Anatomy and Metabolism of the Normal Human Brain Studied by Magnetic Resonance at 1.5 Tesla

Transcription

1 Paul A. Bottomley, Ph.D. Howard R. Hart Jr., Ph.D. William A. Edelstein, Ph.D. John F. Schenck, Ph.D. L. Scott Smith, Ph.D. William M. Leue, B.A. Otward M. Mueller, Ph.D. Roland W. Redington, Ph.D. Anatomy and Metabolism of the Normal Human Brain Studied by Magnetic Resonance at 1.5 Tesla Proton magnetic resonance () images were obtained of the human head in magnetic fields as high as 1.5 Tesla (T) using slotted resonator high radio-frequency 0 (RF) detection coils. The images showed no RF field penetration problems and exhibited an ii (±1)-fold improvement in signal-to-noise ratio over a.i2-t imaging system. The first localized phosphorus 31, carbon 13, and proton chemical shift spectra recorded with surface coils from I the head and body in the same instrument showed relative concentrations of phosphorus metabolites, triglycerides, and, when correlated with proton images, negligible lipid ( - CH2 - ) signal from brain tissue on the time scale of the imaging experiment. Sugar phosphate and phosphodiester concentrations were significantly elevated in the head compared with muscle. This method should allow the combined assessment of anatomy, metabolism, and biochemistry in both the normal and diseased brain. Index terms: Head. magnetic resonance studies, #{149} Magnetic resonance #{149} Magnetic resonance, chemical shift Radiology 1984; 150: R ECENT advances in magnetic resonance () techniques offer two new probes for the detection and diagnosis of disease states. First, proton can provide images of internal human anatomy with excellent soft tissue contrast because of the large differences in relaxation times (1-8). Second, localized phosphorus 3i and carbon 13 chemical shift spectroscopy allows direct access to metabolic processes for the assessment of damaged tissue and its response to therapy (9-14). To date, medical applications of chemical shift spectroscopy have been limited to experimental animals and the limbs of humans because of the severe demands spectroscopy makes of magnet technology, i.e.,the magnetic field strengths used are in excess of 1.4 Tesla (T), with homogeneity better than one part per million (ppm) across the region of examination (15). Conversely, the magnet requirements for proton imaging are much less stringent, and head and body imaging systems typically operate at field strengths below 0.5 T (5). It has been widely speculated that head and body proton imaging is not feasible above about 0.5 T because of the dual problems of radio-frequency (RF) field penetration/phase distortion and coil design at these frequencies (4, i6-20). This implies that imaging and spectroscopy are incompatible. Here we demonstrate that high field proton imaging of the human head at 1.5 T is possible, along with the first localized natural abundance phosphorus 31 and carbon 13 spectra of the head recorded from the same instrument depicting relative metabolite levels in the normal brain (21). Estimates of RF field amplitude, phase shift, and power deposition in the brain based on a homogeneous tissue cylinder model (20) are presented, and the proton imaging signal-to-noise ratio measured at 1.5 T and 0.12 T. METHODS I From the General Electric Corporate Research and Development Center, Schenectady, NY. Received July 21, 1983; accepted and revision requested Oct. 5; revision received Oct. 20. jc Imaging All images and spectra were recorded from healthy volunteers using a 1-rn bore, 1.5-T Oxford Instruments superconducting magnet and a broadband spectrometer operating at 16.1 MHz, 25.6 MHz, and 63.9 MHz for the carbon 13, phosphorus 31, and proton ( H) resonances, respectively. Volunteers experienced no ill effects during or after the experiments. The coil used for head imaging was an enlarged (25-cm-diameter) version of the slotted tube resonator employed by Alderman and Grant for proton decoupling in a standard small-bore, 300-MHz spectrometer (22). The use of conventional half turn saddle coil geometries for head imaging are precluded above about 50 MHz by their self resonance properties (23). The proton images (Fig. 1) are 4-mm-thick transverse sections recorded using our modified spin-warp spin echo imaging method with a partial saturation RF pulse sequence (6), 0.2-second pulse Sequence repetition period (TR), and 8 ms between ir/2 and r RF pulses 441

