MAGNETIC RESONANCE IN MEDICINE

27, 107-1 17 ( 1992)

A Novel Method for Fat Suppression in RARE Sequences N. HIGUCHI,K. HIRAMATSU, AND R.

v. MULKERN*

Department of Radiology, Keio University Hospital: Tokyo, Japan; and * Department of Radiology, Children's Hospital, Boston, Massuchusem 021 15 Received August 27, 1991; revised November 4, 1991; accepted November 5, 1991 Rapid acquisition relaxation-enhanced (RARE) sequences (Hennig d al., Magn. Reson. Med. 3, 823 ( 1986)) utilize one or several Cam-Purcell-Meiboom-Gill (CPMG) echo trains to sample a number of k-space lines each repetition time TR. The technique can rapidly generate multislice T2-weighted images which, as a rule, are strikingly similar in contrast to conventional T,-weighted spin-echo (SE) images. An exception to this rule is the appearance of very bright signal from fat in T2-weighted RARE images as compared to conventional T2-weighted SE images. To reduce this fat signal, we introduce a time , the 90: and first 180," pulse of each echo train such that a phase angle delay, T ~between of7r/2 develops between fat and the reference (water) line at echo maxima. The technique leads to single-acquisition fat suppression without the use of frequency-selectivesaturation pulses and concomitant loss of slices per TR. A Bloch equation analysis is used to identify two major mechanisms contributing to suppression of off-resonance spins such that W T , = 7r/2. Namely, the CPMG sequence becomes a CP sequence with no self-correction properties for imperfect 180" pulses leading to enhanced signal decay, and the raw k-space data matrix become segemented into blocks alternately multiplied by f i , leading to signal dispersion following Fourier transformation. 0 1992 Academic Press, Inc. RATIONALE

The introduction of a proper delay T ~ between , the excitation pulse and refocusing pulse of conventional spin-echo imaging sequences leads to well-known cancellation between selected spectral features like fat and water ( 1 , 2). Since rapid acquisition relaxation-enhanced (RARE) imaging (3-5) uses many RF-refocused spin echoes for individual phase-encoding steps, it is of interest to determine what the effects of similar delays will have on fat and water signal intensities in RARE images. This is especially important since, with the most recently developed RARE sequences, fat remains bright in images generated from late pseudo-echo times in contrast to conventional spinecho images at equivalent echo times ( 5 ) . In this study, we experimentally and theoretically examine the effects of T~ delays on signal intensities from RARE images of chemically shifted species. We find that suppression of fat is quite feasible with a proper delay but that the mechanisms are considerably more intricate than the intravoxel fat / water cancellation associated with standard Dixon-like schemes ( 1, 2). METHODS

During RARE acquisitions ( 3 - 5 ) , horizontal lines in k-space ( 6) are collected using different echoes from Carr-Purcell-Meiboom-Gill (CPMG) echo trains ( 7, 8). In this study, the 16 echoes of each train were phase-encoded with values spanning the 107

0740-3194192 $5.00 Copynght Q 1992 by Academic Press, Inc. All rights of reproduction in any form reserved.

108

HIGUCHI, HIRAMATSU, AND MULKERN

full range of k-space but separated by Ly/16 phase-encode values, where L,, is the total number of phase-encode steps. The phase-encode k-space axis was sampled from the most negative values (early echoes) to the most positive values (late echoes). Subsequently, raw data sets used the first echoes of each train to fill the Ly/ 16 most negative phase-encode lines, the second echoes of each train to fill the next most negative set of L,/ 16 k-space lines, and so on. With this scheme, echoes midway through each echo train were encoded with the smallest magnitude phase-encode values, so that pseudo-echo times (pTEs) ranged from 102 to 170 ms for echo spacings 27, of 12 and 20 ms, respectively. A variable time delay T ~ was , placed between the 90: and first 180," pulse (Fig. 1 ) in order to generate a phase shift ( I , 2) between the reference (water) line and chemically shifted species at the echo centers. Images were acquired with and without this delay to assess chemical-shift suppression capabilities. In several cases, the resulting fat suppression was compared with that available with spectral presaturation and dephasing of the undesired resonance (CHESS pulse suppression ) . Imaging studies were performed on a GE Signa 1.5-T system (GE Medical Systems, Milwaukee, WI). In phantom and brain studies, the 16-echo sequence was repeated 12 times in order to obtain 192 X 256 image matrices. In breathold abdomen studies, 128 X 256 image matrices with a rectangular field of view (FOV) were obtained by repeating the sequence 8 times (23-s acquisition times). The reference frequency was set to the water resonance and images were reconstructed in magnitude mode. An acetone and water phantom study of the T2decay process with and without a 7,delay was performed by generating images from each of the 16 echoes of the CPMG train and measuring signal intensities as a function of echo time TE.

