MAGNETIC RESONANCE IN MEDICINE 26,
342-354 ( 1992)
Partial RF Echo-Planar Imaging with the FAISE Method. 11. Contrast Equivalence with Spin-Echo Sequences PHILIPPE
s. MELKI,FERENCA. JOLESZ,AND ROBERTv. MULKERN*
Departtncxt of Radiology, Brigham and worn en'.^ Hospital, Hurvard Medical School. Boston, Massachusetts; and * Department of Radiology, Children’s Hospital, Harvard Medical School, Boston, MassachusettJ Received July 10, 1991; revised October 10, 1991; accepted October 14, 1991 The fast acquisition interleaved spin-echo (FAISE) sequence and its dual-echo version (DEFAISE) are partial RF echo-planar methods which utilize a specific phase-encode reordering algorithm to manipulate T2 contrast via an operator-controlled pseudo-echo time, pTE. The repetition time, TR, between successive applications of the Carr-PurcellMeiboom-Gill (CPMG) echo trains used in FAKE may be reduced to introduce T I weighting. To quantitatively determine the extent to which FAISE T , and T2 contrast characteristics agree with spin-echo methods, signal intensities from FAISE acquisitions were compared with signal intensities from equivalent CPMG acquisitions. In phantoms and in human heads, the contrast Characteristics of FAISE are found to be highly correlated with that obtained with equivalent CPMG sequences. However, conventional SE sequences generally utilize longer echo spacings than employed with FAISE/CPMG. Thus, echo spacing-dependent mechanisms such as spin-spin coupling and magnetic susceptibility lead to some differences in contrast between conventional SE and FAISE. Finally, FAISE appears to be more sensitive to magnetization transfer effects than conventional SE sequences since more off-resonance irradiation is applied to individual slices during multislice acquisitions. Q 1992 Academic Press. Inc. INTRODUCTION
In clinical MRI, designing an imaging protocol consists in optimizing contrast between common pathological processes and normal surrounding tissues. With conventional SE methods, this optimization is performed by varying the proportion to which various mechanisms contribute to contrast by manipulating repetition times TR and echo times TE. A precise knowledge of how the sequence parameters of any technique influence the contributions from basic contrast mechanisms is essential for image contrast optimization. Partial RF echo-planar techniques, like fast acquisition interleaved spin echo (FAISE) ,have been shown to provide “spin-echo”-like images in significantly reduced acquisition times when compared to conventional SE methods ( 1 ) . This study aims to evaluate how the contrast manipulations achieved by varying the pseudo-echo time pTE, the repetition time TR, and the echo spacing 27, correlates with equivalent contrast manipulations performed with SE methods. The results support the contention that the contrast characteristics of FAISE are extremely similar to those of the slower spin-echo methods and that differences resulting from echo spacing effects are pre0740-3194/92 $5.00 Copyright 0 1992 by Academic Press. Inc. All rights of reproduction in any form reserved
342
THE FAISE METHOD: CONTRAST EQUIVALENCE
343
dictable and well understood. The inevitable T2decay-related artifacts associated with FAISE have been discussed in the accompanying work, here referred to as I. MATERIALS AND METHODS
FAISE, DEFAISE (dual-"echo" FAISE), and CPMG sequences were all implemented on 1.5-T GE SIGNA systems (General Electric Corp. Milwaukee, WI). The FAISE sequence ( I ) permits manipulation of T2contrast through operator selection of the magnitude of the phase-encode gradient used to encode the first echo of the first echo train (see IPS parameter of I). Certain IPS parameters yield minimal phaseencode ghosting noise. We restrict our attention to these values in this work. The resulting pTE, is given by an integral multiple of the echo spacing in the CPMG trains (seeEq. [ 5 ] o f l ) . For both CPMG and FAISE sequences, the phase-encode gradient lobes are applied before each echo readout and compensated following each echo readout. In fact, aside from the phase-encoding line in the pulse diagrams, the CPMG and FAKE sequences are virtually identical. In FAISE, however, the phase-encode gradient is changed from echo to echo according to the algorithm described in I. In the CPMG sequence, the phase-encode gradient is changed from echo train to echo train, allowing individual images to be constructed from each echo. In order to produce dual-contrast FAISE (DEFAISE) images, we have employed echo trains consisting of eight echoes. Each half of the echo train collects data lines for two raw data sets. As with single-contrast FAISE acquisitions, each set of four echoes are phase encoded according to the algorithm described in I. For comparative CPMG and FAISE (or DEFAISE) contrast studies, the TR interval, the echo spacing 27, and the number of echoes N,, were matched. All echoes were collected with the acquisition windows of 8.2 ms. Only one autoprescan was applied, resulting in identical receive/transmit gain levels for each acquisition. An echo spacing of 15 ms was used in most cases but was lengthened in one set of experiments. Quantitative signal intensity comparisons were performed with NiClz phantoms and from head images of normal volunteers. For the T2 relaxation studies, serial FAISE acquisitions produced a set of images associated with the full range of pTE parameters. The corresponding set of CPMG images were obtained following the FAISE acquisitions. For T I studies, the pTE of the FAISE sequence was set at 15 ms, and the TRs were varied. These images were compared with 15-ms TE images of CPMG sequences run with identical TR intervals. To evaluate how the number of echoes per CPMG train influences signal intensity, a relatively long T Iphantom was imaged with the FAISE sequence while varying the number of echoes per train for a number of different TR intervals. In addition, the signal dependence of FAISE and CPMG on echo spacing was studied in corn oil and in a coagulated blood sample. The blood sample had been drawn from a healthy volunteer and stored (unheparinized) in a 20ml syringe for 100 days prior to imaging. These samples were imaged at various pTE/ TE values with echo trains using two different echo spacings. Signal intensities from phantoms and various brain tissues were measured from region of interest placed within both types of images. Correlations between paired measurements were made from images with identical echo spacing, TR. pTE/TE, and number of echoes. From the various sets of CPMG images, T 2values of selected
344
MELKI, JOLESZ, AND MULKERN
materials were calculated from the signal intensity decay with echo time TE. Similarly, the T I values were obtained from signal intensity variations with TR interval using the first echo CPMG images (TE = 15 ms). The effect of off-resonance irradiation during multislice acquisitions was tested with a protocol described by Dixon (2). A multislice four-echo FAISE sequence with a pTE of 15 ms, a 1500-ms repetition time, and a 15-ms echo spacing was employed. The slice thickness was 5 mm with a gap of 17 mm between slice centers. The frequency offset between slices was 1.93 kHz/slice. An uncooked beef steak and a NiC12 phantom were imaged with 1, 3, 5, . . . 15, and 17 slice acquisitions, a single-slice sequence ending this experiment. The slices were excited in the order of their spacial position, and an even number of slices were excited on each side of the center slice. These serial acquisitions were repeated with a dual-echo SE sequence with identical imaging parameters. Analysis of the signal intensities was performed with the central slice image.
