In the weeks leading up to the 2007 International Stroke Conference, stroke imaging experts could have been preparing to debate such cutting-edge topics as whether MRI or CT is best suited for the ischemic mismatch imaging techniques that one day will routinely be used to select patients for thrombolytic therapy.
In the weeks leading up to the 2007 International Stroke Conference, stroke imaging experts could have been preparing to debate such cutting-edge topics as whether MRI or CT is best suited for the ischemic mismatch imaging techniques that one day will routinely be used to select patients for thrombolytic therapy. But 2 weeks before the February conference, the stroke imaging community instead was buzzing about a Jan. 27 Lancet study1 that was decidedly free of bells and whistles: a 356-patient single-site prospective comparison in which conventional gradient-echo and diffusion-weighted MRI outperformed unenhanced CT for the diagnosis of patients with suspected acute stroke.
Since noncontrast CT-based acute stroke diagnosis has become the norm in all but the most progressive centers, the financial and logistical barriers to augmenting or replacing frontline CT with MRI are formidable. Even so, the results of one clinical trial after another suggest the next generation of stroke therapies will almost certainly depend on penumbral imaging, for which researchers have for the most part favored perfusion-diffusion MRI techniques over CT perfusion. For physicians wary of making that philosophical leap, the results of the Lancet study, funded by the National Institute of Neurological Disorders and Stroke, may represent an important intermediate step.
Much of the attention given that study focused on what it did not do. It did not address the issue of penumbral imaging. It acknowledged but did not attempt to solve the issues of hospital logistics that often limit acute stroke patients' access to MRI in even the most elite academic institutions. Because patients who were too medically unstable to undergo MRI were excluded from the study, it did not fully address the issue of patients with severe stroke, in whom rapid diagnosis and triage is most critical. All of these factors must be considered in determining the most effective role for each imaging modality in stroke diagnosis and management.
But what the Lancet study did do was address the issue of MRI for acute stroke at its most basic level. The only stroke therapies currently approved by the FDA-intravenous tissue plasminogen activator within 3 hours of symptom onset and the Merci Retreiver mechanical embolectomy device in patients who are ineligible for intravenous tPA-won their approvals based on clinical trials in which the only imaging technique used in patient selection was unenhanced CT to rule out intracranial hemorrhage.2,3
Other research4-8 has suggested that diffusion-weighted MRI is not only equally capable of identifying hemorrhage but also more accurate for diagnosing acute ischemic stroke, particularly small lesions that are easily missed on CT. But no study had done so in a prospective manner, in a large sample of patients representative of populations seen in clinical practice, and with MRI performed prior to CT in the majority of patients (which theoretically biases the results in favor of CT, because infarcts are more easily detected over time).
Although the pivotal trial of tPA2 was published more than a decade ago, before diffusion-weighted and gradient-echo MR imaging were common practice, it essentially established unenhanced CT as the sole imaging prerequisite for intravenous tPA therapy.
By showing that MRI can rule out intracranial hemorrhage as effectively as unenhanced CT, within the same 3-hour window, the Lancet study challenged this premise. In addition, MRI identified acute ischemic stroke in 46% of 356 patients with suspected acute stroke, compared with 10% identified using unenhanced CT; in 90 patients scanned within 3 hours of symptom onset, acute ischemic stroke was detected in 46% using MRI and only 7% using CT.
"We have a method that for ischemic stroke is five times more effective than the so-called gold standard," said Steven Warach, MD, PhD, senior investigator and chief of stroke diagnostics and therapeutics at the NINDS and an author of the Lancet study. "MRI is more expensive, and hospitals have not made it as accessible as CT. But those are solvable problems. The obstacles are not technical or medical, they're more logistical."
The logistical obstacles, to be sure, are not trivial. Emergency departments in most community hospitals and even many academic centers do not have easy access to an MR scanner, because the scanner is on another floor, it is too often in use by other patients who cannot be rescheduled, or round-the-clock staffing is not available in the event that a patient arrives at 2 a.m. CT also has advantages for patients with contraindications to MRI or those who are medically unstable. About 75% of stroke patients at Massachusetts General Hospital present with minor strokes, according to unpublished data from the hospital. These respond well to intravenous tPA and may not require more sophisticated imaging.
Still, the remaining 25% of patients with severe stroke account for the majority of deaths, morbidity, and costs-up to 10 times more than minor strokes in hospital costs alone, said R. Gilberto Gonzalez, MD, PhD, a professor of radiology at Harvard Medical School and chief of neuroradiology at Massachusetts General Hospital.
