Spectra Display Convention
Bright fat in Turbo Spin Echo sequences
Fatsat & Field Strength
Fatsat - Key points
Problems with Fatsat
Shimming the MAGNETOM
Intrinsic field inhomogeneity
Phase Cycling Gradient Recalled Echo
STIR Short TI Inversion recovery
TI and Turbo Spin Echo
STIR and Gadolinium
STIR - Key Points
Spectrally Selective Inversion
|MTC and Fat Signal||Conclusion|
|Spin Echo as a Fat Suppression Method||References|
|Back to Greg's MRI Home Page|
Fat signal is visible on most MRI images.
Fat suppression techniques work by taking advantage of two characteristic
differences in the MR behaviour of fat and water, Relaxation rates (T1
& T2), and Chemical Shift effects
Fat has a very short T1 relaxation rate, thus it returns a high signal on most T1 weighted sequences. Its T2 value, as observed in SE images, is slightly shorter than that of most tissues, so fat will display moderate to low signal intensity. In sequences that apply multiple 180O pulses rapidly (TSE, FSE), the observed T2 times of fat change significantly .
The MR signals from fat and water have slightly different frequencies. This difference is called the chemical shift. The chemical shift effect results from differences in the distribution of the electron cloud surrounding Hydrogen nuclei in the molecules of` water and fat. The external magnetic field causes an additional drift in the electron clouds of molecules. This drift of charged particles creates a magnetic field that opposes the external field (Lenz's Law) and thus decreases the strength of the magnetic field experienced by nearby nuclei. The electrons are said to "shield" the nucleus. The strength of this opposing field is proportional to the strength of the external field and the density of the electron clouds surrounding the nuclei. A stronger shielding field reduces the experienced magnetic field so the nucleus will precess at a lower Larmor frequency.
Fat/Water Chemical Shift
|Field Strength||Freq. difference|
|1.5 T||223 Hz|
|1.0 T||150 Hz|
|0.5 T||75 Hz|
Chemical shift is expressed in parts per million (ppm), the absolute
frequency difference is proportional to the main magnetic field strength.
Water precesses approximately 3.5 ppm faster than fat.
By convention MR spectra (figure 1) are displayed with a reversed frequency
axis. The higher frequency peaks are plotted to the left of the page. In
many texts the sign of the axis is left out completely leaving you to assume
incorrectly that fat has a higher resonant frequency than water. The confusion
is further compounded by texts that only discuss the difference in frequency
and not the mechanism for it.
The convention results from the methods used to obtain most early NMR spectra, in which the RF system operated at a fixed frequency and the magnetic field was swept from low to high values. Thus species with a higher apparent Gyromagnetic ratio would resonate at the fixed frequency with a lower applied external field, appearing earlier in the field sweep and therefore on the left of the display.(1)
There are two reasons why fat appears bright in virtually all sequences
that rapidly apply multiple RF pulses; Magnetization Transfer (MT) effects
and J-Coupling. The multiple RF pulses act as off-resonance MT pulses,
saturating the bound pool of protons. MT induced signal loss occurs in
most stationary tissues but not much in fat, leaving it relatively brighter.
The more significant effect is based on a phenomenon know as J-coupling
or Scalar Coupling.
J-Coupling or Scalar Coupling is the most significant reason why fat
appears bright on RARE (Turbo) sequences, regardless of their contrast
weightings. Through interaction with the electron cloud, a nucleus can
alter the magnetic field strength experienced by another nucleus on the
same molecule. Therefore various nuclei on the same molecule will precess
at slightly different frequencies, and will interfere constructively and
destructively in a cyclic fashion with each other (beat). In fat molecules,
this results in some T2 shortening, which causes a decrease of signal intensity
in images obtained with TE values between 25-100 msec. In Turbo sequences
refocussing pulses are typically applied with an echo spacing of less than
25 msec. This repeated refocussing neutralises the J-coupling mechanism
of T2 signal loss, and so the fat signal can increase by about 50% compared
to SE images. The J-coupling effect is independent of field strength, so
the increase of fat signal is observed equally when using TSE sequences
on all systems.
|Chemical Shift||T1 times||Miscellaneous|
|Selective Excitation||SPIR||Spin Echo|
|Phase Cycling GRE||MTC|
Commercially available fat suppression techniques are known by a number
of vendor specific acronyms. This section of the paper will group them
by the underlying principle used and try to clear the confusion. Not all
of these methods are discussed in this paper.
