MRI artifacts

Sudil Paudyal
(51)
BSc.MIT Final year
ARTIFACTS IN MRI

5/5/2013 MRI artifacts-sudil 2

• Image Artifact (brit. Artefact) is something observed in a
scientific investigation that is not naturally present but occurs
as a result of the investigative procedure. (oxford dictionary)
• A structure not normally present, but visible as a result of
malfunction in the hardware or software of the device, or a
consequence of environmental influences as heat or humidity
or can be caused by the human body itself.
Introduction

• All MRI images have artifacts to some degree.
• Some are irreversible and may only be reduced while others
can be totally eliminated.
• Knowledge of artifacts a must for technologists to maintain
optimum image quality.
• Causes should be understood and compensated for if possible
to avoid being misjudged as pathology.
• Classified as to their basic principles,
 Physiologic (motion, flow)
 Hardware (electromagnetic spikes, ringing)
 Inherent physics (chemical shift, susceptibility, metal)

Artifacts caused by pt. motion
1. Ghosting and smearing:
 Most common artifacts produced by motion of pt.(voluntary /
involuntary)
– From oesophageal contraction and vascular pulsation during head
and neck imaging,
– From respiration and cardiac activity during thoracic and abdominal
imaging,
– From bowel peristalsis during abdominal and pelvic imaging.
 Occurs when there is magnitude and/or phase deviation from
optimal k-space encoding.

 Magnitude error - when a signal producing voxel changes
position during application of RF pulse.
 Phase errors - Patient motion whenever a transverse component
of magnetization exists and motion is perp. to the applied
magnetic field Bo.

 Appearance of ghosting on final clinical image depends on
where in k-space such phase errors occur:
– If along the x-axis of k-space : frequency encoding direction
– If along y-axis : phase encoding direction
– If in middle of k-space : smearing appearance
– If phase errors are periodic (as in pulsatile motion), periodic ghosting.
 Relatively more common in the phase encoding direction.

Remedy
Total elimination impossible unless imaging a cadaver!!!!
• Various methods
a) Variants of rectilinear k-space filling techniques
 Fast spin echo
 Multisection imaging
 Single shot single section imaging techniques
 Parallel imaging
b) Radial k-space filling techniques
c) Cardiac gating and triggering
d) Respiratory gating
e) Navigator echo

a) Variants of rectilinear k-space filling techniques
 CSE : dominant because of relative insensitivity to field
inhomogeneity.
 FSE/TSE : reduction in TA

 Multisection imaging
 Several interleaved sections are imaged simultaneously
 Helps further decrease acquisition time
 Single shot single section imaging
 Sequences such as the HASTE are more resistant to motion
 Allow more rapid acquisition by filling only half of k-space.

 Parallel imaging :
 involves filling of selected lines in k-space at a predetermined interval.
 Requires use of multichannel, multicoil technology, with each coil
possessing a distinct, known sensitivity profile over the FOV and with
at least two coils placed in the phase encoding direction.
 The no. of phase encoding steps can be reduced by a factor of 1/X,
where X is the parallel imaging factor (2 or 3).
 Thus image acquisition proceed X times faster by filling only one of
every X lines(2nd or 3rd ) of k-space in phase encoding direction and
by known coil sensitivity profiles to synthesize other lines.
Eg: GRAPPA (siemens), SENSE (philips), SMASH

b) Radial k-space filling techniques
 Standard line by line filling of k-space replaced by radial k-space filling
with the use of a multishot radial acquisition technique (eg.
syngoBLADE, Siemens healthcare; PROPELLER, GE healthcare)
 MR imaging datasets acquired in multiple overlapping radial sections,
each of which includes data sampled from the center to the periphery
of k- space.
 A series of low resolution images is first reconstructed from each
radial section and combined to produce a high resolution image.
 As the phase encoding direction varies with each radial section,
ghosting is not propagated in PE direction but is dispersed throughout
radial sections and thus has a lesser effect on final image.
 Also provide correction for in-plane rotation and bulk translational
movement.

c) Cardiac gating and triggering
Cardiac motion periodic, both static and cine images
Cine imaging performed by using a balanced SSFP that
produces tissue contrast based on ratio of T1 to T2, lending
blood a high signal intensity (so called bright blood
imaging).
Cine imaging performed with retrospective cardiac gating
i.e.
a series of MR images of a single anatomic section are acquired
during cardiac cycle, monitored with ECG.
Data within k-space are linked with specific time point in cardiac
cycle.
The acquired datasets are then sorted acc to time stamp to
produce sequential images allowing a dynamic evaluation of
myocardial function.

