AJR 2004; 182:993-1010
© American Roentgen Ray Society
ECG-Gated Cardiac CT
Benoit Desjardins1 and
Ella A. Kazerooni
1 Both authors: Department of Radiology, University of Michigan Medical Center,
1500 E Medical Center Dr., TC-2910A, Ann Arbor, MI 48109-0326.
Received May 5, 2003;
accepted after revision October 21, 2003.
Address correspondence to B. Desjardins.
Introduction
Coronary artery disease is the leading cause of death in the United States
and the leading cause of premature permanent disability
[1]. The latest statistics from
the American Heart Association indicate that in 2003, more than 1 million
Americans experienced an acute coronary event, more than half a million died
from coronary artery disease, and more than 12 million had a history of
symptomatic coronary artery disease. Total cost in the United States to
diagnose and treat coronary artery disease is nearly $115 billion annually
[1].
The definitive diagnosis of coronary artery disease is made by catheter
coronary angiography, an invasive and expensive procedure with associated
morbidity and mortality [2].
Therefore, catheter coronary angiography is often used as second-line
diagnostic test after noninvasive diagnostic procedures, such as a treadmill
exercise tolerance test, that provide indirect evidence of coronary artery
stenosis [3]. In the past few
years, progress has been made in the development of noninvasive diagnostic
alternatives to directly image the coronary arteries. The two most promising
noninvasive approaches are CT and MR coronary angiography.
This review describes the state of the art in cardiac imaging with CT, with
emphasis on coronary artery imaging. After a more detailed description of the
two principal CT technologies used in cardiac imaging—electron beam CT
and MDCT—the technical aspects of ECG-gated cardiac MDCT are described.
Elements relevant to the day-to-day clinical practice of cardiac imaging with
CT are emphasized, with focus on how and why things are done, the clinical
uses of this modern technology, its current limitations, and a comparison with
other imaging techniques such as catheter angiography.
Historical Background
Cardiac imaging using CT is a demanding task. Not only is high spatial
resolution required for imaging small structures such as the coronary
arteries, but high temporal resolution must also be achieved for motion-free
imaging of the heart, given heart rates that may range from 50 to more than
100 beats per minute. Submillimeter spatial resolution is necessary both
in-plane and longitudinally along the z-axis in order to view the
small branches of the coronary arteries. Although temporal resolution of 250
msec is required for motion-free imaging of the heart during diastole, 50 msec
is required for imaging during systole, during which ventricular contraction
causes the greatest cardiac motion
[4]. This statement applies
only for heart rates up to 70 beats per minute. As the heart rate increases
toward 100 beats per minute, a temporal resolution of approximately 150 msec
becomes necessary to image the heart during diastole, because the intervals of
lesser cardiac motion become shorter
[5]. The current CT techniques
target this level of temporal resolution with the use of specific
ECG-synchronized techniques, segmentation, and tailored reconstruction
algorithms. Respiratory motion must also be eliminated for cardiac imaging, so
scanning is optimally performed in a single breath-hold.
CT technology has made remarkable progress during the past three decades,
with successive generations of scanners providing faster imaging at
progressively higher spatial resolution. Following is a short overview on the
history of the developments in CT technology as it relates to cardiac
imaging.
Axial CT
In 1972, the first generation of CT scanners was introduced
[6]. This invention earned
Godfrey Hounsfield and Allan Cormack the Nobel Prize for Medicine in 1979
[7]. Such scanners were not
very efficient by modern standards. They used a pencil-beam X-ray source. One
or two detectors acquired data while the source was translated linearly across
the field of view. The whole system was then rotated by steps of 1° for
each source and detector translation, and this was repeated to span 180°.
Approximately 5 min was needed to acquire one or two tomographic images (the
earliest CT scanners were already dual-slice scanners). This approach was
rapidly improved by using a narrow fan-beam X-ray source and more detectors
per row (second generation)
[8], and then by using a wide
fan-beam X-ray source with rotation of both the source and the detectors
(third generation) [8]. These
latter axial CT scanners could acquire a tomographic image in less than 5 sec.
Cardiac imaging by CT was first reported in 1981
[9] on a third-generation CT
scanner with a 2-sec gantry rotation time, providing an effective temporal
resolution of 500–1,000 msec, still insufficient to achieve motion-free
images of the heart.
Fast Axial CT
In the late 1970s, a stationary detector ring was introduced to remove ring
artifacts in the axial CT images (fourth generation)
[8]. In 1982, both the source
and the detector were made stationary on electron beam CT scanners
[8], which were specifically
created for cardiac imaging and could acquire a tomographic image in 100 msec.
For the first time, electron beam CT provided temporal resolution sufficient
for motion-free imaging of the heart during diastole
[10].
Electron beam CT scanners differ from axial CT scanners. In electron beam
CT, electrons are accelerated in a vacuum funnel and are precisely focused
toward and swept across a 210° tungsten ring anode placed under the
patient. After hitting the target ring, a cone beam of X-ray photons is
emitted. The X-ray photons go through the patient and are captured by two
240° detector rows above the patient. Slice collimation is 3 mm, so 40
slices are needed to cover the entire heart (12 cm), for a total imaging time
of 30 sec. Imaging can usually be performed in one breath-hold. ECG-based
triggering is used for motion-free imaging during diastole. However,
significant motion artifacts still remain in approximately 20% of patients
[11]. Although in-plane
spatial resolution is submillimeter (0.8 x 0.8 mm), the resolution along
the longitudinal z-axis remains limited (3 mm) in order to image the
entire heart in a single breath-hold.