2 - - g. ri ii.v r.t, r.-.e.t R Ild I fll,1 ri t h ru Ii t l 1ie,d rtrdt d,#{236} t I. 5-cm n te rv, Is. \c pt tore, which.i -to1d m.iiiti.itioii tfel hvin ( ct:(n I the (.lr(tid.trte rie 10 (i t,#{236}l),,oij d md f. which in cm ip.rt. c.uii tini I ()2 cnd using. ).2- cond p.irti.ii.itur.i t It fl U I.t..t. t. 1.R fl it. ( h ). I ig& c()!l1t of 2S() \ 2o. I -m ni \ I -ui iii 4-rn ii p1 \ k, a nd.i niiitii d md iicrr ti d or litid nonuniform tin Radiology February 1984

3 Figure 2 CYLINDER RADIUS r0 (m) at 0.10 CYUNDER RADIUS r0 (m) 0) Relative applied RF magnetic field amplitude (a), and phase shift (b) at the center of a homogeneous cylinder of brain tissue of radius r0 as a function of frequency computed using an earlier model (20). I (Kro) is a complex zeroth #{149} order modified Bessel function of the first kind, K is the complex wave number in the tissue, and (Kro) is given by the anctangent of the ratio of the imaginary to the real components of I. The amplitude is normalized to 1.0 at the cylinder surface. Figure 3 H20 PD, PCr Head Head j3.atp Choline I I I I I I ppm ppm ppm a. b. c. a. Hydrogen spectra from the entire human head and thigh recorded using the head imaging coil and a single free induction decay (FID). Peaks correspond to water and alkyl (-CH2-) groups. Chemical shifts are in parts per million (ppm) relative to water. b. Phosphorus 31 spectra from a surface coil placed over the temple and the calf muscle. Chemical shifts are relative to phosphocreatine (PCr). Fifty FIDs repeated at 16-second intervals were averaged for the head spectrum and 20 for the calf muscle spectrum. SP sugar phosphates and 2,3-diphosphoglycerate in blood; Pi inorganic phosphate and 2,3 diphosphoglycerate in blood; PD phosphodiesters; a-, f3-, y-atp = a-,.y-, and y- phosphates of adenosine triphosphate. The broad hump in the head spectrum derives from tissue phospholipids and mineral phosphates in bone. C. Carbon 13 spectrum recorded from a surface coil placed over the temple, averaging 2,000 FIDs repeated at 0.5 seconds. Chemical shifts are relative to tetramethylsilane. RCOOR carboxyl groups; -C=C- double-bonded carbons, glycerol C,,2,3 triglyceride carbons. (TE 16 ms). The displayed images are the amplitudes of the two-dimensional Fourier transformed (FT) spin echo data. The scan time was 102 seconds and the data array size consisted of 256 x 256 independent picture points corresponding to i-mm X i-mm resolution in the imaging plane. Such resolution is exemplified in the magnified section through the carotid arteries (Fig. ie). The proton imaging signal-to-noise ratio (,ti) was measured from a onedimensional projection of a small tube of water placed in the head coil. The noise introduced from the head is accounted for by multiplying the observed signal-to-noise ratio by the square root of the ratio of undamped (head absent) to damped (head present) coil quality factors (Q). With this method, the damped signal-to-noise ratio of the system for a i-ml sample in a 1-Hz bandwidth is given by (24): ;t (Lv).Q (damped) 1/2 crv Q (undamped) Hz112.ml (1) where S is the signal amplitude, i the standard deviation of the noise, K the number of points in the projection of the sample tube containing volume V ml of water, and Lv Hz the frequency interval per point. Volume 150 Number 2 Radiology #{149} 443

