Magnetic Resonance Printed in the USA.
Imaging. All rights
Vol. 9, pp. 469-476, reserved.
1991 Copyright
0730-725X/91 $3.00 + .OO 0 1991 Pergamon Press plc
l Original Contribution
MAGNETIZATION PREPARED RAPID GRADIENT-ECHO (MP-RAGE) MR IMAGING OF THE LIVER: COMPARISON WITH SPIN-ECHO IMAGING EDUARD E.
P. MUGLER, III, JAMES A. BERTOLINA, SPENCER B. GAY, CYNTHIA L. JANUS, AND JAMES R. BROOKEMAN Departmentof Radiology, University of Virginia Health Sciences Center, Charlottesville,Virginia 22908, USA DE LANGE,
JOHN
We have implemented an MR technique that employs a rapid gradient echo sequence, preceded by magnetization preparation pulses to provide Z’t- and Tz-weighted tissue contrast. With this technique, which can be identified as a member of a new family of pulse sequences, generically named Magnetization Prepared RApid Gradient Echo (MP-RAGE), very short repetition times are used, allowing acquisition times of less than one second and images virtually free of motion-induced artifacts during quiet respiration. Fifteen patients with known liver lesions (metastases, hemangiomas, and cysts) were examined using IfI- and T,-weighted 2-dimensional MP-RAGE sequences, and the images were compared with conventional T,- and multi-echo T,-weighted spin-echo (SE) sequences. Signal difference-to-noise ratios (SD/Ns) of the lesions were calculated for all pulse sequences using corresponding axial images and were normalized for voxel volume. The mean normalized SD/Ns of the MP-RAGE sequences were generally comparable to those for the SE sequences. In addition, there were no noticeable respiratory artifacts on the MP-RAGE images whereas these were clearly present on the T,-weighted SE images and to a lesser degree on the Tr-weighted SE images. It is concluded that the MP-RAGE technique could become an important method for evaluating the liver for focal disease.
Keywords: Magnetic resonance (MR), technology; Magnetic resonance (MR), pulse sequences; Magnetic resonance
(MR), rapid imaging; Liver neoplasms, MR studies; Liver, MR studies.
sequence element. With the 2_dimensional(2D) Magnetization Prepared RApid Gradient Echo (MP-RAGE) sequences, MR images with high tissue contrast can be obtained in extremely short imaging times and virtually without motion artifacts. The technique is therefore particularly useful for imaging the abdomen. To evaluate the use of this method in imaging the liver we have compared it with conventional spin-echo (SE) sequences in a preliminary study of 15 patients with focal liver lesions.
INTRODUCTION Recently, Haase et al. introduced a new pulse sequence technique called snapshot FLASH magnetic resonance (MR) imaging. ‘v2The technique can provide T,- and T2-weighted images with superior image contrast at high field strength. This technique, which has also been called TurboFLASH or snapshot GRASS, is characterized by the combination of two distinct pulse sequence segments. The first segment is used to prepare the magnetization for tissue contrast, and the second, which is a fast gradient-echo sequence, is used for data acquisition. This newly developed pulse sequence technique can be recognized as a member of a larger family of pulse sequences which employ a magnetization preparation-rapid gradient-echo acquisition - magnetization recovery cycle as the basic
METHODS AND MATERIALS The MP-RAGE
Pulse Sequences
Gradient-echo pulse sequences are characterized by the fast data sampling that yields high speed MR images with signal-to-noise ratios (SNRs) comparable to
Requests for reprints and correspondence should be addressed to Eduard E. de Lange, M.D., Department of Radiology, Box 170, University of Virginia Health Sciences Center, Charlottesville, VA 22908, USA.
RECEIVED 1215190; ACCEPTED 3/14/91. We thank Geneva Shifflett, Diana Bowman and Kelly Powell for their secretarial assistance; Carol Chowdhry, Ph.D., for the editorial assistance; and Ursula Bunch for the photography.
Acknowledgments-
469
Magnetic Resonance Imaging 0 Volume 9, Number 4, 1991
470
SE images; however, signal difference-to-noise ratios (SD/Ns) are often lower. 3*4By changing TR, TE, and the flip angle, image contrast can be manipulated, but in general contrast relationships between tissues may be complex, and gradient-echo images are neither T,nor Tz-weighted in the sense of SE imaging.3 A rapid gradient-echo (RAGE) image obtained with a short TR, short TE, and a small flip angle displays contrast that is essentially dominated by the proton density of the tissues.2 When such a RAGE sequence is preceded by pulses to prepare the magnetization [magnetization preparation (MP) pulses], the spindensity contrast can be transformed to Tr-weighted or Tz-weighted contrast. 1,2~5 To acquire a T,-weighted MP-RAGE image, a single, nonselective 180” pulse is used to prepare the magnetization. ’ The 180” inversion pulse rotates the longitudinal magnetization from the positive z-axis to the negative z-axis (Fig. 1). If a time delay follows the inversion pulse to allow Tl relaxation to occur and this delay is followed by the RAGE sequence for the
Inversion
Oelay
image acquisition, the image displays T,inversion recovery contrast. The inversion delay time between the MP pulse and the start of the RAGE sequence basically determines the image contrast. To acquire a T2-weighted MP-RAGE image, a combination of three nonselective pulses (90°-180”90’) is needed to prepare the magnetization.2J The first 90” pulse rotates the magnetization from the positive z-axis to the transverse plane. As with SE imaging, a time interval is allowed for T2 decay to occur and T2-dependent contrast to develop between different tissues (Fig. 2). The 180” pulse is applied to refocus the static field inhomogeneity contributions to the dephasing of the magnetization in the transverse plane (i.e., to perform a spin-echo). When the transverse magnetization is refocused (after twice the interval
Metastasis (4
Liver
Fig. 1. T, -weighted MP-RAGE sequence. (A) To provide Tr contrast, a nonselective 180” pulse is employed for magnetization preparation (MP). After a certain inversion delay, the rapid gradient echo (RAGE) sequence is employed for image acquisition. A spoiler gradient before the RAGE sequence is used to dephase any residual transverse magnetization resulting from an imperfect inversion pulse. (B) The 180” MP pulse rotates the longitudinal component of the magnetization (Mz) to the negative z-axis. During the inversion delay time, T, relaxation of the tissues occurs. (The T, of the liver is shorter than the T, of metastasis.) If the RAGE sequence is employed as the longitudinal magnetization of the metastasis passes through the x-y plane, an image is acquired on which the liver is bright and the metastasis black.
Fig. 2. Tz-weighted MP-RAGE sequence. (A) For the T2 contrast a combination of three nonselective pulses is needed. The first 90” pulse about the x-axis rotates the magnetization to the x-y plane. After time is allowed for T, contrast to develop between tissues, the 180” pulse about the y-axis is applied to refocus &hedephasing spins. When the spin-echo is obtained, a second 90” pulse is applied about the x-axis to “store” the T2 contrast in the form of longitudinal magnetization along the negative z-axis. The interval between the two 90” pulses (TE,,,,) is the period during which T, contrast develops. A spoiler gradient is applied to dephase any residual transverse magnetization resulting from imperfect pulses. (B) After the longitudinal magnetiziation is encoded by T2 contrast and rotated to the negative z-axis, the RAGE sequence is employed for data sampling. However, since Tl relaxation occurs after the longitudinal magnetization is encoded, T, contrast adds to T, contrast. As the T2 of the liver is shorter than that of metastasis, the T,-encoded longitudinal magnetization is less than that of metastasis. If the RAGE sequence is employed as the magnetization of the liver passes through the x-y plane, the liver will be black and the metastasis bright (magnitude reconstruction).