Also known as: FATSAT (FS), CHEMSAT, CHESS.
techniques are known generically as Spectral Saturation routines. The method
takes advantage of the difference in resonant frequencies between water
and fat. A 900 RF pulse, tuned to the resonant frequency of
fat is applied flipping the bulk magnetic vector from fat into the transverse
plane. Spoiler gradients are then applied to destroy the phase coherence
of the signal. This saturation routine is followed immediately by the imaging
sequence and the images will only show signal from the remaining water
nuclei. Fat based nuclei will not produce a signal until there has been
time for significant T1 based recovery (approx. 100 msec).
By adjusting the frequency and bandwidth of the saturation pulse water or silicon can be selectively saturated.
The great attraction of the Fatsat technique is that it only modifies the signal of fat, all other contrast relationships and signal characteristics remain the same. Fatsat can be applied to virtually any sequence with any weighting.
Fatsat is simpler and faster to accomplish at higher field strengths because the separation of the fat and water peaks is greater in absolute terms, and wider bandwidth or shorter period pulses can be used.
The application of saturation pulse and gradients takes a considerable time, reducing the time available for acquiring multiple echoes or slice locations within a given TR. On the Siemens VISION scanner the Fatsat routine requires 19 msec per slice in conventional sequences, and 10 msec per slice for Turbo sequences. In many cases TR must be increased or slices sacrificed to implement Fatsat which can cause problems particularly with short TR techniques. Sequence designers usually apply a Fatsat routine preceding each excitation for each slice but when time savings are important they may opt to apply them less frequently.
Fatsat increases the SAR value of a sequence significantly because it must apply an extra 900 pulse for each slice every TR period.
The Fatsat routine is more difficult to apply at lower field strengths
simply because the fat and water resonant frequencies are closer together.
The time required to run the spatial saturation routine is longer at lower
filed strengths for two reasons. To achieve the same bandwidth the RF pulse
period must be increased by the inverse of the change in field strength,
thus a 10 msec pulse at 1.5T will need to be replaced by a 30 msec pulse
at 0.5T 2. The pulse period must be
further increased if the bandwidth is reduced to accommodate the small
absolute chemical shift. Sequence designers must deal with these conflicts,
but it is one reason why Fatsat is not commonly employed for T1 imaging
on lower field strength machines.
saturation pulse is applied to the entire imaging volume but it will only
work if the frequencies of the pulse coincide with the resonant frequency
of fat. Uniform saturation of fat signal all over the image and through
each slice depends on achieving an excellent magnetic field uniformity.
By convention the bandwidth of the Fatsat pulse is equal to the chemical
shift between water and fat, so field variations of more than 1.75 ppm
will degrade the Fatsat. It is also possible to get accidental saturation
of the water signal where the field strength varies positively. Accidental
water saturation is less likely if a narrow Fatsat pulse is used, but this
requires a longer pulse period and will result in a loss of fat saturation
at even lower levels of field variation.
VISION users can choose to include shim coils in the shimming process. The software will then attempt to optimise the shim of the local Field of View in different directions depending on which shim coils are selected. This adds to the quality of the shim as well as to the time needed to get convergence. With software B31A select only from channels A20 A21 B21 B22.(4) If the A20 (Z2) channel is altered you must repeat the frequency adjustment process manually before starting the sequence.
Shimming difficult regions is often unsuccsesful when using an array
coil or surface coil, particularly if there is fat in the near region of
the coil, or the region of interest is displaced from the iso-centre. In
these cases the body coil should be used for the shimming process. This
technique is more fully described in a notes elsewhere on this web site
(Getting Better Fatsat) Changes to the
MAP-shim software are expected in NUMARIS B33A.
Selective Excitation Routines
As was discussed earlier, a frequency selective RF pulse needs a longer time period than a broadband pulse. The binomial excitation technique uses a series of correctly phased broadband RF pulses to selectively excite water by exploiting the phase cycling effect. They achieve selective suppression in a shorter time period than spectral suppression and with lower SAR.
