Terms in this set (101)
When parameter ↑, what is the effect on spatial resolution (SR), SNR, and acquisition time: FOV, slice thickness, image matrix.
↑ FOV → ↓ SR, ↑ SNR, no change of acquisition time.
↑ slice thickness → ↓ SR, ↑ SNR, no change of acquisition time.
↑ image matrix → ↑ SR, ↓ SNR, ↑ acquisition time.
When parameter ↑, what is the effect on spatial resolution (SR), SNR, and acquisition time: signal averages, receiver bandwidth, field strength, partial k-space sampling.
↑ signal averages → no change SR, ↑ SNR, ↑ acquisition time (SNR ↑ as the square root of the # of excitations)
↑ receiver bandwidth→ no change SR, ↓ SNR, and ↓ acquisition time.
↑ field strength→ no change SR, ↑ SNR, no change of acquisition time.
↑ partial k-space sampling→ no change SR, ↓ SNR, ↓ acquisition time.
MRI units and conversion
Tesla (SI) and gauss (cgs)
1 T = 10,000 gauss
Earth's magnetic field is ~ 50 microTesla (0.5 gauss)
What is longitudinal magnetization?
Longitudinal magnetization (T1 relaxation) → is the process of proton relaxation under the influence of the static field.
The MZ longitudinal magnetization vector grows exponentially.
T1 relaxation time = the time required for the MZ vector (MZ magnetization) to reach 63% of its maximum value.
Different tissues exhibit uniquely different T1 relaxation times when exposed to the same static field.
With a subsequent T1 relaxation period, the Mz vector relaxes another 63%
When 5 T1 relaxation periods have elapsed→ the tissue has relaxed to >99% of its maximum value→ M0 is saturated.
What is transverse magnetization?
Loss of synchronous precession is tissue specific→ called T2 relaxation.
T2 relaxation time → defined as the time for the for the transverse magnetization vector (MXY) to ↓ 63% or the time for the transverse vector to ↓ to 37% of its original value.
The net transverse magnetization occurs only when the population of protons is exposed to radio-waves of the same frequency.
What is the 90 degree RF pulse?
The duration of exposure to the RF pulse can be selected so that the # of parallel and antiparallel protons can be equalized→ resulting in no net longitudinal magnetism, MZ = 0→ the only net vector of magnetism is the transverse vector (Mxy )
→ The amount of RF energy required to equalize the parallel and anti-parallel protons (resulting in MZ = 0) → is defined as the 90° RF pulse
A 90° RF pulse→ produces the largest Mxy vector and thus the the strongest MRI signal for that voxel of tissue.
What is free induction decay?
Tissue protons after T1 relaxation are asynchronous.
Following the transmission of a 90° RF pulse→ synchronous precession occurs and the net MZ vector intersects the copper RF coil, inducing an oscillating electric current called the MRI signal.
The observed pulsating AC current in the coil has the same frequency as that of the precessing protons.
After the 90° RF pulse, the magnetization vector rotates at the Larmor frequency in the transverse (xy) plane. FID signal produced is proportional to the xy magnetization vector. The transverse magnetization decays exponentially at time constant T2.
MRI parameters for selected contrast:
PD, T1, T2
PD: long TR (2000-4000 msec), short TE (5-30 msec)
T1: short TR (400-800 msec), short TE (5-30 msec)
T2: long TR (2000-4000 msec), long TE (60-120 msec)
What is echo time (TE)?
Echo time (TE) → operator-selected interval between transmission of a 90° pulse followed by a 180° pulse and the measurement of the MRI signal in a common spin echo pulse sequence.
Longer TE values→ result in ↑ T2 contrast determined by the different T2 relaxation times of the tissues
In the spin echo pulse sequence→ a long TR and long TE result in T2 image contrast.
Example: in the brain, with a long TE, CSF has the greatest MXY vector→ thus appears brightest (CSF protons don't lose their synchronous precession as quickly as protons of white or gray matter).
5 sequences in MRI image formation:
slice-selection→ phase encoding→ refocusing→ frequency encoding → readout.
What happens in slice-selection and phase-encoding sequences?
Slice-Selection→ gradient turns on, altering the resonant frequency of H1 along the slice-select direction.
RF (radiofrequency) pulse is applied while gradient is on.
H1 nuclei having the proper resonant frequency are excited (i.e., in a "slice").
H1 magnetic moments in the excited slice precess in phase.
Phase Encoding→ labels nuclei according to position.
Gradient turns on briefly, pointing along the phase-encode direction.
Precession phase in the excited slice is altered along the phase-encode direction.
What happens in refocusing, frequency-encoding, and readout sequences?
Refocusing→ precession phase of nuclei is altered, allowing readout at TE ("echo time").
Phase is adjusted using an additional RF excitation.
Refocusing remediates some relaxation (e.g., T2*)
Frequency Encoding→ label nuclei according to position.
Gradient turns on during readout.
Precession frequency in the excited slice is altered along the frequency-encoding axis.
Frequency-encoding axis is perpendicular to phase-encoding axis.
Readout→ raw data inserted in "k-space"
How does readout work?
One readout (or one phase-encode) = one line of k-space
The process repeats after a time (TR), with different phase encoding.
K-space is filled row-by-row with each interval TR→ after many readouts, k-space is filled.
Inverse Fourier transform→ converts k-space to image.
"Fast" sequences (e.g. fast spin echo) → may fill several lines of k-space per TR
"Non-cartesian" sequences→ may fill k-space in patterns other than row by row (e.g. radial sequences→ fill k-space as a set of rows passing through the center of k-space, in a star-like pattern).
Image sequences may leave some portion of k-space empty to save imaging time (e.g. "partial fourier" techniques, parallel imaging, etc.)
Describes the resonant frequency of a nucleus, the precession frequency of a nuclear magnetic moment, and the dependence of resonant frequency on magnetic field strength.
w = yB
w = resonant (and precessional) frequency of nucleus.
y = gyromagnetic ratio (intrinsic, constant property of a nuclear species).
B = magnetic field where nuclear moment is located.
If B varies in space, so does w
How do gradients work in the context of the Larmor equation?
