Phosphorus-31 magnetic resonance metabolite imaging in the human body

Phosphorus-31 magnetic resonance metabolite imaging in the human body

Mogneric Resonance Imaging, Vol. IO, pp. 245-256. Printed in the USA. All rights reserved. 1992 Copyright 0 0730-725x/92 $5.00 + .oo 1992 Pergamon P...

1MB Sizes 0 Downloads 48 Views

Mogneric Resonance Imaging, Vol. IO, pp. 245-256. Printed in the USA. All rights reserved.

1992 Copyright 0

0730-725x/92 $5.00 + .oo 1992 Pergamon Press Ltd.

l Original Contribution

PHOSPHORUS-31 MAGNETIC RESONANCE METABOLITE IMAGING IN THE HUMAN BODY DIETER J. MEYERHOFF, ANDREW A. MAUDSLEY, SAUL SCHAEFER, AND MICHAEL W. WEINER Magnetic Resonance Unit, DVA Medical Center, San Francisco, and Departments of Radiology and Medicine, University of California San Francisco, 4150 Clement St. 1lM, San Francisco, CA 94121, USA This work examines the feasibility of three-dimensional phosphorus-31 magnetic resonance spectroscopic imaging (31P MRSI) of metabolites in the human body using nonselective excitation with a single large circular surface coil for transmitting and receiving. The potential and limitations of this approach to clinical imaging are demonstrated on four selected examples: normal liver and heart, hematoma in the calf, and lymphoma in the groin. The obtained metabolite images showed anatomical detail and allowed differentiation of body organs and pathologic tissue from adjacent tissue. Three-dimensionally localized “P spectra were reconstructed from nominal volumes of 4 to 15 cm3. These spectra showed characteristic resonances and metabolite intensity ratios for the tissue of origin demonstrating good three-dimensional localization. We conclude that surface coil 3*P MRSI of body organs to map metabolite distributions is practically feasible with this approach, but due to experimental limitations, clinical utility requires technical improvements.

Keywords: Magnetic resonance, phosphorus studies; Magnetic resonance, spectroscopy; Magnetic resonance, tissue characterization;

Magnetic resonance,

surface coils: Magnetic resonance,

experimental.

Spatial metabolite distribution in living tissue is best visualized as multi-dimensional images of metabolite levels. Proton magnetic resonance imaging (‘H MRI) achieves contrast using the spin density and relaxation properties of water and fat providing primarily anatomical information. Similarly, contrast is achieved in phosphorus Spectroscopic Imaging (31P SI) using the difference in metabolite concentrations and relaxation properties of 31P nuclei, giving both anatomical and biochemical information. However, in contrast to the superb spatial definition achieved by ‘H MRI, similar definition is difficult to attain using 31P NMR spectroscopy due to the intrinsically lower sensitivity of the 31P nucleus and the lower 31P metabolite concentration compared to water and fat. Therefore, it is important to optimize experimental conditions in order to improve sensitivity of the 31P SI experiment. To increase the sensitivity of detection of 31P NMR spec-

troscopy, we have examined the use of 31P SI experiments’v2 together with surface coils.3 We here demonstrate the feasibility of acquiring 31P metabolite images in the human body to detect spatial variations in metabolite concentrations. In the 31P SI experiments described below, nonselective excitation was used, rather than spatially selective pulses, to minimize T2 losses, which was then followed by three-dimensional (3D) phase encoding to provide spatial discrimination with maximum sensitivity.4 This approach resulted in large four-dimensional (4D) data sets, which required the use of a powerful and flexible display utility for rapid review and analysis of 4D-SI data sets, as well as the display of metabolite images as multiple slices through the imaged organ.5 Three-dimensional SI has previously been described of the human brain, heart, liver, and limbs.6-” The 31P SI results of tissues other than the brain were either displayed as individual spectra obtained from voxels located on a fixed grid6 or as coarse gray-scale

RECEIVED 4/17/91; ACCEPTED 8/21/91. This work is supported by the Department of Veterans Affairs Medical Research Service, NIH grants ROl-DK33923 (MWW) and K08-HL-02131 (SS), the San Francisco Chapter of the American Heart Association, the Elsa U.

Pardee Foundation, and by Philips Medical Systems, Shelton, CT. Address all correspondence to Dieter J. Meyerhoff, Ph.D., Magnetic Resonance Unit, DVA Medical Center (1 lM), 4150 Clement St., San Francisco, CA 94121.