0
.& _
T 2 = 955 ms
100
(I)
T2 = 105 ins
w
E
z
o
T 2 = 44 m s
10
30
0
60
90
120
150
180
210
240
PTE (ms)
> 1500
0
x c
T2 = 105 m s
vl
a, C
;
-
T2 = 44 m s
1000
6
T2 = 955 rns
D C
PP
P
Ln
#
,dA
500
J
2
/*‘
0 0
500
1000
1500
CPMG Signal Intensity
FIG. 1. ( a) Signal intensity vs pseudo-echo time pTE for three NIC12 phantoms. The T2values indicated were measured from signal intensity vs echo time TE, as obtained from a 16-echo CPMG sequence. ( b ) Correlation plot of FAISE signal intensity vs CPMG signal intensity for all pTE and TE values and for all three NiCI, phantoms.
345
THE FAISE METHOD CONTRAST EQUIVALENCE RESULTS
Figure la presents a semi-log plot of signal intensity vs pTE for three different NiC12 phantoms. The measurements were made from a series of 16-echo, 16-shot FAISE sequences using a TR of 2000 ms, a 15-ms echo spacing, a 24-cm field of view (FOV), and a 10-mm slice thickness (acquisition time of 34 s ) . A 16-echo CPMG sequence was also used to image these phantoms using the same sequence parameters (acquisition time of 8 min, 56 s ) . The T 2values for each phantom, as obtained from the 16 CPMG images, were calculated to be 44, 105, and 955 ms. The “T2” values calculated from monoexponential fits of the FAISE signal intensity decays with pTE (Fig. la) were found to be 42, 105, and 973 ms, in good agreement with the CPMG-derived T2 values. Figure 1b presents a correlation plot of the FAISE signal intensities vs CPMG signal intensities for all the pTE/TE values used and for all three phantoms (see Table 1 for correlation coefficients). Figures 2a-2f depict head images of a healthy volunteer as acquired with FAISE (Fig. 2a, 2c, and 2e) and CPMG (Fig. 2b, 2d, and 2f) sequences at pTE/TE values of 45,90, and 180 ms. Both sequences utilized 16 echoes, an echo spacing of 15 ms, a TR of 2000 ms, a 5-mm slice thickness, 256 X 256 image matrices, and a 24-cm FOV. Figure 3a presents signal intensity vs pTE for ROIs in gray matter, white matter, cerebrospinal fluid (CSF), and orbital fat. Figure 3b is the correlation plot of signal intensities measured from both CPMG and FAISE acquisitions for all pTE/TE values and ail four tissues. Comparable results were obtained from the DEFAISE sequence and an eight-echo CPMG sequence, as applied to the head of a healthy volunteer. Some of these images are displayed in Figures 4a-4d. Figure 5 a demonstrates typical “T2” decay with pTE which is practically identical with signal decay vs TE, as demonstrated by the correlation plot of Fig. 5b. The following imaging parameters were used for the DEFAISE and CPMG sequences: 22 cm FOV, 5 mm slice thickness, 256 X 256 image matrices, 2000 ms TR, and 15 ms 27 values. Figure 6a presents saturation recovery curves (signal intensity vs TR) obtained from FAISE images of three NiC12 phantoms acquired with a pTE of 15 ms and TR values ranging from 300 to 1500 ms. The signal intensities within each of these phan-
TABLE 1 Results of Linear Least Squares Fit for All Correlation Plots Obtained in This study
In vitro pTE/TE Brain FAISE pTE/TE Brain DEFAISE pTE/TE In vitro TR/TR In vitro N,IN, Blood 21/27 Corn oil 27/27
a
b
cc
Figure
1.007 0.99 1 0.935 0.995 0.926 I .020 0.913
-0.372 5.307 26.682 17.487 5.6 18 5.638 2.929
0.999 0.998 0.996 0.999 0.999 0.998 0.999
lb 3b 5b 6b 7 8a 8b
346
MELKI, JOLESZ, AND MULKERN
FIG. 2. (a-f) Comparison between CPMG (left) and FAKE head images for pTE/TE values of 45, 90, and 180 ms.