In the context of MRI→ a gradient is a magnetic field that varies spatially
Applied to Larmor Equation:
The main field of an MRI scanner is very uniform→ w is uniform
Turning on gradients changes B
Gradients vary spatially→ B varies spatially→ w varies spatially
To form an image, one must be able to identify where tissue is within the scanner.
Solution: label magnetic moments by position, using gradients.
Turning on the magnetic gradient alters the resonant frequency spatially.
Purpose of slice-select gradient, and what happens if it doesn't turn on?
Slice-select gradient→ responsible for specifying excited slices (i.e. slice orientation).
If the direction is changed, the direction of the stack of slices will change as well.
If the slice-select gradient doesn't turn on, basically you will not excite a slice but a huge portion of the patient, because the magnetic field in the scanner is not perfectly uniform.
How is slice-thickness selected?
Slice Thickness→ adjusted by the strength of the gradient and the transmit RF bandwidth.
Stronger gradients = thinner slices
Thicker slices→ result in more tissue averaging within an image voxel→ blurring with increasing thickness.
If the RF pulse is a very narrow bandwidth (close to zero), then the slice will be diminishingly thin.
How are voxels grouped for readout and image formation?
Voxels→ grouped into ensembles
Ensemble→ helps disentangle the location of moments.
Instead of detecting signal from each pixel separately, MRI uses locations paired into ensembles.
Encoding (phase and frequency)→ sets up spatial variations of phase that act like ensembles.
Signal from one set of ensembles = one data point in k-space→ many ensembles are required to fill k-space and construct a full image.
Using ensembles to create an image is more time-efficient than exciting each voxel separately.
During each readout, signal from many different pairs of ensembles is detected sequentially.
2D imaging sequences and advantages over 3D:
2D imaging sequences → acquires images by exciting a stack of thin (2D) slices
Thin "slice" is excited.
Slice-selection gradient "localizes" tissue along one axis of the imaging volume.
Phase and frequency encoding proceeds in each slice.
Phase and frequency encoding localizes tissue along 2 slice axes.
Typically faster scan times than 3D
Phase-encoding artifacts only along one axis.
3D imaging sequences and advantages over 2D:
3D imaging sequences
Thick (3D) "slab" is excited.
The excitation does not localize tissue very well because of slab thickness.
Phase and frequency encoding proceed in slab→ providing tissue localization in 3D.
Frequency encoding→ localizes tissue along one axis.
Phase-encoding→ occurs along two axes perpendicular to frequency-encode.
Phase encoding along first axis→ behaves like 2D phase encoding across the slab face.
Phase encoding along 2nd axis→ behaves like 2D phase encoding along slab length.
Smaller voxels possible with good signal-to-noise (e.g. MRA)
Voxels with equal dimensions possible, for post-acquisition image reformatting.
Artifacts that appear predominantly along the phase-encode axis (e.g. flow artifacts, aliasing) → can appear along 2 different axes in 3D imaging sequences, instead of one axis in 2D imaging.
What artifact occurs when the # of samples in k-space (ensembles) is too small?
Gibbs (truncation) artifact.
Gibbs artifact→ appears as alternating bright and dark lines on the MRI images that are parallel to an adjacent border exhibiting an abrupt change in signal intensity. Can be in phase- and/or frequency-encoding direction.
Origin→ the number of samples in k-space (ensembles) is too small. Fourier-based imaging technique leads to image errors near sharp boundaries.
Lessen or eliminate the artifact:
One can remediate the effect by specifying more phase encodings and more samples acquired during frequency encoding (i.e. increase matrix) → this ↓ the spacing between artifactual lines.
Can also use pre- or post-reconstruction image filtration
Ghosting (Nyquist or N/2) Artifact
Nyquist ghost→ appears as an image copy or multiple copies with decreased intensity that is shifted from the center of the FOV.
Occurs along the phase-encoding direction.
Origin: errors in phase between readouts, due to imperfections in gradient performance (Nyquist). Can also be caused by patient motion (blurring).
Because of the strong demand that an echo-planar-imaging (EPI) sequence places on gradient speed and performance→ ghosts are particularly common in EPI images.
Apply eddy-current correction (Nyquist, or N/2 ghost)
Use flow compensation or apply saturation bands to ↓ the artifact (flow)
Ask the patient to be still (motion).
Use motion-resistance sequences (e.g. SSFSE, partial NEX, PROPELLOR or BLADE or MULTI-VANE).
Use respiratory or cardiac gating, or use breath-holding techniques.
MRI motion artifact
Motion Artifact→ appears as a general blurring of the MRI image, and/or with copies of structures (e.g. vessels, cavity boundaries) duplicated along the phase-encode direction.
Origin: periodic motion faster than the time between different phase encodings.
Motion other than pulsatile flow results in blurring.
One image-formation based technique to identify this artifact is to invert the phase-encode and frequency-encode direction→ motion artifact will flip positions after this inversion of axes.
Chemical shift artifact:
Chemical Shift Artifact→ appears as a bright or black band at the interfaces of fatty tissue surrounded by non-fatty tissue.
The bright band will present on one side of the fatty tissue; the black band will present on the opposite side.
The bands line up along the frequency-encode direction.
The artifact may ↓ the contrast between abnormal lesions and normal tissue, thus ↓ the diagnostic yield of the study.
Origin→ different resonant frequency of fat mimics frequency-encoding.
Method to lessen the effect→ invert the phase-encode and frequency-encode directions.
↑ the bandwidth (Type 1)
For short TE sequences, change TE such that fat and water are mostly in phase (Type 2).
Lengthen TE such that it is greater than ~ 30 msec, and apply shimming over the FOV (Type 2).
Zippers/spike noise artifact:
Stippled/dashed band though the image.
Banding runs primarily along the phase-encoding direction.
Causes→ detection of extraneous RF signals during readout.
Incompatible or malfunctioning medical equipment inside MR scanning room (e.g. syringe pump).
Breached RF shielding of scan room, or door to scan room left open.
Make sure the door to the scanner room is closed.
Have equipment operating in the scan room (e.g. pumps) checked for RF emissions.
Call on medical physics staff or field engineer to find RF leak in shielding.