INTRODUCTION

245

246

Magnetic Resonance Imaging 0 Volume 10, Number 2, 1992

pixels of metabolites overlaid on the graphical representation of the organ. I2 Spectroscopic imaging software has been developed in this laboratory9 which reconstructs spectra from any desired volume of interest or metabolite images. This laboratory has used 3D SI with spin-echo acquisition to obtain 3’P spectroscopic images from human brain using a Birdcage coil with a homogeneous Br field-r3 Here we present a selection of early examples of 31P MRSI with surface coils to demonstrate the current power and limitations of this technique. The 3’P MRSI data sets shown were obtained from the normal heart, normal liver, a hematoma in the calf, and a lymphoma in the groin obtained by 3D phase encoding with a single large surface coil for transmitting and receiving. The examples were selected from a total of about 40 3’P SI experiments performed with surface coils in this laboratory. METHODS A 2-Tesla MRI/MRS research system (Philips Gyroscan, Philips Medical Systems, Shelton, CT) was used with a single-turn surface coil of 14 cm diameter in transmit and receive mode. Multi-slice high resolution MRI was performed before every 31P SI examination to obtain slices of the same or very similar thickness as those obtained with the SI study. The spatial correspondence of the two image data sets was previously confirmed by phantom experiments performed under the same experimental conditions as the in vivo experiments. In these phantom experiments, a small glass vial containing hexamethyl phosphoroustriamide (HMPT) was fixed to a 3 L round glass flask filled with 80 mM aqueous Na2HP04 and was included within the imaging region (field-of-view, FOV) for MRI and SI. The phantom experiments also served to confirm the proper performance of the SI acquisition and processing software. The SI pulse sequence consisted of a nonselective excitation pulse followed by a 1.9 ms delay (for the hematoma study a 2.4 ms delay was used due to the small voxel size) prior to acquisition, during which phase encoding was applied in all three directions using a triangular gradient waveform’ with a maximum gradient field strength of 1.3 mT/m. The values for the three dimensions of the FOV were selected for each body part under investigation to include the primary region of interest and surrounding tissue for anatomical references. The number of phase encoding steps (PES) for each dimension was selected to give the minimal nominal voxel size that yielded useful spectra in reasonable measurement times. Note that the effective voxel dimensions are somewhat larger than the nominal voxel dimensions due to the sinusoidal SI point spread function. From experimental

phantom studies and computer simulations, l4 it was determined that a minimum of 10 phase encoding steps is required in each dimension in order to obtain an image quality which is not severely degraded by Gibb’s ringing, as determined by the variation in image intensity across a uniform object. Reduced flip angle excitation was used together with short scan repetition times to maximize sensitivity. The flip angIe was chosen depending on tissue relaxation times and scan repetition times to give optimal signal intensity in the primary area of interest 4 to 7 cm from the surface coil plane (Ernst angle method). The effective FOV covered by the surface coil was laterally (left-right and caudo-cranial) and vertically (anterior-posterior) determined by the sensitivity of the surface coil and the selected excitation pulse length used. This corresponded approximately to a lateral extent of 18 cm and to an anterior-posterior extent of maximally 10 cm. After standard ‘H MRI with the body or head coil, the magnet was shimmed on the tissue water signal of the sensitive volume of the surface coil. Finally, 31P SI data were acquired without moving the subject between MRI and SI. The total examination time for ‘H MRI, shimming, and 3’P SI was about 2 h. SI data acquisition was performed using 256 time domain sample points, a 2000 Hz spectral width, and 4 scans per phase encoding step. Data processing, display, and analysis were performed on a workstation (microVAX, Digital Equipment Co.) using software developed in house.5,‘3 Spectroscopic images were reconstructed from integrated peak areas of magnitude data from spectral regions selected interactively by positioning of two cursors in a typical spectrum of the SI data set. The final SI data set size after processing was 512 points in the spectral domain (after one zero-fill), 32 x 32 spatial points in the axial plane (x and _Y), which were interpolated to twice the data size for image display, and 16 axial planes perpendicular to the surface coil plane (caudo-cranial, z). The most informative of these planes are displayed in the figures below. All ‘H MR images display either a small white rectangular box, indicating the nominal voxel size and position of the displayed spectra, or/and a large white rectangle, indicating the FOV for the 31P MRSI data acquisition in those cases where the FOV was different from the ‘H MRI FOV. No filtering was used for SI data processing in the three spatial dimensions to demonstrate the relatively sharp delineation of tissue boundaries and the achievable spectral signal-to-noise. Spectra shown were reconstructed from the 3D data set and are from single voxels. The tissue volume from which spectra were obtained was larger than determined by the nominal SI voxel size (FOV/#PES) because of the sinusoidal SI point spread function. Spectra are dis-

31PMR metabolite imaging 0 D.J.