Today, electron beam CT is predominantly used for the detection of calcium
load in the coronary arteries, a prognostic factor in patients with coronary
artery disease [12,
13]. It is also used to a
lesser extent to perform CT coronary angiography
[14] and to evaluate cardiac
function [15]. The spectrum of
applications for electron beam CT is limited, and its physical setup is
cumbersome. It appears to have lost its competitive edge over the latest
all-purpose MDCT scanners, although current technologic improvement in
electron beam CT technique, such as the integration of 4-slice technology,
thinner (1.5-mm) collimation, 30- to 50-msec temporal resolution, and improved
user interfaces could reposition electron beam CT in the future.
Single-Slice Helical CT
Slip-ring technology was first introduced in CT scanners in 1989
[8] and enabled the gantry to
rotate continuously without fixed wires, offering continuous helical scanning
of entire volumes in less than 30 sec, with a 1-sec gantry rotation time and
subsecond image acquisition
[16]. Scanning and image
positions were decoupled, allowing images to be reconstructed at any arbitrary
position along the longitudinal axis. Z-axis resolution was also
improved, enabling improved multiplanar and 3D image reconstruction. General
image quality for a given slice thickness was not significantly different from
that of axial CT, and the radiation dose was also similar, given appropriate
pitch values [17].
Single-slice helical CT could be used for cardiac imaging with
ECG-synchronized protocols
[18]. Temporal resolution was
improved over nonhelical scanners but, at 250–500 msec, remained
insufficient for motion-free cardiac imaging. Almost all patients had motion
artifacts [18]. In-plane
spatial resolution was 0.6 x 0.6 mm. However, to cover the entire volume
of the heart in a single breath hold, 3-mm collimation was required, severely
limiting the z-axis resolution and the ability to evaluate the small
coronary arteries.
Fan-Beam MDCT
CT scanners have been introduced with increasing numbers of detector rows,
referred to as MDCT or multislice CT. In 1993, CT units were introduced with
two rows of detectors and 1-sec rotation time
[19]. In 1994, units with a
0.75-sec gantry rotation time were introduced
[20], which improved volume
coverage and z-axis resolution. In 1998, units with four rows of
detectors were introduced, with 0.5-sec rotation time, in-plane resolution of
0.6 x 0.6 mm, and 1.0- to 1.25-mm collimation
[21]. These units allowed an
eightfold increase in performance compared with single-slice helical CT (four
times the rows, twice the rotation speed). This gain in performance could be
used for faster coverage of the heart by increasing the table feed per
rotation, for covering a greater volume, or for improved temporal resolution
to 125–250 msec. Thinner collimation (1-mm detector) could be used to
improve z-axis resolution. General image quality was better than with
single-slice CT and remained excellent at higher scanning speeds, thereby
reducing motion artifacts.
The first MDCT scanners provided true volume image data and have enabled
motion-free cardiac imaging in diastole for heart rates up to 70 beats per
minute in a single 20- to 40-sec breath-hold. These scanners provided the
first viable alternative to electron beam CT. These systems have superior
spatial resolution to electron beam CT (0.6 x 0.6 x 1.0 mm vs 0.8
x 0.8 x 3 mm) in a single breath-hold, but inferior temporal
resolution (125–250 msec vs 100 msec). ECG gating is used to improve
temporal resolution, but motion artifacts are still present in approximately
50% of patients [22].
Radiation dose issues became more important, with doses up to 10 mSv for
cardiac MDCT versus 1.1 mSv for electron beam CT
[23], and systems were
subsequently designed with variable tube output to decrease total radiation
exposure during the CT examination. In such systems, tube current is modulated
in two ways: by projection angle, with tube current being much lower in
frontal projections than lateral projections because total attenuation
(thickness of the chest) varies according to projection angle, and by ECG,
with tube current being lower during systole because most of the diagnostic
information comes from diastolic images, which present fewer motion
artifacts.
Cone-Beam MDCT
In 2002, the latest generation of MDCT scanners was introduced, with 16
rows of detectors and even thinner collimation (0.625–0.75 mm),
providing for the first time isotropic resolution of 0.5 x 0.5 x
0.6 mm and gantry rotation times of 400–500 msec. Isotropic resolution
means similar resolution along all axes. Arrays of detectors with variable
geometry (isotropic arrays, adaptive arrays) are used. For example, this can
be 24 rows of detectors along the z-axis, with 16 central rows of
0.625-mm collimation flanked on each side by four rows of 1.25-mm collimation,
and 888 columns of detectors along the plane of the gantry (Fig.
1A,
1B). The X-ray tubes are more
powerful, with increased tube current, to provide faster scanning with
improved effective slice thickness.
View larger version (8K):
[in this window]
[in a new window]
[as a PowerPoint slide]
|
Fig. 1A. —Two adaptive geometry configurations for detector rows, with
16 central rows at 0.625-mm collimation flanked on each side by four rows at
1.25-mm collimation. Drawing shows configuration at 16 rows of 0.625-mm
collimation. Only central 16 rows of detectors are used to record a signal at
their native collimation of 0.625 mm.
|
|
View larger version (10K):
[in this window]
[in a new window]
[as a PowerPoint slide]
|
Fig. 1B. —Two adaptive geometry configurations for detector rows, with
16 central rows at 0.625-mm collimation flanked on each side by four rows at
1.25-mm collimation. Drawing shows configuration at 16 rows of 1.25-mm
collimation. Signals of central 16 rows of detectors are combined two by two
to produce an effective collimation of 1.25 mm, whereas signals of peripheral
rows are kept at their native collimation of 1.25 mm.
|
|
These systems provide motion-free cardiac imaging in diastole for heart
rates up to 80 beats per minute, in a 20-sec breath-hold. With segmented
reconstruction algorithms (discussed later in this article), a temporal
resolution of up to 65 msec can be achieved, and motion-free imaging can be
performed for heart rates up to 100 beats per minute. These MDCT scanners can
therefore offer both higher spatial resolution and higher temporal resolution
than electron beam CT and use a small form factor, and they are multipurpose
clinical units. Radiation exposure is relatively high, ranging from 5 to 10
mSv for coronary CT angiography
[23]. This exposure can be
reduced using tube current ECG-synchronized techniques
[24]. Cone-beam artifacts are
present on MDCT with more than four rows of detectors, requiring complex
cone-beam algorithms for image reconstruction
[25–27].