THE FAISE METHOD: CONTRAST EQUIVALENCE
347
FIG.2-Continued
toms were also measured as a function of TR from a series of 1st echo CPMG images. The correlation plot between signal intensities from this study and the FAISE study is displayed in Fig. 6b. As calculated from the CPMG data sets, the T , values were 5 1, 133, and 266 ms. Figure 7 presents the variation of signal intensity taking place within a long T I(556 ms) phantom as the number of echoes is gradually increased to 4, 6, 8, 12, and 16 echoes. These acquisitions were performed with repetition times of 300,500, and 1000 ms and a pTE of 15 ms. Reduction of the signal intensity is observed as the number of echoes per train is increased. The effect is more significant at short TR than at long TR. These results were strongly correlated with identical studies performed with the 15-msTE image of variable echo number CPMG sequences (correlation plot not shown). The effect of echo spacing on signal intensity vs pTE for the coagulated blood sample and a corn oil phantom are demonstrated in Figs. 8a and 8b, respectively. Sixteen-echo, 16-shot FAISE sequences using a 27 of 15 and 30 ms were employed. As may be seen from the figure, the signal intensity decay with pTE is augmented with the 30-ms 27. Sixteen-echo CPMG sequences with 27 values of 15 and 30 ms, respectively, were also used to image these phantoms. An excellent correlation between FAISE and CPMG signal intensities was obtained for all pTE/TE values and for both echo spacings in each phantom (correlation plot not shown). The data from the correlation plots for each study were fit with the linear least squares method to the equation FAISE = a CPMG
+ 6,
[I1
348
MELKI, JOLESZ, AND MULKERN 000
a
White m a t t e r
100
50 30
0
60
90
120
150
180
210
240
PTE (ms)
1000
b
0
x
750--
a, C c
C -
A
Grey matter
A
Fat
-
500-
CSF White matter
F
m
W
v, 4
L L
250-
0 0
250
500 750 CPMG Signal Intensity
1000
FIG.3. ( a ) Signal intensity vs pTE for CSF, white matter, gray matter, and fat. (b) Correlation plot of FAISE signal intensity vs CPMG signal intensity at every TE/pTE for CSF, white matter, gray matter. and fat.
where FAISE and CPMG refer to the signal intensities from each type of sequence, a is the slope of the correlation plot, and b is its y-intercept. Correlation coefficients cc were also calculated for each plot. Table 1 lists the results of these fitting procedures. Note that the slopes are all close to 1, indicating signal intensities from the two sequences are very similar. The largest b values are on the order of noise measurements made outside of the imaged objects. The correlation coefficients are all above 0.99. Using a four-echo FAISE sequence ( pTE of 15 ms) and a double-echo SE sequence ( TEs of 1 5 and 60 ms), the beef steak and NiC12phantom were imaged while gradually increasing the number of slices acquired. Signal intensity measurements in each material were made from the central slice image. Figure 9 presents the ratio of the signal intensity obtained from an n slice acquisition over that obtained from a single-slice acquisition. For both the FAISE and SE acquisitions, a 192 X 256 matrix, a TR of 1500 ms, and a field of view of 24 cm were used. For both FAISE and the dual-SE sequences, increasing the number of slices led to signal loss within the beef steak but
THE FAISE METHOD: CONTRAST EQUIVALENCE
FIG.4. (a-d) Comparison between CPMG (left) and DEFAISE (right) head images.
349
350
MELKI, JOLESZ, AND MULKERN a 1000
A
White m a t t e r
0
Fat
100
15
0
30
45
60
75
90
105
120
PTE (ms)
0 A
White m a t t e r
csF
900.-
100
Grey m a t t e r
A
Subst nigra
0
Fat
500
30
CPMG Signal Intensity
FIG. 5. (a, b ) Signal intensity vs pTE for selected tissues as acquired with the DEFAISE sequence and the corresponding correlation plot.
not within the NiC12 phantom. When a single slice was acquired at the end of the experiment, the signal intensity of the steak returned to its initial value. This signal attenuation with increased number of slices is more significant with the four-echo FAISE sequence than with the dual-SE sequence. DISCUSSION
The FAISE methods are partial RF echo-planar techniques (3) which utilize a specific phase-encode reordering algorithm to manipulate T2contrast ( 1). These rapid scan methods have the ability to produce images with contrast directly comparable to conventional spin-echo methods. In this regard, a comparative contrast mechanism study of FAISE with equivalent multiecho spin-echo sequence (CPMG) was performed in order to quantify the contrast similarity and to appreciate any potential differences that may exist.
35 1
THE FAISE METHOD: CONTRAST EQUIVALENCE a
2250 -x
-
c v)
C
-
1500.-
0 c .-in
m
cn
750.-
LQL
,
L Y i I
A
T, =
0
T,
Y
=
51 ms
133 m s
T, = 266 ms
0 0
500
1000
1500
2( 30
1900
2 10
TR (ms)
400 400
900
1400
CPMG Signal Intensity
FIG.6. (a) Saturation recovery curves for three NiClz phantoms as obtained from FAISE images with pTE fixed at 15 ms. ( b ) Correlation plot of FAlSE and CPMG signal intensities for all TR values and for all three phantoms. Signal intensities were taken from the 15-ms echo time image in the CPMG studies.
The results demonstrate that the actual contrast associated with partial R F echoplanar techniques is highly correlated with those of equivalent CPMG sequences. Particularly, the T2 contrast manipulations performed with the FAISE algorithm through the pTE parameter are closely correlated to those achieved by the conventional TE of the CPMG sequences. The T 2decay curves, saturation recovery curves, FAISE/ CPMG correlation plots, and associated data in Table 1 all document the fact that contrast alterations performed with FAISE techniques are entirely equivalent to those permitted with multiecho SE methods. A close correlation between CPMG and FAISE images is also demonstrated for imaging parameters such as the echo spacing and the echo train length. The results in I and those presented in this work lend support to the idea that FAISE methods can replace conventional SE screening exams. However, this comparative contrast evaluation does not support a complete equivalence between partial R F echo-
352
MELKI, JOLESZ, AND MULKERN
A
--
TR = 1000 rns
TR = 500 m s o TR = 300 rns 0
5
0
15
10
20
Number of echoes (Ne)
FIG.7. Decrease of signal intensity in a NiClz phantom ( T , = 556 ms) FAISE images as the number of echoes is increased. Studies were performed with three different TRs and with pTEs of 15 ms.
