Artefacts In Clinical MRI

Kris Armoogum MSc Department of Medical Physics, Ninewells Hospital Dundee DD2 1QW Kris.Armoogum@tuht.scot.nhs.uk Stephen.Gandy@tuht.scot.nhs.uk 14th Scottish MRI Seminar, Wednesday 19th November 2003

Introduction
Artefacts are parts of reconstructed images that are not present in the true anatomy. Artefacts are dependent on a variety of factors from patient movement to magnetic field inhomogeneities. Artefacts can lead to misdiagnosis if they are not recognised and/or removed. Ideally, we want all image artefacts to be below the level of user's perception.

Main classifications -

1. 2. 3. 4. 5. 6.

Movement Artefacts Geometrical Artefacts Resolution/Sequence Artefacts Bo Artefacts RF Artefacts Noise

1. Movement Artefacts
Motion Artefacts patient Flow artefact inflow/washout effect, diastole, systole, arterial flow Reducing flow artefacts Gradient Moment Nulling Respiratory compensation, triggering, ROPE, navigator echoes

Motion Artefact - Patient

Patient movement as outer areas of k-space acquired May mimic truncation artefact Difference truncation artefact diminishes with distance from the high contrast boundary If related to pulsation of vessels, this can be reduced by applying an anterior sat band

T2W SE Thoracic Spine

Patient Movement Artefact

Smearing of image Particularly in phase direction Solutions Immobilise the patient more effectively Reduce the scan time reduce NSA, breathold, shorter TR, less k-space lines Reduce the scan time (reduced k-space acquisition) e.g. HASTE, TSE with large turbo factor

Flow Artefact Inflow Effect

T1W GE sequence RF pulse saturates the blood momentarily in the slice (yellow) If blood is stationary, long T1 of blood means that no signal available for successive RF pulses to excite hypointense signal If velocity of inflowing blood > z/TR then full inflow occurs and the next RF pulse sees unsaturated spins in the slice Bright Blood signal Other tissues within the slice are saturated, and therefore suppressed
NO FLOW 90o pulse Next TR Dark signal Image

z FLOW 90o pulse Next TR Bright signal

Flow Artefact Washout Effect

Spin-echo sequence - 90o pulse excites spins in the slice In the absence of flow, bright signal is seen because the spins experience both the 90o and 180o pulses In the presence of flow, blood flows out of the slice and does not experience the 180o pulse no rephasing, Black Blood signal D.G. Nishimura "Time-of-Flight MR Angiography." Magn. Reson. Med. 14:194-201 (1990)
TE/2 180o pulse TE Signal 90o pulse Image

NO FLOW

FLOW

90o pulse

180o pulse

No signal

Flow Artefact I - Body

+1 Grad T T -1

S Phase Shift D

Diastole (filling phase), systole (emptying phase). Aortic ghosts in PE direction [because FE step (msec) takes much less time than a PE (sec) step] Physiological modulation

Why only the PE Direction?

Why is motion artefact only seen in the PE direction? A FE step takes much less time (of the order of msec) than a PE step (of the order of seconds) Most motion that occurs during clinical MRI is much slower than the rapid sampling process along the FE axis However, each PE line is separated by the time interval TR which is long enough for blood to flow/move/dephase between successive phase encodings

Phase Shift Effects

S D

m o
MR Signal
FT

o-m o

o+m

Systolic-diastolic switching of the flow velocity (at frequency m) modulates the MR signal (with frequency ) Two frequency sidebands (upper and lower frequency sideband components) appear as ghosts either side of the primary image

Flow Artefact II - Brain

Grad

Velocity profile laminar flow (Re < 2100) Velocity zero at wall, fastest at centre of lumen Continuous spread of velocities
+1

T -1

Continuous Phase Shift

B
A = flow artefact from eye movement B = flow artefact from sagittal sinus

Flow From Sagittal Sinus

Distinguishable from Truncation artefact artefact propagates across the anatomy. Truncation artefact diminishes with distance from the high contrast boundary.

Further Flow Artefacts

A B C

Flow from middle (A) cerebral arteries, (B) eye movement & sagittal sinus, and (C) salivary glands (swallowing)

Flow Popliteal Artery

Popliteal Artery

Popliteal artery flow generates artefacts across the femur. More prominent when fat suppression removes the marrow signal. May disturb interpretation of bone bruising or subchondral cysts.

