See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/51369184

Artifacts in CT: Recognition and Avoidance

Article in Radiographics · November 2004

DOI: 10.1148/rg.246045065 · Source: PubMed

CITATIONS READS
1,494 15,948

2 authors, including:

Nicholas Keat
Imanova Ltd
25 PUBLICATIONS 1,814 CITATIONS

SEE PROFILE

All content following this page was uploaded by Nicholas Keat on 22 June 2016.

The user has requested enhancement of the downloaded file.

 EDUCATION EXHIBIT 1679
RadioGraphics
Artifacts in CT: Recog-
nition and Avoidance1
CME FEATURE
Julia F. Barrett, MSc ● Nicholas Keat, MSc
See accompanying
test at http://
www.rsna.org Artifacts can seriously degrade the quality of computed tomographic
/education (CT) images, sometimes to the point of making them diagnostically
/rg_cme.html
unusable. To optimize image quality, it is necessary to understand why
artifacts occur and how they can be prevented or suppressed. CT arti-
LEARNING facts originate from a range of sources. Physics-based artifacts result
OBJECTIVES
from the physical processes involved in the acquisition of CT data. Pa-
FOR TEST 5
tient-based artifacts are caused by such factors as patient movement or
After reading this
article and taking the presence of metallic materials in or on the patient. Scanner-based
the test, the reader
will be able to: artifacts result from imperfections in scanner function. Helical and
䡲 Identify the various multisection technique artifacts are produced by the image reconstruc-
types of artifacts that tion process. Design features incorporated into modern CT scanners
can appear in CT
images. minimize some types of artifacts, and some can be partially corrected
䡲 Discuss the reasons by the scanner software. However, in many instances, careful patient
why these artifacts
occur.
positioning and optimum selection of scanning parameters are the
䡲 Describe the meth- most important factors in avoiding CT artifacts.
ods of avoiding or ©
RSNA, 2004
suppressing artifacts
available with mod-
ern CT systems.

Index terms: Computed tomography (CT), artifact ● Computed tomography (CT), image quality ● Images, artifact ● Images, quality

RadioGraphics 2004; 24:1679 –1691 ● Published online 10.1148/rg.246045065 ● Content Codes:

1From Imaging Performance Assessment of CT Scanners (ImPACT), St George’s Hospital, Blackshaw Rd, London SW17 0QT, England. Presented
as an education exhibit at the 2003 RSNA scientific assembly. Received April 5, 2004; revision requested May 7; final revision received August 20; ac-
cepted September 1. Both authors have no financial relationships to disclose. Address correspondence to J.F.B. (e-mail: [email protected]).
©
RSNA, 2004
 1680 November-December 2004 RG f Volume 24 ● Number 6

RadioGraphics
Introduction
In computed tomography (CT), the term artifact
is applied to any systematic discrepancy between
the CT numbers in the reconstructed image and
the true attenuation coefficients of the object. CT
images are inherently more prone to artifacts than
conventional radiographs because the image is
reconstructed from something on the order of a
million independent detector measurements. The
reconstruction technique assumes that all these
measurements are consistent, so any error of Figure 1. Changing energy spectrum of an x-ray
beam as it passes through water. The mean energy in-
measurement will usually reflect itself as an error creases with depth. (The attenuated spectra have been
in the reconstructed image. The types of artifact rescaled to be equivalent in size to the unattenuated
that can occur are as follows: (a) streaking, which spectra.)
is generally due to an inconsistency in a single
measurement; (b) shading, which is due to a
group of channels or views deviating gradually
from the true measurement; (c) rings, which are
due to errors in an individual detector calibration;
and (d) distortion, which is due to helical recon-
struction.
It is possible to group the origins of these arti-
facts into four categories: (a) physics-based arti-
facts, which result from the physical processes
involved in the acquisition of CT data; (b) pa-
tient-based artifacts, which are caused by such
factors as patient movement or the presence of
metallic materials in or on the patient; (c) scan-
ner-based artifacts, which result from imperfec-
tions in scanner function; and (d) helical and
multisection artifacts, which are produced by the
image reconstruction process.
In this article, the different types of artifact Figure 2. Attenuation profiles obtained with and
without beam hardening for an x-ray beam passing
within each of these categories will be described through a uniform cylindrical phantom.
with regard to (a) the mechanisms by which they
are generated, (b) the methods employed by CT
equipment manufacturers to suppress them, and Cupping Artifacts.—X rays passing through the
(c) techniques of artifact avoidance available to middle portion of a uniform cylindrical phantom
the operator. are hardened more than those passing though the
edges because they are passing though more ma-
Physics-based Artifacts terial. As the beam becomes harder, the rate at
which it is attenuated decreases, so the beam is
Beam Hardening more intense when it reaches the detectors than
An x-ray beam is composed of individual photons would be expected if it had not been hardened.
with a range of energies. As the beam passes Therefore, the resultant attenuation profile differs
through an object, it becomes “harder,” that is to from the ideal profile that would be obtained
say its mean energy increases, because the lower- without beam hardening (Fig 2). A profile of the
energy photons are absorbed more rapidly than CT numbers across the phantom displays a char-
the higher-energy photons (Fig 1). Two types of acteristic cupped shape (Fig 3a).
artifact can result from this effect: so-called cup-
ping artifacts and the appearance of dark bands or Streaks and Dark Bands.—In very heteroge-
streaks between dense objects in the image. neous cross sections, dark bands or streaks can
appear between two dense objects in an image.
They occur because the portion of the beam that
passes through one of the objects at certain tube
 RG f Volume 24 ● Number 6 Barrett and Keat 1681

