AJR 2005; 185:135-149
© American Roentgen Ray Society
Helical CT for the Evaluation of Acute Pulmonary Embolism
Smita Patel and
Ella A. Kazerooni
Department of Radiology, University of Michigan, 1500 E Medical Center
Dr., TC2910D, Ann Arbor, MI 48109-0326.
Received January 28, 2004;
accepted after revision October 20, 2004.
Address correspondence to S. Patel.
Abstract
OBJECTIVE. In this article, we review the current role of CT
pulmonary angiography and indirect CT venography for the evaluation of
pulmonary thromboembolic disease.
CONCLUSION. With advances in MDCT technology, evaluation of
pulmonary thromboembolic disease can now be performed with combined CT
pulmonary angiography and CT venography as a "one-stop-shopping"
test. CT pulmonary angiography is cost-effective, is accurate, has high
interobserver agreement, and has an added advantage of detecting other
life-threatening diseases in the chest that mimic pulmonary embolism.
Introduction
Pulmonary embolism is the third most common cause of cardiovascular
death, after myocardial ischemia and stroke. In the early 1970s, the incidence
of pulmonary embolism was reported as 630,000 per year with approximately
50,000-100,000 deaths annually in the United States and an untreated mortality
of 30% [1]. In the past few
decades, the incidence of pulmonary embolism has decreased by 45%, whereas
that of deep venous thrombosis is unchanged
[2]. Between 1979 and 1988,
deaths from pulmonary embolism decreased by 30%. This change is likely due to
a combination of factors including changes in diagnostic patterns, decreased
incidence of pulmonary embolism, and decreased case fatality rate
[3].
The diagnosis of pulmonary embolism continues to pose a challenge to both
clinicians and radiologists because the signs and symptoms of pulmonary
embolism are nonspecific. In the original Prospective Investigation of
Pulmonary Embolism Detection (PIOPED) study
[4], only one third of 755
patients who underwent pulmonary angiography for suspected pulmonary embolism
had the diagnosis of pulmonary embolism confirmed. Older imaging tests, such
as chest radiography, ventilation-perfusion (V/Q) scintigraphy, and pulmonary
angiography, suffer from a lack of specificity or are invasive
[5,
6]. CT pulmonary angiography is
a relatively safe, noninvasive test that can be performed quickly in an
emergency setting to directly identify the presence and extent of pulmonary
embolism. CT has also been found cost-effective in various diagnostic
algorithms
[7-9].
Deep venous thrombosis and pulmonary embolism are part of the spectrum of
venous thromboembolic disease. Although deep venous thrombosis was diagnosed
initially on conventional venography, it is now predominantly diagnosed
noninvasively on sonography
[10-14].
CT permits the diagnosis of both pulmonary embolism and deep venous thrombosis
with a single test.
Diagnostic Tests for Pulmonary Embolism
Chest radiographs are predominantly used for excluding other causes of a
patient's signs and symptoms and for the interpretation schema for V/Q scans
[4,
15].
Ventilation-Perfusion Scintigraphy
For the past three decades, combined ventilation and perfusion scans have
been the imaging technique of choice for the diagnosis of pulmonary embolism.
A V/Q scan with normal findings essentially excludes pulmonary embolism,
whereas a high-probability scan is highly specific for pulmonary embolism,
allowing definitive treatment. However, up to 70% of V/Q scans are
nondiagnostic, requiring additional tests to diagnose or exclude pulmonary
embolism. In the PIOPED study
[4], only 14% of patients
studied had a normal V/Q scan and 13% had a high-probability V/Q scan; the
majority, 73%, had an indeterminate or low-probability test result. Of the
patients with pulmonary embolism, only 41% had a high-probability V/Q scan;
the remaining 57% had either an intermediate or low-probability result
[16].
Statistically significant greater accuracy for pulmonary embolism detection
has been reported for CT pulmonary angiography (sensitivity, 94.1%;
specificity, 93.6%; positive predictive value [PPV], 95.5%; negative
predictive value [NPV], 96.2%) than for V/Q scans (sensitivity, 80.8%;
specificity, 73.8%; PPV, 95.5%; NPV, 75.9%) by Blachere et al.
[17]. Similar results were
reported by Grenier and Beigelman
[18]: sensitivities,
specificities, and kappa values with helical CT and scintigraphy were 87%,
95%, and 0.85 and 65%, 94%, and 0.61, respectively. Many believe these results
are sufficient justification for CT pulmonary angiography to replace V/Q
scintigraphy in the diagnostic algorithm for suspected acute pulmonary
embolism.