Volume CT
To further improve the current generation of MDCT scanners requires
increasing both temporal and spatial resolution. Increased temporal resolution
can be achieved by faster gantry rotation. This is an engineering challenge
because the gantry experiences centrifugal forces of 10 G
(gravitational constant) on 0.5-sec rotations
[21]. A realistic limit for
rotation times appears to be approximately 50–200 msec (Kalendar WA,
presented at the 2002 meeting of the European Congress of Radiology). Temporal
resolution can also be improved using better and more complex segmentation
algorithms; however, rotation time must be variable to desynchronize rotation
and cardiac cycle (discussed later in this article), limiting the
applicability of segmented algorithms
[28].
Higher isotropic spatial resolution requires more numerous and smaller
detectors, with thinner, submillimeter slice collimation. Large-area detectors
(40 x 30 cm) or volume CT may replace detector arrays and are currently
under development [29]. A pig
heart imaged on a prototype device (General Electric Medical Systems) with a
flat-panel detector made of a grid of 1024 x 1024 cells each having a
resolution of 200 µm in each axis
[30] shows increased anatomic
detail (Fig. 2). Early results
show that even the fifth-degree coronary artery branches can be visualized
with such a system [31].
However, higher isotropic spatial resolution comes at a cost. As detector
dimension decreases by N, contrast resolution decreases and noise
increases by N2, which is compensated for by an increase
in radiation dose of N4
[21]. Thus, noise and dose
factors may limit future increases in spatial resolution.
View larger version (116K):
[in this window]
[in a new window]
[as a PowerPoint slide]
|
Fig. 2. —Image of excised pig heart obtained with experimental
flat-panel CT unit being developed by General Electric Medical Systems.
Coronary arteries were cannulated and filled with solution of barium sulfate
(Micropaque Colon, Guerbet) before imaging. Coronary CT angiogram
(collimation, 200 x 200 µm; cell pitch, 0.2 mm) shows definition of
coronary arteries with multiple small branches much better than current
technology (16-slice MDCT). (Reprinted with permission from
[30])
|
|
MDCT: General Principles
The basic principles of MDCT are relatively simple
[28]
(Fig. 3A). The X-ray point
source and the detector array are placed on opposite sides of the patient on a
ringlike structure called the gantry. The gantry rotates around the patient,
who is located on a table at its center. The table moves at constant speed
along the axis of the gantry. X rays are emitted toward the patient, penetrate
the patient, and are captured by one or more detectors. This process generates
a series of helical projections of the patient's attenuation properties.
Images representing X-ray attenuation at each point in the volume traversed by
the photons are then mathematically reconstructed from the helical projection
data (Fig. 3B).
View larger version (53K):
[in this window]
[in a new window]
[as a PowerPoint slide]
|
Fig. 3A. —Drawings show helical projections from MDCT. Detectors follow
3D helical path, with table advancing at constant speed while gantry is
rotating. Tube emits X-ray radiation (yellow) that is recorded by
detectors. Resulting set of projections has helical configuration in
space.
|
|
View larger version (23K):
[in this window]
[in a new window]
[as a PowerPoint slide]
|
Fig. 3B. —Drawings show helical projections from MDCT. Images are
reconstructed from projection data by linear interpolation from projections
closest to image plane. Advanced algorithms correct for cone-beam
geometry.
|
|
Effective slice thickness using a single-slice detector scanner is
determined by collimation, whereas in multidetector scanners it is determined
by both the detector configuration and the reconstruction method. The table
travel per complete rotation of the gantry divided by the X-ray beam width is
called the "beam pitch." The beam pitch conforms to the current
industry standard of pitch and is used in this article. In multidetector
scanners, another definition of pitch is sometimes used: the table travel per
complete rotation of the gantry divided by the detector width, called the
"detector pitch." With a (beam) pitch of 0, there is no table
motion and scanning is axial. With a pitch of 1, the table displacement for
each rotation is equal to the z-axis dimension of the array of active
rows of detectors. A pitch greater than 1 implies gaps in the helix of
projections, whereas a pitch between 0 and 1 implies overlap between the
projections (Fig. 4A,
4B,
4C).
Images are reconstructed from linear interpolation of projection data from
rays that are the closest to the image plane
(Fig. 3B), using algorithms to
correct for cone-beam geometry. A minimum of 180° of projection data are
mathematically required to reconstruct a complete image. Thus, slice thickness
increases with pitch, which determines how much the table travels during a
180° rotation. The thinnest slice is equal to the height of a single row
of detectors (the detectors are actually wider than the slice to compensate
for geometric magnification). Thicker slices can be generated mathematically
by combining the thinner slices. The determinants of in-plane resolution are
focal spot size, detector width and geometry, reconstruction algorithm, and
image matrix size. The determinants of longitudinal z-axis resolution
are focal spot size, detector height and collimation, and reconstruction
algorithm. The determinants of temporal resolution are gantry rotation speed
and reconstruction algorithm.
Cardiac Gating
Gating techniques are used to improve temporal resolution and minimize
imaging artifacts caused by cardiac motion. Two approaches to cardiac gating
are typically used: prospective ECG triggering and retrospective ECG
gating.
