quality assurance of mri systems david hearshen slide 1 ... · slide 2 quality control of mri ......
TRANSCRIPT
Quality Assurance of MRI Systems David Hearshen
1
Slide 1
Technical Aspects ofTechnical Aspects ofQuality Control inQuality Control in
Magnetic Resonance Imaging Magnetic Resonance Imaging
___________________________________
___________________________________
___________________________________
___________________________________
___________________________________
___________________________________
___________________________________
Slide 2Quality Control of MRIQuality Control of MRI
SystemsSystems
David Hearshen, Ph.D.
Department of Radiology
Henry Ford Hospital, Detroit, MI
___________________________________
___________________________________
___________________________________
___________________________________
___________________________________
___________________________________
___________________________________
Slide 3
ObjectivesObjectives
To gain a general understanding of MRsystem variability
To be able to set up and manage an MRI QCand Compliance Testing program
To gain a general understanding of MRIArtifacts
___________________________________
___________________________________
___________________________________
___________________________________
___________________________________
___________________________________
___________________________________
Quality Assurance of MRI Systems David Hearshen
2
Slide 4Objectives of an MRI QCObjectives of an MRI QCProgramProgramTo implement a system practical enough for
technologists to utilize on a daily basis
To obtain meaningful results which revealsystem changes before patient care isaffected
To document problems and corrective actionin a manner which satisfies accreditationand regulatory requirements
___________________________________
___________________________________
___________________________________
___________________________________
___________________________________
___________________________________
___________________________________
Slide 5System VariabilitySystem VariabilityMagnetic resonance imaging systems are
subject to variability from:Drift in RF Electronics
RF Coils
RF Transceiver Chain
Magnetic Field Decay
Foreign ferro- or para- magnetic material
Introduction of material producing MRI signal
Gradient system failures
___________________________________
___________________________________
___________________________________
___________________________________
___________________________________
___________________________________
___________________________________
Slide 6
Effects of System VariabilityEffects of System VariabilityRadio Frequency effects
SNR LossOccurs when the RF center frequency is off resonance
Power is insufficient to produce 90 or 180 deg pulses
If the RF coil is detuned
If the RF coil homogeneity is not optimized
NoiseCan increase with failures in preamplifier or amplifier
Can be introduced with failures in the RF shielding(most often door).
___________________________________
___________________________________
___________________________________
___________________________________
___________________________________
___________________________________
___________________________________
Quality Assurance of MRI Systems David Hearshen
3
Slide 7Gradient effectsGradient effects
Geometric DistortionCan occur with when maximum amplitude is not
achieved or gradient is miscalibrated.
Can occur if eddy current compensation changes
Spatially Dependent SNR LossCan occur when local field homogeneity decreases
Can occur when gradient waveforms are notoptimized (eg eddy currents) resulting in incompleterephasing of echo
Phase ErrorsCause misregistration in Phase Encoding direction
Can arise from any of the 3 orthogonal gradients
___________________________________
___________________________________
___________________________________
___________________________________
___________________________________
___________________________________
___________________________________
Slide 8
Magnetic FieldMagnetic Field Inhomogeneity Inhomogeneity
SNR LossCan occur when local total field homogeneity
decreases ⇒ ↓T2*.
Can occur when center frequency is shifted beyondbandwidth of receiver
In Plane Geometric DistortionCan cause regions of low or high signal when
resonant frequency is shifted beyond bandwidth of apixel
___________________________________
___________________________________
___________________________________
___________________________________
___________________________________
___________________________________
___________________________________
Slide 9
Magnetic FieldMagnetic Field Inhomogeneity Inhomogeneity
Geometric Distortion in Slice Select DirectionCan cause misregistration of signal as a function of
slice location, or distortion of slice profile causingincreased “crosstalk”.
Chemical Shift Frequency OffsetsFrequency selective pulses (e.g. lipid or water
saturation, mtc) can have a spatial dependence totheir intended effect.
