- Email: [email protected]
MR imaging at high magnetic fields
European Journal of Radiology 46 (2003) 45 /52 www.elsevier.com/locate/ejrad
MR imaging at high magnetic fields Masaya Takahashi a,*, Hidemasa Uematsu b, Hiroto Hatabu a a b
Department of Radiology, Beth Israel Deaconess Medical Center, Boston, MA 02115, USA Department of Radiology, University of Pennsylvania Medical Center, Philadelphia, PA, USA
Received 12 November 2002; received in revised form 13 November 2002; accepted 14 November 2002
Abstract Recently, more investigators have been applying higher magnetic field strengths (3 /4 Tesla) in research and clinical settings. Higher magnetic field strength is expected to afford higher spatial resolution and/or a decrease in the length of total scan time due to its higher signal intensity. Besides MR signal intensity, however, there are several factors which are magnetic field dependent, thus the same set of imaging parameters at lower magnetic field strengths would provide differences in signal or contrast to noise ratios at 3 T or higher. Therefore, an outcome of the combined effect of all these factors should be considered to estimate the change in usefulness at different magnetic fields. The objective of this article is to illustrate the practical scientific applications, focusing on MR imaging, of higher magnetic field strength. First, we will discuss previous literature and our experiments to demonstrate several changes that lead to a number of practical applications in MR imaging, e.g. in relaxation times, effects of contrast agent, design of RF coils, maintaining a safety profile and in switching magnetic field strength. Second, we discuss what will be required to gain the maximum benefit of high magnetic field when the current magnetic field ( 5/1.5 T) is switched to 3 or 4 T. In addition, we discuss MR microscopy, which is one of the anticipated applications of high magnetic field strength to understand the quantitative estimation of the gain benefit and other considerations to help establish a practically available imaging protocol. # 2002 Elsevier Science Ireland Ltd. All rights reserved. Keywords: Magnetic resonance imaging; Higher magnetic field strength; Contrast agent
1. Introduction Thanks to recent technological development, wholebody magnetic resonance (MR) scanners at higher magnetic field strengths (]/3 T) have been introduced into research and clinical settings. In the beginning, one of the main reasons to install higher fields was its higher sensitivity to the blood oxygenation level-dependent effect for functional MR imaging of the brain [1]. Recently, more investigators applied these higher magnetic field strengths to both research and conventional clinical settings. The expectation for higher magnetic fields in MRI is the improvement in signal-to-noise ratio (SNR) due to higher signal intensity (SI), where the most significant benefit is to decrease the length of time required to obtain images. Then, higher spatial resolution may be achievable. One question is how it improves
* Corresponding author. Tel.: /1-617-667-0198; fax: /1-617-6677021. E-mail address: [email protected] (M. Takahashi).
or practically how beneficial it is when we switch the current magnetic field (5/1.5 T) to 3 or 4 T. Several studies have reported and discussed the advantages of higher magnetic field in, for example, delineation of various brain lesions [1] or cardiac structures [2,3]. Dougherty et al. [2] reported that the SNR of the anterior myocardium at 4 T was 2.9 times higher than that of the same region at 1.5 T. Bernstein et al. demonstrated contrast enhanced imaging at 3 T and concluded that higher spatial resolution at 3 T could improve diagnostic accuracy [4]. In addition, if higher magnetic field can provide better image quality, it may be reasonable to expect a reduction in total injection of contrast agent, for example, in MR angiography which needs to cover a larger area of the peripheral artery [5] or the lung [6,7]. However, such speculation would be difficult to prove as higher magnetic fields change other imaging aspects besides SNR. Many theoretical and experimental studies have been employed to demonstrate the magnetic field dependencies. Besides SNR, the magnetic field-dependence is
0720-048X/02/$ - see front matter # 2002 Elsevier Science Ireland Ltd. All rights reserved. PII: S 0 7 2 0 - 0 4 8 X ( 0 2 ) 0 0 3 3 1 - 5
46
M. Takahashi et al. / European Journal of Radiology 46 (2003) 45 /52
well-documented in tissue relaxation times [8 /10], as well as in MR contrast agent effects (e.g. R1, R2 or R2* relaxivities) [11,12]. SNR depends upon imaging parameters, RF coil sensitivity and machine adjustments, such as magnetic field homogeneity, accuracy in excitation/refocusing pulse settings, etc. These theoretical and experimentally proven properties suggest that imaging parameters must be reconfigured for different magnetic fields. Unlike relaxation time and MR contrast agent effects, the benefit to signal intensity at higher magnetic field should be compared under nearly identical experimental conditions. Therefore, it is imperative to quantify the practical differences in terms of SNR and contrast-to-noise ratios (CNR) between higher and lower (B/1.5 T) magnetic fields. However, the studies of direct comparisons between SNRs and CNRs as an outcome of the combined effect of several magnetic field-dependent parameters at different fields compared with the theoretical values are substantially sparse. Hence, it is still unclear how much benefit we can gain in SNR or what we can/should do in switching a current magnetic field strength (5/1.5 T in most cases) to a higher magnetic field. In this article, we consider the magnetic field dependent alterations, e.g. MR signal on the image, relaxation times, effects of contrast agent, design of RF coil and safety profile. Then, we evaluate the scientific expectations for MR imaging on a higher magnetic field to quantify the scientific and technical issues relative to safe human experimentation. Further, the feasibility of MR microscopy, which is one of the expectations of higher fields, is discussed.
2. SI, SNR and CNR The question of optimum field strength has been a subject of intense controversy for over a decade. The interest in higher fields stems from the fact that SNRs increase with field strength (v ), where SI and noise have different magnetic field-dependencies. SI8 (number of spins) (voltage induced by each spin)
(1)
As shown in Eq. (1), theoretically, the signal intensity from a MR experiment is proportional to the square of the static magnetic field (v2) since both ‘number of spins’ that can be observed and ‘voltage induced by each spin’ increase linearly as magnetic field (v ) increases. Noise is proportional to the static magnetic field (v ), when all noise comes from a sample, resulting in an SNR that is proportional to v in the case. On the other hand, noise is proportional to one-quarter of v (v1/4) when all noise comes from the RF coil, resulting in an SNR that is proportional to v7/4. Therefore, SNR can be expected to increase more than 2.7 (/4/1.5) times at
4 than at 1.5 T. If this is true, since the SNR scales as the square root of the number of image averages, the time needed to obtain the same SNR is reduced by a factor of 8. To confirm this theory, we imaged the brain in a subject at both fields. To make our comparison between the magnetic fields as direct as possible, the same sets of experiments in the same subjects were conducted at both 4 and 1.5 T on the commercially supplied whole-body MR scanners (SignaTM, General Electric Systems, Milwaukee, WI) with the equipped head coils. Fig. 1 shows the T1-weighted images (top) and T2-weighted images (bottom) obtained in the same level of the brain of the same subject. Each image was obtained with a conventional spin echo sequence with the same imaging parameters at 1.5 and 4 T, respectively. These images showed different tissue contrast between the magnetic fields even though the images were acquired with the same set of imaging parameters. In the quantitative measurements of SI, we found that 4 T increased the SI in both white and gray matter (Fig. 1). In addition, those enhancement ratios were also different between the imaging parameters (T1-WI and T2-WI). Thus, 4 T provides a different tissue contrast compared with 1.5 T using the same set of imaging parameters, which might be inconsistent with theoretical values.
3. Relaxation times As discussed above, SNR in biological tissue was found to be in approximate proportion to field strength. However, the practically achievable SNR gain may be somewhat less since the above theory assumes that all parameters except the magnetic field are consistent. One reason for the discrepancy is the increase in T1 relaxation time with increasing field strength. SI is a function of relaxation time that is, in turn, magnetic fielddependent [3]. In theory, T1 value increases in a magnetic field-dependent manner in most biological tissues of which the correlation time (tc) of tissue water is :/10 8 s [13], whereas T2 value does not change (Fig. 2). Comparisons of relaxation times in humans have been published in the literature. Jezzard et al. and Duewell et al. presented a comparison of T1 and T2 relaxation times in human subjects between 1.5 and 4 T in the brain and several peripheral regions [9,10] (Table 1). In any tissue, T1 relaxation times are prolonged at a higher magnetic field, while T2 relaxation times are somewhat shortening. Those results are consistent with previous reports (Fig. 2). To confirm this phenomenon, we conducted the same set of phantom experiments at both 4 and 1.5 T on the same whole-body MR scanners with head coils [14]. Phantoms included different concentrations of Gd-complex aqueous solution with each phantom representing tissue with a different T1 relaxa-
M. Takahashi et al. / European Journal of Radiology 46 (2003) 45 /52
47
Fig. 1. T1- and T2-weighted images of a human subject obtained at 1.5 and 4 Tesla. Each image was acquired with the same set of imaging parameters (TR/TE is indicated in the parentheses), respectively. Note that different magnetic fields provided different image contrast.
