- Email: [email protected]
Fat tissue and fat suppression
Ma~nelk Resonance Printed in the USA,
Imaging. Vol. 11. pp. 385-393, All rights reserved.
1993 Copyright
0
0730-725x/93 WOO + .m 1993 Pergamon Press Ltd.
l Original Contribution
FAT TISSUE AND FAT SUPPRESSION JINTONG MAO,* HONG YAN,~ W. W. BREY,* W.D. Bnx;oo~, JR. ,* J.J. STEINBACH,* AND A. MANCUSG* *Department of Radiology, University of Florida, Gainesville, FL 32610, USA; and tSchool of Electrical Engineering, University of Sydney, NSW 2006, Australia Fat tissues cons&t of fat cells, capillaries, and collagen fibers. In order to completely suppress the signals from fat
tissues in clinical magnetic resonance imaging, the signal from capillaries and collagen flbiersas well as from fat cells should all be suppressed. We have previously reported that fat signal can be uniformly suppressed by applying an optimized presaturation pulse. The inhomogeneously broadened fat peak of tissue spectrum is excited by the optimized pulse and dephased by a subsequent field gradient. The broadened water peak is not affected. In this paper we discuss a technique that suppresses signals from fat tissues completely as well as uniformly. This technique is based on the cancellation of fat and water signals in the same image voxel by combining the optimized selective excitation with the opposite phase imaging technique. Experimental and clinical images demonstrate that the new technique improves the delineation and depiction of anatomy in clinical fat suppression imaging. Keywords: Fat tissue; Fat suppression; Magnetic field inhomogeneity; RF field variation.
INTRODUCTION
voxel by combining the optimized selective excitation with the opposite phase imaging technique.
One of the main problems of all existing techniques for clinical fat suppression is their sensitivity to the magnetic field inhomogeneity. It is known that the inhomogeneity of the magnetic field across a human subject inside a magnet can not be eliminated due to the susceptibility effect. Both water and fat spectral peaks are inhomogeneously broadened.’ This situation worsens in clinical situations because the magnetic field can not be shimmed individually for each patient due to limited scanning time. We have recently reported that the broadened fat peak can be uniformly suppressed by a chemical shift selective (CHESS) sequence with an optimized presaturation pulse, without the requirement of a magnetic field homogeneity higher than that needed for a routine clinical diagnostic MR imaging.’ However, the optimized presaturation technique alone never completely suppresses all signals from the fat tissue because it consists of fat cells, capillaries, and collagen fibers. Capillaries and collagen fibers contribute to the water signals. In this paper we discuss a technique that suppresses signals from fat tissue completely as well as uniformly. This technique is based on the cancellation of fat and water signals in the same image
METHOD Because we try to suppress all the signals from fat tissues, this phenomenon leads us to consider the structure of fat tissues. It is seen from scanning electron micrograph that fat tissue consists of capillaries and other small vessels, collagen fibers, and fat cells, as shown in Fig. 1. An image voxel is usually larger than a fat cell. Thus it contains the capillaries, collagen fibers, and fat cells. In order to suppress the signals from fat tissues, the signals from the capillaries and collagen fibers as well as from the fat cells should all be suppressed. To better visualize the problem involved, three spectra acquired with the same spectroscopic spin echo sequence are shown in Fig. 2. The echo time of the spectroscopic sequence is 15 msec. The in vivo spectrum (A) was obtained with a surface coil from the superficial thick fat tissue of the thigh of an obese volunteer. A Siemens 1.5 T scanner (Erlangen, Germany) was used. A surface coil of diameter of 8 cm was used to localize the fat tissue which is about 8 cm thick. The in vitro spectra (B) and (C) were obtained with a head
RECEIVED8/3/92; ACCEPTED1l/12/92.
Address correspondence 385
to Dr. Jintong Mao.
386
Magnetic Resonance Imaging 0 Volume 11, Number 3, 1993
Fig. 1. Scanning electron micrograph of fat tissues. Capillaries and collagen fibers form a supporting network around fat cells. Usually a voxel for MR imaging is much larger than a fat cell. Thus the voxel contains the capillaries, collagen fibers, and fat cells. (From: Tissues and Organs: A TextAtlas of Scanning Electron Microscopy, by Richard Kessel and Randy Kardon. Copyright 0 1979 by W.H. Freeman and Company. Reprinted by permission.)