___________________________________
___________________________________
___________________________________
___________________________________
___________________________________
___________________________________
___________________________________
Quality Assurance of MRI Systems David Hearshen
4
Slide 10
Daily TestingDaily Testing
ACR MRI Accreditation Recommendation
Magnetic Field Stability (Center Frequency)
Signal to Noise Ratio
Artifact Inspection
___________________________________
___________________________________
___________________________________
___________________________________
___________________________________
___________________________________
___________________________________
Slide 11
Standard Pulse SequencesStandard Pulse Sequences
T1 Weighted T2 Weighted
TR 500 2000
TE min full <20msec 20/80 or 30/90
Matrix 128 or 160 128 or 160
Coil Head Head
#excitations 1 1
FOV 24or 25 24 or 25
___________________________________
___________________________________
___________________________________
___________________________________
___________________________________
___________________________________
___________________________________
Slide 12
Magnetic Field StabilityMagnetic Field Stability(Center Frequency)
Test Object
Procedure
Sequence Parameters
What to record
Factors influencing measurement
___________________________________
___________________________________
___________________________________
___________________________________
___________________________________
___________________________________
___________________________________
Quality Assurance of MRI Systems David Hearshen
5
Slide 13
Center Frequency (Center Frequency (contcont.).)
Typical magnetic field stability is on the order of 0.1ppm per hour or 2.4 ppm per day. At 1T thiscorresponds to approximately 100 Hz decay of thecenter frequency per day and 150 Hz per day at1.5 tesla. A graph of center frequency vs. timeshould indicate steady decay, however, typicalinhomogeneity over a 20 cm DSV is also withinthis range. Therefore, it is important to set upconditions over which reproducible measurementsof the center frequency can be obtained.
___________________________________
___________________________________
___________________________________
___________________________________
___________________________________
___________________________________
___________________________________
Slide 14Center Frequency (Center Frequency (contcont.).)
Additional magnetic fields due to variation inmagnetic susceptibility can distort the line shapeof the resultant water signal and produce an errorin determining the peak position. This isminimized by using a spherically symmetricuniform phantom.
Measurement of center frequency is furthercomplicated by the presence of additional staticmagnetic fields used to adjust the homogeneity forspecific applications (shims). The shim settingscan be modified in two ways.
___________________________________
___________________________________
___________________________________
___________________________________
___________________________________
___________________________________
___________________________________
Slide 15
Center Frequency (Center Frequency (contcont.).)
Routine preventative maintenance for MRI calls forperiodic re-shimming of the magnet using all theavailable set of shim fields. These shims involvecoils designed to vary magnetic field as a functionof spatial coordinates. Usually the set of coilsincludes shims which perturb the magnetic fieldspanning geometry characterized by second orderspherical harmonics. These shim settings areusually not changed by the user.
___________________________________
___________________________________
___________________________________
___________________________________
___________________________________
___________________________________
___________________________________
Quality Assurance of MRI Systems David Hearshen
6
Slide 16
Center Frequency (Center Frequency (contcont.).)
Many manufacturers provide user controlledshimming using the three linear gradient magneticfields, either manually or with a computercontrolled algorithm (autoshimming). Somemanufacturers utilize autoshimming before eachpulse sequence.
___________________________________
___________________________________
___________________________________
___________________________________
___________________________________
___________________________________
___________________________________
Slide 17
Center Frequency (Center Frequency (contcont))
Unaccounted shimming can add an additionalmagnetic field resulting in an increase in thecenter frequency from day to day. It is importanttherefore to use the same setting of the shimmagnetic field gradients for each measurement.
The graph of center frequency over a period of 15months shows an increase in center frequency atseveral time points. The overall decrease inmagnetic field is less than 160 Hz << less thanmanufacturer’s specification
___________________________________
___________________________________
___________________________________
___________________________________
___________________________________
___________________________________
___________________________________
Slide 18Center Frequency over 15 Month PeriodCenter Frequency over 15 Month Period
63860600
63860650
63860700
63860750
63860800
63860850
63860900
7/7/9
4
7/18/9
4
8/1/9
4
8/15/9
4
8/29/9
4
1/16/9
5
2/20/9
5
4/25/9
5
5/8/9
5
8/7/9
5
9/11/9
5
9/26/9
5
___________________________________
___________________________________
___________________________________
___________________________________
___________________________________
___________________________________
___________________________________
Quality Assurance of MRI Systems David Hearshen
7
Slide 19
Center Frequency (Center Frequency (contcont))
ProcedurePrescribe a standard T1 weighted single slice pulse
sequence
Make sure any auto shimming algorithm is turned off.