 For static images to evaluate cardiac structure and not cardiac
func., prospective triggering is used and image acquisition is
usually timed to occur during the end diastole.
 In triggering a certain no. of k-space lines is acquired during
pt. breath holding for a given heartbeat.

d) Respiratory gating
 Infrequently used in clinical imaging as compared to cardiac gating.
 Breath holding method most often used, but may not be feasible for
infants, children, and pt.s with respiratory or cognitive impairment.
 Sequences with short acquisition time may be used, or if breath
holding possible for limited time, exam may be divided into several
brief acquisitions.
 If none is possible, respiratory gating.
 in which respiration related motion is monitored, using bellows or
breathing belt, image acquisition is timed to take place at end
expiration, when there is little or no motion.
 Phase reordering with either COPE or ROPE.

e) Navigator echo
 Most commonly used in abdominal imaging at the interface of lung
and diaphragm
 Application of small, one dimensional spatial encoding gradient in a
plane perpendicular to diaphragm.
 Echo measured at this location allows correction of imaging dataset to
ensure that, only the imaging data acquired, when diaphragm is at its
peak (end expiration), are used in image reconstruction.

Motion artifact contd..
2. Pulsatile flow-related artifacts
 Periodic pulsation of vascular structures leads to ghosting in phase
encoding direction at constant intervals.
 Distance between ghosts depends on diff. between heart rate and TR;
if synched, no ghosting.
 Macroscopic motion eg. Blood in large vessels or CSF in spinal canal
also contribute to ghosting.

Remedy
a) ECG gating :
– time consuming, seldom used except for cardiac imaging
b) Gradient moment nulling:
– Application of additional gradient pulses to correct for phase shifts
among a population of moving protons at the time of echo collection.
– Corrects for constant-velocity motion and helps reduce the signal loss
and ghosting associated with such movement.

c) Saturation pulse:
– Additional RF pulses applied before the sequence in a plane
perpendicular or parallel to imaging plane
– As applied at the beginning of sequence, their use may reduce the no.
of imaging secitons that can be obtained per TR in multisection
acquisitions.
– Also when SAR is already high, use of saturation pulse may result in
excessive heat deposition.

Artifacts caused by field distortions
1. Eddy currents
 source of spatial and temporal distortions in Bo.
 most frequently encountered in clinical DWI because of the high
amplitude and long duration of diffusion- sensitizing gradients.
 When diffusion gradient applied, change in magnetic field creates
electrical currents in neighboring conductive surfaces. Such currents
create smaller magnetic fields (i.e. eddy currents) that modify Bo.
 modern gradient coils equipped with active shielding to avoid these
effects of electrical conduction, eddy currents are not as obtrusive in
routine clinical imaging as they were in the past.

Remedy
 Pre-compensation- A “distorted” gradient waveform is used which
corrects to normal with the eddy current effects.
 Shielded gradients – Active shielding coils between gradient coils and main
gradients.

2. Gradient field nonlinearity
 Occurs in all MR imaging systems and primarily related to gradient
falloff due to the finite size of the gradient coils.
 Predictable.
 Easily corrected by system software, with corrections applied before
the final images are constructed.
 Operator unaware of occurrence of mapping errors due to gradient
field nonlinearity.
 Although post processing techniques can correct distortions resulting
from gradient field nonlinearity, cannot compensate for losses in
spatial resolution that are related to gradient field nonlinearity.

3. RF field inhomogeneities
 Do not cause geometric distortions but contribute to signal
nonuniformity.
 May arise from problems in RF coil construction or standing wave
(dielectric) effects.
 Because the Larmor frequency of protons increases with increasing
field strength, a high-frequency pulse is needed to excite signal
producing protons at MR imaging with high field strengths such as 3 T.
 When the RF wavelength is shorter than the dimensions of the
anatomic structures examined at clinical MR imaging, standing waves
may result.
 Leads to inhomogeneous fat suppression.

