Radiology
HOME HELP FEEDBACK SUBSCRIPTIONS ARCHIVE SEARCH TABLE OF CONTENTS
QUICK SEARCH:   [advanced]


     

This Article
Right arrow Abstract Freely available
Right arrow Figures Only
Right arrow Full Text (PDF)
Right arrow Submit a response
Right arrow Alert me when this article is cited
Right arrow Alert me when eLetters are posted
Right arrow Alert me if a correction is posted
Right arrow Citation Map
Services
Right arrow Email this article to a friend
Right arrow Similar articles in this journal
Right arrow Similar articles in PubMed
Right arrow Alert me to new issues of the journal
Right arrow Download to citation manager
Right arrow reprints & permissions
Citing Articles
Right arrow Citing Articles via Google Scholar
Google Scholar
Right arrow Articles by Graham, S. J.
Right arrow Articles by Henkelman, R. M.
Right arrow Search for Related Content
PubMed
Right arrow PubMed Citation
Right arrow Articles by Graham, S. J.
Right arrow Articles by Henkelman, R. M.
(Radiology. 1999;212:903-910.)
© RSNA, 1999

Technical Developments

Pulsed Magnetization Transfer Imaging: Evaluation of Technique1

Simon J. Graham, PhD and R. Mark Henkelman, PhD

1 From the Departments of Medical Biophysics (S.J.G., R.M.H.) and Medical Imaging (R.M.H.), University of Toronto, Ontario, Canada, and Imaging and Bioengineering Research, Sunnybrook and Women's College Health Sciences Centre, 2075 Bayview Ave, Toronto, Ontario, Canada M4N 3M5. Received September 10, 1998; revision requested November 3; revision received December 2; accepted March 1, 1999. Supported in part by the Terry Fox Foundation and GE Medical Systems of Canada. Address reprint requests to S.J.G. (e-mail: sgraham@sten.sunnybrook.utoronto).


   Abstract
TOP
 Abstract
Introduction
 Materials and Methods
 Results
 Discussion
 References
 
With use of an established physical model, numeric simulations were performed to evaluate current imaging protocols for the two primary applications of magnetization transfer: cerebral magnetic resonance angiography and neuroimaging of white matter disease. The authors found that the current technique is appropriate in the former but suboptimal in the latter. Further clinical investigations could potentially improve magnetization protocols for neuroimaging.

Index terms: Brain, abnormalities, 10.871 • Brain, MR, 10.121417 • Brain, white matter, 10.121417 • Magnetic resonance (MR), magnetization transfer contrast, 10.121417 • Sclerosis, multiple, 10.871


   Introduction
TOP
 Abstract
 Introduction
Materials and Methods
 Results
 Discussion
 References
 
Magnetization transfer (MT) is well known to improve contrast for selective tissues on clinical magnetic resonance (MR) images. Its principal uses are in cases of (a) cerebral MR angiography, in which MT enhances the detectability of small vessels by improving the signal suppression of static brain tissues without affecting blood (1,2), and (b) white matter disease (eg, multiple sclerosis), in which MT enhances lesion conspicuity (3), allows characterization of normal-appearing white matter (4,5), and provides a correlate measure of cognitive decline (6). Other MT applications (eg, musculoskeletal [7] and head and neck [8] imaging of tumors) are also being developed.

Qualitatively, the MT effect is easily described. When radio-frequency (RF) energy is applied several kilohertz off resonance from the Larmor frequency of the water molecules in tissues, longitudinal magnetization associated with tissue proton macromolecules (the semisolid pool) is preferentially saturated in comparison with water protons (the liquid pool), which normally provide the dominant signal intensity in clinical MR imaging. Exchange of longitudinal magnetization between the pools, via macromolecular sites such as hydroxyl or amide groups (9,10), causes signal reduction in the liquid pool. For a variety of tissues, this MT effect is typically 10%–50% of the signal intensity obtained at clinical imaging when MT pulses are not applied (11). Quantitatively, however, the MT effect is a complicated function of the relaxation times and absorption line shapes of the semisolid and liquid pools, their respective volume fractions, and the efficiency of MT exchange. Notwithstanding this complexity, quantitative understanding of MT provides important physical insight regarding how MT image contrast is generated for specific pulse sequences. From the fundamental MT and relaxation properties of tissues, the optimal contrast can be determined for specific clinical applications. Early in the development of MR imaging, similar analyses were performed to optimize T1- and T2-weighted contrast (12,13).

In the past decade, progress has been made in describing the MT properties of tissues mathematically. Initially, a two-pool model was developed to characterize the MT effect in an aqueous agar gel system under continuous RF irradiation off resonance (14). This model also applies for many clinically relevant tissues (gray matter, white matter, muscle, liver, and blood) when irradiation of the semisolid pool is correctly specified (11,15). More recently, the model was extended to include saturation pulses of short duration, which are used in clinical MT imaging to minimize patient power absorption and imaging time (16). In this study, we used this method to evaluate MT imaging protocols for two applications: static signal suppression in cerebral MR angiography and optimization of contrast in white matter disease. In particular, there is need for such an evaluation in white matter neuroimaging; the reported MT effect for normal white matter ranges from 16% (17) to 60% (18), which suggests that there are wide variations in the effectiveness of clinical MT protocols.

