subacute hemorrhage. The axial DW sequence provides diffusion information and images obtained at a *b* value of 0 act as T2-weighted and susceptibility-weighted sequences in case the examination must be terminated after only 3 minutes. The gradient echo sequence is, next to DW imaging, the most critical sequence. It is the most sensitive technique available for detection of hyperacute hemorrhage. Several studies have documented that gradient echo images are at least as sensitive as CT for detection of hemorrhagic infarction (6,7). Animal studies, case reports (8), and my own experience make me believe that gradient echo imaging is more sensitive than CT in detecting subtle hyperacute hemorrhage. Acute hematomas are easily identified on the DW images obtained at a *b* value of 0, but these images are inadequate for detection of subtle hemorrhage encountered in some acute infarcts (9). Gradient echo imaging can also be used to detect acute embolic occlusion of major vessels, the MR equivalent of the “dense vessel sign.” Differentiation between hypointensity due to flow and hypointensity due to acute intraluminal clot is difficult with T2-weighted or FLAIR images. Spatially displaced flow-related enhancement is encountered on gradient echo images in most large patent major vessels producing hyperintensity adjacent to the dark lumen. Absence of this hyperintensity is highly suggestive of clot rather than flow. Gradient echo images also allow for detection of microbleeds that are seen as small punctate foci of hypointensity in patients with hypertensive cerebrovascular disease and cerebral amyloid angiopathy (9). FLAIR imaging allows for adequate assessment of brain parenchyma and for detection of subarachnoid hemorrhage. It is thus particularly useful in identifying stroke mimics such as inflammatory processes, mass lesions, and point-resolved spectroscopy. MRA is the least important sequence. It is, of course, helpful to know the status of the intracranial vessels, but this is not critical in the diagnosis of infarction. The neuroradiologists and stroke neurologists at our hospital are happy with this protocol, but many of our clinical colleagues prefer complete MR and MRA examinations of the head and neck when the patient’s status and system availability allow.

## Patients Who Are Candidates for Thrombolysis

The current criteria for use of intravenous thrombolytics are 1) symptom onset not exceeding 3 hours; 2) involvement of less than one-third the middle cerebral artery distribution; and 3) no CT evidence of hemorrhage. Most patients presenting with acute infarction fail to meet these criteria and therefore are not candidates for thrombolysis. It is important to note that thrombolysis is successful in approximately 75% of patients treated within 3 hours of onset of symptoms (10). Successful, that is, in the sense that there is dissolution of the clot. Success in terms of improved outcome is another matter. Thrombolysis will improve outcome only if there is salvageable brain that is critically underperfused, and the treatment does not produce hemorrhage. If we add this additional inclusion criterion, presence of salvageable brain, it would at first glance appear to shrink the pool of patients who are candidates for thrombolysis even further. But this may not be the case. The current criteria reflect the balance between the benefit of reperfusion and the risk of hemorrhage in the general population. The goal of imaging is to gain sufficient information to “customize” these general criteria to the individual patient. For instance, if we eliminate the patients who are at greatest risk for hemorrhage or those who will not benefit from reperfusion regardless of other criteria, we may find that thrombolysis on balance aids patients treated at 4 or 5 hours, thus dramatically increasing the number of patients who are candidates for thrombolysis. Perfusion imaging identifies the tissue at risk and is thus critical to management. Bolus chase perfusion imaging can be performed with either CT or MR imaging in about 30 seconds (the time it takes the blood to circulate through the brain). There is still uncertainty as to the significance of various perfusion parameters but there is a growing consensus that CBV is a good indicator of initial infarct volume while mean transit time (MTT) and cerebral blood flow (CBF) and possibly loss of flow heterogeneity are more predictive of final stroke volume. Therefore, MTT and CBF are better indicators of brain at risk than CBV.