Remedy
 Shimming- allows precise control of overall homogeneity of RF field.
 Use STIR for Fat sat than spectral tech. like CHESS.
 Dielectric – use phased array coils, software compensation

Aliasing artifacts
• Field of view:
– dimensions of the anatomic region to be included in imaging.
– mathematic product of the acquisition matrix and pixel dimensions.
– Chosen considering size of structure to be evaluated and trade-offs
between SNR and spatial resolution.
– With a constant imaging matrix, a smaller FOV results in higher spatial
resolution but lower SNR.
• Choosing an FOV that is smaller than the area imaged leads to
wraparound or aliasing artifacts.

• Echoes are sampled along the FE direction at the sampling bandwidth
rate, with higher rates resulting in a greater range of frequency sampling.
• Nyquist frequency - highest frequency that may be clearly sampled, with
higher frequencies corresponding to tissues outside the specified FOV.
• High frequencies outside the FOV falsely detected as lower frequencies
which leads to a wraparound artifact on the reconstructed images.

Remedy
• Increase the FOV (decreases resolution).
• Oversampling the data in the frequency direction (standard)
and increasing phase steps in the phase-encoded direction –
phase compensation (time or SNR penalty).
• Swapping phase and frequency direction so phase is in the
narrower direction.
• Use surface coil so no signal detected outside of FOV.
• Saturation pulses may also be applied to structures in the
nonimaged portion of the FOV to reduce signal and, thus,
signal overlap.

Magnetic susceptibility artifacts
 Inaccurate spatial encoding from susceptibility gradients within tissue
leads to distortion of anatomic structures.
 Artifact resulting from the presence of metallic objects, not only distorts
nearby structures but also may result in signal dropout, depending on the
sequence used.
 The effects are field strength dependent, and the potential for metallic
object–related artifacts is greater at 3 T than at lower magnetic field
strengths.
 However, the increase in magnetic field strength from 1.5 to 3 T results in
a significant improvement in SNR, allowing the use of a higher bandwidth
sampling rate and parallel imaging to reduce susceptibility artifacts.
 Worst with long TE and gradient echo sequences.

Remedy
 Imaging at low magnetic field strength, using smaller voxels, decreasing
echo time, and increasing receiver bandwidth.
 Gradient-echo and echo-planar sequences should be avoided, because
they increase susceptibility artifacts.
 The use of spin-echo and particularly fast spin-echo sequences should be
considered.

Chemical shift artifacts
 Protons in fat and water precess at different frequencies in an applied
magnetic field.
 The separation between their resonance frequencies increases with
increasing field strength.
 Eg. At 1.5 T , freqency diff. is 220 Hz, but at 1 T the diff. is 147 Hz and at
lower field strengths (0.5 T or less ) usually insignificant.
 The chemical shift differences lead to spatial misregistration of the MR
signal; i.e. differences in the Larmor frequency are mistaken for
differences in spatial position along the frequency encoding axis.
 The resultant chemical shift artifacts on images manifest most
prominently at fat-water interfaces.

 Amount of chemical shift is expressed in arbitrary units known as parts per
million (ppm) of Bo.
 It’s value is always independent of Bo and equals 3.5 ppm.
 Occurs in the frequency encoding direction only.

Remedy
 Imaging at low magnetic field strength,
 by increasing receiver bandwidth, or by decreasing voxel size.
 T1-weighted- The artifacts tend to be more prominent on T2-
weighted than on T1-weighted images.
 Fat suppression methods often eliminate.
5/5/2013 36MRI artifacts-sudil

Chemical misregistration artifact
 Also produced as a result of the precessional frequency diff.
between fat and water.
 But this is caused because fat and water are in phase at
certain times and out of phase at others, due to difference in
their precessional frequency.
 When they are in phase their signals add constructively and
when out of phase the signals cancel each other out.
 This cancellation effect is known as chemical misregistration
artifact.

 Causes a ring of dark signal around certain organs where fat and
water interfaces occur within the same voxel. Eg. Kidneys
 Mainly occurs in PE direction as it is produced due to phase
difference between fat and water.
 Most degrading to the image in gradient echo pulse sequences
where gradient reversal is very ineffective.

Remedy
 Use SE sequence
 In GRE, select a TE that corresponds to the periodicity of fat and
water
 ie. Select a TE that generates an echo when fat and water are in
phase.

