Brain Revelations: Imaging at 3 T

Matilde Inglese, MD, PhD

High field MRI scanners with a static magnetic field (B ₀ ) of three Tesla (3 T) are now being delivered with full FDA clearance. Indeed, instruments with fields up to 4 T have been given FDA “Non-Significant Risk” clearance, and to date no center has recorded an incident or significant safety risk, even at 7 T in more than 300 normal human subjects at one center. ¹ Consequently, 3 T and higher MR systems are becoming ubiquitous in clinical and basic neuroscience and will assume much of the advanced clinical and research applications role now played by the 1.5-T imagers. Several advanced neuro MR applications, specifically, angiography, diffusion, perfusion, functional imaging (fMRI), and spectroscopy (MRS), would benefit significantly from higher B ₀ s, especially those that are signal-to-noise-ratio (SNR) limited. Due to the high cost and complexity of these systems, it is important to understand whether the use of high-field magnets can be beneficial for the neurosciences. MRS, for example, would benefit from increased spatial and temporal resolution, which could lead to improved metabolite quantification, one of the current major challenges in clinical MRS. Therefore, the purpose of this article is to describe the advantages, disadvantages, and implications of MRS at higher B ₀ s.

MR SPECTROSCOPY OF THE BRAIN

Clinical applications of MRS in the brain are numerous and include characterization of cerebral neoplasms, monitoring therapy-induced changesradiation necrosis versus recurrent tumorepilepsy, infection, demyelinating diseases, trauma, pediatric metabolic disorders, and neurodegenerative processes such as Alzheimer’s disease. ^2,3

Recent publications demonstrate the utility of high field MRS. ^4-6 However, while there may be several obvious advantages in switching to high field spectroscopy, there are also inherent limitations.

The advantages include:

1. Higher Signal to Noise . Due to the excellent spatial resolution of MRI, it is quite easy to appreciate and compare signal and contrast to noise on conventional T1- or T2-weighted images and determine, by eye, if there is appreciable improvement by going to higher B ₀ . In contrast, resolution and SNR are comparatively much lower in MRS. Furthermore, each MRS voxel also has several metabolite signals within it. Consequently, SNR is less visually apparent and usually has to be computed. This is done, for example, by taking the peak amplitude of a metabolite’s signal (eg, the choline) and comparing it with the peak to peak amplitude of the baseline noise (Figure 1). Hence, subtle differences in SNR between different field strengths may not be readily appreciated by eye.

Figure 1. Spectrum from an 81-year-old male with a high-grade glioma. The contrast-enhanced axial T1-weighted image (600/14/1) demonstrates an enhancing mass in the left temporal region. Proton MRS using PRESS (1500/144 [TR/TE]) demonstrates elevated Cho and decreased NAA in a typical tumor spectrum at 1.5 T. SNR can be measured by dividing the amplitude of the metabolite (eg, Cho signal amplitude) by the square root of the amplitude of the baseline noise (noise signal amplitude).

For in vivo spectroscopy, a twofold gain in SNR has been theoretically predicted, for a doubling in field strength, eg, from 1.5 to 3 T. ^7,8 However, investigators were initially surprised and disappointed to find these improvements ranging only from 20% to 46%, as shown in Figures 2, 3, and 4 (below). ^9,10 Indeed, in a realistic experiment, the theoretical doubling of SNR by a twofold increase in B ₀ cannot be achieved due to the use of TRs (times to repeat) of the order of the T1 decay times and TEs (times to echo) of the order of the metabolite T2 decay times and not 0. A further obstacle for clinical spectroscopy at high field is the lack of high-sensitivity, high-homogeneity volume coils. Nevertheless, this area has undergone major developments in recent years, leading to “second-generation” head and body coils for 3 and 4 T imagers. ¹¹

Figure 2. Axial T1-weighted MR images from the section richest in anatomic features in a normal volunteer. Both are superimposed with the outline of the 9 x 9 cm volume of interest (VOIs) and the real part of the corresponding 9 x 9 1H-spectra matrices. The horizontal (1.8-3.5 ppm) scale is common to both fields. A. Examination at 1.5 T. B. Examination at 3 T.

Figure 3. Five spectra from the corresponding dashed boxes in figure 2 scaled to the same maximum NAA peak height. Note: an improved Cr-Cho spectral resolution, better SNR, and flatter baseline between 2 and 3 ppm at 3 T versus 1.5 T.

Although proton ( ¹ H) MRS offers the highest SNR and can be performed with the same hardware used in MRI, MRS research using less sensitive low gyromagnetic (g) ratio nuclei such as phosphorus ( ³¹ P) and carbon ( ¹³ C) has also contributed to the current understanding of brain metabolism. For instance, by attaining superior SNR and spatial resolution at 4 T, a ³¹ P MRS study has demonstrated that human cerebral adenosine triphosphate (ATP) concentrations are lower in gray matter than in white matter. The increasing availability of high-field magnets will undoubtedly encourage an expanded role for low (g) nuclei, mainly because of the sensitivity gains possible with increased B ₀ .

