With multidetector technology and optimal contrast enhancement, CT is stretching its reach in thoracic imaging.
This publication was supported by an educational grant from
Amersham Health, Princeton, NJ. The opinions expressed in this
publication are those of the authors and not necessarily those of
Dr. Bae reports relationships with Tyco Healthcare and
Mallinckrodt through patent agreements and as a consultant. Dr.
Fishman reports relationships with Siemens Medical Solutions and
Amersham Health as a consultant. Dr. Foley reports a relationship
with GE Medical Systems through an investigator agreement. Dr.
Naidich reports a relationship with Siemens Medical Solutions
through its Advisory Board and as a consultant. Dr. Saini reports a
relationship with GE Medical Systems through research support. Dr.
Becker, Dr. Sahani, Dr. Siegel, Dr. Tahktani, and Dr. Zinreich
report that no such relationships exist.
is a Professor of Radiology at the New York University School of
Medicine, New York, NY.
We are living in a time of extraordinary technological prog-ress
in computed tomography (CT). Some 11 years after its introduction,
multislice CT continues to improve, expanding both its range and
accuracy in thoracic imaging. Today, multislice CT can easily
visualize the aorta in acute aortic syndromes, during postoperative
evaluation, and in congenital disease; diagnose pulmonary embolism;
assess regional lung perfusion; and characterize lung nodules,
among other applications. In the future, flat-panel detectors will
undoubtedly revolutionize our approach to CT once again, as will
wide-spread use of computer-assisted diagnosis.
Contrast enhancement plays a central role in thoracic CT and
there are many theoretical considerations governing its use. When
discussing contrast delivery and image processing, it is important
to keep the day-to-day realities of a busy imaging department in
mind, however. In addition to the need to employ straightforward
contrast-enhancement strategies, it is important to remember that
many community radiologists do not have experience using advanced
workstations or have picture archiving and communications systems.
Whether in large medical centers, such as NYU Medical Center, where
we perform >100 CT scans a day, or in smaller institutions,
protocols for contrast delivery must be developed with an eye to
Advances in CT technology have enabled us to perform faster
scans while using thinner collimation to acquire data across larger
and larger volumes of interest. We have progressed from using a
single-detector scanner with 5-mm collimation to examine the main
pulmonary arteries in 1991, to using a 4-slice scanner with 1-mm
collimation to examine many of its branches in 2000. With the
introduction of 8-, 10- and 16-slice CT scanners, it is now
possible to use submillimeter collimation, which enables true
isotropic imaging. One indication of this progress is a reduction
in the number of indeterminate CT pulmonary arteriograms, to 5% to
10%, with multidetector CT scanners.
We, and others, are interested in pushing CT to its limits, and
today that means using 16-detector technology. Table 1 provides
generic guidelines for performing CT angiography (CTA) using 4- or
16-slice scanners. Although these guidelines must be modified for
specific clinical situations, they cover nearly all potential
indications for CTA in the thorax.
We routinely use 1- to 1.25-mm collimators for data acquisition,
and 0.75-mm collimators only rarely. Images are then reconstructed
at 1.5- to 5-mm intervals, as clinically indicated. We typically
use a contrast volume of 100 to 150 mL, with con-centrations of 240
to 300 mgI/mL. We rarely use contrast material with a concentration
as high as 370 mgI/mL. In our practice, the use of monomeric
nonionic contrast material is the norm.
We inject contrast at a maximum rate of 3 to 5 mL/sec. Although
it is feasible to consider the use of faster injection rates, as
has been done in Europe, presently there are no routine indications
for this practice. In early studies of CT examination for pulmonary
embolism, investigators in France reported in-jection rates of 6 to
7 mL/sec using contrast material with a concentration of 120
While this approach proved adequate for identifying central clot,
it proved less efficacious for identifying peripheral segmental and
subsegmental pulmonary arteries.