The least cardiac motion occurs during diastole, when the ventricles are
passively filling. Prospective ECG triggering uses the ECG signal to control
scanning, so that X rays are generated and projection data are acquired only
during cardiac diastole, more than half the rotation of the gantry. The total
number of slices produced per heartbeat during this half rotation of the
gantry is proportional to the number of rows of active detectors. Because an
axial scanning method (rather than helical) is typically used and the table
has to move by the total collimation width after each acquisition, one
heartbeat typically has to be skipped between each acquisition. About 12 cm of
scanning is required to cover most heart sizes, which requires approximately
48 heartbeats for single-slice CT (5-mm collimation), 24 heartbeats for
4-slice MDCT (2.5-mm collimation each row), and 12 heartbeats for 16-slice
MDCT (1.25-mm collimation each row). Thus, multidetector technology can obtain
the entire scan during one breath-hold. Electron beam CT requires about 24
heartbeats and can scan the entire heart in one breath-hold. The start of the
diastolic phase of the cardiac cycle is estimated from the prior three to
seven consecutive heartbeats and occurs approximately 450 msec before the R
wave on the ECG.
Prospective triggering techniques have important limitations. They are
sensitive to heart rate changes and arrhythmias, but they have limited spatial
z-axis resolution in order to cover the entire heart in a single
breath-hold. They are effective only for heart rates of less than 90 beats per
minute and perform poorly with arrhythmias, such as in atrial fibrillation. To
overcome these limitations, retrospective ECG gating techniques are commonly
used, at the expense of a higher radiation dose. Retrospective gating
techniques allow faster continuous cardiac volume coverage, improved
z-axis resolution, and imaging of the entire cardiac cycle for
functional analysis.
In retrospective techniques, partially overlapping MDCT projections are
continuously acquired, and the ECG signal is simultaneously recorded.
Algorithms are then used to sort the data from different phases of the cardiac
cycle by progressively shifting the temporal window of acquired helical
projection data relative to the R wave
(Fig. 5). Every position of the
heart must be covered by a detector row at every point during the cardiac
cycle. This means that the scanner table must not advance more than the total
width of the active detectors for each heartbeat. Helical pitch can be varied
proportionally to the heart rate to achieve continuous volume coverage.
Typical pitch for an average heart rate of 70 beats per minute is 0.3:1, with
a total scanning time of about 20 sec for a 16-slice MDCT scanner using
0.625-mm collimation.
View larger version (11K):
[in this window]
[in a new window]
[as a PowerPoint slide]
|
Fig. 5. —Temporal window for reconstruction from projection data is
approximately 250 msec. Drawing shows that center of window can be located
anywhere during heart cycle. Left box has its center at 10% of R-R interval,
which is during systole. Right box has its center at 70% of R-R interval,
which is during diastole and is most common motion-free imaging temporal
window for heart. On ECG signal, P represents atrial contraction; Q, R, and S
represent ventricular contraction; and T represents ventricular
relaxation.
|
|
Two main algorithmic approaches are used to perform retrospective cardiac
gating: partial scanning and segmented adapted scanning. To reconstruct an
image, a minimum helical projection data segment of 180° must be available
for every fan angle, corresponding to a rotation of 180° plus the breadth
of the fan beam, so approximately two thirds of a full gantry rotation is
required (Fig. 6A). This
technique is called partial scanning. Temporal resolution is therefore two
thirds of rotation time. Parallel beam geometries and rebinning techniques can
be used to decrease the minimum data segment to 180°, with a temporal
resolution of half the rotation time, or 250 msec for 16-slice MDCT scanners
[32]. The partial scanning
technique is typically used in patients with heart rates between 40 and 75
beats per minute.
View larger version (36K):
[in this window]
[in a new window]
[as a PowerPoint slide]
|
Fig. 6A. —Drawings show two types of retrospective reconstruction
algorithms. In partial scanning algorithm, continuous segment of projection
data at single heartbeat is used to reconstruct image.
|
|
To improve temporal resolution, segmented adaptive reconstruction can be
used, which involves combining shorter segments of projection data from two of
more subsequent cardiac cycles
[33]
(Fig. 6B). Temporal resolution
is equal to that of the longest projection data segment. Maximum values for
temporal resolution are 125 msec for segmentation over two cardiac cycles and
65 msec over four cardiac cycles. Either volume coverage or longitudinal
resolution may need to be reduced to maintain a low pitch and still scan in a
single breath-hold.
View larger version (37K):
[in this window]
[in a new window]
[as a PowerPoint slide]
|
Fig. 6B. —Drawings show two types of retrospective reconstruction
algorithms. In segmented adaptive algorithm, different segments of projection
data from same phase of cardiac cycle at successive heartbeats are used to
reconstruct image. Cardiac cycle and gantry rotation must not be synchronized
for different segments to cumulatively cover large enough range of projection
angles to reconstruct image from data.
|
|
Technical Considerations
Several elements of cardiac CT can be optimized to produce the best
possible images at the lowest dose of radiation.
Patient Position
The position of the patient in the CT scanner is important, because a
patient placed off-center in the scanner produces images with suboptimal
temporal resolution. The temporal resolution of any given pixel in an image
depends on the temporal window of the projection data used to reconstruct that
specific pixel. In helical CT using a partial scanning algorithm, a minimum
helical projection data segment of 180° must be available for every fan
angle, so each image is reconstructed from a segment of projection data longer
than 180°. However, not every detector contributes data at each projection
angle to reconstruct an image. For example, if a patient is scanned clockwise
from the 11-o'clock to the 7-o'clock position to produce an image (Fig.
7A,
7B,
7C,
7D), the left side of the
chest in the image has a much tighter temporal window than the right side. The
temporal resolution of the points of the image in this example increases
almost linearly between the edges of the field of view going from the left
side of the chest to the right side
[34]. By positioning the heart
near the center of rotation of the gantry, the temporal resolution of its
elements remains constant and average, rather than going from bad to better as
the X-ray tube rotates on the gantry.
View larger version (62K):
[in this window]
[in a new window]
[as a PowerPoint slide]
|
Fig. 7A. —Drawings show variable temporal resolution across image.
