Allen D. Elster, MD

Gradient-Echo MR Imaging:Techniques and Acronyms’

I

State ofthe Art

This overview briefly traces the his-

tory and nomenclature of gradient-recalled-echo (GRE) techniques usedin magnetic resonance (MR) imaging.

GRE sequences, which are now of-fered commercially by 11 major man-

ufacturers of MR imagers, are pre-

sented to illustrate their structuralsimilarities and subtle differences inimplementation. A classfficationscheme is proposed for these se-quences employing a vendor-inde-

pendent nomenclature. A glossary ofGRE abbreviations and acronymsfound in the current MR marketplace

also is presented.

Index terms: Magnetic resonance (MR)

Magnetic resonance (MR), pulse sequences

Magnetic resonance (MR), technology ‘ State-

of-art reviews

Radiology 1993; 186:1-8

I From the Department of Radiology, Bow-

man Cray School of Medicine, Wake Forest Uni-versity, Medical Center Blvd, Winston-Salem,

NC 27157-1022. Received March 13, 1992; revi-

sion requested May 15; final revision receivedAugust 6; accepted September 10. The author issupported in part by the RSNA Research andEducation Fund as a Winthrop/RSNA Scholar.

Address reprint requests to the author.C: RSNA, 1993

T HROUGHOUT most of the 1980s,spin-echo (SE) techniques domi-

nated the daily practice of clinicalmagnetic resonance (MR) imaging. Bythe middle of the decade, however,interest began to be directed toward anew class of pulse sequences, which

used gradient echoes (GREs). WhereasSEs were produced by pairs of radio-frequency (RF) pulses, GREs wereformed by single RF pulses in conjunc-tion with gradient field reversals.

In the 1990s, GRE techniques havebecome an essential component of themodern MR examination, being of-fered as standard software by everymajor vendor. Because these sequenceshave demonstrated great clinical util-ity in so many diverse settings, a rela-tively large number of GRE variationshave been developed. Unfortunately,no uniform system of nomenclaturefor these sequences has yet beenadopted. As a consequence, almost 40different names, abbreviations, andacronyms for GRE imaging sequences

have been devised by vendors to dif-ferentiate and market their products.

In this overview the history of GREimaging will be briefly recounted. Iwill describe the major varieties ofGRE techniques in common use andpropose a unified, vendor-indepen-dent classification scheme for thesesequences. Where not restricted byconfidentiality agreements, I will alsoillustrate some of the subtle variationsin GRE sequence design unique tocertain brands of imagers, which mayresult in slight differences in their im-aging behavior. As a final contribu-tion, I have constructed a modern lex-icon for GRE terminology, which willmake possible a rapid comparison ofGRE sequences available in the cur-rent MR market (Tables 1, 2).

HISTORY OF GRE

NOMENCLATURE

Although the phenomenon of “nu-

clear induction” was first demon-

I� uIu�flIuIlI

strated by Bloch et a! (1) and Purcellet al (2) in 1946, it was not until 1950

that Hahn (3) recorded the transientMR signal after an RF pulse that wenow call the FID. Later that sameyear, Hahn reported the discovery ofa remarkable new type of MR signal,the SE, which could be generated byapplication of two successive RFpulses (4). Hahn and others also rec-

ognized that a train of three or moreRF pulses could produce a third typeof MR signal called a stimulated echo(4,5).

In 1958, Carr (6) first analyzed whathappens when a long series of closelyspaced (‘r << T2) RF pulses is applied

to a sample (Fig 1). In this scenario,FID signals will occur after each RFpulse and SEs will be produced by

successive pairs of RF pulses. Each set

of three or more RF pulses will in turnproduce stimulated echoes, which

coincide with the SEs when the RFpulses are evenly spaced and no gra-dients are applied for imaging. More-over, if the series of RF pulses is ap-plied sufficiently rapidly, the tails ofthe FIDs and SEs will merge togetherso that a continuous signal of varying

amplitude is produced, and a steady-state free precession will have becomeestablished.

During the 1950s it became increas-ingly recognized that MR signalscould not only be generated by addi-tional RF pulses but also by manipula-tions of the main magnetic field (7). In

1960, Hahn formally proposed usingmagnetic field reversals to induce MRechoes in sea water (8). Abragam, inhis classic 1961 monograph, The Prin-

Abbreviations: FID = free induction decay,FISP = fast imaging with steady-state preces-sion, FLASH = fast low-angle shot, GRE = gra-dient echo, MP-CRE = magnetization-preparedgradient echo, RF = radio frequency, SE = spinecho, 55-CRE = steady-state gradient echo,SS-SE = steady-state spin echo, TE = echo time,TR = repetition time.

Confrast-enhanced FAST (Picker)Contrast-enhanced fast field echo with Ti weighting (Philips)Contrast-enhanced fast field echo with T2 weighting (Philips)55-GRE with SE sanspbing(Elscint)Fourier-acquired steady state (Picker)Field echo (Otsuka, Picker, Philips, Toshiba)Field echo with echo time (TE) set for water/fat signals In opposition

Field even echo by reversal (Picker)Field echo with 1’E set for water and fat signals in phase (Picker)Fast field echo (Philips)Fast GRASS (GE Medical Systems)Fast Imaging with steady-state precision (Siemens)Fastbow-angle shot (Siemens)Field reversal echo (Picker)Fast scan (GE Medical Systems)SS-GRE with RD sampling (Elscint)Fast spoiled GRASS (GE Medical Systems)Gradient field echo (Hitachi)Gradient field echo compensation (Hitachi)Gradient recalled acquisition in the steady state (GE Medical Systems)Gradient echo, gradient-recalled echo