MP-RAGE
MRI of the liver 0 E.E. DE LANGE ET AL.
between the first 90” pulse and the 180” pulse), the second 90” pulse is applied to restore the refocused magnetization to the z-axis, thereby “storing” the T2 contrast of the tissues in the form of longitudinal magnetization. The interval between the two 90” pulses is the period during which the T2 contrast develops between tissues and is labeled TE,,,,. When both 90” pulses are applied about the x-axis (in the rotating frame) and the 180” pulse is applied about the y-axis, the longitudinal magnetization encoded by T2 decay is rotated to the negative z-axis (i.e., driven inversion).6 We found that this technique provides greater contrast between liver lesions and normal liver than does the technique described by Haase in which all preparation pulses are applied about the same axis, causing the longitudinal magnetization to rotate to the positive z-axis (i.e., driven equiIibrium).2 As with the T,-weighted images, a RAGE sequence is employed after the preparation pulses to acquire the image data. MP-RAGE Imaging of the Liver For the 2D MP-RAGE sequences, each image is sampled as a single slice using the RAGE sequence with one acquisition. To obtain clinically useful images of good quality, the acquisition period needs to be long enough to provide adequate SNR. However, if breath-holding is not employed, lengthening the data acquisition period also increases artifacts from motion as sampling of data occurs during the cardiac and respiratory cycles. For our Ti-weighted MP-RAGE images we chose an acquisition time of less than one second to adequately limit motion artifacts during quiet respiration. Employing both theoretical calculations and experimental measurements in normal volunteers,7 we found that an inversion delay of 350 msec, a TR of 7.1 msec, a TE of 3.7 msec, and a flip angle of 10” provided a reasonable compromise for generating strong T,weighted contrast. The flip angle choice represents a trade-off between increased signal-to-noise and contrast-to-noise ratios, and increased artifacts due to the build-up of steady coherences (FLASH type sequence), both of which occur with increased flip angles. We used liver-spleen contrast as a model for liver-lesion contrast in these optimization studies. Also, we employed a rectangular field-of-view (FOV) of 350 x 700 mm with an image matrix of 128 x 256 to obtain isotropic in-plane voxel dimensions. The resulting acquisition time for the RAGE sequence was 909 msec. For the MP-RAGE sequences, data sampling occurs during a Ti-dependent transient and Tl decay, therefore, affects the prepared contrast. This is a particular problem when the goal is to encode T2 contrast,
471
and the T, decay during the RAGE acquisition may destroy the T2 contrast developed with the preparation. However, when T,-weighted MP-RAGE images are obtained with the driven inversion preparation, as is the case with our technique, the T, and T2 decays can complement each other.6 A similar situation arises with the short tau inversion recovery (STIR) technique. To achieve the desired contrast effect from the T2 preparation pulses, the RAGE acquisition time must be substantially shorter than that used for the Ti-weighted images, and must also be approximately the same as or shorter than the T, relaxation time of normal liver. For the T2-weighted MP-RAGE sequence we used a TR of 4.5 msec and only 64 phaseencoding steps to achieve an acquisition time of 288 msec. The TE was 2.3 msec, the flip angle lo”, the matrix 64 x 128, and the FOV 300 x 600 mm. A TE prep of 40 msec was used as this provided essentially no signal from normal liver and relatively high signal from the lesions. Patients and Methods Fifteen patients with known liver lesions were examined with MR imaging using T,- and T2-weighted MP-RAGE pulse sequences with the parameters given in the previous paragraphs. The findings were compared with routine, conventional SE images obtained during the same examination. Ten patients had known metastases from melanoma (n = 1), tumors in the colon or rectum (n = 8), or ovary (n = l), proven either by biopsy or surgery (n = 8), or clinical follow-up and subsequent CT scanning or MR imaging to establish growth of the lesions (n = 2). Four patients had known hemangiomas proved previously on studies, and one patient had multiple liver cysts, proved previously by CT and ultrasound. The studies were performed on a 1.5 Tesla Magnetom 63SP MR imaging system (Siemens Medical Systems, Inc., Iselin, NJ). Ti-weighted SE imaging was performed with TR/TE of 600/15 msec, averaging 4 data acquisitions. The imaging matrix was 128 x 256 and the FOV 420 x 420 mm. The total acquisition time for this sequence was 5.12 min. Multi-echo T2-weighted SE imaging was performed with TR/TE of 2800/40, 80, 120, 160, averaging 2 acquisitions. The matrix was 128 x 256, the FOV was 420 x 420 mm, and the total acquisition time was 11.95 min. Power deposition was monitored for all sequences using the manufacturer’s standard protocol and the power deposition never exceeded the FDA guidelines. Informed consent was obtained for all patients. Scout coronal images were obtained in a multislice mode using a SE 500/15 sequence to determine the off-set for axial acquisitions. All axial images were ob-
472
Magnetic Resonance Imaging 0 Volume 9, Number 4, 1991
tained using a lo-12 mm section thickness. The SE pulse sequences were obtained in a multislice mode of 16 sections with 33% interslice gaps. The MP-RAGE sequences were obtained in a single slice mode using the same 16 slice positions as those for the SE pulse sequences. The SE and the T1-weighted MP-RAGE sequences were obtained in all cases, and the T2weighted MP-RAGE sequence only in 11 of the 15 patients. In some instances additional coronal or sagittal images were obtained using the MP-RAGE sequences. All images were obtained during quiet respiration, and straps were applied across the patient’s abdomen to reduce motion artifacts. No other motion suppression techniques were used. Signal difference-to-noise ratios (SD/Ns) were calculated for all pulse sequences using axial image sections in the same location for each sequence. The axial section that demonstrated the largest dimension of the lesion in the liver was used for SD/N calculations. We chose the image showing the largest extent of the lesion to reduce errors in SD/N measurements from slice misregistration. In patients with multiple lesions, the largest lesion was always chosen for intensity measurements. All lesions evaluated were at least 2.0 cm in diameter. A region of interest (ROI) of at least 60 pixels was used for intensity measurements, the ROI being in the same location on the corresponding image section of each sequence. A region of normal liver of at least 300 pixels was chosen on the same image section as the largest liver lesion, and the inclusion of hepatic vessels was avoided as far as possible. To include the motion artifact contribution to the noise, background noise was measured in a ROI of at least 600 pixels located anterior to the anterior abdominal wall and in line with the given liver lesion along the phaseencoding direction. Contrast between liver masses and normal liver was expressed as an SD/N according to the formula SD/N = 1SI,-S&I /a, where SI, is the mean signal intensity of the lesion, SI, is the mean signal intensity of normal liver, and u is the standard deviation of the background noise. SD/Ns were normalized for voxel volume. The voxel volume normalization was performed by dividing each measured intensity level by the corresponding voxel volume, calculated according to the formula: V= (FOVREAD/MAREAD)(FOVPHASE/ MAPHASE)TH, where V is the voxel volume, MA is the matrix size, and TH is the slice thickness. Mean and standard deviations of the normalized SD/Ns for the four pulse sequences were calculated. The performances of the MP-RAGE and SE sequences were compared by testing for the equality of sample means using a t-test for paired data.’