coil from cylindrical phantoms of fresh pig fat tissues and pure pig fat, respectively. The diameter and length of the cylindrical phantoms are 7 cm and 15 cm, respectively. The fresh pig fat tissue phantom was carefully constructed to ensure that it was completely made out of fat tissues, without any other tissues, for example, the muscles and visible blood vessels. The pure pig fat phantom was bottled after the fat tissue was fried and the pure liquified lipid substance was extracted from the dregs. The in vitro spectrum (B) is similar to the in vivo spectrum (A). This verifies that the surface coil localization technique used to obtain spectrum (A) is adequate in this situation. Spectrum (C) of Fig. 2 reveals spectral peaks located at two different resonance frequencies. Each broadened peak of the spectrum (C)
consists of several sub-peaks. This can be verified by a high field spectrum of adipose tissue (for instance, a spectrum at 7 T).3 But for the purpose of our discussion at present, it will be sufficient to consider only two peaks at 1.5 T. The left peak of the spectrum (C) is contributed by unsaturated olefinic fat. The right peak is contributed by saturated aliphatic fat. The chemical shift, 6, between these two peaks is about 3.5 ppm, which is equivalent to 225 Hz at 1.5 T. The olefinic fat protons approximately have the same frequency as the water protons of the capillaries and collagen fibers. The integral ratio of the two peaks of spectrum (C) is about 0.12. The integral ratio of the two peaks of spectrum (B) is about 0.39. Since the right peaks of both spectra (B) and (C) represent the same amount of the pure fat protons, the olefinic fat represents about 10.7% of the total pure fat protons, which is derived from the expression 0.12/l. 12, where 0.12 represents the left peak and 1.Orepresents the right peak. The olefinic fat also represents about 8.6% (0.12/l .39) of the total fat tissue protons. Because of the similarity of spectra (A) and (B), we expect these percentages to be approximately the same for the in vivo human fat tissues. The ratio of the integrals of the left and right peaks of Fig. 2A is about 0.45. Thus, the amount of the left peak protons of spectrum (A) is about 31% (0.4511.45) of the total fat tissue protons. The difference between the integral ratios of the in vitro and in vivo experiments (0.39 to 0.45) may be due to protons of the circulating blood volume. These ratios may slightly vary due to different individuals and different locations of the fat tissue in the body. The small variation of the integral ratio around 0.45 does not affect the discussion below. The left peak protons of spectrum (A) include the protons of oiefinic fat (about 8.6%) and the water protons of capillaries and collagen fibers (about 22.4070,which is from 31-8.6%). This conclusion is consistent with the report that subcutaneous adipose tissue has a pure fat content of 77% by weight (the weight of fat fraction/total tissue net weight),4 that is, the water content is about 23% by weight. Our purpose is to suppress all the signals from the fat tissues, that is, to suppress both peaks of spectrum (A), but to preserve the water signals from other tissues, whose signals approximately have the same frequency as the olefinic fat peak. It is clear now that the optimized presaturation pulse may only uniformly excite the right peak magnetization of the spectrum (A), which will be destroyed by a subsequent dephasing gradient, but the left peak remains. Thus, the fat signals can be uniformly suppressed, but can not be completely suppressed by only applying a 90” optimized presaturation pulse. In this article fat peak excitation
Fat tissue and fat suppression 0 J. MAO ET
387
AL.
(A)
300
T
250
g 200 \ r-l E 150 G? .r(
m
100.
50.
1
600.
I
400.
,
I
200.
FrequE&
I
-200. / Hz
I
-400.
,
-600.
Fig. 2. Spectra of fat tissues obtained with a spectroscopic spin echo sequence at 1.5 T. The echo time is 15 msec. (A) In vivo spectrum obtained with a surface coil from the superficial thick fat tissue of the thigh of an obese volunteer. (B) and (C) In vitro spectra of fresh pig fat tissue and pure extracted fat, respectively. The left peak of (A) and (B) is from the proton signal of capillaries, collagen fibers, and olefinic fat. The left peak of(C) is from olefinic fat protons. The right peak of each spectrum is from the aliphatic fat protons. (Figure continues on following page.)
388
Magnetic Resonance Imaging 0 Volume 11, Number 3, 1993
900.
800.
700. : > 600.E
‘500.H 2 ,~400.cn 300.-
9
600.
I
400.
I
I
I
I
-400.
-600.
Fig. 2. continued.
means a fat magnetization rotation of a certain flip angle. In the following discussion the left and right peaks are referred to the spectrum (A) of Fig. 2. The left peak is contributed from the protons of olefinic fat, capillaries, and collagen fibers. The right peak is contributed from the aliphatic fat protons. The left peak magnetization can be cancelled by a component of the right peak magnetization if it is selectively rotated by an optimized presaturation pulse of about 117’, which is from 180” - arccos0.45. After about 117’ rotation from the z-direction, the right peak magnetization is resolved into two components. One has the opposite direction and the same magnitude as the left peak magnetization and the other has a transverse orientation as shown in Fig. 3. The left peak magnetization is cancelled by the opposite component of right peak magnetization since they are from the same voxel. This conclusion was proven by fat suppression imaging experiment, using a spin-echo imaging sequence of long TR between successive fat signal excitations to ensure a sufficient recovery of right peak fat magnetization between the excitations. It is possible to completely suppress the fat signal by adequately setting the pulse amplitude if the TR between successive excitations of the right peak is long enough, for instance TR = 1.Oset or longer [The longitudinal relaxation time (T,) of fat tissue is about 284 msec at 1.5 T].4 However, a pulse
sequence with a long repetition time between successive fat excitations may not be very useful for clinical imaging. We need consider a pulse sequence with much short repetition time between the successive fat excitations. In multislice imaging (seven sections or more) with short repetition time (for instance TR is about 500 msec) the situation is different. It is impossible to completely
LEFT
RIGHT 6%
0%
Fig. 3. (A) Simplified vector representation of the spectrum of Fig. 2A. The left and right peak magnetizations from the same voxel are labelled “LEFT” and “RIGHT”, respectively. (B) In spin-echo imaging with long repetition time, the zcomponent of RIGHT has the same magnitude and the opposite direction of LEFT after a 117’ rotation. LEFT is cancelled by the z-component of RIGHT. The xy-component of RIGHT will be eliminated by a subsequent dephasing gradient.