Gradient recalled echo (Resonex)Multiplanar GRASS(GE Medical Systems)Magnetization-prepared rapid GRE (Siemens)Partial flip angle (Toshiba)Partial saturation (Insfrumenthrium)Reversed FISP (Siemens)

Rapidly acquired magnethation-prepared FAST (Picker)RI-spoiled FAST (Picker)Rapid scan (Hitachi)Elseint term for any fast GRE sequenceShort minimum angle shot (Shlmadzu)Spoiled GRASS(GE Medical Systems)Steady-state free precession (GE Medical Systems, Shlmadzu, Toshiba)Small tip angle GRE (Shimadzu)Steady-state technique with refocused FID(Shiznadzu)FAST with Ti confrast (gradient-spoiled) (Picker)Turbo field echo (Philips)

Turbo field echo (Philips)Turbo version OfFLASH (Siemens)Turbo version of SHORT (Elsdnt)

2 #{149}Radiology January 1993

Table 1Com�on of GRE Pulse Sequence Names b�r Vendor

Manufacturer“Basic” GRE

(GRE)

Steady-State

(SS-GRE-FID)

Steady-State

(SS-GRE-SE)

SpOiledGRE

(SP-GRE)Magnetization

Prepared (MP-CRE)

Elacint (Hackensacic NJ) . . . P-SHORT E-SHORT SHORT TUrbO�SHORT*GE Medical Systems (?�xfiI-

waukee, Wis)Hitachi Medical (Tarrytown,

MPGR(1.5 T),PS (03 1)

GFE

GRASS

GFE

SSFP

. . .

SPCR

GFE

FSPGR-prepared,FGR-prepared

RS�N�

Instrun�entarium Imaging(Milwaukee, Wis)

. . . . . . . . . PS ...

Otsuka Electronics (Fort FE PPE . . . . . . ...

Collins, Cob)Philips Medical Systems . . FFE CE-PEE-T2 CE-FFE-T1 TFE�, FASISCAN

(Shelton, Conn)Picker International (High-

land Hts, Ohio)FE, FESUM,

FEDIF

-

FAST CE-FAST RF-FAST,Ti-FAST

RAM�FAST*

Reeonex(Sunnyvale, Calif)Shlmadzu Medical Systems

(Gardena, Calif)

GREGIOSTAGE

. . .

55W. . .

STERF. . .

STAGE...

SMASH

Siemens Medical Systems(Isel1i�, NJ)

. . . FISP PSIF FLASH TUrbOFLASH’,MP-RAGE (3D)

Toshiba America Medical PFI, FE FE . . . . . . TUthO�FE*Systems (Tustin, Cali.f)

Note.-F1D - free induction decay, SS - steady state.e Seque�� with preparatory pulses are notyet comtherdaily available.

ciples of Nuclear Magnetism (9), explic-itly described the feasibility of gener-ating an echo by field or gradientreversal. Using the analogy of runnerswho reverse their direction halfway

through a race, Abragam referred tothese gradient reversal echoes as“racetrack echoes.” Abragam hadsome doubts about the general utility

of this method of echo formation,however, stating, “the point of spoil-ing a very homogeneous field . . . [bygradient reversals] . . . may appearquestionable” (9).

During the 1960s and early 1970s,pulse sequences based on SEs largelycontinued to dominate the MR litera-ture, although research continued ongradient manipulations of the MRsignal and steady-state free preces-sion phenomena (10-13). Neverthe-

less, interest in GRE signal formationlargely floundered until 1976, whenMansfield and Maudsley (14) pro-posed their revolutionary “fast scanimaging.” This new method, a fore-runner of modern echo-planar tech-niques, used gradient reversals togenerate echoes (15). In 1976, Kin-shaw (16) also developed a steady-state imaging method, which wassubsequently implemented in twodimensions and presented in 1981 atthe Bowman Gray International Sym-posium on NMR Imaging (17,18). Bythe early 1980s, several additionaluniversity- and industry-based re-searchers had developed their ownvariations of GRE imaging (19,20). Onthe first Technicare MR i.magers, these

Table 2Acronyms Used in GRE Imaging

Acronym Explanation and Manufacturer

CE-FASTCE-FFE-T1CE-FFE-T2E-SHORTFASTFEFEDIF

PEERF�UMFFEFGRFISPFLASHFREFSF-SHORTPSPGRGFEGFECGRASSGREGREcHOMPGRMP-RAGEFF1PSPSIFRAM-FASTRP-FASTRSSHORTSMASHSPGR55FFSTAGESTERFTi-FASTTFETUrbO-FETUIbOFLASHTWbO.SHORT

RI PULSE RF PULSE El PULSE

aA

Figure 1. A steady-state free precession results when a series ofclosely spaced RF pulses are applied to a sample. The resultant sig-nab is composed of both FID (free induction decay) and SE (spinecho/stimulated echo) components.

RF -4’

Slice

Read ___.�1�__j-�----------t::L._

Phase

Signal

Figure 2. The “basic” GRE sequence. Seetext for details. Dotted gradients along slice-select (section-select) and read provide for

resonant-offset averaging.

Volume 186 #{149}Number 1 Radiology #{149}3

I�

echoes were referred to as “gradient-recalled echoes”; on the early imagersfrom Picker, they were called “fieldreversal echoes” (FREs). Because offield inhomogeneities and gradientpower limitations in these early imag-ers, however, SE techniques graduallysupplanted the initial GRE methods,providing images of markedly supe-nor quality and reproducibility.

By 1983, no official nomenclaturehad yet been adopted for GRE tech-niques. In the first Glossary of NMRTerms (21) published by the AmericanCollege of Radiology, the terms “gra-dient echo” and “field echo” (whichwere both in common usage at thetime) do not even appear. Only themere existence of these techniques

was acknowledged briefly under thedefinitions of “rephasing gradient”

and “spin echo,” where they werereferred to as “time reversal echoes.”