Binomial excitation methods are described numerically with the number
referring to the relative amplitude of the pulse. A bar printed above the
number may be used to indicate a phase reversal.
The simplest is 1:1 excitation also known as jump back excitation. In 1:1 excitation two 450 broadband pulses are applied with an inter-pulse delay (t) equal to the time taken for fat and water to get out of phase. The first pulse is applied along X axis flipping fat and water vectors 450 and aligning their transverse components with the X axis. During the inter-pulse delay the two vectors get progressively further out of phase until the fat vector is lying along the -X axis and the water vector is again along the +X axis. At this point the second pulse is applied with opposite phase to the first (-X), pushing the fat vector back to Mz (flip = 00), while the water vector is pushed to a flip of 900.
The appropriate inter-pulse delay will depend on the field strength and the MR species being considered so it is frequently expressed as the relative phase difference between the vectors expressed in radians. The time needed to achieve 1800 phase separation would be labelled . In essence the binomial excitation routine pulses are phase locked to the target species, with each successive pulse counteracting the previous pulse for that species and adding to the flip angle of the non-target or off-resonance species. In MRI binomial pulse trains are currently used to selectively suppress fat for 3D sequences 5, in spectroscopy they are typically used to eliminate the water peak. The approach can be sped up by decreasing the inter-pulse delay to/2 and applying the second pulse along the -Y axis (6), work is also aimed at applying spatially selective binomial excitation routines to replace spectral saturation . Higher order binomial excitation routines (1:2:1 or 1:3:3:1) achieve a more selective suppression but take more time because of the additional inter-pulse delays.
Binomial Excitation - Key points
The phase cycling effect is commonly exploited with T1 weighted spoiled gradient echo sequences (SGE, FLASH, SPGR) when imaging the abdomen and marrow spaces. In Phase (IP) or Opposed Phase (OOP) images can be selected by choosing an appropriate TE. Images obtained when the fat and water vectors are in-phase will have similar contrast to SE images. When the TE coincides with an opposed phase period a distinct dark line (Indian ink artefact) is noted at borders of fatty and water based tissue. This occurs where the voxel contains equal amounts of fat and water spins, achieving a complete cancellation of signal. In voxels where the mix is not 50:50 varying degrees of signal loss will occur compared to the IP image. This signal loss can be used to identify fatty infiltration of the liver, and tumour infiltration of marrow spaces.
T2* decay causes a rapid decrease in Mxy as TE extends in
GRE sequences, so it is important to image quality to use the earliest
TE possible for the OOP image. At low field strengths this is easily achieved
although the OOP time is late. At 1.5T the first OOP time occurs at 2.2
msec and partial echo techniques are often employed to implement this short
TE. While this approach sacrifices some SNR, the loss is less than waiting
until the next OOP period and suffering the T2* decay.
STIR is an acronym for Short TI Inversion Recovery. STIR sequences use an inversion pulse and time delay (TI) to null longitudinal magnetisation (Mz) so that fat will not contribute signal in any subsequent signal generation routine. STIR is most commonly explained with reference to an inversion recovery spin echo (IRSE) sequence but the approach is also applicable to TSE, GRE, GRASE TurboFLASH and EPI acquisitions.
The initial 1800 inversion pulse flips longitudinal magnetisation to the -Mz, after which longitudinal recover occurs with the MZ evolving through Mz=0 back to thermal equilibrium. After a time delay (Inversion Time TI) a 900 pulse is applied converting Mz into detectable Mxy. Contrast between tissue depends entirely on T1 differences at this stage. In order to null a tissue the TI must be chosen to coincide with the T1 curve of the tissue crossing the Mz=0 line. For the situation where TR - TI is long (more than 5 times T1) (7), this can be calculated as : TI null = ln2 x T1 or 0.69 x T1
creating STIR images the reconstruction process is only sensitive to the
magnitude of the Mz of a tissue, not whether it is positive or negative.