Segment of projection data from 11-o'clock to 7-o'clock position is used. At
extremities of that segment, only partial projection data are used to
reconstruct image. In this specific image, left side of chest has much tighter
temporal window than right side. At 11-o'clock position, data from right chest
are not used in reconstruction.
|
|
View larger version (67K):
[in this window]
[in a new window]
[as a PowerPoint slide]
|
Fig. 7B. —Drawings show variable temporal resolution across image.
Segment of projection data from 11-o'clock to 7-o'clock position is used. At
extremities of that segment, only partial projection data are used to
reconstruct image. In this specific image, left side of chest has much tighter
temporal window than right side. At 2-o'clock (B) and 4-o'clock
(C) positions, all data are used.
|
|
View larger version (67K):
[in this window]
[in a new window]
[as a PowerPoint slide]
|
Fig. 7C. —Drawings show variable temporal resolution across image.
Segment of projection data from 11-o'clock to 7-o'clock position is used. At
extremities of that segment, only partial projection data are used to
reconstruct image. In this specific image, left side of chest has much tighter
temporal window than right side. At 2-o'clock (B) and 4-o'clock
(C) positions, all data are used.
|
|
View larger version (62K):
[in this window]
[in a new window]
[as a PowerPoint slide]
|
Fig. 7D. —Drawings show variable temporal resolution across image.
Segment of projection data from 11-o'clock to 7-o'clock position is used. At
extremities of that segment, only partial projection data are used to
reconstruct image. In this specific image, left side of chest has much tighter
temporal window than right side. At 7-o'clock position, data from right chest
are not used.
|
|
Heart Rate
The heart rate of the patient is an important determinant of image
reconstruction quality and for selecting scanning parameters such as pitch and
reconstruction algorithms (pitch is covered in the next section). If the heart
rate is too high (> 100 beats per minute), a temporal resolution of less
than 150 msec becomes necessary for motion-free cardiac imaging during
diastole. Segmented adaptive reconstruction can be used to improve temporal
resolution, but at the expense of volume coverage or longitudinal resolution,
to keep scanning time in one breath-hold. If neither can be sacrificed,
ß-blocker medication can be administered orally or IV 1 hr before
scanning to reduce the heart rate
[35].
If segmented adaptive reconstruction is used, improved temporal resolution
can be achieved only if the patient's heart rate is not synchronized with the
rotation cycle of the gantry. If they are synchronized, the same cardiac phase
corresponds to the same angular segment of projection data at every rotation,
and it is therefore not possible to build the 180° span of projection data
required to reconstruct individual images.
A patient's heart rate may vary during scanning. The heart rate decreases
after breath-holding [36]. If
the heart rate varies during scanning after the rotation time and pitch are
selected by the operator, the temporal resolution will vary from image to
image.
Pitch
Helical pitch is a parameter that is selected before the examination is
acquired. Helical pitch is varied proportionally to the heart rate to achieve
continuous volume coverage (Fig.
8). If the heart rate increases, the pitch can increase. If the
heart rate decreases, the pitch must decrease. If the pitch is too high given
the heart rate, gaps in the image data set are present. A pitch that is too
low implies increased radiation exposure and increased duration of the
breath-hold. Before scanning, the heart rate of the patient should be assessed
under breath-holding conditions, and the lowest expected heart rate should be
used to select the pitch. The latest scanners can automatically adjust the
pitch when variations in the heart rate are detected.
Collimation
Scans should be acquired at the best possible spatial resolution for the
specific task at hand. Coronary angiography requires the best resolution. The
left main artery has a diameter of 4 mm, and the distal left anterior
descending and circumflex arteries have diameters of 1 mm
[37]. Submillimeter spatial
resolution, both in-plane and longitudinally, is therefore required for
coronary CT angiography. Other examinations, such as calcium scoring, do not
require longitudinal spatial resolution this fine and are performed with
slightly larger collimation. It is always possible to reconstruct thicker
slices from the projection data of thinner slices, but thinner slices require
higher radiation exposure.
Dosimetry
Radiation exposure for cardiac CT is relatively high because of continuous
overlapping scanning and the use of retrospective cardiac gating. Coronary CT
angiography has a 5- to 10-mSv exposure, which is more than the typical 2 mSv
of traditional diagnostic catheter coronary angiography
[23]. Calcium scoring has an
exposure of 1–2 mSv
[23]. Modern scanners
incorporate fine control of tube current with respect to cardiac cycle, and
this will become more important as spatial resolution is improved in future
units. Radiation dose should be high enough to maintain an appropriate
contrast-to-noise ratio for diagnostic quality images, but no higher.
Clinical Indications
The two most important clinical indications for cardiac CT are calcium
scoring and coronary CT angiography. CT may also be used to characterize
coronary plaque and to evaluate cardiac function, myocardial perfusion,
infarcts, tumors, pericardial disease, postsurgical complications, and
congenital malformations [38].
The exact roles of coronary CT angiography for atherosclerotic heart disease
and patient selection criteria for coronary CT angiography are under
investigation.
Coronary Artery Calcium Imaging
Coronary artery plaques can be calcified. Even small amounts of
calcification can be identified on CT. Since the advent of electron beam CT,
coronary calcium has been a popular but controversial noninvasive method to
assess coronary artery disease. The load of coronary calcium increases by
about 15–25% per year without treatment, and either slows or stops
during statin therapy [39].
Large amounts of coronary calcium are a sensitive, but not a specific, marker
for coronary artery disease
[12]. Yet the value of
coronary calcium as an independent or superior risk factor to predict cardiac
events has not been fully established
[40]. The current American
College of Cardiology and American Heart Association guidelines
[12] describe two indications
for calcium scoring: to detect coronary calcium in patients with atypical
chest pain, and to quantify and follow up calcified coronary plaque in
asymptomatic patients with other positive cardiovascular risk factors.