180

180

180

180

G

FIG.1. RARE pulse sequence with variable T~ delay between the 90" pulse and the first 180" pulse. The initial 90" pulse is displaced to the left by a time 5 , (cross-hatched pulse) along with its associated sliceselective lobes (not shown displaced in diagram) when fat is to be suppressed.

FAT SUPPRESSION IN RARE SEQUENCES

109

RESULTS

Figures 2a-2d are RARE images of a vial containing acetone and a vial containing water. The images were obtained with no delay, a T, delay of 1.2 ms, a presaturation / dephase (CHESS) interval centered on the acetone resonance, and the combination of a 1.2-ms T, delay and presaturation/dephase interval. Manual prescan of the phantom revealed a frequency separation of approximately 190 Hz between the water and acetone peaks. Figure 3a is a plot of mean signal intensities and standard deviations from ROIs within the water and acetone phantoms as a function of 7,. The signal intensity in the acetone phantom decreased to less than 1% of the zero-delay signal intensity at T, values around 1.3 ms. The signal intensity from water never deviated more than 2% throughout the T, range. Figure 3b is a plot of signal intensity vs echo time in the water phantom (diamonds) and the acetone phantom (circles), as obtained from a 16-echo CPMG data set. Closed symbols are from a data set acquired with a 1.2-ms T , delay while the open symbols were acquired with no delay. The transverse decay times T 2 ,obtained for water and acetone without a T, delay, were greater than 2000 ms. The water T2 was uninfluenced by the 1.2-ms T, delay while the acetone T 2 dropped to approximately 53 ms. Figures 4a and 4b are RARE brain images obtained without and with a 1.1-ms 7 , delay, respectively. A total of 18 slice locations were acquired in 54 s with each sequence. It is apparent that bright signal from subcutaneous and retrobulbar orbital fat is markedly suppressed with the use of a 1.1-ms T, delay. Figures 5a-5c are axial breathold acquisition RARE images of the abdomen obtained with no fat suppression scheme, a T, delay of 1.2 ms, and a chemical-shift fat suppression pulse. The 2 5 s TR accomodated 12 slice locations with the first two schemes while only 9 were available in the same TR when CHESS pulses were used. Both suppression schemes resulted in considerable loss of signal from fatty tissue, though in an inhomogeneous fashion. The T, delay suppressed fat around the kidneys somewhat better than the CHESS method which provided better fat suppression in the outer body layers. DISCUSSION

It has been previously noted that heavily T2-weighted RARE images from sequences with 15-ms echo spacings have bright fat, as opposed to their conventional spin-echo counterparts ( 5 ) . We note that in the original RARE work of Hennig et al. echo spacings were over 33 ms and bright fat was not reported ( 3 ) .The cause of the bright fat reported in ( 5 ) was suggested to be due to an effective lengthening of T2among J-coupled protons at the shorter echo spacings, an effect discussed decades ago by Allerhand ( 9 ) . To demonstrate that this effect is responsible for the bright fat from a complete theoretical analysis using known or approximate J-coupling constants, relevant CPMG echo spacings, and chemical shifts is beyond the scope of the present work. However, convincing experimental support for this contention may be found in an earlier work by Hinks and Henkelman warning against the use of organic materials as MRI phantoms ( 10). These investigators demonstrated with CPMG studies of propylene glycol ( CH3CH(OH ) C H2 0 H) how J-coupling reduces the effective T 2of 80 ms at echo spacings of 10 ms to less than 40 ms at echo spacings