0
30
60
90
120
150
180
210
240
PTE (ms)
b
0
30
60
90
120
150
180
210
240
pTE (rns)
FIG.8. (a, b) Effect of echo spacing on signal intensity decay with pTE for a clotted blood sample (a) and a corn oil sample (b).
THE FAISE METHOD CONTRAST EQUIVALENCE
353
1125~ NiCl2 phantom, dual SE
A
x V NiC12 phantom,
c
vi
1 0 0 0 1 - I - -p0 c F
0 .1 .' 1
m
LL
0
0875.-
07501
0
Steak, dual SE
0
Steak, FAISE
.\.\..
\o-O,
F
I
FAISE
- 4- - L - I - -X-
\b\.
UI
3
-x
*--.
0 . 0 '
I
- p -
!
I
:
I
0
4-
/ I
I
planar methods images and routine SE methods. Indeed, T2 and spin density image weightings are commonly acquired using dual-echo SE sequences utilizing assymetric, longer interecho intervals (e.g., TE 30/80) than those employed with the fast techniques. In certain mediums, the signal intensity depends not solely on the echo time TE, but also on the echo spacing ( 4 - 6 ) . As demonstrated with the clotted blood sample (Fig. 8a) in which magnetic.suceptibility differences are present ( 4 , 5 ) the use of a short echo spacing (e.g., 15 vs 30 ms) leads to a lengthening of the apparent T2 and therefore, to increased signal intensity at any given pTE or TE. Such effects are important to indentify since they may play an important role in the detection of blood products and brain iron. Another predictable difference between SE images acquired with longer echo spacings and comparable R F echo-planar acquisitions is represented by an increased signal intensity from fat containing tissue. As demonstrated in Fig. 8b, the use of a short echo spacing is responsible for increased signal within the oil sample. Decreased contributions to the T2decay process from spin-spin interactions most probably accounts for this effect (6). However, detailed studies linking homo/ hetero nuclear J-coupling constants with signal attenuation in FAISEICPMG experiments are required before definitive statements can be made regarding the precise mechanisms responsible for bright fat. As demonstrated in Fig. 7, increasing the number of echoes results in a reduced signal intensity within both FAISE and CPMG images (see Table 1 ). The effect is well-known (8, 9) and is a consequence of the retardation of full longitudinal recovery until after the last 180" pulse in a given train. Thus, increasing echo train lengths can lead to subtle T I contrast differences between FAISE and saturation recovery SE sequences. A more incidental contrast mechanism leading to differences in contrast between FAISE and conventional SE sequences is the magnetization transfer contrast ( MTC) phenomenon (2, 7). With the increased number of R F pulses occurring during FAISE acquisitions, more off-resonance irradiation of recovering slices is taking place than in conventional multislice SE methods. In Fig. 9, the signal attenuation taking place within the beef steak (muscle) as the number of slices is increased is not observed in the NiC12 phantom. In addition, this attenuation was consistently smaller in images
354
MELKI, JOLESZ, AND MULKERN
acquired with the dual-SE sequence than in the image collected with a four-echo FAISE sequence. The increased sensitivity of FAISE to MTC is a natural consequence of the increased RF power deposition and may provide a useful method for accessing MTC contrast. CONCLUSION
This study demonstrates that image contrast with FAISE is governed by pure T 2 and T I contrast mechanisms, as in conventional multiecho SE sequences. However, noticeable contrast differences essentially related to the use of different echo spacings may be anticipated between conventional SE methods and partial RF echo-planar techniques in certain situations. If a proper compromise is made between image artifact and relevant image weightings, partial RF echo-planar techniques have the potential to largely replace conventional SE examinations. ACKNOWLEDGMENTS This work was supported in part by I'lnstitut National de la Sante et de la recherche Medical (INSERM, France) and by NIH Grant PO1 CA 41 167. One of us (P.S.M.) is grateful for the hnancial support obtained from the INSERM while on leave from the unit6 316 directed by Professeur Ltandre Pourcelot (C.H.U. Bretonneau, Tours, France). REFERENCES 1. P. S. MELKI,R. V. MULKERN, L. P. PANYCH,AND F. A. JOLESZ,J. Mugn. Reson. Imaging 1, 319
(1991).
2. T. W. DIXON,H. ENGELS,M. CASTILLO,AND M. SHARDASHTI, Mugn. Reson. Imaging8,417 ( 1990). 3. J. HENNIG,A. NAURETH, AND H. FRIEDBURG, Mugn. Reson. Med. 3, 823 (1986). 4. K. R. THULBORN, J. C. WATERTON,P. M. MATTHEWS, AND G. K. RADDA,Biochim. Biophys. Acta 714,265 (1982). 5. K. WEINGARTEN, R. D. ZIMMERMAN, P. T. CAHILL,AND M. D. F. DECK,AJNR 12,475 ( 1991). 6. A. ALLERHAND, J. Chem. Phys. 44, 1 (1966). 7. S. D. WOLF AND R. S. BALABAN, Magn. Reson. Med. 10, I35 ( 1989).
8. T. R. NELSON,R. E. HENDRICK, AND W. R.HENDEE,Mugn. Reson. Imaging 2,285 ( 1984). 9. S. J. RIEDERER, S. A. BOBMAN, J. N. LEE,F. FARZANEH, AND H. 2. WANG,J. Compul. Assist. Tomogr. 10, 103 (1986).