With MR imaging, the perfusion examination is added to the end of the fast stroke study. Perfusion data are obtained for virtually the entire brain. The ischemic penumbra is the brain tissue with normal diffusion and decreased perfusion surrounding the core region of infarction (defined as the region that is hyperintense on DW images) that is likely to go on to infarction if hypoperfusion is not rapidly corrected. An additional region of mild hypoperfusion or normal perfusion with decreased hemodynamic reserve may surround the zone of abnormal perfusion. The MR perfusion maps are easy to generate and interpret. Parametric maps of cerebral blood volume, time to peak, and relative cerebral blood flow are typically produced. These maps are usually adequate for assessment of perfusion in acute infarction, but they can be inaccurate when there is diffuse or multifocal hypoperfusion, because they document differences between portions of the brain rather than absolute perfusion measurements. In addition, it is difficult to evaluate changes in perfusion on serial examinations. In patients with acute infarction, this is not an issue but in patients with spasm secondary to subarachnoid hemorrhage relative perfusion data are inadequate.

The workup of hyperacute stroke with multimodal CT provides most of the information obtained from MR imaging. Noncontrast CT is followed by CTP and CTA of the cervical and cranial vessels (11,12). CTA allows for accurate assessment of both extra- and intracranial circulation. CTP is currently limited to four sections, although the introduction of 16-row scanners should make it possible to obtain greater brain coverage. The core area of irreversible infarction is the region of hypoattenuation identified on the source images from the CTA examination. (The extent of the core infarct identified on the CT source images has been shown to correlate with core infarct identified on DW imaging [12]). Normally perfused brain enhances after contrast material administration more than hypoperfused brain, which accentuates the hypoattenuation seen on nonenhanced CT scans. If hypoattenuation on noncontrast CT scans is a reflection of hypoperfusion, the findings on multimodal CT can be viewed as a continuum. Profoundly hypoperfused infarcted brain is visible on noncontrast scans. Hypoperfusion visible on CT source images is less severe but still sufficient to cause infarction. Hypoperfusion that can be identified only on perfusion maps represents salvageable brain at risk for infarction. CTP has an advantage over MR perfusion, because it is possible to obtain absolute rather than relative perfusion values. CBF and MTT measurements are therefore more accurate with CT than MR imaging. Absolute perfusion measurements are obtained with CT by performing deconvolution computations that subtract attenuation contributions from small arteries and veins. Deconvolution is more difficult to achieve with MR imaging, in part, because susceptibility effects of contrast material in small vessels alter signal intensity in the adjacent brain. It is likely that some of the technical issues involved in generating absolute perfusion data from MR imaging will be overcome in the near future; but for the moment CTP, is superior in this regard. The multimodal CT examination provides all essential information for making therapeutic decisions (intravenous or intra-arterial thrombolysis or both), including extent of infarction, location of vascular occlusion, presence of gross hemorrhage, and status of perfusion in brain surrounding the infarct. Advocates of CT point out that the examination is faster, more easily obtained, and less expensive than MR imaging.

On the other hand, there are practical and theoretical difficulties with CT. It is true that the CT examination is faster than the MR examination (5 minutes versus 20 minutes), but at the end of acquisition the MR study can be made ready for interpretation in about 5 minutes (the time it takes to generate the perfusion maps). Much more time and expertise are required for CT. Construction of CTA images from source images has become much easier and faster in the past few years, but it still takes an experienced radiologist or specially trained technologist at least 5–10 minutes. MRA images are automatically generated and instantaneously available. Despite the theoretical advantage that absolute perfusion measures obtained with CT have over the relative perfusion measures obtained with MR imaging, it is unclear how these absolute measures aid in therapeutic decisions. The parametric maps take time to produce and quantitative interpretation is a tedious process. (At our institution, it takes a minimum of 15 minutes to construct and evaluate the perfusion maps.) Therefore the time it takes to perform the examination and generate angiograms and perfusion maps is actually longer and much more labor intensive with CT than MR imaging. In practice, the initial CT interpretation is made by evaluating the source images from both the CTA and the CTP examinations. If there is time, the parametric perfusion maps are visually inspected. Careful quantitative assessment of perfusion is frequently performed after the clinical decision has been made. Evaluation of source images requires experience and expertise. This approach works well in academic centers, but I suspect it will be less successful in the community hospital setting. The addition of CTA and CTP has not changed the fact that CT is more easily performed but MR imaging more easily interpreted.