Anticipating that those high-risk patients will benefit most from therapies that require MRI-based selection makes investing in MRI seem like a good idea at places like the University of California, San Diego, where radiology department chair William G. Bradley Jr., MD, boasts that two 3-Tesla scanners are being installed adjacent to emergency departments specifically for use on stroke patients.
"The future really is with MR," Bradley said. "CT is going to be the workhorse, but I think we'll learn a lot more with MRI."
MRI-based diagnosis of acute intracranial hemorrhage is based on the so-called deoxyhemoglobin border sign,9,10 a hypointense rim around an area of hyperintensity (well visualized on gradient-echo images), typically surrounded by edema (Figure 1). The Lancet study found no significant difference between CT and MRI for detecting acute intracranial hemorrhage, with 25 positive diagnoses made using CT (along with 3 false negatives) and 23 positive diagnoses made with MRI (4 false negatives).
These results are consistent with those of the Hemorrhage and Early MR Imaging Evaluation (HEME study) published by NINDS and University of California, Los Angeles, researchers in 2004,11 in which MR and CT were found to be equivalent for the diagnosis of acute intracerebral hemorrhage; in 4 cases, acute hemorrhagic transformation was seen on MRI but not on CT. That same year, using CT as the gold standard for comparison, researchers from the University of Heidelberg reported that MRI was 100% sensitive and accurate for intracerebral hemorrhage in the hands of experienced readers and 95% sensitive when read by medical students.12
The relative sensitivity of MRI and CT for intracranial hemorrhage is something that radiologists at UCSD have been tracking since MRI became accessible to the emergency department in the summer of 2003. But despite finding in favor of MRI in that regard, Bradley acknowledges that the circumstances under which intravenous tPA was approved present a potential liability issue for anyone who uses MRI alone to screen patients before administering the therapy.
"You pretty much have to do CT today," he said. "We've shown we can exclude hemorrhage very well with MR, but you run some risk if there's a problem."
Even research demonstrating the diagnostic equivalence of MRI and CT for ruling out hemorrhage may not fully address this issue, in the absence of a revised clinical trial of intravenous tPA using diffusion-weighted MRI alone instead of unenhanced CT alone. Such a revisiting of the original clinical trial protocol is unlikely, primarily because the emergence of perfusion-based penumbral imaging has convinced clinicians and researchers that the existence of salvageable tissue-not the number of hours from symptom onset, which was a key inclusion criterion in the 1995 tPA trial-should determine whether a patient receives thrombolytic therapy. And although both modalities can be used to generate the perfusion maps on which penumbral estimates are based, MR-based perfusion techniques have thus far spent the most time in the research spotlight.
The concept of the ischemic penumbra, initially studied in monkeys and confirmed using PET in humans,13 is based on the pattern of cerebral blood flow reduction that results from the ischemic event. Severely diminished levels of cerebral blood flow characterize the irreversibly infarcted core and differentiate it from the ischemic penumbra, or region of potentially salvageable tissue, which also experiences reduced blood flow but can be sustained for a number of hours by collateral circulation. During this time, thrombolytic therapy is most likely to be effective.
MRI-based penumbral imaging, which has emerged as a surrogate for the clinically impractical PET, uses diffusion-weighted imaging (DWI) techniques to define the ischemic core and perfusion-weighted imaging (PWI) techniques to define the extent of cerebral blood flow reduction (Figure 2). The mismatch between the 2, or the region between the perimeter of the core and the perimeter of the hypoperfused area, represents the ischemic penumbra.
Because penumbral imaging has shown that mismatch persists for many hours after the 3-hour window for intravenous tPA has closed, multiple clinical trials of new therapeutic options14-17 and expansion of the tPA window18,19 are testing the concept of treatment based on MRI-determined penumbra rather than time from symptom onset (see accompanying article). If these physiology-based trials lead to FDA approvals, a much higher percentage of stroke patients would be eligible for treatment than the 3% to 8.5% who currently receive intravenous tPA.20 And if selection criteria for the approved therapies require diffusion and perfusion imaging, that would almost certainly mean a much bigger role for MRI in acute stroke management.
"It'll change the whole imaging paradigm," said Anthony J. Furlan, MD, director of the vascular neurology program and associate director of the cerebrovascular center at The Cleveland Clinic.
To be sure, MRI-based penumbral imaging isn't perfect. Unlike PET, MRI provides an estimate rather than a true measure of blood flow. Coupled with the heterogeneous distribution of ischemia, this means that diffusion MRI may in fact overestimate the amount of irreversibly infarcted tissue and perfusion MRI may overestimate the extent of hypoperfusion.