Simulations were performed for spoiled gradient-echo (GRE) imaging, a pulse sequence often used in conjunction with MT. The predicted signals, based on measurements of the MT and relaxation properties of excised tissues at body temperature, were compared with those in the literature.


   Materials and Methods
TOP
 Abstract
 Introduction
 Materials and Methods
Results
 Discussion
 References
 
MR Parameters of Tissues
To simulate the MR signals produced at pulsed MT imaging, several fundamental MR parameters must be estimated for the tissues of interest. Proton-density estimates for blood and gray and white matters were obtained from previously reported water content measurements (19), and for the latter two tissues, these estimates were modified to exclude the vascular component (20). Whereas numerous reports of the relaxation properties of tissues exist, MT parameters have previously been measured at room temperature only (11,15) and may not reflect physiologic conditions. Therefore, initial experiments at body temperature were performed with ex vivo tissues to measure MR parameters for simulation purposes. Blood was obtained from a single human subject, with the approval of our local ethics committee and with informed consent, and was oxygenated by means of vigorous shaking in air. Homogeneous samples of gray matter from the parietal cortex and white matter from the corpus callosum were excised from freshly slaughtered cattle. Samples were approximately 0.5 cm3 in volume and were placed in sealed borosilicate glass tubes (10-mm outside diameter). No other sample preparation was performed. All measurements were initiated within 90 minutes after tissues were extracted. Because such measurements are typically quite precise, experiments were repeated with only two samples of each tissue to assess reproducibility.

Experiments were conducted at 1.5 T by using a superconducting magnet with a bore diameter of 20 cm (Nalorac Cryogenics, Martinez, Calif), controlled with a programmable console (SMIS, Surrey, England). Samples were maintained at body temperature with heated forced air. Experiments were performed without imaging to measure bulk sample properties. The rectangular pulse duration was approximately 10 µsec for a 90° pulse. Three MR experiments were performed: (a) a Carr-Purcell-Meiboom-Gill, or CPMG, sequence (21,22) to measure T2 (repetition time [TR] of 10,000 msec, 1-msec echo spacing, 5,000 even echoes sampled, four signals averaged); (b) an inversion-recovery sequence (21) to measure T1 relaxation (31 inversion time intervals spaced logarithmically from 1 to 32,000 msec, free induction decay amplitude acquired at each inversion time interval, two signals averaged, 10,000 msec between 90° and inversion pulses); and (c) continuous-wave RF irradiation off resonance (14) to measure the MT effect (5-second duration, 27 frequencies off resonance spaced logarithmically from 0.01 to 225 kHz, seven continuous-wave irradiation amplitudes increasing from pulse amplitude, or f1, of 83 Hz by factors of two). Because the duration of the MT experiments was approximately 84 minutes, the total set of MT data for blood was obtained by measuring seven well-mixed blood samples for 12 minutes each to eliminate potential error due to sedimentation of red blood cells. Measurements of the sample temperature indicated that negligible RF power absorption occurred during the course of these experiments (data not shown).

Analysis of these data first involved estimation of average relaxation times, or <T1> and <T2>, for each tissue by means of monoexponential fits to the relaxation data. For white matter, average T2 was estimated by means of a weighted sum of relaxation times obtained from a biexponential fit. MT parameters and relaxation times for the liquid pool A, T1A and T2A, were estimated from a least squares fit of the MT data to an established two-pool model describing the MT effect for continuous-wave RF irradiation (14) and knowledge of the average T1 and T2 values. An iterative procedure was included to estimate the absorption line shape of the semisolid pool (15), and it was performed at each irradiation frequency off resonance. (T1 for the semisolid pool B, T1B, was not estimated precisely from the data and was assumed to be 1,000 msec for all tissues. This assumption is justified because the T1B value has little influence on pulsed MT simulations [16]. Sensitivity analysis [data not shown] indicated that a 10% change in T1B produced a change of approximately 0.1% in MT effect.) All errors were reported either as the average SD obtained from the fitting procedure or as half the difference between the mean parameter estimates for the two samples of each tissue, whichever was largest.

Pulsed MT with Excitation, Absorption, Saturation, and Evolution, or "EASE"
Simulation of pulsed MT contrast requires an understanding of four key concepts: excitation, absorption, saturation, and evolution. These concepts are illustrated in Figure 1, and mathematic treatment is provided elsewhere (16). In response to an MT pulse, magnetization in the liquid pool, MA, nutates about the associated effective B1 field in a manner very similar to the well-known RF excitation described with the Bloch equations (Fig 1, A). The associated decrease in longitudinal magnetization can be mistaken with MT and is known as the direct effect. Absorption of RF energy by both pools, however, is characterized by the absorption line shapes, gA and gB, and the MT pulse amplitude, f1, applied at a given frequency offset from resonance, {Delta} (Fig 1, B). Because the semisolid pool has a very small T2 value (approximately 10 µsec), the line shape gB is much broader than that of gA, and absorption occurs preferentially in the semisolid pool and becomes less efficient (requires more power) as the offset frequency increases. Fast T2 relaxation in the semisolid pool also ensures that RF absorption manifests as saturation (attenuation) of longitudinal magnetization, MBz, with negligible generation of transverse magnetization (Fig 1, C). When the MT pulse is turned off, magnetization in both pools evolves owing to T1 recovery and MT exchange, characterized by a rate constant, R (Fig 1, D). The equilibrium magnetization in the liquid pool is taken as unity, whereas the relative size of the semisolid pool is indicated as MBo. The evolution occurs in two phases: a fast phase, dominated by the rate R, in which the liquid pool replenishes the longitudinal magnetization in the semisolid pool quite rapidly, and a slow phase, dominated by T1A, in which the coupled system returns to equilibrium.



