CNS perfusion refers to the delivery of blood to the brain,
along with its essential nutrients and oxygen. There are several MR
techniques for evaluating CNS perfusion.
This article will focus on dynamic susceptibility
gadolinium-enhanced contrast mechanisms.
The minimum normal cerebral blood flow is approximately 55
mL/100 g/min (Figure 1). Through a wide range of lower blood flow
rates, the brain is able to maintain normal function. Among the
brain's autoregulatory mechanisms for responding to low-flow
states, the first is usually an increase in collateral flow, which
on perfusion studies is reflected in a longer mean transit time
(MTT). Initially, there is also vasodilation of the capillary bed,
with a corresponding increase in cerebral blood volume (CBV).
Eventually, however, perfusion pressure drops low enough that
vascular collapse ensues and cerebral blood flow (CBF) is
At a threshold of approximately 15 to 20 mL/100 g/min, neuronal
dysfunction becomes evident. The neurons are stunned, but if
perfusion is reestablished quickly enough, they can recover fully.
Tissue in this reversible injury zone is known as the ischemic
penumbra, a word derived from the astrological term describing the
part of a planetary body lost in a partial eclipse.
The brain is particularly susceptible to a shortfall in its
blood supply, because it does not have long-term energy stores. If
the duration and degree of ischemia are severe enough, infarction
will be the ultimate result. A complete interruption of blood flow
is followed by irreversible injury within just a few minutes. In
many situations, however, the low-flow state or vascular occlusion
develops over time, presenting an opportunity to intervene and
salvage tissue. Even in acute stroke, MR perfusion techniques can
capture and characterize blood flow changes in the critical hours
after symptom onset, offering a chance to intervene.
Gadolinium-enhanced MR studies enable us to evaluate the
critical components of perfusion and define the hemodynamic
conditions in the brain. There are several goals for MR perfusion
imaging of stroke.
The first is the characterization of the dynamic pathophysiology of
the ischemic state. The second is the use of such data to select
patients for therapy, identifying those most likely to benefit from
thrombolysis, for example, and withholding therapy from those in
whom the risks outweigh the potential benefits. A third goal is to
follow-up the effects of an intervention, such as stenting or
At our institution, perfusion imaging is conducted in the
context of a more comprehensive stroke evaluation. We start with an
MR of the brain, including diffusion-weighted imaging (Table 1). We
use time-of-flight MRA in the head and the neck, then a
gadolinium-enhanced double-dose elliptic-centric MRA study to look
at the aortic arch, the origin of the great vessels, and the
Next we do perfusion imaging, using a single-dose gadolinium
bolus tracking study captured with a gradient-echo/echo-planar
sequence. We inject the gadolinium contrast media at 3 to 5 mL/sec
through a standard intravenous catheter, capturing the wash-in and
wash-out of the contrast over approximately 1 minute. We then
produce parameter maps and signal intensity-time curves to
establish region-by-region perfusion characteristics. Finally, we
obtain post-contrast images of the brain to look for intravascular
stasis and, perhaps, subacute parenchymal ischemia, which are both
better visualized after contrast than without it.
The use of gradient-echo/echo-planar perfusion sequences may
result in susceptibility artifact in certain regions of the brain,
such as the brain stem. We have found in clinical practice that
approximately two-thirds of strokes are in the distribution of the
middle cerebral artery, where susceptibility is not a major
problem. When we suspect that a patient has an infarct of the brain
stem or posterior fossa; however, we do perfusion imaging with a
spin-echo sequence to reduce artifact and couple it with a double
dose of gadolinium.
Perfusion Curves and Maps
The patient in Figure 2 underwent MR imaging approximately 6
hours after an acute infarction of the brain territory supplied by
the left middle cerebral artery. On CT and gadolinium-enhanced T1,
T2, and fluid-attenuated inversion recovery (FLAIR) MR images
(Figure 2A), evidence suggestive of stroke is very subtle.
Diffusion-weighted imaging (DWI), however, demonstrates the
"lightbulb" sign of restricted diffusion, a hallmark of acute and
usually irreversible ischemia. Indeed, today diffusion-weighted
imaging conducted with echo-planar sequences is the most sensitive
way to detect ischemia.
Selected slices from the patient's perfusion study demonstrate
wash-in of a bolus of gadolinium contrast media (Figure 2B). Areas
with normal perfusion darken in seconds, but only during the final
half-minute of the sequence is it possible to observe the slow
arrival of contrast in the infarcted left hemisphere.
Perfusion data can be used to create signal intensity-time
curves for selected regions of interest. When evaluating perfusion
dynamics, the important attributes to consider are those related to
the temporal course of blood delivery, as they reflect the time it
takes for gadolinium to arrive and then wash through the
microcirculation. These parameters include time to peak and various
estimates of transit time. In Figure 2C, the signal intensity-time
curve shows a precipitous drop in signal intensity upon arrival of
gadolinium in the normal right side of the brain, followed quickly
by wash-out. By comparison the arrival of contrast in the ischemic
region is delayed, and the washout takes longer.