Acquisition
A typical protocol used for 16-slice MDCT is as follows: axial 16-row mode,
1.25-mm collimation, 20-mm table feed between each acquisition, reconstruction
of eight images at 2.5-mm thickness per table location, 0.4- to 0.5-sec
rotation, scanning from the carina to cardiac apex, X-ray tube at 120 kV and
300–320 mAs, field of view of 25 cm, and prospective gating with a delay
of 70%. The protocol may be executed twice for reproducibility assessment.
Ideally, a continuous volume containing the heart should be scanned as rapidly
as possible in one breath-hold to minimize motion, with thin slices and the
lowest radiation exposure providing a sufficiently high contrast-to-noise
ratio. Cardiac gating should be used to avoid gaps. In this case, prospective
cardiac gating is used. A CT phantom can be placed underneath the patient to
calibrate the attenuation values in the final images
[13]. Initial scout
posteroanterior and lateral views are obtained to locate the position of the
heart. After a period of hyperventilation, the patient is asked to
breath-hold. The approximate image acquisition time by axial cine protocol is
15–16 sec, the exposure time is 3.4 sec, and the total radiation dose is
1–2 mSv [23].
Postprocessing
The total coronary calcium load is determined semiautomatically. A
threshold of 130 H and higher is used, and all clusters of calcifications in
the coronary arteries are identified by a postprocessing algorithm and
quantified. Clusters smaller than two pixels are considered to be noise and
are not included in the final assessment. For semiquantitative assessment of
coronary artery calcium, the Agatston calcium score is the most widely used
[41]. The total score is the
sum of all scores of calcium clusters on all images of the scan (Fig.
9A,
9B,
9C,
9D). The score of each single
lesion on each image is the product of its area multiplied by a cofactor
reflecting calcium density in each lesion. More precisely, the cofactor
indicates maximal attenuation in the lesion (1, 130–199 H; 2,
200–299 H; 3, 300–399 H; and 4, 
400 H). The Agatston score is
highly dependent on the imaging parameters and shows variable reproducibility,
especially for small amounts of calcium, with variations in the scores by 20%
or more from scan to scan
[42]. Volumetric
quantification algorithms are also available for calcium scoring
[43]. They have a higher
reproducibility (7–25% variability)
[43] and could replace the
Agatston score in the future.
View larger version (94K):
[in this window]
[in a new window]
[as a PowerPoint slide]
|
Fig. 9A. —55-year-old man with multiple cardiovascular risk factors who
underwent 16-slice MDCT (collimation, 16 x 1.25 mm; rotation time, 500
msec) for calcium scoring. Scores from all lesions on all images are summed to
generate total calcium score. In this case, total score was 750. Images at
four levels are presented. MDCT scans show calcification in left main coronary
artery (LMA) and left anterior descending (LAD) arteries.
|
|
View larger version (93K):
[in this window]
[in a new window]
[as a PowerPoint slide]
|
Fig. 9B. —55-year-old man with multiple cardiovascular risk factors who
underwent 16-slice MDCT (collimation, 16 x 1.25 mm; rotation time, 500
msec) for calcium scoring. Scores from all lesions on all images are summed to
generate total calcium score. In this case, total score was 750. Images at
four levels are presented. MDCT scans show calcification in left main coronary
artery (LMA) and left anterior descending (LAD) arteries.
|
|
View larger version (105K):
[in this window]
[in a new window]
[as a PowerPoint slide]
|
Fig. 9C. —55-year-old man with multiple cardiovascular risk factors who
underwent 16-slice MDCT (collimation, 16 x 1.25 mm; rotation time, 500
msec) for calcium scoring. Scores from all lesions on all images are summed to
generate total calcium score. In this case, total score was 750. Images at
four levels are presented. MDCT scans show calcification in right coronary
(RCA) and left circumflex (LCX) arteries.
|
|
View larger version (133K):
[in this window]
[in a new window]
[as a PowerPoint slide]
|
Fig. 9D. —55-year-old man with multiple cardiovascular risk factors who
underwent 16-slice MDCT (collimation, 16 x 1.25 mm; rotation time, 500
msec) for calcium scoring. Scores from all lesions on all images are summed to
generate total calcium score. In this case, total score was 750. Images at
four levels are presented. MDCT scans show calcification in right coronary
(RCA) and left circumflex (LCX) arteries.
|
|
Comparison with Electron Beam CT
Although calcium scoring has been performed since the 1990s using electron
beam CT, it is slowly being replaced by scoring with MDCT because of the
greater availability of the latter type of units. MDCT has a poorer temporal
resolution than does electron beam CT (250 vs 100 msec), so images by MDCT
have more motion artifacts. However, MDCT has higher spatial resolution than
electron beam CT, lower image noise, and similar radiation exposure (1.0 vs
0.7 mSv) [23]. Noise can be
decreased in MDCT by varying tube current and voltage, which is useful for
obese patients. Increase in overlap of projection data (decreasing pitch)
results in higher reproducibility than electron beam CT
[44].
Coronary Artery Lumen Imaging
By far the most exciting new development of MDCT is noninvasive motion-free
imaging of the lumen of the coronary arteries. The latest 16-slice technology
[45] enables submillimeter
isotropic resolution in all three axes, and therefore excellent reformatting
in any imaging plane. This technology makes possible evaluation of many small
structures, such as distal submillimeter coronary artery branches, that was
not possible with 4-slice CT. It also makes feasible reliable visualization
and assessment of grafts, stent patency, and in-stent stenosis.
Today most coronary CT angiography is still experimental. CT angiography is
clinically indicated for evaluation of the patency of coronary vessels after
intervention, for evaluation for a suspected anomalous coronary artery origin,
and in patients with atypical chest pain. CT angiography may be useful for
follow-up of patients undergoing statin therapy
[46], to assess cardiac
valves, and in other cardiac diseases such as tumors and pericardial
disease.