It is assumed that fat will have the shortest T1 in the area of interest
so while all other tissues have negative Mz at the TI needed to null fat
they are displayed with a pixel brightness proportional to their absolute
(modulus) value. This type of display is variously called Modulus (IRM)
or absolute (IRABS), and is different to the approach used with IR sequences
employing more moderate inversion times. Thus in a STIR image, tissues
with short T1 are displayed darker than tissues with long T1, which is
the opposite of what is expected in a conventional T1 weighted image. As
TE is increased T2 effects increase the contrast between tissues because
tissues with short T1 typically have relatively short T2. For this reason
STIR sequences are often said to have additive T1 and T2 contrast. Late
echo STIR images closely resemble Fatsat T2 weighted images and can be
applied interchangeably in many applications. Remember however that overall
signal levels still fall as TE is increased.
|STIR - Key Points
This method is a combination of the spectral saturation and STIR routines. It is available on Philips scanners where it is designated Spectral Inversion Recovery (SPIR)(10), and GE scanners with the designation SPECtral Inversion At Lipids (SPECIAL). (11) The idea is to apply a spectrally selective pulse to flip fat spins then, after the time interval that lets the Mz of fat reach zero, the excitation pulse is applied and the signal of the water spins give most of the signal.
Time is the major disadvantage of this fat suppression method. The routine is best applied at least once per TR, and for multi-slice sequences once per slice per TR. The best fat suppression is achieved with a 180 degree spectral inversion pulse and a time delay equal that used in normal STIR imaging ( approximately 150 msec at 1.5 T plus the finite time required for the inversion pulse) This is clearly impractical for anything but extremely long TR techniques.
In practice the "inversion" pulse used ranges between slightly more than 90 degrees but significantly less than 180 degrees the required delay time can be acceptably short. In most scanners using this method the radiographer has control of the inversion flip angle and the system determines the optimum delay time. The routine is applied in a "segmented "fashion allowing some small variation in the degree of fat suppression across slices.
SPIR will be as sensitive to local field inhomogeneity as spectral saturation
routines. It is suitable for use after Gd contrast because only the fat
spins are affected by the routine
Fat suppression by subtraction of the pre-contrast image from the contrast
images particularly appropriate in the breast where the anatomy is prone
to extreme magnetic field inhomogeneity even after patient shimming, and
where the dynamic nature of contrast uptake precludes STIR imaging with
sufficient spatial and temporal resolution.
Movement will mar subtraction images so patient comfort, immobilisation and procedure speed are very important. The MR receiver gain setting must not be altered between pre and post contrast sequences so that global signal intensity changes caused by the contrast are not neutralised.
Creating subtraction images from a dynamic string of T1 3DFLASH sequences
with a fast high dose(.166mmol/kg) injection of Gd-DTPA allows improved
lesion detection and discrimination between malignant and benign breast
lesions (12). The approach has
also been applied with 1.5 second serial scanning to investigate pelvic
Magnetisation Transfer Contrast pulses are frequently used to diminish
background signal in angiography sequences and post contrast. Fat is largely
unaffected by these pulses and so its signal increases relative to the
non-fatty tissue. This can be counterproductive in neck angiography and
in detecting enhancing lesion in or near fat. In these instances MTC is
At present the bright fat signal on T2 weighted TSE images is often noted as the most obvious difference to a T2 weighted SE image, but increasingly TSE sequences are becoming the standard method of acquiring T2 weighted images. In that event, Spin Echo becomes is a niche sequence, and the fat signal suppression due to scalar coupling effects and a lack of MTC induced contrast must be considered amongst its characteristics. SE offers mild fat signal suppression in a T2 weighted image with reduced SAR loading rates, increased sensitivity to blood breakdown products and a substantial increase in sequence time.
MRI Radiographers have access to a range of fat suppression methods,
which work by exploiting particular characteristics of the MR signals from
fat and water. Understanding the mechanisms of each method provides insight
into appropriate applications and situations where artefacts can be anticipated.
The widespread use of TSE sequences for T2 weighted imaging has increased
the situations where high fat signal may obscure lesion visibility. Controlling
fat/water contrast can enhance the sensitivity and specificity of an MRI
examination in many clinical settings, so the MR Radiographer must be familiar
with available fat suppression techniques and know when to apply them.
Paper prepared initially for the first New Zealand MR Users' Meeting,
Auckland November 1996, then expanded for the Siemens MR User meeting Sydney
|Click Me to Mail to Me
If you would like this paper presented at your MRI meeting or you have comments or suggestions, please contact the author by e-mail.