342-354 ( 1992)
Partial RF Echo-Planar Imaging with the FAISE Method. 11. Contrast Equivalence with Spin-Echo Sequences PHILIPPE
s. MELKI,FERENCA. JOLESZ,AND ROBERTv. MULKERN*
Departtncxt of Radiology, Brigham and worn en'.^ Hospital, Hurvard Medical School. Boston, Massachusetts; and * Department of Radiology, Children’s Hospital, Harvard Medical School, Boston, MassachusettJ Received July 10, 1991; revised October 10, 1991; accepted October 14, 1991 The fast acquisition interleaved spin-echo (FAISE) sequence and its dual-echo version (DEFAISE) are partial RF echo-planar methods which utilize a specific phase-encode reordering algorithm to manipulate T2 contrast via an operator-controlled pseudo-echo time, pTE. The repetition time, TR, between successive applications of the Carr-PurcellMeiboom-Gill (CPMG) echo trains used in FAKE may be reduced to introduce T I weighting. To quantitatively determine the extent to which FAISE T , and T2 contrast characteristics agree with spin-echo methods, signal intensities from FAISE acquisitions were compared with signal intensities from equivalent CPMG acquisitions. In phantoms and in human heads, the contrast Characteristics of FAISE are found to be highly correlated with that obtained with equivalent CPMG sequences. However, conventional SE sequences generally utilize longer echo spacings than employed with FAISE/CPMG. Thus, echo spacing-dependent mechanisms such as spin-spin coupling and magnetic susceptibility lead to some differences in contrast between conventional SE and FAISE. Finally, FAISE appears to be more sensitive to magnetization transfer effects than conventional SE sequences since more off-resonance irradiation is applied to individual slices during multislice acquisitions. Q 1992 Academic Press. Inc. INTRODUCTION
In clinical MRI, designing an imaging protocol consists in optimizing contrast between common pathological processes and normal surrounding tissues. With conventional SE methods, this optimization is performed by varying the proportion to which various mechanisms contribute to contrast by manipulating repetition times TR and echo times TE. A precise knowledge of how the sequence parameters of any technique influence the contributions from basic contrast mechanisms is essential for image contrast optimization. Partial RF echo-planar techniques, like fast acquisition interleaved spin echo (FAISE) ,have been shown to provide “spin-echo”-like images in significantly reduced acquisition times when compared to conventional SE methods ( 1 ) . This study aims to evaluate how the contrast manipulations achieved by varying the pseudo-echo time pTE, the repetition time TR, and the echo spacing 27, correlates with equivalent contrast manipulations performed with SE methods. The results support the contention that the contrast characteristics of FAISE are extremely similar to those of the slower spin-echo methods and that differences resulting from echo spacing effects are pre0740-3194/92 $5.00 Copyright 0 1992 by Academic Press. Inc. All rights of reproduction in any form reserved
342
THE FAISE METHOD: CONTRAST EQUIVALENCE
343
dictable and well understood. The inevitable T2decay-related artifacts associated with FAISE have been discussed in the accompanying work, here referred to as I. MATERIALS AND METHODS
FAISE, DEFAISE (dual-"echo" FAISE), and CPMG sequences were all implemented on 1.5-T GE SIGNA systems (General Electric Corp. Milwaukee, WI). The FAISE sequence ( I ) permits manipulation of T2contrast through operator selection of the magnitude of the phase-encode gradient used to encode the first echo of the first echo train (see IPS parameter of I). Certain IPS parameters yield minimal phaseencode ghosting noise. We restrict our attention to these values in this work. The resulting pTE, is given by an integral multiple of the echo spacing in the CPMG trains (seeEq. [ 5 ] o f l ) . For both CPMG and FAISE sequences, the phase-encode gradient lobes are applied before each echo readout and compensated following each echo readout. In fact, aside from the phase-encoding line in the pulse diagrams, the CPMG and FAKE sequences are virtually identical. In FAISE, however, the phase-encode gradient is changed from echo to echo according to the algorithm described in I. In the CPMG sequence, the phase-encode gradient is changed from echo train to echo train, allowing individual images to be constructed from each echo. In order to produce dual-contrast FAISE (DEFAISE) images, we have employed echo trains consisting of eight echoes. Each half of the echo train collects data lines for two raw data sets. As with single-contrast FAISE acquisitions, each set of four echoes are phase encoded according to the algorithm described in I. For comparative CPMG and FAISE (or DEFAISE) contrast studies, the TR interval, the echo spacing 27, and the number of echoes N,, were matched. All echoes were collected with the acquisition windows of 8.2 ms. Only one autoprescan was applied, resulting in identical receive/transmit gain levels for each acquisition. An echo spacing of 15 ms was used in most cases but was lengthened in one set of experiments. Quantitative signal intensity comparisons were performed with NiClz phantoms and from head images of normal volunteers. For the T2 relaxation studies, serial FAISE acquisitions produced a set of images associated with the full range of pTE parameters. The corresponding set of CPMG images were obtained following the FAISE acquisitions. For T I studies, the pTE of the FAISE sequence was set at 15 ms, and the TRs were varied. These images were compared with 15-ms TE images of CPMG sequences run with identical TR intervals. To evaluate how the number of echoes per CPMG train influences signal intensity, a relatively long T Iphantom was imaged with the FAISE sequence while varying the number of echoes per train for a number of different TR intervals. In addition, the signal dependence of FAISE and CPMG on echo spacing was studied in corn oil and in a coagulated blood sample. The blood sample had been drawn from a healthy volunteer and stored (unheparinized) in a 20ml syringe for 100 days prior to imaging. These samples were imaged at various pTE/ TE values with echo trains using two different echo spacings. Signal intensities from phantoms and various brain tissues were measured from region of interest placed within both types of images. Correlations between paired measurements were made from images with identical echo spacing, TR. pTE/TE, and number of echoes. From the various sets of CPMG images, T 2values of selected
344
MELKI, JOLESZ, AND MULKERN
materials were calculated from the signal intensity decay with echo time TE. Similarly, the T I values were obtained from signal intensity variations with TR interval using the first echo CPMG images (TE = 15 ms). The effect of off-resonance irradiation during multislice acquisitions was tested with a protocol described by Dixon (2). A multislice four-echo FAISE sequence with a pTE of 15 ms, a 1500-ms repetition time, and a 15-ms echo spacing was employed. The slice thickness was 5 mm with a gap of 17 mm between slice centers. The frequency offset between slices was 1.93 kHz/slice. An uncooked beef steak and a NiC12 phantom were imaged with 1, 3, 5, . . . 15, and 17 slice acquisitions, a single-slice sequence ending this experiment. The slices were excited in the order of their spacial position, and an even number of slices were excited on each side of the center slice. These serial acquisitions were repeated with a dual-echo SE sequence with identical imaging parameters. Analysis of the signal intensities was performed with the central slice image.
