Stroke is a leading cause of permanent disability and death in the United States. Noncontrast computed tomography (CT) is the mainstay for imaging of acute stroke in the emergency setting, while more definitive characterization of stroke, often in the subacute stage, is achieved in a more limited fashion with magnetic resonance imaging, positron emission tomography, and single-photon emission computed tomography. However, the combination of noncontrast head CT, CT angiography, and the developing technique of first-pass CT perfusion imaging provides a rapid, well-tolerated and widely available means of characterizing acute stroke and selecting appropriate candidates for thrombolytic therapy.
received her MD-PhD degree from the University of Texas Medical
Branch, Galveston, in 1999. She is currently a Co-Chief Resident
at Wake Forest University, Winston-Salem, NC, and plans to pursue
a Fellowship in Magnetic Resonance or Body Imaging.
Stroke is a leading cause of permanent disability and
death in the United States. Noncontrast computed tomography (CT)
is the mainstay for imaging of acute stroke in the emergency
setting, while more definitive characterization of stroke, often
in the subacute stage, is achieved in a more limited fashion with
magnetic resonance imaging, positron emission tomography, and
single-photon emission computed tomography. However, the
combination of noncontrast head CT, CT angiography, and the
developing technique of first-pass CT perfusion imaging provides
a rapid, well-tolerated and widely available means of
characterizing acute stroke and selecting appropriate candidates
for thrombolytic therapy.
Stroke is the leading cause of permanent disability and third
leading cause of death in the United States.
In the emergency setting, noncontrast computed tomography (CT) is
the mainstay for imaging of acute stroke. The primary role of CT is
to exclude both hemorrhagic infarct (an absolute contraindication
to thrombolytic therapy) and tumor mimicking the symptoms of
stroke. Noncontrast CT can often demonstrate areas of acute
cerebral ischemia or infarction, manifested by regions of
hypoattenuation, but these can be difficult to detect within the
first few hours,
making more precise localization and characterization of stroke
difficult (Figure 1A).
Yet, the first few hours are precisely when such information is
most critical, as current guidelines suggest initiating intravenous
thrombolytic therapy within 3 hours of stroke onset, and
intra-arterial thrombolytic agents within 6 hours; intra-arterial
thrombolysis is most effective in patients with proximal
large-vessel occlusions, making timely localization of the site of
vessel occlusion a priority (Figure 1B).
Diffusion-weighted magnetic resonance imaging (MRI) is a sensitive
marker for acute cerebral ischemia and is used in acute and
subacute stroke imaging, but this use is primarily restricted to
The 24-hour-a-day availability of MRI remains very limited,
particularly in community and general hospitals, where the majority
of stroke patients initially seek medical care.
Furthermore, MRI is less sensitive than CT for acute hemorrhage, an
absolute contraindication for the administration of thrombolytics.
The ideal technology for imaging acute stroke would combine ready
availability with the ability to provide timely anatomic and
functional characterization of cerebral ischemia, allowing more
accurate selection of candidates for intervention.
A three-tiered approach: CT imaging of acute
A three-tiered approach to CT imaging of acute stroke
incorporates noncontrast CT (NCCT) of the head, CT angiography
(CTA) and CT perfusion imaging (CTP). CT angiography and CT
perfusion imaging can be added to the standard noncontrast head CT
protocol with little increase in scanning time (<10 to 15
minutes total scanner time), particularly with faster
multi-detector CT scanners.
The majority of stroke patients tolerate iodinated contrast
and experimental animal models have demonstrated the safety of
iodinated contrast material in the setting of acute stroke.
Furthermore, both CTA and CTP can be performed using single- and
multichannel CT scanners widely available in the emergency setting,
and commercially available processing software.
Table 1 summarizes techniques used in noncontrast CT, CTA, and
CTP protocols. A more detailed description is included below.
Noncontrast head CT
Standard NCCT of the head can be performed using existing
protocols. Briefly, with the patient in a headholder, 2.5- to 5-mm
contiguous axial slices from the skull base to the vertex are
obtained at approximately 120 kVp and 170 mA. After evaluating the
scan at standard window width and level settings (eg, 90 HU, 40
HU), window width and level settings can also be narrowed (8 HU, 32
HU) to improve detection of regions of relative hypo-attenuation,
which may correspond to acute infarct (Figure 2A).