Fat tissue and fat suppression 0 J. MAO ET AL.
suppress the fat signal in multislice imaging with a short TR by only adjusting the pulse amplitude. The reason is the steady state effect of the right peak magnetization, since the actual repetition time between successive excitations of the aliphatic fat peak is very short and the aliphatic fat magnetization is not sufficiently recovered before next excitation. In the above example the repetition time between the successive fat excitations is about 78.6 msec, which is much shorter than the longitudinal relaxation time of fat tissue. In order to suppress the signal from fat tissue completely in multislice imaging with short TR, the technique of changing the amplitude of an optimized presaturation pulse must be supplemented by another technique, as described below. The opposite phase imaging technique proposed by Dixon’ may be chosen as the supplemental technique. The left- and right peak fat magnetizations are in phase and constructively interfere immediately after a 90” excitation or at the maximum of a Hahn spin echo since they are from the same voxel. The magnetizations are out of phase and destructively interfere after a time period, to = l/26, where 6 is the chemical shift between the two peaks. At 1.5 T, to = 2.25 msec. In this way, the in- and opposite-phase images are obtained.5 However, the water signals from capillaries and collagen fibers of fat tissue have not been considered in current publications of fat suppression study.4-9 Wehrli et al.‘j and Chan et al.’ considered the suppression of the signal from olefinic fat protons, but not the signals from capillaries and collagen fibers. The cancellation between water and fat magnetizations was discussed in these papers only for the junctional zone between subcutaneous fat and other tissues which primarily contain water, but not for the subcutaneous fat tissue itself.
389
Chan et al.’ used the opposite phase imaging technique to eliminate the signal from the olefinic fat protons. We have found that the same concept of opposite phase imaging can be used in a short TR, multislice sequence to suppress the signals from capillaries and collagen fibers, as illustrated in Fig. 4. The z-component of the steady-state right peak fat magnetization can have the same magnitude as the left peak fat magnetization by adjusting the amplitude of the presaturation pulse such that it is equivalent to a flip angle of about 63” (arccos0.45) for the steady-state magnetization of the right peak. Thus, the signals of all of the components of fat tissue can be suppressed completely and uniformly by combining the techniques of optimized selective presaturation and opposite phase imaging. RESULTS In our clinical practices, the same setting of the flip angle of the optimized selective presaturation pulse has been used with both surface and volume coils and in different body areas for different patients. No individual flip angle adjustment is necessary. Imaging operation procedure for fat suppression is the same as for regular imaging. The only needed adjustment is for the amplitudes of the RF pulses of 63”, !XY’,and 180”. This is usually accomplished by “autotune” capability of commercial system. Two examples, Figs. 5 and 6, illustrate the comparison of the images acquired by the optimized presaturation pulse with and without opposite phase imaging technique. In these two figures, (A)-(D) were obtained only with the optimized presaturation pulse; (E)-(H) were obtained by the combined technique of the optimized pulse with the opposite phase technique. Images (A) and (E), similarly images (B) and (F), and so on, are from the same slice. The
RIGHT
LEFr (4
@I
(C)
(D)
Fig. 4. (A) Simplified vector representation of the spectrum of Fig. 2A, the same as Fig. 3A. (B) RIGHT is in steady state when a train of presaturation pulses with short TR is applied. The z-component of RIGHT can be adjusted to the magnitude of LEFT by setting the flip angle of the presaturation pulse to about 63”. The xy-component of RIGHT will be eliminated by a subsequent adequate dephasing field gradient. (C) The whole LEFT magnetization and the z-component of RIGHT are turned to transverse plane by a 90” slice selective pulse. (D) After the time period to, LEFT and the z-component of RIGHT (now they are in xy plane) have precessed to the opposite direction and they cancel each other since they originate from the same voxel.
390
391
Fat tissue and fat suppression 0 J. MAO ET AL.