RESULTS
The SD/Ns normalized for voxel volume are plotted in Fig. 3. The mean SD/N values and standard deviations are also given in Fig. 3. Comparing the T,MP-RAGE to the SE sequences, no statistically significant differences were found except for TE 160 TZ-weighted SE sequence (n = 15, p < .Ol). In this case, there was a statistically significant increase in the mean for the MP-RAGE versus the SE sequence. Comparing the T,-MP-RAGE to the SE sequences, no statistically significant differences were found except for TE 80 Tz-weighted SE sequence (n = 11, p < .Ol). In this case, there was a statistically significant increase in the mean for the SE versus the MPRAGE sequence. In all cases the lesions displayed relatively low signal intensity on T,-MP-RAGE images and high signal intensity on the T,-MP-RAGE images (Figs. 4, 5). There was no noticeable difference in the signal characteristics of the metastases, hemangiomas, or cysts. There was no noticeable blurring from respiratory motion on the MP-RAGE images, although blurring was obvious on the multiecho T,-weighted SE images and to a lesser degree on some of the T,-weighted SE images. DISCUSSION
The conspicuity of focal liver lesions on MR images varies with field strength. At low and intermediate field strength (0.02-0.6 Tesla) the short TR, short TE spin-echo technique is superior in displaying focal liver lesions, a reflection of the significant differences in T,values between focal lesions and liver.‘-” As field strength increases, T,differences between abnormal and normal tissues decrease, 12*13 and lesions may become less conspicuous with short TR, short TE pulse sequences. Studies have shown that at high field strength (1.5 T), conspicuity of focal hepatic disease is superior on long TR, short TE and long TR, long TE pulse sequences. 9,10At high field strength, however, artifacts from the moving anterior abdominal wall during respiration are more prominent than at Iow to mid field strength. ‘* Although signa averaging during short TR, short TE pulse sequences can effectively suppress motion artifacts without increasing imaging times to extreme proportions, the length of the examination makes signal averaging impractical for pulse sequences with long TRs. Therefore, other means of suppressing respiratory artifacts have been used for the long TR techniques. These include methods where the image acquisition is triggered by the respiratory cycle; techniques that reorder the phase-
MP-RAGE MRI of the liver 0 E.E. SD/N
DE LANGE
473
ET AL.
10 0 S
MAN 1.6 SD 2.0
7
(n=15) A
8
MEAN 2.2 SD 1.0 (n=15) A MEAN 1.5 SD 1.5
MEAN 2.3
(n=l5) A A A MEAN 0.8 SD 0.4 (n=ll)
T,-w-&\Cg
SB ma/L6
T*-w-RACII
A
d t.
*
MEAN 1.5 SD 1.3 (n=15) 4
(n=15) t.
a
A
:
4 a
I
2
8
%
8
9E Psw,sa
SE 2wo,*oo
51 zaoo,4a
MEAN 0.0 SD 0.9
: d : 2.
__~
sg zaoo,ml
Fig. 3. SD/N of 15 liver lesions for each pulse sequence normalized for voxel volume.
encoding sequence in relation to the respiratory cycle14; and the use of gradient moment nulling15 and/or spatial presaturation. l6 An alternative strategy in imaging focal liver lesions has been the use of short TR spin-echo”,‘* or gradient-echo3*18**9acquisitions which can be completed in a single breath-hold. Although only a small number of patients was evaluated, our preliminary findings indicated that SD/Ns of the MP-RAGE images obtained in less than one second per slice with one acquisition were in the same range as those of the 4 acquisition T1-weighted and the 2 acquisition Tz-weighted SE images obtained in approximately 5 and 12 min (for 16 slices), respectively (Fig. 3). Therefore, conspicuity of lesions on the fast MP-RAGE images was generally comparable to that for the conventional SE images. One of the drawbacks of our study is that the TR used for the T1weighted SE sequence was relatively long. We used this TR to provide complete coverage of the liver with one sequence. However, we realize that if shorter TRs were used, the SD/N values would probably increase. Also, we used no motion-suppression techniques other than the straps over the abdomen. As our MR imager does not have the capability of respiratory ordered phase encoding, we were not able to use this technique for motion reduction. In addition, we did not use gradient moment nulling to reduce motion artifacts during liver imaging because with this technique the vessels display high signal intensity. As focal lesions usually display similar high signal intensity on TZweighted images, it is more difficult to distinguish the lesions from the vessels, particularly when the lesions are small. However, the Tz-weighted MP-RAGE sequence also displayed the vessels with high signal intensity creating a similar problem for lesion detection.
In this preliminary study, the relatively long TR used for the T,-weighted SE and the absence of motion artifact reduction strategies in the T,-weighted acquisitions represent a bias in favor of the MP-RAGE technique. On the other hand, the MP-RAGE sequences are only in the early stages of development and optimization, and significant improvements would be expected in the future. Thus, we feel that the preliminary findings with the MP-RAGE sequences do speak well for potential applications of the technique. In future studies, it will be important to compare MPRAGE with the other spin-echo and gradient-echo techniques which can be performed within a single breathhold. In our study only relatively large lesions (>2.0 cm) were evaluated to avoid errors in SD/N calculations resulting from the changes in anatomy occurring during respiration. Although no anatomic changes occur when images are obtained during breath holding, we purposely did not use breath holding during the MPRAGE images because the anatomic positions obtained would not necessarily correlate with the SE images obtained during respiration. However, to obtain consecutive slice sections as needed for clinical imaging, breath holding or respiratory triggering would be appropriate. Since small lesions were not evaluated in this study, it is not known whether conspicuity of these lesions is also better on the 7’,-weighted MPRAGE images compared with SE images. Therefore, the sensitivity of MP-RAGE technique in detecting small lesions is still unknown. For the T,-weighted MP-RAGE sequence we used a TE,,,, of 40 msec because it provided high contrast between liver and normal spleen; however, the optimal TE,,,, for liver lesion characterization is yet to be determined, and it is likely
Magnetic Resonance Imaging 0 Volume 9, Number 4, 1991
474
(A)
CD)
(Cl
Fig. 4. MR images of a liver metastasis from colorectal carcinoma. (A) Transverse T,-weighted MP-RAGE image obtained in approximately 900 msec. Using a delay of 350 msec between the 180” magnetization preparation pulse and the start of the RAGE sequence, the liver displays relatively high signal intensity and the lesion (arrow) low signal intensity. There are virtually no motion-induced artifacts from respiration. (B) Corresponding 7”-weighted MP-RAGE image at the same slice position as (A) obtained in approximately 290 msec. Using a driven inversion preparation with a TE,,,, of 40 msec, the normal liver displays low signal intensity and the lesion (arrow) displays high signal intensity. There are no noticeable artifacts from motion. (C) Corresponding conventional SE 600/15 image obtained in approximately 5 min. The lesion (arrow) is almost isointense to the liver. Note increased image degradation from anterior wall motion. (D) SE 2800180 image. Lesion (arrow) displays relatively high signal intensity compared to the surrounding liver.
that two or more TE,,,, values may be needed for tissue characterization. Another problem with the T,weighted MP-RAGE sequence is that the RAGE acquisition occurs during a Tr-dependent transient, and, therefore, the acquisition of T2-weighted images using the MP-RAGE technique is somewhat problematical.
However, modifications to the technique have already which may solve this problem. These include the use of reordered phase-encoding ” and a “multi-shot” approach which acquires the image data by concatenating several prepare-acquire cycles.2’ In addition to modifying the contrast introduced by been proposed
MP-RAGE
MRI of the liver 0 E.E. DE LANGE ET AL.