Phase Shift Effects

m o
FT

o-max

o+max

MR Signal Complex modulation frequency, made up of many component waves FT results in a spread of upper and lower sidebands - artefact Worse for GE images, due to bright blood inflow effect SE dark blood inflow effect less severe phase artefacts

Reducing Phase Flow Effects - I

Low bandwidth
+G

High bandwidth

+2G

Grad

2T 2T
-G

Grad

T T

-2G

Large Phase Shift Phase Shift

Small

Use sequence with a high bandwidth shorter TE Amount of dephasing generally goes up with TE2 Shorter TE minimises velocity induced phase effects

Gradient Moment Nulling

Gradient Moment Nulling (GMN) technique used for flow compensation +1 -2

Phase shift

+1

Stationary tissue (0o) unaffected Constant velocity blood (1o motion) rephased by +1-2+1 gradients Constant acceleration blood (2o) need 1+3-3+1 gradients to rephase Jerk motion (3o) need +1-4+6-4+1 gradients to rephase 2o flow comp 3o flow comp

+3 +1
Stationary tissue Constant velocity blood Constant acceleration blood

+6 +1 -4 -4 +1

-1 -3

Flow Compensation
Thoracic spine

Vertebral foramen

Two cords appear to be present in first T2W TSE image. The extra cord is flow artefact from pulsatile CSF flow. First order (+1-2+1 gradient) flow compensation (Gradient Moment Nulling) results in the RHS image.

Swallowing Motion
Cervical spine

S
CIV CV

A T

Any patient motion during a scan can cause PE artefacts (A-P above). Left image - artefact generated by patient swallowing during data acquisition - increased signal intensity in the spinal cord. Eliminated by applying presaturation RF pulses to the anatomy that was generating the artefact. Sat band visible on RHS image.

Respiratory Artefact
Periodic respiratory motion ghosting above and below the body
Remove by breathold imaging (<18sec scan time). Increase the NSA (anatomy SNR improved relative to ghosts). NSA 4 to 6 ~ respiratory compensation.

Respiratory Gating/Triggering
Bellows pressure = electrical trigger

Reduces respiratory artefact

Bellows placed over abdomen. Sequence TR gated via patient breathing rate. Equivalent to TR ~ 4000ms, (breathing rate 15/min) so only able to generate PDW and T2W scans (long TR reduces T1 effect). Signal acquired when chest wall is in same position - minimises ghost images.

Inhalation

Exhalation

Respiratory Compensation
Reduces respiratory artefact

ROPE Respiratory Ordered Phase Encoding (also uses bellows).

Typical TR for a T1W sequence = 500 msec. Typical breathing rate = 15 breaths per min, i.e. 4000 msec period. Can therefore fit 8 TRs (8 PE steps) per breathing cycle. Outer k-lines (image boundary detail) acquired at peak inhalation. Central k-lines (signal, contrast) are acquired at peak exhalation.

K-space
+64 -64

Inhalation Exhalation

Respiratory Gating or ROPE ?

Gating: + simple technique Gating: effective TR very long (cannot do T1W) ROPE: + shorter scan times ROPE: residual ghosts if patient breathes deeply

Navigator Echoes
Two slice selective directions and FE in the third direction (of motion). Small column of tissue excited across the diaphragm. Spin echo sequence acquires series 1D images of the diaphragm boundary over time. Stack images side-by-side - intensity difference between diaphragm and lung indicates respiratory motion. Navigator echo is interleaved within main scan sequence. Data for main image can then be adjusted for respiratory motion by using data acquired during specific range of diaphragm motion.

2. Geometrical Artefacts
Phase wrap Partial volume Cross talk Magic Angle artefact

Phase Wrap - Aliasing

Regions outside FOV still produce a signal if in proximity to receiver coil. Anatomy outside FOV is mapped inside FOV. Corrected by - larger FOV or apply presat pulses to undesired tissue. No Phase Wrap double the FOV; but because PE steps is doubled need to half number of averages to keep scan time constant. Aliasing in FE direction can occur, but eliminated by filters (no time waste).

+200o = -160o

180o

-180o -160o

+180o +200o

EQUIVALENT

Partial Volume Effect

Partial volume occurs if slice thickness > thickness of tissue of interest If small structure is entirely contained within the slice thickness along with other tissue of differing signal intensities then 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. Typically would use 3mm slices for cranial nerves and 5-10mm slices for liver.

VII (Facial) and VIII (Acoustic) cranial nerves

Cross Talk
Perfect RF pulse is a sinc function (FT = top hat) Real RF pulse is a truncated sinc (FT = top hat with rounded edges) Inter-slice cross talk could cause increased T1 weighting and reduced SNR.

How Does Cross Talk Occur?