Spike artifacts→ appear as a series of bands alternating between light and dark, superimposed on the image
Origin→ come from a single "spike" or clump of data in k-space, elevated by noise.
The pattern in a spike artifact resembles an ensemble during phase/frequency encoding.
What artifact occurs when the FOV in the phase-encoding direction is smaller than the anatomy that is being imaged?
Aliasing ("Wrap-Around") Artifact→ appears as an anatomical structure, from outside the FOV, that wraps around to the other side of the image.
This wrap-around→ predominantly occurs along the phase-encoding direction → for this and other reasons, the phase-encoding direction is often chosen along the short axis of the tissue being imaged.
Origin→ FOV in the phase-encoding direction is smaller than the anatomy that is being imaged.
Method to lessen the artifact: exchange the frequency-encode direction with the phase-encode direction.
In general, if this is a concern, one should choose the short axis of the body in the image to be the phase-encoding direction.
Aliasing can be mitigated by ensuring that the FOV properly contains the tissue that has been excited by the sequence.
Alternatively, spatial saturation bands can be applied to null the signal from tissue outside the FOV.
Magnetic susceptibility artifact and approaches to reduce it:
Dramatic changes in magnetic susceptibility.
Presence of metal objects in patient or scanner (these artifacts will tend to be very obvious).
Echo-planar (EPI) sequences.
Remove metal if possible
Try to use non-gradient echo sequences (best).
↑ bandwidth (fair) ↑ matrix (fair) ↓ slice thickness (fair)
For EPI and minimal susceptibility artifact, try using parallel imaging, or ↑ the parallel imaging acceleration rate.
Fat-saturation failure artifact
Appearance→ varying signal intensity in fatty tissues when fat-saturation is applied.
Causes→ resonant frequency of fat is not uniform in the image.
MRI scanner with poorly-shimmed magnetic fields.
Anatomy is distant from magnet isocenter (e.g. shoulder).
If the resonant frequency of fat doesn't match the special RF pulse → fat-sat will fail, and some fat will be bright on the image.
Failures can result from poor uniformity of the bulk magnetic field of the scanner.
Apply shimming within the region of interest, and try not to include air in the shimming volume.
Move the anatomy as close as possible to the isocenter.
For fingers→ ask the patient to hold them together.
Large-scale signal non-uniformity.
Shading often near the center of the FOV.
Causes→ variation in tissue conductivity.
Common Occurrences→ more prominent with ↑ field strength scanners (e.g. 3T)
Abdominal Imaging Spinal Imaging
Imaging of large water volume (e.g. pregnancy or ascites) or obese patients.
Use dielectric pads (↑ electrical conductivity near the patient).
Engage multi-channel transmission, if available on scanner.
MRI voxel volume
Voxel volume = slice thickness x (FOVPhase / Matrix SizePhase) x (FOVRead/Matrix SizeRead)
↑ FOV, ↑ slice thickness, and ↓ matrix size→ ↑ voxel volume
↑ frequency and slice-select gradient strengths result in ↓ voxel volume
↓ voxel volume→ leads to ↑ potential spatial resolution
↓ voxel volume also results in ↓ SNR and ↓ pixel size
Thus ↑ matrix size, ↓ FOV, and ↓ slice thickness lead to ↑ potential spatial resolution
What determines pixel size in MRI?
Pixel size→ corresponds to in-plane resolution
Pixel size = FOV / matrix size
Pixel size (In-plane resolution)→ inversely proportional to the product of the gradient amplitude and duration.
The more intense (higher amplitude) a gradient or the longer a gradient is applied (↑ product) → the smaller the pixel size and the better the spatial resolution.
As with matrix size and FOV, the pixel size can be different along the 2 axes → thus, the pixel size along the readout axis will be controlled by the product of the readout gradient amplitude and duration.
Determinants of slice thickness and effect with increasing slice thickness:
Slice thickness→ determines the 3rd dimension of the voxel
Depends on the RF pulse bandwidth and waveform (shape of the RF pulse) applied simultaneously with slice selection gradient.
↑ slice thickness by a factor of 2
→ results in ↑ voxel volume by a factor of 2.
→ Signal is proportional to voxel volume, which ↑ by a factor of 2, so SNR is doubled.
→ it also ↓spatial resolution
Effect of increasing or decreasing matrix size?
Image matrix size→ corresponds to the image width and height in pixels
Larger matrix size with constant FOV→ results in smaller pixel size.
↓ matrix size by a factor of 2 (256 x 256 vs 512 x 512) → cuts the time to acquire the images by a factor of 2, ↑ voxel volume by a factor of 4, and results in ↑ SNR and ↓spatial resolution (as voxel volume ↑)
SNR relationship to field strength:
↑ in the field→ ↑ difference between the populations of parallel and antiparallel spins→ ↑ potential signal
→ induces a signal in the RF coil that varies with the square of the B0 field
→ counterbalanced by a linear ↑ in the noise with B0
Voxel size and SNR relationship:
Voxel Size: ↑ voxel size results in linear ↑ in signal intensity
Assuming a uniform proton density→ the # of excited spins is proportional to the voxel size, and so is the signal intensity.
Signal ↑ linearly with ↑ voxel size.
Noise is independent of the voxel volume, and is associated with the coil-related factors.
Selective RF pulses→ yield imperfect slice profiles, whose edges aren't straight and extend beyond the nominal slice thickness.
With multi-slice technique and with contiguous slices, a selective RF pulse can thus partially excite the adjacent slices.
Likewise, if several interlacing slice stacks cross, the zone of intersection will be partially excited.
This "cross-talk" → will cause a modification in slice signal and/or a loss of signal through partial saturation in the slice or zone of intersection.
Slice-select gradient→ determines slice thickness (↑ gradient→ thinner slices), but does not produce slice cross-talk.
RF coil quality factor
As the sensitive volume of the coil ↓, the total noise detected by the coil from structures adjacent to the selected slice plane also ↓
→ Thus, a smaller coil volume will yield a better SNR
Noise sensitivity→ is related to the RF coil sensitive volume, and is relatively independent of the reconstructed voxel size, while the signal is dependent on the voxel size.
A local coil (or better, a surface coil) → has a higher SNR than a body coil due to the smaller volume of the surface coil (making it sensitive to a smaller region of the object).