Despite these practical limitations, it is clear that multimodal CT can be used to make therapeutic decisions concerning thrombolysis, and therefore at present it is an effective tool for assessment of hyperacute infarction. Over the next 5 years, we can expect CT to improve with of the introduction of scanners with even more detectors. (I fully expect to see a 128-row scanner described as “faster-than-light speed,” causing Albert Einstein to rise from his grave and strike us all down.) Workstation hardware and software advances will make it possible to generate high-quality CTA and CTP data with greater speed and ease. Research will determine which perfusion measures are most useful, but I suspect that these advances will not lead to a dramatic improvement in stroke outcome.

New therapeutic options are needed if we are to improve stroke outcome dramatically. To devise new therapies, we need more than just perfusion information. We need to know about the status of the brain itself. MR imaging already provides this information. DW and T2-weighted sequences allow for the direct and independent assessment of cytotoxic and vasogenic edema, respectively. The simple scheme of a core area of irreversible infarction surrounded by a penumbra of hypoperfusion was developed in animal models. In practice, human strokes are much more complex. The value of diffusion information is sometimes downplayed in the discussion of stroke treatment, because diffusion abnormalities usually reflect irreversible brain damage. Complete reversal of diffusion abnormalities is a rare phenomenon that occurs in fewer than 5% of infarcts, but diffusion information still may affect outcome and therefore therapy. It is common to describe DW examinations as if they were pregnancy tests, either unequivocally positive or negative, but there is a difference. Whereas a woman cannot be a “little bit pregnant,” a brain *can be* a little bit infarcted. There is a continuum of diffusion abnormalities seen in patients with infarction. To be sure, most infarcts reveal severe apparent diffusion coefficient (ADC) decreases (> 70%), but in some cases ADC changes are less marked. Reversible diffusion abnormalities occur most frequently when initial ADC reductions are mild (13). The severity of diffusion changes within an infarct correlates with clinical outcome independent of infarct size. In clinical practice, serial DW examinations reveal significant variation in both the initial ADC and subsequent evolution of diffusion changes. I have found that initial signal intensity on DW images (indicative of cytotoxic edema) is variable, and over the first few days, it is not uncommon to see heterogeneous changes with increasing signal intensity in some portions of the infarct, and on occasion, rapid resolution of hyperintensity in other regions. What does it mean when an infarct has only mild diffusion changes? It probably reflects heterogeneity within the “core” infarct with a mixture of living and dead (or dying) cells. This heterogeneity explains the fact that transient ischemic attacks may produce mild but persistent decreases in ADC (undetectable on DW images) in the affected portion of the brain (14). Presumably a small number of cells are destroyed, enough to lower ADC but not enough to produce permanent neurological damage. Therefore, the “core” infarct may contain viable cells. Therapies aimed at protecting these viable cells or interrupting the events leading from cell damage to cell death could improve outcome. Hyperintensity on FLAIR images (indicative of vasogenic edema) also varies depending on time of appearance and severity. This variation depends upon restoration of flow to vessels with damaged endothelial cells. Assessment of the integrity of the endothelium (possibly by combining perfusion and T2 information) might therefore help to predict the likelihood of hemorrhagic transformation. Proponents of CT could point out that all of this speculation is very interesting, but at least for the moment it is of no clinical relevance and they would be right.

Or, would they?

The story is not quite over. In every space opera there is a hook—a “Luke, I am your father” moment—which alters everything and sets up the next installment. In our little story, that hook is hemorrhage. Gradient echo images can reveal hemorrhagic transformation of infarction not detectable on CT scans. Current criteria for thrombolysis are based on CT evidence of hemorrhage. Should those criteria be extended to include hemorrhage invisible on CT scans but depicted on MR images? Gradient echo images also reveal the presence of microbleeds, small punctate foci of chronic hemorrhage seen in patients with hypertension and cerebral amyloid angiopathy. Microbleeds imply vascular fragility and are associated with an increased risk for infarction, hemorrhagic infarction, and frank hematomas (15). It is possible that patients with microbleeds should not be candidates for thrombolysis or at least intravenous thrombolysis. This all seems like bad news, and another reason to render someone ineligible for the only treatment known to improve outcome in stroke. Although this may be true, it is possible that elimination of patients most at risk for hemorrhage will allow for more aggressive treatment of the remaining patients. We also need to develop a less defeatist attitude toward hemorrhage. Therapies aimed at limiting hemorrhage and its negative consequences (eg, edema) might allow for improved outcome. In any event, if it turns out that the detection of these hemorrhages is therapeutically important, the debate will be over and MR imaging will have “won.” Future investigations will surely ask and answer these questions.