After noncontrast CT, CTA of the vertebrobasilar system and
circle of Willis is performed. The lateral scout image can be used
to select a coverage area extending from the skull base to the
50 mL (370 to 400 mg I/mL) of nonionic iodinated contrast material
is injected via power injector at 2.5 to 4 mL/sec via an 18- or
20-gauge antecubital intravenous line.
After a 20- to 25-second delay, the scan is initiated in the
caudal-cranial direction. The remainder of the scanning protocol
will be dictated by the limitations of the scanner--ie, single
channel versus multidetector. Please refer to Table 1 for more
Postprocessing typically involves the creation of overlapping
limited-volume maximum intensity projection (OLIVE MIP) reformats
of the source data, which can often be performed by the
technologist in a semiautomated fashion.
Such reformatted images significantly improve the sensitivity for
arterial occlusion, particularly in the case of distal-branch
They can be reviewed at an offline workstation with
three-dimensional (3D) display capability, or can be sent to an
integrated picture archiving and communications system (PACS)
station for review alongside the NCCT. The source images can be
reconstructed in 3-mm slices for qualitative estimates of whole
brain perfusion (Figure 2B).
Dynamic CT perfusion
Dynamic CT perfusion images are generated by tracking over time
the passage of a bolus of contrast material ("first pass") through
the cerebral vasculature. Typically, the selected location contains
all three vascular territories (anterior cerebral, middle cerebral,
and posterior cerebral arteries), usually near the level of the
basal ganglia. Alternately, an abnormality detected on NCCT can be
used for localization.
40 mL (370 mg I/mL) is administered at 4 to 5 mL/sec via an 18- or
After a 5-second scan delay, the scan is initiated in cine mode
(typically 1 image/sec) at 80 kVp, 200 mA. A low-voltage protocol
has the effect of increasing relative enhancement (due to
dependence on iodine) and decreasing patient radiation exposure. On
a standard helical CT scanner, one 10-mm slice can be evaluated;
multidetector scanners allow greater coverage and the ability to
obtain multiple slices.
To reduce lens irradiation, most institutions select a level or
scan angle that avoids direct irradiation of the eye.
The collected data are either processed directly on the scanner or
transferred to an offline workstation. The majority of commercially
available software packages use a complicated mathematical model
known as deconvolution analysis to map multiple parameters,
including cerebral blood flow (CBF), cerebral blood volume (CBV),
and mean transit time (MTT). The deconvolution model has the
advantage of allowing for lower (3 to 6 mL/sec), and more
clinically acceptable, injection rates than does an alternative,
the maximum slope model (injection rates, 10 to 20 mL/sec).
Typically, the operator is required to manually select arterial and
venous regions of interest (ROIs) as input functions for the
analysis (Figure 3A). The unaffected anterior cerebral artery,
middle cerebral artery (MCA), and superior sagittal sinus are often
used as reference vessels for the ROIs.
The first pass of contrast material is tracked from the arterial to
venous side of the circulation (Figure 3B).
The analysis yields several measurements. Cerebral blood flow is
the volume flow rate of the blood through the cerebral vasculature
per unit time (normal: 50 to 60 mL/100 g/min). Cerebral blood
volume (CBV) is the amount of blood in a given amount of tissue at
any given time (normal: 4 mL/100 g of tissue). Finally, mean
transit time (MTT) is the average time it takes for blood to travel
from the arterial to the venous side of the brain circulation; it
is also a factor of CBV and CBF (normal: 5 seconds) [above
characterizations of CBF, CBV and MTT adapted from Ortiz and
These parameters are mapped on a color-coded composite image
(Figures 3C and D), enabling qualitative and quantitative
evaluation of decreased perfusion.
A reduction in CBF of greater than 34% relative to the
or an absolute CBF value of <13 mL/100g/min
have been shown to be markers for severe acute cerebral ischemia,
likely to progress to infarct in the absence of thrombolytic
intervention. Nonviable or infarcted tissue, which is unlikely to
respond to intervention and is, indeed, at high risk for
reperfusion injury, may be categorized by CBV; a CBV below 1.5 to
2.5 mL/100 g appears to be related to irreversible infarct.
The difference between the regions of severe ischemia and
irreversible infarct represents the ischemic penumbra, or the
tissue that may be salvageable with therapeutic intervention.