Electron beam CT has been reported to have high sensitivity, specificity,
and negative predictive value for assessment of the proximal and mid segments
of the coronary arteries [47,
48]. The limited longitudinal
spatial resolution of electron beam CT (1.5–3.0 mm) in a single
breath-hold limits its assessment of distal coronary arteries. Images are also
degraded by motion, and imaging of several coronary segments (e.g., the mid
right coronary artery and the circumflex artery) is often of poor quality
because of the atrial contraction at end-diastole. Studies have found that
approximately 25% of coronary artery segments must be excluded because of poor
image quality [47,
49]. The same is true for
4-slice MDCT, for which visibility of coronary segments is optimal for heart
rates only below 65 beats per minute
[50].
Because of its shorter acquisition time, superior spatial resolution
(0.625-mm longitudinal resolution), and temporal resolution approaching that
of electron beam CT, 16-slice MDCT is currently superior to electron beam CT
and 4-slice MDCT for the evaluation of coronary artery disease
[51], especially in coronary
segments with diameters smaller than 2 mm.
At least four elements affect the quality of the images produced by CT
angiography. First, the presence of heavy calcification creates streak
artifacts and markedly limits the assessment of vessel lumen. The possible
development of calcium subtraction techniques may minimize or eliminate this
artifact. Second, excessive motion can markedly degrade image quality. The
duration of the period of low motion decreases at high heart rates, making
imaging more difficult. A heart rate greater than 75 beats per minute and
arrhythmia are therefore limiting. The motion of the left main, left anterior
descending, and circumflex coronary arteries mainly follows the motion of the
left ventricle. The motion of the right coronary artery is related to the
contraction of the right atrium and ventricle and is not negligible at
end-diastole. Good assessment of the mid portion of the right coronary artery
is especially problematic in many cases
[52], a problem also seen with
coronary MR angiography. Third, coronary CT angiography becomes less reliable
for small vessels with a lumen diameter of 0.5–1.5 mm. Fourth, a
breath-hold of approximately 30 sec may be required for adequate coronary
imaging, depending on the heart size and heart rate. Although such a
breath-hold is not a problem for normal test subjects (or with CT phantoms
[45]), it can prove difficult
for patients with severe cardiac disease.
Acquisition
A typical protocol for 16-slice MDCT is as follows: helical 16-row mode,
0.625-mm collimation (1.25-mm collimation if scanning time is > 30 sec),
400- to 600-msec rotation, pitch of 0.275:1–0.3:1, scanning from the
carina to the cardiac apex, X-ray tube at 120 kV and 340–400 mAs, field
of view of 25 cm, and retrospective gating. Initial scout frontal and lateral
images are obtained and the position of the heart is identified. After a
period of hyperventilation, the patient is asked to breath-hold. Unenhanced CT
through the coronary arteries is performed for the purpose of calcium scoring
and to confirm location of the heart. This is followed by a timing bolus to
determine the scanning delay for coronary CT angiography. A single axial scan
at 5-mm collimation approximately 1 cm below the carina is obtained every 2
sec to determine peak aortic root enhancement (Fig.
10A,
10B,
10C,
10D,
10E). A region of interest is
placed in the root of the aorta, near the origin of the left main coronary
artery. A test bolus of a total of 15–20 mL of iohexol 300
(Omnipaque-300, Nycomed Imaging) is administered at a rate of 4 mL/sec.
Additional parameters are field of view, 25 cm; and X-ray tube, 120 kV and 40
mAs. The image acquisition and exposure time is 12–15 sec, and the total
radiation dose is 0.33 mSv (Goodsitt MM, unpublished data). The scanning delay
is typically equal to the time from the start of the injection to peak
enhancement of the ascending aorta, plus 3–5 sec for filling of the
distal coronary arteries.
View larger version (137K):
[in this window]
[in a new window]
[as a PowerPoint slide]
|
Fig. 10A. —51-year-old man who underwent 16-slice MDCT (collimation, 16
x 0.625 mm; rotation time, 500 msec). Sequential images every 2 sec show
progression of timing bolus. Region of interest (circle) is centered
on aortic root to determine peak enhancement for timing purposes. Total of
15–20 mL of contrast agent is administered at rate of 4 mL/sec, and
images are acquired until enhancement in region of interest reaches predefined
threshold. MDCT scans show that after 2 sec of injection, only superior vena
cava is enhanced (A); after 10 sec of injection, pulmonary arteries
enhance (B); after 20 sec, aorta shows early enhancement (C);
and after 30 sec, aorta fully enhances at value greater than 100 H
(D).
|
|
View larger version (149K):
[in this window]
[in a new window]
[as a PowerPoint slide]
|
Fig. 10B. —51-year-old man who underwent 16-slice MDCT (collimation, 16
x 0.625 mm; rotation time, 500 msec). Sequential images every 2 sec show
progression of timing bolus. Region of interest (circle) is centered
on aortic root to determine peak enhancement for timing purposes. Total of
15–20 mL of contrast agent is administered at rate of 4 mL/sec, and
images are acquired until enhancement in region of interest reaches predefined
threshold. MDCT scans show that after 2 sec of injection, only superior vena
cava is enhanced (A); after 10 sec of injection, pulmonary arteries
enhance (B); after 20 sec, aorta shows early enhancement (C);
and after 30 sec, aorta fully enhances at value greater than 100 H
(D).
|
|
View larger version (144K):
[in this window]
[in a new window]
[as a PowerPoint slide]
|
Fig. 10C. —51-year-old man who underwent 16-slice MDCT (collimation, 16
x 0.625 mm; rotation time, 500 msec). Sequential images every 2 sec show
progression of timing bolus. Region of interest (circle) is centered
on aortic root to determine peak enhancement for timing purposes. Total of
15–20 mL of contrast agent is administered at rate of 4 mL/sec, and
images are acquired until enhancement in region of interest reaches predefined
threshold. MDCT scans show that after 2 sec of injection, only superior vena
cava is enhanced (A); after 10 sec of injection, pulmonary arteries
enhance (B); after 20 sec, aorta shows early enhancement (C);
and after 30 sec, aorta fully enhances at value greater than 100 H
(D).