So stay tuned for *Stroke Wars, Episode V: Attack of the Microbleeds.*

## References

1.  Mullins ME, Schaefer PW, Sorenson AG, et al. **CT and conventional and diffusion-weighted imaging in acute stroke: study in 691 patients at presentation in the emergency department.** Radiology 2002;224:353–360
    
    [PubMed](http://www.ajnr.org/lookup/external-ref?access_num=12147827&link_type=MED&atom=%2Fajnr%2F25%2F8%2F1304.atom) 
    
    [Web of Science](http://www.ajnr.org/lookup/external-ref?access_num=000177025400010&link_type=ISI) 

2.  Koenig M, Klotz E, Luka B, et al. **Perfusion CT of the brain: diagnostic approach for early detection of ischemic stroke.** Radiology 1998;209:85–93
    
    [PubMed](http://www.ajnr.org/lookup/external-ref?access_num=9769817&link_type=MED&atom=%2Fajnr%2F25%2F8%2F1304.atom) 
    
    [Web of Science](http://www.ajnr.org/lookup/external-ref?access_num=000076024700018&link_type=ISI) 

3.  Roberts HC, Dillon WP, Furlan AJ, et al. **Computed tomographic findings in patients undergoing intra-arterial thrombolysis for acute ischemic stroke due to middle cerebral artery occlusion: results from the PROACT II trial.** Stroke 2002;33:1557–1565
    
    [Abstract/FREE Full Text](http://www.ajnr.org/lookup/ijlink/YTozOntzOjQ6InBhdGgiO3M6MTQ6Ii9sb29rdXAvaWpsaW5rIjtzOjU6InF1ZXJ5IjthOjQ6e3M6ODoibGlua1R5cGUiO3M6NDoiQUJTVCI7czoxMToiam91cm5hbENvZGUiO3M6OToic3Ryb2tlYWhhIjtzOjU6InJlc2lkIjtzOjk6IjMzLzYvMTU1NyI7czo0OiJhdG9tIjtzOjIwOiIvYWpuci8yNS84LzEzMDQuYXRvbSI7fXM6ODoiZnJhZ21lbnQiO3M6MDoiIjt9) 

4.  Fiebach JR, Schellinger PD, Jansen O, et al. **CT and diffusion-weighted MR imaging in randomized order: diffusion-weighted imaging results in higher accuracy and lower interrater variability in the diagnosis of hyperacute ischemic stroke.** Stroke 2002;33:2206–2210
    
    [Abstract/FREE Full Text](http://www.ajnr.org/lookup/ijlink/YTozOntzOjQ6InBhdGgiO3M6MTQ6Ii9sb29rdXAvaWpsaW5rIjtzOjU6InF1ZXJ5IjthOjQ6e3M6ODoibGlua1R5cGUiO3M6NDoiQUJTVCI7czoxMToiam91cm5hbENvZGUiO3M6OToic3Ryb2tlYWhhIjtzOjU6InJlc2lkIjtzOjk6IjMzLzkvMjIwNiI7czo0OiJhdG9tIjtzOjIwOiIvYWpuci8yNS84LzEzMDQuYXRvbSI7fXM6ODoiZnJhZ21lbnQiO3M6MDoiIjt9) 

5.  Mullins ME, Lev MH, Schellingerhout D et al. **Influence of availability of clinical history on detection of early stroke using unenhanced CT and diffusion-weighted MR imaging.** AJR Am J Roentgenol 2002;179:223–228
    
    [PubMed](http://www.ajnr.org/lookup/external-ref?access_num=12076941&link_type=MED&atom=%2Fajnr%2F25%2F8%2F1304.atom) 

6.  Linfante I, Llinas RH, Caplan LR, Warach S. **MRI features of intracerebral hemorrhage within 2 hours from symptom onset.** Stroke 1999;30:2263–2267
    