played after Lorentzian filtering and individual phasing and are not corrected for the initial missing data points. The resulting baseline distortion, which was most apparent in liver spectra, was removed using a baseline correction routine of the NMRl software (New Methods Research Inc., Syracuse, NY). This routine is a finite automation procedure to delimit spectral data into sections of baseline, upward and downward slope, and top of peaks. The result of this automation and an extrapolation procedure were then used to calculate a synthetic baseline. Spectra of individual studies are displayed with the same vertical scale. Table 1 contains variable acquisition and processing parameters for the presented studies. RESULTS

A T, -weighted, non-gated transverse ‘H MR scout image through the liver of a normal volunteer in supine position is shown in Fig. 1A. The surface coil used for 3’P data acquisition was underneath the subject’s back centered approximately under the right lobe of the liver. The large rectangle superimposed on the proton MRI corresponds to the FOV of 200 x 300 mm2 of the SI metabolite maps, while the small rectangle (labeled 1) indicates a nominal voxel size. The ‘H MR image corresponds in caudo-cranial direction and slice thickness (21 mm) to the axial spectroscopic images shown in Figs. 1B and C, which were obtained using 10 phase encoding steps in anterior-posterior direction and 14 phase encoding steps in left-right direction. Figure 1B shows a spectroscopic image reconstructed from the integrated intensity of the PME/Pi/ PDE region of a liver spectrum (from -1 to +8 ppm relative to hepatic cr-ATP at -9.92 ppm) to emphasize signal coming solely from the liver (Resting muscle contains comparatively little Pi and phosphoesters). Fig. 1C shows the corresponding PCr distribution map of muscle, reconstructed from - 1 to -3 ppm of a mus-

Table 1. Acquisition Parameter FOV in X, y, z (cm) Initial data size in x, y, Z* Nominal voxel size (cm3) Total SI time (min) Excitation angle (“) Scan repetition time (s) Lorentzian spectral filter (Hz)

and processing Liver

20, 30, 30 10, 14, 14 9.2 45 40 0.35 14

*Same as number of phase encoding steps (PES). tCardiac gated.

MEYERHOPF ET AL.

241

cle spectrum. An edge-detected copy of the ‘H MRI is superimposed on the phosphate images to enable easier correlation of images. The high signal intensity in the PME/Pi/PDE spectroscopic image (Fig. 1B) corresponds to the liver region on the MRI, while the spectroscopic image of PCr outlines muscle in the right (left on the MRI) thoracic wall and right back muscle close to the surface coil underneath the subject’s liver. The area of lower signal intensity in the PCr image corresponds to the left back muscle which is at a greater lateral distance to the surface coil. The signal void area between the back muscles corresponds to the spine, where hardly any PCr was detected. No PCr signal was detected in the liver. Individual 31P SI spectra obtained from single voxels without spatial smoothing (nominal voxel size before spatial zero-filling was 9 cm3) within the sensitive region of the coil are displayed in Fig. 2. The spectra are from (A) back muscle, (B) back muscle and liver tissue, and (C) liver tissue alone about 7 cm above the surface coil plane from the nominal voxel indicated on the ‘H MRI. Note the relative high signal-to-noise ratio of spectrum 2C obtained from the liver compared to spectrum 2A from back muscle close to the coil. This is because the excitation pulse length used was optimized for a region about 7 cm above the surface coil in the posterior segment of the right lobe of the liver. The SI data set was acquired with a scan repetition time of 350 ms within 45 min. The relatively high S/N of these spectra indicate that even lower spatial resolution or shorter total acquisition times are possible. Figure 3 shows the result of a gated ‘H MRI and 31P SI study of a normal human heart. Figure 3A is one of four transverse ‘H MR scout images through the heart. The superimposed large rectangle indicates the SI FOV, while the smaller rectangle (labeled 1) indicates the nominal voxel size and the position for one of the spectra. The subject was positioned supine which was more comfortable than the prone position for the

parameters

for the presented

Heart 22, 28, 24 10, 10, 10 14.8 66 50 ca. lt 18

studies Calf

20, 20, 20 12, 12, 10 5.6 40 30 0.4 1

Lymphoma 30, 36, 20 12, 14, 10 12.9 50 36 0.45 4

248

Magnetic Resonance Imaging 0 Volume

10, Number

2, 1992

(A)

09

Fig. 1. (A) Non-gated transverse ‘H MR scout image (TR/TE 450/30) through the liver of a normal volunteer. ‘I‘he superimps Ised large rectangle corresponds to the FOV used for 31P SI data acquisition and display, whi .le the small ret :tangle 31P spectroscopic images reconstructed from (B) the PME 1/P i/PDE (labe :led 1) indicates the nominal voxel size. Corresponding regic m of a liver spectrum and (C) from the PCr resonance of a back muscle spectrum. Superimposec ;1 is the corrz :spconding -detected ‘H MR image of Fig. 1A.

“P MR metabolite imaging 0 D.J.