110

HIGUCHI, HIRAMATSU, AND MULKERN

FIG. 2. Images of acetone (smaller vial) and water phantoms. Acquisition parameters for the 16-echo, 12-shot sequence were pTE, 96 ms; TR, 2500 ms; 24-cm FOV; 5-mm slice thickness; and 192 X 256 image matrix. ( a ) Standard RARE image; ( b ) delay time 7, = 1.2 ms; ( c) standard acquisition with chemical saturation pulse centered on acetone; ( d ) combination of chemical saturation pulse and r, = 1.2 ms. All images have been filmed with identical window settings chosen to accent the background noise (phaseencode direction is vertical)

FAT SUPPRESSION IN RARE SEQUENCES

111

FIG. 2-Conlinued

greater than 25 ms. This is precisely the range of echo spacings and J-coupling interactions germane to current RARE implementations and conventional spin-echo studies and clearly provides sufficient evidence for expecting bright fat in the former and not in the latter.

112

HIGUCHI, HIRAMATSU, AND MULKERN

.. ................

1000 -

2500

.

._ x

--

W

600-

I

e

C

-

g

x

:.

C a,

1500-

.-rn

la la 6*4*******+*** 8oo--

1000-

v,

500 --

07

****+

*** * *

+*;

..............

-C 400~~~~0~0000000000~

cn 5 200-

0 --200 1

FAT SUPPRESSION IN RARE SEQUENCES

113

FIG.4.( a ) RARE brain images ofa healthy volunteer obtained without a .r,delay. ( b ) Same slice acquired with a .rcdelay of 1.1 ms. Acquisition parameters were TR, 4000 ms; pTE, 96 ms; 5-mm slice thickness; 20-cm FOV; 192 X 256 image matrices; 18 slice locations; and an acquisition time of 54 s/sequence.

to have a flip angle close to i ~ Let. the actual flip angle be T ( 1 - S), where 6 is a dimensionless parametei representing me imperfection of the ?r pulse. To first order in 6, the rotation matrix representing the refocusing pulse is

114

HIGUCHI, HIRAMATSU, AND MULKERN

FIG. 5. RARE breathold acquisition images of the abdomen. Images were obtained with the 16-echo, eight-shot sequence with an acquisition time of 23 s. ( a ) No fat suppression scheme. ( b ) RARE with T~ delay of 1.2 ms. ( c ) RARE with chemical-shift satur;+:-- y*’n fat. A total of 12 slice locations were obtained with the first two methods while the fat suppression pulse reduced the total number of slices to 9. Acquisition parameters were TR, 2500 ms; pTE, 96 ms; and a 128 X 256 rectangular FOV (40-cmlong axis).

-

115

FAT SUPPRESSION IN RARE SEQUENCES

FIG. 5-Continued

1 @=O 0

0 -1 -6ir

0 [I1

6a. -1

Calculating all three components of the magnetization within each echo interval (to first order in 6 ) leads to the observation that the z-magnetization m grows with echo number n , according to

+

+(n

M =

7r6[sin W(T

rn

-7r6n cos( wT)sin( wTC)

T,)

-

l ) c o s ( w ~ ) s i n ( w ~ , ) ] for odd n

[2a]

and =

for even n .

[ 2b]

These equations reveal that for “on-resonance’’ spins, w = 0, there is no growth of zmagnetization with echo number n , reflecting the self-correcting properties of the Meiboom-Gill modification ( 7) to the Carr-Purcell multiecho sequence (8).However, for spins obeying the condition WT, = ir/2, the growth of z-magnetization with n is maximized, providing an efficient conduit for the return of transverse magnetization to the “unobservable” longitudinal axis, hence the apparent T2 shortening for spins with w r C = x / 2 (Fig. 3b). The analysis further reveals that thef- and g-magnetization at the center of each echo are given by

f = cos( W T , )

