Diffusion-weighted imaging (DWI) MRI is significantly more accurate than conventional noncontrast computed tomography (CT) for diagnosing acute ischemic stroke (AIS) within the first 12 hours after symptom onset, according to a new evidence-based guideline from the AAN.
No Caption Available.
Subcommittee member Steven Warach, MD, PhD, chief of the Section on Stroke Diagnostics and Therapeutics at the NINDS, reviewed the findings with Neurology Today.
One of our primary duties as physicians is to make the correct diagnosis, because all subsequent decisions about patient management follow from our diagnosis. Typically, patients with acute neurological impairment undergo routine noncontrast CT imaging to evaluate for infarct and exclude hemorrhage, as well as other possible structural lesions that can mimic stroke. However, we found that the collective evidence in the medical literature shows that DWI is more accurate and superior in other ways to CT. Of course, the availability, cost, and staffing requirements for MRI enter into decisions about which test to perform during the acute period, but hopefully, this parameter will help guide such decisions.
DWI was found to be superior to CT in patients presenting within 12 hours of symptom onset, on the basis of one Class I study and three Class II studies, the two highest ratings for evaluating such studies.
The Class I study was a prospective blinded trial covering a full range of patients presenting to an emergency department with a suspicion of acute stroke — they took all comers, the full range of patients, including those who turned out to be stroke mimics. In 221 patients scanned within 12 hours of onset, the MRI readers correctly diagnosed acute ischemic stroke four times more often than those screening CT scans – in 94 patients versus 22. Similar differences were also seen in a subset of 90 patients scanned within 3 hours of onset. Relative to final hospital discharge diagnosis recorded by the treating vascular neurologist in the medical records — and based on the available follow-up imaging and clinical course — the sensitivity, specificity, and accuracy of DWI was 77 percent, 96 percent, and 86 percent, respectively. For CT, it was 16 percent, 97 percent, and 55 percent, respectively, so you can see the difference.
Early signs of an AID are often too subtle to be easily detected by CT and interpreted by readers, except the stroke specialist or neuroradiologist. With DWI, these signs show up earlier and much more clearly — almost like a light bulb, as one colleague puts it, and are easier to detect for specialists and non-specialists.
Historically, the availability, cost, and staffing requirements for MRI have been cited as reasons why that test may not be performed during the acute period. Most hospitals have the required MRI capability. The problem is that many hospitals are reluctant to use MRI, having established protocols that rely on CT as a front-line screening tool, fully scheduled MRI scanners, and MRI units and trained personnel are not made immediately available for emergencies or after hours. MRI is more expensive than a CT scan, but the cost of routine MRI is not as prohibitive as it might seem. A very common practice in stroke care is that an initial CT scan is followed by a later MRI scan. One MRI would obviate the need for a CT, and therefore be less expensive than doing both.
All of these are all solvable problems. I have worked at hospitals where there has been a shift from routine CT screening to DWI MRIs in stroke diagnosis, and it was not that difficult, provided there is a willingness to change screening practices. There is an emerging literature supporting the cost effectiveness of MRI for treating stroke and more research needs to be done to assess that. Processes need to be put in place to permit direct, immediate access of patients from emergency departments to MRI scanners, and hopefully, this parameter will help guide such changes.
For potential candidates for thrombolytic therapy, time is critical and brain imaging must be immediately available. In increasing numbers of hospitals, MRI is immediately available for these cases, and there is evidence that MRI may be more cost-effective than CT even for thrombolytic decision-making. A recent publication in Neurology from a major US stroke center reported that 14 percent of tPA-treated patients, screened with CT, proved to be non-stroke upon further work-up. Although the risk of administering thrombolytics to stroke mimics appears to be low, a screening MRI is likely to minimize treatment of non-stroke with tPA. However, if a patient has a medical contraindication to MRI, which may be 10-15 percent of patients, or MRI is not immediately available, then CT must be performed to screen for thrombolysis decision-making. For the majority of stroke patients, however, who are not thrombolytic candidates, delays to wait for the MRI to become available would be justifiable.
Finding good comparative data on PWI was far more difficult. All of the published studies were Class IV trials, so the data are not nearly as solid as with DWI. The need for better studies is an important direction for future research. Bear in mind that PWI shows a different aspect of stroke pathology that practitioners may find useful for diagnosis and management. It shows focal ischemia in the brain by flagging areas with an absence or drastic reduction of blood flow. At this point there are not enough reliable data to adequately evaluate the utility of PWI or make a recommendation.