0
.& _
T 2 = 955 ms
100
(I)
T2 = 105 ins
w
E
z
o
T 2 = 44 m s
10
30
0
60
90
120
150
180
210
240
PTE (ms)
> 1500
0
x c
T2 = 105 m s
vl
a, C
;
-
T2 = 44 m s
1000
6
T2 = 955 rns
D C
PP
P
Ln
#
,dA
500
J
2
/*‘
0 0
500
1000
1500
CPMG Signal Intensity
FIG. 1. ( a) Signal intensity vs pseudo-echo time pTE for three NIC12 phantoms. The T2values indicated were measured from signal intensity vs echo time TE, as obtained from a 16-echo CPMG sequence. ( b ) Correlation plot of FAISE signal intensity vs CPMG signal intensity for all pTE and TE values and for all three NiCI, phantoms.
345
THE FAISE METHOD CONTRAST EQUIVALENCE RESULTS
Figure la presents a semi-log plot of signal intensity vs pTE for three different NiC12 phantoms. The measurements were made from a series of 16-echo, 16-shot FAISE sequences using a TR of 2000 ms, a 15-ms echo spacing, a 24-cm field of view (FOV), and a 10-mm slice thickness (acquisition time of 34 s ) . A 16-echo CPMG sequence was also used to image these phantoms using the same sequence parameters (acquisition time of 8 min, 56 s ) . The T 2values for each phantom, as obtained from the 16 CPMG images, were calculated to be 44, 105, and 955 ms. The “T2” values calculated from monoexponential fits of the FAISE signal intensity decays with pTE (Fig. la) were found to be 42, 105, and 973 ms, in good agreement with the CPMG-derived T2 values. Figure 1b presents a correlation plot of the FAISE signal intensities vs CPMG signal intensities for all the pTE/TE values used and for all three phantoms (see Table 1 for correlation coefficients). Figures 2a-2f depict head images of a healthy volunteer as acquired with FAISE (Fig. 2a, 2c, and 2e) and CPMG (Fig. 2b, 2d, and 2f) sequences at pTE/TE values of 45,90, and 180 ms. Both sequences utilized 16 echoes, an echo spacing of 15 ms, a TR of 2000 ms, a 5-mm slice thickness, 256 X 256 image matrices, and a 24-cm FOV. Figure 3a presents signal intensity vs pTE for ROIs in gray matter, white matter, cerebrospinal fluid (CSF), and orbital fat. Figure 3b is the correlation plot of signal intensities measured from both CPMG and FAISE acquisitions for all pTE/TE values and ail four tissues. Comparable results were obtained from the DEFAISE sequence and an eight-echo CPMG sequence, as applied to the head of a healthy volunteer. Some of these images are displayed in Figures 4a-4d. Figure 5 a demonstrates typical “T2” decay with pTE which is practically identical with signal decay vs TE, as demonstrated by the correlation plot of Fig. 5b. The following imaging parameters were used for the DEFAISE and CPMG sequences: 22 cm FOV, 5 mm slice thickness, 256 X 256 image matrices, 2000 ms TR, and 15 ms 27 values. Figure 6a presents saturation recovery curves (signal intensity vs TR) obtained from FAISE images of three NiC12 phantoms acquired with a pTE of 15 ms and TR values ranging from 300 to 1500 ms. The signal intensities within each of these phan-
TABLE 1 Results of Linear Least Squares Fit for All Correlation Plots Obtained in This study
In vitro pTE/TE Brain FAISE pTE/TE Brain DEFAISE pTE/TE In vitro TR/TR In vitro N,IN, Blood 21/27 Corn oil 27/27
a
b
cc
Figure
1.007 0.99 1 0.935 0.995 0.926 I .020 0.913
-0.372 5.307 26.682 17.487 5.6 18 5.638 2.929
0.999 0.998 0.996 0.999 0.999 0.998 0.999
lb 3b 5b 6b 7 8a 8b
346
MELKI, JOLESZ, AND MULKERN
FIG. 2. (a-f) Comparison between CPMG (left) and FAKE head images for pTE/TE values of 45, 90, and 180 ms.