RadioGraphics

Figure 3. CT number profiles obtained across the

center of a uniform water phantom without calibration
correction (a) and with calibration correction (b).

Figure 4. CT image shows streaking artifacts

due to the beam hardening effects of contrast
medium.

has been used. In the chest scan shown in Figure

4, the contrast medium has caused artifacts that
might be mistaken for disease in nearby anatomy.

Built-in Features for Minimizing Beam

Hardening.—Manufacturers minimize beam
Figure 5. CT images of a skull phantom recon-
hardening by using filtration, calibration correc-
structed without bone correction (a) and with bone
correction (b). tion, and beam hardening correction software.
Filtration: A flat piece of attenuating, usually
metallic material is used to “pre-harden” the
beam by filtering out the lower-energy compo-
nents before it passes through the patient. An ad-
ditional “bowtie” filter further hardens the edges
of the beam, which will pass through the thinner
parts of the patient.
Calibration correction: Manufacturers cali-
brate their scanners using phantoms in a range of
sizes. This allows the detectors to be calibrated
with compensation tailored for the beam harden-
ing effects of different parts of the patient. Figure
3b demonstrates the elimination of cupping arti-
facts by this means in a phantom. Since patient
anatomy never exactly matches a cylindrical cali-
bration phantom, in clinical practice there may be
Figure 6. CT images of the posterior fossa show the either a slight residual cupping artifact or a slight
dark banding that occurs between dense objects when “capping” artifact, with a higher central CT value
only calibration correction is applied (a) and the reduc- due to overcorrection.
tion in artifacts when iterative beam hardening correc- Beam hardening correction software: An itera-
tion is also applied (b). (Reprinted, with permission,
from reference 1.)
tive correction algorithm may be applied when
images of bony regions are being reconstructed.
This helps minimize blurring of the bone–soft
positions is hardened less than when it passes tissue interface in brain scans (Fig 5) and also
through both objects at other tube positions. This reduces the appearance of dark bands in nonho-
type of artifact can occur both in bony regions of mogeneous cross sections (Fig 6).
the body and in scans where a contrast medium
 1682 November-December 2004 RG f Volume 24 ● Number 6

RadioGraphics

Figures 7, 8. (7) Mechanism of partial volume artifacts, which occur when a dense object lying off-
center protrudes part of the way into the x-ray beam. (8) CT images of three 12-mm-diameter acrylic
rods supported in air parallel to and approximately 15 cm from the scanner axis. (a) Image obtained with
the rods partially intruded into the section width shows partial volume artifacts. (b) Image obtained with
the rods fully intruded into the section width shows no partial volume artifacts.

Avoidance of Beam Hardening by the Op-

erator.—It is sometimes possible to avoid scan-
ning bony regions, either by means of patient po-
sitioning or by tilting the gantry. It is important to
select the appropriate scan field of view to ensure
that the scanner uses the correct calibration and
beam hardening correction data and, on some
systems, the appropriate bowtie filter.