2. Higher Spectral Resolution . Spectral resolution is essentially determined by the starting and ending points of a peak. The ability to resolve different adjacent metabolite peaks not only allows accurate identification of each metabolite, but in both clinical and research settings, it affects the quantification of each metabolite, and hence, the differentiation of normal from “pathological” peaks.

Figure 4. Left: Two axial slices from two patients with multiple sclerosis, superimposed with the MRS volume-of-interest (VOI) and the real part of the 1H spectra matrices from them. Top, a: The 8 x 10 cm2 VOI and 8 x 10 spectra at 1.5 T (54 min acquisition). Bottom, b: the 7 x 8 cm2 VOI and 7 x 8 spectra array at 3 T (27 min acquisition). Right: Columns of six spectra taken from the dashed box on the corresponding images and spectra matrices on the left at each, 1.5 and 3 T fields. Note: the correspondence of the spectra, via metabolite level deficits, with both the gross anatomical features in this slice (ventricles) and with the periventricular lesions in both subjects. Note that the SNR at 3 T is similar to that at 1.5 T even though the measurement was half as long. (Click the image for a larger version.)

At higher field strength, there is improvement in resolution between metabolites. ^9,10,12 This is reflected by improved baseline separation of choline (Cho) and creatine (Cr), 0.2 parts per million (ppm) apart at 3 T versus 1.5 T (see Figures 3 and 4). The overall result is improvement in spectral resolution that means better ability in identifying and quantifying individual peaks. This will also improve the ability to study the group of glutamate and glutamine peaks at shorter TE between 2.05 and 2.5 ppm. It is thought that at 3 T, there will be less spectral overlap of these amino acids and glucose as well. ^4,10 This has important clinical implication in acute cerebral ischemia, multiple sclerosis, Alzheimer’s disease, and hepatic encephalopathy. In the treatment of epilepsy with antiepileptic drugs such as vigabatrin, levels of gamma-aminobutyric acid (GABA) can be monitored more readily at high field strength than at 1.5 T because of improved spectral resolution. ⁴

3. Higher Spatial and Temporal Resolution . Clinical MRS can be performed with commercially available single-voxel localization, or 2D and 3D chemical shift imaging (CSI) multi-voxel sequences (the latter two generally referred to as spectroscopic imaging [SI]). Reducing the MRS voxel size will substantially lower the SNR. At 3 T and higher, smaller and smaller voxels may be possible at the expense of the intrinsic field-dependent SNR gain. Voxels below 1 cm ³ (a limiting size at 1.5 T) can be obtained with good SNR in similar acquisition times. ⁶ Reduced voxel size is a significant advantage in studies of small brain structures, eg, the hippocampus, basal ganglia, MS plaques, as shown in Figure 4, and small neoplasms. ¹³ This has potential application in patients with temporal lobe epilepsy and, in part, in patients with Alzheimer’s disease for differentiating a normal from “normal-appearing” hippocampus (on conventional T2-weighted MRI), which may already have early neuronal losses detectable only by a decrease in N-acetylaspartate (NAA) observable with ¹ H-MRS. ⁴

One restriction to performing MRS routinely in clinical practice has been long measurement times (up to almost a hour). Reduced acquisition times at higher B ₀ s will no doubt be beneficial to the clinical radiologist, MR technician, hospital administrator, and, not least, the patient. This is demonstrated in Figure 4, where similar SNR to that at 1.5 T is obtained at 3 T at half the measurement time. It means that not only will total examination times be reduced, but multiple MRS protocols and locations can be acquired at different echo times, which may find clinical utility in stroke, infections, and tumor characterization. ¹⁴ Furthermore, larger brain volumes can be studied in similar measurement times as current single-voxel protocols with multi-voxel 2D or 3D SI sequences. Shorter acquisition times will also facilitate dynamic or kinetic MRS, eg, to follow drug uptake and metabolism in real time. Indeed, when using substrates enriched with MR-active stable isotope (eg, ¹³ C or ¹⁵ N), MRS can be used to monitor the flow of the isotope into different metabolic intermediates. With high field ¹³ C MRS, for example, SNR is sufficient to allow localized detection of label incorporation into multiple molecules, including GABA, glycogen, aspartate, glutathione, glutamate, glutamine, and NAA. To improve SNR further, ¹³ C and ¹⁵ N signals are often detected indirectly through the scalar coupling that exists between these nuclei and attached ¹ H. Such heteronuclear techniques allow isotopically labeled molecules to be detected with the high sensitivity of ¹ H-MRS.

4. Improved Baseline and Quantification . Higher field strength will also lead to a flatter baseline (Figures 2, 3, 4). This is particularly evident in the 2.0 to 3.0 ppm region, between the NAA and Cr peaks (Figure 3). At the intermediate TE of 135 ms, there is a general x 2 decrease in T2 decay time of all the metabolites at the higher, 3 T, field strength. This leads to general attenuation of all peaks but, in particular, the total disappearance of the broader peaks, eg, glutamine, glutamate, and lipids. The flatter baseline makes for more reliable estimation of peak area and hence more precise quantification. ¹⁵ Quantification of metabolites has always been a problematic issue with spectroscopy. At higher field strength, the improvement in signal to noise, flatter baseline, and better spectral resolution (Figures 2-4) among peaks should all help to improve quantification. ¹⁶ Improvements in shimming at higher field will also enhance the lineshapes of metabolite peaks. Increased SNR and spectral resolution have been shown to increase metabolite level quantification by 39% at 4 T compared with 1.5 T. ¹³

DRAWBACKS

Higher magnet field strength, however, entails certain limitations.