View larger version (15K):
[in this window]
[in a new window]
  		Figure 1. Schematics depict pulsed MT with excitation, absorption, saturation, and evolution, or EASE. A, Excitation of the liquid pool magnetization. B1eff = effective B1 field, MA = magnetization in the liquid pool. B, Absorption line shapes for the liquid pool (gA) and the semisolid pool (gB). {Delta} = frequency off resonance, {omega} = frequency. In C and D, MAo = relative size of the liquid pool, MBo = relative size of the semisolid pool, MAz = saturation (attenuation) of longitudinal magnetization in the liquid pool, MBz = saturation (attenuation) of longitudinal magnetization in the semisolid pool, t = time. C, Saturation of the semisolid pool. D, Evolution of both pools after completion of an MT pulse. (See Materials and Methods for details.)

 
The signal decrease in the liquid pool due to one MT pulse is typically only several percent, but repetitive application of MT pulses produces a pronounced net decrease over time. In general, the net signal decrease in the liquid pool due to MT exchange depends on (a) the product of R and the fractional size of the semisolid pool with respect to the liquid pool, MBo, divided by the rate of T1 recovery in the liquid pool (23) and (b) the MT pulse parameters: pulse amplitude, duration, offset frequency, and repetition time.

Pulsed MT Imaging
Although the excitation, absorption, saturation, and evolution principles are applicable to arbitrary pulse sequences, MT pulses used with spoiled GRE imaging are of primary interest because this combination currently receives the most clinical application. The pulse sequence used in simulations in this study contained three distinct elements (Fig 2): MT pulse plus spoiler gradient (with total time duration, T); small flip angle excitation characterized by the angle, {theta}, which produces transverse magnetization for the liquid pool but has negligible effect on the semisolid pool; and TR, which incorporates time for spatial encoding and evolution of image contrast. (The inclusion of MT pulses inevitably compromises the minimum TR achievable with spoiled GRE imaging). Transverse magnetization is assumed to be completely spoiled prior to application of the subsequent MT saturation pulse.



View larger version (6K):
[in this window]
[in a new window]
  		Figure 2. Schematic depicts simplified spoiled GRE sequence including MT pulses. Key parameters influencing contrast are MT pulse amplitude, f1; frequency off resonance, {Delta}; imaging flip angle, {theta}; and repetition time, TR. The minimum TR interval is compromised by the time duration, T, required for MT saturation. t = time.

 
Simulations were conducted (MATLAB; MathWorks, Natick, Mass) with use of the standard MT pulse implemented at 1.5 T (Signa; GE Medical Systems, Milwaukee, Wis) (pulse envelope described by an apodized, single-lobe sinc function of 14-msec duration and 150-Hz bandwidth). Therefore, image contrast depended on four parameters: MT pulse amplitude, f1, and offset frequency, {Delta}, and the conventional spoiled GRE parameters, {theta} and TR. To investigate the relative importance of each of these parameters and the associations between them, simulations were performed for each clinical application by using a broad range of possible imaging protocols (f1, 0.2–1.0 kHz; {Delta}, 0.5–4.0 kHz; {theta}, 5°–40°; TR, 20–125 msec). The spoiled GRE signal intensity without MT pulses, SI, was also calculated according to the appropriate equation (24):

with the equilibrium magnetization for the liquid pool again taken as unity.

Simulations of Cerebral MR Angiography
The pulse sequence shown in Figure 2 is easily used to simulate imaging contrast for cerebral MR angiography. Current state-of-the-art clinical imaging protocols combine pulsed MT, multiple offset thin-slab acquisition, or MOTSA (25); application of the three-dimensional fast spoiled GRE sequence; and tilted, optimized, nonsaturating, excitation, or TONE, pulses for ramped small flip angle excitation across each slab (26). A typical slab thickness of 20 mm was assumed, and the ratio of {theta} values at the entrance and exit of the slab was 1:2.2 according to clinical prescriptions (27). Time-of-flight contrast between flowing blood and static tissues was assessed for vessels in the center of the slab. While magnetization associated with static tissues was allowed to reach steady state, blood was assumed to flow uniformly perpendicular through the slab and thus was subject to markedly fewer MT pulses and ramped RF excitation, which is dependent on blood velocity and TR (25). To assess image contrast for small 0.5-mm-diameter vessels, an empirical model relating vessel diameter and blood velocity was used (28) that predicted an average flow velocity of 37 mm/sec.

Because MR angiographic applications are subject to imaging time constraints, simulations were performed with TRs of 45 msec and 26 msec without MT pulses for comparison. These conditions represent the minimum TR values obtainable in a previous clinical study (27). Blood at the center of the image slab therefore received either six or 10 excitations with or without MT, respectively. In the absence of MT, over the middle half of the slab, the blood signal varied plus or minus 20% from the value at the center, assuming a flip angle of 25°.