TE = 20 TR = 600 SLICE 1
90o 180o

PE2

SLICE 2

90o 180o

PE2

SLICE N

90o 180o

PE2

Typical TR for T1W scan = 600ms, typical TE = 20ms. Theoretically possible to acquire 30 slices within the TR. Cross talk region between slices 1 and 2 experiences RF excitation from slice 1, then slice 2. Effective TR is 20ms giving loss of signal due to lack of T1 recovery. Solution - interleave slices. A 3D sequence avoids the problem altogether contiguous slices.
10-20% interslice gap

Magic Angle Artefact (54.7o)

Collagen fibril orientation w.r.t. B0 field. T2 lengthening at Magic Angle. Result is that the T2W image becomes hyperintense at the magic angle. Magic Angle is solution to: 3cos2-1=0 (from dipolar Hamiltonian mathematical theory) Magic angle imaging of the median nerve (brain) which has a high collagen content.

Median Nerve in brain

At Magic Angle

3. Resolution/Sequence Artefacts
Truncation artefact Chemical Shift artefact (Types I and II)

Truncation Artefact - Brain

128 x 256 256 x 256

Also known as Gibbs (ringing) artefact. Usually occurs in the PE direction at high contrast borders. Due to undersampling of high spatial frequencies (sharp edged borders) Remedied by taking more samples (e.g. 256 PE steps). Truncation artefact causes ring-down effect because F.T. of truncated sinc function has ripples at the edges.

F.T. t

Truncation Artefact or Syrinx?

Problematic down centre of spinal cord could be misinterpreted as a syrinx

CV

Syrinx (fluid filled cavity in spinal cord)

Chemical Shift Artefact Type I

T1W Lumbar spine

T1W image of lumbar spine. Low BW sequence used. Frequency shift of a few pixels is visible at the base of each vertebra (black line). Vertebra-disc boundary detail is lost at the top of each vertebra. Observation of small disc herniations in L spine difficult.

L III

L IV

Chemical Shift (I) - Explained

Displacement of yolk (fat) on LHS image

Egg (low BW)

Egg (high BW)

Protons from different molecules (eg: fat & water) precess at different frequencies. Protons in H2O precess slightly faster than those in fat, (diff. is 3.5 ppm) Chemical shift = 3.5ppm = 224Hz at 1.5T [ 0 = .B0 :: (42.6MHz/T)(1.5T) :: 64MHz :: 3.5ppm x 64MHz = 224Hz ] LHS = 12.5kHz (low BW), 256 resolution. Chemical shift is 4.6 pixels [ 224 / (12.5kHz/256) ] Chemical shift also occurs between silicone & fat/water (Breast MRI) Modify CS by using fat suppression, increase the bandwidth, swap freq and phase directions, or lower the Bo field (impractical)!

Chemical Shift Type II Artefact

Out of Phase
Liver Thoracic aorta

In Phase

Worked example

Applies to Gradient Echo techniques, (not in SE because of 180 refocusing pulse). Fat and water proton resonant frequencies differ by 3.5ppm. For an imaging field strength of 1.5T, == 64 MHz (from 0 = .B0 ). Difference between fat and water proton resonant frequencies is therefore about 224 Hz, ( diff ). The phase of the fat and water spin vectors will thus coincide at 1/diff, which is 4.6 ms. If a TE of 4.6 ms is used, then the fat and water components of the signal will be in phase. If a TE of 6.9 ms (4.6 + 2.3) is used then the fat and water components of the signal will be out of phase.

Chemical Shift Type II Artefact

F
3.5 PPM

W o
4.6ms WF F 2.3ms W

Phase cancellation artefact gradient echo sequences Water precesses slightly faster than fat (phase difference between them) Phase differences accumulate between water and fat signal Vary the TE, f+W (in phase), f-w (out of phase black boundary artefact) At 1.5T, f-w occurs in 4.6ms multiples, starting at about 2.3ms (then 6.9, 11.5, 16.1 ms) - artefact At 1.5T, f+w occurs at 4.6ms (then 9.2, 13.8, 18.4 ms) no artefact Dixon technique ip+op images = water image, ip-op = fat image The artefact can occur in both encoding directions Not a problem in SE images since 180o pulse refocuses chemical shift

4. Bo Artefacts
Susceptibility artefacts Metallic artefacts Bo Inhomogeneity

Susceptibility Artefacts
Occur when two materials with different magnetic susceptibility () lie together, (tissue-air & tissue-fat). Local Bo changes cause spin dephasing at the boundary causing signal loss. Haemosiderin (end stage of haemorrhage) deposits (high ) local susceptibility changes in tissue. Susceptibility artefacts can be useful - bony trabeculae (low ). Use a FSE and keep TE short to minimise susceptibility artefacts.

Metallic Artefacts
Similar to susceptibility artefacts. Metals have much higher susceptibility than tissue. Large Bo inhomogeneities around object causing signal loss and distortion. Implants absorb RF energy, so local field varies. RF problems affect SE sequences as well as GE.