Set up parameters for the auto pre-scan and load the pre-set values of the shims.
Perform the clinical auto pre-scan procedure.
___________________________________
___________________________________
___________________________________
___________________________________
___________________________________
___________________________________
___________________________________
Slide 20
Center Frequency (Center Frequency (contcont))
ProcedureRecord the time of the measurement, resultant center
frequency, and shim file settings.Optionally record the transmitter and receiver gains. If
the same shim file setting was used subtract theprevious days frequency from the current value andcompare with the specified decay.
If the measurements were not acquired at the same timeon successive days, adjust for total decay time.
Action level: decay within manufacturer specification.
___________________________________
___________________________________
___________________________________
___________________________________
___________________________________
___________________________________
___________________________________
Slide 21
Signal to Noise RatioSignal to Noise Ratio
Test ObjectsMaterial
GeometryRF Homogeneity
Magnetic Susceptibility Effects
Positioning
Coil
___________________________________
___________________________________
___________________________________
___________________________________
___________________________________
___________________________________
___________________________________
Quality Assurance of MRI Systems David Hearshen
8
Slide 22
Signal to Noise RatioSignal to Noise Ratio
In measuring signal to noise ratio, a uniformtest object is necessary in order to minimizeloss of signal due to variation in magneticsusceptibility as well as spatial variation inthe signal intensity due to RFinhomogeneity.
Solutions which are not comparable to tissueconductivity may not load the coil properly.
___________________________________
___________________________________
___________________________________
___________________________________
___________________________________
___________________________________
___________________________________
Slide 23
Signal to Noise Ratio (Signal to Noise Ratio (contcont.).)
The choice of coil is determined by the largestvolume of procedures performed clinically, thebest homogeneity characteristics, and the ease ofboth acquiring and analyzing the data. Though insome instances the head coil is not always used forthe largest number of procedures, it usually hassuperior homogeneity compared to surface coilsused in spine or extremity imaging.
___________________________________
___________________________________
___________________________________
___________________________________
___________________________________
___________________________________
___________________________________
Slide 24
Signal to Noise Ratio (Signal to Noise Ratio (contcont.).)
Example Pulse sequenceTR - 500 to 800 milliseconds
TE - 10 to 20 milliseconds
Single 10mm slice at isocenter.
Single excitation, 256 x 160 matrix
FOV - 20 to 24cm.
___________________________________
___________________________________
___________________________________
___________________________________
___________________________________
___________________________________
___________________________________
Quality Assurance of MRI Systems David Hearshen
9
Slide 25
Signal to Noise Ratio (Signal to Noise Ratio (contcont.).)
These parameters are used to generate a T1weighted image. Specific TR, TE etc., canbe taken from a common clinical protocol,provided the T1 of the test object material iswithin the range of normal tissue. Oncechosen, these parameters should beremained fixed as they all affect theresultant signal to noise ratio.
___________________________________
___________________________________
___________________________________
___________________________________
___________________________________
___________________________________
___________________________________
Slide 26
Signal to Noise Ratio (Signal to Noise Ratio (contcont.).)
Use of the minimum TE as calculated by theimaging system is not recommended sincethis value will change with changes tosoftware and hardware.
Test object dimension and FOV should bechosen to fill greater than 85% of the usablefield view of the coil in order to adequatelytake into account RF homogeneity effects.
___________________________________
___________________________________
___________________________________
___________________________________
___________________________________
___________________________________
___________________________________
Slide 27
Signal to Noise Ratio (Signal to Noise Ratio (contcont.).)