To time image acquisition, we use either real-time bolus
tracking or a 20-mL test bolus of contrast material. In the latter
case, we acquire images at 5, 10, 15, and 20 seconds, as a guide to
optimizing contrast enhancement. In our experience, the main
purpose for a test bolus is to determine whether initial flow rates
provide optimal contrast enhancement of the central pulmonary
arteries. In cases of suboptimal contrast enhancement, we increase
the flow rate to a maximum of 5 mL/sec. Although a test bolus is
effective in ensuring a good-quality study, in our experience,
bolus tracking has proved more reliable for gauging when to acquire
Multidetector CT can be used to evaluate virtually any clinical
condition involving the aorta. While acute aortic syndromes, such
as aortic aneurysm and dissection, are among the most common, CT is
also useful in postoperative examinations and in the evaluation of
congenital disease. More recently, attention has focused on the use
of CTA for the evaluation of stent placement in patients with
Unlike imaging of the abdominal aorta, in which identification
of branch vessels is of critical importance, in the evaluation of
the thoracic aorta we routinely use thicker sections (5 mm). In
select cases, in which identification of bronchial, intercostal, or
anomalous vessels is needed, we may obtain 1- to 3-mm sections. In
all cases, additional routine or curved multiplanar sections are
reconstructed to simulate the appearance of the aorta as seen
angio-graphically. Depending on available technology, we also use
both maximum intensity projection images (MIPs) or volumetric
reconstructions (Figure 1).
Although volumetric renderings, in particular, elegantly
demonstrate aortic anatomy, they are clearly dependent on
optimization of contrast administration. Equally important is the
interface between the radiologist and advanced imaging technology.
While it is possible to optimize contrast enhancement throughout
the entire abdomen and chest, it is unlikely that CTA will gain
widespread acceptance by general radiologists until easier, more
automated methods of advanced imaging techniques are introduced.
Only then will multislice CT reach its full potential.
CT angioscopy is another promising application, not only for
evaluating the aorta but, potentially, for evaluating both the
pulmonary and coronary arteries as well (Figure 2).
Pulmonary arterial disease
The ability to perform increasingly sophisticated CTA has
revolutionized the diagnosis of diseases of the pulmonary arteries
and veins. Evaluation of acute pulmonary embolism, for example,
represents the most important opportunity today for clinical
expansion of CT in the chest. A decade ago, CT had virtually no
role in the diagnosis of pulmonary embolism; now it is one of the
most commonly requested scans in our institution.
With the aid of contrast enhancement, we are able to image
through the entire thorax, routinely visualizing small pulmonary
emboli even in fourth-, fifth-, and sixth-generation pulmonary
arteries. The volume-rendered images in Figure 3 depict a small
clot in the right pulmonary artery to the right upper lobe. Such
superb spatial resolution, as well as the ability to acquire data
through the entire thorax using a single contrast injection, has
now become common. Figure 2 demonstrates the potential for
pulmonary angioscopy to be used as an alternative method to
evaluate patients with suspected pulmonary embolism, specifically
those who might require embolectomy.
To image small pulmonary arteries, it is necessary to acquire
the thinnest possible sections. When data are acquired using a
single-detector CT and 3-mm-thick slices, subsegmental artery
visualization is possible in just 37% to 39% of patients.
By comparison, when multidetector CT is used to acquire data in
1.25-mm thick slices, visualization is possible in 71% to 76% of
patients. Now, with the capability of acquiring data in 0.75-mm
thick sections throughout the thorax, our ability to make more
precise diagnoses is even better.
A key issue in pulmonary arterial CT is the optimal delay
between contrast administration and the initiation of scanning. A
study by Hartmann et al
compared the use of a fixed 20-second scan delay with the use of
contrast transit time in a total of 107 patients. With a
single-detector CT scanner, investigators found no difference
between the two groups in image quality. They concluded that a
20-second fixed delay was acceptable for the evaluation of
In our experience, however, use of a fixed scan delay results in
a level of contrast enhancement insufficient to make subtle
diagnoses in segmental and subsegmental pulmonary arteries in
approximately 5% of cases, even with a 4-detector scanner.