    [Abstract/FREE Full Text](http://www.ajnr.org/lookup/ijlink/YTozOntzOjQ6InBhdGgiO3M6MTQ6Ii9sb29rdXAvaWpsaW5rIjtzOjU6InF1ZXJ5IjthOjQ6e3M6ODoibGlua1R5cGUiO3M6NDoiQUJTVCI7czoxMToiam91cm5hbENvZGUiO3M6OToic3Ryb2tlYWhhIjtzOjU6InJlc2lkIjtzOjEwOiIzMC8xMS8yMjYzIjtzOjQ6ImF0b20iO3M6MjA6Ii9ham5yLzI1LzgvMTMwNC5hdG9tIjt9czo4OiJmcmFnbWVudCI7czowOiIiO30=) 

7.  Schellinger PD, Jansen, Fiebach JB, et al. **A standardized MRI stroke protocol: comparison with CT in hyperacute intracerebral hemorrhage.** Stroke 1999;30:765–768
    
    [Abstract/FREE Full Text](http://www.ajnr.org/lookup/ijlink/YTozOntzOjQ6InBhdGgiO3M6MTQ6Ii9sb29rdXAvaWpsaW5rIjtzOjU6InF1ZXJ5IjthOjQ6e3M6ODoibGlua1R5cGUiO3M6NDoiQUJTVCI7czoxMToiam91cm5hbENvZGUiO3M6OToic3Ryb2tlYWhhIjtzOjU6InJlc2lkIjtzOjg6IjMwLzQvNzY1IjtzOjQ6ImF0b20iO3M6MjA6Ii9ham5yLzI1LzgvMTMwNC5hdG9tIjt9czo4OiJmcmFnbWVudCI7czowOiIiO30=) 

8.  Packard AS, Kase CS, Aly AS, Barest GD. **“Computed tomography–negative” intracerebral hemorrhage: case report and implications for management.** Arch Neurol 2003;60:1156–1159
    
    [CrossRef](http://www.ajnr.org/lookup/external-ref?access_num=10.1001/archneur.60.8.1156&link_type=DOI) 
    
    [PubMed](http://www.ajnr.org/lookup/external-ref?access_num=12925375&link_type=MED&atom=%2Fajnr%2F25%2F8%2F1304.atom) 
    
    [Web of Science](http://www.ajnr.org/lookup/external-ref?access_num=000184692600020&link_type=ISI) 

9.  Lin DM, Filippi CG, Steever AB, Zimmerman RD. **Detection of intracranial hemorrhage: comparison between gradient-echo images and b0 images obtained from diffusion-weighted echo-planar sequences.** AJNR Am J Neuroradiol 2001;22:1275–1281
    
    [Abstract/FREE Full Text](http://www.ajnr.org/lookup/ijlink/YTozOntzOjQ6InBhdGgiO3M6MTQ6Ii9sb29rdXAvaWpsaW5rIjtzOjU6InF1ZXJ5IjthOjQ6e3M6ODoibGlua1R5cGUiO3M6NDoiQUJTVCI7czoxMToiam91cm5hbENvZGUiO3M6NDoiYWpuciI7czo1OiJyZXNpZCI7czo5OiIyMi83LzEyNzUiO3M6NDoiYXRvbSI7czoyMDoiL2FqbnIvMjUvOC8xMzA0LmF0b20iO31zOjg6ImZyYWdtZW50IjtzOjA6IiI7fQ==) 

10. Molina CA, Alvarez-Sabín J, Montaner J, et al. **Thrombolysis-related hemorrhagic infarction: a marker of early reperfusion, reduced infarct size, and improved outcome in patients with proximal middle cerebral artery occlusion.** Stroke 2002;33:1551–1556
    
    [Abstract/FREE Full Text](http://www.ajnr.org/lookup/ijlink/YTozOntzOjQ6InBhdGgiO3M6MTQ6Ii9sb29rdXAvaWpsaW5rIjtzOjU6InF1ZXJ5IjthOjQ6e3M6ODoibGlua1R5cGUiO3M6NDoiQUJTVCI7czoxMToiam91cm5hbENvZGUiO3M6OToic3Ryb2tlYWhhIjtzOjU6InJlc2lkIjtzOjk6IjMzLzYvMTU1MSI7czo0OiJhdG9tIjtzOjIwOiIvYWpuci8yNS84LzEzMDQuYXRvbSI7fXM6ODoiZnJhZ21lbnQiO3M6MDoiIjt9) 

