Diffusion-weighted imaging (DWI) and perfusion-weighted imaging (PWI) are sensitive imaging techniques that have only recently become clinically available for the detection of acute ischemic stroke. DWI is based on the microscopic motion of water molecules (diffusion) and can evaluate the extent and location of ischemic lesions within minutes. PWI evaluates blood flow to the brain parenchyma and can detect microvascular perfusion abnormalities. In combination, these techniques may identify tissue at risk for infarction and may aid in the appropriate selection of patients for thrombolytic therapy. The purpose of this article is to review the basic principles of DWI and PWI and to discuss the application of these techniques for the diagnosis and management of patients with acute ischemic stroke.
Dr. Bochar received her MD from the University of
Illinois College of Medicine in 1997. She is currently a
third-year radiology resident at Emory University School of
Medicine, Atlanta, GA.
Ischemic stroke is an important cause of morbidity and mortality
worldwide; it is the third leading cause of death
and affects more than 400,000 individuals per year in the United
Until recently, there has been no approved treatment for acute
ischemic cerebral infarction. Although neuroprotective agents have
been shown to be beneficial in animal models with middle cerebral
no significant benefits have been proven in humans. In 1996,
intravenous (IV) tissue plasminogen activator (t-PA) became the
first United States Food and Drug Administration (FDA)-approved
medication for the treatment of acute stroke. This approval was
prompted by the results of the National Institute of Neurologic
Disorders and Stroke (NINDS) t-PA Study Group, which demonstrated
improvement in clinical outcome when IV t-PA was administered
within 3 hours of the onset of acute stroke.
The first European Cooperative Acute Stroke Study (ECASS I)
demonstrated improvement in clinical outcome and some functional
measures in a subgroup of stroke patients without computed
tomography (CT) signs of extensive infarction when IV recombinant
t-PA (rt-PA) was administered within 6 hours from the onset of
Given the narrow window of 3, or potentially 6, hours from the
onset of stroke symptoms to the administration of IV t-PA, the
timely diagnosis of acute ischemic stroke is essential. With the
recent development of improved magnetic resonance (MR) techniques,
including new diffusion-weighted imaging (DWI) and
perfusion-weighted imaging (PWI) techniques, the rapid and accurate
detection of acute infarction is now possible.
--Diffusion-weighted MR technology has been available for several
decades. Stejskal and Tanner
described diffusion-weighted sequences as early as 1965. Although
the clinical application of DWI in evaluating neurologic disorders
was first reported in the mid 1980s,
DWI has only recently become practical in the clinical setting.
This is largely a result of the development of echo-planar
diffusion sequences, which have dramatically decreased imaging
times and motion artifacts. DWI is now a rapid imaging modality,
with the ability to obtain images of the whole brain in less than a
minute. DWI can also detect hyper-acute ischemic changes promptly
after the onset of ischemic stroke. In animal models, ischemic
changes have been observed on DWI within minutes,
and in humans, changes have been reported as early as 30 minutes
after the onset of stroke symptoms.
The lesions on diffusion-weighted images have been shown to
correlate with regions of ischemia and infarction on histologic and
DWI is based on the microscopic movement (Brownian motion) of
water molecules. By applying strong MR diffusion-sensitizing
gradient pulses, water protons can be labeled during dephasing and
In normal diffusion, water protons move between the dephasing and
the rephasing gradients, resulting in signal attenuation. If there
is restricted diffusion, such as in acute infarction, the signal
attenuation is decreased, resulting in hyperintensity on
The degree of diffusion-weighting is called the b value. If the
b value is zero, there is no diffusion weighting. The images are
similar to T2-weighted images, and the cerebral spinal fluid (CSF)
appears bright. Conversely, if the b value is high, there is heavy
diffusion-weighting. The CSF appears dark since the water molecules
in the CSF have unrestricted diffusion.
It is important to note that diffusion-weighted images are a
combination of diffusion information and T2 signal intensity. If a
lesion, such as a chronic infarct, is hyperintense on T2-weighted
images, the lesion may also appear hyperintense on DWI despite a
lack of restricted diffusion. This phenomenon has been referred to
as "T2 shine-through"
and can be eliminated by the generation of an apparent diffusion
coeffient (ADC) map.
The apparent diffusion coeffient is a measure of water diffusion
and is calculated by differences in the rate of change of signal
intensity at various b values.
It is termed "apparent" because the measured value does not
indicate pure diffusion, but reflects capillary perfusion and other
An ADC map is created by having signal intensities that are equal
to the magnitude of the ADC values.