ACR Accreditation “T1” Series
TR 500
TE 20 msec
Matrix 256
Coil Head
#excitations 1
FOV 24or 25
Time 2:16
Use of 256 matrix is notnecessary for SNR,however, if using theACR phantom, othertests may be analyzed.The extra minute is smallcompared to set up time(aligning the phantom).
___________________________________
___________________________________
___________________________________
___________________________________
___________________________________
___________________________________
___________________________________
Quality Assurance of MRI Systems David Hearshen
10
Slide 28
Signal to Noise Ratio (Signal to Noise Ratio (contcont.).)Noise Characteristics
Theoretically, noise in magnetic resonance israndom. In practice however, there can besystematic contributions to background noisearising mainly from unwanted phase shiftsacquired in the RF transceiver chain. Sincephase is used in the image reconstructionalgorithm to supply spatial information, thesephase shifts produce spatial variation in thisbackground signal (“ghosts”).
___________________________________
___________________________________
___________________________________
___________________________________
___________________________________
___________________________________
___________________________________
Slide 29
Signal to Noise RatioSignal to Noise Ratio
In measuring signal to noise ratio, a uniformtest object is necessary in order to minimizeloss of signal due to variation in magneticsusceptibility as well as spatial variation inthe signal intensity due to RFinhomogeneity.
Solutions which are not comparable to tissueconductivity may not load the coil properly.
___________________________________
___________________________________
___________________________________
___________________________________
___________________________________
___________________________________
___________________________________
Slide 30
AnalysisAnalysis
If S is the mean of the ROI from the original imageand σN is the standard deviation of an ROI fromthe noise image, then
Record the SNR along with means and standarddeviations.
Action level: SNR within manufacturer specification
SNRS
N
= 2σ
___________________________________
___________________________________
___________________________________
___________________________________
___________________________________
___________________________________
___________________________________
Quality Assurance of MRI Systems David Hearshen
11
Slide 31
Signal to Noise Ratio ( Signal to Noise Ratio (contcont.).)
Noise Estimation: Method 1While it may be attractive to measure the noise of
a region of interest from a non signal producingarea of the image field of view outside the testobject using a single image, this practice mayintroduce systematic noise into the calculationof the SNR. If a single acquisition is desired,estimate noise from multiple regions in theimage.
___________________________________
___________________________________
___________________________________
___________________________________
___________________________________
___________________________________
___________________________________
Slide 32
Procedure1. Scan test object once using T1 series.
2. Choose ROI that encompasses at least 75% ofthe test object. Measure and record mean andstandard deviation of the ROI
3. Calculate PSG from ACR accreditation.
4. If PSG acceptable, use average of 4 ROI σ’s toestimate noise.
Benefit is that PSG measurement can be tracked
Signal to Noise Ratio ( Signal to Noise Ratio (contcont.).)___________________________________
___________________________________
___________________________________
___________________________________
___________________________________
___________________________________
___________________________________
Slide 33 Signal to Noise Ratio ( Signal to Noise Ratio (contcont.).)
ROIs forsystematic noisemeasurement
Susceptibilityfrom airbubble
ACR T1 Series Slice #7
Signal ROI
___________________________________
___________________________________
___________________________________
___________________________________
___________________________________
___________________________________
___________________________________
Quality Assurance of MRI Systems David Hearshen
12
Slide 34
Signal to Noise Ratio (Signal to Noise Ratio (contcont.).)
SNR and PSG for three systemsFixed 11.5T Fixed 2 1.5T Mobile 1.0 TMean σ Mean σ Mean σ
L 12.08 5.82 11.22 5.87 20.96 11.77R 11.18 6.87 10.59 6.15 22.59 10.40T 10.42 5.13 9.93 5.07 20.28 10.18B 9.04 4.41 10.47 5.14 19.14 10.04Signal 1168.00 50.35 1407.19 40.32 1451.46 50.00
Noise Ave 10.68 5.56 10.55 5.56 20.74 10.60
PSG 0.001627 0.000501 0.001423
SNR 297.2202 358.0868 193.6942
___________________________________
___________________________________
___________________________________
___________________________________
___________________________________
___________________________________
___________________________________
Slide 35
Signal to Noise Ratio (Signal to Noise Ratio (contcont.).)