Fig. 3. Cross-sectional T1-weighted image of a fixed excised spinal cord of the larval sea lamprey. Image was obtained at 9.4 T experimental machine; resolution was 9/9 mm resolution. See Ref. [27].
Fig. 2. Magnetic field dependency in T1 and T2 relaxation times, modified from Ref. [13].
tion time. In this study, the trains of spin echo images with varied TRs or TEs were obtained with the same commercial clinical scanners with the head coils described above. The relaxation times (T1, T2) for all phantoms were determined at both 1.5 and 4 T from the fitting curves. The results in this confirmatory study demonstrated that any T1 relaxation times were prolonged (1.10/1.47 times) at 4 T compared with those at
1.5 T, while T2 values were identical or slightly shortened (Table 2). Further, a standard contrast-enhanced MR angiographic sequence (3D spoiled gradient recalled acquisition or SPGR) sequence with the same imaging parameters was utilized to confirm changes in SI. Peak SNRs at 4 T increased at least 2.21 times higher compared with those at 1.5 T. Moreover, peak CNRs at 4 T increased at least 1.59 times higher compared with those at 1.5 T in the range of Gd concentrations expected during clinical use. In addition, those enhancements of SNR and CNR were a function of a flip angle that we used. Based on those results, using higher
M. Takahashi et al. / European Journal of Radiology 46 (2003) 45 /52
48
Table 1 Comparison of T1 and T2 relaxation times in human subject [9,10] Tissue
T1 (s)
T2 (ms)
1.5 T
4T
1.5 T
4T
0.9 /1.3* 0.7 /1.1* 0.98 0.31 0.29
1.72 1.04 1.83 0.39 0.42
77 /90 62 /80 31 47 47
63 50 26 38 42
a
Brain Gray matter White matter Muscleb Fatb Bone marrowb a
Lezzard et al. [9]. Duewell et al. [10]. * From previous literature. b
Table 2 Comparison of T1 and T2 relaxation time in gadolinium doped water solution at room temperature, modified Ref. [14] Gd concentration (mmol/l)
0 0.125 0.5 1.25 2.5 5
T1 (ms)
T2 (ms)
1.5 T
4T
1.5 T
4T
2556 1067 419 191 123 67
3636 1566 562 253 142 81
1643 911 348 160 84 43
1504 862 351 160 83 42
At room temperature.
magnetic fields seems to be beneficial in CNRs as well as in SNRs even without optimization of imaging parameters at each magnetic field. A relationship between the SI of a gradient echo sequence, the relaxation time and the optimal flip angle (ao: Ernst angle), can be expressed as follows: SI b×
[1 exp(TR=T1)] × exp(TE=T2) × sin a 1 exp(TR=T1) × cos a
(2)
and cos ao exp(TR=T1)
reflected in the TR, the SNR per unit time is optimized with an Ernst angle pulse and the shortest achievable value of TR/T1. The necessity of optimization of imaging parameters was presented in a previous work. Keiper et al. [15] compared the usefulness in the diagnosis of white matter abnormalities in multiple sclerosis patients following the optimization of imaging parameters between 1.5 and 4 T. Their results demonstrated that MR imaging at 4 T (512 /256 matrix) could depict smaller lesions that could not be detected at 1.5 T (256 /192 matrix), implying that the higher resolution at 4 T provides higher accuracy of diagnosis in the same patients with almost identical total scan time. Although T2 values were substituted for T2* in the phantom study because T2 and T2* values should be theoretically identical in phantoms in each magnetic field [16], it is considered to be different from the conditions in some tissues where the T2* value is much shorter than the T2 value in some tissues. A magnitude of susceptibility (g ) is proportional to the magnetic field as shown in the following equation [17]: Dx B0 g (4) 2 RGz where Dx is the difference in magnetic susceptibility of adjoining substances, B0 (/v ) is the static magnetic field, R is the cross section radius and Gz is the read-out gradient. However, this effect on T2* depends on T2 in tissue since 1/T2* is a function of T2 and T2? (R2*/ R2/R2?) [18]. The shorter T2 and T2* values at a higher magnetic field may cause a larger decrease in the SNR and CNR than would be expected in some tissue, such as the lung. Previously, we found that the CNR increased in the central arteries of the lung, but did not increase in the pulmonary peripheral arteries at 4 T as the dose of contrast agent increased, ranging from 0.05 to 0.2 mmol/kg body weight [19]. Therefore, the optimal imaging parameters for the clinical application should be carefully considered, particular when an undesirable T2* effect may be involved.
(3)
where b is the scaling factor and a is the flip angle. SI is determined by its relaxation times (T1 and T2*) in individual tissue conditions in any imaging sequence. This implies that the same intensity will not be obtained with the same set of imaging parameters due to the alternation of relaxation times at different magnetic field. Since T1 values at higher magnetic field are longer than those at lower magnetic field, the TR, presumably as well as the flip angle, should be longer (smaller for flip angle) to optimize the SNR of the same sample at the higher field. Using longer TR, the advantage in SI at a higher field would be less in unit time. In other words, since the primary limitation imposed by long T1 relaxation time at higher magnetic field strength is
4. Relaxivities of Gd-complex The R1 relaxivity of MR contrast agent is dependent upon various parameters, such as the type of contrast agent [20], temperature and tissue environment as well as magnetic field strength [11,12]. R1 relaxivity of a paramagnetic contrast agent is higher at lower field strength [11]. R2 and R2* values should be theoretically identical in phantoms in each magnetic field [16]. In the phantom study described above, the authors attempted to compare the effects of contrast agent. For an accurate determination of the efficacy of Gd-complex (R1, R2 and R2*), only some of the relaxation times
M. Takahashi et al. / European Journal of Radiology 46 (2003) 45 /52
(T1, T2) that could be excellently fitted to the curve (r / 0.995) were reciprocally plotted against the concentrations of Gd at both 4 and 1.5 T. As a result, R1 and R2 relaxivity values were determined to be 2.95 and 4.82 (l ×/ s 1 ×/mmol 1) at 4 T and 3.89 and 4.67 (l ×/s1 ×/mmol 1) at 1.5 T, respectively. R1 at 4 T was lower ( :/25%) than R1 at 1.5 T, while the R2 at 4 T was almost that at 1.5 T (Table 3). Hence, we found that R1 relaxivity decreases as the magnetic field strength increases, while R2 relaxivity does not change as much, which is consistent with previous reports [16]. Unlike Gd-complex, R2 and R2* might be considerably changed depending upon the type of contrast agent (e.g. super paramagnetic iron oxide: SPIO), application root and/or tissues. This suggested that we should also consider the use of the MR contrast agent, though it is not clear whether this change is substantially effective in current clinical usage at higher magnetic field.