Catheter Pulmonary Angiography
Since the late 1960s, pulmonary angiography has been considered the most
accurate test for the evaluation of pulmonary embolism and the reference test
with which new diagnostic techniques are compared
[19,
20]. Pulmonary angiography is
invasive and underused and has a small—but definite—risk
[21-23].
Two studies 12 years apart in 1,240 patients showed that only 12-14% of
patients with an inconclusive diagnosis of pulmonary embolism on V/Q
scintigraphy subsequently underwent pulmonary angiography
[5,
6]. Many patients with
suspected pulmonary embolism are treated with anticoagulants on the basis of
clinical suspicion and nonspecific test results
[5]. Recent advances in CT
technology show that pulmonary angiography is an imperfect reference test: It
is less accurate than previously thought, particularly at the subsegmental
level [16].
Baile et al. [24] evaluated
the accuracy of pulmonary angiography and CT pulmonary angiography in a
porcine model using methacrylate beads injected into the pulmonary arteries
via the jugular vein to simulate emboli. Both catheter pulmonary angiography
and CT pulmonary angiography were performed; postmortem sections of the
pulmonary arteries served as the reference test. The sensitivity and PPV for
1-mm-collimation helical CT were 87% (95% confidence interval [CI], 79-93%)
and 81% (95% CI, 73-88%), respectively. For pulmonary angiography, the
sensitivity was 87% (95% CI, 79-93%) and PPV was 88% (95% CI, 80-93%), which
is not significantly different (p = 0.42)
[24]. In an earlier study of
angiography in a porcine model, researchers reported a 20% false-negative rate
especially when there was a partially occluding thrombus
[25]. Interobserver
variability for pulmonary angiography is considerable, especially at the
subsegmental level [4]. Overall
interobserver variability can approach 10-15% with pulmonary angiography and
is higher for smaller vessels
[4,
16].
CT Pulmonary Angiography
Incidentally detected pulmonary embolism was initially reported on
nonhelical CT studies performed for other clinical indications. In 1978,
Sinner [26] first reported the
diagnosis of pulmonary embolism on CT in a case report, and in 1982, Sinner
[27] reported a series of 21
consecutive patients with pulmonary embolism seen on CT. Subsequently,
visualization of pulmonary embolism on CT was reported in 1980 by Godwin et
al. [28] in the central
pulmonary arteries and in 1984 by Breatnach and Stanley
[29] in the segmental
pulmonary arteries. CT began to be used to evaluate the extent of pulmonary
embolism in patients with known diagnosis of pulmonary embolism, but was not
specifically used as a diagnostic test for pulmonary embolism until the advent
of helical CT.
In 1992, Remy-Jardin et al.
[30] reported the first
prospective study comparing single-detector helical CT at 5-mm collimation
with selective pulmonary angiography as the reference test in 42 patients with
central pulmonary embolism. Their results—100% sensitivity and 96%
specificity (one false-positive CT study due to asymmetry in pulmonary artery
perfusion from increased pulmonary arterial resistance confirmed at pulmonary
angiography)—showed promise for the use of CT. Teigen et al.
[31,
32] reported similar results
on electron beam CT for the detection of central pulmonary embolism.
Subsequent studies comparing single-detector CT with pulmonary angiography
showed a sensitivity of 53-100% and specificity of 78-100%
(Table 1).
Although CT pulmonary angiography specificity has been consistently high,
one of the major questions regarding CT pulmonary angiography is its
sensitivity for subsegmental emboli. For example, in a study by Goodman et al.
[33], a sensitivity of 86% was
reported when evaluating the central arteries in 20 patients, but it dropped
to 63% when subsegmental vessels were included. However, this was a small
series. Patients with an indeterminate probability V/Q scan result were
recruited and not consecutive patients with suspected pulmonary embolism.
Eleven (55%) of the 20 patients had pulmonary embolism on angiography, much
higher than the percentage of all patients evaluated for suspected pulmonary
embolism.
In the past several years, CT technology has evolved from single-detector
CT to MDCT and from 4-MDCT to 64-MDCT. This has significantly improved the
visualization of small pulmonary arteries on CT pulmonary angiography studies
and should translate into improved sensitivity for the detection of
subsegmental pulmonary embolism in clinical practice. Gantry rotation speed
has also decreased, from 1 to 0.4 sec, leading to a shorter breath-hold and
less respiratory motion artifact. The method of CT pulmonary angiography
interpretation has also evolved from hard-copy films to soft-copy computer
workstation review using active scrolling. Improved sensitivity for the
detection of small emboli over hard-copy review has been shown
[34]. At many institutions, CT
has replaced V/Q scanning for the evaluation of pulmonary embolism.