To compare the efficiency of cerebral MR angiography with different MT pulse parameters and without MT, the signal difference between blood and static tissue was calculated and divided by the square root of TR. This figure of merit is proportional to the signal-difference–to-noise ratio independent of acquisition time. The calculation accounts for the increase in minimum imaging time that occurs when MT pulses are used, which could in principle be used to perform signal averaging when MT pulses are absent.

Simulations of White Matter Neuroimaging
In this example, simulations were performed with a TR value of 20 msec, which approximates the minimum TR that is currently achievable when MT pulses are included, and also for a TR value of 125 msec, a typical value used for MT imaging of multiple sclerosis (5). Results of imaging protocols were compared on the basis of the MT ratio difference between white and gray matters per square root of TR, in which the MT ratio was expressed as the percentage signal decrease due to MT relative to the signal achieved without MT at the same TR value.


   Results
TOP
 Abstract
 Introduction
 Materials and Methods
 Results
Discussion
 References
 
MR Parameters of Tissues
Estimated relative proton density and estimated relaxation time and MT parameters obtained by means of measurements at body temperature are summarized in the Table for ex vivo blood, gray matter, and white matter tissues. The estimates show the expected general relationships. The T1A and T2A values for white matter were reduced with respect to those of gray matter and correlate with the reduced proton density and water content in white matter (29). The relaxation times for blood are characteristically large, and in particular the T2A value of 310 msec was representative of oxygenated rather than deoxygenated blood, similar to previous measurements (30,31). Considering the MT parameters, the semisolid pool is largest for white matter and smallest for blood, and the MT rate constants have the same ranking as those in a previous study (11). Interestingly, the rate constants were found to be larger by a factor of two than those measured at room temperature (11). This observation is important because it emphasizes that previous estimates of MT parameters from data acquired at room temperature are inappropriate when used to predict contrast in vivo. The specific mechanism responsible for the temperature dependence of MT exchange remains to be determined, although diffusion probably plays a role.


View this table:
[in this window]
[in a new window]
  		Estimated Relaxation Times and MT Parameters for ex Vivo Tissues at Body Temperature (Bo = 1.5 T)
 
In Figure 3, the extracted (15) absorption line shape, gB, is plotted for each tissue versus offset frequency, {Delta}. The area under each line shape equals one and for clarity, error bars (obtained by varying {chi}2 for the fit by ±1 SD) are indicated for white matter only. The absorption line shape for blood is much narrower than those for gray and white matters, displaying a fivefold larger amplitude on resonance ({Delta} = 0) and a half width at half maximum of approximately 500 Hz. The line shapes for white and gray matters are equivalent within experimental error and still have appreciable amplitude at 10 kHz off resonance (data not shown).



View larger version (15K):
[in this window]
[in a new window]
  		Figure 3. Schematic depicts estimated absorption line shapes of the semisolid pool, gB, for ex vivo blood, white matter, and gray matter at body temperature. The area of each line shape is normalized to one, and for clarity, error bars (obtained by varying {chi}2 for the fit by ±1 SD) are shown for only white matter. The absorption line shape for blood is much narrower than that for white or gray matter. {Delta} = frequency off resonance.

 
Cerebral MR Angiography
Imaging efficiency characterized by the signal difference {Delta}Mxy (blood signal minus static tissue signal) per square root of time is shown in Figure 4 for a range of MT imaging protocols at TRs of 45 msec (thick lines) and for protocols of 26 msec (thin lines) without MT pulses. For clarity, only MT protocols that were quite close to maximizing the signal difference per square root of time are shown. White matter was taken as the static tissue because it has a much shorter T1 than does gray matter and is more difficult to suppress. Figure 4, top, which illustrates the relationship with flip angle, indicates that use of an MT pulse of modest amplitude (f1 = 450 Hz, {Delta} = 1.2 kHz) resulted in an improvement of 11% in signal difference per square root of time compared with that for imaging without an MT pulse. Because the MT pulse increases the available minimum TR (in this case, 45 vs 26 msec), the imaging flip angle must be increased for maximum efficiency, from 25° without MT pulses to 30° with MT pulses. The gain in efficiency is only weakly dependent on MT pulse amplitude and off-resonance frequency (Fig 4, middle and bottom, thick lines), requiring f1 and {Delta} values above thresholds of only approximately 300 Hz and 1.5 kHz, respectively.



View larger version (25K):
[in this window]
[in a new window]
  		Figure 4. Schematics depict signal difference, {Delta}Mxy, (blood signal minus static tissue signal) per square root of TR for a range of MT protocols that are near optimum (TR = 45 msec [thick line]). Relationships are shown with top, flip angle, {theta}; middle, MT pulse amplitude, f1; and bottom, frequency off resonance, {Delta}. Imaging without MT pulses is indicated by thin lines (TR = 26 msec). Bottom: The dotted line is obtained when none of the tissues of interest exhibits MT exchange.

 
The dotted line in Figure 4, bottom, indicates the predicted signal difference per square root of time if neither blood nor static tissue exhibited MT exchange (ie, the signal difference due to only the direct effect in the two tissues). A substantial portion of the signal difference was due to the direct effect for frequencies off resonance below approximately 1.5 kHz. At larger frequencies off resonance, however, the direct effect diminished and tended toward the signal difference obtained without MT pulses, which indicated that the maximum gain in sensitivity occurred when MT exchange was maximized and the direct effect was minimized.