6. Safety consideration Theoretical calculations of the interaction of high magnetic fields with human subjects have been reviewed. To date, no hazardous physical or physiological phenomena have been shown. The mechanism considered included orientation of macromolecules and membranes, effects on nerve conduction, electrocardiograms and electroencephalograms, and blood flow. The most current clinical MR imagers at lower magnetic field (5/1.5 T) equip up to 25 mT/m. If higher magnetic fields are to be used to archive higher spatial resolution, the gradient strength must increase. In the combination of higher statistic magnetic field and gradients, strength may be an issue in some applications due to limitations in the current FDA guidelines for specific absorption rate (SAR). SAR is defined as follow: SAR
5. RF coil The application of higher magnetic field strengths to MR imaging (particular in whole body imaging) is more demanding because of the difficulty in building RF coils since the penetration of radio frequency into the tissue becomes harder [3,21]. It is necessary to understand the relationship between SNR and RF coil, since an incomplete RF coil may sacrifice the advantage in SNR at increased magnetic field strength. RF coil characteristics, especially a receive coil, significantly impact SNR. SNR increases with decreasing coil diameter. Thus, the coil sensitivity of the head coil is :/3-fold higher than that of the body coil. The surface coil with smaller diameter gains more sensitivity, whereas the SNR drops off very rapidly with increasing depth from the surface. To cover these difficulties, an array of surface coils must be developed. Reported by Wright et al. [22], another idea to increase coil sensitivity and further improve SNR is to reduce coil temperature, thus lowering its resistance and thermal noise voltages and increasing its Q , while keeping the sample at room temperature. The cryogenic SNR gain would be greatest for coil and sample configurations having QL/QU close to 1. Table 3 Comparison of relaxivities of gadolinium (Gd)-complex in aqueous solution, modified Ref. [14] Relaxivities
1.5 T
4T
R1 R2 R2*
3.89 4.67 4.55
2.95 4.82 4.67
* mmol 1 s 1 at room temperature.
49
sjEj2 2r
t
TR
NP NS
(5)
where s is conductivity, E is the electric field, r is tissue density, t is pulse duration and NP and NS are number of pulses and image slices, respectively. Since E is proportional to static magnetic field, SAR greatly increases at higher magnetic field, which may limit the application in number of slices, selection of flip angle, etc. Additionally, RF energy is absorbed more effectively at higher frequencies; RF absorption, as expressed by SAR, must be carefully monitored. This could be a major concern in any application at high field strength as Bottomley et al. previously suggested [21].
7. MR microscopy In using a higher magnetic field, the investigators expect images with higher spatial resolution to be more beneficial in research and clinical settings. Recently, transgenic and genetically engineered mouse models have been used increasingly and have led to important advances in many scientific communities. Thus, there has been an increased demand to image mice in vivo with a microscopic method. Whereas micro-computed tomography (m-CT) can potentially generate higherresolution images [23], MR is unique in that it is also able to provide detailed information on anatomy and function of soft tissues. Many excellent works have been reported in transgenic mice model genotype to phenotype [24,25], in which most of the cases were conducted with an experimental image scanner at a high magnetic field. Besides transgenic imaging, measurement of apparent diffusion coefficients of water molecules was
50
M. Takahashi et al. / European Journal of Radiology 46 (2003) 45 /52
performed in a single neuron [26]. More recently, a single axon in an excised lamprey spinal cord (Fig. 3) [27] was demonstrated and an apparent diffusion coefficient of the single axon was measured [28]. The techniques of very high resolution MR imaging have been developed largely in the past decade [26]. MR microscopy has been developed in trabecular bone imaging at the clinical magnetic field strength B/1.5 T. There are at least two reasons for applying MR microscopy to bone imaging. First, high signal contrast is raised between bones and surrounding tissue (bone marrow). Second, a smaller RF coil can be designed for the wrist or ankle. It is amenable to micromorphometry ex vivo and in vivo in laboratory animals and even in humans [29 /32]. The authors have provided new and interesting information on the use of quantitative in vivo MR microscopy and spectroscopy in conjunction with digital image processing to evaluate the epiphyseal and metaphyseal tissues of rabbits treated with dexamethasone at a 1.5 T clinical scanner [33]. One of the difficulties of imaging microarchitecture in vivo is achievement of sufficient resolution and SNR to resolve individual structures. The capability for direct visualization will have implications for acquiring sufficient image quality in vivo. As we discuss below, the higher the SNR, the longer the total scan time. However, we need to consider the interface between the desire for reasonable data acquisition times and adequate SNR, in particular in vivo imaging. Here, we discuss the imaging factors regulating spatial resolution in considering the protocol for a high resolution image at high magnetic field. SNR is the most limiting factor to increased spatial resolution in MR imaging, since scan time for a given SNR scales as the inverse sixth power of the linear voxel dimension [34]. SNR is primarily a function of voxel size, which is determined by the number of samples in the phase encoding and frequency encoding directions (in-plane resolution), the slice thickness (d) and the field of view in both directions, frequency (FOVf) and phase (FOVp) encodings. Hence, decreasing voxel size either by decreasing field of view or slice thickness or by increasing the matrix size, decreases SNR [35]. Therefore, any parameter determining voxel volume will also affect SNR, where SNR depends upon FOV quadratically in the same matrix size. The noise averaging process of repeated sampling has important implications on SNR. In theory [35], SNR increases as the square root of the number of samples collected. This holds true for the number of phase encodings (Np), the number of frequency encodings (Nf) as well as the number of signal acquisitions or number of excitations (NEX). The combining factors above can be summarized as follows:
pffiffiffiffiffiffiffiffiffiffiffi FOVf FOVp NEX d SNR pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi Nf × Np
(6)
Thus, it is true that changing the number of frequency and phase encodings affects both voxel size and signal averaging. The net effect is an inverse square root relationship between SNR and the product of phase and frequency encoding. By contrast, SNR scales as the square root of the number of excitations. Hence, doubling SNR requires quadrupling of NEX, which exacts a scan time penalty. Spatial resolution is typically expressed in terms of pixel size that is determined as the ratio of FOV divided by the number of phase or frequency encodings. Hence, we can decide spatial resolution in either of two ways: by manipulating the FOV or the matrix size. Changing slice thickness also affects resolution, albeit in a different way. Increasing slice thickness causes increased partial volume blurring. The effect of pixel size on image is demonstrated in Fig. 4. Fig. 4(A) shows a 3D projection image of the distal femoral epiphysis of a live rabbit obtained at a 1.5 T clinical scanner with 98/98/300 mm3 spatial resolution. The total scan time was :/20 min. Fig. 4(B) demonstrates a 3D projection of the small trabecular bone specimen from the proximal tibiae in rats. Locations have been matched to a cube highlighted in Fig. 1(A). The imaging was performed on a 9.4 T experimental machine with a total scan time of :/55 min to afford 39 mm isotropic voxel. Comparing these two images, the reduction in voxel size from 98 /98/300 mm3 in vivo conducted at 1.5 T (Fig. 4A) to 39 mm3 ex vivo at 9.4 T (Fig. 4B) presently entails an approximate 50-fold SNR penalty. This could not be recovered fully by the magnetic statistic field increase from 1.5 to 9.4 T when the RF coil insert was the same and had the same sensitivity. This obviously means that the RF power is dissipated beneficially by increasing Q in smaller RF coils for SNR gain. Basically, RF coil sensitivity (Q dumpling) is reciprocal in proportion to its diameter [21]. According to the discussion above, the achievable benefit in SNR might be a factor of 4 after optimization of imaging parameters when the magnetic field strength is changed from 1.5 to 3 or 4 T in clinical situations, where the RF coils have almost the same sensitivity between the fields. This contributes to image resolution increasing at a factor of up to 4 in the same scan time, e.g. 128/128 to 256 /256 matrix. In this case, the slice thickness should be kept constant. When we prefer to save scan time, we can reduce the number of excitations by half, since SNR is proportional to the square of the total scan time. If we want further advances, development of RF coil, image acquisition, restoration and processing techniques should be involved. We estimated the benefits for MR imaging when the current MR scanner with magnetic field strength of
M. Takahashi et al. / European Journal of Radiology 46 (2003) 45 /52
51
Fig. 4. (A) Three dimensional (3D) projection image of the distal femoral epiphysis of a live rabbit covering the volume analyzed. Projection direction is inferior to superior at an angle of 308 relative to the femoral anatomic axis. (B) 3D projection image of the small trabecular bone specimen from the proximal tibiae in rats. Locations have been matched to a cube highlighted in (A). (A) and (B) were obtained on 1.5 T clinical and 9.4 T experimental machines in the total scan time of :/20 and 55 min, respectively.