There is no accuracy data for MDCT available to date. The multicenter
PIOPED II study, which completed recruitment of 1,068 patients in 2003, is
designed to evaluate the accuracy of MDCT for pulmonary embolism detection,
and should yield important information on the test characteristics of
MDCT.
CT Technique
Patients should hyperventilate before the CT pulmonary angiography
acquisition by taking several large breaths in and out to maximize
breath-holding for the scan. CT pulmonary angiography is performed in a single
breath-hold. The acquisition includes the entire lungs on the fastest
scanners. The breath-hold ranges from 5 to 30 sec, depending on the scanner
type, and is considerably shorter with 16-MDCT than single-detector CT. Other
advantages of MDCT are increased z-axis coverage and decreased
partial volume averaging.
The timing of contrast bolus administration is critical to obtain optimal
opacification of the pulmonary arteries. A fixed scanning delay time of 20-25
sec can be used or a timing bolus can be used by injecting 15-20 mL of
contrast material and placing a region of interest in the pulmonary trunk to
obtain a time-density curve from which the scan delay can be calculated.
Alternatively, bolus tracking with a cursor in the main pulmonary artery that
triggers scanning at a preset threshold can be used. A timing method should be
used in patients with suspected or known cardiac dysfunction because the
optimum scan delay time can be 40 sec or more. Scanning is performed with a
100- to 125-mL bolus of IV contrast material injected at a rate of 4 mL/sec
using a power injector. Table 2 details the CT techniques used on different generations of helical CT
scanners.
Single-Detector CT Technique
Performing CT pulmonary angiography on a single-detector CT scanner
requires a trade-off between collimation and coverage. Even with this,
anatomic coverage is limited to approximately 10-12 cm, usually from aortic
arch to dome of the higher hemidiaphragm. A CT pulmonary angiography scan at
3-mm collimation and pitch of 1.3-1.6 covers 12 cm in the z-axis
using a gantry rotation speed of 1 sec
(Table 2). Using narrower
collimation or increasing z-axis coverage is not possible because of
limitations on CT tube cooling. Imaging the remainder of the pulmonary
arteries above and below the normal z-axis coverage requires either
an additional time delay or wider collimation, resulting in suboptimal
opacification of the arteries.
MDCT Technique
MDCT pulmonary angiograms are obtained at 1- to 1.5-mm collimation
throughout the entire thorax. MDCT protocols for 4-, 8-, and 16-MDCT scanners
used in the PIOPED II study are listed in
Table 2. The scanning time
ranges from 18 to 28 sec on 4-MDCT and from 8 to 13 sec on 16-row MDCT. This
allows high-resolution imaging of small pulmonary arteries throughout the
entire thorax in a shorter time and with less respiratory motion than
single-detector CT.
CT Findings of Pulmonary Embolism
CT findings of acute pulmonary embolism are related to the identification
of emboli that may or may not be surrounded by contrast material
(Table 3 and Figs.
1A, and
1B). Secondary findings of
pulmonary embolism may also be seen (Table
3 and Figs. 2 and
3).
View larger version (78K):
[in this window]
[in a new window]
[as a PowerPoint slide]
|
Fig. 1A —42-year-old man who was hypoxic on room air; patient was
paraplegic from spinal cord injury due to high-speed motorcycle crash. CT scan
shows bilateral central pulmonary embolism (long thick arrows),
subsegmental emboli (long thin arrow), and right lower lobe superior
segment pulmonary infarct (short arrows). Note
"tram-track" sign in inferior segmental artery of lingula
(arrowheads).
|
|
View larger version (74K):
[in this window]
[in a new window]
[as a PowerPoint slide]
|
Fig. 1B —42-year-old man who was hypoxic on room air; patient was
paraplegic from spinal cord injury due to high-speed motorcycle crash. CT scan
shows rim sign in left lower lobe pulmonary artery (long arrow),
small left effusion (arrowheads), and pulmonary infarct (short
arrows).
|
|
View larger version (124K):
[in this window]
[in a new window]
[as a PowerPoint slide]
|
Fig. 2 —67-year-old woman with glioblastoma multiforme and
right-sided chest pain. Sagittal CT reformation image shows subsegmental
pulmonary emboli (long arrows) and large wedge-shaped pulmonary
infarct posteriorly (short arrows).
|
|
View larger version (70K):
[in this window]
[in a new window]
[as a PowerPoint slide]
|
Fig. 3 —36-year-old woman with history of recurrent pulmonary
embolism. CT scan obtained using lung window settings shows mosaic
attenuation: areas of ground-glass attenuation with enlargement of pulmonary
arteries and areas of low attenuation due to diminished blood flow from
presence of emboli.
|
|
Advantages of Helical CT
A considerable advantage of CT over both V/Q scintigraphy and pulmonary
angiography is the ability to depict other conditions that clinically mimic
pulmonary embolism, such as acute pneumonia, lung abscess (Figs.