Gray Matter–to–White Matter Contrast
A similar family of curves is shown in Figure 5, which illustrates the MT ratio difference between white and gray matters per square root of time for different MT protocols with TRs of 20 and 125 msec. Again, only protocols producing MT ratio differences that were close to the most efficient at each TR value are shown. The maximum MT ratio difference was slightly less at longer TR (16% for TR of 20 msec, with respect to gray matter, 11% for TR of 125 msec [data not shown]). Expressed per square root of time, therefore, the MT ratio difference generated at TR of 20 msec was more efficient by a factor of approximately 3.6. The flip angle value for maximum MT was smaller at TR of 20 msec ({theta} = 15°) and required adjustment within 5° for best efficiency, whereas {theta} values for TR of 125 msec provided good sensitivity from 20° to 40° (Fig 5, top). A similar effect was exhibited with respect to MT pulse amplitude and frequency off resonance (Fig 5, middle and bottom). The maximum MT ratio difference at TR of 20 msec required f1 amplitudes of ap- proximately 300–500 Hz at 1.5–2.5 kHz off resonance, whereas TR of 125 msec required f1 greater than approximately 500 Hz at 1.5–4.0 kHz off resonance.



View larger version (26K):
[in this window]
[in a new window]
  		Figure 5. Schematics depict MT ratio difference, {Delta}MTR (white matter minus gray matter) per square root of TR for MT protocols near optimum efficiency at TRs of 20 msec (thick line) and 125 msec (thin line). Relationships are shown with top, flip angle, {theta}; middle, MT pulse amplitude, f1; and bottom, frequency off resonance, {Delta}. Bottom: The dotted lines predict the MT ratio difference obtained if the direct effect is neglected.

 
In Figure 5, bottom, dotted lines illustrate the MT ratio difference that would be observed if the direct effect of RF absorption on the liquid pool was neglected. For the optimized imaging protocols at each TR value, MT exchange accounted for approximately 95% of the signal difference between white and gray matters.


   Discussion
TOP
 Abstract
 Introduction
 Materials and Methods
 Results
 Discussion
References
 
With a detailed understanding of MT exchange, simulations can be used to evaluate the MT contrast produced with imaging protocols for various clinical applications. This approach fulfills a practical need. Simulations can be more comprehensive than actual imaging experiments, have no initial effect on local MR imaging resources, and usually do not require patient databases. They can also be used to investigate protocols that are currently unachievable with clinical imagers and parameters that are difficult or impossible to measure physically (eg, dotted lines in Figs 4, bottom and 5, bottom). Thus, simulations can provide impetus for improved MR imaging technology. They are, however, simplifications of reality. How should the simulation results in this study influence clinical practice? For the two applications investigated, investigators in previous clinical studies have attempted to optimize MT technique for cerebral MR angiography (27) and assessment of white matter disease (18,32).

Cerebral MR Angiography
Previously, Goodrich et al (27) compared results with multiple offset thin-slab acquisition, or MOTSA, plus tilted, optimized, nonsaturating, excitation, or TONE, methods with MT pulses to those without MT pulses, with use of the same TR values investigated in our study and the same MT pulse envelope. For fixed MT pulse amplitude and frequency off resonance (f1 = 470 Hz, {Delta} = 1.2 kHz), various tilted, optimized, nonsaturating, excitation pulse flip angles and slab thicknesses were investigated. For 10 patients, the largest peak signal-difference–to-noise ratio with MT for 0.5- to 0.75-mm-diameter vessels occurred with a slab thickness of 16 mm and represented a gain of 29% compared to results with analogous imaging without MT pulses. When expressed as signal-difference–to-noise ratio per unit time, these two imaging protocols were in fact roughly equivalent in efficiency. In comparison, the simulations predicted 45% increased signal difference when MT pulses were used compared with when MT pulses were absent and predicted a modest gain in signal difference per unit time of 11%. Considering the inherent assumptions in the simulations, the simulation and clinical results were in quite good agreement. Furthermore, the simulation results indicate that vessel contrast is quite insensitive to the choice of MT protocol and suggest that appropriate technique is currently used in this clinical application (27). MT pulse amplitudes above a threshold of 400 Hz and offset frequencies above 1.5 kHz all prove satisfactory.

Imaging of White Matter Disease
McGowan et al (32) developed an MT imaging protocol that has been adopted in numerous neuroimaging studies (3,5,33). In initial experiments, sinc pulses were used, with nominal bandwidth of 100 Hz and duration of 19 msec; from investigations of several different f1 amplitudes, f1 of 160 Hz was found to produce the largest MT effect for frequency offsets ranging from 3 to 15 kHz. This MT pulse was coupled with GRE imaging (TR of 100 msec, minimum echo time, {theta} of approximately 10°) to achieve primarily proton-density contrast in the absence of MT pulses. The average specific absorption rate of this pulse sequence was approximately 2.0 W/kg and generated an average MT ratio for normal white matter of 43% ± 1 (mean ± SD) (33). Alternatively, Finelli et al (18) proposed a technique with a 10-msec single-period sinusoid envelope and f1 of 700 Hz and {Delta} of 4 kHz coupled with a more rapid spoiled GRE sequence (TR of 25 msec, echo time of 4.4 msec, and flip angle of 6°). In this case, the average specific absorption rate was 4.0 W/kg, which produced an MT ratio of 59% ± 2 in normal white matter.