B/1.5 T is replaced with a higher field up to 3 or 4 T. A number of practical implications in the imaging of biological tissues at higher field strength must be considered. The gain in SNR from the higher magnetic field strength may be substantially offset by prolonged T1 relaxation times, thus optimization of the imaging parameters is important. Although it was not discussed in this review, higher magnetic field must produce better frequency resolution of near degenerated resonances that are not resolvable at lower field in magnetic resonance spectroscopy. Another motivation for high magnetic field is the ability to use other nuclei (e.g. 23Na, 39 K), rather than protons, for which sensitivities are not sufficient to be observed at lower magnetic field. Recently, whole-body MR scanners with much higher magnetic fields (8 /10 T) have been developed and some have been applied in human studies. In conclusion, MR imaging at a higher magnetic field strength (]/3 Tesla) will be opening a new arena. The appropriate optimizations, such as image acquisition, development in RF coil design and image processing algorithms with adequate safety profiles, would expand the applications.
Acknowledgements We would like to thank Dr Shigeru Kiryu at Beth Israel Deaconess Medical Center/Harvard Medical School for his assistance in article preparation.
References [1] Thulborn KR. Clinical rationale for very-high-field (3.0 Tesla) functional magnetic resonance imaging. Top Magn Reson Imaging 1999;10:37 /50. [2] Dougherty L, Connick TJ, Mizsei G. Cardiac imaging at 4.0 Tesla. Magn Reson Med 2001;45:176 /8.
[3] Noeske R, Seifert F, Rhein KH, Rinneberg H. Human cardiac imaging at 3 T using phased array coils. Magn Reson Med 2000;44:978 /82. [4] Bernstein MA, Huston J, III, Lin C, Gibbs GF, Felmlee JP. Highresolution intracranial and cervical MRA at 3.0 T: technical considerations and initial experience. Magn Reson Med 2001;46:955 /62. [5] Boos M, Lentschig M, Scheffler K, Bongartz GM, Steinbrich W. Contrast-enhanced magnetic resonance angiography of peripheral vessels. Different contrast agent applications and sequence strategies: a review. Invest Radiol 1998;33:538 /46. [6] Hatabu H, Gaa J, Kim D, Li W, Prasad PV, Edelman RR. Pulmonary perfusion and angiography: evaluation with breathhold enhanced three-dimensional fast imaging steady-state precession MR imaging with short TR and TE. Am J Roentgenol 1996;167:653 /5. [7] Hany TF, Schmidt M, Hilfiker PR, Steiner P, Bachmann U, Debatin JF. Optimization of contrast dosage for gadoliniumenhanced 3D MRA of the pulmonary and renal arteries. Magn Reson Imaging 1998;16:901 /6. [8] Duewell SH, Wolff SD, Wen H, Balaban RS, Jezzard P. MR imaging contrast in human brain tissue: assessment and optimization at 4 T. Radiology 1996;199:780 /6. [9] Jezzard P, Duewell S, Balaban RS. MR relaxation times in human brain: measurement at 4 T. Radiology 1996;199:773 /9. [10] Duewell SH, Ceckler TL, Ong K, et al. Musculoskeletal MR imaging at 4 T and at 1.5 T: comparison of relaxation times and image contrast. Radiology 1995;196:551 /5. [11] Lauffer RB. Magnetic resonance contrast media: principles and progress. Magn Reson Q 1990;6:65 /84. [12] Vander Elst L, Laurent S, Muller RN. Multinuclear magnetic resonance characterization of paramagnetic contrast agents. The manifold effects of concentration and counterions. Invest Radiol 1998;33:828 /34. [13] Gadian DG. In: Gadian DG, editor. Nuclear magnetic resonance and its applications to living systems. New York: Oxford University Press, 1982. [14] Uematsu H, Dougherty L, Takahashi M, et al. A direct comparison of signal behavior between 4.0 Tesla and 1.5 Tesla: a phantom study. Eur J Radiol 2002;45:154 /159. [15] Keiper MD, Grossman RI, Hirsch JA, et al. MR identification of white matter abnormalities in multiple sclerosis: a comparison between 1.5 T and 4 T. Am J Neuroradiol 1998;19:1489 /93. [16] Fernandez-Seara MA, Wehrli FW. Postprocessing technique to correct for background gradients in image-based R*(2) measurements. Magn Reson Med 2000;44:358 /66.
52
M. Takahashi et al. / European Journal of Radiology 46 (2003) 45 /52
[17] Ludeke KM, Roschmann P, Tischler R. Susceptibility artefacts in NMR imaging. Magn Reson Imaging 1985;3:329 /43. [18] Takahashi M, Wehrli FW, Hwang SN, Wehrli SL. Relationship between cancellous bone induced magnetic field and ultrastructure in a rat ovariectomy model. Magn Reson Imaging 2000;18:33 /9. [19] Uematsu H, Dougherty L, Takahashi M, et al. Pulmonary MR angiography with contrast agent at 4 Tesla: a preliminary result. Magn Reson Med 2001;46:1028 /30. [20] Takahashi M, Tsutsui H, Murayama C, Miyazawa T, FritzZieroth B. Neurotoxicity of gadolinium contrast agents for magnetic resonance imaging in rats with osmotically disrupted blood /brain barrier. Magn Reson Imaging 1996;14:619 /23. [21] Bottomley PA, Andrew ER. RF magnetic field penetration, phase shift and power dissipation in biological tissue: implications for NMR imaging. Phys Med Biol 1978;23:630 /43. [22] Wright AC, Song HK, Wehrli FW. In vivo MR micro imaging with conventional radiofrequency coils cooled to 77 K. Magn Reson Med 2000;43:163 /9. [23] Hildebrand T, Laib A, Muller R, Dequeker J, Ruegsegger P. Direct three-dimensional morphometric analysis of human cancellous bone: microstructural data from spine, femur, iliac crest, and calcaneus. J Bone Miner Res 1999;14:1167 /74. [24] Johnson GA, Cofer GP, Gewalt SL, Hedlund LW. Morphologic phenotyping with MR microscopy: the visible mouse. Radiology 2002;222:789 /93. [25] Sze RW, Chan CB, Dardzinski BJ, et al. Three-dimensional MR microscopy of a transgenic mouse model of dilated cardiomyopathy. Pediatr Radiol 2001;31:55 /61. [26] Aguayo JB, Blackband SJ, Schoeniger J, Mattingly MA, Hintermann M. Nuclear magnetic resonance imaging of a single cell. Nature 1986;322(6075):190 /1.
[27] Wright AC, Wehrli SL, Zhang G, et al. Visualization of individual axons in excised lamprey spinal cord by magnetic resonance microscopy. J Neurosci Methods 2002;114:9 /15. [28] Takahashi M, Hackney DB, Zhang G, et al. Magnetic resonance microimaging of intra-axonal water diffusion in live excised lamprey spinal cord. Proc Natl Acad Sci USA 2002;99:16192 /6. [29] Ma J, Wehrli FW, Song HK. Fast 3D large-angle spinecho imaging (3D FLASE). Magn Reson Med 1996;35:903 / 10. [30] Majumdar S, Newitt D, Mathur A, et al. Magnetic resonance imaging of trabecular bone structure in the distal radius: relationship with X-ray tomographic microscopy and biomechanics. Osteoporos Int 1996;6:376 /85. [31] Gordon CL, Webber CE, Christoforou N, Nahmias C. In vivo assessment of trabecular bone structure at the distal radius from high-resolution magnetic resonance images. Med Phys 1997;24:585 /93. [32] Takahashi M, Wehrli FW, Wehrli SL, et al. Effect of prostaglandin and bisphosphonate on cancellous bone volume and structure in the ovariectomized rat studied by quantitative 3D NMR microscopy. J Bone Miner Res 1999;14:680 /9. [33] Takahashi M, Wehrli FW, Hilaire L, Zemel BS, Hwang SN. In vivo NMR microscopy allows short-term serial assessment of multiple skeletal implications of corticosteroid exposure. Proc Natl Acad Sci USA 2002;99:4574 /9. [34] Callaghan PT. Principles of nuclear magnetic resonance microscopy. Oxford: Clarendon Press, 1991. [35] Wehrli FW, Ford JC, Chung HW, et al. Potential role of nuclear magnetic resonance for the evaluation of trabecular bone quality. Calcif Tissue Int 1993;53(Suppl 1):S162 /9.