6%
475
(W
Fig. 5. MP-RAGE images of a hemangioma. (A) Transverse T, -weighted image demonstrates the 4-cm lesion (arrow) which displays low signal intensity. (B) Corresponding T,-weighted image shows the lesion (arrow) displaying high signal intensity. The images were obtained during respiration and there are no noticeable motion artifacts.
the preparation pulses, the TI decay that occurs during the gradient-echo acquisition may also introduce artifacts into the image. Assuming standard phase encoding is employed, the combined effects of T, decay during the acquisition and the acquisition process generally cause a filter function in the associated phaseencoding direction that is real and asymmetric. This results in a point spread function (PSF) that is complex (hermitian). This PSF produces a tissue-dependent blurring and may create edge artifacts at tissue interfaces. We are investigating reordered phase encoding as a means to eliminate the edge artifacts, to alleviate the effects on the image contrast of the acquisition process and T, decay, and to provide increased SD/NS.~’ One drawback of the MP-RAGE technique is that in its present implementation the in-plane resolution is less than that of conventional SE images. Efforts are being made to further improve the sequences without significantly increasing the imaging times. A 3-dimensional data acquisition, as recently described, may improve image quality considerably.22 Meanwhile, since 2D MP-RAGE imaging provides high contrast images in short imaging times and without significant motion artifact, it is probable that this technique may become an important method to evaluate the liver for focal disease. In particular, because of the fast imaging times, the technique may be useful for dynamic imaging of the liver using gadolinium DTPA for tissue enhancement, and we are investigating this method. Nevertheless, the sensitivity and specificity in detecting and characterizing focal hepatic lesions using the MP-RAGE sequences have yet to be determined, and
we would advise that, until more experience is gained with these MP-RAGE sequences, some caution be observed in interpreting the contrast obtained in view of the transient nature of the tissue magnetization during the acquisition period. A particular advantage of the MP-RAGE technique is that it can be implemented on most current MR imagers with only minor modifications to the machine hardware and software. This is because the gradient strengths and switching times required are much less demanding on the MR system than those required for echo planar imaging. We implemented the sequences on our standard commercial scanner without any hardware additions or modifications. Implementation of the sequences is not restricted to high-field MR systems and such sequences could be implemented on lower field systems as well. However, the intrinsically lower SNR and shorter T, values may be a drawback to the effective use of this technique at low field strength.
REFERENCES Haase, A.; Matthaei, D.; Bartkowski, R.; Diihmke, E.; Leibfritz, D. Inversion recovery snapshot FLASH MR imaging. J. Comput. Assist. Tomogr. 13: 1036-1040; 1989. Haase, A. Snapshot FLASH MRI. Applications to r,, T2, and chemical-shift imaging. Mugn. Reson. Med. 13:77-89; 1990. Unger, E.C.; Cohen, M.S.; Gatenby, R.A.; Clair, M.R.; Brown, T.R.; Nelson, S.J.; McGlone, J.S. Single breath-holding scans of the abdomen using FISP and
476
4.
5.
6.
7.
8. 9.
10.
11.
12.
13.
Magnetic Resonance Imaging 0 Volume 9, Number 4, 1991 FLASH at 1ST. J. Comput. Assist. Tomogr. 12:575583; 1988. Frahm, J.; Haase, A.; Matthaei, D. Rapid NMR imaging of dynamic processes using the FLASH technique. Magn. Reson. Med. 3:321-327; 1986. Kiefer, B.; Deimling, M.; Finelli, D. Ultrafast measurement of TI - and Tz-weighted images with “Snapshot”FLASH. Book of Abstracts, Society of Magnetic Resonance in Medicine 1:367; 1989. Conturo, T.E.; Kessler, R.M.; Beth, A.H. Cooperative T, and T2 effects on contrast using a new driven inversion spin-echo (DISE) MRI pulse sequence. Magn. Reson. Med. 15:397-419; 1990. Mugler, J.P., III; Brookeman, J.R. Optimization of pulse sequence parameters in “Snapshot-FLASH” imaging. Magn. Reson. Imaging 8(Sl):ll5; 1990. Glantz, S.A. Primer of Biostatistics (2nd ed.), New York: McGraw-Hill; 1987. Reinig, J.W.; Dwyer, A.J.; Miller, D.L.; Frank, J.A.; Adams, G.W.; Chang, A.E. Liver metastases: Detection with MR imaging at 0.5 and 1.5 T. Radiology 170:149153; 1989. Foley, W.D.; Kneeland, J.B.; Cates, J.D.; Kellman, G.M.; Lawson, T.L.; Middleton, W.D.; Hendrick, R.E. Contrast optimization for the detection of focal hepatic lesions by MR imaging at 1.5 T. AJR 149:1155-l 160; 1987. Reinig, J.W.; Dwyer, A.J.; Miller, D.L.; White, M.; Frank, J.A.; Sugarbaker, P.H.; Chang, A.E.; Doppman, J.L. Liver metastases detection: Comparative sensitivities of MR imaging and CT scanning. Radiology 162:43-47; 1987. Stark, D.D. Liver. In: Stark, D.D.; Bradley, W.G. (Eds.), Magnetic Resonance Imaging. St. Louis: Mosby; 1988: pp. 934-1054. Bottomley, P.A.; Hardy, C.J.; Argersinger, R.E. The field dependence of T, contrast between normal and pathologic tissue: a review (abstr). In: Book of Abstracts: Society of Magnetic Resonance in Medicine (vol.
2.), Berkley, Calif: Society of Magnetic Resonance in Medicine; 1986: pp. 407-408. 14. Bailes, D.R.; Gilderdale, D. J. ; Bydder, G.M.; Collins, A.G.; Firmin, D.N. Respiratory ordered phase encoding (ROPE): Method for reducing respiratory motion artifacts in MR imaging. J. Comput. Assist. Tomogr. 9: 835-838; 1985. 15. Pattany, P.M.; Phillips,
J.J.; Chiu, L.C.; Lipcamon, J.D.; Duerk, J.L.; McNally, J.M.; Mohapatra, S.N. Motion artifact suppression technique (MAST) for MR imaging. J. Comput. Assist. Tomogr. 11:369-377; 1987. 16. Felmlee, J.P.; Ehman, R.L. Spatial presaturation: A method for suppressing flow artifacts and improving depiction of vascular anatomy in MR imaging. Radiofogy 164:559-564; 1987. 17. Mirowitz, S.A.; Lee, J.K.T.; Brown, J.J.; Eilenberg,
S.S.; Heiken, J.P.; Perman, W.H. Rapid acquisition spin-echo (RASE) MR imaging: A new technique for reduction of artifacts and acquisition time. Radiology 175: 131-135; 1990. 18. Winkler, M.L.; Thoeni, R.F.; Luh, N.; Kaufman,
19.
20.
21.
22.
L.; Margulis, A.R. Hepatic neoplasia: Breath-hold MR imaging. Radiology 170:801-806; 1989. Edelman, R.R.; Hahn, P.F.; Buxton, R.; Wittenberg, J.; Ferrucci, J.T.; Saini, S.; Brady, T.J. Rapid MR imaging with suspended respiration: Clinical application in the abdomen. Radiology 161:125-132; 1986. Mugler, J.P., III; Spraggins, T.A. Improving image quality in snapshot FLASH and 3D MP RAGE sequences by employing reordered phase encoding. Works-in-Progress, Society of Magnetic Resonance in Medicine, 1990. Edelman, R.R.; Wallner, B.; Singer, A.; Atkinson, D.J.; Saini, S. Segmented TurboFLASH: Method for breath-hold MR imaging of the liver with flexible contrast. Radiology 177:515-521; 1990. Mugler, J.P., III; Brookeman, J.R. Three-dimensional magnetization-prepared rapid gradient-echo imaging (3D MP RAGE) Magn. Reson. Med. 15:152-157; 1990.