The large difference in MT ratios obtained with these two different imaging protocols can be easily explained theoretically. For a single ideal MT pulse that completely saturates the semisolid pool and leaves the liquid pool unaffected, the critical time, or tcrit, can be calculated when maximum signal decrease occurs in the liquid pool owing to MT exchange (16). From the Table, the critical time is approximately 100 msec for blood and approximately 70 msec for gray and white matters. The MT pulse repetition period should, therefore, be much less than the critical time to accumulate the MT effect efficiently over many pulses. Thus, imaging with shorter TR interval produces larger MT ratio values. If TR is much greater than the critical time, then MT effect is accumulated inefficiently; pulse power must be markedly increased to produce more direct effect. It is only because T1 recovery in the liquid pool occurs more rapidly than MT exchange under these conditions that appreciable MT ratio values are obtained at longer TR (16).

The simulation results suggest an alternative MT imaging protocol for neuroimaging to that previously advocated. Both McGowan et al (32) and Finelli et al (18) focused on maximizing the MT effect for white matter. However, if white matter MT is maximized, then it may also be maximized for white matter that is partially demyelinated or exhibits partial axonal loss. What is ultimately important is the MT ratio contrast between normal white matter and white matter lesions. With use of gray matter to represent axonal loss, we found that MT ratio contrast is maximized between white and gray matters with technique that does not provide the maximum MT effect in both tissues. Furthermore, this protocol, found by means of simulation (f1 = 400 Hz, {Delta} = 2 kHz, {theta} = 15°, TR = 20 msec), exhibits an average specific absorption rate less than 4.0 W/kg when compared with that of Finelli et al (18).

General Recommendations
The results from these simulations were obtained on the basis of a number of simplifying assumptions. In particular, (a) MR properties of ex vivo tissues at body temperature were used to approximate those of living tissues in physiologic conditions, with bovine gray and white matters substituted for the human analogues; (b) T2* effects were neglected; (c) owing to the lack of detailed MR measurements of the MR properties of white matter lesions (which are variable in character), gray matter was used as a substitute; and (d) time-of-flight signal enhancement for blood accounted neither for vessel narrowing, branching, and tortuosity nor for the nonlocalized MT saturation in clinical application (27) (ie, where each MT pulse actually saturates the entire head, not just a single slab). The error introduced by these assumptions is difficult to assess. Assumption d, however, means that flowing blood is likely subject to more MT and imaging pulses than were simulated. This could account for the theoretically greater MT over that actually observed clinically (27).

Nevertheless, the simulation results were sufficiently compelling that they highlight the need to optimize MT imaging protocols by means of clinical studies and to investigate image contrast further in patient populations. For cerebral MR angiography, results in such studies have produced good MT imaging technique (27). There are several basic guidelines for conducting such investigations that are easily implemented in any MR imaging laboratory.

1. Choose the minimum TR that allows an acceptable specific absorption rate. Use of MT pulses alters the flip angle {theta} that produces best image contrast per unit time, so image quality must be investigated as a function of {theta}. This can be performed adequately with use of the standard MT pulse provided with the MR imager. Noise levels should be reduced to acceptable levels by means of signal averaging.

2. Choose an MT pulse: Both binomial on-resonance and shaped off-resonance pulses are acceptable. Findings in previous simulations and clinical studies indicate that appreciable MT effect can be obtained by means of both binomial on-resonance and shaped off-resonance pulses (16,18,34). The direct effect must be minimized in both cases, which is simpler with shaped pulses because they are less sensitive to static field inhomogeneity (34). For different shaped pulses, MT effect scales linearly with average pulse power (16,35); standard choices, such as sinc or Gaussian pulse envelopes should provide desired MT contrast within current specific absorption rate limits. To minimize the required power, the off-resonance frequency should be chosen just large enough to produce negligible direct effect, and results of simulations indicate that offset frequencies of 1.5–2.0 kHz are appropriate.

3. Investigate image contrast as a function of MT pulse amplitude. Weak MT pulses produce poor MT image contrast, whereas excess power produces undesirable direct effect. For neuroimaging, findings in this study suggest that the desirable MT pulse amplitude is insensitive to the imaging flip angle, so that {theta} determined in guideline 1 should be sufficient. If MT pulse amplitude must be increased, then the TR value may also have to be increased slightly for specific absorption rate to remain acceptable, and guideline 1 should be revisited.

The simulation results also suggest that the MT protocol for cerebral MR angiography is flexible (a range of MT pulse parameters produce equivalent contrast). It may be possible to determine the best MT protocol for imaging white matter disease and then to use this protocol for cerebral MR angiography without much loss of efficiency. Such an approach would facilitate pulse sequence implementation but requires verification with further study.