Partial Volume
There are a number of ways in which the partial Figure 9. CT image of a shoulder phantom shows
volume effect can lead to image artifacts. These streaking artifacts caused by photon starvation.
artifacts are a separate problem from partial vol-
ume averaging, which yields a CT number repre-
sentative of the average attenuation of the materi-
als within a voxel.
One type of partial volume artifact occurs
when a dense object lying off-center protrudes
partway into the width of the x-ray beam. In Fig-
ure 7, the divergence of the x-ray beam along the
z axis has been greatly exaggerated to demon-
strate how such an off-axis object can be within
Figure 10. Tube current modulation as a function of
the beam, and therefore “seen” by the detectors,
tube angle. mA ⫽ milliamperage.
when the tube is pointing from left to right but
outside the beam, and therefore not seen by the
detectors, when the tube is pointing from right to direction, for example in the posterior fossa. To
left. The inconsistencies between the views cause limit image noise, thicker sections can be gener-
shading artifacts to appear in the image (Fig 8a). ated by adding together several thin sections.
Partial volume artifacts can best be avoided by
using a thin acquisition section width. This is Photon Starvation
necessary when imaging any part of the body A potential source of serious streaking artifacts is
where the anatomy is changing rapidly in the z photon starvation, which can occur in highly at-
tenuating areas such as the shoulders (Fig 9).
When the x-ray beam is traveling horizontally, the
attenuation is greatest and insufficient photons
 RG f Volume 24 ● Number 6 Barrett and Keat 1683

RadioGraphics

Figure 11. Projection data

as they might appear for a
horizontal x-ray beam passing
through the shoulders. Dia-
grams show the data in their
original form (a) and with
adaptive filtration (b).

Figure 12. The three components of multidimensional adaptive filtration: averaging of adjacent in-
plane detector readings (a), averaging of each detector reading at successive projection angles (b), and
broadening of the z filter for high-attenuation angles (c). Black line in c ⫽ reconstruction position.

reach the detectors. The result is that very noisy Adaptive Filtration.—Some manufacturers use
projections are produced at these tube angula- a type of adaptive filtration to reduce the streak-
tions. The reconstruction process has the effect of ing in photon-starved images. This software cor-
greatly magnifying the noise, resulting in horizon- rection smooths the attenuation profile in areas
tal streaks in the image. of high attenuation before the image is recon-
If the tube current is increased for the duration structed (Fig 11).
of the scan, the problem of photon starvation will A multidimensional adaptive filtration tech-
be overcome, but the patient will receive an un- nique is currently being developed for use on
necessary dose when the beam is passing through multisection scanners. For the small proportion
less attenuating parts. Therefore, manufacturers of projection data that exceed a selected attenu-
have developed techniques for minimizing photon ation threshold, smoothing is carried out be-
starvation. tween adjacent in-plane detectors (Fig 12a) and
between successive projection angles (Fig 12b),
Automatic Tube Current Modulation.—On while the z filter used in helical reconstruction
some scanner models, the tube current is auto- is broadened for high-attenuation projection
matically varied during the course of each rota- angles to allow more photons to contribute to
tion, a process known as milliamperage modulation.
This allows sufficient photons to pass through the
widest parts of the patient without unnecessary
dose to the narrower parts (Fig 10).
 1684 November-December 2004 RG f Volume 24 ● Number 6

RadioGraphics

Figure 13. Original axial CT images (top) and coronal reformatted images (bottom) in their original
form (a) and after reconstruction with multidimensional adaptive filtration (b). (Courtesy of Willi Kalender,
PhD, University of Erlangen, Germany.)

the reconstruction (Fig 12c). Figure 13 demon-

strates the degree to which streaking is reduced
while maintaining spatial resolution with the
technique (2).