Also, coil proximity to the tissue improves the signal strength.
How do we use TR and TE to emphasize T1 contrast?
→ the TR has to be set to short values (comparable to the T1 of the tissues)
→ and the TE has to be set to low values in order to minimize differences in contrast due to T2.
How do we use TR and TE to emphasize T2 and PD contrast?
In order to emphasize T2 contrast:
→ The TE has to be set to long values (comparable to the T2 of the tissues).
→ and the TR must be set to long values in order to minimize differences in contrast due to T1.
In order to obtain proton-density (PD) contrast:
TE is kept low.
TR is kept high to minimize both T2 and T1-weighted contrast.
How does Gradient-Echo create signal, and what determines image-weighting?
Gradient echo techniques use a single RF pulse and no 180° rephasing pulse
→ thus relaxation due to fixed field inhomogeneities is not reversed
→ and the loss of signal results from T2* effects (pure T2 + static field inhomogeneities).
→ the signal obtained is thus T2*-weighted rather than T2-weighted due to this magnetic susceptibility sensitivity.
Utilizes low flip angles (i.e. <90 degrees)
For a gradient-echo sequence, image weighting will depend on:
The flip angle for T1 weighting→ the greater the angle, the more T1-weighting
The flip angle is usually much smaller than the routine 90° flip angle used in spin-echo imaging.
The TE for T2
weighting→ the shorter the TE, the less T2
Spin-Echo (SE) Imaging Technique:
Each SE sequence starts with a 90° RF excitation pulse→ tips longitudinal magnetization into transverse magnetization in a selected slice.
Phase and frequency gradient pulses→ used to codify spatial location in a selected slice.
The spins dephase for a short time, TE/2
A 180° RF pulse is then applied, which rephases spins, causing a "spin echo"-180° RF pulse occurs halfway between the 90° RF and the echo.
A "spin" echo occurs when tissue spins rephase at TE.
After a time TR (time of repetition), which allows recovery of the longitudinal magnetization→ the signal is repeated.
The sequence must be repeated to fill each line of k-space.
One row of k-space is filled during each TR.
Effect of ↑ TR and TE in spin-echo imaging?
↑ TR→ ↓ the T1 weighting because it allows the longitudinal magnetization to recover more completely between excitations.
↑ TE → ↑ the T2-weighting because it allows transverse magnetization to lose phase coherence for a longer period of time.
As TE ↑ → differences in signal between short T1 tissues (e.g. fat) and long T2 tissues (e.g. CSF) become more apparent.
Fast Spin-Echo (FSE) technique:
The RF excitation pulse is followed by a "train" of refocusing pulses.
A "train" of echoes occurs→ each echo fills one row of k-space (# of lines = # of echoes)
Total scan time ↓ by the echo train length.
During echo-train collection, the transverse magnetic signal gradually ↓ for each line of the echo-train (T2-blurring)
→ T2-blurring alters the T2 weighting and ↓ the sharpness of the image
Phase rewinding gradients→ rephase the transverse magnetization → makes the transverse magnetization more consistent between phase-encode steps and minimizes artifacts.
Applications of FSE:
FSE is useful when inhomogeneity and magnetic susceptibilities in the magnetic field preclude obtaining high-resolution images by faster pulse sequences.
FSE can be modified to allow for faster 3D imaging techniques, but this results in T2-blurring of the images due to the use of long echo-train acquisition.
Fat suppression→ can help narrow the DDX, as fat and water are both hyperintense on T2-weighted FSE images.
Gradient Recalled Echo (GRE) technique:
GE starts with a small (<90°) excitation pulse → tips longitudinal magnetization, creating some transverse magnetization in a selected slice.
Phase and frequency gradient pulses→ used to codify the spatial location in a selected slice.
Negative frequency gradient pulse is applied→ causes rapid dephasing of transverse magnetization.
Positive frequency gradient pulse then rephases transverse magnetization → produces a "gradient" echo at TE.
Gradient recalled echoes are also called "field echoes."
The sequence must be repeated for each line of k-space.
GRE: benefits, problem, image contrast
GRE is more sensitive to magnetic field inhomogeneity and magnetic susceptibility than is SE.
Can be an advantage→ imaging hemorrhage
Can be a disadvantage→ loss of signal and geometric distortion.
Benefits of GRE:
Fast image acquisition time.
Lower specific absorption rate (SAR) = less heating of patient tissues
Problem with GRE: ↓ signal-to-noise (SNR) ratio
Contrast in GRE Imaging:
GRE→ produces T2
, T1, and mixed-weighted (T1/T2
2*, T2/T1) images depending mainly on flip angle, TE, TR, and the specific GRE technique used.
Spoiled GRE → allows for very short TR values and high T1-weighted contrast
Residual transverse magnetization is destroyed (spoiled) after each echo→ thus the formation of each echo begins solely with longitudinal magnetization (the amount of longitudinal magnetization reflects T1 relaxation rates)
Spoiling occurs by either:
Gradient spoiling: strong spoiler gradient.
RF spoiling (more common): phase changes in RF excitation
Fast Spoiled GRE:
Fast Spoiled GRE→ allows fast 3D imaging with high resolution
After the RF excitation pulse, the frequency encode gradient dephases and then rephases the transverse magnetization several times → produces a train of gradient echoes → Each echo fills 1 row of k-space.
Preparatory inversion pulse→ used to produce T1 weighting.
Acquisition time ↓ by the echo train length.
Coherent GRE and balanced coherent GRE:
Residual transverse magnetization is NOT destroyed (spoiled) after each echo..
Transverse magnetization approaches a "steady state" mixture of longitudinal and transverse magnetization.
The FID data and/or the echo data can be acquired.
Image contrasts→ include T1/T2*, T1, and T2
Balanced Coherent GRE
All 3 gradients are balanced such that dephasing of spins is compensated for by an equal rephasing of spins
Image contrast: includes T2/T1
Good contrast between blood and muscle (e.g. cardiac imaging)
High resolution and less magnetic susceptibility than standard balanced coherent GRE images → at the expense of longer TR and thus longer acquisition time.