4A, and
4B), pneumothorax,
pneumo-mediastinum, pleural or pericardial effusion, aortic dissection,
cardiovascular disease, mediastinitis, mediastinal abscess, esophageal
rupture, and malignancy or interstitial pulmonary fibrosis; those other
conditions have been reported in 11-70% of CT examinations performed for
suspected acute pulmonary embolism
[35-41].
With 16-MDCT and ECG-gating, it may also be possible to identify occlusion of
a coronary artery or nonenhancement of the myocardium in patients with acute
myocardial infarction [42].
This information may be important because clinical signs and symptoms of
pulmonary embolism and myocardial infarction overlap.
View larger version (77K):
[in this window]
[in a new window]
[as a PowerPoint slide]
|
Fig. 4A —24-year-old man who presented with end-stage renal disease
secondary to diabetes, right atrial mass, and 3-week history of pneumonia. CT
scan shows large central pulmonary embolism (arrow) in right main
pulmonary artery. Note pericardial effusion (arrowheads).
|
|
View larger version (110K):
[in this window]
[in a new window]
[as a PowerPoint slide]
|
Fig. 4B —24-year-old man who presented with end-stage renal disease
secondary to diabetes, right atrial mass, and 3-week history of pneumonia. CT
scan shows cavitary right lower lobe mass (long white arrow)
representing lung abscess with adjacent empyema (short white arrows).
Note incidental right atrial myxoma (black arrow) and small
pericardial effusion (arrowheads).
|
|
Pitfalls and Limitations of CT
As with any diagnostic test, there are pitfalls and limitations
[43-45].
The pitfalls of CT pulmonary angiography can be divided into technical and
interpretive pitfalls, as listed in Table
4.
Technical Pitfalls
The most significant technical pitfall is poor contrast opacification of
the pulmonary arteries (Fig.
5). This may occur with poor cardiac function and can be overcome
using a timing bolus. Improper coordination of the total contrast injection
dose and injection flow rate may lead to a pseudo filling defect in the
pulmonary artery that mimics pulmonary embolism. Motion artifact can create a
pseudo filling defect caused by doubling of vessels
[46]. When a
high-spatial-frequency reconstruction algorithm is used, high attenuation is
seen around vessels, mimicking pulmonary embolism. A soft-tissue
reconstruction algorithm should be used.
View larger version (79K):
[in this window]
[in a new window]
[as a PowerPoint slide]
|
Fig. 5 —33-year-old man who had undergone pelvic surgery for trauma.
CT scan shows poor bolus of contrast material in pulmonary arteries and
beam-hardening artifact (arrow) from high amount of contrast material
in superior vena cava, which accounts for low attenuation in right main
pulmonary artery, making evaluation for subtle pulmonary embolism
difficult.
|
|
In large patients, images obtained at thin collimation are grainy because
of increased noise, which may obscure small arteries. Increasing the scanning
collimation from 1.25- to 2.5-mm or reconstructing 1.25-mm acquired data at a
2.5-mm thickness may improve visibility in such cases.
When contrast material is dense in the superior vena cava, streak artifacts
from beam hardening may obscure portions of the right main and upper lobe
pulmonary arteries and may even mimic pulmonary embolism. Using a saline push
of 20-25 mL immediately after the IV contrast injection may reduce the density
of the contrast material, as does scanning in the caudal-to-cranial
direction.
A pulmonary arterial flow artifact in which there is abrupt bilateral
short-segment loss of pulmonary arterial opacification is thought to be caused
by the inhomogeneous admixture of contrast material from the superior vena
cava and unopacified blood from the inferior vena cava within the right
atrium. This artifact is associated with inspiration immediately before
imaging and is caused by transient interruption of the contrast column in the
pulmonary arteries; this phenomenon has also been referred to as the
"stripe sign" [47,
48].
CT pulmonary angiography can be a challenging technique to perform in ICU
patients because of respiratory motion, suboptimal bolus with poor cardiac
reserve, and streak artifact from lines and tubes. However, in one series of
50 consecutive ICU patients with suspected pulmonary embolism, 76% of CT
pulmonary angiography examinations were of good to excellent quality
[49]. A Swan-Ganz balloon
catheter may cause streak artifact that creates the false appearance of emboli
or may totally obscure an embolism (Fig.