|
|
View larger version (148K):
[in this window]
[in a new window]
[as a PowerPoint slide]
|
Fig. 10D. —51-year-old man who underwent 16-slice MDCT (collimation, 16
x 0.625 mm; rotation time, 500 msec). Sequential images every 2 sec show
progression of timing bolus. Region of interest (circle) is centered
on aortic root to determine peak enhancement for timing purposes. Total of
15–20 mL of contrast agent is administered at rate of 4 mL/sec, and
images are acquired until enhancement in region of interest reaches predefined
threshold. MDCT scans show that after 2 sec of injection, only superior vena
cava is enhanced (A); after 10 sec of injection, pulmonary arteries
enhance (B); after 20 sec, aorta shows early enhancement (C);
and after 30 sec, aorta fully enhances at value greater than 100 H
(D).
|
|
View larger version (12K):
[in this window]
[in a new window]
[as a PowerPoint slide]
|
Fig. 10E. —51-year-old man who underwent 16-slice MDCT (collimation, 16
x 0.625 mm; rotation time, 500 msec). Sequential images every 2 sec show
progression of timing bolus. Region of interest (circle) is centered
on aortic root to determine peak enhancement for timing purposes. Total of
15–20 mL of contrast agent is administered at rate of 4 mL/sec, and
images are acquired until enhancement in region of interest reaches predefined
threshold. Graph illustrates that in this case 100 H was reached at 30 sec
(image 15). Scanning delay used for coronary CT angiography is 30 sec + 5 sec
to allow contrast agent to fill arteries.
|
|
The next step is the actual coronary CT angiography. A total dose of
100–150 mL of iohexol 300 is administered through an antecubital vein
using a power injector at a rate of 4 mL/sec, using the scanning delay
determined from the timing bolus. The rapid scanning time of the 16-slice MDCT
scanner reduces the duration of the single breath-hold scan and therefore
reduces the amount of IV contrast material necessary for the examination. This
step is also performed after a period of hyperventilation. The image
acquisition and exposure time is 15–30 sec.
Postprocessing
Postprocessing steps include retrospective reconstruction of the images at
a resolution of 0.6 x 0.6 x 0.625 mm, at phases 70%, 75%, and 80%
of the R-R interval. A partial scanning algorithm is used if the heart rate is
less than 60 beats per minute; otherwise, segmented adaptive scanning is used.
The sequences of images at all three reconstructed phases are transferred to a
processing workstation for further analysis with specialized software. The
three sequences are initially compared, and the sequence showing the least
amount of motion is selected for further processing. In practice, the sequence
of images reconstructed at 75% is the most often used for further processing,
so the other two sequences are examined only if the former is contaminated by
too much motion.
Initial processing of the acquired data produces an image of the global
coronary tree, using 3D surface shading and 3D volume rendering. Examination
of the coronary arteries is then performed, with the assistance of coronary
artery analysis software, using 2D multiplanar reformations, 2D maximum
intensity projections, and curved coronary artery reformations (Fig.
11A,
11B,
11C,
11D). The American College of
Cardiology–American Heart Association guidelines for segmental anatomy
[53] and lesion morphology
[54] are used for lesion
characterization. Each stenosis is graded by percentage of diameter, with
reference to the coronary artery diameter proximal and distal to the stenosis.
Bypass grafts are included in the scan and are evaluated from their proximal
to distal anastomoses. Multiple findings can be identified in the coronary
arteries, such as stents, aneurysms, or an aberrant origin (Figs.
12,
13,
14,
15,
16).
View larger version (135K):
[in this window]
[in a new window]
[as a PowerPoint slide]
|
Fig. 11A. —50-year-old man with atypical chest pain who underwent
16-slice MDCT (collimation, 16 x 0.625 mm; rotation time, 500 msec;
table feed, 6 mm/sec) for coronary CT angiography. Multiplanar reformatting
illustrates left anterior descending coronary artery (A) and right
coronary artery (B).
|
|
View larger version (124K):
[in this window]
[in a new window]
[as a PowerPoint slide]
|
Fig. 11B. —50-year-old man with atypical chest pain who underwent
16-slice MDCT (collimation, 16 x 0.625 mm; rotation time, 500 msec;
table feed, 6 mm/sec) for coronary CT angiography. Multiplanar reformatting
illustrates left anterior descending coronary artery (A) and right
coronary artery (B).
|
|
View larger version (46K):
[in this window]
[in a new window]
[as a PowerPoint slide]
|
Fig. 11C. —50-year-old man with atypical chest pain who underwent
16-slice MDCT (collimation, 16 x 0.625 mm; rotation time, 500 msec;
table feed, 6 mm/sec) for coronary CT angiography. Straightening of coronary
artery by powerful algorithms assists visualization and aids in quantifying
stenosis.
|
|
View larger version (121K):
[in this window]
[in a new window]
[as a PowerPoint slide]
|
Fig. 11D. —50-year-old man with atypical chest pain who underwent
16-slice MDCT (collimation, 16 x 0.625 mm; rotation time, 500 msec;
table feed, 6 mm/sec) for coronary CT angiography. Surface shading of heart
shows 3D course of coronary arteries on its surface. Asterisk indicates right
ventricle (RV) outflow. AA = ascending thoracic aorta, LAD = left anterior
descending, RA = right atrium, RCA = right coronary artery.
|
|
View larger version (132K):
[in this window]
[in a new window]
[as a PowerPoint slide]
|
Fig. 12. —60-year-old man who underwent 16-slice MDCT (collimation, 16
x 0.625 mm; rotation time, 500 msec; table feed, 6 mm/sec) for coronary
CT angiography. MDCT scan shows dense calcifications in left main
(arrow) and left anterior descending (arrowheads) coronary
arteries.
|
|
Top
Introduction
Historical Background