MR imaging at high magnetic fields Masaya Takahashi a,*, Hidemasa Uematsu b, Hiroto Hatabu a a b
Department of Radiology, Beth Israel Deaconess Medical Center, Boston, MA 02115, USA Department of Radiology, University of Pennsylvania Medical Center, Philadelphia, PA, USA
Received 12 November 2002; received in revised form 13 November 2002; accepted 14 November 2002
Abstract Recently, more investigators have been applying higher magnetic field strengths (3 /4 Tesla) in research and clinical settings. Higher magnetic field strength is expected to afford higher spatial resolution and/or a decrease in the length of total scan time due to its higher signal intensity. Besides MR signal intensity, however, there are several factors which are magnetic field dependent, thus the same set of imaging parameters at lower magnetic field strengths would provide differences in signal or contrast to noise ratios at 3 T or higher. Therefore, an outcome of the combined effect of all these factors should be considered to estimate the change in usefulness at different magnetic fields. The objective of this article is to illustrate the practical scientific applications, focusing on MR imaging, of higher magnetic field strength. First, we will discuss previous literature and our experiments to demonstrate several changes that lead to a number of practical applications in MR imaging, e.g. in relaxation times, effects of contrast agent, design of RF coils, maintaining a safety profile and in switching magnetic field strength. Second, we discuss what will be required to gain the maximum benefit of high magnetic field when the current magnetic field ( 5/1.5 T) is switched to 3 or 4 T. In addition, we discuss MR microscopy, which is one of the anticipated applications of high magnetic field strength to understand the quantitative estimation of the gain benefit and other considerations to help establish a practically available imaging protocol. # 2002 Elsevier Science Ireland Ltd. All rights reserved. Keywords: Magnetic resonance imaging; Higher magnetic field strength; Contrast agent
1. Introduction Thanks to recent technological development, wholebody magnetic resonance (MR) scanners at higher magnetic field strengths (]/3 T) have been introduced into research and clinical settings. In the beginning, one of the main reasons to install higher fields was its higher sensitivity to the blood oxygenation level-dependent effect for functional MR imaging of the brain [1]. Recently, more investigators applied these higher magnetic field strengths to both research and conventional clinical settings. The expectation for higher magnetic fields in MRI is the improvement in signal-to-noise ratio (SNR) due to higher signal intensity (SI), where the most significant benefit is to decrease the length of time required to obtain images. Then, higher spatial resolution may be achievable. One question is how it improves
* Corresponding author. Tel.: /1-617-667-0198; fax: /1-617-6677021. E-mail address: [email protected] (M. Takahashi).
or practically how beneficial it is when we switch the current magnetic field (5/1.5 T) to 3 or 4 T. Several studies have reported and discussed the advantages of higher magnetic field in, for example, delineation of various brain lesions [1] or cardiac structures [2,3]. Dougherty et al. [2] reported that the SNR of the anterior myocardium at 4 T was 2.9 times higher than that of the same region at 1.5 T. Bernstein et al. demonstrated contrast enhanced imaging at 3 T and concluded that higher spatial resolution at 3 T could improve diagnostic accuracy [4]. In addition, if higher magnetic field can provide better image quality, it may be reasonable to expect a reduction in total injection of contrast agent, for example, in MR angiography which needs to cover a larger area of the peripheral artery [5] or the lung [6,7]. However, such speculation would be difficult to prove as higher magnetic fields change other imaging aspects besides SNR. Many theoretical and experimental studies have been employed to demonstrate the magnetic field dependencies. Besides SNR, the magnetic field-dependence is
0720-048X/02/$ - see front matter # 2002 Elsevier Science Ireland Ltd. All rights reserved. PII: S 0 7 2 0 - 0 4 8 X ( 0 2 ) 0 0 3 3 1 - 5
46
M. Takahashi et al. / European Journal of Radiology 46 (2003) 45 /52
well-documented in tissue relaxation times [8 /10], as well as in MR contrast agent effects (e.g. R1, R2 or R2* relaxivities) [11,12]. SNR depends upon imaging parameters, RF coil sensitivity and machine adjustments, such as magnetic field homogeneity, accuracy in excitation/refocusing pulse settings, etc. These theoretical and experimentally proven properties suggest that imaging parameters must be reconfigured for different magnetic fields. Unlike relaxation time and MR contrast agent effects, the benefit to signal intensity at higher magnetic field should be compared under nearly identical experimental conditions. Therefore, it is imperative to quantify the practical differences in terms of SNR and contrast-to-noise ratios (CNR) between higher and lower (B/1.5 T) magnetic fields. However, the studies of direct comparisons between SNRs and CNRs as an outcome of the combined effect of several magnetic field-dependent parameters at different fields compared with the theoretical values are substantially sparse. Hence, it is still unclear how much benefit we can gain in SNR or what we can/should do in switching a current magnetic field strength (5/1.5 T in most cases) to a higher magnetic field. In this article, we consider the magnetic field dependent alterations, e.g. MR signal on the image, relaxation times, effects of contrast agent, design of RF coil and safety profile. Then, we evaluate the scientific expectations for MR imaging on a higher magnetic field to quantify the scientific and technical issues relative to safe human experimentation. Further, the feasibility of MR microscopy, which is one of the expectations of higher fields, is discussed.
2. SI, SNR and CNR The question of optimum field strength has been a subject of intense controversy for over a decade. The interest in higher fields stems from the fact that SNRs increase with field strength (v ), where SI and noise have different magnetic field-dependencies. SI8 (number of spins) (voltage induced by each spin)
(1)
As shown in Eq. (1), theoretically, the signal intensity from a MR experiment is proportional to the square of the static magnetic field (v2) since both ‘number of spins’ that can be observed and ‘voltage induced by each spin’ increase linearly as magnetic field (v ) increases. Noise is proportional to the static magnetic field (v ), when all noise comes from a sample, resulting in an SNR that is proportional to v in the case. On the other hand, noise is proportional to one-quarter of v (v1/4) when all noise comes from the RF coil, resulting in an SNR that is proportional to v7/4. Therefore, SNR can be expected to increase more than 2.7 (/4/1.5) times at
4 than at 1.5 T. If this is true, since the SNR scales as the square root of the number of image averages, the time needed to obtain the same SNR is reduced by a factor of 8. To confirm this theory, we imaged the brain in a subject at both fields. To make our comparison between the magnetic fields as direct as possible, the same sets of experiments in the same subjects were conducted at both 4 and 1.5 T on the commercially supplied whole-body MR scanners (SignaTM, General Electric Systems, Milwaukee, WI) with the equipped head coils. Fig. 1 shows the T1-weighted images (top) and T2-weighted images (bottom) obtained in the same level of the brain of the same subject. Each image was obtained with a conventional spin echo sequence with the same imaging parameters at 1.5 and 4 T, respectively. These images showed different tissue contrast between the magnetic fields even though the images were acquired with the same set of imaging parameters. In the quantitative measurements of SI, we found that 4 T increased the SI in both white and gray matter (Fig. 1). In addition, those enhancement ratios were also different between the imaging parameters (T1-WI and T2-WI). Thus, 4 T provides a different tissue contrast compared with 1.5 T using the same set of imaging parameters, which might be inconsistent with theoretical values.
3. Relaxation times As discussed above, SNR in biological tissue was found to be in approximate proportion to field strength. However, the practically achievable SNR gain may be somewhat less since the above theory assumes that all parameters except the magnetic field are consistent. One reason for the discrepancy is the increase in T1 relaxation time with increasing field strength. SI is a function of relaxation time that is, in turn, magnetic fielddependent [3]. In theory, T1 value increases in a magnetic field-dependent manner in most biological tissues of which the correlation time (tc) of tissue water is :/10 8 s [13], whereas T2 value does not change (Fig. 2). Comparisons of relaxation times in humans have been published in the literature. Jezzard et al. and Duewell et al. presented a comparison of T1 and T2 relaxation times in human subjects between 1.5 and 4 T in the brain and several peripheral regions [9,10] (Table 1). In any tissue, T1 relaxation times are prolonged at a higher magnetic field, while T2 relaxation times are somewhat shortening. Those results are consistent with previous reports (Fig. 2). To confirm this phenomenon, we conducted the same set of phantom experiments at both 4 and 1.5 T on the same whole-body MR scanners with head coils [14]. Phantoms included different concentrations of Gd-complex aqueous solution with each phantom representing tissue with a different T1 relaxa-
M. Takahashi et al. / European Journal of Radiology 46 (2003) 45 /52
47
Fig. 1. T1- and T2-weighted images of a human subject obtained at 1.5 and 4 Tesla. Each image was acquired with the same set of imaging parameters (TR/TE is indicated in the parentheses), respectively. Note that different magnetic fields provided different image contrast.