4 Spectroscopy Phosphorus 31 and carbon 13 spectra were recorded using two 6.5-cmdiameter, tuned, flat spiral surface coils, each serving as both transmitter and receiver. The spatial selectivity of these coils is approximately confined to a volume subtended by the coil circumference and one radius deep from the center of the coil (5, 9). Phosphorus 31 and carbon 13 spectra of the head were obtained after 13 and 16 minutes averaging, respectively, with the surface coils pressed against the temples at the same level seen in Figure ig, thereby intersecting an approximately 70-cc volume of brain and surface tissue. Phosphorus 31 spectra from the calf muscle were recorded in 5 minutes and were excited with 90- second RF pulses repeated every 16 seconds (TR) and Fourier transformed with a 20-ms time constant exponential filter. Carbon 13 spectra were recorded with 0.5-second pulse spacing and a 20-ms filter. This choice of pulse repetition periods ensured little distortion of spectra due to the differing T1 relaxation times of the component peaks, at considerable expense to the signalto-noise ratio (6). Brain T,s were about 3.5 seconds for phosphocreatine (PCr); 1 second for adenosine triphosphate (ATP), phosphodiester (PD), and sugar phosphate (SP) phosphorus 31 peaks; and second for the alkyl (-CH2-) carbon-13 peaks. Phosphoflis 31 and carbon 13 spectra are hydrogen coupled. The RF pulse length was adjusted to roughly a r pulse at the surface, thereby minimizing signal contributions from surface tissue (9). The magnet was shimmed for optimum homogeneity using the proton signal observed from the head with the same surface coils (1 0) and resistive shim gradient magnetic field coils. Relative concentrations of metabolites are estimated from the areas under the corresponding peaks after baseline deconvolution (9). Imaging RESULTS Using equation (1), the damped signal-to-noise ratio (/ ) obtained at 64 MHz was 11,000 (±1,000) Hz 12 m1, compared with 1,000 (±100) Hz1 2 m1 from our 5.1-MHz system (24), which utilizes a saddle-shaped RF coil design. Plots of the RF magnetic field amplitude and phase shift at the center of homogeneous brain tissue cylinders of radii ro up to 0.2 m as a function of frequency, computed assuming a homogeneous tissue cylinder model (20), are depicted in Figure 2. The amplitude is given by [Io(Kro)J where Jo is a modified Bessel function of the first kind and zeroth order, and K, the complex wave number defined in Bottomley and Andrew (20), is a function of tissue resistivity and the dielectric constant. Rat brain tissue resistivity and dielectric constant dispersions were used for calculating the curves (20). The phase shift is the arctangent of the ratio of the imaginary and real components of Jo. The model gives only a qualitative indication of RF amplitude and phase distortions. It assumes an RF field directed coaxially to the tissue cylinder, rather than the transverse case appropniate here (17). Power deposition in the head is proportional to the square of both the frequency and cylinder radius in the cylindrical model (25). The maximum (surface) estimated specific absorption rate in imaging experiments is about.1 W/kg from the curves described in Bottomley and Edeistein (25), well below the suggested 2 W/kg of the Bureau of Radiological Health guideline. Spectroscopy Figure 3a shows proton spectra obtamed from a single free induction decay (FID) recorded with the imaging head coil from the entire human head (top) and the thigh (lower). The two well-resolved peaks correspond to protons in water and alkyl groups (5). The field homogeneity across the whole head is about 0.5 ppm, as judged by the line width of the water peak. Phosphorus 31 and carbon 13 surface coil spectra from the head and calf muscle are depicted in Figure 3b and c, assuming peak assignments from anima! and human limb studies (9-14). Chemical shifts are in parts per million relative to phosphocreatine and tetramethylsilane in phosphorus 31 and carbon 13 spectra, respectively. The broad hump in the phosphorus 31 head spectrum centered at approximately 6 ppm probably derives from tissue phospholipids and mineral phosphates in the skull. In Figure 3b, the ratios of phosphorus 31 metabolite concentrations for the head are approximately [$-ATP]/[PCr].86 ±.2,[PD]/[PCr] =.87 ±.2, [Pij/[PCr].26 ±.15, [SP]/ [PCr] =.53 ±.2, compared with muscle values of.4 ±.08,.07 ±.1,.1 ±.07, and.12 ±.1, respectively. Imaging DISCUSSION If the spectrometer system noise contributions and dielectric losses in the sample are minimized, the