Echo-Planar Imaging (EPI): technique, contrast, applications
After the RF excitation pulse→ frequency encode gradient cycles continuously.
Echoes are acquired to reconstruct an image after a single RF excitation (single-shot)
Reduced image matrix often used (e.g. 32 x 32, 64 x 64, 128 x 128)
Very short acquisition times (<50 msec)
Image contrast→ generally T2*
T2* decay→ causes the magnitude of echoes to ↓ during image acquisition.
Uses→ functional MRI (fMRI) and diffusion imaging
Inversion Recovery (IR) SE technique:
A 180° RF pulse inverts tissue magnetization→ suppresses selected tissue
During the inversion time (TI) → T1 relaxation processes cause different tissues to return toward equilibrium magnetization values at different rates.
The amount of T1 relaxation that occurs during TI will affect image contrast.
A standard spin echo acquisition sequence is then performed.
Allows for high signal intensity of fluid or edema.
Benefits of IR SE:
TI can be chosen such that the tissue magnetic moment for a given type of tissue will be very small when the 90° SE excitation occurs→ so that type of tissue won't contribute signal (i.e. it will be suppressed)
STIR (short tau IR)
Fat signal (short T1) is suppressed using short TI
FLAIR (fluid attenuated IR)
Long TI can be chosen to suppress fluid (CSF) → producing T2-weighted contrast, known as T2-FLAIR (AKA FLAIR)
TI can be chosen on the order of T1 for soft-tissue for producing T1 contrast = T1 FLAIR → thought to produce better T1 contrast images than traditional T1-weighted FSE images.
Compared to other tissue suppression techniques, IR is much less susceptible to magnetic field inhomogeneity and magnetic susceptibilities.
A 180° inversion pulse is used to "prepare" tissue magnetization before the image acquisition sequence begins.
Useful for creation of heavily T1-weighted images without a dominant contribution from fat (e.g. brain, liver, MSK)
GRE technique allows ultrafast 2D/3D imaging with high resolution
3D MRI Technique:
3D MRI Technique
Slices imaged in a 3D stack undergo RF excitation and RF detection simultaneously.
The "slice-select" gradient is used twice in 3D techniques:
First = "slab select" → used to excite tissues in all slices.
Second = "slice select" → used to spatially encode each slice with a unique set of phases and frequencies.
Image weighting can be controlled by preparatory inversion pulses.
3D MRI: advantages and disadvantages
3D MRI Benefits
Allows for thinner slices (<1mm)
Slices are contiguous.
↑ SNR compared to that with 2D acquisition.
3D voxels can be isotropic (e.g. 1mm x 1mm x 1mm) → Isotropic voxels→ enable multiplanar reconstruction with good resolution in all planes.
3D MRI Problems
Many 3D techniques are GRE-based→ thus susceptible to magnetic field inhomogeneity and magnetic susceptibilities.
Artifact-prone (e.g. ghosting, distortion, signal voids, motion).
Acquisition time: 2D vs. 3D
2D MRI: acquisition time is proportional to: TR x Np x NEX
3D MRI: acquisition time proportional to: TR x Np x NEX x S
TR = repetition time (msec)
Np = # of phase encodes
NEX = # of excitations or averages
S = # of slices
What factors affect acquisition times for modern MRI techniques?
Echo train length (ETL): inversely related with acquisition time
In fast MRI techniques, acquisition time is ↓ by a factor equal to the ETL
Example: ↑ the ETL from 4 to 16→ ↑ the # of k-space that are filled during each TR period from 4 to 16 → ↓ acquisition time by a factor of 4.
Parallel imaging acceleration factor: ↓ acquisition time
Np is ↓ through spatial sensitivity of multiple coils.
The image acquisition time is ↓ by the acceleration factor.
Rectangular FOV: ↓ acquisition time
Anatomic coverage is ↓ in the phase-encode direction→ ↓ Np and thus ↓ acquisition time.
3 categories of clinical MRI systems:
Low-field systems → B0 < 0.5 T
Have larger bore sizes and can permit larger patient weight limits → advantage for large patients.
Also permit open design variety → allows shifting the patient sideways to bring imaged anatomy closer to the magnet's isocenter.
Mid-field systems→ B0 between 0.5 T and 1.0 T
High-field systems→ B0 > 1.5 T
Offer better resolution, predominantly due to ↑ SNR
Generation of high magnetic fields over a large volume is challenging → thus these tend to have smaller bore sizes than with low-field systems.
Calculation of magnetic field inhomogeneity
Inhomogeneity of a magnetic field inside the magnet is specified relative to a reference value→ the strength of the magnetic field at the magnet's isocenter.
Inhomogeneity = (actual value of the magnetic field at a given point - value of the field at isocenter) / (value of the magnetic field at isocenter)
→ specified in parts-per-million (ppm), obtained by multiplying the value of the ratio by 1 million (106)
How is quality of the magnetic field specified and measured?
Quality of the magnetic field→ specified by indicating the largest value of inhomogeneity inside a spherical volume of specified diameter→ DSV (diameter of spherical volume)
Due to variable computation methods, the peak-to-peak (PP) method is often used
PP method→ reports the difference between the largest and smallest value of the magnetic field within the specified DSV
Typical MRI magnets today have inhomogeneity in the range of 5-10 ppm PP over 40 cm DSV
The larger the body part, the more critical the need for homogeneity
What is shimming, and in what clinical circumstances does poor shimming often become evident?
Shimming→ the process of improving the homogeneity of the magnetic field inside a large magnet → 2 categories: active and passive
Normally difficult to detect poor shimming, as it presents as geometric distortions that are hard to detect visually.
Exception: Applications that require precise geometric localization (e.g. robotic surgery or XRT planning) → here, geometric distortions can become glaring.
Poor shimming is also more easily recognized in applications that rely on very high homogeneity to produce acceptable image quality→ such as all fat-saturation techniques; or image protocols that rely on EPI acquisition schemes, such as DWI.
Passive shimming: technique, advantages, disadvantages.
Passive shimming → relies on strategic placement of small pieces of soft iron around the magnet bore
→ introduces small, local deviations in the distribution of the magnetic field.
→ If performed correctly, these small changes will counteract the inhomogeneities of the main field → more uniform field overall.