coronal head images of Fig. 5 are obtained from the adjacent slices of optic nerve of a volunteer by a head coil. The fat signals are suppressed, except for some regions near the nasal cavity and the teeth where high susceptibility effect causes a strong field distortion. The optic nerve delineation is not affected by the very high susceptibility effect. The overall inhomogeneity of the image intensity is caused by the inhomogeneity of the RF field. The contrast enhanced axial lumbar spine images of Fig. 6 are obtained from a patient by a spine surface coil. The scan parameters are the same for the images of Fig. 5 and Fig. 6: TR = 420 msec, number of slice = 7, NEX = 2, slice thickness = 4 mm, matrix = 192 x 256, TE = 15 msec for the inphase images and TE = 17.25 msec for the opposite phase images. Both optic nerve and lumbar spine are well depicted on the fat suppressed images. The images obtained by the combined technique show the improvement when all of the components of fat tissue are both uniformly and completely suppressed. All the images were scaled to emphasize the background noise level. The signal intensity in the fat suppressed area, for instance the subcutaneous fat of the lumbar region and the area around the optic nerve, is comparable with the background noise in the images obtained by the combined technique. Our clinical images have shown that the quality of the fat suppressed images is maintained within a moderate range of repetition time. There is no obvious change of the images when the repetition time is changed, for instance, from 420 to 550 msec. Thus the combined technique has improved the quality of the fat suppressed images. DISCUSSION
It should be noted that the in vitro spectrum (B) and (C) of Fig. 2 from pig fat tissue and fried pig fat were obtained at room temperature of 22” C while in vivo spectrum (A) was obtained at body temperature of 36.5”C. The temperature difference affects the relaxation time. However, it does not affect the above conclusion about the water and fat quantitative constitution of the fat tissue since the repetition time of the spin-echo spectroscopic sequence was long enough (TR = 3 set or longer) to allow full recovery of magnetization. Some confusion may occur related to our fat suppression technique of cancelling the water signal by the fat signal. Although the water signals from collagen fiber and capillaries of fat tissues are eliminated in our discussion, no water signals from other tissues are suppressed if some other tissues locate in the same image voxel as the fat tissues. The water signal cancellation
Magnetic Resonance Imaging 0 Volume 11, Number 3, 1993
392
0)
G-0
Fig. 6. Contrast enhanced T,fat suppressed axial lumbar spine images acquired by a surface spine coil with the same experimental conditions: TR = 420 msec, slice thickness = 4 mm, NEX (number of excitations) = 2, matrix size = 192 x 256, TE = 15 ms for image (A)-(D) and TE = 17.25 msec for (E)-(H). (A)-(D) Four adjacent lumbar spine images with fat suppression by an optimized presaturation pulse. The overall inhomogeneity of the image intensity is caused by inhomogeneity of the RF field. (E)-(H) Four adjacent lumbar spine images with fat uniformly and completely suppressed. The signal intensity of the fat suppressed subcutaneous area is comparable with the background noise in (E)-(H) and it is less affected by the inhomogeneity of the RF field, although the signal intensity of the muscle has been affected by the RF field inhomogeneity.
Fat tissue and fat suppression 0 J. MAO ETAL.
occurs only in the fat tissue itself due to its water and fat constitution. Thus, a pathological process or infiltrative lesion in fat tissue-will not be obscured by elimination of water signal of the fat tissues. Our technique works well in general clinical situations. However, a few exceptions exist where air-tissue or bone-tissue susceptibility differences generate a very large local field gradient. For instance in the regions near the nasal cavity and the teeth, the fat signal is not uniformly suppressed, as shown in Fig. 5, although this local high distortion does not affect the delineation of the optic nerve. The multiple point Dixon techniques have similar difficulty. *-lo However, the multipoint Dixon technique has the disadvantages of extra imaging time and low S/N efficiency. Our technique requires only one image acquisition and does not lower the S/N ratio, and may be clinically more efficient. For a spin echo sequence, the opposite-phase image is obtained by simply delaying the signal acquisition and the read-out gradient the time period to. The situation is different for a gradient-echo sequence. Since the phase relation between the two components at the echo time for the gradient echo sequence varies according to the specific gradient echo sequence under consideration, the acquisition time delay needs to be determined individually for each specific gradient echo sequence. This situation is clearly described by Wehrli et aL6 For the gradient echo sequence, the slice-selective pulse usually is not a 90” pulse. Thus, both vectors of Fig. 4B may only represent the transverse components of fat tissue magnetizations. The complete cancellation of the left and right peak magnetizations needs to be further investigated for gradient echo sequences.
Acknowledgment-This Foundation.
393 work is supported
by The Whitaker
REFERENCES Schneider, E.; Glover, G.H. Rapid in vivo proton shim-
5. 6.
7.
8.
9.
10.
ming. Magn. Reson. Med. 18:335; 1991. Mao, J.; Yan, H.; Bidgood, Jr., W.D. Fat suppression with an improved selective presaturation pulse. Magn. Reson. Imaging 10:49; 1992. Maynor, C.H.; Charles, H.C.; Herfkens, R.J.; Suddarth, S.A.; Johnson, G.A. Chemical shift imaging of atherosclerosis at 7.0 Tesla. Invest. Radiol. 24:52; 1989. Szumowski, J.‘; Simon, J.H. Proton chemical shift imaging. In: D.D. Stark, W.G. Bradley @is). Magnetic Resonance Imaging, 2nd ed. St. Louis: Mosby Year Book; 1992: p. 479. Dixon, W.T. Simple proton spectroscopic imaging. Radiology 153:189; 1984. Wehrli, F.W.; Perkins, T.G.; Shimakawa, A.; Roberts, F. Chemical shift-induced amplitude modulations in images obtained with gradient refocusing. Magn. Reson. Imaging 5:157; 1987. Chan, T.W.; Listerud, J.; Kressel, H.Y. Combined chemical-shift and phase-selective imaging for fat suppression: Theory and initial clinical experience. Radiology 181: 41; 1991. Glover, G.H.; Schneider, E. Three-point Dixon technique for true water/fat decomposition with B0 inhomogeneity correction. Magn. Reson. Med. 18:371; 1991. Glover, G.H. Multipoint Dixon technique for water and fat proton and susceptibility imaging. .I. Magn. Reson. Zmag. 1:521; 1991. Haacke, E.M., Patrick, J.L., Lenz, G.W., Parrish, T. The separation of water and lipid components in the presence of field inhomogeneities. Rev. Magn. Reson. Med. 1:123; 1986.
