Artifacts are relatively common in computed tomographic (CT) imaging, although their cause may not always be obvious. This article describes common CT artifacts (such as streaks, rings, and black and white bands) and addresses causes of artifacts, their effects on the quality of the radiographic image, and procedures that can be used to reduce the presence of such artifacts.
Dr. Al-Shakhrah
is an Associate Professor of Medical Physics, University of
Jordan, Amman, Jordan; and
Dr.
Al-Obaidi
is
a Professor of Medical Physics, Mu'tah University, Al-Karak,
Jordan.
A computed tomographic (CT) image is a display of the anatomy of
a thin slice of the body developed from multiple X-ray absorption
measurements made around the body's periphery. Unlike conventional
tomography, in which the image of a thin slice is created by
blurring out the information from unwanted regions, the CT image is
constructed mathematically, using data arising only from the
section of interest. Generating such an image is confined to cross
sections of the anatomy that are oriented essentially perpendicular
to the axial dimension of the body.
1
An artifact is any distortion or error in an image that is
unrelated to the subject being studied. Artifacts are relatively
common in CT imaging and may be considered as a source, or type, of
noise. Their cause may not always be obvious. However, there are a
number of different effects that may be responsible for artifacts
in CT.
Because artifacts in CT arise as a result of the interaction
between the subject and the machine, it is useful to classify the
artifacts by the nature of the error made in the scanning process.
In CT, artifacts may be produced by:
* Errors in X-ray attenuation measurements;
* Alterations in the energy spectrum of the X-ray beam (beam
hardening) as it passes through the patient;
* The presence of high-density foreign materials in the
body;
* Partial-volume averaging effect;
* Motion of the patient;
* Quantum mottle (noise);
* Malfunction of the detector arising from errors in detector
calibrations and balance, geometric effects, or a machine
peculiarity; or
* Inadequate temperature, humidity, or the presence of small
dust particles within the computer that causes an inadequacy in the
reconstruction algorithm.
This article will present common artifacts such as streaks,
rings, and black and white bands that appear in CT studies. This
review addresses causes of artifacts, their effects on the quality
of the radiographic image, and procedures that can be used to
reduce the presence of such artifacts. The reduction of artifacts
enhances CT interpretive accuracy and helps to establish a correct
diagnosis.
Method
During a 1-year period, 7197 CT studies were performed in a
large (600-bed) teaching hospital. Within this total number of
studies, 432 repeat studies were performed due to image artifacts.
These repeat studies were collected and classified according to the
cause of the artifacts viewed. The study included both inpatients
and outpatients, but not all CT studies of uncooperative patients
from intensive and coronary care units were repeated. The CT
systems employed in this study included a Siemens Somotom Plus 4
spiral CT (Siemens Medical Solutions, Erlangen, Germany) and a
Philips Tomoscan AV spiral CT (Philips Medical Systems, Best,
Netherlands).
Results
Repeat CT studies were performed due to artifacts in 432
patients, and repeat CT studies were obtained in 6% of all CT
studies performed during the review period. Major artifacts were
found in the form of streaks, rings, and black and white bands.
Figures 1 through 8 provide examples of artifacts on CT scans in
different regions and organs in which artifacts can clearly be
seen. Figure 9 presents the causes of artifacts as well as the
incidence of each artifact found in the images studied. Regardless
of cause, most CT artifacts found in the study were manifest as
streaks in the CT image.
Discussion
Joseph
2
has reported two reasons for streak artifacts. First, each
individual measurement involves the evaluation of a single ray or
straight-line path through the slice. A second, more subtle cause
for streak artifacts arises when there is an abrupt discrepancy or
inconsistency between views, as might be seen with patient
motion.
The following discussion will address our results and their
causes, according to each artifact etiology.
Motion
Patient motion has a devastating effect on image quality. This
was the primary reason for the development of a body unit (spiral
CT) that could complete a scan during a patient's breath-hold.
The artifacts found in our study caused by motion were
manifested as black or white bands, dark spots, loss of resolution,
or distortion of anatomy
3
(Figure 1). These artifacts account for 15% of repeated studies.
Clinically, such artifacts are important not only because they
degrade image quality, but also because they can sometimes be
mistaken for pathologic changes, such as bronchiectasis.
4
Theoretically, motion artifacts can be reduced by fast scanning,
5
gating,
6
tube alignment,
7
corrective reconstruction,
8
or postprocessing of the scan.
9
Knowing the pattern, magnitude, and frequency of motion in
advance would allow the use of an algorithm to remove the motion
from the projection data and then reconstruct a motion-free or, at
least, a motion-reduced image.
10
A technical problem is image misregistration due to variation in
breath-holding from one scan to the next.
11
Misregistration leads to failure to image part of the lung, such
that a lesion might be partially or entirely excluded on the final
scan. This problem is eliminated by using spiral CT during
breath-holding.
12,13
X-ray geometry, geometric oversights
Precise control of fan-beam position is important for the
production of high-quality CT images; failure to maintain fan-beam
orientation can produce image artifacts.
14
Performance of CT systems relies mainly on geometric precision and
measurement quality. Inaccurate geometry, inaccurate alignment of
the X-ray tube with the detectors, or incorrect data can produce
artifacts and blurring that limit spatial resolution (Figure 2). In
our study, artifacts due to geometric causes accounted for 5% of
repeat studies. A precise geometric calibration procedure is
required, and some corrections must be applied to the raw
attenuation data in order to obtain accurate measurements. An X-ray
cone-beam CT system has been developed by Rizo and Martin.