Shortly thereafter, interest in GREimaging was rekindled by a group of

German researchers under the direc-tion of A. Haase and J. Frahm, whoproposed the FLASH (fast low-angleshot) technique (22). FLASH was

unique compared to previous GREmethods in two major respects: (a) Its

gradient structure was specificallydesigned to produce spoiling or dis-ruption of transverse coherences,thereby allowing spin-density- or Ti-weighted images to be obtained, and

(b) it used RF flip angles of less than90#{176},reducing saturation effects and

ii providing another variable with

which to manipulate image contrast.Shortly thereafter, several othergroups of investigators proposed GREtechniques that used low flip anglepulses but whose gradient structurepreserved transverse coherences(FAST, Fourier-acquired steady statetechnique; FISP, fast imaging withsteady-state precession; and GRASS,gradient-recalled acquisition in thesteady state) (23-29). Confusion im-mediately began to arise, however,since the structures of these se-quences were under constant revi-

sion, often being modified substan-

tially while their original names wereretained. Perhaps the best example ofthis phenomenon is the evolutionarychanges experienced by the FLASHand FISP sequences, whose currentcommercial implementations as Sie-mens products only remotely resem-ble their initial descriptions in the sci-entific literature from the mid-i980s.

Since i986, we have witnessed anexplosive growth in the number andcomplexity of GRE pulse sequences.Today, the major manufacturers ofMR imagers offer nearly 40 GRE vari-ants, each with a different name orclever acronym selected for marketingpurposes. I will now attempt to de-scribe how each of these GRE Se-quences works, categorizing theminto functionally similar groups and

proposing a vendor-independent no-menclature. Because it is not possibleto provide a comprehensive mathe-matical analysis of these sequences inthe limited space available, the inter-ested reader is referred to several cx-cellent technical reviews that haverecently been published (30-36).

THE “BASIC” GRE SEQUENCE

The most basic of all GRE tech-

niques offered by commercial vendors

today has a structure similar to that ofa conventional SE sequence, except

that the 180#{176}refocusing pulse is miss-ing. As the timing diagram (Fig 2) forsuch a sequence illustrates, a single

RF pulse (with arbitrary tip angle a) isfirst applied to a section (which hasbeen selected by simultaneous activa-tion of the section-selection [alsoknown as “slice-selection”] gradient).Dephasing of spins into a frequency-dependent spatial pattern is providedby the first negative (downward)-going lobe of the read gradient inFigure 2. Subsequently, the positive-

(upward)-going lobe of the same gra-dient reverses this dispersion of spins,resulting in their refocusing into a

GRE. Spatial encoding by phase isprovided by applying differentstrengths of phase-encode gradient

during each RF cycle. This “basic”GRE sequence is identical in structureto the original FLASH sequence pro-posed by Haase et al (22). The reader

should note, however, that the“FLASH” sequence offered as a prod-

uct by Siemens has undergone severalmodifications and refinements andthus can no longer be classified as a“basic” GRE technique.

The fundamental differences be-tween GRE and SE imaging are ad-dressed in numerous textbooks and

review articles (31,32,37-40). Gradientrefocusing of the MR signal corrects

only for phase shifts induced by theaction of the gradient itself. Specifi-

cally, phase shifts resulting from fieldinhomogeneities, static tissue suscep-tibility gradients, and chemical shiftsare not canceled at the center of theGRE as they are in the ideal SE exper-iment. The transverse decay of the

GRE signal is therefore determined by

the effective spin-spin relaxation time(T2*), which is a reflection of both“true” T2 and inhomogeneity effects.Unwanted phase dispersions bymeans of T2* processes may be mini-mized by using small voxels and short

TEs (4i).Commercial implementations of the

“basic” GRE sequence are availableon most imagers and are offered un-der familiar names such as MPGR, FE,and PFI (Table i). The relative sim-plicity of this sequence makes it easilyadapted for such features as cardiacgating, gradient-moment nulling, andthree-dimensional Fourier transform(3DFT) acquisition. Most manufactur-ers who offer this basic GRE sequencealso permit the user to adjust the RFflip angle (a). Manipulation of the RFflip angle allows one to obtain uniqueimage contrasts, as well as to alter thelevel of equilibrium magnetization.

The potential advantages of partialflip angle MR imaging have been

M

z

M�

y

Figure 3. A small flip angle pulse (a) may

generate an appreciable transverse componentof magnetization (M5) while disturbing thelongitudinal component (Mr) only slightly.

C,C A

PC-

�B�#{176}#{176}

a. b.

Figure 4. Resonant offset effects. (a) Spins A, B, and C have different resonant offsets (�3).(b) The effect of an RF pulse (a) depends on the initial resonant offsets. A’, B’, and C’ are thefinal positions of A, B, and C following this pulse.

4 #{149}Radiology January 1993

thoroughly described in numerous

technical and clinical reports (37,38,42). Because of fundamental trigono-metric properties, a small flip angle

(a) pulse may create an appreciabletransverse magnetization while dis-

turbing the longitudinal magnetiza-tion only slightly (Fig 3). For example,a 15#{176}RF pulse acting on a magnetiza-tion M initially aligned in the z direc-tion creates a transverse componentof size (sin 15#{176})x M, or 0.26M. Mean-while, the longitudinal component

along the z axis has barely been dis-turbed, reduced only 3% to (cos 15#{176})xM, or 0.97M. The fact that the long-

itudinal magnetization has beenlargely preserved means that little

saturation has occurred; for short TRsequences, a significantly stronger MRsignal may thus be obtained by usingsmall flip angle (rather than 90#{176})RFpulses.