6). To avoid this artifact, pull the catheter out of the pulmonary
artery and place it in the heart or superior vena cava before CT pulmonary
angiography.
View larger version (118K):
[in this window]
[in a new window]
[as a PowerPoint slide]
|
Fig. 6 —76-year-old man with acute dyspnea and hypotension after
sigmoid colectomy. CT scan shows Swan-Ganz catheter (black arrow) in
right main pulmonary artery, with adjacent streak artifact. Note
low-attenuation abnormality posterior to right main pulmonary artery
(white arrow), which may represent pulmonary embolism or
beam-hardening artifact from Swan-Ganz catheter. Moderate-sized bilateral
pleural effusions with adjacent atelectasis are also seen.
|
|
Interpretative Pitfalls
Normal anatomic structures or abnormalities of other structures, such as
the bronchi, can be mistaken for pulmonary embolism
[44]. This interpretation
error can largely be avoided using a computer workstation to scroll through
the anatomy to confirm that what is seen is a pulmonary artery. To avoid
confusing a pulmonary vein and pulmonary artery, actively scroll along the
vessel to determine whether it drains into the left atrium, as a pulmonary
vein would, or merges to form the hilum of the lung and the main pulmonary
artery.
Intersegmental lymph nodes adjacent to pulmonary arteries can be confused
for emboli (Fig. 7); however,
this is less of a problem with thin-collimation acquisition on fast MDCT
scanners. For example, in an early study of CT pulmonary angiography by
Remy-Jardin et al. [30] in
1992, nine intersegmental lymph nodes were interpreted as pulmonary embolism
in three of 41 patients. However, using thin collimation, overlapping
reconstructions, active scrolling with soft-copy review, and multiplanar
reconstructions, this pitfall can be avoided.
View larger version (82K):
[in this window]
[in a new window]
[as a PowerPoint slide]
|
Fig. 7 —48-year-old man with acute onset of shortness of breath and
pleuritic chest pain. CT scan shows low-attenuation abnormality posterior to
both upper lobe segmental pulmonary arteries (arrows). This normal
lymph node tissue may be confused for pulmonary embolism, particularly if
hard-copy images are used for interpretation.
|
|
Mucous plugging in dilated bronchi appears as tubular branching
low-attenuation structures that may mimic pulmonary embolism. A false-positive
diagnosis of pulmonary embolism can be avoided by using active scrolling and
following the bronchi proximally to the central tracheobronchial tree (Figs.
8A, and
8B).
View larger version (88K):
[in this window]
[in a new window]
[as a PowerPoint slide]
|
Fig. 8A —68-year-old man with metastatic prostate cancer, left-sided
chest pain, and dyspnea. Axial CT scan shows multiple subsegmental filling
defects that mimic subsegmental emboli in right lower lobe. However, these are
mucoid-impacted subsegmental bronchi (arrows). Note small enhancing
arteries adjacent to dilated mucous-filled bronchi.
|
|
View larger version (88K):
[in this window]
[in a new window]
[as a PowerPoint slide]
|
Fig. 8B —68-year-old man with metastatic prostate cancer, left-sided
chest pain, and dyspnea. Sagittal reformatted CT scan shows large central
right hilar tumor (long arrows) that is causing mucoid impaction of
segmental and subsegmental lower lobe bronchi (short arrows).
Low-attenuation branching mucoid impaction mimics pulmonary embolism.
|
|
Accuracy of CT Pulmonary Angiography
Most of the published data on the diagnostic test characteristics of CT
pulmonary angiography have been collected on single-detector helical CT
(Table 1), with sensitivity
ranging from 53% to 100% and specificity ranging from 78% to 100%
[30,
33,
35-38,
46,
50-59].
Pulmonary angiograms were used as the reference test in many of these studies.
In several studies, there was selection bias because the study population did
not consist of consecutive patients with clinically suspected pulmonary
embolism, leading to wide variability in sensitivity and specificity.
There is still some reluctance to accept helical CT for the evaluation of
pulmonary embolism, particularly when CT findings are negative. Other tests
such as lower limb compression sonography or pulmonary angiography are
considered the next investigation of choice. In a meta-analysis of 12 studies
of CT pulmonary angiography using single-detector CT in 1,250 patients,
Safriel and Zinn [60] reported
overall sensitivity and specificity for CT pulmonary angiography as 74.1% and
89.5%, respectively. Those authors concluded that helical CT is an appropriate
first-line test for patients with suspected pulmonary embolism.