Undersampling
The number of projections used to reconstruct a
CT image is one of the determining factors in im-
age quality. Too large an interval between projec-
tions (undersampling) can result in misregistra-
tion by the computer of information relating to
sharp edges and small objects. This leads to an
effect known as view aliasing, where fine stripes
appear to be radiating from the edge of, but at a
distance from, a dense structure (Fig 14). Stripes
appearing close to the structure are more likely to
be caused by undersampling within a projection,
Figure 14. CT image of a Teflon block in a
which is known as ray aliasing. water phantom shows aliasing (arrow) due to un-
Aliasing may not have too serious an effect on dersampling of the edge of the block.
the diagnostic quality of an image, since the
evenly spaced lines do not normally mimic any
anatomic structures. However, where resolution need to be avoided as far as possible. View alias-
of fine detail is important, undersampling artifacts ing can be minimized by acquiring the largest
possible number of projections per rotation. On
some scanners, this can be achieved only by using
a slower rotation speed, while on others the num-
ber of projections is independent of rotation
speed. Ray aliasing can be reduced by using spe-
cialized high-resolution techniques, such as quar-
ter-detector shift or flying focal spot, which man-
ufacturers employ to increase the number of
samples within a projection.
 RG f Volume 24 ● Number 6 Barrett and Keat 1685

RadioGraphics

Figure 15. CT images of a patient with metal spine implants, reconstructed without any correc-
tion (a) and with metal artifact reduction (b). (Courtesy of Siemens, Forchheim, Germany.)

scanning commences. For nonremovable items,

such as dental fillings, prosthetic devices, and sur-
gical clips, it is sometimes possible to use gantry
angulation to exclude the metal inserts from scans
of nearby anatomy. When it is impossible to scan
the required anatomy without including metal
objects, increasing technique, especially kilovolt-
age, may help penetrate some objects, and using
thin sections will reduce the contribution due to
partial volume artifact.

Software Corrections for Metal Artifacts.—

Streaking caused by overranging can be greatly
reduced by means of special software corrections.
Manufacturers use a variety of interpolation tech-
niques to substitute the overrange values in at-
tenuation profiles. The effectiveness of one such
technique is illustrated in Figure 15. The useful-
Figure 16. CT image of the head shows mo- ness of metal artifact reduction software is some-
tion artifacts. times limited because, although streaking distant
from the metal implants is removed, there still
remains a loss of detail around the metal-tissue
Patient-based Artifacts interface, which is often the main area of diagnos-
tic interest. Beam hardening correction software
Metallic Materials should also be used when scanning metal objects
The presence of metal objects in the scan field can to minimize the additional artifacts due to beam
lead to severe streaking artifacts. They occur be- hardening.
cause the density of the metal is beyond the normal
range that can be handled by the computer, result- Patient Motion
ing in incomplete attenuation profiles. Additional Patient motion can cause misregistration artifacts,
artifacts due to beam hardening, partial volume, which usually appear as shading or streaking in
and aliasing are likely to compound the problem the reconstructed image (Fig 16). Steps can be
when scanning very dense objects. taken to prevent voluntary motion, but some in-
voluntary motion may be unavoidable during
Avoidance of Metal Artifacts by the Opera-
tor.—Patients are normally asked to take off re-
movable metal objects such as jewelry before
 1686 November-December 2004 RG f Volume 24 ● Number 6

RadioGraphics

Figure 17. CT images of the body created with conventional reconstruction (a) and with motion arti-
fact correction (b). (Reprinted, with permission, from reference 3.)

body scanning. However, there are special fea-

tures on some scanners designed to minimize the
resulting artifacts.