Therefore, we now prefer to use real-time bolus tracking.
A study by Kirchner et al
involved 179 patients undergoing lung CT using a 4-slice scanner
and real-time bolus tracking. The study design had certain
limitations. For example, although investigators used contrast
material with a concentration of 300 mgI/mL, they injected at a
rate of only 2 mL/sec. In addition, most patients were undergoing
evaluation for thoracic neoplasia rather than pulmonary
Nonetheless, the study yielded several interesting observations.
Investigators triggered imaging when contrast enhancement reached
100 HU. Although, on average, scanning was initiated after a delay
of 21 seconds, the range was quite wide, 12 to 48 seconds,
reflecting substantial variation in patient physiology.
Investigators used an average of 10 monitoring images, a number
that would not be expected to substantially increase radiation
dose. They found no correlation between bolus geometry and patient
age, body surface area, or even weight in this study group. In
addition, contrast volume averaged only 48 mL (range 38 to 71 mL).
It is unlikely that such a low volume would be feasible for CT of
the aorta or pulmonary arteries, however.
It is worth noting that in CT of the chest, single-phase imaging
is the norm. The complicated multiphasic contrast injections that
characterize CT of the liver, for example, are generally
Perfusion imaging, although still in its infancy, shows
considerable promise as a correlative method for assessing
pulmonary embolism. Wildberger et al
reported preliminary data on the use of multidetector CT to produce
perfusion-weighted color maps of lung paren-chyma. Using a 4-slice
CT scanner, 120 mL of contrast material, and a 30-second scan
delay, investigators acquired data in 1.25-mm thick slices and
reconstructed it in increments of 0.8 mm. From source images, they
performed an automated 3D segmentation of the lungs, followed by
vessel cutting and adaptive filtering. The filtered volume data
were used to create a color-coded display of lung density, which
was fused with the original CT images.
In the 4 patients who did not have evidence of pulmonary
embolism on previous multislice CT, perfusion imaging showed a
homogeneous distribution of color-coded densities in the lung
(Figure 4A). In the 2 patients with established pulmonary embolism,
perfusion-weighted color maps showed low-density values distal to
the occluded pulmonary arteries (Figure 4B). Investigators
concluded that lung densitometry with color-coded display of
densities in the lung parenchyma may provide additional information
in the evaluation of pulmonary embolism.
Whether lung perfusion imaging will become clinically practical
is uncertain. Clearly, it is technologically feasible to create
such images, and to examine not just the embolus, but lung
physiology as well.
The potential to combine pulmonary angiography and pulmonary
venography represents another area of active research.
The published literature offers no consensus on the clinical
indications for such studies or the best way to perform them.
Nonetheless, they demonstrate that we are capable of finding clot
in the peripheral veins and suggest that CT venography may be an
important adjunct in the imaging of pulmonary embolism. Determining
the optimal method for contrast enhancement, including scan delay,
will require further investigation.
Computer-assisted diagnosis of pulmonary embolism also has great
potential. In Figure 5, segmented views of the pulmonary arteries
to the right lower lobe show the presence of multiple filling
defects that were identified using computer-assisted diagnosis.
This tool will likely become important, particularly for detecting
small subsegmental pulmonary emboli.
The best ways to administer contrast and project data are also
the subject of intense research. One approach would be to use
bicolor coding of multiplanar reconstructions of the pulmonary
arteries to display differences in contrast density suggestive of
embolism (Figure 5).
In addition to routine applications, CTA can be of value in
evaluating virtually any intrathoracic disease in which abnormal
vasculature is present. In addition to assessing thoracic aortic
aneurysms and dissections, CTA has also proved valuable for
assessing aortitis, identifying congenital anomalies, and
evaluating posttraumatic thoracic aortic injuries.
Similarly, although CTA has most often been used to assess acute
pulmonary embolism, less frequent indications include chronic
pulmonary embolism, pulmonary arteritis, pulmonary vascular
neoplasms, and traumatic pulmonary vascular injuries, among others.