THE FAISE METHOD: CONTRAST EQUIVALENCE
347
FIG.2-Continued
toms were also measured as a function of TR from a series of 1st echo CPMG images. The correlation plot between signal intensities from this study and the FAISE study is displayed in Fig. 6b. As calculated from the CPMG data sets, the T , values were 5 1, 133, and 266 ms. Figure 7 presents the variation of signal intensity taking place within a long T I(556 ms) phantom as the number of echoes is gradually increased to 4, 6, 8, 12, and 16 echoes. These acquisitions were performed with repetition times of 300,500, and 1000 ms and a pTE of 15 ms. Reduction of the signal intensity is observed as the number of echoes per train is increased. The effect is more significant at short TR than at long TR. These results were strongly correlated with identical studies performed with the 15-msTE image of variable echo number CPMG sequences (correlation plot not shown). The effect of echo spacing on signal intensity vs pTE for the coagulated blood sample and a corn oil phantom are demonstrated in Figs. 8a and 8b, respectively. Sixteen-echo, 16-shot FAISE sequences using a 27 of 15 and 30 ms were employed. As may be seen from the figure, the signal intensity decay with pTE is augmented with the 30-ms 27. Sixteen-echo CPMG sequences with 27 values of 15 and 30 ms, respectively, were also used to image these phantoms. An excellent correlation between FAISE and CPMG signal intensities was obtained for all pTE/TE values and for both echo spacings in each phantom (correlation plot not shown). The data from the correlation plots for each study were fit with the linear least squares method to the equation FAISE = a CPMG
+ 6,
[I1
348
MELKI, JOLESZ, AND MULKERN 000
a
White m a t t e r
100
50 30
0
60
90
120
150
180
210
240
PTE (ms)
1000
b
0
x
750--
a, C c
C -
A
Grey matter
A
Fat
-
500-
CSF White matter
F
m
W
v, 4
L L
250-
0 0
250
500 750 CPMG Signal Intensity
1000
FIG.3. ( a ) Signal intensity vs pTE for CSF, white matter, gray matter, and fat. (b) Correlation plot of FAISE signal intensity vs CPMG signal intensity at every TE/pTE for CSF, white matter, gray matter. and fat.
where FAISE and CPMG refer to the signal intensities from each type of sequence, a is the slope of the correlation plot, and b is its y-intercept. Correlation coefficients cc were also calculated for each plot. Table 1 lists the results of these fitting procedures. Note that the slopes are all close to 1, indicating signal intensities from the two sequences are very similar. The largest b values are on the order of noise measurements made outside of the imaged objects. The correlation coefficients are all above 0.99. Using a four-echo FAISE sequence ( pTE of 15 ms) and a double-echo SE sequence ( TEs of 1 5 and 60 ms), the beef steak and NiC12phantom were imaged while gradually increasing the number of slices acquired. Signal intensity measurements in each material were made from the central slice image. Figure 9 presents the ratio of the signal intensity obtained from an n slice acquisition over that obtained from a single-slice acquisition. For both the FAISE and SE acquisitions, a 192 X 256 matrix, a TR of 1500 ms, and a field of view of 24 cm were used. For both FAISE and the dual-SE sequences, increasing the number of slices led to signal loss within the beef steak but
THE FAISE METHOD: CONTRAST EQUIVALENCE
FIG.4. (a-d) Comparison between CPMG (left) and DEFAISE (right) head images.
349
350
MELKI, JOLESZ, AND MULKERN a 1000
A
White m a t t e r
0
Fat
100
15
0
30
45
60
75
90
105
120
PTE (ms)
0 A
White m a t t e r
csF
900.-
100
Grey m a t t e r
A
Subst nigra
0
Fat
500
30
CPMG Signal Intensity
FIG. 5. (a, b ) Signal intensity vs pTE for selected tissues as acquired with the DEFAISE sequence and the corresponding correlation plot.
not within the NiC12 phantom. When a single slice was acquired at the end of the experiment, the signal intensity of the steak returned to its initial value. This signal attenuation with increased number of slices is more significant with the four-echo FAISE sequence than with the dual-SE sequence. DISCUSSION
The FAISE methods are partial RF echo-planar techniques (3) which utilize a specific phase-encode reordering algorithm to manipulate T2contrast ( 1). These rapid scan methods have the ability to produce images with contrast directly comparable to conventional spin-echo methods. In this regard, a comparative contrast mechanism study of FAISE with equivalent multiecho spin-echo sequence (CPMG) was performed in order to quantify the contrast similarity and to appreciate any potential differences that may exist.
35 1
THE FAISE METHOD: CONTRAST EQUIVALENCE a
2250 -x
-
c v)
C
-
1500.-
0 c .-in
m
cn
750.-
LQL
,
L Y i I
A
T, =
0
T,
Y
=
51 ms
133 m s
T, = 266 ms
0 0
500
1000
1500
2( 30
1900
2 10
TR (ms)
400 400
900
1400
CPMG Signal Intensity
FIG.6. (a) Saturation recovery curves for three NiClz phantoms as obtained from FAISE images with pTE fixed at 15 ms. ( b ) Correlation plot of FAlSE and CPMG signal intensities for all TR values and for all three phantoms. Signal intensities were taken from the 15-ms echo time image in the CPMG studies.
The results demonstrate that the actual contrast associated with partial R F echoplanar techniques is highly correlated with those of equivalent CPMG sequences. Particularly, the T2 contrast manipulations performed with the FAISE algorithm through the pTE parameter are closely correlated to those achieved by the conventional TE of the CPMG sequences. The T 2decay curves, saturation recovery curves, FAISE/ CPMG correlation plots, and associated data in Table 1 all document the fact that contrast alterations performed with FAISE techniques are entirely equivalent to those permitted with multiecho SE methods. A close correlation between CPMG and FAISE images is also demonstrated for imaging parameters such as the echo spacing and the echo train length. The results in I and those presented in this work lend support to the idea that FAISE methods can replace conventional SE screening exams. However, this comparative contrast evaluation does not support a complete equivalence between partial R F echo-
352
MELKI, JOLESZ, AND MULKERN
A
--
TR = 1000 rns
TR = 500 m s o TR = 300 rns 0
5
0
15
10
20
Number of echoes (Ne)
FIG.7. Decrease of signal intensity in a NiClz phantom ( T , = 556 ms) FAISE images as the number of echoes is increased. Studies were performed with three different TRs and with pTEs of 15 ms.