"Mismatch is a very simplistic view of acute cerebral ischemia," said David Liebeskind, MD, neurology director of stroke imaging and associate neurology director of the UCLA Stroke Center. "It's only representative of the moment in time when you happen to be injecting the contrast media. So doing a diffusion-perfusion study at 1 point is not necessarily going to tell you what's going to happen 3 hours later. When you add that no 2 patients are identical, it becomes difficult to classify or stratify based on exact thresholds of mismatch."
Perhaps not surprisingly, questions abound as to which of the many perfusion-weighted imaging parameters provide the best estimate and what degree of mismatch should define penumbra. In a review of 11 papers describing DWI-PWI findings in patients with acute stroke, researchers from the University of Edinburgh identified 5 different definitions of mismatch and 7 different methods for measuring perfusion.21 In their own 46-patient study, the same researchers also found that approximately half of patients without DWI-PWI mismatch experienced lesion growth (11 of 21 when cerebral blood flow was used as the perfusion parameter, and 5 of 13 when mean transit time was used), suggesting that MRI-based penumbral imaging does not identify all patients who might benefit from stroke therapy.22
"Most studies of thrombolysis and mismatch did not include patients without mismatch, so we do not know whether they will benefit from thrombolysis or not," said Joanna M. Wardlaw, MD, a professor of neuroradiology at the University of Edinburgh and coauthor of both studies.
Researchers are continuing to refine the techniques used to define DWI-PWI mismatch. A multivariate voxel-based model developed at UCLA, which achieved 81% accuracy in initial reports,23 is currently being incorporated into the MR-RESCUE (MR and recanalization of stroke clots using embolectomy) study.17 Marc Fisher, MD, a professor of neurology at the University of Massachusetts Medical School, is among those leading efforts to assess the feasibility in animal models of using arterial spin labeling perfusion to develop probability density maps.
Despite the limitations of DWI-PWI mismatch and the lack of consensus as to how precisely it should be used, however, stroke physicians do agree that therapy based on a penumbral estimate-even an imperfect estimate-is still preferable to the current time-dependent paradigm.
"We're trying to get information that helps us decide whether to treat somebody," Gonzalez said. "We just need to know that there's a major penumbra. To be practical, it has to be easy to evaluate. Is it going to be scientifically perfect? No. But you don't need that."
Studies have shown perfusion CT to be comparable to diffusion-perfusion MRI for defining ischemic penumbra24,25 (Figure 3). Perfusion CT is being included as a selection tool in the second installment of the phase III Desmoteplase in acute ischemic stroke (DIAS-2) trial.
Like its MRI counterpart, CT perfusion provides only an estimate of ischemic penumbra. Interpretation of CT perfusion maps is generally thought to be more technically difficult than that of diffusion- and perfusion weighted images. A 2004 study from the Barrow Neurological Institute found postprocessing of CT perfusion data by 3 experienced technologists resulted in less reproducibility than the authors felt would be necessary for clinical practice.26
"CT perfusion is not as easy as one would think," said Furlan, who has been an investigator with multiple clinical trials, including DIAS-2. "We see some variability between centers."
Still, the fact that CT is far more accessible than MRI in most emergency rooms means that the future of stroke imaging will likely include penumbral imaging using both modalities.
"What I envision is that smaller hospitals will do CT perfusion, and bigger hospitals will do MRI," Fisher said.
He noted that telemedicine may also come to play an important role when the CT perfusion images obtained at smaller centers that have limited stroke expertise need to be interpreted.
That scenario is essentially an extension of the "drip and ship" model used in many areas, where CT is done and intravenous tPA initiated within the 3-hour window at a small hospital, while ineligible patients and severe cases-which represent approximately 25% of stroke patients, in whom intravenous tPA is unlikely to be effective-are transported to a major stroke center for MR and angiographic imaging and intra-arterial thrombolysis or mechanical embolectomy.
But if the current evolution in ambulance triage protocols unfolds in such a way that patients with severe strokes are taken directly to major stroke centers, the potential for utilization of penumbral imaging-and improved patient outcomes-will be that much greater.
"To me, this is the key to the future of stroke care," Fisher said. "We need to increase the number of patients we treat, which means we need to treat them later after onset, and it doesn't do any good to treat them after the tissue is gone."
Jordana Bieze Foster is a freelance writer based in Masschusetts and a frequent contributor to Applied Neurology.