Imaging. All rights
Vol. 9, pp. 469-476, reserved.
1991 Copyright
0730-725X/91 $3.00 + .OO 0 1991 Pergamon Press plc
l Original Contribution
MAGNETIZATION PREPARED RAPID GRADIENT-ECHO (MP-RAGE) MR IMAGING OF THE LIVER: COMPARISON WITH SPIN-ECHO IMAGING EDUARD E.
P. MUGLER, III, JAMES A. BERTOLINA, SPENCER B. GAY, CYNTHIA L. JANUS, AND JAMES R. BROOKEMAN Departmentof Radiology, University of Virginia Health Sciences Center, Charlottesville,Virginia 22908, USA DE LANGE,
JOHN
We have implemented an MR technique that employs a rapid gradient echo sequence, preceded by magnetization preparation pulses to provide Z’t- and Tz-weighted tissue contrast. With this technique, which can be identified as a member of a new family of pulse sequences, generically named Magnetization Prepared RApid Gradient Echo (MP-RAGE), very short repetition times are used, allowing acquisition times of less than one second and images virtually free of motion-induced artifacts during quiet respiration. Fifteen patients with known liver lesions (metastases, hemangiomas, and cysts) were examined using IfI- and T,-weighted 2-dimensional MP-RAGE sequences, and the images were compared with conventional T,- and multi-echo T,-weighted spin-echo (SE) sequences. Signal difference-to-noise ratios (SD/Ns) of the lesions were calculated for all pulse sequences using corresponding axial images and were normalized for voxel volume. The mean normalized SD/Ns of the MP-RAGE sequences were generally comparable to those for the SE sequences. In addition, there were no noticeable respiratory artifacts on the MP-RAGE images whereas these were clearly present on the T,-weighted SE images and to a lesser degree on the Tr-weighted SE images. It is concluded that the MP-RAGE technique could become an important method for evaluating the liver for focal disease.
Keywords: Magnetic resonance (MR), technology; Magnetic resonance (MR), pulse sequences; Magnetic resonance
(MR), rapid imaging; Liver neoplasms, MR studies; Liver, MR studies.
sequence element. With the 2_dimensional(2D) Magnetization Prepared RApid Gradient Echo (MP-RAGE) sequences, MR images with high tissue contrast can be obtained in extremely short imaging times and virtually without motion artifacts. The technique is therefore particularly useful for imaging the abdomen. To evaluate the use of this method in imaging the liver we have compared it with conventional spin-echo (SE) sequences in a preliminary study of 15 patients with focal liver lesions.
INTRODUCTION Recently, Haase et al. introduced a new pulse sequence technique called snapshot FLASH magnetic resonance (MR) imaging. ‘v2The technique can provide T,- and T2-weighted images with superior image contrast at high field strength. This technique, which has also been called TurboFLASH or snapshot GRASS, is characterized by the combination of two distinct pulse sequence segments. The first segment is used to prepare the magnetization for tissue contrast, and the second, which is a fast gradient-echo sequence, is used for data acquisition. This newly developed pulse sequence technique can be recognized as a member of a larger family of pulse sequences which employ a magnetization preparation-rapid gradient-echo acquisition - magnetization recovery cycle as the basic
METHODS AND MATERIALS The MP-RAGE
Pulse Sequences
Gradient-echo pulse sequences are characterized by the fast data sampling that yields high speed MR images with signal-to-noise ratios (SNRs) comparable to
Requests for reprints and correspondence should be addressed to Eduard E. de Lange, M.D., Department of Radiology, Box 170, University of Virginia Health Sciences Center, Charlottesville, VA 22908, USA.
RECEIVED 1215190; ACCEPTED 3/14/91. We thank Geneva Shifflett, Diana Bowman and Kelly Powell for their secretarial assistance; Carol Chowdhry, Ph.D., for the editorial assistance; and Ursula Bunch for the photography.
Acknowledgments-
469
Magnetic Resonance Imaging 0 Volume 9, Number 4, 1991
470
SE images; however, signal difference-to-noise ratios (SD/Ns) are often lower. 3*4By changing TR, TE, and the flip angle, image contrast can be manipulated, but in general contrast relationships between tissues may be complex, and gradient-echo images are neither T,nor Tz-weighted in the sense of SE imaging.3 A rapid gradient-echo (RAGE) image obtained with a short TR, short TE, and a small flip angle displays contrast that is essentially dominated by the proton density of the tissues.2 When such a RAGE sequence is preceded by pulses to prepare the magnetization [magnetization preparation (MP) pulses], the spindensity contrast can be transformed to Tr-weighted or Tz-weighted contrast. 1,2~5 To acquire a T,-weighted MP-RAGE image, a single, nonselective 180” pulse is used to prepare the magnetization. ’ The 180” inversion pulse rotates the longitudinal magnetization from the positive z-axis to the negative z-axis (Fig. 1). If a time delay follows the inversion pulse to allow Tl relaxation to occur and this delay is followed by the RAGE sequence for the
Inversion
Oelay
image acquisition, the image displays T,inversion recovery contrast. The inversion delay time between the MP pulse and the start of the RAGE sequence basically determines the image contrast. To acquire a T2-weighted MP-RAGE image, a combination of three nonselective pulses (90°-180”90’) is needed to prepare the magnetization.2J The first 90” pulse rotates the magnetization from the positive z-axis to the transverse plane. As with SE imaging, a time interval is allowed for T2 decay to occur and T2-dependent contrast to develop between different tissues (Fig. 2). The 180” pulse is applied to refocus the static field inhomogeneity contributions to the dephasing of the magnetization in the transverse plane (i.e., to perform a spin-echo). When the transverse magnetization is refocused (after twice the interval
Metastasis (4
Liver
Fig. 1. T, -weighted MP-RAGE sequence. (A) To provide Tr contrast, a nonselective 180” pulse is employed for magnetization preparation (MP). After a certain inversion delay, the rapid gradient echo (RAGE) sequence is employed for image acquisition. A spoiler gradient before the RAGE sequence is used to dephase any residual transverse magnetization resulting from an imperfect inversion pulse. (B) The 180” MP pulse rotates the longitudinal component of the magnetization (Mz) to the negative z-axis. During the inversion delay time, T, relaxation of the tissues occurs. (The T, of the liver is shorter than the T, of metastasis.) If the RAGE sequence is employed as the longitudinal magnetization of the metastasis passes through the x-y plane, an image is acquired on which the liver is bright and the metastasis black.
Fig. 2. Tz-weighted MP-RAGE sequence. (A) For the T2 contrast a combination of three nonselective pulses is needed. The first 90” pulse about the x-axis rotates the magnetization to the x-y plane. After time is allowed for T, contrast to develop between tissues, the 180” pulse about the y-axis is applied to refocus &hedephasing spins. When the spin-echo is obtained, a second 90” pulse is applied about the x-axis to “store” the T2 contrast in the form of longitudinal magnetization along the negative z-axis. The interval between the two 90” pulses (TE,,,,) is the period during which T, contrast develops. A spoiler gradient is applied to dephase any residual transverse magnetization resulting from imperfect pulses. (B) After the longitudinal magnetiziation is encoded by T2 contrast and rotated to the negative z-axis, the RAGE sequence is employed for data sampling. However, since Tl relaxation occurs after the longitudinal magnetization is encoded, T, contrast adds to T, contrast. As the T2 of the liver is shorter than that of metastasis, the T,-encoded longitudinal magnetization is less than that of metastasis. If the RAGE sequence is employed as the magnetization of the liver passes through the x-y plane, the liver will be black and the metastasis bright (magnitude reconstruction).