At first it may seem paradoxical thatreducing the flip angle (and hence thefraction of longitudinal magnetizationdeflected into the transverse plane)would result in a stronger MR signalthan could be obtained by using 90#{176}pulses. One must realize, however,

that the magnetization tipped to cre-ate the MR signal arises from thesteady-state longitudinal magnetiza-tion (Mi) that exists immediately be-fore each RF pulse. The signal ob-tamed from a fast GRE sequence thusrepresents a “balancing act” betweenfactors that maintain the steady-statelongitudinal magnetization and thosethat increase the fraction of magneti-

zation that is tipped into the trans-verse plane. In general, therefore, the

MR signal in the “basic” GRE sequencewill not be maximized at a = 90#{176}but

at a value known as the Ernst angle,which depends on the ratio TR/T1(where TR = repetition time) and istypically much smaller than 90#{176}(il,12).

So far, our simplified analysis of theaction of the RF pulse (Fig 3) has as-sumed that the steady-state magneti-zation M has no net transverse com-ponent (ie, M is aligned with the zaxis at the time of the RF pulse). Boththe basic GRE sequence and the

spoiled GRE sequence (discussed sub-sequently) generally satisfy this as-sumption and can be classified as inco-herent steady-state techniques. Theterm “incoherent” implies that only a

longitudinal steady state has beenestablished (ie, the transverse compo-nents are dispersed in phase and are

incoherent).

Under special conditions, however,a coherent steady state may becomeestablished, in which both the longi-tudinal and transverse components

reach a steady state. In this situation,the steady-state magnetization M isnot strictly aligned with the z axis butalso has a projection in the transverseplane. Pulse sequences that generatethis coherent steady-state signal willbe called SS-GRE sequences and arediscussed in greater detail below. Aswe will describe, special structuring ofthe gradient waveforms must be per-formed in order to maintain this co-herent steady state. Furthermore, thesignal behavior, image contrast, andoptimal flip angle for these sequenceswill depend not only on TR/T1, butalso on T2 and a new variable (�3), theangle through which the spin pre-

cesses in the transverse plane be-

tween two consecutive RF pulses. Theangle 13 is called the resonant offsetangle, phase angle, or precession an-gle (Fig 4a). The value of �3 will typi-cally vary with position and dependon multiple local factors including thenet effect of imaging gradients, staticfield inhomogeneities, and the trans-mitted phase of the RF pulse (35). Fig-ure 4b illustrates graphically how ana#{176}RF pulse affects a given spin de-pending on the resonant offset angleof that spin.

As long as our prototype “basic”

GRE sequence is operated in a multi-section mode with relatively long TRvalues (providing intrinsic “spoiling”or disruption of transverse coher-ences), resonant offset effects prove tobe of little concern. If operated in asingle-section mode with short TRvalues (eg, TR < 150 msec), however,

an uncontrolled partial steady-statefree precession may become estab-lished (27,30). Position-dependentclustering of resonant offset angles

may then occur. If such a position-dependent distribution of resonantoffset angles exists across a section atthe end of a cycle, then the next RFpulse will have an effect on this signalthat is also position-dependent (43).As a result, bands of varying signal

intensity (“FLASH bands”) may ap-pear, which significantly degrade thefinal image.

The practical method to reducethese resonant-offset artifacts is toextend the duration of the read gradi-

ent following echo collection (dottedlines, Fig 2). The read gradient is lefton long enough to ensure that a fullrange of resonant offset angles (from0#{176}to 360#{176})exists across each voxel(44). This uniform integration of sig-nab prior to the next RF pulse meansthat contribution to the steady-statemagnetization will be averaged orsmoothed over a full range of reso-

nant offsets. Accordingly, position-dependent variations in phase and

signal intensity will be minimized.Haacke and Tkach (34) have recom-mended that the term “resonant off-set averaged steady state (ROAST)”

be applied to such methods. While Iprefer to avoid the adoption of yet

another acronym, the phrase “reso-nant offset averaging” is a very goodone and is appropriately descriptiveof this process. Prolongation of thereadout gradient to accomplish reso-nant offset averaging is apparentlyused by all manufacturers who offerthe basic GRE sequence as an option(eg, Picker’s FE sequences). As a mi-

a

ARF

Slice

Read �nders

Phase�(i,

Signal

Figure 5. Steady-state GRE sequence withFID sampling (55-CRE-FID). See text for de-tails.

RF

Slice

Read

Phase

Signal

Figure 6. Steady-state GRE sequence with

SE sampling (SS-CRE-SE). See text for details.

Volume i86 #{149}Number 1 Radiology #{149}5

nor variation, some manufacturerswill also apply a second constant de-phasing gradient along the section-select axis to provide resonant offsetaveraging within the plane of the sec-tion (eg, GE Medical Systems’ MPGRsequence).

In practice, the basic GRE sequenceis operated in the multisection mode

with values of TR sufficiently long sothat transverse coherences are effec-tively averaged out or disrupted(spoiled). When used under theseconditions, the basic GRE sequencewill display image contrast similarto that of the spoiled GRE sequencediscussed below. For this reason, not

all manufacturers offer the basicGRE sequence we have described, butrather provide gradient- or RF-spoiled GRE techniques, which canalso be used in long TR, multisection

applications. No matter whichmethod is employed, the equilibriumsignal obtained will be proportionalto spin density, modified by the ef-fects of T2* relaxation and Ti satura-tion. Increased T2* contrast sensitivityis obtained principally by lengthening

TE. Ti saturation effects are regulatedby flip angle and TR (45).