Imaging. Vol. 11. pp. 385-393, All rights reserved.
1993 Copyright
0
0730-725x/93 WOO + .m 1993 Pergamon Press Ltd.
l Original Contribution
FAT TISSUE AND FAT SUPPRESSION JINTONG MAO,* HONG YAN,~ W. W. BREY,* W.D. Bnx;oo~, JR. ,* J.J. STEINBACH,* AND A. MANCUSG* *Department of Radiology, University of Florida, Gainesville, FL 32610, USA; and tSchool of Electrical Engineering, University of Sydney, NSW 2006, Australia Fat tissues cons&t of fat cells, capillaries, and collagen fibers. In order to completely suppress the signals from fat
tissues in clinical magnetic resonance imaging, the signal from capillaries and collagen flbiersas well as from fat cells should all be suppressed. We have previously reported that fat signal can be uniformly suppressed by applying an optimized presaturation pulse. The inhomogeneously broadened fat peak of tissue spectrum is excited by the optimized pulse and dephased by a subsequent field gradient. The broadened water peak is not affected. In this paper we discuss a technique that suppresses signals from fat tissues completely as well as uniformly. This technique is based on the cancellation of fat and water signals in the same image voxel by combining the optimized selective excitation with the opposite phase imaging technique. Experimental and clinical images demonstrate that the new technique improves the delineation and depiction of anatomy in clinical fat suppression imaging. Keywords: Fat tissue; Fat suppression; Magnetic field inhomogeneity; RF field variation.
INTRODUCTION
voxel by combining the optimized selective excitation with the opposite phase imaging technique.
One of the main problems of all existing techniques for clinical fat suppression is their sensitivity to the magnetic field inhomogeneity. It is known that the inhomogeneity of the magnetic field across a human subject inside a magnet can not be eliminated due to the susceptibility effect. Both water and fat spectral peaks are inhomogeneously broadened.’ This situation worsens in clinical situations because the magnetic field can not be shimmed individually for each patient due to limited scanning time. We have recently reported that the broadened fat peak can be uniformly suppressed by a chemical shift selective (CHESS) sequence with an optimized presaturation pulse, without the requirement of a magnetic field homogeneity higher than that needed for a routine clinical diagnostic MR imaging.’ However, the optimized presaturation technique alone never completely suppresses all signals from the fat tissue because it consists of fat cells, capillaries, and collagen fibers. Capillaries and collagen fibers contribute to the water signals. In this paper we discuss a technique that suppresses signals from fat tissue completely as well as uniformly. This technique is based on the cancellation of fat and water signals in the same image
METHOD Because we try to suppress all the signals from fat tissues, this phenomenon leads us to consider the structure of fat tissues. It is seen from scanning electron micrograph that fat tissue consists of capillaries and other small vessels, collagen fibers, and fat cells, as shown in Fig. 1. An image voxel is usually larger than a fat cell. Thus it contains the capillaries, collagen fibers, and fat cells. In order to suppress the signals from fat tissues, the signals from the capillaries and collagen fibers as well as from the fat cells should all be suppressed. To better visualize the problem involved, three spectra acquired with the same spectroscopic spin echo sequence are shown in Fig. 2. The echo time of the spectroscopic sequence is 15 msec. The in vivo spectrum (A) was obtained with a surface coil from the superficial thick fat tissue of the thigh of an obese volunteer. A Siemens 1.5 T scanner (Erlangen, Germany) was used. A surface coil of diameter of 8 cm was used to localize the fat tissue which is about 8 cm thick. The in vitro spectra (B) and (C) were obtained with a head
RECEIVED8/3/92; ACCEPTED1l/12/92.
Address correspondence 385
to Dr. Jintong Mao.
386
Magnetic Resonance Imaging 0 Volume 11, Number 3, 1993
Fig. 1. Scanning electron micrograph of fat tissues. Capillaries and collagen fibers form a supporting network around fat cells. Usually a voxel for MR imaging is much larger than a fat cell. Thus the voxel contains the capillaries, collagen fibers, and fat cells. (From: Tissues and Organs: A TextAtlas of Scanning Electron Microscopy, by Richard Kessel and Randy Kardon. Copyright 0 1979 by W.H. Freeman and Company. Reprinted by permission.)