By measuring the area under the signal intensity-time curve
during the first pass of the contrast bolus, it is possible to
calculate relative cerebral blood volume. The width of the curve
that depicts the wash-in and wash-out of gadolinium contrast
conceptually represents an estimate of the mean transit time (MTT).
The next step is to calculate the net cerebral blood flow in
milliliters per 100 grams of brain per minute, using the central
volume principle and an equation that, in its simplest form, can be
represented as: CBF = CBV/MTT. Cerebral blood volume is given in
milliliters per 100 grams of brain, and transit time, in seconds or
minutes. In practice, this is a complicated mathematical problem
that requires computation of additional derivatives of the raw
curves and can be approached in several ways.
The best way to calculate perfusion parameters, and which to use
on a routine basis, are the subject of intense research.
Fortunately, the rich data sets characteristic of MR perfusion
studies offer many hemodynamic parameters to study. Ultimately a
measure of the timing of blood delivery, an estimate of blood
volume, and a calculation of overall cerebral blood flow will be
optimal to fully characterize brain perfusion.
Color-coded parameter maps enable rapid identification of areas
of low or slow blood flow, which are depicted in red. In Figure 2D,
top row, the mean transit time is much longer in the left
hemisphere when compared with the right, and the cerebral blood
volume map shows complete vascular collapse, with low blood volume.
This defect is perfectly matched in anatomic extent to the
diffusion-weighted image, indicating an absence of potentially
reversible ischemia, or penumbra, in the surrounding tissue.
The color maps in Figure 2 are created with standard commercial
software. We have developed in-house software that rapidly creates
similar maps, while automatically deriving the arterial input
This additional piece of information can be used to mathematically
process the images more precisely, in a process called
deconvolution. The result is a family of black-and-white maps
(Figure 2D, bottom row) that depict not only mean transit time and
cerebral blood volume, but also cerebral blood flow.
In many cases, imaging can identify patients with ongoing
ischemia who may potentially benefit from therapy. A small or
modest diffusion abnormality, coupled with a very large perfusion
abnormality--a perfusion-diffusion mismatch--identifies an area
with very low blood flow but no infarction (Figure 3). Such an area
is a realistic target for rapid intervention.
The patient in Figure 4 was experiencing transient ischemic
attacks in the right hemisphere of the brain. Diffusion-weighted
imaging (Figure 4A) demonstrates a few abnormalities in the deep
white matter, while perfusion imaging reveals a very prolonged
transit time and very low blood flow in a larger surrounding area.
The color maps clearly show prolonged transit through the right
hemisphere (Figure 4B). Angiography revealed a tight carotid
stenosis, which was corrected with stenting. The same-day
post-stent perfusion study shows a complete reversal of the
perfusion abnormality (Figure 4C). On clinical follow-up, the
patient demonstrated no neurological deficits.
A comprehensive approach to CNS imaging can be used to follow
the effects of other interventions, such as thrombolysis, as well
as to evaluate vascular compromise in trauma. For example, Figure 5
depicts a patient with a skull fracture resulting from a rollover
motor vehicle accident. There is no apparent blood flow on the
T2-weighted images (Figure 5A), and a carotid occlusion can be seen
down to the level of the bifurcation on angiography (Figure 5B).
Perfusion parameter maps (Figure 5C), however, demonstrate that
there is no resting perfusion abnormality and no need for
The ability to perform perfusion studies in combination with MRA
enables global assessment of the patient with stroke or CNS
compromise. It is an approach that can be summarized by the "4
P's": We look at the
, which are the great vessels supplying blood to the brain; the
, and the
, the surrounding area of ischemia that requires intervention if
infarction is to be avoided. This global approach is possible only
through the use of gadolinium contrast media. Its use will
undoubtedly be expanded to include other applications of CNS
imaging in the future. *
THOMAS GRIST, MD:
Thank you very much, Dr. Rowley, for an excellent discussion of the
role of perfusion techniques in the CNS system. In the specific
technique for perfusion, you mentioned that you're using a gradient
echo-planar perfusion sequence. I noticed in some of the images
that it looked like there was a little susceptibility artifact at
the base of the brain.
Some people are using spin-echo echo-planar imaging (EPI), I
think with perhaps a higher dose of contrast agent, because the
spin-echo EPI may be less sensitive to the T2* effects. But on the
other hand, it reduces the susceptibility of artifact. Do you have
any comments on that?