STEADY-STATE (55)-GRESEQUENCES

SS-GRE sequences are designed to

operate in either single-section or

�wzaders’

.

multisection modes by using ex-

tremely short TR values (eg, TR < 50

msec) and to produce a coherentsteady-state free precession. Thissteady state is controlled and main-tamed in two ways: (a) through theuse of resonant offset averaging onthe read and section-select gradients(as in the basic GRE sequence) and

(b) through the application of re-

winder gradients on the phase-en-

code axis, discussed in detail below.Recall that the steady-state signal

can be considered to be the sum oftwo components: an FID occurringearly in the cycle (just after each RFpulse) and a stimulated echo/SE oc-

curring late in the cycle (just beforethe next RF pulse). In theory it should

be possible to recover both signalssimultaneously and coherently, pro-

vided that the gradient profiles areperfectly balanced (the so-called origi-nal or true FISP experiment) (29). Inpractice, however, unpredictable

phase errors make this perfect balancedifficult to achieve (46,47). Accord-ingly, all steady-state sequences com-mercially available today sample (byGRE formation) either the FID or thestimulated echo/SE components, butnot both. It should be noted, how-ever, that Siemens has developed amethod to extract both signals simul-taneously by collecting pairs ofechoes acquired with and withoutphase alternation of the RF (48); thus,it is likely that “true FISP” will beavailable as a commercial product in

the near future (Laub G, personalcommunication).

SS-GRE Sequences withFID Sampling

Coherent steady-state sequencesthat sample the FID component areamong the most widely used of all

GRE techniques and are implementedin a nearly identical fashion by everymanufacturer. Familiar names for

these sequences include GRASS, FISP,FAST, and FFE (Table 2). The struc-

ture of these sequences is similar tothat of the basic GRE technique previ-

a ously described. A prototype 55-GRE--A---- FID sequence modeled after Picker’s

FAST sequence is shown in Figure 5.�1#{149}J�L_ Note that the imaging gradients have

been purposely left unbalanced alongboth the section-select and read axesto produce resonant offset averaging(and hence artifact reduction). Alsoobserve that the phase-encode gradi-ent pulses have been applied twiceper cycle, the second time with re-versed polarity. These second phase-

encode gradient steps are known as

rewinders, and their purpose is to

ensure stability of the phase of theMR signal in each repetition intervaland to aid the development of coher-ent transverse magnetization (32).Without these rewinders, resonantoffsets would vary from cycle to cycle(since the phase-encode step chang-es). Phase-encoded information in

one cycle could spill over into thenext cycle, generating unwantedstimulated echoes or interference(FLASH) bands across the image.

Image contrast for a small flip angle

55-GRE-FID sequence exhibits thesame proportionality to spin-density

and T2* effects as the basic GRE se-quence. However, because steady-

state longitudinal and transverse

components of magnetization exist atthe end of each cycle, repetitive RFpulses cause a “mixing” or exchange

of magnetization among the compo-nents (49). Mathematical analysis ofthese effects predicts that image con-

trast will also depend in a compli-cated manner on flip angle, Ti, T2,

and T2/Ti, as well as RF phase rela-tionships and the net effect of imag-ing gradients (50-54). As the flip angleis increased, signal behavior and im-

age contrast are primarily determinedby T2/Ti.

SS-GRE Sequences withSE Sampling

GRE sequences that sample the SE

(and stimulated echo) components of

the coherent steady-state signal havea pulse timing diagram (Fig 6) that is amirror image of the SS-GRE FID se-

quence. This “time reversal” of gradi-ent applications is reflected in theacronym Siemens uses for this se-

quence, PSIF (which is just a reversalof the letters in FISP). GE MedicalSystems calls their version of this se-quence SSFP (which could really ap-ply to any short TR sequence). Elscintrefers to their sequence as E-SHORT;

the “E” presumably refers to record-ing of the spin/stimulated echo com-ponent of the steady-state signal. In

our vendor-independent nomencla-ture, we will refer to this sequence asSS-GRE-SE.

Because the echo appears to occurbefore the RF pulse in the timing dia-

gram, it is not immediately apparenthow an echo is created at all by this

unusual sequence. Indeed, if the se-

quence were run for only one cycle,no echo would be recorded. Whenone studies phase relationships overtwo cycles, however, an echo willbe produced from magnetizationbrought down into the transverse

a

A (+1- phase change)

Read �

Phase

Signal

Figure 7. Spoiled GRE sequence. See text

for details.

RF -v

Figure 8. Magnetization prepared GRE Se-

quence (MP-CRE). See text for details.

6 #{149}Radiology January 1993

plane by an RF pulse in the precedingcycle. The effective echo time is thus

TR plus TE, since an entire additionalcycle (of length TR) has passed priorto echo collection. This relatively longevolutionary period before echo col-lection allows for natural transversedecay of the magnetization to occur.

Images from the SS-SE-FID sequence

thus appear to have a dominant “T2-like” weighting. Because T2 contrast

is apparently “enhanced” by this

technique, the acronym “contrast-

enhanced FAST,” or CE-FAST, wasthe original name given to this se-quence (55), and it has been retainedby Picker for the name of their commer-cial version. The “contrast-enhanced”concept of T2 weighting is also re-flected in the acronym used byPhilips: CE-FFE-T2.

SPOILED GRE TECHNIQUES

The term “spoiling” refers to thepurposeful disruption of transversecoherences that persist from cycle to

cycle (32). All MR manufacturers now

offer some form of spoiled-GRE (SP-GRE) sequence as part of their com-mercial packages (Table i). Familiarnames include FLASH, SPGR, RF-

FAST, and Ti-FAST.