frequency dependence of the signalto-noise ratio is given by ii (aa2 v1 2 +$v2 b5) 2 (2) where a and b are the coil and sample radii, respectively; a and /3 are constants; the term in the numerator reflects the frequency dependence of the signal; and the two terms in the denominator represent noise contributions from the coil and eddy current (inductive) losses in the sample, respectively (26). Some additional frequency dependence is introduced by the dispersion of tissue resistivities (20). For brain, this causes approximately a 1.8-fold increase in /3 between 5 and 64 MHz, while for muscle the. change in /3 is negligible (20). Since patient inductive losses are essentially irreducible, improvements in the signal-to-noise ratio can be achieved only by increasing v (or field strength) or reducing the coil noise. The coil noise may be considered optimum when its noise contribution is insignificant compared with the contribution of the body. The change in Q in the damped and undamped coil (assuming negligible dielectric losses) gives an indication of this. For example, the unloaded 64-MHz slotted tube resonator Q of 224 was reduced to 25 when loaded with the head, implying that the eddy-current losses in the head were eight times the losses in the coil. Under such conditions, the first term in the denominator of equation (2) may be neglected and the predicted frequency dependence of 4 is roughly linear. However, at 5.1 MHz, the observed losses in the head were only 0.4 times the coil losses. Thus the observed 11-fold improvement in i/i between 5.1 MHz and 64 MHz is in fair accord with equation (2). Note that the principal noise source in an optimized imaging system is ideally the sample rather than the instrument regardless of operating frequency. Images showed no discernible artifacts attributable to RF magnetic field penetration/attenuation effects or phase distortion associated with induced eddy currents, in qualitative agreement with calculations based on the homogeneous tissue cylinder model (20). The curves (Fig. 2) show approximately a 20% increase rather than attenuation in amplitude at the center at 60 MHz, assuming a maximum brain cylinder radius of about.075 m estimated from Figure 1. The phase shift is about 18#{176}. The increase in amplitude would not produce power deposition hot spots near the center because the flux linking eddy current 444 #{149} Radiology February 1984

5 loops decreases rapidly as the area of the loops falls to zero (20, 25). In practice, tissue interfaces should further ameliorate these problems by restnicting the radius of areas subtended by the eddy current paths. Such considerations, along with our use of the amplitude spectra to avoid the phase distortion as suggested earlier (20), also support a reduced effect on whole body scans obtained at these frequencies. Examination of published proton spin lattice relaxation times (T,s) mdicates no significant degradation of T1 contrast among normal and diseased tissues at high frequencies (7, 8, 27-28). Since 1 increases with field strength, the contrast-to-noise ratio is still increasing for most tissues at 64 MHz (8). The trend toward longer T,s may, however, necessi tate extended repetition periods for T1 imaging pulse sequences. Spectroscopy The proton alkyl peak in the thigh spectrum (Fig. 3a) is usually attributed to tissue lipids. Its virtual absence in the head spectrum is both intriguing and fortuitous. The brain is rich in lipid, which constitutes about 16% of white matter and 6% of gray matter (29), and this difference in composition has been linked to the excellent discnimination between gray and white matter observed in images of the brain (30). Our results do not support the role of lipid as a direct gray/white matter contrast discniminant in images, but lipid concentrations may indirectly influence contrast by altening the water proton relaxation times. Moreover, it is likely that the alkyl proton spin-spin relaxation times (T2) in the brain are sufficiently short (<5 ms) so that they do not contribute to the signal acquired during imaging schemes such as ours (6) that employ spin echo pulse sequences. This is consistent with an extremely broad proton alkyl peak in Figure 3a, and the absence of any chemical shift ghost artifacts (31) in the brain due to this peak in Figure 1. Some ghost artifact from subcutaneous lipid is, however, detectable at the periphery of the images upon magnification, and this component may, in fact, be entirely responsible for the broad alkyl peak in the proton spectrum. Surface tissue may therefore also be the likely source of the alkyl peaks in the carbon 13 spectrum (Fig. 3c). Phosphorus 31 spectra recorded from the left and right sides of the head and from 10 other normal subjects show no significant variations. However, different levels of choline and glycerol were observed in carbon 13 spectra from another individual, again suggestive of a varying amount of surface tissue contamination of brain spectra. Sugar phosphates and phosphodiesters occur at significantly elevated levels in the phosphorus 31 head spectra relative to those of muscle (9, 13) where these peaks are barely detectable. The measured ratios of phosphorus 31 metabolites in the head spectra agree with biochemical assays from rat brains (13). Higher sugar phosphate levels reflect the greater energy demands of the brain, supplied through glycolysis. Phosphodiesters such as phosphatidylethanolamine and lecithin (phosphatidylcholine) are major constituents of myelin in the brain (32) and are metabolically active. Some variation in PD amplitude might therefore occur in demyelinating disorders and in the development of the juvenile brain. CONCLUSION These experiments demonstrated the feasibility of high field imaging, its compatibility with in vivo spectroscopy, and the superior signal-tonoise ratios that may be realized. High-resolution imaging of the major arteries offers hope for the detection and quantitation of atherosclerosis and heart disease. Ratios of PCr and ATP to inorganic phosphate provide a quantitative assessment of tissue health, varying with ischemia (12, 13), cancerous tissue (33), metabolic disorders (11), and drug therapy (10). Intensities of glycerol, choline, and fatty acid carbon 1 3 resonances (determined from alkyl, carboxyl, and doublebonded carbon peaks) yield information about the relative contributions of tissue triglycerides and membranes (14). The use of carbon 13-labeled components gives access to other metabolic pathways such as the deposition and mobilization of glycogen (14, 34). Improvements to the system that we envision include installation of threedimensional or simultaneous multiple section FT imaging (5); three-dimensional FT chemical shift spectroscopic imaging (5, 35, 36), which obviates the necessity for spatial localization with surface coils; and whole-body high field strength imaging. Acknowledgments: We thank S. Karr, D. Vatis, M. O Donnell, R. B. Harms, W. Adams, J. Piel, R. Argersinger, D. R. Eisner, and A. Argesta for their valuable assistance. General Electric Corporate Research and Development Center P.O. Box 8 Schenectady. NY References 1. Doyle FH, Gove JC, Pennock JM, et al. Imaging of the brain by nuclear magnetic resonance. Lancet 1981 ; 2: Bydder GM, Steiner RE, Young JR. et al. Clinical N imaging of the brain: 140 cases. AJR 1982; 139: Alfidi RJ. Haaga JR. El Ynusef SJ. et al. Work in progress. Preliminary experimental results in humans and animals with a superconducting, whole-body, N scanner. Radiology 1982; 143: Crooks L, Arakawa M, Hoenninger J. et al. Work in Progress. Nuclear magnetic resonance whole-body imager operating at 3.5 KGauss. Radiology 1982; 143: Bottomley PA. N imaging techniques and applications: a review. Rev Sci Instrum 1982; 53: Edelstein WA, Bottomley PA. Hart HR. Smith LS. Signal, noise, and contrast in nuclear magnetic resonance (N) imaging. J Comp Assist Tomogr 1983; 7: Ling CR. Foster MA, Hutchison JMS. Comparison of N water proton T1 relaxation times of rabbit tissues at 24 MHz and 2.5 MHz. Phys Med Biol 1980; 25: Hart HR, Bottomley PA, Edelstein WA. et al. Nuclear magnetic resonance imaging: contrast to noise ratio as a function of strength of magnetic field. AJR (in press). 