Prolongation of mean transit time (MTT) also occurs with cerebral
ischemia, with a threshold value of >6 seconds.
While MTT is one of the most qualitatively sensitive markers of
cerebral ischemia, its quantitative use is limited primarily to the
calculation of CBF (CBF = CBV/MTT).
The source images from the CTA can also be reconstructed in 3-mm
slices to yield perfused blood volume estimates.
These images highlight hypoperfused tissue as regions of decreased
enhancement; the presence of a hypodense lesion on the CTA source
images correlates strongly with irreversible infarct, even in
conjunction with a normal NCCT.
Unenhanced images can be subtracted from the CTA source images to
yield "perfused blood volume maps."
This protocol obviates a second contrast bolus and shortens the
total scan time, both of which may be important considerations in
this patient population. However, it cannot provide the qualitative
measurements of CBV and MTT that are provided by first-pass CT
Validation of CTA and CTP
Experimental models and clinical studies
Experimental animal models of CT perfusion have provided some
direct validation of the ability of CT perfusion to accurately
predict the final extent of cerebral infarct. Hamberg et al used a
primate stroke model to demonstrate that CBV or CBF functional maps
obtained as soon as 30 minutes after balloon occlusion of the MCA
correlated strongly with final infarct size.
The combination of CTA and CTP was shown in one study to improve
the accuracy of diagnosing cerebral ischemia when compared with a
combination of noncontrast CT and clinical examination,
particularly in localizing the stroke (40% improvement) and the
site of vascular occlusion (38% improvement).
CT perfusion accurately predicts the final infarct size,
has been used in the acute setting to estimate the extent of the
and correlates strongly with clinical outcome.
The accuracy of CT perfusion for detecting acute infarct has also
been validated against a number of other imaging modalities,
including diffusion MRI
and xenon CT.
Limitations of CTA and CTP
The most important shortcoming of CT perfusion is the limited
anatomic coverage necessitated by the demands of dynamic data
acquisition. A scan limited to 1 or 2 slices may miss an area of
perfusion abnormality, particularly when a normal noncontrast CT
precludes a more specific slice selection.
As multidetector CT technology continues to develop, this
limitation is likely to become a less important consideration, as
significantly increased scan speeds on 16- and 32-channel scanners
will allow the acquisition of data from a larger volume of brain
without a loss of temporal resolution. Roberts et al
have suggested a modification of the CTP technique (using a
4-channel multidetector CT), to incorporate a "toggling-table"
approach. Briefly, the use of alternating table positions enables
acquisition of 2 different sets of CT perfusion images, each
consisting of 2 1-cm slices. Total coverage is increased to 4 cm,
double that of a conventional technique. Another option is to
repeat the CT perfusion protocol at 2 anatomic locations, the
additional location typically being superior to the first.
However, this has the disadvantage of increasing contrast dose and
radiation exposure, as well as requiring substantially more
postprocessing. Finally, the CTA source images can be used to
generate a whole-brain perfusion map, as previously described.
However, these generate only CBV data by extrapolation of density
change in a reference vessel, and are not as sensitive for
quantitative measurement of perfusion.
CTA, CTP, and thrombolytic therapy
One of the areas of greatest promise for CTA and CTP is in the
selection of appropriate candidates for thrombolytic therapy. The
benefit is twofold: CT angiography enables more accurate
localization of arterial occlusion, while perfusion data enables
risk stratification and prediction of treatment outcome.
CT angiography demonstrates good agreement with catheter
angiography for large-vessel occlusion.
Accurate localization and confirmation of vessel occlusion has
important implications for therapy; in the Prourokinase in Acute
Cerebral Thromboembolism (PROACT) trial, 62% of patients triaged to
catheter angiography because of strong clinical suggestion of
proximal MCA occlusion did not demonstrate such an occlusion on
angiography, and were excluded from treatment.
Clearly, a minimally invasive, rapid screening technique such as
CTA has the potential to prevent patients from undergoing
unnecessary invasive procedures.
The potential benefits of thrombolytic therapy--chiefly, the
prevention of incapacitating disability--must always be weighed
against the risk of intracranial hemorrhage arising from the
reperfusion of irreversibly infarcted tissue.