Noise Estimation: Method 2An accepted practice is to image the test object
twice, repeating the acquisition within a fewminutes of the first. The noise is estimatedfrom an ROI in an image formed by subtractingthe first acquisition from the second. Theassumption is that the signal (and noise) in thetwo images are uncorrelated.
___________________________________
___________________________________
___________________________________
___________________________________
___________________________________
___________________________________
___________________________________
Slide 36Noise ImageNoise Image
___________________________________
___________________________________
___________________________________
___________________________________
___________________________________
___________________________________
___________________________________
Quality Assurance of MRI Systems David Hearshen
13
Slide 37
Signal to Noise Ratio (Signal to Noise Ratio (contcont.).)
Procedure1. Scan test object twice using same parameters.
2. Subtract the second image from the first on apixel by pixel basis. (This may require theimages to be downloaded to a work station).
3. Choose ROI that encompasses at least 75% ofthe test object. Measure and record mean andstandard deviation of the ROI in both theoriginal images and the subtracted noise image.
___________________________________
___________________________________
___________________________________
___________________________________
___________________________________
___________________________________
___________________________________
Slide 38SNR Method 2SNR Method 2
P la ne S σ S NRAxia l 1185 11.5 145.7255S a g itta l 1184 11.62 144.0989Co ro na l 1174 11.5 144.3728
1.5T Body Coil
___________________________________
___________________________________
___________________________________
___________________________________
___________________________________
___________________________________
___________________________________
Slide 39
SNR over an 15 Month PeriodSNR over an 15 Month Period
A graph of SNR in the head coil over 15 months show day to dayvariation of SNR with a mean =104.3 and σ=2.6
90
92
94
96
98
100
102
104
106
108
110
7/7/9
4
7/18/9
4
8/1/9
4
8/15/9
4
8/29/9
4
1/16/9
5
2/20/9
5
4/25/9
5
5/8/9
5
8/7/9
5
9/11/9
5
9/26/9
5
___________________________________
___________________________________
___________________________________
___________________________________
___________________________________
___________________________________
___________________________________
Quality Assurance of MRI Systems David Hearshen
14
Slide 40Systematic BackgroundSystematic Background
Axial
Body Coil
FOV 36 cm
___________________________________
___________________________________
___________________________________
___________________________________
___________________________________
___________________________________
___________________________________
Slide 41Systematic BackgroundSystematic Background
Sagittal Coronal
___________________________________
___________________________________
___________________________________
___________________________________
___________________________________
___________________________________
___________________________________
Slide 42
Weekly TestingWeekly Testing
ACRProcessor sensitometry
OtherPhase Stability
Magnetic Field Homogeneity
Signal - to - Noise Ratio in orthogonal planes
Image Uniformity
___________________________________
___________________________________
___________________________________
___________________________________
___________________________________
___________________________________
___________________________________
Quality Assurance of MRI Systems David Hearshen
15
Slide 43
ProcessorProcessor Sensitometry Sensitometry
Though analogous to other digital imagingmodalities, printing of MR films can beproblematic due to variation in the output ofthe MR system. This variation is due toVariable gain in the transceiver subsystem
Variation in tissue contrast (T1, T2, )
Variation in operator selectable parameters
___________________________________
___________________________________
___________________________________
___________________________________
___________________________________
___________________________________
___________________________________
Slide 44
ProcessorProcessor Sensitometry Sensitometry
Example ProtocolDisplay SMPTE pattern
Inspect SMPTE image on the monitorNote 5% and 95% levels visibility
Note any artifacts
Film image using preset window and levelWindow width 256, Window level 128
Scan output image and recordSpeed, Contrast, Base+Fog
___________________________________
___________________________________
___________________________________
___________________________________
___________________________________
___________________________________
___________________________________
Slide 45Modified SMPTE ImageModified SMPTE Image
___________________________________
___________________________________
___________________________________
___________________________________
___________________________________
___________________________________
___________________________________
Quality Assurance of MRI Systems David Hearshen
16
Slide 46
0.00
0.50
1.00
1.50
2.00
2.50
3.00
1 4 7 10 13 16 19
Dens ity Ste p
Op
tica
l D
en
sit
y
Tone Scale CalibrationTone Scale Calibration___________________________________
___________________________________
___________________________________
___________________________________
___________________________________
___________________________________
___________________________________
Slide 47
ProcessorProcessor Sensitometry Sensitometry
This procedure allows a standardized way ofcomparing the image on the monitor with theoutput of the laser printer. The film densities maybe scanned, plotted, and interpreted in a mannerconsistent with other processors used in thedepartment, however, this test does not narrow thesource of a potential problem to either themonitor, MRI console output (either video ordigital), laser printer performance, or processorperformance.