An advantage of MDCT is thin-collimation scanning with better visualization
of small pulmonary arteries, particularly at the segmental, subsegmental, and
smaller levels [40,
61,
62]. Eighty-nine percent of
segmental and 75% of subsegmental pulmonary arteries are well visualized with
1.25-mm-collimation 4-MDCT compared with 75% segmental and 36% subsegmental
pulmonary arteries with 3-mm-collimation single-detector CT
[40]. There is further
incremental improvement using 16-MDCT. Ninety-four percent of segmental and
88% of subsegmental pulmonary arteries are well visualized using 16-MDCT
(Patel et al., 2003 Society for Computed Body Tomography and Magnetic
Resonance annual meeting). There is not only improved visualization of the
subsegmental pulmonary arteries using 1-mm collimation, but also improved
interobserver agreement about the presence or absence of emboli
[40,
61].
Subsegmental emboli in patients with cardiopulmonary compromise may have
greater prognostic implications than in patients without cardiopulmonary
compromise. The presence of subsegmental emboli may be an indicator of a
thrombus burden in the deep veins of the legs, representing future emboli.
Using 3D reconstruction and 1-mm scan collimation, Coche et al.
[63] visualized 96% of
subsegmental pulmonary arteries in an ideal group of 20 patients with no lung
parenchymal abnormality or artifacts, excellent contrast bolus, and
z-axis coverage of the entire thorax including all subsegmental
pulmonary arteries.
Improved visualization of the peripheral pulmonary arteries is also seen in
patients with underlying pulmonary disease
[62]
(Fig. 9), predominantly
because of the faster scanning times, thinner collimation, and the homogeneous
pulmonary artery contrast enhancement capability of MDCT
[40,
61,
62,
64]. Paddle wheel and
multiplanar volume reformations allow continuous display of the pulmonary
arteries and may be used as an adjunct when interpreting CT pulmonary
angiography. This may lead to further incremental improvement in vascular
conspicuity, particularly for vessels that run oblique to the imaging plane
[65,
66]. CT plays a role in
evaluating the evolution of pulmonary embolism, both for resolution of emboli
over time and for chronic thromboembolic disease that may ensue.
View larger version (102K):
[in this window]
[in a new window]
[as a PowerPoint slide]
|
Fig. 9 —53-year-old man with end-stage ischemic cardiomyopathy and
bronchiolitis obliterans. CT scan shows extensive bilateral air-space disease.
Despite severity of parenchymal disease, segmental pulmonary embolism
(arrow) is shown in left lower lobe.
|
|
Interobserver Variability
CT
Interobserver agreement on a per-patient basis for the detection of acute
pulmonary embolism at CT pulmonary angiography is moderately high, with kappa
values ranging from 0.59 to 0.94
[17,
18,
38,
46,
53,
55,
57,
58,
67-72].
Chartrand-Lefebvre et al. [67]
showed good overall interobserver (
= 0.85) and intraobserver (
= 0.87) agreement. Not surprisingly, interobserver agreement was better at the
lobar level ( = 0.70) than at the segmental level ( = 0.47).
Global and vascular territory analyses using CT pulmonary angiography and
comparing with pulmonary angiography yielded kappa values for the main, lobar,
segmental, and subsegmental pulmonary arteries of 0.91, 0.78, 0.56, and 0.21,
respectively, using 3-mm-collimation single-detector CT
[72]. In a larger group of 299
patients using single-detector CT at 3-mm collimation, Perrier et al.
[46] reported excellent
interobserver agreement ( = 0.82-0.90)
[45].
Thinner collimation improves interobserver agreement, with a kappa value of
0.98 using 2-mm collimation versus 0.94 with 3-mm collimation (p <
0.05) [38]. Significantly
improved interobserver agreement is noted at the subsegmental level with
4-MDCT compared with single-detector CT
[40,
61]. When reviewers disagree,
it is usually because of suboptimal scan quality secondary to poor IV contrast
bolus, extensive motion artifact, small subsegmental arteries, extensive
pulmonary parenchymal disease, or partial volume averaging.
CT Versus V/Q Scintigraphy
There is considerable inter- and intraobserver variability (25-30%) in the
interpretation of V/Q scans for pulmonary embolism, with poor interobserver
( 
0.5) and intraobserver agreement
[73]. Despite modifications of
interpretation schemes, such as the Biello criteria, there has been no
significant improvement in interobserver agreement
[74]. Significantly better
interobserver agreement has been reported with CT
[17,
18,
36,
53,
55,
68,
75,
76]
(Table 5).