Low ADC values reflect restricted diffusion and appear hypointense
on an ADC map (but hyperintense on DWI). Likewise, nonrestricted
diffusion appears hyperintense on an ADC map (but hypointense on
DWI). ADC values in acute ischemic lesions are often compared with
the ADC values of the nonaffected contralateral brain parenchyma,
and the ratio of the two is referred to as the ADC ratio (ADCr).
ADC values in acute ischemic lesions are typically 30% to 40% lower
than in the contralateral brain parenchyma.
Serial evaluations of cerebral infarcts in humans have demonstrated
a decrease in the ADC values in the first week, followed by a
period of "pseudonormalization" of the ADC values around 5 to 10
days. After the first week, the ADC values may become elevated,
corresponding to increased diffusion secondary to encephalomalacia
as seen histologically.
By utilizing the ADC values and maps in combination with the
diffusion-weighted images, acute ischemic lesions can readily be
distinguished from chronic ones (figure 1).
Anisotropy is the phenomenon in which diffusion is not equal in
all directions. In the brain, increased diffusion is seen along
white matter tracts. Greater diffusion is observed when the
diffusion-sensitizing gradients are applied parallel to the white
matter tracts, and slower diffusion is observed when the gradients
are applied perpendicular to the white matter tracts.
Diffusion anisotropy may result in hyperintensity when the
gradients are applied perpendicular to the white matter tracts, and
this may lead to the misinterpretation of the diffusion-weighted
images. A simple way to counteract the effects of anisotropy is to
combine the individual, mutually orthogonal diffusion-weighted
images, resulting in a "trace" isotropic image.
The exact mechanism of restricted diffusion during acute
ischemia is unclear, but appears to be related to the failure of
cellular energy metabolism.
Sodium-potassium adenosine-triphosphatase (Na-K ATPase) within the
cell membrane is disrupted by ischemia. Absolute tissue perfusion
under 15 to 20 cc/100g/min is the threshold for Na-K ATPase
dysfunction and has also been shown to be the threshold for
restricted diffusion. In the presence of Na-K ATPase dysfunction,
there is influx of sodium and water into the cells, with resulting
cytotoxic edema, "compartmentalization" of the water molecules, and
--PWI is also referred to as perfusion imaging (PI) or
hemodynamically weighted imaging (HWI). It evaluates the blood flow
in the cerebral microvasculature and detects areas of perfusion
abnormalities. There are two major methods currently used to
determine PWI: bolus-tracking PWI (BT-PWI) and arterial
spin-labeling techniques (ASL-PWI).
Bolus-tracking, which is also called first-pass bolus method or
susceptibility-based perfusion imaging, requires the IV injection
of a paramagnetic contrast agent, namely a gadolinium chelate. The
difference between the increased magnetic susceptibility of the
gadolinium and the lower susceptibility of the surrounding tissues
creates local field inhomogeneities, which result in signal loss on
T2*-weighted images (figure 2). From the data obtained, time-signal
intensity curves are created, and various hemodynamic parameters
can be calculated. The most common parameters to evaluate perfusion
on PWI are time-to-peak (TTP), mean transit time (MMT), and
regional or relative cerebral blood flow (rCBF) and volume (rCBV).
Although absolute CBF and CBV can be calculated experimentally,
these parameters require the calculation of arterial input as well
as lengthy processing.
Arterial spin-labeling PWI is also known as spin-tagging or
time-of-flight PWI. It is termed noninvasive, as exogenous contrast
is not required. Water protons in arterial blood are labeled with
an inversion pulse and compared to a control image without an
ASL-PWI results in a more direct measurement of CBF than
bolus-tracking and provides quantification of CBF and qualitative
DWI has been shown to be very accurate in the diagnosis of acute
with reported sensitivities of 88% to 100% and specificities of 86%
to 100% in clinical studies. Lansberg et al
compared DWI with CT within 7 hours of stroke onset. CT was
observed to have only a 42% to 63% sensitivity, compared with 100%
sensitivity for DWI in this series. A strong correlation was also
noted between the final infarct volumes and the acute DWI lesion
volumes, but no correlation was noted between the CT volumes and
the final infarct volumes. Lansberg et al
also compared DWI to conventional MRI within 48 hours of acute
stroke. DWI correctly identified at least one lesion in 94% of the
patients, compared to 71% to 80% for conventional MRI. Inter-rater
reliability was also observed, and was reported to be good
(kappa=0.8) for DWI, but only moderate (kappa=0.5-0.6) for
conventional MRI. Lesion conspicuity and observer confidence were
also significantly improved by the addition of DWI.