___________________________________
___________________________________
___________________________________
___________________________________
___________________________________
___________________________________
___________________________________
Slide 48Processor QCProcessor QC
___________________________________
___________________________________
___________________________________
___________________________________
___________________________________
___________________________________
___________________________________
Quality Assurance of MRI Systems David Hearshen
17
Slide 49SNR ResultsSNR Results
75
80
85
90
95
100
105
110
115
120
125
7/7/9
4
7/18/
94
8/1/
94
8/15/9
4
8/29/9
4
1/16/9
5
2/20/
95
4/25/
95
5/8/9
5
8/7/9
5
9/11/9
5
9/26/9
5
+/- 10%
mean =104.3 and σ=2.6
___________________________________
___________________________________
___________________________________
___________________________________
___________________________________
___________________________________
___________________________________
Slide 50
-200
-150
-100
-50
0
50
100
150
200
7/7/9
4
7/18/
94
8/1/9
4
8/15
/94
8/29/
94
1/16/
95
2/20
/95
4/25
/95
5/8/9
5
8/7/
95
9/11
/95
9/26
/95
∆υ
1.0T
1.5T
Daily Change in FrequencyDaily Change in Frequency___________________________________
___________________________________
___________________________________
___________________________________
___________________________________
___________________________________
___________________________________
Slide 51Review of Test ResultsReview of Test Results
Technologist - Daily ChecklistInspect QC phantom for artifacts
Inspect SMPTE Pattern for artifacts (monitor)
Inspect SMPTE Pattern for artifacts (film)
Graph Center Frequency, SNR, Sensitometry
PhysicistReview of Daily QC
Monthly or Quarterly (ACR Semi-Annual)
Review Artifacts, Action Limits -ad hoc
___________________________________
___________________________________
___________________________________
___________________________________
___________________________________
___________________________________
___________________________________
Quality Assurance of MRI Systems David Hearshen
18
Slide 1
Artifact Inspection
MRI is susceptible to artifacts:Low SNR (and low energy) compared to other
imaging modalities
Ability to directly modify spatial frequencysampling
Sensitivity to metallic objects
Patient – RF coil interaction is variable
___________________________________
___________________________________
___________________________________
___________________________________
___________________________________
___________________________________
___________________________________
Slide 2
Artifact Inspection
Artifacts may appear differently in test objectsand patients because of differences in:Magnetic susceptibility
Coil loading
T1, T2 of test object material
High contrast between plastic (no signal) andfluid
___________________________________
___________________________________
___________________________________
___________________________________
___________________________________
___________________________________
___________________________________
Slide 3EPI 1/2 FOV Phase Artifact
___________________________________
___________________________________
___________________________________
___________________________________
___________________________________
___________________________________
___________________________________
Quality Assurance of MRI Systems David Hearshen
19
Slide 4
Artifact Inspection
A single source may cause artifacts to appeardifferentlyUsing different coils
Using different pulse sequence parametersAnatomic plane
pulse sequence type (eg spin echo vs. gradient echo)
timing (TR, TE, gating)
Spatial resolution (FOV, matrix size)
Using different materials
___________________________________
___________________________________
___________________________________
___________________________________
___________________________________
___________________________________
___________________________________
Slide 5
Gradient Echo vs. SE of Bleed
___________________________________
___________________________________
___________________________________
___________________________________
___________________________________
___________________________________
___________________________________
Slide 6
RF Artifacts
External RF NoiseSpecific frequency usually one pixel in
width
Phase Incoherent
Independent of anatomic plane
May change position with FOV orbandwidth
___________________________________
___________________________________
___________________________________
___________________________________
___________________________________
___________________________________
___________________________________
Quality Assurance of MRI Systems David Hearshen
20
Slide 7External RF Noise
___________________________________
___________________________________
___________________________________
___________________________________
___________________________________
___________________________________
___________________________________
Slide 8External RF Noise
___________________________________
___________________________________
___________________________________
___________________________________
___________________________________
___________________________________
___________________________________
Slide 9
RF Artifacts
Transmit Receive switch failureInductive coupling of transmit coil and
surface coil
Spatial variation in RF power