CT Versus Pulmonary Angiography
Pulmonary angiography is less accurate than previously thought,
particularly at the subsegmental level
[24]. Interobserver agreement
for the central arteries is 89% but is only 13-66% for subsegmental arteries
[16,
77,
78]. Comparing dual-section CT
with selective pulmonary angiography as a reference standard in 158 patients,
Qanadli et al. [58] found that
interobserver agreement was slightly better with CT ( = 0.78-0.94) than
pulmonary angiography ( = 0.67-0.89); those authors concluded that
helical CT could replace pulmonary angiography in most patients.
Clinical Outcome After a Negative CT Pulmonary Angiogram
When pulmonary embolism is diagnosed on the basis of CT pulmonary
angiography findings, specificity is high. Therefore, a positive diagnosis of
pulmonary embolism on CT is usually accepted. When CT pulmonary angiography is
negative for pulmonary embolism, there may be greater hesitancy to withhold
anticoagulation therapy and accept a negative CT pulmonary angiography result
as a true-negative. This reluctance is because of lingering questions about
the sensitivity of CT for the detection of subsegmental pulmonary embolism.
However, many studies have reported that a negative CT pulmonary angiogram for
pulmonary embolism is comparable to a negative catheter pulmonary angiogram in
terms of patient outcome (Table
6).
After a negative catheter angiogram, fewer than 2% of patients develop
pulmonary embolism. Two published series of 380 and 167 patients after a
negative catheter pulmonary angiogram reported a 1.6% and 1.7% incidence of
pulmonary embolism over the next 6-12 months
[79,
80]. Similar results have been
reported after a negative CT pulmonary angiogram, as listed in
Table 6, for a total of 4,233
patients with a weighted average incidence of 1.3% for venous thrombotic
disease and 0.4% for fatal pulmonary embolism
[68,
81-94].
For example, in a study from the Mayo Clinic of 993 patients with negative CT
pulmonary angiography findings, 0.5% of patients developed pulmonary embolism
and 0.3% developed fatal pulmonary embolism after negative CT pulmonary
angiography. Thus, in most patients with suspected acute pulmonary embolism
and no symptoms of deep venous thrombosis, anticoagulation therapy can be
safely withheld after negative CT pulmonary angiography.
Radiation Exposure from CT Pulmonary Angiography
Using an anthropomorphic phantom, Resten et al.
[95] reported that average
doses for single-detector CT pulmonary angiography were five times smaller
than those for catheter digital subtraction pulmonary angiography
[95]. Although radiation
exposure is higher with MDCT, a potential benefit of MDCT compared with
single-detector CT is improved visualization of the segmental and subsegmental
pulmonary arteries and greater accuracy of pulmonary embolism detection
[96].
In a recent study by Kuiper et al.
[97], the average effective
dose for 4-MDCT pulmonary angiography was 4.2 mSv compared with 7.1 mSv for
digital subtraction angiography. In pregnant patients, the mean fetal dose
with single-detector CT was recently reported as less than that for V/Q
scanning at varying gestational ages: 100-370 mGy for V/Q scanning versus
3.3-20.2 mGy (first trimester), 7.9-76.7 mGy (second trimester), and
51.3-130.8 mGy (third trimester) for CT
[98]. These doses are well
below that considered safe for fetal exposure.
Cost Effectiveness
CT pulmonary angiography is less expensive than both catheter pulmonary
angiography and V/Q scintigraphy. A study by van Erkel et al.
[7] reported a
cost-effectiveness decision model for the diagnosis of pulmonary embolism
using six diagnostic strategies, four imaging techniques (V/Q, ultrasound
venography, helical CT, pulmonary angiography), and the D-dimer
blood test. This model included both the cost of diagnosing and treating
pulmonary embolism and deep venous thrombosis, the accuracy and complications
of the tests, and the prognosis in treated and untreated patients. Helical CT
in any combination reduced mortality and improved cost-effectiveness in the
diagnostic workup of suspected pulmonary embolism. Three years later, the same
authors [8] reported that the
most cost-effective strategy for evaluating thromboembolic disease was lower
extremity sonography followed by CT pulmonary angiography if sonography was
negative.
More recently, Perrier et al.
[9] evaluated
cost-effectiveness stratified by clinical probability for pulmonary embolism.
Single-detector CT as a single test was not cost-effective. However, using
4-MDCT and assuming greater than 85% sensitivity for pulmonary embolism, CT
was the most cost-effective strategy for all clinical probabilities when
combined with lower extremity sonography and the D-dimer test.