Partial volume artifacts
 Partial volume occurs if slice thickness > thickness of tissue of
interest
 Occurs when multiple tissue types are encompassed within a single
voxel
 If small structure is entirely contained within the slice thickness
along with other tissue of differing signal intensities,
 the resulting signal displayed on the image is a combination of
these two intensities.
 This reduces contrast of the small structure.
 If the slice is the same thickness or thinner than the small structure,
only that structures signal intensity is displayed on the image.
 Volume averaging is most likely to occur in the slice-selection
direction of the image.

Remedy
 Decreasing voxel size, particularly reducing section thickness.
 Three-dimensional Fourier transform imaging is particularly useful,
because it provides thin sections with no intervening gaps and is
conductive to reformatting in alternate imaging planes.
 Multiplanar imaging option helps to clarify.

Signal truncation artifacts
 Occur in regions of boundary between high and low signal intensity
 Caused by approximation errors in Fourier transform analysis.
 As the signal is sampled over a limited period of time, some data
are necessarily omitted (truncated) in k-space, causing the signal
intensity of a given pixel on the final image to vary from its ideal
signal intensity.
 Commonly seen at the interface of the low-signal intensity spinal
cord with high-signal-intensity CSF on T2WI of the spine, mimic
spinal canal dilatation (ie. Hydromyelia,syrinx).
 Appears as a periodic “ringing” at high contrast interfaces.

128x256
256x256
Remedy
• Use of higher resolution imaging matrix and filtration methods.(under
sampling avoided)
• Gradient reorientation will displace the artifacts to another portion of the
image.

Slice-overlap (cross-slice) artifacts
• Loss of signal seen in an image from a multi-angle, multi-slice
acquisition.
• If the slices obtained at different disk spaces are not parallel, then
the slices may overlap when two levels are done at the same time,
e.g., L4-5 and L5-S1.
• The level acquired second will include spins that have already been
saturated.
• This causes a band of signal loss crossing horizontally in image,
usually worst posteriorly.
• Therefore, overlap of sections within areas of diagnostic interest
should be carefully avoided.

Remedy
 Avoid steep change in angle between slice groups.
 Use separate acquisitions.
 Use small flip angle.

Cross-excitation artifacts
• The imperfect shape of RF slice profiles leads to the unintended
excitation of adjacent tissue.
• This excitation results in the saturation of such tissue
• Manifest as decreased signal intensity and decreased contrast that
can hinder lesion detection.

Remedy
• Introduce an intersection gap that is 10% to 50% of the
prescribed section thickness.
• Interleaved image acquisition, in which odd-numbered
sections are initially acquired,followed by acquisition of even-
numbered sections.
• optimized RF pulses that have a more rectangular slice profile
can be implemented.

Cross-talk Artifact
 Perfect RF pulse is a sinc function (FT = ‘top hat’)
 Real RF pulse is a truncated sinc (FT = ‘top hat with rounded
edges’)
 Result of imperfect slice excitation of adjacent slices causing
reduction in signal over entire image.

• Inter-slice cross talk could cause increased T1 weighting and
reduced SNR.
• May be reduced by using gap, interleaving slices and
optimized (but longer) RF pulses.

Magic Angle Effects
 Produced by the particular physical properties of fibrillary tissues and their
interaction with the static magnetic field.
 Seen most frequently in tendons and ligaments that are oriented at a 55o
angle to the main magnetic field.
 Due to dipolar interactions that reduce their T2 relaxation time.
 Normal dipolar interactions between the H+’s in water molecule aligned in
tendons shortens T2, causing loss of signal.
 T2 relaxation time is lengthened and maximal when these fibrillary
structures are at a 55° angle to B0.
 Maximal for short TE.

Remedy
 Lengthening TE
 T1 weighted imaging

Zipper Artifacts
• Most are related to hardware or software problems
• May occur in either frequency or phase direction.
• Zipper artifacts from RF entering room during image
acquisition are oriented perpendicular to the frequency
direction and easily controlled.

Spike artifact
• Caused by one ‘bad’ data point in k-space.
• Fig. shows one data point in k-space, which is out of the
ordinary.
• The resulting image show diagonal lines throughout the
image.
Remedy - Repeat the scan.

Zebra stripes
• Observed along the periphery of gradient-echo images (abrupt
transition in magnetization at the air-tissue interface)
• Increased by aliasing that results from the use of a relatively small
field of view.
• May also occur when pt. touches the coil or a result of phase wrap.
Remedy
• expanding the FOV, using SE pulse sequences.
• using oversampling techniques to reduce aliasing.