Metallic Artefact

Small metal flake in lumbar spinal canal

Bo Inhomogeneity and FatSat

Unsuccessful Fat suppression in T2W breast images. Result of poor Bo field homogeneity. Artefacts arise because of inability to distinguish fat and water frequencies locally. Usually more prominent in images with a large FOV or off-axis. Solution improve the magnet shimming. Modern magnets auto shimming for very reliable fatsat.

5. RF Artefacts
Ghosting RF interference Stimulated Echoes RF Coil artefacts Steady State artefacts

Ghosting
Arises from any structure that moves during acquisition of data eg: chest wall, pulsatile movement of vessels, swallowing etc.) Ghosts displaced along PE axis due to inherent time delay between phase encoding and readout. Number and intensity depends upon period of modulation and the TR.
Inhalation Exhalation B A P1 P2

Chest wall

Moving anatomy is mismapped into the FOV.

Quadrature Ghost
Occurs due to differences in the gain of real and imaginary receiver channels Phase errors between the two quadrature RF receive channels can also cause this Ghost is displaced diagonally across the centre in both PE and FE directions Solution - ? phase alternating

RF Interference
Zipper artefact appears as bright and dark zipper lines along PE. External RF picked up by coils (e.g RF breakthrough waveguide filters). Pulse oximeters (monitors the percentage of haemoglobin saturated with oxygen) use RF can be picked up by MR coils. RF from within the MR system may be coherent bright spot on image. Mains RF modulated by 50Hz regularly spaced faint zipper artefacts across image.

RF breakthrough

Zipper artefact

Herring-Bone Artefact
Occurs due to the presence of a spike of noise (or an arc from a static discharge) in the raw data. FT (series of spikes) which is convolved with the image data. Probably due to breakdown of RF system (poor RF decoupling). Best solution rescan the image.

Halo Artefact
Results from signal clipping caused by overflow on the ADCs. Occurs if receiver gain is incorrectly set. Signal becomes too large for the ADC range and information in the centre of k-space is lost. Unusual - unless receiver gain is manually set.

Stimulated Echoes (STE)

1st 90o pulse Dephasing 2nd 90o pulse Hahn Echo

z x y

Lag

y
Lead

3rd 90o pulse

Stim Echo

1st pulse forms transverse magnetisation 2nd pulse remaining transverse components form Hahn echo 3rd pulse converts longitudinal magnetisation to transverse magnetisation, and components re-phase to form stimulated echo

STE Coherence Pathways

Phase 90o 90o 90o
(Longitudinal)

H=Hahn Echo S=Stimulated Echo

H
SHOULDER

STE

STE has different spatial encoding and contrast Avoid STE by using spoiler gradients to destroy residual transverse magnetisation, or use rewinder gradients to prevent the STE occurring in the sampling window Can also widen the bandwidth, or alter the TE to avoid STE

RF Coil Artefacts
One of the arrays of a phased array coil is out of phase with the other coils. Bands of signal addition and cancellation. Solution call engineer!

Sagittal Pelvis

Surface Coil Flare

Axial abdomen

The result of signal saturation at edge of surface coil. Optimal signal is further in from edge. Solution Surface Coil Intensity Correction (SCIC) algorithm that reduces the high intensity fat signal nearest the coil for improved visualisation. SCIC is very useful for correcting sagittal and axial spine images.

Steady State Imaging - Artefacts

Coronal abdomen

Moire Fringes

Common on True FISP, balanced FFE, FIESTA (fully balanced gradients). Related to variation of steady state condition due to Bo inhomogeneities. Aliasing of one side of the body to the other results in superimposition of signals of different phases that alternatively add and cancel. Equivalent to introducing a systematic error to the flip angle. Require a short TE and good shimming otherwise bands ~ 1/B0 Solution phase alternation of RF pulse

Noise

Random Noise
Noise can be considered an artefact since it is unwanted. Grainy, snowy, no recognisable pattern. Solution improve the SNR Increase slice thickness, increase TR, reduce TE, decrease bandwidth, decrease pixel resolution, increase the FOV, increase phase steps, increase the number of averages Remember trade-offs (scan time [2D] = TR x NY x NEX).

And finally

Observer Artefact
Self explanatory Otherwise known as Upside-down Error Solution apply for time off !

References
MRI from Picture to Proton: Donald W. McRobbie, Elizabeth A.Moore, Martin J.Graves and Martin R.Prince Cambs Uni Press All you need to know about MRI Physics: Moriel NessAiver For further information Kris.Armoogum@tuht.scot.nhs.uk Stephen.Gandy@tuht.scot.nhs.uk