Additional indications include vascular anomalies, such as those
that occur in patients with pulmonary sequestration or
arteriovenous malformations (Figure 6), and abnormalities of the
systemic and pulmonary veins, including the superior vena caval
Although nodule enhancement does not come under the heading of
CT angiography, it is an appropriate topic for any discussion of
the practical applications of contrast-enhanced thoracic CT. Nodule
enhancement has been well described in the literature.
The data suggest that the use of contrast enhancement in the
evaluation of pulmonary nodules is extremely helpful for
differentiating benign from malignant disease.
In a multicenter trial, Swensen et al
used contrast-enhanced CT to study 356 lung nodules 5 to 40 mm in
diameter. Nearly half of the nodules were malignant. Investigators
found that in malignant nodules, enhancement averaged 38.1 HU, as
compared with 10.0 HU in benign nodules (
<0.001). Using 15 HU as the cutoff, they found nodule
enhancement to have a 98% sensitivity and 58% specificity for
Evaluation of nodule enhancement is not widely done, however. In
part, this results from uncertainty about how the technique should
be applied. It is hoped that eventually it will be common to scan
each nodule as a volume and to do whole-volume determinations of
enhancement at various time phases. It may also be possible to
conduct permeability and perfusion studies of pulmonary
nodules--additional applications in which optimization of contrast
enhancement will play a crucial role.
With the introduction of multidetector CT scanners capable of
acquiring thin sections through the entire thorax in a single
breath-hold, CTA has now entered the era of routine clinical
practice. As outlined above, virtually any study requiring
evaluation of intrathoracic vessels may now be performed as a CT
angiogram. This has truly revolutionized our approach to diagnostic
imaging, as evidenced by the preferred use of CT in patients with a
clinical suspicion of an acute aortic syndrome or, especially, a
pulmonary embolism. The result has been near complete replacement
of the need for routine angiography in our department. It is
anticipated that with more sophisticated methods of data
acquisition, we will be able to routinely evaluate lung perfusion
as well as more accurately characterize focal lung pathology.
(A and B) Volumetric rendering of a Type B aortic dissection
showing both the true and false lumens to advantage.
(A) Volumetric and (B) angioscopic renderings of clot in the left
interlobar pulmonary artery (arrow). Use of virtual angioscopy
may prove of value in selected cases in which embolectomy is
Volumetric renderings of the central pulmonary arteries show
discrete filling defects in both (A) the truncus anterior (arrow)
and (B) the right interlobar pulmonary arteries (arrow).
(A) Normal lung perfusion is characterized by a normal
color-coded anteroposterior lung perfusion gradient in a patient
imaged in the supine position. (B) Perfusion image in a patient
with multiple pulmonary emboli. Perfusion imaging enables color
mapping of the extent of perfusion defects that would not be
evaluable on routine CT images. (Reprinted with permission from
Wildberger et al.
(A and B) Images through the basilar pulmonary arteries using
computer-assisted diagnosis (CAD) to identify multiple pulmonary
emboli (PE) (arrows). (Images provided courtesy of Carol Novak,
Siemens Corporate Research, Princeton, NJ).
Pulmonary arteriovenous malformation. (A) Axial and (B) sagittal
volumetrically rendered views show to advantage both the enlarged
feeding artery and draining vein.
ELLIOT K. FISHMAN, MD:
Thanks, David. In terms of pulmonary embolism, whether it's with
CTA in the future, or just now, one of the problems people
typically have is flow-related phenomenon, determining if it is
really a clot or is it just flow? What are your thoughts about
choosing contrast? Would an iso-osmolar contrast agent be better in
DAVID P. NAIDICH, MD:
Well, we don't really have any experience with iso-osmolar contrast
in that context: I can't really answer that. But I think the
question of why it is that some contrast studies of the pulmonary
arteries don't look as enhanced as they should relative to that 4%
or 5% of cases in which doing more detailed timing and volume
acquisition really would have played a role.