Advantage: relatively inexpensive
Disadvantage→ very slow, requiring special hardware and software.
The resulting final effect is static
If changes are required at a later time, re-shimming will require substantial downtime (up to several days) to recalculate the shim requirements and modify the amounts of shimming material inside the magnet's bore.
Active shimming: technique, advantages, disadvantages
Active shimming→ performed using an extra set of small, oddly-shaped coils, strategically positioned inside the main magnet.
The size, location, and shape of each coil are carefully chosen so that the magnetic field produced by a given coil is proportional to a given power distance from magnetic isocenter→ called a shim order.
The shimming process→ consists of running small DC currents through these coils and finely tweaking these currents until optimal homogeneity of the main field is achieved.
The # of coils rapidly ↑ with the shim order.
Disadvantage→ the process of shimming is tedious and time-consuming, as well as approximate.
Performance of a shimming system→ improves with ↑ # of coils used, but so do the cost and complexity of the process.
How can shimming be performed in real-time on newer scanners?
Gradient coils necessary for imaging act in the same manner as 1st order shim coils
Normally, large current pulses are sent to the gradient coils→ generate the short-lived, strong magnetic field gradients required for encoding of spatial information during imaging sequences.
By sending small DC currents through those coils, one can perform the 1st order shim as well.
Combining these 2 approaches permits performing 1st order shimming in real time (i.e. during the exam).
→ this improves the homogeneity of the magnetic field on a per-patient basis→ ↑ image quality
Most new scanners offer this functionality
What are the components of the most common magnet used in all high-field (>/= 1.5T) scanners?
Superconductive Magnet→ an electromagnet whose coils are wound with superconductive wire.
Superconductive wire→ typically consists of several strands of special material (usually NbTi or less commonly Nb3Sn alloy) embedded in a copper matrix forming an external shell.
After the wire is cooled below its transition temperature→ the alloy strands lose their electrical resistance→ become superconductive.
The superconductive coils are immersed in liquid helium (He) to maintain steady and stable operational temperature (4.2 K)
Main advantage and disadvantage of superconductor magnet design?
Main advantage of superconductive design → significant savings in the cost of electricity needed to run the machine (compared to resistive magnet).
After energizing with DC current, the coils are shunted with a superconductive relay switch, thus completing the current loop with zero electrical resistance
→ The electrical power supply can then be disconnected and the current in the coils will flow steadily for practically forever
→ thus, obviates the need for external source of electrical current→ electricity cost savings.
Main disadvantage→ substantial costs of maintaining the magnet's cryosystem (liquid He)
Also, when the magnet is fully operational, a large amount of energy is stored in the magnetic field produced by the coil→ release of the energy (e.g. during a quench) → could cause significant damage.
Quench = a rapid loss of superconductivity in the wire (~ 5 minutes)
→ as a result, all the energy stored in the magnetic field is rapidly dispersed into the wire as heat→ causes rapid evaporation of the liquid He coolant.
→ the magnet then loses its magnetic field and requires extensive work to restore it to an operational state.
What are magnetic fringe fields, and what type of MRI magnet design is used to combat them?
Magnetic fringe fields→ refer to magnetic field produced by the magnet but located outside the body, and can extend a large distance from the magnet's isocenter.
Actively shielded magnets→ engineered to reduce the extent in space of magnetic fringe fields, and also limit effects of eddy currents.
These utilize 2 coils sets→ carry currents in opposite directions.
Outer coil→ drastically ↓ fringe fields outside the magnet, while only slightly ↓ the field inside the bore, produced by the inner coil.
The ↓ field from inner coil is compensated for by ↑ the current in the inner, main magnet coil
The combination of effects of the coils results in equivalent nominal field strength (e.g. 1.5 T) with drastically ↓ fringe field.
Magnetic field gradient (MFG) strength, and effects of higher maximum MFG strength
Gradient strength→ critical in assessing the robustness of the MFG hardware.
MFG strength uses units of 1 milliTesla per meter (1 mT/m)
Describes an MFG that ↑ the strength of the static magnetic field by one mT per every meter of distance traveled from the system isocenter in the direction of the gradient.
Maximum MFG strength should be specified for only a single channel (i.e. only one gradient coil activated at a time)
Advantage of a system with higher maximum MFG strength → the ability to write and execute faster scanning protocols.
Applicable to breath-hold applications and real-time (e.g. cardiovascular) imaging.
Disadvantage of very high gradient strength→ potential impact on patient safety (induction of peripheral nerve stimulation).
Slew-rate of MFG coils:
The current in the MFG coils cannot be turned on or off instantaneously.
The slew-rate = the maximum rate of current increase in the MFG coil = maximum gradient strength / gradient rise time
Given in T/m per second, or T/sm
Industry standard rise time (time required to increase current from zero to the maximum value) = 1 ms
Danger of using very high slew rates→ potential for inducing peripheral nerve stimulation in the patient.
What is a side-effect of rapidly changing the MFG?
When magnetic field changes within a space, parasitic currents are induced in any conductor located within that space.
→ these currents (eddy currents) → create their own magnetic fields that oppose the changes in the original magnetic field.
Eddy currents result in deterioration of MR image quality → geometric distortion, lack of sharpness, and strange ghosting patterns commonly occur.
Due to the diverse manifestations of eddy currents, they are very difficult to specifically identify→ may explain a sudden deterioration of image quality for no obvious reason.
Method to limit the detrimental effects of eddy currents→ use of actively-shielded MFG coils.
Effects of MFG field non-linearity:
MFG distortions→ can create bizarre-looking features at the edges of large FOV, and poor performance of protocols that require off-center FOV (e.g. imaging of the shoulder).
Gradient non-linearity can also affect the planarity of structures.
In the presence of non-linear gradients→ the surface of an imaged slice is no longer planar, but forms a concave or saddle-shaped form.
→ structures visible on the image are no longer co-planar (things close to the isocenter are at a different location along the slice select direction than things at the periphery of the FOV)
RF power is proportional to:
RF power→ ↑ in proportion to the coil volume and to the square of the main magnetic field.
Example: using a smaller head coil compared to a larger body coil requires much less power. Using a 3T magnet instead of a 1.5T magnet also requires increased power for equivalent performance.