ATP

from three Fig. 2. Single voxel 31P SI spectra obtained regions: (A) back muscle, (B) back muscle and liver, and (C) liver, the latter region being indicated on the ‘H MR image of Fig. 1A. Nominal voxel size was 9.2 cm3. No smoothing was applied in the spatial domain, while the spectral domain was exponentially multiplied with a 14 Hz line broadening.

long examination time of 2 h. The bright regions on the ‘H MR image were closest to the surface coil which was strapped to the subject’s chest, centered about 3 cm to the left of the subject’s sternum. The 31P spectroscopic images shown in Figs. 3B and C correspond to the FOV outlined by the white rectangle on the ‘H MRI. The SI data set was obtained in 66 min with an excitation pulse length optimized at the location of the septum 50 mm below the surface coil plane. The 31P SI acquisition was gated to 250 ms after the R-wave of the electrocardiogram, in mid-systole, when left ventricular wall thickness would be maximal and signal from intra-cavity blood would be minimal. The scan repetition time was about 1 s in this examination. Figure 3B is composed of the PCr and ATP resonances of the sensitive area of the coil and indicates high signal intensities in the chest, interrupted by a signal void area in the region of the sternum. The highest signal intensity is from the anterior myocardium, left ventri-

MEYERHOFF ET AL.

249

cle, and septum, the latter coinciding with the region where the excitation pulse length was optimized. Note the weak signal intensity in the area of the right ventricular myocardium. The superimposed edge-detected ‘H MRI clearly identifies the region of highest PME signal intensity in Fig. 3C as from the right ventricular chamber. Due to the much greater wall thickness of the left ventricle in systole, the lower PME signal intensity is observed in the left ventricular chamber. The spatial resolution for data acquisition was 2.2 x 2.8 x 2.4 cm3, corresponding to a nominal volume of 15 cm3. This nominal resolution allows spectra from 12 neighboring, nonoverlapping nominal voxels to be reconstructed from the single transverse plane through the heart displayed in Fig. 3A. During 3’P image processing the data set was zero-filled in the spatial domain to 32 x 32 x 16 (x, y, z) voxels and further interpolated to 64 x 64 (x, y) just for transverse image display. Figure 4 shows single voxel spectra obtained without smoothing in the spatial domain and with 18Hz exponential line broadening in the spectral domain. The three spectra are from: (A) ventricular chamber blood, (B) the anterior myocardium, and (C) an area labeled 1 on the ‘H MRI (largely septum). The PCr over fl-ATP peak area ratios obtained from the heart spectra were 1.7 and 1.0, respectively. The blood spectrum showed mainly signal from 2,3-diphosphoglycerate (2,3-DPG). The relatively low PCr signal intensities and PCr/ATP ratios of individual spectra show that myocardium and septum can be clearly distinguished from chest muscle which exhibits much higher relative PCr signal intensities. Figure 5 shows the study of a patient with a 3-weekold sports injury, resulting in a hematoma in the right calf. Figure 5A displays a transverse ‘H MRI slice through the calf muscle at the position of the hematoma obtained with the calf in a standard head imaging coil. In this examination the ‘H MRI FOV corresponded to the FOV of the 31P SI experiment and the superimposed small rectangle indicates the nominal voxel size for “P SI data acquisition. The corresponding slice of the PCr spectroscopic image obtained in 40 min with the surface coil located under the calf is displayed in Fig. 5B. Note the drop-off of signal intensity away from the surface coil, starting at approximately the center of the calf where the excitation pulse length was optimized. As identified by the superimposed edgedetected ‘H MRI, the signal void areas on the PCr image spatially correspond to the femur (top) andmore surprisingly-to the hematoma and an adjacent region below the hematoma that does not look normal on the ‘H MRI. Figure 5C is a ratio image obtained from an image of the PME/Pi/PDE region (+2 to +8 ppm relative to the PCr resonance at 0 ppm) over that

250

Magnetic Resonance Imaging 0 Volume 10, Number 2, 1992

@I

(C)

Fig. 3. (A) Gated transverse ‘H MR image (TR/TE approx. 800/30) through the heart of a normal volunteer. The superimposed large rectangle corresponds to the FOV used for 3’P SI data acquisition and display, while the small rectangle 31P spectroscopic images reconstructed from (B) the sum of PCr (labeled 1) indicates the nominal voxel size. Corresponding and ATP resonances in skeletal muscle and myocardium and (C) from the 2,3-DPG resonance of a chamber blood spectrum. Superimposed on Fig. 3C is the corresponding edge-detected ‘H MR image of Fig. 3A. Note that the brightest signal in the PCr plus ATP image (Fig. 3B) is from left ventricle and the septum, the area to which the excitation pulse length was optimized.

of PCr, in which the high intensity area within the calf spatially corresponds to the center of the hematoma on the ‘H MRI. This low signal-to-noise ratio image and corresponding spectra (see below) show that the hyperintensity in the hematoma region is due to some low

PME/Pi/PDE signal intensity in the absence of PCr in the hematoma. Hyperintensities outside the calf are due to noise. The spectrum in Fig. 6A, reconstructed from the single voxel in the hyperintense hematoma region indicated on the MRI, supports this interpretation

“P MR metabolite imaging 0 D.J.