HOWARD ROWLEY, MD:
Yes, I think that either approach can be successful, depending on
the susceptibility artifacts. Of course, susceptibility is a
dual-edged sword. You need some susceptibility to see the perfusion
changes. But the more susceptibility you have in your sequence, the
greater the extent of the skull-based and brain-stem region
artifacts. We've found that in clinical practice, two-thirds of the
strokes are in the middle cerebral distribution, so susceptibility
isn't a major problem. When we suspect a brain stem or posterior
fossa infarct, we have chosen to do a spin-echo planar technique.
In that situation, we would use a double-dose of gadolinium, and
process it in a similar way. There are several groups around the
country that prefer that method. Spin echo also gives you better
sensitivity to the capillary bed, as opposed to some large veins,
at least if you don't use deconvolution.
So then at the brain stem is where you're using it, brain-stem spin
echo. Then you give a double dose of contrast agent, just for the
MARTIN PRINCE, MD:
You mentioned that you use a first significant dose of contrast for
your 3D MRA of the arch carotid vessels. Then after that, the
perfusion study is done. So, does that then limit the amount of
gadolinium you have available to use on the perfusion study? Also,
what effect does the gadolinium that has already been introduced
into the circulation have for your MRA study?
Great question. Well, I think one reason that we have chosen the
gradient echo technique is that it only requires a single dose of
contrast. We performed the gadolinium-enhanced MRA with a double
dose, 0.2 per kilo. Then, we follow it with a single dose, which
keeps us within a triple-dose range. However, in situations in
which we might need a spin echo technique, for example, we'd be up
to four times the standard dose, which we still feel very
comfortable with. Also, one of the benefits of the perfusion
sequence is that you can perform that after other standard imaging,
or after gadolinium-enhanced MRA, because you are only looking at
the relative change compared with the baseline. So baseline signal
changes don't interfere with the measurement of perfusion.
With regard to the dose rate for these perfusion sequences, in
order to identify the arterial input function, it's nice to have as
compact a bolus as possible. So at what rate do you give the
gadolinium? Can you inject this by hand; or do you use a power
We strongly prefer a power injector, because of the reliability and
sequencing of the time that we acquire the data. So our standard
procedure is to use a single dose injected through a standard IV
catheter. We like at least a 22-gauge catheter. We'll go as low as
2 mL per second in an infant (we've done a number of infants
successfully), up to about 5 mL, if we have a large patient with a
large-bore IV. The arterial input function can be derived manually
using a region of interest drawn near a vessel or on a vessel; or,
as in our case, it can be acquired automatically. To some extent,
that removes the problem of needing to have a very tight bolus,
because the arterial input function is used to deconvolve that
curve, no matter what its shape.
RICHARD SEMELKA, MD:
Do you find that it's essential to have an imaging window in which
to perform the studyparticularly looking for the penumbra from
the start of patient symptoms to when you image them to have
maximum therapeutic benefit? What is the time that you look at?
That's a terrific question, Richard. The only FDA-approved therapy
right now for stroke is IV tPA within 3 hours. So technically, we
would be doing these studies within 3 hours, to try to help
determine what's going on with the patient and whether there is a
target for intervention. However, I think that there have been a
number of studies, including the intra-arterial PROACT study, which
have shown time windows out to 6 hours. There have also been many
case series and anecdotes showing success in individual cases, at
least carefully selected, out to 12 or even 20 hours.
The important point is that we're not looking as much at time
windows anymore. We're actually looking at physiology, and we are
looking at the individual patient, and it's a status of their
collateral blood flow that may allow one patient to go many hours
and still have a successful intervention. Another patient who might
be only at 1 hour could potentially be excluded. So it's the
physiologic window as opposed to time window that I think is
important, and that's what perfusion is giving us.
Well, even in this time window, there is both the time required to
acquire the data, and then there is the time required to do the
color mapping and various other post-processing. What is your sense
of how long it takes you from the time you've completed the data
acquisition, to the time when you can confidently say whether or
not there is a penumbra?
The sequence itself takes about 72 seconds. We run on the console
using our in-house software, which uses an algorithm that produces
these maps within 1 minute. Those maps are deconvolved. With our
vendor-supplied software, we are able to pull it across a network,
which is actually a rate-limiting step. It takes 4 or 5 minutes for
a large study like this to come across to my workstation. Then,
even my fellows and residents can perform this offline analysis in
2 or 3 minutes per case. It's actually quite easy, once you've done
a couple of them.
You also mentioned that the number of images is extraordinary. It
can't possibly be that you could film such a study and get it to
the interventionalist who might act on the results. How do you
provide this information to the referring physicians?
There are two components of that. One is the review of these
studies, which I think is best done on a workstation, as opposed to
on film. We don't film the source images, the 504 images that are
standard for our sequence. But instead, we have 14 slices for each
of the different parameter maps. Those are filmed as part of
clinical routine, put in the jacket, and conveyed to the referring