What determines the precision of the ADC conversion process?
Sampling bit depth.
Specifies the # of bits used to store the digital values of the sampled signal.
This precision is often referred to as "quantization error," since all analog values are rounded off to integer values with this accuracy.
All MRI ADC systems have sampling bit depth between 12 and 16 bits.
Inadequate bit depth→ obliterates much of the information about image details (mostly residing in the very low strength part of the signal) by introducing quantization round-off errors→ images appear less sharp.
MRI bandwidth and relationship with SNR:
Bandwidth (BW) = sampling rate / # of samples
Smaller bandwidth→ ↑ SNR, but also can cause spatial distortions and ↑ chemical shift.
Larger bandwidth→ ↓ SNR, but allows faster imaging
Proper selection of digital BW has a major impact on the final image quality →BW influences SNR, motion artifacts, the magnitude of fat shift, and the selection of available TE values.
If there is a mismatch between the digital and analog BW settings, huge artifacts will likely occur.
Rule of thumb→ use as wide a BW as the image SNR allows to minimize other limitations.
Specific absorption rate (SAR)
Specific Absorption Rate (SAR) → describes the RF power absorbed by the patient's body.
Expressed in units of W/kg→ represents the average energy deposited in one kilogram of tissue per second, averaged over some set scanning time.
Proportional to # of images acquired per unit time and depends on patient dimensions, RF waveform, tip angle, and coil type.
Blood Oxygen Level Dependent (BOLD) contrast- used in fMRI
When the brain is imaged using MR acquisitions sensitive to differences in local magnetic field (T2*-weighted) → the resultant signal is ↑ regions with higher concentrations of oxygenated hemoglobin
It is necessary to have a conditional contrast (e.g. task vs rest) to detect the BOLD signal changes.
In resting-state fMRI (RSfMRI): Low-frequency (<0.1 Hz) BOLD fluctuations→ may show strong correlations at rest in brain networks that would operate together during the performance of a task.
Basic DWI sequence technique:
2 magnetic field gradient pulses of equal area are placed equidistant and on opposite sides of a 180° pulse in a spin-echo pulse sequence.
For a given gradient pulse duration (d) → the amount of diffusion-weighting that is applied in a pulse sequence is determined by the amplitude of the diffusion-weighting gradients (G) that are applied.
Typically, lower amplitudes are chosen to image slower diffusion.
The grayscale contrast in a DW image with a non-zero b value (typically 1000 units) → determined by a combination of the T2 relaxation time and the DC of the extracellular water.
However, a computed map of the ADC isolates the DC from the T2 relaxation time.
Calculation of ADC→ requires 2 or more acquisitions with different diffusion weightings obtained by changing the amplitude of the diffusion gradient.
Biological mechanism of restricted diffusion in cerebral ischemia?
Acute cerebral ischemia→ if the cerebral blood flow is ↓, the cell membrane ion pump fails→ excess Na enters the cell
This is followed by a net movement of water from the extracellular (EC) to the intracellular (IC) compartment → resulting in cytotoxic edema.
Relative to the EC space, diffusion of the IC water molecules is restricted by IC organelles → this restricted diffusion results in ↓ ADC and ↑ signal intensity on DWI
Restricted diffusion in brian tumors
Brain tumors→ the central portions of some primary and secondary brain tumors, if not yet necrotic, may exhibit restricted diffusion as they outgrow their blood supply and become ischemic.
Once the tumors are necrotic, they will not demonstrate restricted diffusion→ thus ADC appears higher in the center.
2 major categories of MRA techniques:
Dark blood→ obtained using a spin-echo based technique where the moving blood experiences the 90° pulse, but has moved out of the slice plane before experiencing the 180° pulse.
→ Thus, signal from the moving blood is not dephased→ so it appears dark in the image.
Stationary tissue→ experiences both the 90° and 180° pulses→ so it appears bright relative to the moving blood.
Bright blood scans→ obtained using a gradient-echo (GRE) pulse sequence.
The moving blood is "fresh" relative to the stationary tissue, which is saturated by repeated application of RF pulses.
→ As a result, the fresh blood appears bright relative to the stationary tissue.
How is contrast generated in magnetization transfer (MT) MRI?
MT contrast arises from the application of an additional off-resonance RF pulse → this pulse selectively saturates macromolecule-bound protons and partially saturates MRI-visible free water protons through magnetization exchange between these 2 distinct proton pools.
This exchange of magnetization→ allows the presence of the macromolecular proton pool and the degree of exchange to be observed through the effect on the observed water signal.
Basic categorization of MRI contrast agents?
Positive Contrast Materials → cause hyperintensity on T1-weighted images
Positive agent→ stimulates an ↑ in spin flip transitions→ results in ↓ T1 values and ↑ brightness on T1-weighted images.
Also known as relaxation agents.
Negative Contrast Materials → produce substantial magnetic inhomogeneity due to magnetic susceptibility.
Magnetic susceptibility→ perturbs the Larmor frequency of protons → results in a loss of phase coherence and ↓ T2 values (hypointensity)
Negative agents→ also known as shift agents, chemical shift agents, or frequency agents.
Gadolinium-based contrast agents increase risk of NSF in patients with:
Acute or chronic severe renal insufficiency (GFR < 30 mL/min/1.73m2)
Acute renal insufficiency of any severity due to the hepatorenal syndrome or in the perioperative transplantation period.
What are FDA limits for static field exposure?
Significant risk constituted as:
Main static magnetic field > 8T for adults, children, and infants older than 1 month
> 4T for neonates and infants < 1 month old.
Mechanism of burns from RF heating of conductors:
RF heating of conductors→ could lead to potential burns
This is most likely when conductors on the patient's body form loops → induced voltage causes current induction within the loops and causes heating.
The same may occur with implanted devices.
Factors affecting specific absorption rate (SAR)
Factors Affecting SAR
SAR ↑ with the square of the magnetic field strength
SAR limits can be easily surpassed with high field (> 3T) system
SAR ↑ with the square of the flip angle
Fast sequences typically use a low flip angle (10-40°), but TR is also usually low → energy deposition as a function of time may still be high.