(4

1

V-9

ATP T

Fig. 4. 3’P SI spectra obtained from single voxels of: (A) right ventricular chamber blood, (B) anterior myocardium just below the chest muscle, and (C) the area indicated by the small rectangle on the ‘H MRI of Fig. 3A, largely comprising septum. Nominal voxel size was 15 cm3. No smoothing was applied in the spatial domain, while the spectral domain was exponentially line broadened with I8 Hz.

in that

no PCr (and ATP) and only very weak resonances in the PME/Pi/PDE region are detected. The absence of PCr and ATP indicate the absence of any viable tissue in the hematoma, while the very low intensity signals between 5 and 7 ppm are most likely due to inorganic phosphate and/or PME, probably 2,3-DPG. While fresh blood gives rise to intense signal from 2,3-DPG (see Fig. 4A from heart chamber blood) the very low intensity of resonances in the PME/Pi/PDE region of the hematoma spectrum may indicate low concentrations of fresh blood, low MR visibility of blood in the hematoma, or clotted blood. The spectrum in Fig. 6B was reconstructed from a single voxel in a normal appearing region of the calf muscle to the right of the fibula about 5 cm above the surface coil (nominal voxel size: 5.6 cm3). Note the splitting of the ATP resonances due to phosphorusphosphorus scalar coupling. In other “P SI studies of the calf of normal subjects, similarly obtained in 40 min with the 14 cm

251

MEYERHOFF ET AL.

surface coil (not shown here), the achieved spatial resolution of nominal 3.5 cm3 (1.8 x 1.4 x 1.4 cm3) allowed resolution of the fibula (1 x 1.4 cm2 on ‘H MR images) as a signal void area on spectroscopic images reconstructed from the PCr resonance. Figure 7 shows the results of a 31P SI study of a lymphoma in the groin of a 40-year-old male patient obtained in 50 min SI acquisition time. The transverse ‘H MRI of this patient is shown in Fig. 7A. The lesions are relatively homogeneous in signal appearance on this T2 weighted ‘H MR image. The FOV for 31P SI is indicated on the ‘H MR image by the large rectangle, while the small rectangle (labeled 1) indicates the nominal voxel size for this examination. The surface coil is located above the major subcutaneous lesion with its center indicated by the rectangular signal area, corresponding to the HMPT filled glass phantom in the surface coil plane center. Phosphorus metabolite images of the tumor are shown in Figs. 7B and 7C. The solid tumor is clearly visible as a homogeneous high intensity region on the image of Fig. 7C reconstructed from the PME/Pi region of a tumor spectrum (from +2 to +6 ppm relative to a-ATP at -9.92 ppm) and as low intensity region on the PCr image of Fig. 7B. The two areas of high intensity on the PCr image correspond to the two main muscle groups adjacent to the tumor within the effective FOV of the surface coil. High S/N spectra could be obtained from four slices through the tumor region, from different regions within the tumor and from healthy tissue adjacent to the tumor. The spectra displayed in Fig. 8 are single voxel spectra obtained without spatial filtering from a nominal voxel of 13 cm3 from (A) pelvic muscle to the right of the tumor (from the voxel position indicated on the ‘H MRI) and (B) the tumor center. Note the absence of a PCr signal from the tumor spectrum and the presence of strong ATP signals from the tumor core, indicating well perfused tissue. The very high PME peak in the tumor spectrum at 4.44 ppm (relative to a-ATP at -9.92 ppm) has been tentatively assigned on the basis of its chemical shift to AMP and/or phosphoethanolamine. Is The relatively low Pi level and the relatively high ATP and PME levels indicate rapidly growing tumor tissue without significant necrosis throughout the tumor. Tumor necrosis was also not consistent with the appearance of the lesion on ‘H MR images. DISCUSSION The examples presented here demonstrate the practical feasibility of obtaining phosphorus metabolite maps of tissues in the body using nonselective pulses and phase encoding in three directions with large surface coils for sending and receiving. Volume definition