coil from cylindrical phantoms of fresh pig fat tissues and pure pig fat, respectively. The diameter and length of the cylindrical phantoms are 7 cm and 15 cm, respectively. The fresh pig fat tissue phantom was carefully constructed to ensure that it was completely made out of fat tissues, without any other tissues, for example, the muscles and visible blood vessels. The pure pig fat phantom was bottled after the fat tissue was fried and the pure liquified lipid substance was extracted from the dregs. The in vitro spectrum (B) is similar to the in vivo spectrum (A). This verifies that the surface coil localization technique used to obtain spectrum (A) is adequate in this situation. Spectrum (C) of Fig. 2 reveals spectral peaks located at two different resonance frequencies. Each broadened peak of the spectrum (C)
consists of several sub-peaks. This can be verified by a high field spectrum of adipose tissue (for instance, a spectrum at 7 T).3 But for the purpose of our discussion at present, it will be sufficient to consider only two peaks at 1.5 T. The left peak of the spectrum (C) is contributed by unsaturated olefinic fat. The right peak is contributed by saturated aliphatic fat. The chemical shift, 6, between these two peaks is about 3.5 ppm, which is equivalent to 225 Hz at 1.5 T. The olefinic fat protons approximately have the same frequency as the water protons of the capillaries and collagen fibers. The integral ratio of the two peaks of spectrum (C) is about 0.12. The integral ratio of the two peaks of spectrum (B) is about 0.39. Since the right peaks of both spectra (B) and (C) represent the same amount of the pure fat protons, the olefinic fat represents about 10.7% of the total pure fat protons, which is derived from the expression 0.12/l. 12, where 0.12 represents the left peak and 1.Orepresents the right peak. The olefinic fat also represents about 8.6% (0.12/l .39) of the total fat tissue protons. Because of the similarity of spectra (A) and (B), we expect these percentages to be approximately the same for the in vivo human fat tissues. The ratio of the integrals of the left and right peaks of Fig. 2A is about 0.45. Thus, the amount of the left peak protons of spectrum (A) is about 31% (0.4511.45) of the total fat tissue protons. The difference between the integral ratios of the in vitro and in vivo experiments (0.39 to 0.45) may be due to protons of the circulating blood volume. These ratios may slightly vary due to different individuals and different locations of the fat tissue in the body. The small variation of the integral ratio around 0.45 does not affect the discussion below. The left peak protons of spectrum (A) include the protons of oiefinic fat (about 8.6%) and the water protons of capillaries and collagen fibers (about 22.4070,which is from 31-8.6%). This conclusion is consistent with the report that subcutaneous adipose tissue has a pure fat content of 77% by weight (the weight of fat fraction/total tissue net weight),4 that is, the water content is about 23% by weight. Our purpose is to suppress all the signals from the fat tissues, that is, to suppress both peaks of spectrum (A), but to preserve the water signals from other tissues, whose signals approximately have the same frequency as the olefinic fat peak. It is clear now that the optimized presaturation pulse may only uniformly excite the right peak magnetization of the spectrum (A), which will be destroyed by a subsequent dephasing gradient, but the left peak remains. Thus, the fat signals can be uniformly suppressed, but can not be completely suppressed by only applying a 90” optimized presaturation pulse. In this article fat peak excitation
Fat tissue and fat suppression 0 J. MAO ET
387
AL.
(A)
300
T
250
g 200 \ r-l E 150 G? .r(
m
100.
50.
1
600.
I
400.
,
I
200.
FrequE&
I
-200. / Hz
I
-400.
,
-600.
Fig. 2. Spectra of fat tissues obtained with a spectroscopic spin echo sequence at 1.5 T. The echo time is 15 msec. (A) In vivo spectrum obtained with a surface coil from the superficial thick fat tissue of the thigh of an obese volunteer. (B) and (C) In vitro spectra of fresh pig fat tissue and pure extracted fat, respectively. The left peak of (A) and (B) is from the proton signal of capillaries, collagen fibers, and olefinic fat. The left peak of(C) is from olefinic fat protons. The right peak of each spectrum is from the aliphatic fat protons. (Figure continues on following page.)
388
Magnetic Resonance Imaging 0 Volume 11, Number 3, 1993
900.
800.
700. : > 600.E
‘500.H 2 ,~400.cn 300.-
9
600.
I
400.
I
I
I
I
-400.
-600.
Fig. 2. continued.
means a fat magnetization rotation of a certain flip angle. In the following discussion the left and right peaks are referred to the spectrum (A) of Fig. 2. The left peak is contributed from the protons of olefinic fat, capillaries, and collagen fibers. The right peak is contributed from the aliphatic fat protons. The left peak magnetization can be cancelled by a component of the right peak magnetization if it is selectively rotated by an optimized presaturation pulse of about 117’, which is from 180” - arccos0.45. After about 117’ rotation from the z-direction, the right peak magnetization is resolved into two components. One has the opposite direction and the same magnitude as the left peak magnetization and the other has a transverse orientation as shown in Fig. 3. The left peak magnetization is cancelled by the opposite component of right peak magnetization since they are from the same voxel. This conclusion was proven by fat suppression imaging experiment, using a spin-echo imaging sequence of long TR between successive fat signal excitations to ensure a sufficient recovery of right peak fat magnetization between the excitations. It is possible to completely suppress the fat signal by adequately setting the pulse amplitude if the TR between successive excitations of the right peak is long enough, for instance TR = 1.Oset or longer [The longitudinal relaxation time (T,) of fat tissue is about 284 msec at 1.5 T].4 However, a pulse
sequence with a long repetition time between successive fat excitations may not be very useful for clinical imaging. We need consider a pulse sequence with much short repetition time between the successive fat excitations. In multislice imaging (seven sections or more) with short repetition time (for instance TR is about 500 msec) the situation is different. It is impossible to completely
LEFT
RIGHT 6%
0%
Fig. 3. (A) Simplified vector representation of the spectrum of Fig. 2A. The left and right peak magnetizations from the same voxel are labelled “LEFT” and “RIGHT”, respectively. (B) In spin-echo imaging with long repetition time, the zcomponent of RIGHT has the same magnitude and the opposite direction of LEFT after a 117’ rotation. LEFT is cancelled by the z-component of RIGHT. The xy-component of RIGHT will be eliminated by a subsequent dephasing gradient.