11. Ezzeddine MA, Lev MH, McDonald CT. **CT angiography with whole brain perfused blood volume imaging: added clinical value in the assessment of acute stroke.** Stroke 2002;33:959–966
    
    [Abstract/FREE Full Text](http://www.ajnr.org/lookup/ijlink/YTozOntzOjQ6InBhdGgiO3M6MTQ6Ii9sb29rdXAvaWpsaW5rIjtzOjU6InF1ZXJ5IjthOjQ6e3M6ODoibGlua1R5cGUiO3M6NDoiQUJTVCI7czoxMToiam91cm5hbENvZGUiO3M6OToic3Ryb2tlYWhhIjtzOjU6InJlc2lkIjtzOjg6IjMzLzQvOTU5IjtzOjQ6ImF0b20iO3M6MjA6Ii9ham5yLzI1LzgvMTMwNC5hdG9tIjt9czo4OiJmcmFnbWVudCI7czowOiIiO30=) 

12. Schramm, Schellinger PD, Jochen B, et al. **Comparison of CT and CT angiography source images with diffusion-weighted imaging in patients with acute stroke within 6 hours after onset.** Stroke 2002;33:2426–2432
    
    [Abstract/FREE Full Text](http://www.ajnr.org/lookup/ijlink/YTozOntzOjQ6InBhdGgiO3M6MTQ6Ii9sb29rdXAvaWpsaW5rIjtzOjU6InF1ZXJ5IjthOjQ6e3M6ODoibGlua1R5cGUiO3M6NDoiQUJTVCI7czoxMToiam91cm5hbENvZGUiO3M6OToic3Ryb2tlYWhhIjtzOjU6InJlc2lkIjtzOjEwOiIzMy8xMC8yNDI2IjtzOjQ6ImF0b20iO3M6MjA6Ii9ham5yLzI1LzgvMTMwNC5hdG9tIjt9czo4OiJmcmFnbWVudCI7czowOiIiO30=) 

13. Schaefer PW Hassankhani A, Koroshetz WJ, et al. **Partial reversal of DWI abnormalities in stroke patients undergoing thrombolysis: evidence of DWI and ADC thresholds.** Stroke 2002;33:357–364
    
    

14. Kamal AK, Segal AZ, Ulug AM. **Quantitative diffusion-weighted MR imaging in transient ischemic attacks.** AJNR Am J Neuroradiol 2002;23:1533–1538
    
    [Abstract/FREE Full Text](http://www.ajnr.org/lookup/ijlink/YTozOntzOjQ6InBhdGgiO3M6MTQ6Ii9sb29rdXAvaWpsaW5rIjtzOjU6InF1ZXJ5IjthOjQ6e3M6ODoibGlua1R5cGUiO3M6NDoiQUJTVCI7czoxMToiam91cm5hbENvZGUiO3M6NDoiYWpuciI7czo1OiJyZXNpZCI7czo5OiIyMy85LzE1MzMiO3M6NDoiYXRvbSI7czoyMDoiL2FqbnIvMjUvOC8xMzA0LmF0b20iO31zOjg6ImZyYWdtZW50IjtzOjA6IiI7fQ==) 

15. Nighoghossian N, Hermier M, Adeleine P, et al. **Old microbleeds are a potential risk factor for cerebral bleeding after ischemic stroke: a gradient-echo T2*-weighted brain MRI study.** Stroke 2002;33:735–742
    
    [Abstract/FREE Full Text](http://www.ajnr.org/lookup/ijlink/YTozOntzOjQ6InBhdGgiO3M6MTQ6Ii9sb29rdXAvaWpsaW5rIjtzOjU6InF1ZXJ5IjthOjQ6e3M6ODoibGlua1R5cGUiO3M6NDoiQUJTVCI7czoxMToiam91cm5hbENvZGUiO3M6OToic3Ryb2tlYWhhIjtzOjU6InJlc2lkIjtzOjg6IjMzLzMvNzM1IjtzOjQ6ImF0b20iO3M6MjA6Ii9ham5yLzI1LzgvMTMwNC5hdG9tIjt9czo4OiJmcmFnbWVudCI7czowOiIiO30=) 

*   Copyright © American Society of Neuroradiology