252

Magnetic Resonance Imaging 0 Volume 10, Number 2, 1992

(4

09

u

Fig. 5. (A) Transverse ‘H MR scout image (TR/TE 450/30) through the calf of a patient with a hematoma due to a 3-weekold sports injury. The small rectangle in the hematoma region indicates the nominal voxel size for 3’P SI data acquisition. Corresponding 3’P spectroscopic images reconstructed from (B) PCr and from (C) the ratio of PME/Pi/PDE over PCr. The weak resonances between 5 and 7 ppm and the absence of PCr in the hematoma create the appearance of the ratio image. The FOV used for 3’P SI data acquisition and display corresponds to the FOV for the ‘H MR image of Fig. 5A. Superimposed on the metabolite images is the corresponding edge-detected ‘H MR image of Fig. 5A.

in all three directions is obtained by phase encoding only, without the need of additional volume preselection by, for example, slice selection gradients and with the accuracy of lateral volume definition. This ap-

preach provides valuable and spatially accurate spectral (biochemical) information with high sensitivity and visualization of three-dimensional metabolite distributions to assess spatial variations under normal

,‘P MR

metabolite imaging 0 D.J.

MEYERHOFF ET AL.

253

(A) PCr

I

ATP a

Y

P

LEP

I”

(1

IO

xl

PPM

Fig. 6. 3LP SI spectra obtained from single voxels of (A) the hematoma (from the voxel indicated on the ‘H MRI) and (B) healthy muscle to the right of the fibula. Note the very low signal-to-noise ratio of spectrum A obtained from the hematoma (see text). Nominal voxel size was 5.6 cm 3. No smoothing was applied in the spatial domain, while the spectral domain was exponentially line broadened with 1 Hz.

and pathologic conditions. The application of surface coils causes a modulation of the spectroscopic image intensity by the typical sensitivity profile of a surface coil, which is determined by its B1 field and by inhomogeneous partial saturation due to the excitation angle used. Nevertheless, the metabolite images pre-

sented here show anatomical detail over a wide region, which can be directly related to a high-resolution ‘H MR image obtained with standard imaging coils. The subject need not be moved between the MRI and SI

parts of the examination which lasts a total of about 2 h. Useful biochemical information obtained from high signal-to-noise spectra is limited to a smaller region within the sensitive area of the surface coil, though with large surface coils this covers sufficient volumes to enable imaging of large body regions. When the SI data are viewed as an image, due to the viewers’ greater tolerance to lower S/N in images, useful spectroscopic metabolite images can generally be reconstructed from a larger region than can individual spectra. This is illustrated in the normal human heart study: while the metabolite image reconstructed from myocardial PCr and ATP resonances in Fig. 3B shows right ventricular myocardium, the 31P spectra obtained from this region are of very low signal-to-noise. The use of nonselective excitation with a surface coil causes modulation of the intensity of the metabolite images as a function of B1 field and flip-angle response. Methods have been proposed to minimize

inhomogeneous excitation. Bottomley et al. I6 have described the use of a much larger excitation coil and a small receive coil to acquire 2D SI data from the human heart, and Garwood et al. l7 have used B, insensitive pulses for low-angle excitation and short scan repetition times. While many of these or similar pulses for low-angle excitation may be too long for “P SI data acquisition (because of the relatively short spinspin relaxation times for some 31P metabolites), they will find wide application in ‘H SI.” The use of an adiabatic half-passage pulse for one-dimensional spectroscopic imaging and 90” excitation with surface coils was recently presented by Brown et al.19 The delay for phase-encoding between signal excitation and FID acquisition in these SI experiments causes some TT signal losses and also spectral baseline distortion due to Fourier transformation with missing data points. Baseline distortion is especially evident in SI spectra obtained from the liver, which contains large amounts of immobile phospholipids. These contribute to the broad underlying signal (sometimes termed the “hump”) in liver spectra acquired without missing data points. ‘“,20,21For molar quantitation of metabolites from peak areas this distortion requires correction. Different approaches are currently used, such as backwards calculation of missing data points. 1o,22Time domain fitting for spectral analysis is also being investigated to address this problem.22*23 Spin-echo acquisition, which does not suffer from

254

Magnetic Resonance Imaging 0 Volume 10, Number 2, 1992

@I

(Cl

Fig. 7. (A) First echo of a transverse T, weighted ‘H MR image (TR/TE 1800/20/80) through the center of a subcutaneous lymphoma in the groin of a male patient. The superimposed large rectangle corresponds to the FOV used for 3’P SI data acquisition and display, while the small rectangle over the pelvic muscle (labeled 1) indicates the nominal voxel size for 3’P SI data acquisition. Corresponding 31P spectroscopic images reconstructed from (B) the PCr resonance in muscle and from (C) the PME and Pi resonances of a lymphoma spectrum. Superimposed on the metabolite images is the corresponding edgedetected ‘H MR image of Fig. 7A.

missing data points, has been used in this laboratory with a homogenous head coil in brain studies.’ This approach, however, cannot be used with surface coils because of the limited excitation profile of the spin-

echo pulse sequence. Finally, improvements in gradient performance will allow significant reduction of the initial delay period requiring less elaborate baseline corrections procedures.