Although the sensitivity and specificity of DWI in the detection
of acute ischemic stroke are very high, false-negatives and
false-positives have been observed. Ay et al
reported 27 of 782 consecutive patients with stroke-like deficits
who had normal DWI images. When compared with the final diagnosis,
37% of these 27 patients were found to have a "stroke-mimic,"
including migraine, seizure, functional disorder, transient global
amnesia, and brain tumors. The remaining 63% were shown to have
ischemic events including transient ischemic attacks, prolonged
reversible deficits, and brainstem lacunar (3/27) and hemispheric
(3/27) infarctions. All three of the patients with hemispheric
infarctions but normal initial diffusion-weighted images had
abnormalities on PWI. Presumably, the tissues in the region of the
PWI abnormalities were ischemic but viable, and in the presence of
prolonged ischemia, the lesions progressed to infarction. This
emphasizes the importance of the addition of PWI to DWI. Other
investigators have also described false-negative DWI in patients
with presumed very small brainstem and deep gray nuclei
potentially because these tiny lacunar infarctions may be below the
resolution of DWI.
False-positive diffusion-weighted images have been seen in patients
with tumor or cerebral abscess.
These two disease entities can usually be easily distinguished from
acute ischemic stroke when conventional MR images, including T1-
and T2-weighted images, are obtained in combination with the
diffusion-and perfusion-weighted images.
Lesion volumes and clinical outcomes
The evaluation of ischemic lesion volumes with DWI and PWI has
proven to be significant in several regards. Both techniques have
provided an insight into the dynamic nature of infarct evolution.
Schwamm et al
reported that infarctions tend to increase in size during the first
several days after stroke onset and reach a maximum volume at an
average of 70 hours, and that final infarct volumes tended to be
smaller than the maximum infarct volumes. Numerous investigations
have shown that both DWI and PWI are good predictors of final
infarct volumes, with a strong correlation existing between lesion
volumes on initial diffusion- and perfusion-weighted images and
final infarct volumes (r values ranging from 0.479 to 0.99).
Karonen et al
specifically observed that of the various perfusion parameters, the
final infarct volumes correlated best with rCBV. The degree of
DWI-PWI mismatch also correlates well with infarct growth. Finally,
lesion volumes on DWI and PWI have been shown to have an excellent
correlation with clinical outcomes.
Several stroke scales have been devised to provide an index of
the neurologic status of patients after ischemic stroke. Such
scales can be used to assess the severity of the stroke, that is,
the neurological impairment, within the first few days. When
applied after the first few months of a stroke, the stroke scales
can quantify both functional outcomes and neurologic recovery.
Some of the more common stroke scales include the National
Institutes of Health Stroke Scale (NIHSS), the Canadian Neurologic
Scale, The Middle Cerebral Artery Neurologic Score, the Barthel
Index, the Rankin Scale, and the European Stroke Scale. There is a
good correlation (with r values of 0.44 to 0.88) between the DWI
and PWI lesion volumes and the stroke scale severity and outcome
The correlation is weaker in penetrator artery disease, but
stronger in cortical infarction.
When PWI is performed in combination with DWI, different
patterns of abnormalities are seen.
The most common pattern is one in which the perfusion abnormality
is larger than, and surrounds, the DWI abnormality. This is
referred to as a DWI-PWI mismatch, and is often associated with
large vessel (such as proximal middle cerebral artery [MCA])
occlusion. The area of abnormal perfusion surrounding the infarcted
DWI core is thought to represent ischemic tissue at risk for
infarction, or the "ischemic penumbra."
This tissue is potentially viable if perfusion is restored.
The degree of initial DWI-PWI mismatch correlates well with the
degree of infarct growth. Barber et al
reported that the presence of a DWI-PWI is a significant predictor
of infarct growth, and in the absence of a DWI-PWI mismatch, no
significant lesion enlargement occurred.
Other common patterns of DWI-PWI abnormalities include a DWI
lesion that is equal to or larger than the perfusion abnormality
and a DWI lesion in the absence of a perfusion abnormality. In the
presence of small vessel disease (such as distal MCA or perforator
infarctions), the initial PWI lesion volumes approximate the
initial DWI lesion volumes, and there is only slight infarct growth
with time. The finding of a DWI abnormality without a perfusion
abnormality suggests presence of infarction with reperfusion. No
significant infarct growth occurs in this situation. These findings
again emphasize the importance of combining perfusion imaging along
with diffusion imaging to potentially viable tissues that are at
risk for infarction.