While the final images produced bythese sequences may appear nearly

identical, several different spoilingmethods (49) are used by the various

MR vendors (Fig 7). Siemens’ FLASHtechnique involves the application ofvariable spoiler gradients along thesection-select axis. The amplitude ofthese spoilers is varied linearly fromview to view in the Siemens product,

whereas some other manufacturersuse a gradient “look-up” table con-taming an array of optimized values.

In 1989, GE Medical Systems ap-

plied a different spoiling strategywith the release of its SPGR sequence.In this technique, the phase of the RF

carrier is semirandomly changed fromview to view, effectively preventingthe buildup of transverse phase co-herences. In theory at least, RF spoil-ing is superior to variable gradient

amplitude spoiling in that it is spa-

tially invariant and does not generateeddy currents.

Picker has also implemented an RF

spoiling technique on their new line

of MR systems, called RF-FAST. Thistechnique is very similar to SPGR, butit uses a different algorithm for select-

ing RF phase offsets. Picker also re-tains a second type of spoiled GREsequence on many of its lower-field-strength systems called Ti-FAST,

which is a gradient-spoiled technique

similar to Siemens’ FLASH.Elscint, Instrumentarium, and Sie-

mens have developed their own tech-niques for RF spoiling, which can be

implemented on imagers that do nothave digitally controlled RF amplifiersubsystems. Although representatives

from each company have been reluc-tant to reveal the details of their

methods, the RF spoiling is appar-

ently achieved following the methodof Zur et al (49) by allowing semiran-dom phase shifts of the RF oscillatorto accumulate between views, thuspreventing the buildup of transverse

coherences.

RAPID (PREPARED) GRESEQUENCES

If an SS-GRE sequence is run withvery short TR values (ie, TR � T2*),

neither a longitudinal nor a trans-verse steady state has enough time tobecome fully established during the

course of an imaging experiment (36).Because image acquisition has nottaken place under steady-state condi-tions, nonuniform weighting of thedata along the phase-encoded axeswill occur. Loss of image resolution

along this direction results. Further-

more, because the TR values are soshort, small flip angles must be usedto minimize saturation and to pre-

serve the signal-to-noise ratio. As aresult, image contrast in these se-

quences is dominated by spin-densityeffects and is thus relatively poor.Rapid GRE sequences typically dem-

onstrate inadequate contrast betweensoft tissues of similar composition and

are thus not very sensitive in the de-tection of pathologic abnormalities. Inspite of these potential drawbacks,the overwhelming benefit of imagingspeed has made the rapid GRE se-quence a viable technique.

The simplest form of rapid GRE

sequence is merely an 55-GRE-FIDtechnique run with very short TR val-

ues. This type of sequence is suitablefor breath-hold abdominal studiesand for dynamic contrast material-enhanced studies performed to mea-sure perfusion (56). Several vendorsoffer this simplest form of rapid GREsequence (eg, Picker’s RAM-FAST,Philips’ Turbo FFE, Siemens’ Turbo-FLASH).

Although rapid GRE sequencesgenerally produce images with rela-tively poor tissue contrasts, interest-ing Ti or T2 contrast behavior may berestored by applying a preparatorypulse (or pulses) in the interval before

rewiaders’

Qi .-.

“hardl8O” a -a

A �A A

Slice �‘spoiler

Read

Phase

Signal

Preparatory Period A :��

�/J�

data collection is begun. I will refer tosequences modified in this fashion asmagnetization-prepared GRE (MP-GRE)

techniques. Both Siemens and GE

Medical Systems currently offer MP-

GRE sequences as commercially avail-able products, marketed under thefamiliar names of MP-RAGE and fast-SPGR-prepared, respectively.

The simplest preparatory pulse issimply a nonselective (“hard”) i80#{176}pulse, which inverts the tissue mag-netization across the sample (57,58).After an inversion time delay, a rapid

SS-GRE-FID sequence is performed,as illustrated in Figure 8. Image con-trast is determined by the effectiveinversion time delay for this se-quence, which is the time interval be-tween the 180#{176}pulse and the centralphase-encoding step. Because the lon-gitudinal magnetization (and henceTi contrast) may be changing duringthe MP-GRE sequence, it is poten-

tially important to be able to havecontrol over the ordering of the

phase-encode steps. Final image con-trast will depend on the precise orderin which the phase-encode lines havebeen sampled (56). Additionally, ifrapid Ti relaxation occurs during dataacquisition, segmentation of the totalsequence into several steps, includingwaiting periods, may be necessary.

By changing the preparatory periodto 90#{176}/i80#{176}/-90#{176}set of pulses, T2 con-trast can be obtained (59). This is the

Volume 186 #{149}Number 1 Radiology #{149}7

so-called driven equilibrium versionof the sequence. Other preparatoryperiod variations are possible includ-ing schemes to produce stimulatedechoes, magnetization transfer effects,chemical shift effects, and diffusion

sensitization (60,61). An exuberantgrowth in MP-GRE sequences overthe next few years is anticipated, with

many new variations and namesforthcoming. Whatever distinctionsonce existed between rapid GRE andecho-planar techniques will likely

continue to fade away as new fastimaging strategies are developed thatcombine features of both approaches.