RF Overflow Artifacts (Clipping)
 Causes a nonuniform, washed-out appearance to an image.
 Occurs when the signal received from the amplifier exceeds
the dynamic range the analog-to-digital converter causing
clipping.
 Autoprescanning usually adjusts the receiver gain to prevent
this from occurring.

Moire Fringes
 An interference pattern most commonly seen when doing gradient
echo images.
 One cause is aliasing of one side of the body to the other results in
superimposition of signals of different phases that add and cancel.
 Can also be caused by receiver picking up a stimulated echo.
 Similar to the effect of looking though two window screens.

Central Point Artifact
• A focal dot of increased or decreased signal in the center of an
image.
• Caused by a constant offset of the DC voltage in the
amplifiers.
Remedy
 Requires recalibration by engineer
 Maintain a constant temperature in equipment
room for amplifiers.

Quadrature ghost artifact
• Another amplifier artifact caused by unbalanced gain in the two
channels of a quadrature coil.
• Combining two signals of different intensity causes some
frequencies to become less than zero causing 180 degree “ghost.”

Entry slice (Inflow) phenomenon
• Unsaturated spins in blood or CSF entering the initial slices results in
greater signal then reduces on subsequent slices.
• Characterized by bright signal in a blood vessel (artery or vein) at entry
site.
• May be confused with thrombus.
• The use of gradient echo flow techniques can be used to differentiate
entry slice artifacts from occlusions.
• Can cause spatial saturation to reduce.
• Mechanism for TOF angiography.

Shading artifacts
 Produces a loss of signal intensity in one part of image.
 Main cause is uneven excitation of nuclei within the pt. due to RF
pulses applied at flip angles other than 90* and 180*
 Also caused by abnormal loading on coil or by coupling of coil at
one point. This may occur with large pt. who touches one side of
the body coil and couples it at that point.
 Appear as foci of relatively reduced signal intensity involving a
portion of the image.
 Abnormalities contained in the shaded portion of the MR image
may be obscured.
 Also be caused by inhomogeneities in the main magnetic field that
can be improved by shimming.

Remedy
• Always ensure that coil is loaded correctly i.e correct size of coil for
anatomy under examination, and pt. is not touching the coil at any point.
• Use of foam pads or water bags bet.n coil and patient
• Ensure that appropriate pre-scan parameters have been obtained before
the scan, as these determine the correct excitation frequency and
amplitude of applied RF pulses.

Conclusion…
• Artifacts in MR images are an inevitable truth.
• MRI arifacts occur because one or more of the assumptions underlying the
imaging principles have been violated.
• Some can be reduced while others can be totally eliminated.
• Artifact correction methods usually involve one or more of the following:
 Hardware calibration
 Scanning parameter optimization
 Special pulse sequence design
 Signal and image postprocessing
• By understanding the mechanism of their production and their effects on the final
image, technologists should considerably try to minimize these artifacts with the
use of reduction techniques.
• Ideally, we want all image artifacts to be below the level of user's perception.
• Artifact correction is an active area of research today, and will continue to be in
the future as advances in MRI technology reveal new image information and new
kinds of artifacts.

Lastly…
Observer artifact:
Self explanatory
Otherwise known as “Upside down Error”
Apply for time off from duty.

Find more at….
1. Morelli JN, Runge VM, Ai F, et.al. An Image-based Approach
to Understanding the Physics of MR Artifacts. RadioGraphics
2011; 31:849–866
2. MRI in practice, 2nd edn. By Catherine Westbrook and
Carolyn Kaut
3. MRI artifacts, PPT presentation by Ray Ballinger
4. MRI physics course; Artifacts and suppression technique by
Jerry Allisson
5. www.mr-tip.com
6. www.mritutor.org

On examining a knee with prosthesis in situ, what artifact is
expected?
Difference between chemical shift and chemical
misregistration
Truncation artifact and reduction
List different motion compensation options.
Why is motion artifact seen only in PE direction?
What are the sources of zipper artifact?
Why does aliasing occur?
Difference between cross talk and cross excitation.

THANK YOU!!!

MRI artifacts

More Related Content

What's hot (20)

Viewers also liked (20)

Similar to MRI artifacts (20)

More from Sudil Paudyal (9)

Recently uploaded (20)

MRI artifacts

Editor's Notes