MP-RAGE
MRI of the liver 0 E.E. DE LANGE ET AL.
between the first 90” pulse and the 180” pulse), the second 90” pulse is applied to restore the refocused magnetization to the z-axis, thereby “storing” the T2 contrast of the tissues in the form of longitudinal magnetization. The interval between the two 90” pulses is the period during which the T2 contrast develops between tissues and is labeled TE,,,,. When both 90” pulses are applied about the x-axis (in the rotating frame) and the 180” pulse is applied about the y-axis, the longitudinal magnetization encoded by T2 decay is rotated to the negative z-axis (i.e., driven inversion).6 We found that this technique provides greater contrast between liver lesions and normal liver than does the technique described by Haase in which all preparation pulses are applied about the same axis, causing the longitudinal magnetization to rotate to the positive z-axis (i.e., driven equiIibrium).2 As with the T,-weighted images, a RAGE sequence is employed after the preparation pulses to acquire the image data. MP-RAGE Imaging of the Liver For the 2D MP-RAGE sequences, each image is sampled as a single slice using the RAGE sequence with one acquisition. To obtain clinically useful images of good quality, the acquisition period needs to be long enough to provide adequate SNR. However, if breath-holding is not employed, lengthening the data acquisition period also increases artifacts from motion as sampling of data occurs during the cardiac and respiratory cycles. For our Ti-weighted MP-RAGE images we chose an acquisition time of less than one second to adequately limit motion artifacts during quiet respiration. Employing both theoretical calculations and experimental measurements in normal volunteers,7 we found that an inversion delay of 350 msec, a TR of 7.1 msec, a TE of 3.7 msec, and a flip angle of 10” provided a reasonable compromise for generating strong T,weighted contrast. The flip angle choice represents a trade-off between increased signal-to-noise and contrast-to-noise ratios, and increased artifacts due to the build-up of steady coherences (FLASH type sequence), both of which occur with increased flip angles. We used liver-spleen contrast as a model for liver-lesion contrast in these optimization studies. Also, we employed a rectangular field-of-view (FOV) of 350 x 700 mm with an image matrix of 128 x 256 to obtain isotropic in-plane voxel dimensions. The resulting acquisition time for the RAGE sequence was 909 msec. For the MP-RAGE sequences, data sampling occurs during a Ti-dependent transient and Tl decay, therefore, affects the prepared contrast. This is a particular problem when the goal is to encode T2 contrast,
471
and the T, decay during the RAGE acquisition may destroy the T2 contrast developed with the preparation. However, when T,-weighted MP-RAGE images are obtained with the driven inversion preparation, as is the case with our technique, the T, and T2 decays can complement each other.6 A similar situation arises with the short tau inversion recovery (STIR) technique. To achieve the desired contrast effect from the T2 preparation pulses, the RAGE acquisition time must be substantially shorter than that used for the Ti-weighted images, and must also be approximately the same as or shorter than the T, relaxation time of normal liver. For the T2-weighted MP-RAGE sequence we used a TR of 4.5 msec and only 64 phaseencoding steps to achieve an acquisition time of 288 msec. The TE was 2.3 msec, the flip angle lo”, the matrix 64 x 128, and the FOV 300 x 600 mm. A TE prep of 40 msec was used as this provided essentially no signal from normal liver and relatively high signal from the lesions. Patients and Methods Fifteen patients with known liver lesions were examined with MR imaging using T,- and T2-weighted MP-RAGE pulse sequences with the parameters given in the previous paragraphs. The findings were compared with routine, conventional SE images obtained during the same examination. Ten patients had known metastases from melanoma (n = 1), tumors in the colon or rectum (n = 8), or ovary (n = l), proven either by biopsy or surgery (n = 8), or clinical follow-up and subsequent CT scanning or MR imaging to establish growth of the lesions (n = 2). Four patients had known hemangiomas proved previously on studies, and one patient had multiple liver cysts, proved previously by CT and ultrasound. The studies were performed on a 1.5 Tesla Magnetom 63SP MR imaging system (Siemens Medical Systems, Inc., Iselin, NJ). Ti-weighted SE imaging was performed with TR/TE of 600/15 msec, averaging 4 data acquisitions. The imaging matrix was 128 x 256 and the FOV 420 x 420 mm. The total acquisition time for this sequence was 5.12 min. Multi-echo T2-weighted SE imaging was performed with TR/TE of 2800/40, 80, 120, 160, averaging 2 acquisitions. The matrix was 128 x 256, the FOV was 420 x 420 mm, and the total acquisition time was 11.95 min. Power deposition was monitored for all sequences using the manufacturer’s standard protocol and the power deposition never exceeded the FDA guidelines. Informed consent was obtained for all patients. Scout coronal images were obtained in a multislice mode using a SE 500/15 sequence to determine the off-set for axial acquisitions. All axial images were ob-
472
Magnetic Resonance Imaging 0 Volume 9, Number 4, 1991
tained using a lo-12 mm section thickness. The SE pulse sequences were obtained in a multislice mode of 16 sections with 33% interslice gaps. The MP-RAGE sequences were obtained in a single slice mode using the same 16 slice positions as those for the SE pulse sequences. The SE and the T1-weighted MP-RAGE sequences were obtained in all cases, and the T2weighted MP-RAGE sequence only in 11 of the 15 patients. In some instances additional coronal or sagittal images were obtained using the MP-RAGE sequences. All images were obtained during quiet respiration, and straps were applied across the patient’s abdomen to reduce motion artifacts. No other motion suppression techniques were used. Signal difference-to-noise ratios (SD/Ns) were calculated for all pulse sequences using axial image sections in the same location for each sequence. The axial section that demonstrated the largest dimension of the lesion in the liver was used for SD/N calculations. We chose the image showing the largest extent of the lesion to reduce errors in SD/N measurements from slice misregistration. In patients with multiple lesions, the largest lesion was always chosen for intensity measurements. All lesions evaluated were at least 2.0 cm in diameter. A region of interest (ROI) of at least 60 pixels was used for intensity measurements, the ROI being in the same location on the corresponding image section of each sequence. A region of normal liver of at least 300 pixels was chosen on the same image section as the largest liver lesion, and the inclusion of hepatic vessels was avoided as far as possible. To include the motion artifact contribution to the noise, background noise was measured in a ROI of at least 600 pixels located anterior to the anterior abdominal wall and in line with the given liver lesion along the phaseencoding direction. Contrast between liver masses and normal liver was expressed as an SD/N according to the formula SD/N = 1SI,-S&I /a, where SI, is the mean signal intensity of the lesion, SI, is the mean signal intensity of normal liver, and u is the standard deviation of the background noise. SD/Ns were normalized for voxel volume. The voxel volume normalization was performed by dividing each measured intensity level by the corresponding voxel volume, calculated according to the formula: V= (FOVREAD/MAREAD)(FOVPHASE/ MAPHASE)TH, where V is the voxel volume, MA is the matrix size, and TH is the slice thickness. Mean and standard deviations of the normalized SD/Ns for the four pulse sequences were calculated. The performances of the MP-RAGE and SE sequences were compared by testing for the equality of sample means using a t-test for paired data.’