15
The machine was designed to control small parts limited to a few
centimeters, with a high spatial resolution close to 30 microns.
They introduced the machine setup and described the calibration
computing resources involved in the system. They also discussed the
performance on experimental data.
15
Detector artifacts
Artifacts from errors in detector calibrations and balance are
common. A malfunction of any one detector would incorrectly
backproject along the data ring to produce the artifacts. If
detectors are not matched or intercalibrated accurately, the
backprojection for each data ring would be slightly different,
causing multiple rings (Figure 3). These artifacts caused by
equipment malfunction can be eliminated by repair or good
preventive maintenance.
Beam hardening
Beam-hardening artifacts result from the preferential absorption
of low-energy photons from the beam. The effect is more pronounced
in areas of large attenuation, such as bone. The artifact is seen
as a shadow beneath ribs, for example, or increased shadows in the
mediastinum or skull (Figure 4). This type of artifact accounted
for 21% of the repeat scans in our study. This effect occurs
throughout the image but usually is not perceived except where
there is a great deal of hardening, such as in the vicinity of
bone. This effect can be compensated for by the use of special
filters or a special correction algorithm.
1
Recent experiments using CT and transmission radiography show
that gadolinium (Gd) agents can increase image contrast by up to a
factor of 2 when compared with more commonly used iodinated agents
on an equi-molar basis. It has also been suggested that
beam-hardening artifacts may be reduced with the use of Gd. This
hypothesis was tested by Ruth and Joseph
16
on three different CT scanners using a circular water equivalent
phantom with a contrast-filled tube inserted. It was found that the
artifacts were 1.3 to 1.8 times more pronounced with the iodinated
contrast when compared with gadopentetate dimeglumine
(Gd-DTPA).
High-density foreign material
The presence of objects that have an exceptionally high or low
attenuation can create streaking artifacts by forcing the detectors
to operate in a nonlinear response region.
1
Figure 5 demonstrates the star pattern caused by high-density
foreign materials. In our study, this type of artifact caused 25%
of repeat studies, one of the most common reasons for image
degradation. A small metal fragment produces a star pattern, and
the star effect is accentuated by any motion. The only way to avoid
this problem with current mathematical reconstructions is to change
the angle of the slice to exclude the foreign body, but this
approach might also exclude pathology. A similar pattern can be
produced by gas, for example, in the gastric fundus, but the effect
is less marked.
4
Partial-volume averaging
When tissues of widely different absorption occupy the same
voxel, the beam attenuation is proportional to the average value of
the attenuation coefficient of the voxel. A volume average is
computed for such voxels, leading to the partial-volume error. The
scan of the skull-brain shown in Figure 6 demonstrates this effect.
Images generated with helical scanning are degraded by
partial-volume artifacts caused by an increased slice thickness
when compared with conventional CT scanning.
18
Partial-volume effects on measurements of CT numbers may be
minimized by the use of thin sections and by the selection of a
section that lies in the center of the object of interest for
measurement.
19
Quantum mottle (noise) artifacts
In a study of 20 examiniations, soft-tissue imaging degraded by
scattering artifacts was reported in 14 examinations.
20
Quantum mottle is dose-related image noise that has the appearance
of granular steaks arising from high-attenuation structures, such
as the shoulder region. A generalized adaptive median filter (GAMF)
was introduced by Hsieh
21
for more robust noise suppression and edge preservation in CT to
combat severe streaking artifacts resulting from excessive X-ray
quantum noise (Figure 7).
Temperature and humidity
Some computer components are more sensitive to extremes of
temperature and humidity than is conventional X-ray equipment. For
this reason, manufacturer's recommendations regarding air
conditioner installation should be closely followed. Whenever
possible, a backup air-conditioning system should be available. The
low-temperature limits below which solid-state devices cannot
operate appropriately will probably never be encountered in a
hospital installation. Failure from moderately increased
temperature, as well as humidity, will frequently take the form of
unexplained malfunctions of the computer,
19
including increasing numbers of artifacts and inappropriate
responses to instructions (Figure 8).
One aspect of the computer environment that must be considered
is dust particle size. The particle size of cigarette smoke is only
slightly larger than the size of the air gap between the playback
head and disk drive of the computer system. In addition, its
relatively small size makes it particularly difficult to filter
from the atmosphere. For this reason, smoking in the computer room
should be strictly forbidden.
Conclusion
In our study, CT artifacts were found to be produced by: the
presence of high- density foreign materials in the body (25%); an
error in X-ray attenuation measurements, as a result of alterations
in the energy spectrum of the X-ray beam (beam hardening) as it
passes through the patient (21%); partial-volumeaveraging effects
(16%); motion of the patient (15%); quantum mottle (noise) (7%);
malfunction of the detector arising from errors in detector
calibrations and balance (6%); geometric effects or a machine
peculiarity (5%); and by inadequate temperature, humidity, or the
presence of small dust particles within the computer causing an
inaccuracy in the reconstruction algorithm (5%). Regardless of the
causes of the artifacts found in this study, most CT artifacts
manifested as streaks. Films that were repeated due to artifacts
accounted for 6% of the total 7197 films taken during the 1-year
period. Patient cooperation, use of thin sections, repair and/or
good preventive maintenance, a clean computer environment, and
suitable temperature and humidity can reduce artifacts to a
minimum.
Acknowledgment
The authors thank Mr. Jameel Al-Sarayrah for his assistance in
collecting CT studies and in classifying imaging artifacts
according to etiology.