CONCLUSION

The large number of GRE se-

quences and acronyms can be madeless confusing if one groups them intofunctional categories based on theirgeneral structure. This review pro-vides a simplified nomenclature thatis vendor-independent and adequatelycategorizes all GRE sequences corn-rnercially available on clinical MR im-agers at the present time. I encourageMR vendors to adopt more uniform

and descriptive terms for their GREsequences of the future and, when

given a choice, to select an accuratescientific designation for their sequenceinstead of an enigmatic acronym. #{149}

Acknowledgments: I express sincere apprecia-tion to Beth Hales for her assistance in the prep-aration of this manuscript and to Dick Moranfor sharing his perspective on the “early days”of NMR. I also thank Drs David Sherrill (GEMedical Systems), Scott Hinks (GE Medical Sys-tems), Linda Eastwood (Picker), John Steidley(Philips), Roger Radke (Siemens), SheldonSchaffer (Hitachi), Z. Goldman (Elscint), DonKarle (Shimadzu), Len Platt (Otsuka), and R.Halbach (Instrumentarium) for explaining thesubtleties of GRE implementations on theirproducts. Because of the continued rapidchanges in this field, the reader should consultwith the manufacturers directly for more de-tailed descriptions of individual pulse se-quences that become available after this articlegoes to press.

References1. Bboch F, Hansen WW, Packard M. The

nuclear induction experiment. Phys Rev1946; 70:474-485.

2. Purcell EM, Torrey HC, Pound RV. Reso-nance absorption by nuclear magnetic mo-ments in a solid. Phys Rev 1946; 69:37-38.

3. Hahn EL. Nuclear induction. Phys Rev1950; 77:297-298.

4. Hahn EL. Spin echoes. Phys Rev 1950;80:580-594.

5. Woessner DE. Effects of diffusion in flu-clear magnetic resonance spin-echo experi-ments. J Chem Phys 1961; 34:2057-2061.

6. Carr HY. Steady-state free precession innuclear magnetic resonance. Phys Rev1958; 112:1693-1701.

7. Anderson AG, Garwin RL, Hahn EL, Hor-ton JW, Tucker GL, Walker RM. Spin

echo serial storage memory. J Appi Phys1955; 26:1324-1338.

8. Hahn EL. Detection of sea-water motionby nuclear precession. J Geophys Res 1960;65:776-777.

9. Abragam A. The principles of nuclearmagnetism. Oxford, England: Clarendon,1961; 62-63.

10. Stejskal EO, TannerJE. Spin diffusionmeasurements: spin echoes in the presenceof a time-dependent field gradient. J ChemPhys 1965; 42:288-292.

11. Ernst RR, Anderson WA. Application ofFourier transform spectroscopy to mag-netic resonance. Rev Sd Instrum 1966; 37:93-101.

12. Kaiser R, Bartholdi E, Ernst RR. Diffusionand field-gradient effects in NMR Fourierspectroscopy. J Chem Phys 1974; 60:2966-2979.

13. Freeman R, Hill HDW. Phase and inten-sity anomalies in Fourier transform NMR.Magn Reson 1971; 4:366-383.

14. Mansfield F, Maudsley AA. Planar and

line-scan spin imaging by NMR. ProcXIXth Congress Ampere, Heidelberg 1976;247-252.

15. Mansfield P. Pykett IL. Biological andmedical imaging by NMR. J Magn Reson1978; 29:355-373.

16. Hinshaw WS. Image formation by nu-clear resonance: the sensitive pointmethod. J Appl Phys 1976; 47:3709-3721.

17. NewhouseJH, Pykett IL, Brady TJ, et al.NMR scanning of the abdomen: prelimi-nary results in small animals. In: WitcofskiRL, Karstaedt N, Partain CL, eds. NMR im-aging: proceedings of an internationalsymposium on nuclear magnetic resonanceimaging, October 1-3, 1981. Winston-Salem, NC: Bowman Gray School of Medi-cine, 1982; 121-124.

18. Buonanno FS, Pykett IL, Vielma TJ, et al.Proton NMR imaging of normal and abnor-mal brain: experimental and clinical obser-vations. In: Witcofski RL, Karstaedt N, Par-tam CL, eds. NMR imaging: proceedings ofan international symposium on nuclearmagnetic resonance imaging, October 1-3,1981. Winston-Salem, NC: Bowman GraySchool of Medicine, 1982; 147-157.

19. Edelstein WA, Hutchinson JMS,Johnson G,Redpath TW. Spin warp NMR imagingand application to whole body imaging.Phys Med Biol 1980; 25:751-756.

20. Young IR, Bailes DR, Collins AG, Gilder-dale DJ. Image options in NMR. In: Wit-cofski RL, Karstaedt N, Partain CL, eds.NMR Imaging: proceedings of an interna-tional symposium on nuclear magnetic res-onance imaging, October 1-3, 1981. Wins-ton-Salem, NC: Bowman Gray School ofMedicine, 1982; 93-100.

21. Axel L, Margulis AR, Meaney TF, eds.

Glossary of NMR terms. Chicago: AmericanCollege of Radiology, 1983; 20-24.

22. Haase A, FrahmJ, Matthaei D, H#{228}nickeW,Merboldt KD. FLASH imaging: rapidNMR imaging using low flip angle pulses. JMagn Reson 1986; 67:258-266.

23. Kumar NG, Karstaedt N, Moran PR. Fastscan 2DFT imaging at 0.15T by field echoesand small pulse angle excitations (abstr).Magn Reson Imaging 1986; 4:108-109.

24. UtZJA, Herfkens RD. Glover G, Peic N.Three second clinical NMR images using agradient recall acquisition in a steady-statemode (GRASS) (abstr). Magn Reson Imag-ing 1986; 4:106.

25. Young IR, Payne JA, Collins AG, BydderGM. MRI: the development of T2 contrastwith rapid field echo sequences. Magn Re-son Med 1987; 4:333-340.

26. Bydder GM, Young IR. Clinical use of the

partial saturation and saturation recoverysequences in MR imaging. J Comput AssistTomogr 1985; 9:1020-1032.

27. FrahmJ, Hanicke W, Merboldt KD.Transverse coherence in rapid FLASHNMR imaging. J Magn Reson 1987; 27:307-314.

28. Van der Meulen P. Groen JP, Cuppen JJM.Very fast MR imaging by field echoes andsmall angle excitation. Magn Reson Imag-ing 1985; 3:297-299.

