MAGNETIC RESONANCE IN MEDICINE
26,23 1-240 ( 1992)
Pulsed Saturation Transfer Contrast BOB S. Hu,* STEVENM. CONOLLY, GRAHAM A. WRIGHT, DWIGHTG. NISHIMURA,AND ALBERTMACOVSKI Magnetic Resonance Systems Research Laboratory, 120 Durand, Stanford University, Stanford, California 94305; and *Division of Cardiovascular Medicine, Department of Medicine, Stanford University, Stanford, California 94305 Received April 5, 1991; revised September 3, 1991; accepted September 27, 1991
In vivo 'H conventional NMR image contrast generation usually relies on the macroscopic T , and T2relaxation parameters of the tissues of interest. Recently cross-relaxation related image contrast has been reported by Wolff and Balaban in animal models. Due primarily to the broad lineshape of the intended saturation spin pool and the use of off-resonance irradiation, high specific absorption rate and an auxiliary RF amplifier have been necessary to produce these images. The relatively long spin-lattice relaxation property of this spin pool, however, suggests the use of pulse methods to achieve saturation. In this paper, we show that short-T, spin pools can be selectively saturated with short intense R F pulses. Cross-relaxation time constants can be measured using the technique of saturation recovery. In vivo magnetization-transfer-weightedimages can be produced using pulses on commercial whole-body imagers without additional hardware. o 1992 Academic press, Inc. INTRODUCTION
The phenomenon of magnetization transfer between distinct pools of spins has been described as a mechanism of biological tissue relaxation ( 1-5). These mechanisms appear to be present in a wide variety of tissues and are thought to be related to the establishment of an exchangeable separate spin environment by macromolecules. The early work performed in spectrometers on hydrated protein samples as well as ex vivo samples of biological tissues established the exchange of magnetization between pools of relatively mobile long-T2 species and more restricted short-T2species. Furthermore, this phenomenon was shown to affect the macroscopically measured bulk relaxation times. A number of the essential elements of the magnetization transfer phenomenon have been described, including the possible exchange mechanisms, rate of exchange, macroscopic relaxation times, as well as estimates of the amount of exchangeable protons in various tissues (1-7). Wolff and Balaban ( 5 ) first produced in vivo images with magnetization-transferweighted contrast (MTC). They also coined the terms free ( H f )and restricted ( H , ) proton pools to describe the exchange compartments. Their technique took advantage of the broad lineshape of the short-T2 species by performing continuous irradiation several kilohertz off resonance to achieve selective saturation. Sufficient saturation of the short-T2species, normally unobservable, is then indirectly observed via exchange with and subsequent partial saturation of the long T2species. However, continuous 23 1
0740-3194/92 $5.00 Copyright 0 1992 by Academic Press, Inc. All rights of reproduction in any form reserved.
HU ET AL.
off-resonance irradiation leads to the practical problems of high-power deposition (specific absorption rate; SAR) and the need for an auxiliary RF amplifier on conventional whole-body imagers. Dixon ( 8 )has demonstrated the presence of magnetization transfer in conventional whole-body imagers by the application of multiple off-resonance pulses; however, the effects obtained were small. On the basis of the reported phenomenon of saturation transfer we analyze the possible imaging options available on conventional magnetic resonance imaging units to enhance and exploit the magnetization transfer effects. Starting from the Bloch equations, we show that conventional pulses produce effects on the bound proton pools which border on the large and small signal domains of Rabi. The differing behaviors of the observable versus the short-T2 proton pools during a constant RF pulse lead to the design of pulses which allow the production of magnetization-transferweighted images with efficient on-resonance pulses. Production of in vivo images using this technique on a commercial imaging system is demonstrated. METHODS
Theory Modifying the Bloch equations to include the effects of spin coupling between two systems of spins is generally accomplished by the addition of two spin-exchange terms ( I , 9-11). While the exact mechanism of exchange can be due to a large number of factors, the same phenomenological equations for longitudinal magnetization can be used to describe the observed macroscopic process. We will generally follow the notation introduced by Forsen and Hoffman as modified by Wolff and Balaban. Selective Saturation Because the reported exchange time constants are at least an order of magnitude longer than a typical RF pulse in whole-body imagers ( 3 , 5), the exchange processes do not appreciably affect the system response during such excitation. Neglecting the exchange process, T, recovery, and off-resonance effects, the solution for the Bloch equations in the presence of a constant continuous RF field is the classic rotating frame equation ( 11). Assuming the initial condition that = 0, the M , magnetization can be put in the form
+ !!sinh(Pt) P
It is clear that at low irradiating field strengths the solution is purely real and leads to exponential decays. As the B1 field strength increases 0 becomes imaginary so that M, becomes a decaying sinusoid with frequency approximately proportional to w I . We normally work at the latter limit (the Rabi large-signal domain) because conventional MRI units typically operate with peak w IT29 The extremely short T , ( t 2 0 0
PULSED SATURATION TRANSFER CONTRAST
ps) of the restricted proton
pool ( 5, 7), however, guarantees pure decay in the presence of typical peak RF strengths of less than 1 kHz equivalent in clinical imagers. Thus, given a pulse of proper amplitude and duration, the net effect for observable protons of relatively long T2 is an angle rotation while the bound proton with its short T 2is largely decay. Given this analysis a properly designed zero degree or transparent pulse can lead to selective saturation of the restricted proton pool while leaving the free proton pool relatively undisturbed.
Cross Relaxation Once the initial condition of selective saturation of H, is achieved, the exchange behavior of the longitudinal magnetization can be observed. In the absence of an irradiation field the modified Bloch equations can be reduced to the following pair of first order equations:
T 1 and T I ,are the longitudinal relaxation times expected without exchange. T~ and rr are the cross-relaxation times. Subscripts of r and f refer to H , and Hf proton pools
respectively. In fact, the exchange process is only observable if the exchange time constants are significantly shorter than the T1time constants ( 9 , 12). This leads to additional simplifications.Given the initial condition of M,, = 0, a monoexponentially decaying M f ican be derived:
The steady-state relationships, given the initial selective saturation condition can then be expressed as follow: M,,(co)
These solutions suggest that knowledge of M , (0) and M A (00 ) gives an estimate of the size of the free and restricted proton pools and the r f to 7 , ratio. Furthermore, the rate of exchange can then be used to estimate the absolute value of the exchange constants. This general analysis of exchange spectra was first reported by Bloembergen et al. in connection with their explanation of the experimental results of cross-saturation effects in paramagnetic salts ( 9 ) . SAR Analysis The large amplitude of the pulses required can lead to significant power deposition. While it is known that a given steady-state magnetization in a continuous wave experiment is maintained by the power of the irradiation, the integrated power can be
HU ET AL.
reduced in approaching a given desired magnetization by optimizing the irradiation amplitude. Analytically, Eq. [ 21 can be rewritten in a form which emphasizes the role of the total energy deposition:
E represents the energy of the pulse. 7 is the duration of the pulse. At the point of critical damping, 4 ( E / 7 ) T := 1, the residual longitudinal magnetization can be rewritten as
Mz(0)exp[-2ET2]( 1 +2ET2).
Similarly, at a given E if 4 ( E / 7 )T ;