Fat tissue and fat suppression 0 J. MAO ET AL.
suppress the fat signal in multislice imaging with a short TR by only adjusting the pulse amplitude. The reason is the steady state effect of the right peak magnetization, since the actual repetition time between successive excitations of the aliphatic fat peak is very short and the aliphatic fat magnetization is not sufficiently recovered before next excitation. In the above example the repetition time between the successive fat excitations is about 78.6 msec, which is much shorter than the longitudinal relaxation time of fat tissue. In order to suppress the signal from fat tissue completely in multislice imaging with short TR, the technique of changing the amplitude of an optimized presaturation pulse must be supplemented by another technique, as described below. The opposite phase imaging technique proposed by Dixon’ may be chosen as the supplemental technique. The left- and right peak fat magnetizations are in phase and constructively interfere immediately after a 90” excitation or at the maximum of a Hahn spin echo since they are from the same voxel. The magnetizations are out of phase and destructively interfere after a time period, to = l/26, where 6 is the chemical shift between the two peaks. At 1.5 T, to = 2.25 msec. In this way, the in- and opposite-phase images are obtained.5 However, the water signals from capillaries and collagen fibers of fat tissue have not been considered in current publications of fat suppression study.4-9 Wehrli et al.‘j and Chan et al.’ considered the suppression of the signal from olefinic fat protons, but not the signals from capillaries and collagen fibers. The cancellation between water and fat magnetizations was discussed in these papers only for the junctional zone between subcutaneous fat and other tissues which primarily contain water, but not for the subcutaneous fat tissue itself.
389
Chan et al.’ used the opposite phase imaging technique to eliminate the signal from the olefinic fat protons. We have found that the same concept of opposite phase imaging can be used in a short TR, multislice sequence to suppress the signals from capillaries and collagen fibers, as illustrated in Fig. 4. The z-component of the steady-state right peak fat magnetization can have the same magnitude as the left peak fat magnetization by adjusting the amplitude of the presaturation pulse such that it is equivalent to a flip angle of about 63” (arccos0.45) for the steady-state magnetization of the right peak. Thus, the signals of all of the components of fat tissue can be suppressed completely and uniformly by combining the techniques of optimized selective presaturation and opposite phase imaging. RESULTS In our clinical practices, the same setting of the flip angle of the optimized selective presaturation pulse has been used with both surface and volume coils and in different body areas for different patients. No individual flip angle adjustment is necessary. Imaging operation procedure for fat suppression is the same as for regular imaging. The only needed adjustment is for the amplitudes of the RF pulses of 63”, !XY’,and 180”. This is usually accomplished by “autotune” capability of commercial system. Two examples, Figs. 5 and 6, illustrate the comparison of the images acquired by the optimized presaturation pulse with and without opposite phase imaging technique. In these two figures, (A)-(D) were obtained only with the optimized presaturation pulse; (E)-(H) were obtained by the combined technique of the optimized pulse with the opposite phase technique. Images (A) and (E), similarly images (B) and (F), and so on, are from the same slice. The
RIGHT
LEFr (4
@I
(C)
(D)
Fig. 4. (A) Simplified vector representation of the spectrum of Fig. 2A, the same as Fig. 3A. (B) RIGHT is in steady state when a train of presaturation pulses with short TR is applied. The z-component of RIGHT can be adjusted to the magnitude of LEFT by setting the flip angle of the presaturation pulse to about 63”. The xy-component of RIGHT will be eliminated by a subsequent adequate dephasing field gradient. (C) The whole LEFT magnetization and the z-component of RIGHT are turned to transverse plane by a 90” slice selective pulse. (D) After the time period to, LEFT and the z-component of RIGHT (now they are in xy plane) have precessed to the opposite direction and they cancel each other since they originate from the same voxel.
390
391
Fat tissue and fat suppression 0 J. MAO ET AL.