Fig. 3. Cross-sectional T1-weighted image of a fixed excised spinal cord of the larval sea lamprey. Image was obtained at 9.4 T experimental machine; resolution was 9/9 mm resolution. See Ref. [27].
Fig. 2. Magnetic field dependency in T1 and T2 relaxation times, modified from Ref. [13].
tion time. In this study, the trains of spin echo images with varied TRs or TEs were obtained with the same commercial clinical scanners with the head coils described above. The relaxation times (T1, T2) for all phantoms were determined at both 1.5 and 4 T from the fitting curves. The results in this confirmatory study demonstrated that any T1 relaxation times were prolonged (1.10/1.47 times) at 4 T compared with those at
1.5 T, while T2 values were identical or slightly shortened (Table 2). Further, a standard contrast-enhanced MR angiographic sequence (3D spoiled gradient recalled acquisition or SPGR) sequence with the same imaging parameters was utilized to confirm changes in SI. Peak SNRs at 4 T increased at least 2.21 times higher compared with those at 1.5 T. Moreover, peak CNRs at 4 T increased at least 1.59 times higher compared with those at 1.5 T in the range of Gd concentrations expected during clinical use. In addition, those enhancements of SNR and CNR were a function of a flip angle that we used. Based on those results, using higher
M. Takahashi et al. / European Journal of Radiology 46 (2003) 45 /52
48
Table 1 Comparison of T1 and T2 relaxation times in human subject [9,10] Tissue
T1 (s)
T2 (ms)
1.5 T
4T
1.5 T
4T
0.9 /1.3* 0.7 /1.1* 0.98 0.31 0.29
1.72 1.04 1.83 0.39 0.42
77 /90 62 /80 31 47 47
63 50 26 38 42
a
Brain Gray matter White matter Muscleb Fatb Bone marrowb a
Lezzard et al. [9]. Duewell et al. [10]. * From previous literature. b
Table 2 Comparison of T1 and T2 relaxation time in gadolinium doped water solution at room temperature, modified Ref. [14] Gd concentration (mmol/l)
0 0.125 0.5 1.25 2.5 5
T1 (ms)
T2 (ms)
1.5 T
4T
1.5 T
4T
2556 1067 419 191 123 67
3636 1566 562 253 142 81
1643 911 348 160 84 43
1504 862 351 160 83 42
At room temperature.
magnetic fields seems to be beneficial in CNRs as well as in SNRs even without optimization of imaging parameters at each magnetic field. A relationship between the SI of a gradient echo sequence, the relaxation time and the optimal flip angle (ao: Ernst angle), can be expressed as follows: SI b×
[1 exp(TR=T1)] × exp(TE=T2) × sin a 1 exp(TR=T1) × cos a
(2)
and cos ao exp(TR=T1)
reflected in the TR, the SNR per unit time is optimized with an Ernst angle pulse and the shortest achievable value of TR/T1. The necessity of optimization of imaging parameters was presented in a previous work. Keiper et al. [15] compared the usefulness in the diagnosis of white matter abnormalities in multiple sclerosis patients following the optimization of imaging parameters between 1.5 and 4 T. Their results demonstrated that MR imaging at 4 T (512 /256 matrix) could depict smaller lesions that could not be detected at 1.5 T (256 /192 matrix), implying that the higher resolution at 4 T provides higher accuracy of diagnosis in the same patients with almost identical total scan time. Although T2 values were substituted for T2* in the phantom study because T2 and T2* values should be theoretically identical in phantoms in each magnetic field [16], it is considered to be different from the conditions in some tissues where the T2* value is much shorter than the T2 value in some tissues. A magnitude of susceptibility (g ) is proportional to the magnetic field as shown in the following equation [17]: Dx B0 g (4) 2 RGz where Dx is the difference in magnetic susceptibility of adjoining substances, B0 (/v ) is the static magnetic field, R is the cross section radius and Gz is the read-out gradient. However, this effect on T2* depends on T2 in tissue since 1/T2* is a function of T2 and T2? (R2*/ R2/R2?) [18]. The shorter T2 and T2* values at a higher magnetic field may cause a larger decrease in the SNR and CNR than would be expected in some tissue, such as the lung. Previously, we found that the CNR increased in the central arteries of the lung, but did not increase in the pulmonary peripheral arteries at 4 T as the dose of contrast agent increased, ranging from 0.05 to 0.2 mmol/kg body weight [19]. Therefore, the optimal imaging parameters for the clinical application should be carefully considered, particular when an undesirable T2* effect may be involved.
(3)
where b is the scaling factor and a is the flip angle. SI is determined by its relaxation times (T1 and T2*) in individual tissue conditions in any imaging sequence. This implies that the same intensity will not be obtained with the same set of imaging parameters due to the alternation of relaxation times at different magnetic field. Since T1 values at higher magnetic field are longer than those at lower magnetic field, the TR, presumably as well as the flip angle, should be longer (smaller for flip angle) to optimize the SNR of the same sample at the higher field. Using longer TR, the advantage in SI at a higher field would be less in unit time. In other words, since the primary limitation imposed by long T1 relaxation time at higher magnetic field strength is
4. Relaxivities of Gd-complex The R1 relaxivity of MR contrast agent is dependent upon various parameters, such as the type of contrast agent [20], temperature and tissue environment as well as magnetic field strength [11,12]. R1 relaxivity of a paramagnetic contrast agent is higher at lower field strength [11]. R2 and R2* values should be theoretically identical in phantoms in each magnetic field [16]. In the phantom study described above, the authors attempted to compare the effects of contrast agent. For an accurate determination of the efficacy of Gd-complex (R1, R2 and R2*), only some of the relaxation times
M. Takahashi et al. / European Journal of Radiology 46 (2003) 45 /52
(T1, T2) that could be excellently fitted to the curve (r / 0.995) were reciprocally plotted against the concentrations of Gd at both 4 and 1.5 T. As a result, R1 and R2 relaxivity values were determined to be 2.95 and 4.82 (l ×/ s 1 ×/mmol 1) at 4 T and 3.89 and 4.67 (l ×/s1 ×/mmol 1) at 1.5 T, respectively. R1 at 4 T was lower ( :/25%) than R1 at 1.5 T, while the R2 at 4 T was almost that at 1.5 T (Table 3). Hence, we found that R1 relaxivity decreases as the magnetic field strength increases, while R2 relaxivity does not change as much, which is consistent with previous reports [16]. Unlike Gd-complex, R2 and R2* might be considerably changed depending upon the type of contrast agent (e.g. super paramagnetic iron oxide: SPIO), application root and/or tissues. This suggested that we should also consider the use of the MR contrast agent, though it is not clear whether this change is substantially effective in current clinical usage at higher magnetic field.