The true sensitivity of DWI for diagnosing AIS is not 100 percent, but probably between 85 percent and 95 percent. Many of the initial case series and small CT comparative studies reported a near 100 percent sensitivity for DWI in the hyperacute stage of stroke in highly selected subsets of patients known stroke, however, cases of DWI-negative stroke were increasingly reported, especially for very early stroke, and patients with mild deficits (NIH Stroke Scale less than 4) and brainstem locations. But if you take all of the studies we reviewed collectively, maybe 5 percent of true AIS cases were negative on MRI. There may be also false positives in MRIs because there can be image artifacts on DWI that could be mistaken for small regions of AIS, especially in the early hours after a stroke when the light bulb is still a little dim, but that was less common than false negatives. It is important to remember that while brain imaging is an essential component of acute stroke work-up, it is not the final word; the clinician integrates a careful history and physical exam along with the scan results to arrive at the clinical diagnosis that will guide treatment decisions.
We found that baseline DWI volume probably predicts baseline clinical stroke severity and final lesion volume in anterior-circulation stroke syndromes, and is possibly accurate in predicting clinical outcome as measured by the NIH Stroke Scale (NIHSS) and Barthel Index — but not baseline NIHSS scores in posterior-circulation stroke syndromes. Baseline PWI volume possibly predicts baseline clinical stroke severity, but more study is needed. Some studies have reported on the use of the combination of DWI and PWI to identify an ischemic penumbra, a region of at-risk but potentially salvageable brain, but more study is required to determine its value in predicting outcome or response to recanalization therapy.
The sensitivity of CT varies, depending on the imaging features of infarction, stroke severity, exam time from clinical onset, study population, and other variables. Sensitivity estimates range from 12 percent to 92 percent, with an overall estimate of 40 percent to 60 percent for the six-hour time window. A post-hoc analysis of CT data from the NINDS tPA study yielded a 31 percent sensitivity for early infarct signs.
DWI-positive scans in TIA are common, and acute ischemic DWI lesions are present in 40.1 percent of patients with the clinical diagnosis of a TIA, according to some studies, a finding that correlates with symptom duration. Use of DWI has been a major impetus to redefine transient clinical deficits as stroke on the basis of DWI evidence of cerebral infarction. Only one of the TIA studies involved DWI performed within 24 hours of symptom onset, but a recent study estimated the epidemiologic impact of DWI-based diagnosis would result in a reduced annual TIA incidence of 33 percent, and an increase in stroke incidence of 7 percent in the United States as a result of a more accurate classification.
Although the experienced reader can discriminate ischemic from hemorrhagic stroke on DWI, the diagnosis of ICH by MRI is best done by gradient echo (GRE) pulse sequences. We did not specifically address intracerebral hemorrhage or hemorrhagic stroke in this assessment. However, there is good evidence in the literature supporting the equivalence of GRI and CT for acute ICH detection. With regard to DWI and ICH detection, one prospective multicenter study evaluated the ability of DWI to discriminate between ischemic and hemorrhagic strokes in 62 patients assessed within six hours with DWI, with a mean scan time of three hours. Baseline CT, follow-up imaging, and each patient's clinical course were then used to validate the diagnosis of ischemic stroke or ICH. Three blinded raters independently rated all scans, and they were able to correctly identify all ischemic strokes and ICHs. Because the aim of that study was to evaluate whether MRI with DWI could differentiate ischemic stroke from ICH within the first six hours, the spectrum of patients and controls was narrow.
Although DWI is established as superior to CT for AIS detection, my opinion is that DWI plus GRE would be the minimum MRI requirement to replace non-contrast CT for acute stroke screening. For either MRI or CT, the addition of vascular imaging and perfusion imaging to the exam are common in clinical practice, but further studies are needed to adequately define and make recommendations about their specific, respective roles in acute stroke diagnosis and management.
This question goes beyond the scope of this assessment, but from my knowledge of the literature I know there would be inadequate data to give an evidence based recommendation.
Vascular imaging (CTA or MRA) add an important dimension to understanding the individual patient pathology, however they are only adequate for showing relevant occlusions of the larger intracranial vessels (for example, MCA M1 or M2 segments, distal ICA, basilar), and that is a minority of all strokes.
There has been recent attention quantifying the increased cancer risk with CT and CTA, and some patients with renal failure are contraindicated for CTA because of the risk of contrast induced nephropathy — CTA requires contrast injection.
Some practitioners use CTA source images as a “poor man's DWI”; it is actually showing a kind of poor brain perfusion (low cerebral blood volume) that predicts infarction. But again only for larger strokes, so saying it is “as informative” as MRI is not quite correct. However, for large, severe strokes one could argue that it may be as informative as MRI (again, there is no hard evidence to support this, just expert opinion); but those are the minority of strokes. On the whole CTA does provide valuable vascular information, but is not equivalent to MRI; and does have its risks.
Schellinger PD, Bryan RN, Warach S, et al. Evidence-based guideline: The role of diffusion and perfusion MRI for the diagnosis of acute ischemic stroke. Report of the therapeutics and Technology Assessment Subcommittee of the American Academy of Neurology. 2010;75: 177–186.