Other Pulse Sequences
This study focused on use of MT pulses with spoiled GRE imaging, the predominant pulse sequence used in MT protocols. However, appreciable deliberate MT effect can also be generated with spin-echo imaging (34) as long as enough interleaved sections are acquired per TR interval to ensure that MT pulses are applied with a period much less than the critical time. This pulse sequence is more complicated to simulate because some saturation of the semisolid pool is caused by refocusing pulses at varying frequencies off resonance during multisection interleaving (36). An accurate comparison of pulsed MT with spoiled GRE versus spin-echo pulse sequences is beyond the scope of this article. Interestingly, use of MT pulses with fast spin-echo imaging produces little deliberate (17) MT effect, because the echo train length typically exceeds the critical time. However, unintended MT contrast is also produced when fast spin-echo imaging is performed without an MT saturation pulse, owing to the many refocusing pulses (36,37). This suggests that calculation of MT ratios is inadvisable with use of fast spin-echo imaging.

In conclusion, simulation results indicate that current use of MT in imaging protocols is close to optimal for cerebral MR angiography, whereas neuroimaging protocols can likely be made more effective. It is hoped that this report will encourage clinical neuroradiologists to investigate improvement of MT protocols for white matter lesions.


   Footnotes
 
Abbreviations: GRE = gradient echo MT = magnetization transfer RF = radio frequency TR = repetition time

Author contributions: Guarantor of integrity of entire study, R.M.H.; study concepts and design, R.M.H., S.J.G.; definition of intellectual content, R.M.H., S.J.G.; literature research, S.J.G., R.M.H.; experimental studies, S.J.G.; data acquisition and analysis, S.J.G.; statistical analysis, S.J.G.; manuscript preparation and editing, S.J.G.; manuscript review, R.M.H.


   References
TOP
 Abstract
 Introduction
 Materials and Methods
 Results
 Discussion
 References
 