RESULTS
The SD/Ns normalized for voxel volume are plotted in Fig. 3. The mean SD/N values and standard deviations are also given in Fig. 3. Comparing the T,MP-RAGE to the SE sequences, no statistically significant differences were found except for TE 160 TZ-weighted SE sequence (n = 15, p < .Ol). In this case, there was a statistically significant increase in the mean for the MP-RAGE versus the SE sequence. Comparing the T,-MP-RAGE to the SE sequences, no statistically significant differences were found except for TE 80 Tz-weighted SE sequence (n = 11, p < .Ol). In this case, there was a statistically significant increase in the mean for the SE versus the MPRAGE sequence. In all cases the lesions displayed relatively low signal intensity on T,-MP-RAGE images and high signal intensity on the T,-MP-RAGE images (Figs. 4, 5). There was no noticeable difference in the signal characteristics of the metastases, hemangiomas, or cysts. There was no noticeable blurring from respiratory motion on the MP-RAGE images, although blurring was obvious on the multiecho T,-weighted SE images and to a lesser degree on some of the T,-weighted SE images. DISCUSSION
The conspicuity of focal liver lesions on MR images varies with field strength. At low and intermediate field strength (0.02-0.6 Tesla) the short TR, short TE spin-echo technique is superior in displaying focal liver lesions, a reflection of the significant differences in T,values between focal lesions and liver.‘-” As field strength increases, T,differences between abnormal and normal tissues decrease, 12*13 and lesions may become less conspicuous with short TR, short TE pulse sequences. Studies have shown that at high field strength (1.5 T), conspicuity of focal hepatic disease is superior on long TR, short TE and long TR, long TE pulse sequences. 9,10At high field strength, however, artifacts from the moving anterior abdominal wall during respiration are more prominent than at Iow to mid field strength. ‘* Although signa averaging during short TR, short TE pulse sequences can effectively suppress motion artifacts without increasing imaging times to extreme proportions, the length of the examination makes signal averaging impractical for pulse sequences with long TRs. Therefore, other means of suppressing respiratory artifacts have been used for the long TR techniques. These include methods where the image acquisition is triggered by the respiratory cycle; techniques that reorder the phase-
MP-RAGE MRI of the liver 0 E.E. SD/N
DE LANGE
473
ET AL.
10 0 S
MAN 1.6 SD 2.0
7
(n=15) A
8
MEAN 2.2 SD 1.0 (n=15) A MEAN 1.5 SD 1.5
MEAN 2.3
(n=l5) A A A MEAN 0.8 SD 0.4 (n=ll)
T,-w-&\Cg
SB ma/L6
T*-w-RACII
A
d t.
*
MEAN 1.5 SD 1.3 (n=15) 4
(n=15) t.
a
A
:
4 a
I
2
8
%
8
9E Psw,sa
SE 2wo,*oo
51 zaoo,4a
MEAN 0.0 SD 0.9
: d : 2.
__~
sg zaoo,ml
Fig. 3. SD/N of 15 liver lesions for each pulse sequence normalized for voxel volume.
encoding sequence in relation to the respiratory cycle14; and the use of gradient moment nulling15 and/or spatial presaturation. l6 An alternative strategy in imaging focal liver lesions has been the use of short TR spin-echo”,‘* or gradient-echo3*18**9acquisitions which can be completed in a single breath-hold. Although only a small number of patients was evaluated, our preliminary findings indicated that SD/Ns of the MP-RAGE images obtained in less than one second per slice with one acquisition were in the same range as those of the 4 acquisition T1-weighted and the 2 acquisition Tz-weighted SE images obtained in approximately 5 and 12 min (for 16 slices), respectively (Fig. 3). Therefore, conspicuity of lesions on the fast MP-RAGE images was generally comparable to that for the conventional SE images. One of the drawbacks of our study is that the TR used for the T1weighted SE sequence was relatively long. We used this TR to provide complete coverage of the liver with one sequence. However, we realize that if shorter TRs were used, the SD/N values would probably increase. Also, we used no motion-suppression techniques other than the straps over the abdomen. As our MR imager does not have the capability of respiratory ordered phase encoding, we were not able to use this technique for motion reduction. In addition, we did not use gradient moment nulling to reduce motion artifacts during liver imaging because with this technique the vessels display high signal intensity. As focal lesions usually display similar high signal intensity on TZweighted images, it is more difficult to distinguish the lesions from the vessels, particularly when the lesions are small. However, the Tz-weighted MP-RAGE sequence also displayed the vessels with high signal intensity creating a similar problem for lesion detection.
In this preliminary study, the relatively long TR used for the T,-weighted SE and the absence of motion artifact reduction strategies in the T,-weighted acquisitions represent a bias in favor of the MP-RAGE technique. On the other hand, the MP-RAGE sequences are only in the early stages of development and optimization, and significant improvements would be expected in the future. Thus, we feel that the preliminary findings with the MP-RAGE sequences do speak well for potential applications of the technique. In future studies, it will be important to compare MPRAGE with the other spin-echo and gradient-echo techniques which can be performed within a single breathhold. In our study only relatively large lesions (>2.0 cm) were evaluated to avoid errors in SD/N calculations resulting from the changes in anatomy occurring during respiration. Although no anatomic changes occur when images are obtained during breath holding, we purposely did not use breath holding during the MPRAGE images because the anatomic positions obtained would not necessarily correlate with the SE images obtained during respiration. However, to obtain consecutive slice sections as needed for clinical imaging, breath holding or respiratory triggering would be appropriate. Since small lesions were not evaluated in this study, it is not known whether conspicuity of these lesions is also better on the 7’,-weighted MPRAGE images compared with SE images. Therefore, the sensitivity of MP-RAGE technique in detecting small lesions is still unknown. For the T,-weighted MP-RAGE sequence we used a TE,,,, of 40 msec because it provided high contrast between liver and normal spleen; however, the optimal TE,,,, for liver lesion characterization is yet to be determined, and it is likely
Magnetic Resonance Imaging 0 Volume 9, Number 4, 1991
474
(A)
CD)
(Cl
Fig. 4. MR images of a liver metastasis from colorectal carcinoma. (A) Transverse T,-weighted MP-RAGE image obtained in approximately 900 msec. Using a delay of 350 msec between the 180” magnetization preparation pulse and the start of the RAGE sequence, the liver displays relatively high signal intensity and the lesion (arrow) low signal intensity. There are virtually no motion-induced artifacts from respiration. (B) Corresponding 7”-weighted MP-RAGE image at the same slice position as (A) obtained in approximately 290 msec. Using a driven inversion preparation with a TE,,,, of 40 msec, the normal liver displays low signal intensity and the lesion (arrow) displays high signal intensity. There are no noticeable artifacts from motion. (C) Corresponding conventional SE 600/15 image obtained in approximately 5 min. The lesion (arrow) is almost isointense to the liver. Note increased image degradation from anterior wall motion. (D) SE 2800180 image. Lesion (arrow) displays relatively high signal intensity compared to the surrounding liver.
that two or more TE,,,, values may be needed for tissue characterization. Another problem with the T,weighted MP-RAGE sequence is that the RAGE acquisition occurs during a Tr-dependent transient, and, therefore, the acquisition of T2-weighted images using the MP-RAGE technique is somewhat problematical.
However, modifications to the technique have already which may solve this problem. These include the use of reordered phase-encoding ” and a “multi-shot” approach which acquires the image data by concatenating several prepare-acquire cycles.2’ In addition to modifying the contrast introduced by been proposed
MP-RAGE
MRI of the liver 0 E.E. DE LANGE ET AL.