Avoidance of Motion Artifacts by the Opera-

tor.—The use of positioning aids is sufficient to
prevent voluntary movement in most patients.
However, in some cases (eg, pediatric patients), it
may be necessary to immobilize the patient by
means of sedation. Using as short a scan time as
possible helps minimize artifacts when scanning
regions prone to movement. Respiratory motion
can be minimized if patients are able to hold their
breath for the duration of the scan.
The sensitivity of the image to motion artifacts
depends on the orientation of the motion. There- Figure 18. CT image of the body obtained with the
fore, it is preferable if the start and end position of patient’s arms down but outside the scanning field
shows streaking artifacts.
the tube is aligned with the primary direction of
motion, for example, vertically above or below a
patient undergoing a chest scan. Specifying body which helps reduce the severity of motion arti-
scan mode, as opposed to head scan mode, may facts. The use of partial scan mode can also re-
automatically incorporate some motion artifact duce motion artifacts, but this may be at the ex-
reduction in the reconstruction. pense of poorer resolution.
Software correction: Most scanners, when
Built-in Features for Minimizing Motion used in body scan mode, automatically apply re-
Artifacts.—Manufacturers minimize motion duced weighting to the beginning and end views
artifacts by using overscan and underscan modes, to suppress their contribution to the final image.
software correction, and cardiac gating. However, this may lead to more noise in the verti-
Overscan and underscan modes: The maxi- cal direction of the resultant image, depending on
mum discrepancy in detector readings occurs be- the shape of the patient. Additional, specialized
tween views obtained toward the beginning and motion correction is available on some scanners.
end of a 360° scan. Some scanner models use The effectiveness of one such technique in cor-
overscan mode for axial body scans, whereby an recting artifacts due to motion of a fluid interface
extra 10% or so is added to the standard 360° is demonstrated in Figure 17.
rotation. The repeated projections are averaged, Cardiac gating: The rapid motion of the heart
can lead to severe artifacts in images of the heart
and to artifacts that can mimic disease in associ-
 RG f Volume 24 ● Number 6 Barrett and Keat 1687

RadioGraphics arms are outside the scan field, they are not
present in the image, but their presence in some
views during scanning has led to such severe arti-
facts throughout the image as to significantly de-
grade its usefulness. Similar effects can be caused
by dense objects such as an intravenous tube con-
taining contrast medium lying outside the scan
field. Blocking of the reference channels at the
sides of the detector array may also interfere with
data normalization and cause streaking artifacts.
To avoid artifacts due to incomplete projec-
tions, it is essential to position the patient so that
no parts lie outside the scan field. Scanners de-
signed specifically for radiation therapy planning
have wider bores and larger scan fields of view
than standard scanners and permit greater versa-
Figure 19. Formation of a ring artifact tility in patient positioning. They also allow scan-
when a detector is out of calibration. ning of exceptionally large patients who would
not fit within the field of view of standard scan-
ners.
Some manufacturers monitor the reference
data channels for inconsistencies and avoid using
reference data that appear suspicious. As an alter-
native, the CT system may be designed with ref-
erence detectors on the tube side or with ray
paths within the gantry to eliminate possible in-
terference with reference data.

Scanner-based Artifacts

Ring Artifacts
If one of the detectors is out of calibration on a
third-generation (rotating x-ray tube and detector
assembly) scanner, the detector will give a consis-
Figure 20. CT image of a water-filled
tently erroneous reading at each angular position,
phantom shows ring artifacts.
resulting in a circular artifact (Fig 19). A scanner
with solid-state detectors, where all the detectors
ated structures, for example, dissected aorta. To are separate entities, is in principle more suscep-
overcome these difficulties, techniques have been tible to ring artifacts than a scanner with gas de-
developed to produce images by using data from tectors, in which the detector array consists of a
just a fraction of the cardiac cycle, when there is single xenon-filled chamber subdivided by elec-
least cardiac motion. This is achieved by combin- trodes.
ing electrocardiographic gating techniques with Rings visible in a uniform phantom (Fig 20) or
specialized methods of image reconstruction (4). in air might not be visible on a clinical image if a
wide window is used. Even if they are visible, they
Incomplete Projections would rarely be confused with disease. However,
If any portion of the patient lies outside the scan they can impair the diagnostic quality of an im-
field of view, the computer will have incomplete age, and this is particularly likely when central
information relating to this portion and streaking detectors are affected, creating a dark smudge at
or shading artifacts are likely to be generated. the center of the image.
This is illustrated in Figure 18, which shows a
patient scanned with the arms down instead of
being raised out of the way of the scan. As the
 1688 November-December 2004 RG f Volume 24 ● Number 6

RadioGraphics

Figure 22. Series of CT images from a helical scan of the abdomen shows helical artifacts (arrows).
(Reprinted, with permission, from reference 5.)

Figure 21. Consecutive axial CT images from a heli-

cal scan of a cone-shaped phantom lying along the
scanner axis. (Reprinted, with permission, from refer-
ence 5.)