9. Ackerman JJH, Grove TH, Wong GG. Gadian DG, Radda GK. Mapping of metabolites in whole animals by 31P N using surface coils. Nature 1980; 283: Nunnally RL, Bottomley PA. Assessment of pharmacological treatment of myocardial infarction by phosphorus-31 N with surface coils. Science 1981; 211: , 11. Ross BD, Radda GK, Gadian DG, Rocker G, Esini M, Falconer-Smith J. Examination of a case of suspected McArdles syndrome by 31P nuclear magnetic resonance. New Engl J Med 1981; 304: Shoubridge EA, Bniggs RW, Radda GK. 3 P N saturation transfer measurements of the steady state rates of creatine kinase and AlP synthetase in the rat brain. FEBS Lett 1982; 140: Bottomley PA, Kogure K. Namon R, Alonso OF. Cerebral energy metabolism in rats studied by phosphorus nuclear magnetic resonance using surface coils. Magn Reson lmag 1982; 1: Alger JR. Sillerud LO, Behar KL. et al. In vivo carbon-13 nuclear magnetic resonance studies of mammals. Science 1981; 214: Bottomley PA. Instrumentation for whole body N imaging. In: Witcofski RL, Karstaedt N, Partain CL, eds. Proc Internat Symp on N imaging. Winston-Salem: Dept of Radiol, Bowman Gray School of Medicine. 1982: Hoult Dl. Radiofrequency coil technology in N scanning. In: Witcofski RL, Karstaedt N, Partain CL, eds. Proc Internat Symp on N Imaging. Winston-Salem: Dept of Radiol. Bowman Gray School of Medicine, 1982: Mansfield P, Morris PG. N imaging in biomedicine. Advances in Magnetic Resonance, Suppl. 2. New York: Academic, 1982: Brownell GL, Budinger TF, Lauterbun PC. McGeer PL. Positron tomography and nuclear magnetic resonance imaging. Science 1982; 215: Pykett IL. N imaging in medicine. Sci Am 1982; 246: Bottomley PA, Andrew ER. RF magnetic field penetration, phase shift and power dissipation in biological tissue: implications Volume 150 Number 2 Radiology #{149} 445

6 for N imaging. Phys Med Biol 1978; 23: &ttomley PA. Hart HR. Edelstein WA, et al. N imaging/spectroscopy system to study both anatomy and metabolism. Lancet, July 30, 1983: Alderman DW. Grant DM. An efficient decoupler coil design which reduces heating in conductive samples in susperconducting spectrometers. J Magn Reson 1979; 36: Hoult DI, Richards RE. The signal-to-noise ratio of the nuclear magnetic resonance experiment. J Magn Reson 1976; 24: Edelstein WA, Bottomley PA, Pfeifer LM. A signal-to-noise calibration method for N imaging systems. Med Phys (in press). 25. Bottomley PA, Edelstein WA. Power deposition in whole-body N imaging. Med Phys 1981; 8: Hoult DI, Lauterbur PC. The sensitivity of the zeugmatographic experiment involving human samples. J Magn Reson 1979; 34: Coles BA. Dual frequency proton spin relaxation measurements on tissues from normal and tumor-bearing mice. J NatI Canc Inst 1976; 57: Escanye JM, Canet D, Robert J. Frequency dependence of water proton longitudinal N relaxation times in mouse tissues at 20#{176}C.Biochim. Biophys. Acta 1982; 721: Suzuki K. Chemistry and metabolism of brain lipids. In: Albers RW, Siegel GJ, Katzman R, Agranoff BW, eds. Basic neurochemistry. Boston: Little, Brown and Co, 1972: Gore JC. The meaning and significance of relaxation in N imaging. In: Witcofski RL, Karstaedt N, Partain CL, eds. Proc Internat Symp on N Imaging. Winston- Salem: Dept. of Radiology, Bowman Gray School of Medicine, 1982: Bottomley PA. A versatile magnetic field gradient control system for N imaging. J Phys E: Sci Instrumm 1981; 14: Norton WT. Myelin. In: Albers RW, Siegel GJ. Katzman R, Agranoff BW, eds. Basic neurochemistry. Boston: Little, Brown and Co. 1972: Ng IC, Evanochko WT. Hiramoto RN, et al. 31P N spectroscopy of in vivo tumors. Magn Reson 1982; 49: Stevens AN, Iles RA, Morris PG, GriffithsJR. Detection of glycogen in a glycogen storage disease by carbon-13 nuclear magnetic resonance. FEBS Lett 1982; 150: Brown TR, Kincaid BM, Ugurbil K. N chemical shift imaging in three dimensions. Proc Natl Acad Sci USA 1982; 79: Maudsley AA, Hilal SK, Perman WH, Simon HE. Spatially resolved high resolution spectroscopy by four dimensional N. Magn Reson 1983; 51: #{149} Radiology February 1984