6%
475
(W
Fig. 5. MP-RAGE images of a hemangioma. (A) Transverse T, -weighted image demonstrates the 4-cm lesion (arrow) which displays low signal intensity. (B) Corresponding T,-weighted image shows the lesion (arrow) displaying high signal intensity. The images were obtained during respiration and there are no noticeable motion artifacts.
the preparation pulses, the TI decay that occurs during the gradient-echo acquisition may also introduce artifacts into the image. Assuming standard phase encoding is employed, the combined effects of T, decay during the acquisition and the acquisition process generally cause a filter function in the associated phaseencoding direction that is real and asymmetric. This results in a point spread function (PSF) that is complex (hermitian). This PSF produces a tissue-dependent blurring and may create edge artifacts at tissue interfaces. We are investigating reordered phase encoding as a means to eliminate the edge artifacts, to alleviate the effects on the image contrast of the acquisition process and T, decay, and to provide increased SD/NS.~’ One drawback of the MP-RAGE technique is that in its present implementation the in-plane resolution is less than that of conventional SE images. Efforts are being made to further improve the sequences without significantly increasing the imaging times. A 3-dimensional data acquisition, as recently described, may improve image quality considerably.22 Meanwhile, since 2D MP-RAGE imaging provides high contrast images in short imaging times and without significant motion artifact, it is probable that this technique may become an important method to evaluate the liver for focal disease. In particular, because of the fast imaging times, the technique may be useful for dynamic imaging of the liver using gadolinium DTPA for tissue enhancement, and we are investigating this method. Nevertheless, the sensitivity and specificity in detecting and characterizing focal hepatic lesions using the MP-RAGE sequences have yet to be determined, and
we would advise that, until more experience is gained with these MP-RAGE sequences, some caution be observed in interpreting the contrast obtained in view of the transient nature of the tissue magnetization during the acquisition period. A particular advantage of the MP-RAGE technique is that it can be implemented on most current MR imagers with only minor modifications to the machine hardware and software. This is because the gradient strengths and switching times required are much less demanding on the MR system than those required for echo planar imaging. We implemented the sequences on our standard commercial scanner without any hardware additions or modifications. Implementation of the sequences is not restricted to high-field MR systems and such sequences could be implemented on lower field systems as well. However, the intrinsically lower SNR and shorter T, values may be a drawback to the effective use of this technique at low field strength.
REFERENCES Haase, A.; Matthaei, D.; Bartkowski, R.; Diihmke, E.; Leibfritz, D. Inversion recovery snapshot FLASH MR imaging. J. Comput. Assist. Tomogr. 13: 1036-1040; 1989. Haase, A. Snapshot FLASH MRI. Applications to r,, T2, and chemical-shift imaging. Mugn. Reson. Med. 13:77-89; 1990. Unger, E.C.; Cohen, M.S.; Gatenby, R.A.; Clair, M.R.; Brown, T.R.; Nelson, S.J.; McGlone, J.S. Single breath-holding scans of the abdomen using FISP and
476
4.
5.
6.
7.
8. 9.
10.
11.
12.
13.
Magnetic Resonance Imaging 0 Volume 9, Number 4, 1991 FLASH at 1ST. J. Comput. Assist. Tomogr. 12:575583; 1988. Frahm, J.; Haase, A.; Matthaei, D. Rapid NMR imaging of dynamic processes using the FLASH technique. Magn. Reson. Med. 3:321-327; 1986. Kiefer, B.; Deimling, M.; Finelli, D. Ultrafast measurement of TI - and Tz-weighted images with “Snapshot”FLASH. Book of Abstracts, Society of Magnetic Resonance in Medicine 1:367; 1989. Conturo, T.E.; Kessler, R.M.; Beth, A.H. Cooperative T, and T2 effects on contrast using a new driven inversion spin-echo (DISE) MRI pulse sequence. Magn. Reson. Med. 15:397-419; 1990. Mugler, J.P., III; Brookeman, J.R. Optimization of pulse sequence parameters in “Snapshot-FLASH” imaging. Magn. Reson. Imaging 8(Sl):ll5; 1990. Glantz, S.A. Primer of Biostatistics (2nd ed.), New York: McGraw-Hill; 1987. Reinig, J.W.; Dwyer, A.J.; Miller, D.L.; Frank, J.A.; Adams, G.W.; Chang, A.E. Liver metastases: Detection with MR imaging at 0.5 and 1.5 T. Radiology 170:149153; 1989. Foley, W.D.; Kneeland, J.B.; Cates, J.D.; Kellman, G.M.; Lawson, T.L.; Middleton, W.D.; Hendrick, R.E. Contrast optimization for the detection of focal hepatic lesions by MR imaging at 1.5 T. AJR 149:1155-l 160; 1987. Reinig, J.W.; Dwyer, A.J.; Miller, D.L.; White, M.; Frank, J.A.; Sugarbaker, P.H.; Chang, A.E.; Doppman, J.L. Liver metastases detection: Comparative sensitivities of MR imaging and CT scanning. Radiology 162:43-47; 1987. Stark, D.D. Liver. In: Stark, D.D.; Bradley, W.G. (Eds.), Magnetic Resonance Imaging. St. Louis: Mosby; 1988: pp. 934-1054. Bottomley, P.A.; Hardy, C.J.; Argersinger, R.E. The field dependence of T, contrast between normal and pathologic tissue: a review (abstr). In: Book of Abstracts: Society of Magnetic Resonance in Medicine (vol.
2.), Berkley, Calif: Society of Magnetic Resonance in Medicine; 1986: pp. 407-408. 14. Bailes, D.R.; Gilderdale, D. J. ; Bydder, G.M.; Collins, A.G.; Firmin, D.N. Respiratory ordered phase encoding (ROPE): Method for reducing respiratory motion artifacts in MR imaging. J. Comput. Assist. Tomogr. 9: 835-838; 1985. 15. Pattany, P.M.; Phillips,
J.J.; Chiu, L.C.; Lipcamon, J.D.; Duerk, J.L.; McNally, J.M.; Mohapatra, S.N. Motion artifact suppression technique (MAST) for MR imaging. J. Comput. Assist. Tomogr. 11:369-377; 1987. 16. Felmlee, J.P.; Ehman, R.L. Spatial presaturation: A method for suppressing flow artifacts and improving depiction of vascular anatomy in MR imaging. Radiofogy 164:559-564; 1987. 17. Mirowitz, S.A.; Lee, J.K.T.; Brown, J.J.; Eilenberg,
S.S.; Heiken, J.P.; Perman, W.H. Rapid acquisition spin-echo (RASE) MR imaging: A new technique for reduction of artifacts and acquisition time. Radiology 175: 131-135; 1990. 18. Winkler, M.L.; Thoeni, R.F.; Luh, N.; Kaufman,
19.
20.
21.
22.
L.; Margulis, A.R. Hepatic neoplasia: Breath-hold MR imaging. Radiology 170:801-806; 1989. Edelman, R.R.; Hahn, P.F.; Buxton, R.; Wittenberg, J.; Ferrucci, J.T.; Saini, S.; Brady, T.J. Rapid MR imaging with suspended respiration: Clinical application in the abdomen. Radiology 161:125-132; 1986. Mugler, J.P., III; Spraggins, T.A. Improving image quality in snapshot FLASH and 3D MP RAGE sequences by employing reordered phase encoding. Works-in-Progress, Society of Magnetic Resonance in Medicine, 1990. Edelman, R.R.; Wallner, B.; Singer, A.; Atkinson, D.J.; Saini, S. Segmented TurboFLASH: Method for breath-hold MR imaging of the liver with flexible contrast. Radiology 177:515-521; 1990. Mugler, J.P., III; Brookeman, J.R. Three-dimensional magnetization-prepared rapid gradient-echo imaging (3D MP RAGE) Magn. Reson. Med. 15:152-157; 1990.