Avoidance and Software Corrections Figure 23. CT image of a 12-mm-diameter

acrylic sphere supported in air, obtained with
The presence of circular artifacts in an image is an
0.6-mm section acquisition and beam pitch of
indication that the detector gain needs recalibra- 1.75, shows windmill artifact.
tion or may need repair services. Selecting the
correct scan field of view may reduce the artifact
by using calibration data that fit more closely to there are additional artifacts that can occur in he-
the patient anatomy. lical scanning due to the helical interpolation and
All modern scanners use solid-state detectors, reconstruction process. The artifacts occur when
but their potential for ring artifacts is reduced by anatomic structures change rapidly in the z direc-
software that characterizes and corrects detector tion (eg, at the top of the skull) and are worse for
variations. higher pitches.
If a helical scan is performed of a cone-shaped
Helical and Multisection CT Artifacts phantom lying along the z axis of the scanner, the
resultant axial images should appear circular.
Helical Artifacts in the Ax- In fact, their shape is distorted because of the
ial Plane: Single-Section Scanning weighting function used in the helical interpola-
In general, the same artifacts are seen in helical tion algorithm (Fig 21). For some projection
scanning as in sequential scanning. However, angles, the image is influenced more by contribu-
tions from wider parts of the cone in front of the
scan plane; for other projection angles, contribu-
tions from narrower parts of the cone behind the
 RG f Volume 24 ● Number 6 Barrett and Keat 1689

RadioGraphics

Figure 24. (a) Fan beam acquisition as used in single-section scanners. (b) Cone beam acquisition as used in mul-
tisection scanners.

tion scanners. The typical windmill-like appear-

ance of such artifacts (Fig 23) is due to the fact
that several rows of detectors intersect the plane
of reconstruction during the course of each rota-
tion. As helical pitch increases, the number of
detector rows intersecting the image plane per
rotation increases and the number of “vanes” in
the windmill artifact increases.
Z-filter helical interpolators are commonly
used on multisection scanners to replace the two-
point interpolators usually used on single-section
scanners. One of the benefits of z-filter interpola-
tors is that they reduce the severity of windmill
artifacts, especially when the image reconstruc-
tion width is wider than the detector acquisition
Figure 25. Volume of data collected by an outer row width. Artifacts may also be slightly reduced by
of detectors (left) and an inner row (right) on a 16-sec- using noninteger pitch values relative to detector
tion scanner. acquisition width, such as pitches of 3.5 or 4.5 on
a four-section scanner (6). This is because z-axis
scan plane predominate. Thus, the orientation of sampling density is optimized for noninteger
the artifact changes as a function of the tube posi- pitches.
tion at the center of the image plane. In clinical
images, such as the series of liver images shown in Cone Beam Effect
Figure 22, helical artifacts can easily be misinter- As the number of sections acquired per rotation
preted as disease. increases, a wider collimation is required and the
To keep helical artifacts to a minimum, steps x-ray beam becomes cone-shaped rather than fan-
must be taken to reduce the effects of variation shaped (Fig 24). Figure 25 shows an exaggerated
along the z axis. This means using, where pos- view of the x-ray beam and detectors along the z
sible, a low pitch, a 180° rather than 360° helical axis. As the tube and detectors rotate around the
interpolator if there is a choice, and thin acquisi- patient (in a plane perpendicular to the diagram),
tion sections rather than thick. Sometimes, it is the data collected by each detector correspond to
still preferable to use axial rather than helical im- a volume contained between two cones, instead of
aging to avoid helical artifacts (eg, in brain scan- the ideal flat plane. This leads to artifacts similar
ning). to those caused by partial volume around off-axis
objects. The artifacts are more pronounced for
Helical Artifacts in the outer detector rows than for the inner ones
Multisection Scanning (Fig 26), where the data collected correspond
The helical interpolation process leads to a more more closely to a plane.
complicated form of axial image distortion on
multisection scanners than is seen on single-sec-
 1690 November-December 2004 RG f Volume 24 ● Number 6

RadioGraphics

Figure 26. CT images from data collected by an outer detector row (a)
and an inner detector row (b) show cone beam artifacts around a Teflon rod,
which was positioned 70 mm from the isocenter at an angle of 60° to the
scanner axis.