31P

MR metabolite imaging 0 D.J.

255

MEYERHOFF ET AL.

PCr

(4

d

PME ’

ATP

b pi

3

PDE

Fig. 8. 3’P SI spectra obtained from single voxels of: (A) pelvic muscle, corresponding to the voxel indicated on the ‘H MRI of Fig. 7A, and (B) lymphoma. Nominal voxel size was 13 cm 3. No smoothing was applied in the spatial domain, while the spectral domain was exponentially line broadened with 4 Hz.

Technical limitations exist with the current surface coil 3D SI approach. Therefore, further clinical application requires technical improvements, especially in the field of inhomogeneous signal excitation and reception. There are several clinical situations in which maps of metabolite distributions are potentially useful. For example, studies of tumors would enable visualization of the metabolic profile of the tumor compared to the adjacent normal tissue. Measurements of the spatial variation of 31P metabolites in the human heart could define the anatomic distribution and transmural variation in metabolites during induced myocardial ischemia in patients with coronary artery disease, the extent of infarcted and viable myocardium following of memyocardial infarction,24 and the heterogeneity tabolite levels in global myocardial processes such as left ventricular hypertrophy and/or dilated cardiomyopathy.25,26 Further optimization of surface coil 3D SI acquisition and processing will extend the usefulness of this technique. For example, correction of metabolite maps for the B, sensitivity profile of the surface coil will provide more accurate representation of metabolite distributions, while higher field magnets will reduce imaging times and/or improve spatial resolution. Considering that the same 14 cm surface coil was used in studies of normal human liver”’ and correcting for different volume sizes and slightly different locations within the liver and with respect to the surface coil, the S/N of the 3’P SI liver spectrum of Fig. 2 compares very favorably with liver spectra obtained

with the ISIS technique. 20,2’ It is clear that surface coil 3D SI is currently very useful in obtaining spectral information. In addition, it is a feasible technique for the acquisition of metabolite distribution maps from a variety of human organs and pathologies in clinically acceptable total examination times. Further utilization and technical improvements are necessary before its role in diagnostic imaging can be fully assessed. REFERENCES

1. Brown, T.R.; Kincaid, B.M.; Ugurbil, K. NMR chemical shift imaging in three dimensions. Proc. Natl. Acad. Sci. USA 79:3523-3526; 1982. 2. Maudsley, A.A.; Hilal, SK.; Perman, W.H.; Simon, H.E. Spatially resolved high-resolution spectroscopy by “Four-Dimensional” NMR. J. Magn. Reson. 5 1: 147152; 1983. 3. Meyerhoff, D.J.; Maudsley, A.A.; Twieg, D.B.; Weiner, M.W. Phosphorus metabolite mapping of human heart, liver, and kidney by surface coil spectroscopic imaging. Sot. Magn. Reson. Med. 1:140; 1990. (Abstract) 4. Brunner, P.; Ernst, R.R. Sensitivity and performance time in NMR Imaging. J. Magn. Reson. 3:83-106; 1979. 5. Maudsley, A.A.; Elliott, M.A.; Weiner, M.W. Data display and analysis for 4D spectroscopic imaging. Sot. Magn. Reson. Med. 1:571; 1990. (Abstract) 6 Bottomley, P.A.; Charles, H.C.; Roemer, P.B.; Flamig, D.; Engeseth, H.; Edelstein, W.A.; Mueller, O.M. Human in vivo phosphate metabolite imaging with 31P NMR. Magn. Reson. Med. 7:319-336; 1988. 7. Tropp, J.S.; Sugiura, S.; Derby, K.A.; Suzuki, Y.; Haw-

256

8.

9.

10.

Il.

12.

13.

14.

15.

16.