___________________________________
___________________________________
___________________________________
___________________________________
___________________________________
___________________________________
___________________________________
Quality Assurance of MRI Systems David Hearshen
21
Slide 10Transmit - Surface Coil Artifact
Phased Array Spine
___________________________________
___________________________________
___________________________________
___________________________________
___________________________________
___________________________________
___________________________________
Slide 11ADC Overflow (Clipping)
Occurs when the signal is larger thanthe maximum ADC value, usually if gainis set too high.
___________________________________
___________________________________
___________________________________
___________________________________
___________________________________
___________________________________
___________________________________
Slide 12
Non Phase Encoded Signal
Stimulated Echoes
FID (inhomogeneous B1, gradients)
Signal from outside FOVUsually if phase and frequency directions
swapped.
Small FOV in larger body part
___________________________________
___________________________________
___________________________________
___________________________________
___________________________________
___________________________________
___________________________________
Quality Assurance of MRI Systems David Hearshen
22
Slide 13Stimulated Echo Interference
Fringe spacing inverselyproportional to differencein echo timing.
___________________________________
___________________________________
___________________________________
___________________________________
___________________________________
___________________________________
___________________________________
Slide 14Axial, Sagittal Lumbar Spine
___________________________________
___________________________________
___________________________________
___________________________________
___________________________________
___________________________________
___________________________________
Slide 15Homogeneity ArtifactsLocal inhomogeneity
causes shifts in position in frequency encodingdirection
causes dephasing within a voxel
causes frequency selective pulses to be spatiallyshifted
Generally independent of pulse sequenceparameters exceptworse on gradient echo vs spin echo
saturation pulses
___________________________________
___________________________________
___________________________________
___________________________________
___________________________________
___________________________________
___________________________________
Quality Assurance of MRI Systems David Hearshen
23
Slide 16Homogeneity Artifacts
Shim Misadjusted
SE GRE
___________________________________
___________________________________
___________________________________
___________________________________
___________________________________
___________________________________
___________________________________
Slide 17Homogeneity/Susceptibility Artifact
Susceptibility
Homogeneity
___________________________________
___________________________________
___________________________________
___________________________________
___________________________________
___________________________________
___________________________________
Slide 18 Metallic Artifacts - SE Spine ___________________________________
___________________________________
___________________________________
___________________________________
___________________________________
___________________________________
___________________________________
Quality Assurance of MRI Systems David Hearshen
24
Slide 19Metallic Artifacts - FSE Spine
SpatialSat
___________________________________
___________________________________
___________________________________
___________________________________
___________________________________
___________________________________
___________________________________
Slide 20Metallic Artifacts - SE Spine
WireRemoved
___________________________________
___________________________________
___________________________________
___________________________________
___________________________________
___________________________________
___________________________________
Slide 21Metallic Artifacts SE Knee
Coronal
___________________________________
___________________________________
___________________________________
___________________________________
___________________________________
___________________________________
___________________________________
Quality Assurance of MRI Systems David Hearshen
25
Slide 22Metallic Artifacts FSE Knee
CoronalFat Sat
___________________________________
___________________________________
___________________________________
___________________________________
___________________________________
___________________________________
___________________________________
Slide 23Metallic Artifacts FSE Knee
SagittalFat Sat
___________________________________
___________________________________
___________________________________
___________________________________
___________________________________
___________________________________
___________________________________
Slide 24
Gradient Artifacts
Geometric DistortionSimilar to Metallic Artifact but usually
over FOV
Direction DependentIn plane Frequency vs. Phase
Slice profile distortion