Evaluation for Deep Venous Thrombosis
Large proximal deep venous thrombosis may lead to fatal pulmonary embolism
and residual deep venous thrombosis in proximal lower limb veins may lead to
recurrent pulmonary embolism and pulmonary hypertension
[1,
99]. Until the mid to late
1980s, conventional venography was routinely used to diagnose deep venous
thrombosis [100]. Currently,
compression sonography, a widely available, noninvasive, and inexpensive test,
is the imaging technique of choice
[101-104].
The sensitivity and specificity of sonography for veins above and including
the popliteal vein compared with conventional venography range from 92% to
100% and from 80% to 100%, respectively
[105]. However, for calf
veins, sensitivity drops to 11-92%
[106].
Compression sonography is limited in the evaluation of pelvic and abdominal
veins, obese patients, and those with complex venous anatomy. MR venography
has excellent sensitivity and specificity, but is less readily available and
is more expensive; it is usually reserved for patients with poor renal
function or iodinated contrast allergy in whom sonography is technically
difficult
[107-109].
Incidental deep venous thrombosis has been noted on conventional CT
examinations for years [110].
Direct CT venography has shown high sensitivity, specificity, and
interobserver agreement, but is invasive
[111,
112]. Indirect helical CT
venography can be combined with CT pulmonary angiography for the noninvasive
evaluation of both pulmonary embolism and deep venous thrombosis
[86,
113].
Indirect CT Venography
Combined CT pulmonary angiography and CT venography was first described by
Loud et al. in 1998 [113].
The same authors [86]
subsequently compared CT venography with lower extremity venous sonography as
a reference standard in 71 patients, 19 of whom had deep venous thrombosis
revealed on both CT venography and sonography; the sensitivity and specificity
for femoropopliteal deep venous thrombosis was 100%.
At least a dozen studies have been published comparing CT venography with
sonography as the reference (Table
7). Sensitivity ranges from 71% to 100%; specificity, 94-100%;
PPV, 67-100%; and NPV, 97-100%
[81-88,
90,
91]. When CT venography is
compared with sonography or conventional venography
(Table 8), the weighted average
sensitivity is 94.5% and specificity is 98.2%. Moderately good interobserver
agreement, with kappa values of 0.59-0.88, has been reported
[70,
84,
87].
In a large study of CT pulmonary angiography with CT venography, 58 of 650
patients had both pulmonary embolism and deep venous thrombosis, and 31
patients had isolated deep venous thrombosis
[90]. In a large multicenter
study using CT pulmonary angiography and CT venography in 541 patients, deep
venous thrombosis was present in 8% of patients. Deep venous thrombosis was
correctly identified on CT venography, but was missed on sonography in four
patients; there were no false-negative CT venograms
[82]. Another advantage of CT
venography is evaluating pelvic and abdominal veins. Kappa values for
interobserver agreement of deep venous thrombosis on CT venography are
0.56-0.88 [81,
82,
87]
(Table 8). Alternative
diagnoses or other findings clinically mimicking deep venous thrombosis may be
seen [85].
CT Venography Technique
CT venography is performed after CT pulmonary angiography with the patient
supine. After a scan delay of 2.5-4 min after the start of the injection bolus
for CT pulmonary angiography, scans are obtained from the iliac crests to the
tibial plateaus or from the diaphragm to the proximal calves in either the
caudal-to-cranial direction or the cranial-to-caudal direction. Axial or
helical scanning technique is used at 5- to 10-mm collimation. Some authors
acquire slices every 20-50 cm when the axial technique is used. We use helical
technique with a 7.5-mm collimation, a pitch of 1.375:1, and table speed of
27.5 mm/sec and scan from the tibial plateaus to the iliac crests
(Table 2). Radiation dose
increases significantly when thinner sections are obtained at shorter
intervals. The study is interpreted on the workstation using the same
technique as that used to interpret CT pulmonary angiography, but with a
narrow window width.
The additional radiation dose imparted to the patient by addition of CT
venography to CT pulmonary angiography is less of a concern for older
patients. The radiation dose must be weighed against the benefit of a single
test to evaluate for venous thromboembolism. The radiation dose is reduced by
using sequential technique with images acquired at intervals of a few
centimeters; however, a short-segment deep venous thrombosis could be
overlooked with this technique. Low-dose techniques have been tried, but
images can be noisy especially through the pelvis and in obese patients. The
use of elastic stockings has been shown to significantly increase venous
enhancement of the deep veins of the lower extremities during CT venography
[114].
Top
Abstract
Introduction