DWI and PWI play a major role in the evaluation and follow-up of
patients for thrombolytic therapy. Since t-PA is associated with an
increased risk of intracranial hemorrhage,
it is important to select patients who will most likely benefit
from treatment and who are the least likely to experience
complications. Patients who are the most likely to benefit from IV
t-PA include those with an acutely occluded major intracranial
vessel, especially an occluded proximal MCA; and patients with
"tissue at risk," as demonstrated by DWI and PWI.
Preliminary studies have shown that early recanalization after
thrombolytic therapy can save tissues at risk for further
infarction. Patients with early recanalization demonstrated
significantly smaller infarct volumes and significantly better
outcomes when compared with patients with vessel occlusion without
In addition to aiding in the selection of patients who may
benefit from thrombolytic therapy, DWI and PWI may help exclude
patients who are at an increased risk for complications. The
presence of intracranial hemorrhage, brain edema or mass effect,
and involvement of >33% of an MCA territory are all
contraindications to IV t-PA.
Preliminary investigations of standardized multimodality stroke
protocols including DWI, PWI, conventional MRI, and magnetic
resonance angiography (MRA) found MRI to be as reliable as CT in
the detection of acute intracranial hemorrhage (ICH),
thus potentially eliminating the need for CT in the evaluation of
acute stoke. Additionally, there is an increased sensitivity in DWI
when compared with CT for the detection of >33% MCA territory
DWI can also differentiate acute from chronic lesions,
which is essential since only acute lesions can benefit from
Besides facilitating patient selection for thrombolytic
administration, DWI and PWI are also helpful in patient follow-up
after such therapy. When combined with conventional MRI, DWI and
PWI can evaluate for potential complications including intracranial
hemorrhage. DWI and PWI can also determine the efficacy of
thrombolytic therapy by evaluating lesion volumes following
Since the therapeutic window for IV t-PA is very narrow, at 3,
or potentially 6, hours from the onset of symptoms, diagnostic
images must be obtained quickly. Sunshine et al
reported "ultrafast MR imaging" of stroke patients. In this series,
the mean time from entering the emergency department to the
beginning of imaging was < 45 minutes, and the mean imaging time
was < 15 minutes, including DWI, PWI, conventional MRI, and
post-processing. Although these time frames seem idealistic and not
practically obtainable, it is hoped that in the future there will
be widespread availability of emergency MR imaging, including DWI
and PWI, to rapidly and accurately diagnose acute ischemic
While the cost effectiveness of IV t-PA has been established,
as of yet, no published investigations have determined the cost
effectiveness of DWI and PWI in the evaluation of acute ischemic
stroke. There are, however, several theoretical cost benefits
associated with the use of DWI and PWI. First, expenses may be
reduced by potentially eliminating the need for CT to exclude
intracranial hemorrhage. Second, DWI and PWI may aid in the
appropriate selection of patients for thrombolytic therapy.
Patients without a DWI-PWI mismatch are much less likely to have
"tissue at risk" for infarction, and the expense and unnecessary
risk of IV t-PA may be spared in these patients. Patients with an
unacceptably increased risk for complications from thrombolytic
therapy can be identified, and the potential for intracranial
hemorrhage can be prevented. Finally, since DWI combined with PWI
is very sensitive at detecting acute ischemic lesions and since
there is a strong correlation between the perfusion and diffusion
lesion volumes and the final infarct volumes, the number of
follow-up studies with CT or conventional MRI may be reduced.
Large-scale investigations are required to validate these
In summary, DWI and PWI are relatively new and accurate imaging
modalities for the timely diagnosis of acute ischemic infarction
and hypoperfusion. Lesions on diffusion- and perfusion-weighted
images have been found to correlate well with infarct growth, final
infarct volume, stroke severity, and clinical outcomes. DWI and PWI
may eventually become incorporated into acute stroke management
algorithms for evaluating potential candidates for thrombolytic
therapy by detecting "tissue at risk" for infarction and by
identifying patients at an increased risk for complications. DWI
and PWI can also be utilized for patient follow-up after
thrombolytic therapy to exclude potential complications and to
evaluate the response to treatment. Further studies are needed to
confirm the sensitivity and specificity of DWI and PWI in
combination with multimodality MRI to exclude intracranial
hemorrhage; to evaluate the change in perfusion and diffusion
abnormalities following thrombolytic therapy and recanalization;
and to determine the cost effectiveness of DWI and PWI in the
evaluation of acute ischemic stroke.
I am very grateful to Chad Holder, MD, and Richard Woodcock, MD,
(both from Emory University School of Medicine, Atlanta, GA) for
their thoughtful input and for providing images for this