Figure 27. CT images of a phantom, obtained by using four-section acquisition and stan-
dard reconstruction (a), 16-section acquisition and standard reconstruction (b), and 16-
section acquisition and cone beam reconstruction (c). (Courtesy of Siemens.)

Figure 28. (a) Sagittal reformatted image from axial CT data obtained with 5-mm collimation and a
5-mm reconstruction interval. (b) Sagittal reformatted image from single-section helical CT data ob-
tained with 5-mm collimation and a 2.5-mm reconstruction interval.

Cone beam effects get worse for increasing Multiplanar and Three-
numbers of detector rows. Thus, 16-section scan- dimensional Reformation
ners should potentially be more badly affected by Major improvements in multiplanar and three-di-
artifacts than four-section scanners. However, mensional reformation have come about since the
manufacturers have addressed the problem by introduction of helical scanning and, to an even
employing various forms of cone beam recon- greater extent, with multisection scanning. The
struction instead of the standard reconstruction faster speed with which the required volume can be
techniques used on four-section scanners. The scanned means that the effects of patient motion are
effectiveness of one such technique is demon- much reduced, and the use of narrower acquisition
strated in the phantom study shown in Figure 27. sections and overlapping reconstructed sections
leads to sharper edge definition on reformatted im-
ages.
 RG f Volume 24 ● Number 6 Barrett and Keat 1691

RadioGraphics

Figure 29. Original axial CT image (a) and coronal reformatted image (b) of the sinuses, ob-
tained with a 16-section scanner by using thin acquisition sections. (Courtesy of Siemens [7].)

from the axis of rotation because the noise inho-

mogeneity is worse off-axis.

Summary
Artifacts originate from a range of sources and
can degrade the quality of a CT image to varying
degrees. Design features incorporated into mod-
ern scanners minimize some types of artifact, and
some can be partially corrected by the scanner
software. However, there are many instances
where careful patient positioning and the opti-
mum selection of scan parameters are the most
important factors in avoiding image artifacts.
Figure 30. Maximum intensity projection image ob-
tained with helical CT shows zebra artifacts.
References
1. Hsieh J. Image artifacts: appearances, causes and
Stair Step Artifacts.—Stair step artifacts appear corrections. In: Computed tomography: principles,
design, artifacts and recent advances. Bellingham,
around the edges of structures in multiplanar and Wash: SPIE Press, 2003; 167–240.
three-dimensional reformatted images when wide 2. Kachelriess M, Watzke O, Kalender WA. General-
collimations and nonoverlapping reconstruction ized multi-dimensional adaptive filtering for con-
intervals are used. They are less severe with helical ventional and spiral single-slice, multi-slice, and
scanning, which permits reconstruction of overlap- cone-beam CT. Med Phys 2001; 28:475– 490.
3. Seeram E. Image quality. In: Computed tomogra-
ping sections without the extra dose to the patient phy: physical principles, clinical applications and
that would occur if overlapping axial scans were quality control. 2nd ed. Philadelphia, Pa: Saunders,
obtained (Fig 28). Stair step artifacts are virtually 2001; 174 –199.
eliminated in multiplanar and three-dimensional 4. Barrett JF, Keat N, Platten D, Lewis MA, Edyvean
reformatted images from thin-section data obtained S. Cardiac CT scanning. MHRA Report 03076.
London, England: Medicines and Healthcare Prod-
with today’s multisection scanners (Fig 29). ucts Regulatory Agency, 2003.
5. Wilting JE, Timmer J. Artifacts in spiral-CT images
Zebra Artifacts.—Faint stripes may be apparent and their relation to pitch and subject morphology.
in multiplanar and three-dimensional reformatted Eur Radiol 1999; 9:316 –322.
images from helical data because the helical inter- 6. Taguchi K, Aradate H. Algorithm for image recon-
struction in multi-slice helical CT. Med Phys 1998;
polation process gives rise to a degree of noise 25:550 –561.
inhomogeneity along the z axis. This “zebra” ef- 7. Somatom sessions special issue II. Forchheim, Ger-
fect (Fig 30) becomes more pronounced away many: Siemens, 2001.

This article meets the criteria for 1.0 credit hour in category 1 of the AMA Physician’s Recognition Award. To obtain
credit, see accompanying test at http://www.rsna.org/education/rg_cme.html.
View publication stats