6. Safety consideration Theoretical calculations of the interaction of high magnetic fields with human subjects have been reviewed. To date, no hazardous physical or physiological phenomena have been shown. The mechanism considered included orientation of macromolecules and membranes, effects on nerve conduction, electrocardiograms and electroencephalograms, and blood flow. The most current clinical MR imagers at lower magnetic field (5/1.5 T) equip up to 25 mT/m. If higher magnetic fields are to be used to archive higher spatial resolution, the gradient strength must increase. In the combination of higher statistic magnetic field and gradients, strength may be an issue in some applications due to limitations in the current FDA guidelines for specific absorption rate (SAR). SAR is defined as follow: SAR
5. RF coil The application of higher magnetic field strengths to MR imaging (particular in whole body imaging) is more demanding because of the difficulty in building RF coils since the penetration of radio frequency into the tissue becomes harder [3,21]. It is necessary to understand the relationship between SNR and RF coil, since an incomplete RF coil may sacrifice the advantage in SNR at increased magnetic field strength. RF coil characteristics, especially a receive coil, significantly impact SNR. SNR increases with decreasing coil diameter. Thus, the coil sensitivity of the head coil is :/3-fold higher than that of the body coil. The surface coil with smaller diameter gains more sensitivity, whereas the SNR drops off very rapidly with increasing depth from the surface. To cover these difficulties, an array of surface coils must be developed. Reported by Wright et al. [22], another idea to increase coil sensitivity and further improve SNR is to reduce coil temperature, thus lowering its resistance and thermal noise voltages and increasing its Q , while keeping the sample at room temperature. The cryogenic SNR gain would be greatest for coil and sample configurations having QL/QU close to 1. Table 3 Comparison of relaxivities of gadolinium (Gd)-complex in aqueous solution, modified Ref. [14] Relaxivities
1.5 T
4T
R1 R2 R2*
3.89 4.67 4.55
2.95 4.82 4.67
* mmol 1 s 1 at room temperature.
49
sjEj2 2r
t
TR
NP NS
(5)
where s is conductivity, E is the electric field, r is tissue density, t is pulse duration and NP and NS are number of pulses and image slices, respectively. Since E is proportional to static magnetic field, SAR greatly increases at higher magnetic field, which may limit the application in number of slices, selection of flip angle, etc. Additionally, RF energy is absorbed more effectively at higher frequencies; RF absorption, as expressed by SAR, must be carefully monitored. This could be a major concern in any application at high field strength as Bottomley et al. previously suggested [21].
7. MR microscopy In using a higher magnetic field, the investigators expect images with higher spatial resolution to be more beneficial in research and clinical settings. Recently, transgenic and genetically engineered mouse models have been used increasingly and have led to important advances in many scientific communities. Thus, there has been an increased demand to image mice in vivo with a microscopic method. Whereas micro-computed tomography (m-CT) can potentially generate higherresolution images [23], MR is unique in that it is also able to provide detailed information on anatomy and function of soft tissues. Many excellent works have been reported in transgenic mice model genotype to phenotype [24,25], in which most of the cases were conducted with an experimental image scanner at a high magnetic field. Besides transgenic imaging, measurement of apparent diffusion coefficients of water molecules was
50
M. Takahashi et al. / European Journal of Radiology 46 (2003) 45 /52
performed in a single neuron [26]. More recently, a single axon in an excised lamprey spinal cord (Fig. 3) [27] was demonstrated and an apparent diffusion coefficient of the single axon was measured [28]. The techniques of very high resolution MR imaging have been developed largely in the past decade [26]. MR microscopy has been developed in trabecular bone imaging at the clinical magnetic field strength B/1.5 T. There are at least two reasons for applying MR microscopy to bone imaging. First, high signal contrast is raised between bones and surrounding tissue (bone marrow). Second, a smaller RF coil can be designed for the wrist or ankle. It is amenable to micromorphometry ex vivo and in vivo in laboratory animals and even in humans [29 /32]. The authors have provided new and interesting information on the use of quantitative in vivo MR microscopy and spectroscopy in conjunction with digital image processing to evaluate the epiphyseal and metaphyseal tissues of rabbits treated with dexamethasone at a 1.5 T clinical scanner [33]. One of the difficulties of imaging microarchitecture in vivo is achievement of sufficient resolution and SNR to resolve individual structures. The capability for direct visualization will have implications for acquiring sufficient image quality in vivo. As we discuss below, the higher the SNR, the longer the total scan time. However, we need to consider the interface between the desire for reasonable data acquisition times and adequate SNR, in particular in vivo imaging. Here, we discuss the imaging factors regulating spatial resolution in considering the protocol for a high resolution image at high magnetic field. SNR is the most limiting factor to increased spatial resolution in MR imaging, since scan time for a given SNR scales as the inverse sixth power of the linear voxel dimension [34]. SNR is primarily a function of voxel size, which is determined by the number of samples in the phase encoding and frequency encoding directions (in-plane resolution), the slice thickness (d) and the field of view in both directions, frequency (FOVf) and phase (FOVp) encodings. Hence, decreasing voxel size either by decreasing field of view or slice thickness or by increasing the matrix size, decreases SNR [35]. Therefore, any parameter determining voxel volume will also affect SNR, where SNR depends upon FOV quadratically in the same matrix size. The noise averaging process of repeated sampling has important implications on SNR. In theory [35], SNR increases as the square root of the number of samples collected. This holds true for the number of phase encodings (Np), the number of frequency encodings (Nf) as well as the number of signal acquisitions or number of excitations (NEX). The combining factors above can be summarized as follows:
pffiffiffiffiffiffiffiffiffiffiffi FOVf FOVp NEX d SNR pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi Nf × Np
(6)
Thus, it is true that changing the number of frequency and phase encodings affects both voxel size and signal averaging. The net effect is an inverse square root relationship between SNR and the product of phase and frequency encoding. By contrast, SNR scales as the square root of the number of excitations. Hence, doubling SNR requires quadrupling of NEX, which exacts a scan time penalty. Spatial resolution is typically expressed in terms of pixel size that is determined as the ratio of FOV divided by the number of phase or frequency encodings. Hence, we can decide spatial resolution in either of two ways: by manipulating the FOV or the matrix size. Changing slice thickness also affects resolution, albeit in a different way. Increasing slice thickness causes increased partial volume blurring. The effect of pixel size on image is demonstrated in Fig. 4. Fig. 4(A) shows a 3D projection image of the distal femoral epiphysis of a live rabbit obtained at a 1.5 T clinical scanner with 98/98/300 mm3 spatial resolution. The total scan time was :/20 min. Fig. 4(B) demonstrates a 3D projection of the small trabecular bone specimen from the proximal tibiae in rats. Locations have been matched to a cube highlighted in Fig. 1(A). The imaging was performed on a 9.4 T experimental machine with a total scan time of :/55 min to afford 39 mm isotropic voxel. Comparing these two images, the reduction in voxel size from 98 /98/300 mm3 in vivo conducted at 1.5 T (Fig. 4A) to 39 mm3 ex vivo at 9.4 T (Fig. 4B) presently entails an approximate 50-fold SNR penalty. This could not be recovered fully by the magnetic statistic field increase from 1.5 to 9.4 T when the RF coil insert was the same and had the same sensitivity. This obviously means that the RF power is dissipated beneficially by increasing Q in smaller RF coils for SNR gain. Basically, RF coil sensitivity (Q dumpling) is reciprocal in proportion to its diameter [21]. According to the discussion above, the achievable benefit in SNR might be a factor of 4 after optimization of imaging parameters when the magnetic field strength is changed from 1.5 to 3 or 4 T in clinical situations, where the RF coils have almost the same sensitivity between the fields. This contributes to image resolution increasing at a factor of up to 4 in the same scan time, e.g. 128/128 to 256 /256 matrix. In this case, the slice thickness should be kept constant. When we prefer to save scan time, we can reduce the number of excitations by half, since SNR is proportional to the square of the total scan time. If we want further advances, development of RF coil, image acquisition, restoration and processing techniques should be involved. We estimated the benefits for MR imaging when the current MR scanner with magnetic field strength of
M. Takahashi et al. / European Journal of Radiology 46 (2003) 45 /52
51
Fig. 4. (A) Three dimensional (3D) projection image of the distal femoral epiphysis of a live rabbit covering the volume analyzed. Projection direction is inferior to superior at an angle of 308 relative to the femoral anatomic axis. (B) 3D projection image of the small trabecular bone specimen from the proximal tibiae in rats. Locations have been matched to a cube highlighted in (A). (A) and (B) were obtained on 1.5 T clinical and 9.4 T experimental machines in the total scan time of :/20 and 55 min, respectively.