Magnetic Resonance Imaging 0 Volume 10, Number 2, 1992 rysko, C.; Yamagata, H.; Klein, J.E.; Ortendahl, D.A. ; Kaufman, L.; Acosta, G.F. Characterization of MR spectroscopic imaging of the human head and limb at 2.OT. Radiology 169:207-212; 1988. Bailes, D.R.; Bryant, D.J.; Case, H.A.; Collins, A.G.; Cox, I.J.; Hall, A.S.; Harman, R.R.; Khenia, S.; McArthur, P.; Ross, B.D., Young, I.R. In vivo implementation of three-dimensional phase-encoded spectroscopy with a correction for field inhomogeneity. J. Magn. Reson. 77:&O-470; 1988. Maudsley, A.A.; Twieg, D.B.; Sappey-Marinier, D.; Hubesch, B.; Hugg, J. W.; Matson, G.B.; Weiner, M.W. Spin-echo spectroscopic imaging in the human brain. Magn. Reson. Med. 14:415-422; 1990. Twieg, D.B.; Meyerhoff, D.J.; Hubesch, B.; Roth, K.; Sappey-Marinier, D.; Boska, M.D.; Gober, G.R.; Schaefer, S.; Weiner, M.W. Phosphorus-31 magnetic resonance spectroscopy in humans by spectroscopic imaging: Localized spectroscopy and metabolite imaging. Magn. Reson. Med. 12:291-305; 1989. Vigneron, D.B.; Nelson, S.J.; Murphy-Boesch, J.; Kelley, D.A.C.; Kessler, H.B., Brown, T.R.; Taylor, J.S. Chemical shift imaging of human brain: Axial, sagittal, and coronal P-31 metabolite images. Radiology 177: 643-649; 1990. Cox, I.J.; Bryant, D.J.; Collins, A.G.; George, P.; Harman, R.R.; Hall, A.S.; Hodgson, H.J.F.; Khenia, S.; McArthur, P.; Spencer, D.H.; Young, I.R. Four-dimensional chemical shift MR imaging of phosphorus metabolites of normal and diseased human liver. J. Camp. Assist. Tom. 12(3):369-376; 1988. Hugg, J.W.; Matson, G.B.; Twieg, D.B.; Maudsley, A.A.; Sappey-Marinier, D.; Weiner, M.W. 31P spectroscopic imaging of normal and pathological human brains. Magn. Reson. Imag. 10:227-243; 1992. Maudsley, A.A.; Hugg, J. W.; Fernandez, E. J.; Matson, G.B.; Weiner, M.W. Application of reduced k-space sampling in spectroscopic imaging. (Abstract) Sot. Magn. Reson. Med. 1:186; 1991. Goerenstein, D.C.; Shah, D.O. Appendixes: Selective compilation of 3’P NMR data. In: Goerenstein, D.G. (Ed.). Phosphorus-31 NMR: Principles and applications. London: Academic Press; 1984: 549-591. Bottomley, P.A.; Hardy, C.J.; Roemer, P.B. Phosphate metabolite imaging and concentration measurements in

17.

18.

19.

20.

21.

22.

23.

24.

25.

26.

human heart by nuclear magnetic resonance. Magn. Reson. Med. 141425-434; 1990. Staewen, R.S.; Johnson, A.J.; Ross, B.D.; Parrish, T.; Merkle, H.; Garwood, M. 3D-FLASH imaging using a single surface coil and a new adiabatic pulse, BIR-4. Znvest. Radio/. 25:559-567; 1990. Fernandez, E.J.; Maudsley, A.A.; Higuchi, T.; Weiner, M. W. ‘H spectroscopic imaging of rat brain at 7 Tesla. (Abstract) Experimental NMR Conference, St. Louis, MO; 1991. Brown, T.R.; Buchthal, S.D.; Murphy-Boesch, J.; Nelson, S.J.; Taylor, J.S. A multislice sequence for “P in vivo spectroscopy. ID chemical shift imaging with an adiabatic half-passage pulse. J. Mugn. Reson. 82:629633; 1989. Meyerhoff, D.J.; Karzcmar, G.S.; Matson, G.B.; Boska, M.D.; Weiner, M.W. Non-invasive quantitation of human liver metabolites using image-guided 3 1P magnetic resonance spectroscopy. NMR Biomed. 3:17-22; 1990. Meyerhoff, D.J.; Boska, M.D.; Thomas, A.M.; Weiner, M. W. Alcoholic liver disease: Quantitative image-guided P-31 MR spectroscopy. Radiology 173:393-400; 1989. Barkhuijsen, H.; De Beer, R.; Bovee, W.M.M.J.; van Ormondt, D. Retrieval of frequencies, amplitudes, damping factors, and phases from time domain signals using a linear least squares procedure. J. Magn. Reson. 61:465-474; 1985. Nelson, S.; Brown, T.R. A new method for automatic quantification of lD-spectra with low signal-to-noise ratio. J. Magn. Reson. 75:229-243; 1987. Bottomley, P.A.; Herfkens, R. J.; Smith, L.S.; Bashore, T.M. Altered phosphate metabolism in myocardial infarction: P-31 MR spectroscopy. Radiology 165:703707; 1987. Schaefer, S.S.; Gober, J.R.; Schwartz, G.G.; Twieg, D.B.; Weiner, M.W.; Massie, B. In vivo phosphorus-31 spectroscopic imaging in patients with global myocardial disease. Am. J. Card. 65:1154-1161; 1990. Weiss, R.G.; Bottomley, P.A.; Hardy, C.J.; Gerstenblith, G. Regional myocardial metabolism of highenergy phosphates during isometric exercise in patients with coronary artery disease. New Engl. J. Med. 323: 1593-1600; 1990.