The practice of initiating thrombolytic therapy within 6 hours of
symptom onset is intended to exclude patients in whom the ischemic
penumbra has, in the face of ongoing hypoperfusion, progressed to
infarction. With perfusion CT, however, CBF can be calculated
directly and used to select appropriate candidates for thrombolytic
therapy. It has been suggested that the ideal target for
intervention is tissue with a CBF of 8 to 25 mL/100g/min.
Direct measurement of CBF may exclude some patients who seek
medical attention within 6 hours, as well as include some whose
symptoms are of longer standing, providing a more rational basis
for risk stratification. Some initial reports suggest that CTA and
CTP may also be useful immediately after intervention, as a tool
for assessing recanalization of the involved vessel.
Imaging of acute stroke: CT versus MR
Definitive imaging of acute ischemic stroke has focused largely
on diffusion-weighted magnetic resonance imaging (DW-MRI), which
relies on the phenomenon of restricted diffusion of water in areas
of cerebral infarction, likely resulting from cytotoxic edema.
Diffusion-weighted MRI accurately reflects the extent of cerebral
infarction within the first few hours of clinical stroke (Figure
Furthermore, whole-brain MR perfusion imaging can be used to
generate cerebral blood flow (CBF) maps. The mismatch between
perfusion and diffusion defects has been used to characterize the
However, a number of recent studies have demonstrated that a
combination of noncontrast CT, CTA, and CTP enable comparable
characterization of acute stroke as that achieved by MRI. Accurate
identification of the ischemic penumbra
and assessment of lesion volume
did not differ between the two modalities. Furthermore, CT
perfusion has the potential to more accurately quantify cerebral
blood flow, because unlike MRI, it is characterized by a linear
relationship between attenuation change and contrast dye
concentration, and CT demonstrates better spatial resolution.
The main advantage of MR perfusion imaging is its ability to
evaluate the entire brain, instead of being constrained to a more
limited anatomic area.
However, there are additional considerations beyond these
technical factors. The majority of patients with acute stroke
receive treatment at a local or small regional hospital, not at a
large academic center. Handschu et al
reported that of 103 patients admitted to a general hospital for
acute stroke, only one-third underwent brain imaging and only one
underwent MRI. MRI continues to have limited availability,
particularly in the emergency setting, and the majority of stroke
patients are initially admitted to hospitals without MR scanners.
CT is much more widely accessible, and is already being used in the
majority of patients imaged for acute stroke to exclude
intracranial hemorrhage; thus, a framework is already in place for
the widespread utilization of CTA and CTP. Its potential to more
rationally select candidates for thrombolytic therapy suggests it
may also play a role in broadening the use of interventional
therapy for acute ischemia.
PET and SPECT
Positron emission tomography (PET) performed with
F-labeled ligands is considered the gold standard for absolute
brain perfusion, and was used to generate normal and ischemic
threshold values for cerebral blood flow.
However, the short half-life of these agents essentially
necessitates an on-site cyclotron and markedly limits the
availability of PET for perfusion imaging. Single-photon emission
computed tomography (SPECT) is much more widely available and can
be performed in a somewhat timely fashion in centers with adequate
nuclear medicine departments. Multiple studies have demonstrated
the ability of SPECT to measure relative cerebral blood flow and to
predict clinical outcome in patients with acute stroke.
However, SPECT is unable to identify intracranial hemorrhage or
definitively localize the site of a large-vessel occlusion, and its
spatial resolution is much poorer than that of CT and MR. Thus,
while SPECT might be able to provide perfusion data, noncontrast CT
and CTA would still be required for full characterization of acute
stroke. Given this, it is much more time- and cost-efficient to
perform an exam that combines NCCT, CTA, and CTP, rather than
separate nuclear medicine and CT studies.
There are only a few hours during which thrombolytic therapy may
be used in the treatment of acute cerebral ischemia. The
combination of noncontrast head CT, CT angiography, and dynamic CT
perfusion provides a rapid, well-tolerated, and widely available
means of characterizing acute stroke and selecting appropriate
candidates for intervention. No other single imaging modality,
whether MRI, PET, or SPECT, matches the widespread availability and
diagnostic capacity of CT. CT truly comes closest to providing
"one-stop shopping" in the imaging of acute stroke.
The author thanks Michael Lev, MD, for providing many of the
images used in this paper and Jonathan Burdette, MD, for assistance
with the manuscript.