B/1.5 T is replaced with a higher field up to 3 or 4 T. A number of practical implications in the imaging of biological tissues at higher field strength must be considered. The gain in SNR from the higher magnetic field strength may be substantially offset by prolonged T1 relaxation times, thus optimization of the imaging parameters is important. Although it was not discussed in this review, higher magnetic field must produce better frequency resolution of near degenerated resonances that are not resolvable at lower field in magnetic resonance spectroscopy. Another motivation for high magnetic field is the ability to use other nuclei (e.g. 23Na, 39 K), rather than protons, for which sensitivities are not sufficient to be observed at lower magnetic field. Recently, whole-body MR scanners with much higher magnetic fields (8 /10 T) have been developed and some have been applied in human studies. In conclusion, MR imaging at a higher magnetic field strength (]/3 Tesla) will be opening a new arena. The appropriate optimizations, such as image acquisition, development in RF coil design and image processing algorithms with adequate safety profiles, would expand the applications.
Acknowledgements We would like to thank Dr Shigeru Kiryu at Beth Israel Deaconess Medical Center/Harvard Medical School for his assistance in article preparation.
References [1] Thulborn KR. Clinical rationale for very-high-field (3.0 Tesla) functional magnetic resonance imaging. Top Magn Reson Imaging 1999;10:37 /50. [2] Dougherty L, Connick TJ, Mizsei G. Cardiac imaging at 4.0 Tesla. Magn Reson Med 2001;45:176 /8.
[3] Noeske R, Seifert F, Rhein KH, Rinneberg H. Human cardiac imaging at 3 T using phased array coils. Magn Reson Med 2000;44:978 /82. [4] Bernstein MA, Huston J, III, Lin C, Gibbs GF, Felmlee JP. Highresolution intracranial and cervical MRA at 3.0 T: technical considerations and initial experience. Magn Reson Med 2001;46:955 /62. [5] Boos M, Lentschig M, Scheffler K, Bongartz GM, Steinbrich W. Contrast-enhanced magnetic resonance angiography of peripheral vessels. Different contrast agent applications and sequence strategies: a review. Invest Radiol 1998;33:538 /46. [6] Hatabu H, Gaa J, Kim D, Li W, Prasad PV, Edelman RR. Pulmonary perfusion and angiography: evaluation with breathhold enhanced three-dimensional fast imaging steady-state precession MR imaging with short TR and TE. Am J Roentgenol 1996;167:653 /5. [7] Hany TF, Schmidt M, Hilfiker PR, Steiner P, Bachmann U, Debatin JF. Optimization of contrast dosage for gadoliniumenhanced 3D MRA of the pulmonary and renal arteries. Magn Reson Imaging 1998;16:901 /6. [8] Duewell SH, Wolff SD, Wen H, Balaban RS, Jezzard P. MR imaging contrast in human brain tissue: assessment and optimization at 4 T. Radiology 1996;199:780 /6. [9] Jezzard P, Duewell S, Balaban RS. MR relaxation times in human brain: measurement at 4 T. Radiology 1996;199:773 /9. [10] Duewell SH, Ceckler TL, Ong K, et al. Musculoskeletal MR imaging at 4 T and at 1.5 T: comparison of relaxation times and image contrast. Radiology 1995;196:551 /5. [11] Lauffer RB. Magnetic resonance contrast media: principles and progress. Magn Reson Q 1990;6:65 /84. [12] Vander Elst L, Laurent S, Muller RN. Multinuclear magnetic resonance characterization of paramagnetic contrast agents. The manifold effects of concentration and counterions. Invest Radiol 1998;33:828 /34. [13] Gadian DG. In: Gadian DG, editor. Nuclear magnetic resonance and its applications to living systems. New York: Oxford University Press, 1982. [14] Uematsu H, Dougherty L, Takahashi M, et al. A direct comparison of signal behavior between 4.0 Tesla and 1.5 Tesla: a phantom study. Eur J Radiol 2002;45:154 /159. [15] Keiper MD, Grossman RI, Hirsch JA, et al. MR identification of white matter abnormalities in multiple sclerosis: a comparison between 1.5 T and 4 T. Am J Neuroradiol 1998;19:1489 /93. [16] Fernandez-Seara MA, Wehrli FW. Postprocessing technique to correct for background gradients in image-based R*(2) measurements. Magn Reson Med 2000;44:358 /66.
52
M. Takahashi et al. / European Journal of Radiology 46 (2003) 45 /52
[17] Ludeke KM, Roschmann P, Tischler R. Susceptibility artefacts in NMR imaging. Magn Reson Imaging 1985;3:329 /43. [18] Takahashi M, Wehrli FW, Hwang SN, Wehrli SL. Relationship between cancellous bone induced magnetic field and ultrastructure in a rat ovariectomy model. Magn Reson Imaging 2000;18:33 /9. [19] Uematsu H, Dougherty L, Takahashi M, et al. Pulmonary MR angiography with contrast agent at 4 Tesla: a preliminary result. Magn Reson Med 2001;46:1028 /30. [20] Takahashi M, Tsutsui H, Murayama C, Miyazawa T, FritzZieroth B. Neurotoxicity of gadolinium contrast agents for magnetic resonance imaging in rats with osmotically disrupted blood /brain barrier. Magn Reson Imaging 1996;14:619 /23. [21] Bottomley PA, Andrew ER. RF magnetic field penetration, phase shift and power dissipation in biological tissue: implications for NMR imaging. Phys Med Biol 1978;23:630 /43. [22] Wright AC, Song HK, Wehrli FW. In vivo MR micro imaging with conventional radiofrequency coils cooled to 77 K. Magn Reson Med 2000;43:163 /9. [23] Hildebrand T, Laib A, Muller R, Dequeker J, Ruegsegger P. Direct three-dimensional morphometric analysis of human cancellous bone: microstructural data from spine, femur, iliac crest, and calcaneus. J Bone Miner Res 1999;14:1167 /74. [24] Johnson GA, Cofer GP, Gewalt SL, Hedlund LW. Morphologic phenotyping with MR microscopy: the visible mouse. Radiology 2002;222:789 /93. [25] Sze RW, Chan CB, Dardzinski BJ, et al. Three-dimensional MR microscopy of a transgenic mouse model of dilated cardiomyopathy. Pediatr Radiol 2001;31:55 /61. [26] Aguayo JB, Blackband SJ, Schoeniger J, Mattingly MA, Hintermann M. Nuclear magnetic resonance imaging of a single cell. Nature 1986;322(6075):190 /1.
[27] Wright AC, Wehrli SL, Zhang G, et al. Visualization of individual axons in excised lamprey spinal cord by magnetic resonance microscopy. J Neurosci Methods 2002;114:9 /15. [28] Takahashi M, Hackney DB, Zhang G, et al. Magnetic resonance microimaging of intra-axonal water diffusion in live excised lamprey spinal cord. Proc Natl Acad Sci USA 2002;99:16192 /6. [29] Ma J, Wehrli FW, Song HK. Fast 3D large-angle spinecho imaging (3D FLASE). Magn Reson Med 1996;35:903 / 10. [30] Majumdar S, Newitt D, Mathur A, et al. Magnetic resonance imaging of trabecular bone structure in the distal radius: relationship with X-ray tomographic microscopy and biomechanics. Osteoporos Int 1996;6:376 /85. [31] Gordon CL, Webber CE, Christoforou N, Nahmias C. In vivo assessment of trabecular bone structure at the distal radius from high-resolution magnetic resonance images. Med Phys 1997;24:585 /93. [32] Takahashi M, Wehrli FW, Wehrli SL, et al. Effect of prostaglandin and bisphosphonate on cancellous bone volume and structure in the ovariectomized rat studied by quantitative 3D NMR microscopy. J Bone Miner Res 1999;14:680 /9. [33] Takahashi M, Wehrli FW, Hilaire L, Zemel BS, Hwang SN. In vivo NMR microscopy allows short-term serial assessment of multiple skeletal implications of corticosteroid exposure. Proc Natl Acad Sci USA 2002;99:4574 /9. [34] Callaghan PT. Principles of nuclear magnetic resonance microscopy. Oxford: Clarendon Press, 1991. [35] Wehrli FW, Ford JC, Chung HW, et al. Potential role of nuclear magnetic resonance for the evaluation of trabecular bone quality. Calcif Tissue Int 1993;53(Suppl 1):S162 /9.