0
30
60
90
120
150
180
210
240
PTE (ms)
b
0
30
60
90
120
150
180
210
240
pTE (rns)
FIG.8. (a, b) Effect of echo spacing on signal intensity decay with pTE for a clotted blood sample (a) and a corn oil sample (b).
THE FAISE METHOD CONTRAST EQUIVALENCE
353
1125~ NiCl2 phantom, dual SE
A
x V NiC12 phantom,
c
vi
1 0 0 0 1 - I - -p0 c F
0 .1 .' 1
m
LL
0
0875.-
07501
0
Steak, dual SE
0
Steak, FAISE
.\.\..
\o-O,
F
I
FAISE
- 4- - L - I - -X-
\b\.
UI
3
-x
*--.
0 . 0 '
I
- p -
!
I
:
I
0
4-
/ I
I
planar methods images and routine SE methods. Indeed, T2 and spin density image weightings are commonly acquired using dual-echo SE sequences utilizing assymetric, longer interecho intervals (e.g., TE 30/80) than those employed with the fast techniques. In certain mediums, the signal intensity depends not solely on the echo time TE, but also on the echo spacing ( 4 - 6 ) . As demonstrated with the clotted blood sample (Fig. 8a) in which magnetic.suceptibility differences are present ( 4 , 5 ) the use of a short echo spacing (e.g., 15 vs 30 ms) leads to a lengthening of the apparent T2 and therefore, to increased signal intensity at any given pTE or TE. Such effects are important to indentify since they may play an important role in the detection of blood products and brain iron. Another predictable difference between SE images acquired with longer echo spacings and comparable R F echo-planar acquisitions is represented by an increased signal intensity from fat containing tissue. As demonstrated in Fig. 8b, the use of a short echo spacing is responsible for increased signal within the oil sample. Decreased contributions to the T2decay process from spin-spin interactions most probably accounts for this effect (6). However, detailed studies linking homo/ hetero nuclear J-coupling constants with signal attenuation in FAISEICPMG experiments are required before definitive statements can be made regarding the precise mechanisms responsible for bright fat. As demonstrated in Fig. 7, increasing the number of echoes results in a reduced signal intensity within both FAISE and CPMG images (see Table 1 ). The effect is well-known (8, 9) and is a consequence of the retardation of full longitudinal recovery until after the last 180" pulse in a given train. Thus, increasing echo train lengths can lead to subtle T I contrast differences between FAISE and saturation recovery SE sequences. A more incidental contrast mechanism leading to differences in contrast between FAISE and conventional SE sequences is the magnetization transfer contrast ( MTC) phenomenon (2, 7). With the increased number of R F pulses occurring during FAISE acquisitions, more off-resonance irradiation of recovering slices is taking place than in conventional multislice SE methods. In Fig. 9, the signal attenuation taking place within the beef steak (muscle) as the number of slices is increased is not observed in the NiC12 phantom. In addition, this attenuation was consistently smaller in images
354
MELKI, JOLESZ, AND MULKERN
acquired with the dual-SE sequence than in the image collected with a four-echo FAISE sequence. The increased sensitivity of FAISE to MTC is a natural consequence of the increased RF power deposition and may provide a useful method for accessing MTC contrast. CONCLUSION
This study demonstrates that image contrast with FAISE is governed by pure T 2 and T I contrast mechanisms, as in conventional multiecho SE sequences. However, noticeable contrast differences essentially related to the use of different echo spacings may be anticipated between conventional SE methods and partial RF echo-planar techniques in certain situations. If a proper compromise is made between image artifact and relevant image weightings, partial RF echo-planar techniques have the potential to largely replace conventional SE examinations. ACKNOWLEDGMENTS This work was supported in part by I'lnstitut National de la Sante et de la recherche Medical (INSERM, France) and by NIH Grant PO1 CA 41 167. One of us (P.S.M.) is grateful for the hnancial support obtained from the INSERM while on leave from the unit6 316 directed by Professeur Ltandre Pourcelot (C.H.U. Bretonneau, Tours, France). REFERENCES 1. P. S. MELKI,R. V. MULKERN, L. P. PANYCH,AND F. A. JOLESZ,J. Mugn. Reson. Imaging 1, 319
(1991).
2. T. W. DIXON,H. ENGELS,M. CASTILLO,AND M. SHARDASHTI, Mugn. Reson. Imaging8,417 ( 1990). 3. J. HENNIG,A. NAURETH, AND H. FRIEDBURG, Mugn. Reson. Med. 3, 823 (1986). 4. K. R. THULBORN, J. C. WATERTON,P. M. MATTHEWS, AND G. K. RADDA,Biochim. Biophys. Acta 714,265 (1982). 5. K. WEINGARTEN, R. D. ZIMMERMAN, P. T. CAHILL,AND M. D. F. DECK,AJNR 12,475 ( 1991). 6. A. ALLERHAND, J. Chem. Phys. 44, 1 (1966). 7. S. D. WOLF AND R. S. BALABAN, Magn. Reson. Med. 10, I35 ( 1989).
8. T. R. NELSON,R. E. HENDRICK, AND W. R.HENDEE,Mugn. Reson. Imaging 2,285 ( 1984). 9. S. J. RIEDERER, S. A. BOBMAN, J. N. LEE,F. FARZANEH, AND H. 2. WANG,J. Compul. Assist. Tomogr. 10, 103 (1986).