29. Oppelt A, Graumann R, Barfus H, et al.FISP: a new fast MRJ sequence. Electro-medica 1986; 3:15-18.

30. Patz S. Steady-state free precession: anoverview of basic concepts and applica-tions. Adv Magn Reson Imaging 1989;1:73-102.

31. FrahmJ, Gyngell ML, Hanicke W. Rapidscan techniques. In: Stark DD, Bradley WGJr. eds. Magnetic resonance imaging. 2nded. St Louis: Mosby-Year Book, 1992; 165-203.

32. Wehrli FW. Fast-scan magnetic reso-nance: principles and applications. MagnReson Q 1990; 6:165-236.

33. Haacke EM, Frahm J. Editorial: a guide tounderstanding key aspects of fast gradient-echo imaging. JMRI 1991; 1:621-624.

34. Haacke EM, Tkach JA. Fast MR imaging:techniques and clinical applications. AJR1990; 155:951-964.

35. Haacke EM, Wielopolski PA, Tkach JA. A

comprehensive technical review of shortTR fast magnetic resonance imaging. RevMagn Reson Med 1991; 3:53-169.

36. Cohen MS. Weisskoff RM. Ultra-fast im-aging. Magn Reson imaging 1991; 9:1-37.

37. Mills TC, Ortendahi DA, Hylton NM,Crooks LE, Carlson JW, Kaufman L. Par-tial flip angle MR imaging. Radiology 1987;162:531-539.

38. Winkler ML, Ortendahl DA, Mills TC, et al.Characteristics of partial flip angle and gra-dient reversal MR imaging. Radiology1988; 166:17-26.

39. Tkach JA, Haacke EM. A comparison ofspin echo and gradient echo sequences.

Magn Reson imaging 1988; 6:373-389.40. Posse S, Aue WP. Susceptibility artifacts

in spin-echo and gradient-echo imaging.Magn Reson 1990; 88:473-492.

41. Haacke EM, Tkach JA, Parrish TB. Reduc-ing T2* dephasing in gradient field-echoimaging. Radiology 1989; 166:266-270.

42. Buxton RB, Edelman RR, Rosen BR, Wis-mer GL, Brady TJ. Contrast in rapid MRimaging: Ti- and T2-weighted imaging.Comput Assist Tomogr 1987; 11:1-16.

43. Hawkes RC, Patz S. Some factors whichinfluence the steady state in steady-statefree precession. Magn Reson Imaging 1989;5:117-127.

44. Van der Meulen P, GroenJP, Tinus AMC,Bruntink G. Fast field echo imaging: anoverview and contrast calculations. MagnReson Imaging 1988; 6:355-368.

45. Hendrick RE, Kneeland JB, Stark DD.Maximizing signal-to-noise and contrast-to-noise ratios in FLASH imaging. MagnReson Imaging 1987; 5:117-127.

46. Zur Y, Wood ML, Neuringer U. Motion-insensitive steady-state free precession im-aging. Magn Reson Med 1990; 3:140-145.

47. Haacke EM, Wielopolski PA, TkachJA,Modic MT Steady-state free precession inthe presence of motion: an application tocerebrospinal fluid. Radiology 1991; 175:545-552.

48. Bruder H, Fischer H, Graumann R, Diem-ling M. A new steady-state imaging se-quence for simultaneous acquisition of twoMR images with clearly different contrast.Magn Reson Med 1988; 7:35-42.

8 #{149}Radiology January 1993

49. Zur Y, Wood ML, Neuringer U. Spoilingof transverse magnetization in steady-statesequences. Magn Reson Med 1991; 21 :251-263.

50. Sekihara K. Steady-state magnetizationsin rapid NMR imaging using small flip an-gles and short repetition intervals. IEEETrans Med Imaging 1987; 6:157-164.

51. Zur Y, Stokar 5, Bendel P. An analysis offast imaging sequences with steady-statetransverse magnetization refocusing. MagnReson Med 1988; 6:175-193.

52. Patz S. Some factors that influence thesteady state in steady-state free precession.Magn Reson Imaging 1988; 6:405-413.

53. Buxton RB, Fisel CR, Chien D, Brady TJ.Signal intensity in fast NMR imaging withshort repetition times. J Magn Reson 1989;83:576-585.

54. Gyngell ML. The steady-state signals in

short-repetition-time sequences. J MagnReson 1989; 81:474-483.

55. Gyngell ML. The application of steady-state free precession in rapid 2DFT NMR

imaging: FAST and CE-FAST sequences.Magn Reson Imaging 1988; 6:415-419.

56. Chien D, Edelman RR. Ultrafast imagingusing gradient echoes. Magn Reson Q1991; 1:31-56.

57. Haase A, Matthaei W, Bartkowski R,Duhmke E, Leibfritz D. Inversion-recov-ery snapshot FLASH MR imaging. J Corn-put Assist Tomogr 1989; 13:1036-1039.

58. MuglerJP III, Brookeman JR. Three-di-mensional magnetization-prepared rapidgradient-echo imaging. Magn Reson Med1990; 15:152-157.

59. van Uijen CMJ, den BoefJH. Driven equi-librium radiofrequency pulses in NMR im-aging. Magn Reson Med 1984; 1:502-513.

60. Haase A. Snapshot FLASH MRI: applica-tions to Ti, T2, and chemical-shift imaging.Magri Reson Med 1990; 13:77-89.

61. Jones RA, Southon TE. A magnetizationtransfer preparation scheme for snapshotFLASH imaging. Magn Reson Med 1991;19:483-488.