  1. Pike GB, Hu BS, Glover GH, Enzmann DR. Magnetization transfer time-of-flight magnetic resonance angiography. Magn Reson Med 1992; 25:372-379.[Medline]
  2. Furst G, Hofer M, Steinmetz H, et al. Intracranial stenoocclusive disease: MR angiography with magnetization transfer and variable flip angle. AJNR 1996; 17:1749-1757.[Abstract]
  3. Grossman RI, Gomori JM, Ramer KN, Lexa FJ, Schnall MD. Magnetization transfer: theory and clinical applications in neuroradiology. RadioGraphics 1994; 14:279-290.[Abstract]
  4. Filippi M, Campi A, Dousset V, et al. A magnetization transfer imaging study of normal-appearing white matter in multiple sclerosis. Neurology 1995; 45:478-482.[Abstract/Free Full Text]
  5. Van Buchem M, McGowan JC, Kolson DL, Polansky M, Grossman RI. Quantitative volumetric magnetization transfer analysis in multiple sclerosis: estimation of macroscopic and microscopic disease burden. Magn Reson Med 1996; 36:632-636.[Medline]
  6. Rovaris M, Filippi M, Falautano M, et al. Relations between MR abnormalities and patterns of cognitive impairment in multiple sclerosis. Neurology 1998; 50:1601-1608.[Abstract]
  7. Li KCP, Hopkins KL, Moore SG, et al. Magnetization transfer contrast MRI of musculoskeletal neoplasms. Skeletal Radiol 1995; 24:21-25.[Medline]
  8. Markkola AT, Aronen HJ, Paavonen T, et al. Spin lock and magnetization transfer imaging of head and neck tumours. Radiology 1996; 200:369-375.[Abstract/Free Full Text]
  9. Kucharczyck W, MacDonald PM, Stanisz GJ, Henkelman RM. Relaxivity and magnetization transfer of white matter lipids at MR imaging: importance of cerebrosides and pH. Radiology 1994; 192:521-529.[Abstract/Free Full Text]
  10. Liepinsh E, Otting G. Proton exchange rates from amino acid side chains: implications for image contrast. Magn Reson Med 1996; 35:30-42.[Medline]
  11. Morrison C, Henkelman RM. A model for magnetization transfer in tissues. Magn Reson Med 1995; 33:476-482.
  12. Henkelman RM, Hardy P, Poon PY, Bronskill MJ. Optimal pulse sequence for imaging hepatic metastases. Radiology 1986; 161:727-734.[Abstract/Free Full Text]
  13. McVeigh ER, Henkelman RM, Bronskill MJ. Optimization of survey protocols for MRI. Magn Reson Med 1990; 13:177-191.[Medline]
  14. Henkelman RM, Huang X, Xiang Q, Stanisz GJ, Swanson SD, Bronskill MJ. Quantitative interpretation of magnetization transfer. Magn Reson Med 1993; 29:759-766.[Medline]
  15. Li JG, Graham SJ, Henkelman RM. A flexible magnetization transfer line shape derived from tissue experimental data. Magn Reson Med 1997; 37:866-871.[Medline]
  16. Graham SJ, Henkelman RM. Understanding pulsed magnetized transfer. JMRI 1997; 7:903-912.
  17. Van Waesberghe JHTM, Castelijns JA, Lazeron RHC, Lycklama à Nijeholt GJ, Barkhof F. Magnetization transfer contrast (MTC) and long repetition time spin-echo MR imaging in multiple sclerosis. Magn Reson Imaging 1998; 16:351-358.[Medline]
  18. Finelli DA, Hurst GC, Amantia P, Jr, Gullapali RP, Apicella A. Cerebral white matter: technical development and clinical applications of effective magnetization transfer (MT) power concepts for high-power, thin-section, quantitative MT examinations. Radiology 1996; 199:219-226.[Abstract/Free Full Text]
  19. Ernst T, Kreis R, Ross BD. Absolute quantitation of water and metabolites in the human brain. I. Compartments and water. J Magn Reson 1993; 102(series B):1-8.
  20. Woodbury DM. Physiology of body fluids. In: Ruch TC, Patton HD, Scher AM, eds. Physiology and biophysics: circulation, respiration, and fluid balance. 20th ed. Philadelphia, Pa: Saunders, 1974.
  21. Carr HY, Purcell EM. Effects of diffusion on free precession in nuclear magnetic resonance experiments. Phys Rev 1954; 94:630-638.
  22. Meiboom S, Gill D. Modified spin-echo method for measuring nuclear relaxation times. Rev Sci Instr 1958; 29:688-691.
  23. Henkelman RM, Morrison CS. A model of magnetization transfer for various tissues in vitro (abstr) In: Proceedings of the Third Meeting of the International Society for Magnetic Resonance in Medicine. Berkeley, Calif: International Society for Magnetic Resonance in Medicine, 1995; 108.
  24. Van der Meulen P, Groen JP, Tinus AMC, Bruntink G. Fast field echo imaging: an overview and contrast calculations. Magn Reson Imaging 1988; 6:355-368.[Medline]
  25. Parker DL, Yuan C, Blatter DD. MR angiography by multiple thin slab acquisition. Magn Reson Med 1991; 17:434-451.[Medline]
  26. Atkinson D, Brant-Zawadski M, Gillan G, Purdy D, Laub G. Improved MR angiography: magnetization transfer suppression with variable flip angle excitation and increased resolution. Radiology 1994; 190:890-894.[Abstract/Free Full Text]
  27. Goodrich KC, Blatter DD, Parker DL, Du YP, Meyer KJ, Bernstein MA. A quantitative study of ramped radio frequency, magnetization transfer, and slab thickness in three-dimensional time-of-flight magnetic resonance angiography in a patient population. Invest Radiol 1996; 31:323-332.[Medline]
  28. Du YP, Parker DL, Davis WL, Cao G, Buswell HR, Goodrich KC. Experimental and theoretical studies of vessel contrast-to-noise ratio in intracranial time-of-flight MR angiography. JMRI 1996; 1:99-108.
  29. Fullerton GD. Physiologic basis of magnetic relaxation In: Stark DD, Bradley WG Jr, eds. Magnetic resonance imaging. 2nd ed. St Louis, Mo: Mosby–Year Book, 1992.
  30. Thulborn KR, Waterton JC, Matthews PM, Radda GK. Oxygenation dependence of the transverse relaxation time of water protons in whole blood at high field. Biochem Biophys Acta 1982; 288:354-361.
  31. Wright GA, Hu BS, Macovski A. Estimating oxygen saturation of blood in vivo with MR imaging at 1.5T. JMRI 1991; 1:275-283.
  32. McGowan JC, Schnall MD, Leigh JS. Magnetization transfer imaging with pulsed off-resonance saturation: variation in contrast with saturation duty cycle. JMRI 1994; 4:79-82.
  33. Loevner LA, Grossman RI, Cohen JA, Lexa FJ, Kessler D, Kolson DL. Microscopic disease in normal-appearing white matter on conventional MR images in patients with multiple sclerosis: assessment with magnetization-transfer measurements. Radiology 1995; 195:511-515.
  34. Pike GB, Glover GH, Hu BS, Enzmann DR. Pulsed magnetization transfer spin-echo MR imaging. JMRI 1993; 3:531-539.
  35. Hua J, Hurst GC. Analysis of on- and off-resonance magnetization transfer techniques. JMRI 1995; 5:113-120.
  36. Santyr GE. Magnetization transfer effects in multislice MR imaging. Magn Reson Imaging 1993; 11:521-532.[Medline]
  37. Constable RT, Anderson AW, Zhong J, Gore JC. Factors influencing contrast in fast spin-echo MR imaging. Magn Reson Imaging 1992; 10:497-511.[Medline]




This Article
Right arrow Abstract Freely available
Right arrow Figures Only
Right arrow Full Text (PDF)
Right arrow Submit a response
Right arrow Alert me when this article is cited
Right arrow Alert me when eLetters are posted
Right arrow Alert me if a correction is posted
Right arrow Citation Map
Services
Right arrow Email this article to a friend
Right arrow Similar articles in this journal
Right arrow Similar articles in PubMed
Right arrow Alert me to new issues of the journal
Right arrow Download to citation manager
Right arrow reprints & permissions
Citing Articles
Right arrow Citing Articles via Google Scholar
Google Scholar
Right arrow Articles by Graham, S. J.
Right arrow Articles by Henkelman, R. M.
Right arrow Search for Related Content
PubMed
Right arrow PubMed Citation
Right arrow Articles by Graham, S. J.
Right arrow Articles by Henkelman, R. M.

HOME HELP FEEDBACK SUBSCRIPTIONS ARCHIVE SEARCH TABLE OF CONTENTS
RADIOLOGY RADIOGRAPHICS RSNA JOURNALS ONLINE