Although performing bolus tracking has become less of a problem,
we have discovered that using a time-delay test bolus may be as
We try to use as little contrast as possible to maximize
delivery. So we initially use 3 mL/sec, and then acquire data at 5,
10, 15, and 20 seconds. I have found that there is a certain
percentage of cases, not surprisingly, in which using that
acquisition protocol, you can see that none of those images are
really sufficient for good contrast enhancement.
The result of that has been when we use a timed bolus with
resulting poor vascular opacification, we can increase the rate of
injection from 3, to 4, or even 5 mL/sec, an option we lose with
bolus tracking. Ultimately, I don't think there is going to be a
solution, because I think part of the reason for suboptimal
opacification in select cases is both anatomic and, more
importantly, physiologic. Especially considering branch or crossing
points of smaller vessels and those with an oblique course, it is
just difficult to be certain that one is not missing a small
peripheral pulmonary embolus.
Has anyone had experience using an iso-osmolar contrast agent with
pulmonary emboli, compared with other kinds of contrast?
I can't comment, as we haven't been using it.
We use iso-osmolar contrast routinely in a lot of pulmonary emboli
studies. Dr. Pannu was doing some work that's just being analyzed
in some animal models, looking at the opacification of vessels and
homogeneity of vessels with different contrast agents. Our
experience in trying to document this in an animal model has been
that with an iso-osmolar agent, you get a more homogeneous
opacification of vessel. We've had fewer problems with determining
if a finding is an embolus or is flow-related.
I agree with you that in some patients, especially patients with
high blood volumes, such as postpartum patients being evaluated for
pulmonary emboli, it's impossible to get a good study. I don't care
how you inject the contrast and how you time it; in some cases, it
seems that nothing gets very bright, no matter what you do.
Even more than a postpartum patient, a woman in her third trimester
is also very hard to evaluate. In the third trimester, the heart
rate is 150% or 200% of what it normally would be, with increased
blood vol-ume. But, even in the most unexpected patients in whom
you don't know of any history of heart disease nor of any history
of even pulmonary physiologic abnormalities, sometimes you never
quite get the opacification necessary for good interpretation.
SANJAY SAINI, MD:
David, I don't do thoracic imaging, but I understand that the
density of contrast material and the superior vena cava can create
That's something that has been mentioned. I suppose, initially, if
people aren't used to evaluating PA studies they may be a little
confusing. It is an artifact that's easily recognizable as an
artifact, and it really has not interfered for some time now. It
used to be that people would use this to argue for scanning from
below up. But I don't think this is really necessary.
Does that mean that using higher density contrast material may be
less useful, because of that artifact?
Again, I'm not sure that streak artifacts specifically from the
superior vena cava are ever going to be that much of a problem.
Once you know what to look for, it's really not an issue.
W. DENNIS FOLEY, MD:
This could be an argument for a saline chase.
: I want to add one question. Do you have any comments on EKG
gating, particularly for the left lower lobe and pulmonary emboli?
Particularly with the low mAs that you are now using, could it be a
The problem with prospective gating is you need to know which cases
would likely benefit. Although, predictably, there are always going
to be problems looking at the lingular pulmonary arteries, because
cardiac motion with most state-of-the-art prospective gating
prolongs studies and makes them more difficult to perform. Most
important, prospective gating may limit volume acquisition in a
KYONGTAE T. BAE, MD, PhD:
What threshold do you use for bolus tracking?
We've arbitrarily been using 125 HU.
MARILYN SIEGEL, MD:
When do you use dual-phase or arterial-venous-phase imaging in the
chest? Do you ever use it for some of the cardiac studies
We don't really use it. Dual-phase imaging is rarely necessary in
the thorax: the sole exception is
nodule enhancement/perfusion studies. For these cases, a saline
chase would theoretically be most important to ensure uniform
contrast delivery. To my knowledge, this remains a problem, as I
believe that there is now only one manufacturer that markets a dual
injector. Is that correct?
There are now two or three of them.
It has been suggested that a saline chaser be used without a dual
injector; unfortunately, this has proved too complicated to gain