___________________________________
___________________________________
___________________________________
___________________________________
___________________________________
___________________________________
___________________________________
Quality Assurance of MRI Systems David Hearshen
26
Slide 25Geometric Distortion
___________________________________
___________________________________
___________________________________
___________________________________
___________________________________
___________________________________
___________________________________
Slide 26
Phase Artifacts
Gradient effectsSpatially dependent SNR loss
GhostsDifferentiate from motion
Non gradient effectsPhase Modulation Ghosts
DC offset
___________________________________
___________________________________
___________________________________
___________________________________
___________________________________
___________________________________
___________________________________
Slide 27 Phase Modulation Artifact
Weak Ghost displaced in PE directionPeriodicity dependent upon synchronization with TRLine voltage, harmonicsGradient instability
___________________________________
___________________________________
___________________________________
___________________________________
___________________________________
___________________________________
___________________________________
Quality Assurance of MRI Systems David Hearshen
27
Slide 28EPI 1/2 FOV Phase Artifact
___________________________________
___________________________________
___________________________________
___________________________________
___________________________________
___________________________________
___________________________________
Slide 29Phase Contrast Venogram
___________________________________
___________________________________
___________________________________
___________________________________
___________________________________
___________________________________
___________________________________
Slide 30 Gradient Instability Artifact
___________________________________
___________________________________
___________________________________
___________________________________
___________________________________
___________________________________
___________________________________
Quality Assurance of MRI Systems David Hearshen
28
Slide 31
Data Acquisition Errors
Data corrupted during digitization before FFT(eg electrical transient)
Bad memory locations
Pattern depends on what data points aremodified
___________________________________
___________________________________
___________________________________
___________________________________
___________________________________
___________________________________
___________________________________
Slide 32Data Acquisition Errors
Modifying a fewpoints during digitization of asingle echo.
___________________________________
___________________________________
___________________________________
___________________________________
___________________________________
___________________________________
___________________________________
Slide 33Signal Corrupted During Scan
Similar to a data acquisition error, the signal may bemodified by discrete events during any part of thepulse sequence. Their effect upon the image dependsupon when and how many events occur.
ν ν
Not atisocenter
___________________________________
___________________________________
___________________________________
___________________________________
___________________________________
___________________________________
___________________________________
Quality Assurance of MRI Systems David Hearshen
29
Slide 34Signal Corrupted During Scan
ν
Noise appears to be phase incoherent at specific frequencies onupper right.
___________________________________
___________________________________
___________________________________
___________________________________
___________________________________
___________________________________
___________________________________
Slide 35Signal Corrupted During Scan
ννAxial Sag
High frequency or high contrast residual image of legTwo dimensional pattern
___________________________________
___________________________________
___________________________________
___________________________________
___________________________________
___________________________________
___________________________________
Slide 36Artifact Summary
Artifact vary with imaging conditionsPulse sequence details, coil, patient
RF ArtifactsExternal noise phase incoherent, single pixelOther artifacts not along orthogonal directionsPhase modulation - ghosts in PE direction
Gradient ArtifactsInstabilities - Ghosts in PE directionDistortion - direction dependentSeverity, appearance direction dependent
___________________________________
___________________________________
___________________________________
___________________________________
___________________________________
___________________________________
___________________________________
Quality Assurance of MRI Systems David Hearshen
30
Slide 37
Acknowledgments
Carl Gregory Biomedical Magnetic Resonance Laboratory
University of Illinois, Urbana, Illinoishttp://bmrl.med.uiuc.edu:8080/~cgregory/notebook/artif
acts.html
___________________________________
___________________________________
___________________________________
___________________________________
___________________________________
___________________________________
___________________________________