[3a1

g = (- 1 )“+‘sin(WT,)

[3b1

1 I6

HIGUCHI, HIRAMATSU, AND MULKERN

+

and the complex signal intensity, f ig, becomes modulated by the factor exp (iwTC(- l ) n + l ) . Thus, for spins with W T , = ~ / 2our , RARE acquisition scheme leads to k-space segmentation into blocks of L y / 16 k-space lines alternately multiplied by ki, resulting in dispersion of signal following Fourier transformation. Simulations based on this analysis demonstrate the effects of both T , shortening and dispersion. Calculated phase-encode profiles (magnitude vs transform variable Q ) of a 5-cm “box” with input T 2values of 2000 ms for on-resonance spins and 50 ms for off-resonance spins is depicted in Fig. 6. These simulations are based on expressions previously developed to model phase-encode artifacts in RARE ( 11) but are specifically for a l6-ech0, eight-shot sequence with a 15-ms echo spacing and an FOV of 17.5 cm. The well-defined box of the on-resonance spins is observed to become ghosted and substantially reduced in intensity when off-resonance ( w 7 , = ~ / 2 con) ditions apply. Note how the residual ghosting pattern in the simulation of the offresonance phantom mimics the residual signal from the acetone phantom in Fig. 2b. Finally, the method proposed for fat suppression is not insensitive to field inhomogeneities, whatever their source, as evidenced by the residual fat signal in the abdomen image (Fig. 5b). This can be anticipated since over large fields of view or areas in which magnetic susceptibility differences are prominent, a single 7,value cannot fulfill N I T , = 7r/2 for all lipid protons. As such, combining a T , shift with CHESS pulses may prove to be the most fruitful approach when maximal fat suppression is desired over increased number of slices per TR. CONCLUSIONS

Bright fat signal in T,-weighted RARE images most probably occurs due to the use of many closely spaced refocusing pulses which reduce contributions from spin-spin splittings to the T 2decay process ( 9 , 10). A simple method involving an appropriate 7, delay between the excitation pulse and first refocusing pulse serves to reduce this

0-0

WTC =

0-• W T C =

F -5 -3.200

- 1.400

0 n/2

T 0.400

2.200

FT Variable R

FIG,6. Simulations of a 5-cm “box” profile along the phase-encode direction as generated with a Bloch 2 equation analysis of a 16-echo, eight-shot R A R E sequence with and without the T~ delay for a ~ / fat-water 2 profile was placed into the simulation empirically. phase shift. The enhanced T, decay for the WT, = ~ / delay

FAT SUPPRESSION IN RARE SEQUENCES

117

fat signal in single-acquisition studies. The method does not lead to any loss in the total number of slices per TR nor does it add additional RF pulses to a sequence already associated with high RF power deposition levels. REFERENCES 1. 2. 3. 4.

5.

6. 7.

8. 9. 10. 11.

W. T. DIXON,Radiology 153, 189 (1984). E. YAMAMOTO AND H. KOHNO,Phys. Med. Biol. 3,713 ( 1986). J. HENNIG,A. NAURETH,AND H. FRIEDBURG, Mugn. Reson. Med. 3,823 (1986). R. V. MULKERN,S . T. S. WONG, C . WINALSKI,AND F. A. JOLESZ,Magn. Reson. Imuging 8, 557 (1990). P. s. MELKI,R. V. MULKERN,L. P. PANYCH,AND F. A. JOLESZ,J. Magn. Reson. Imaging 1, 319 (1991). D. B.TWIEG,Med. Phys. 10,610 (1983). S. MEIBOOMAND D. GILL,Rev. Sci. Instrum. 29,688 (1958). H. Y. CARRAND E. M. PURCELL,Phys. Rev. 94,630 ( 1954). A. ALLERHAND,J. Chem. Phys. 44, 1 (1966). R. S. HINKSANDR. M. HENKELMAN, Med. Phys. 15,61 (1988). R. V. MULKERN, P. s. MELKI,P.JAKAB,N.HIGUCHI,AND F. A. JOLESZ,Med. Phys. 18,1032 (1991).