coronal head images of Fig. 5 are obtained from the adjacent slices of optic nerve of a volunteer by a head coil. The fat signals are suppressed, except for some regions near the nasal cavity and the teeth where high susceptibility effect causes a strong field distortion. The optic nerve delineation is not affected by the very high susceptibility effect. The overall inhomogeneity of the image intensity is caused by the inhomogeneity of the RF field. The contrast enhanced axial lumbar spine images of Fig. 6 are obtained from a patient by a spine surface coil. The scan parameters are the same for the images of Fig. 5 and Fig. 6: TR = 420 msec, number of slice = 7, NEX = 2, slice thickness = 4 mm, matrix = 192 x 256, TE = 15 msec for the inphase images and TE = 17.25 msec for the opposite phase images. Both optic nerve and lumbar spine are well depicted on the fat suppressed images. The images obtained by the combined technique show the improvement when all of the components of fat tissue are both uniformly and completely suppressed. All the images were scaled to emphasize the background noise level. The signal intensity in the fat suppressed area, for instance the subcutaneous fat of the lumbar region and the area around the optic nerve, is comparable with the background noise in the images obtained by the combined technique. Our clinical images have shown that the quality of the fat suppressed images is maintained within a moderate range of repetition time. There is no obvious change of the images when the repetition time is changed, for instance, from 420 to 550 msec. Thus the combined technique has improved the quality of the fat suppressed images. DISCUSSION
It should be noted that the in vitro spectrum (B) and (C) of Fig. 2 from pig fat tissue and fried pig fat were obtained at room temperature of 22” C while in vivo spectrum (A) was obtained at body temperature of 36.5”C. The temperature difference affects the relaxation time. However, it does not affect the above conclusion about the water and fat quantitative constitution of the fat tissue since the repetition time of the spin-echo spectroscopic sequence was long enough (TR = 3 set or longer) to allow full recovery of magnetization. Some confusion may occur related to our fat suppression technique of cancelling the water signal by the fat signal. Although the water signals from collagen fiber and capillaries of fat tissues are eliminated in our discussion, no water signals from other tissues are suppressed if some other tissues locate in the same image voxel as the fat tissues. The water signal cancellation
Magnetic Resonance Imaging 0 Volume 11, Number 3, 1993
392
0)
G-0
Fig. 6. Contrast enhanced T,fat suppressed axial lumbar spine images acquired by a surface spine coil with the same experimental conditions: TR = 420 msec, slice thickness = 4 mm, NEX (number of excitations) = 2, matrix size = 192 x 256, TE = 15 ms for image (A)-(D) and TE = 17.25 msec for (E)-(H). (A)-(D) Four adjacent lumbar spine images with fat suppression by an optimized presaturation pulse. The overall inhomogeneity of the image intensity is caused by inhomogeneity of the RF field. (E)-(H) Four adjacent lumbar spine images with fat uniformly and completely suppressed. The signal intensity of the fat suppressed subcutaneous area is comparable with the background noise in (E)-(H) and it is less affected by the inhomogeneity of the RF field, although the signal intensity of the muscle has been affected by the RF field inhomogeneity.
Fat tissue and fat suppression 0 J. MAO ETAL.
occurs only in the fat tissue itself due to its water and fat constitution. Thus, a pathological process or infiltrative lesion in fat tissue-will not be obscured by elimination of water signal of the fat tissues. Our technique works well in general clinical situations. However, a few exceptions exist where air-tissue or bone-tissue susceptibility differences generate a very large local field gradient. For instance in the regions near the nasal cavity and the teeth, the fat signal is not uniformly suppressed, as shown in Fig. 5, although this local high distortion does not affect the delineation of the optic nerve. The multiple point Dixon techniques have similar difficulty. *-lo However, the multipoint Dixon technique has the disadvantages of extra imaging time and low S/N efficiency. Our technique requires only one image acquisition and does not lower the S/N ratio, and may be clinically more efficient. For a spin echo sequence, the opposite-phase image is obtained by simply delaying the signal acquisition and the read-out gradient the time period to. The situation is different for a gradient-echo sequence. Since the phase relation between the two components at the echo time for the gradient echo sequence varies according to the specific gradient echo sequence under consideration, the acquisition time delay needs to be determined individually for each specific gradient echo sequence. This situation is clearly described by Wehrli et aL6 For the gradient echo sequence, the slice-selective pulse usually is not a 90” pulse. Thus, both vectors of Fig. 4B may only represent the transverse components of fat tissue magnetizations. The complete cancellation of the left and right peak magnetizations needs to be further investigated for gradient echo sequences.
Acknowledgment-This Foundation.
393 work is supported
by The Whitaker
REFERENCES Schneider, E.; Glover, G.H. Rapid in vivo proton shim-
5. 6.
7.
8.
9.
10.
ming. Magn. Reson. Med. 18:335; 1991. Mao, J.; Yan, H.; Bidgood, Jr., W.D. Fat suppression with an improved selective presaturation pulse. Magn. Reson. Imaging 10:49; 1992. Maynor, C.H.; Charles, H.C.; Herfkens, R.J.; Suddarth, S.A.; Johnson, G.A. Chemical shift imaging of atherosclerosis at 7.0 Tesla. Invest. Radiol. 24:52; 1989. Szumowski, J.‘; Simon, J.H. Proton chemical shift imaging. In: D.D. Stark, W.G. Bradley @is). Magnetic Resonance Imaging, 2nd ed. St. Louis: Mosby Year Book; 1992: p. 479. Dixon, W.T. Simple proton spectroscopic imaging. Radiology 153:189; 1984. Wehrli, F.W.; Perkins, T.G.; Shimakawa, A.; Roberts, F. Chemical shift-induced amplitude modulations in images obtained with gradient refocusing. Magn. Reson. Imaging 5:157; 1987. Chan, T.W.; Listerud, J.; Kressel, H.Y. Combined chemical-shift and phase-selective imaging for fat suppression: Theory and initial clinical experience. Radiology 181: 41; 1991. Glover, G.H.; Schneider, E. Three-point Dixon technique for true water/fat decomposition with B0 inhomogeneity correction. Magn. Reson. Med. 18:371; 1991. Glover, G.H. Multipoint Dixon technique for water and fat proton and susceptibility imaging. .I. Magn. Reson. Zmag. 1:521; 1991. Haacke, E.M., Patrick, J.L., Lenz, G.W., Parrish, T. The separation of water and lipid components in the presence of field inhomogeneities. Rev. Magn. Reson. Med. 1:123; 1986.