SAR ↓ with ↑ (lengthened) TR → this ↑ the overall time in which energy is deposited, thus ↓ deposited power
SAR ↑ with the # of RF pulses per cycle (TR), for example, in high echo train length spin echo sequences.
SAR ↑ with ↑ patient size in order to deposit signal throughout a larger interrogated volume.
Methods of ↓ SAR:
The use of quadrature coils (circularly polarizing or CP) rather than linear coils for transmission of RF → ↓ SAR by approximately ½
↓ flip angle for refocusing pulses (i.e. using 160° rather than 180° RF pulses for spin echo sequences) → especially effective for fast spin echo sequences.
↑ (lengthening TR) → thereby ↓ the duty cycle.
Imaging fewer slices per TR or overall sequence.
↓ the echo train length (turbo factor) with fast or turbo spin echo sequences.
If possible, avoiding use of the body (bore) transmit coil when other transmit/receive coils may be substituted (e.g. head or knee).
SAR Exposure Standards: FDA Limits
According to FDA, SAR should be limited to produce less than a 1° C body-core temperature increase.
The following significant risk criteria should be adhered to, with values less than:
4 W/kg averaged over the whole body for 15 minutes.
3 W/kg averaged over the head for 10 minutes
8 W/kg in any gram of tissue in the head or torso for 15 minutes
12 W/kg in any gram of tissue in the extremities for 15 minutes.
Biological effects of gradient coils in tissue:
Rapidly-fluctuating magnetic fields→ may induce currents in tissues.
These currents may exceed the nerve depolarization threshold → resulting in peripheral nerve stimulation (PNS)
PNS→ most likely to occur in pulse sequences with high gradient duty cycle → like EPI
Retinal stimulation→ results in harmless visual flashes of light called magnetophosphenes.
Gradient Exposure Standards: FDA Limits
Gradient Exposure Standards:
FDA limits gradient dB/dt to that "sufficient to produce severe discomfort or painful stimulation."
Threshold for instantaneous acoustic trauma is 140 dB
Use of the highest duty cycle sequences (e.g. EPI and DTI) permits acoustic noise at the patient to near these levels.
Properly fitted earplugs offer about 10-30 dB protection
Headphones offer an additional 10-30 dB.
In tandem, they ↓ SPL by at least 40 dB, well below the recommended 99 dBA limit
FDA significant risk criteria for sound pressure level (SPL) is:
Peak unweighted SPL > 140 dB and...
Weighted root mean square (RMS) sound pressure level > 99 dBA with hearing protection in place.
MRI Site Access Restrictions:
Zone 1 (general public):
Zone 2 (unscreened MR patients): the interface between Zone 1 and the strictly controlled Zones III and IV
Patients are greeted, screening questions answered here.
Zone III (screened MR patients and personnel): this comprises the area where interactions between the MR scanner's magnetic fields with ferromagnetic objects or equipment could result in severe injury or death.
The static magnetic field in Zone III may exceed 50 𝞵T (5 gauss)
The area should be clearly demarcated and marked as potentially hazardous
Unscreened non-MR personnel shouldn't have access
Zones IV (screened MR patients under constant direct supervision of trained MR personnel):
This is synonymous with the MR scanner room.
Should be demarcated and clearly marked by signage as being potentially hazardous due to strong magnetic fields.
Levels of MR personnel
2 levels of MR personnel:
Level 1→ those having passed minimal MR safety education to work within Zone III
Level 2→ those more extensively trained or educated in MR safety → can enter into and work within Zone IV
All others are considered non-MR personnel
Ferromagnetic object (iron, steel, cobalt, nickel) screening:
Ferromagnetic objects must be restricted from entering Zone III
Sites should have immediate access to a strong handheld magnet (> 1000 gauss) to text for grossly detectable ferromagnetic forces.
All portable metallic or partially metallic objects brought into Zone IV are to be labeled according to FDA/ASTM criteria.
MR Safe, not MR safe, or MR conditional (until the composition of the item is known to be non-metallic)
Intracranial aneurysm clips→ MRI shouldn't be performed until there has been documentation that the clip is either MR safe or MR conditional
Documented titanium clips can be accepted for scanning without other testing.
What is the critical factor in MRI zonal design?
The location of the 0.5 mT (5Gs) boundary → inside which the static magnetic field exceeds 0.5 mT (about 10X Earth's magnetic field).
The 0.5 mT (5 Gs) limit→ adopted by FDA as a safety boundary defining the restricted access space of the MRI scanner
Divides "unrestricted public space" from the "restricted access space" of the MRI suite.
Arterial spin labeling (ASL)
Non-contrast MRI technique where endogenous blood in the arterial supply is saturated or labeled and then allowed to perfuse the tissue in the imaging slice.
A 2nd scan is acquired at the same location with unlabeled blood protons.
Small signal loss in the images with the labeled blood- the difference image between the labeled and unlabeled scans will be proportional to the CBF.
Phase-contrast MRI (pc-MRI):
Technique for noninvasive assessment of hemodynamics.
The PC image provides a method of quantifying flowing blood through vessels. By measuring the phase shift that occurs as protons in the blood move through a magnetic field, the velocity and direction of the blood can be obtained.
Stationary tissues appear gray, while tissue moving through the plane appears as shades of either white or black, depending on the direction.
What is the predominant factor in determining T1 contrast?
TR- affects contrast between tissues that differ in T1 values
Long TR- permits magnetization in all tissues to fully recover- generate no T1-weighting.
Short TR- only tissues with short T1 values fully recover their longitudinal magnetization and contribute a signal.
Short TR times are less than ~300 ms at 1.5T and ~450 ms at 3T.
What is the predominant factor in determining T2 contrast?
Short TE values will result in little loss of transverse magnetization (i.e. little T2 decay)- thus produce no differences (contrast) between tissues that have different T2 values.
Short TE values- minimal T2-weighting
Long TE values- will reduce the intensity of transverse magnetization for tissues with short T2 much more than for tissues with long T2.
Long TE values are typically > 60 ms
What are the effects of increasing MR field strength?
Increasing field strength increases T2 relaxation time, SNR, and RF energy deposition in the patient. May also increase certain artifacts.