1. Susceptibility and Eddy Current Artifacts . Higher B ₀ s will incur increased susceptibility from paramagnetic substances and blood products, and at air/tissue and tissue/bone interfaces, eg, near the air-filled sinuses and near the skull base. Therefore, obtaining good spectra from lesions near the skull base or close to the calvarium will be more difficult at high fields. These problems can be reduced by improved automatic local shimming routines using second and higher order shim-coils. ^6,13

Eddy current artifacts may also be more apparent at high field due to increased torque and thrust on the gradient coil. ¹³ This may lead to spectral distortion ^17,18 and loss in SNR. However, ongoing improvements in hardware design will no doubt overcome these problems in the near future.

2. Larger Chemical Shift Misregistration Errors for 2D and 3D SI . Due to the clinical need for greater coverage of the brain, all vendors are now offering 2D and 3D SI packages. The selective pulses used for volume localization will progressively incur (with both field and chemical-shift increase) worse volume misregistration at higher B ₀ s. ^4,19 This ultimately means that the volume of spectroscopic information measured will not be the same as that displayed on the MRI. Nevertheless, software improvements are being made by the major manufacturers to reduce this problem.

3. SAR/RF Penetration/Safety Issues . The higher radio-frequency power specific-absorption rate (SAR) at higher magnetic field strength, which indicates energy absorption in the subject, can be a restriction in neuro-spectroscopy. This is because more power is needed to affect a 90° pulse, at higher fields, eg, twice as much power at 7 T as at 4 T. ²⁰ Therefore, longer TRs or less selective pulses may have to be used at the expense of longer protocols or poorer localization, respectively, to adhere to the FDA limit of 3.2 watt/kg weight (SAR) in the head.

4. Quantification . The observed metabolite ratios depend on their intrinsic individual T1 and T2 relaxation times. While the former are similar at 1.5 and 3 T, the T2 relaxation times are approximately two times shorter at 3 T. ²¹ The normal metabolite ratios at 1.5 T may, therefore, be different at 3 T. ²² Consequently, it is important to gather normative data at 3 T and compare it with previously acquired 1.5 T data. Furthermore, due to the overall T2 reduction with B ₀ increase, long TE MRS studies suffer commensurately greater SNR losses at 3 T than at 1.5 T. ^10,22 Therefore, for long, TE spectroscopy (greater than 144 ms), there is minimal gain in going to high field. ⁹ Consequently, it may no longer be useful to perform MRS at long echo times (270-288 ms). ⁴

5. Costs . The main challenge to widespread installation of high-field systems may ultimately be the overall cost. The state of the art 3 T scanners cost at least $1 million more than current 1.5 T imagers, and the site placement costs, due to shielding and weight requirements, are also higher. Despite the obvious advantages of high-field imaging, it may be difficult to justify such expenditures in major research institutions, let alone in private practice radiology. However, once the clinical utility of 3 T imaging is firmly established, to the extent that it becomes the “standard of care,” there may potentially be greater competition and pressure between the major vendors to make their product more affordable.

To date, there have been some efforts to combine MR imaging and MR spectroscopic techniques in order to distinguish surgical from nonsurgical lesions in the brain. ²³ Although conventional MRI provides exquisite anatomic delineation of intracranial pathology, occasionally it has led radiologists or surgeons to misdiagnose lesions that should have been treated conservatively rather than surgically. For example, there is overlap in the MRI appearance of many ring-enhancing lesions such as gliomas, metastases, inflammatory lesions, demyelinating lesions, subacute ischemia, and some AIDS-related lesions. It goes without saying that neurosurgical intervention carries substantial morbidity, mortality, financial, and potential emotional cost to the patient and family. Consequently, this has tremendous impact not only for the patient and physician but for the cost and efficiency of health care delivery.

We have recently shown that combined information from spectroscopic and imaging data might be useful in confirming preoperatively a diagnosis of a nonsurgical lesion such as a tumefacting demyelinating lesion (TDL) or stroke. ²³ This can help make a preoperative diagnosis that not only provides the neurosurgeon with a confident treatment plan for the patient but allays considerable patient anxiety. On the other hand, supplementing anatomic information with metabolic and physiologic information from ¹ H-MRS and perfusion imaging increases sensitivity and specificity in neuro-diagnosis, avoiding unnecessary surgery.

In conclusion, the use of high-field MR techniques has several features that are highly desirable for the neuroscience community. Given the wide range of neurological MR applications at 3 T, it is essential to understand the strength and weaknesses of each type of system, and use them accordingly. This is particularly the case for MRS. There are many factors that will affect the quality of spectra at different B ₀ s, and, therefore, it is important to keep in mind the current limitations and challenges.

Matilde Inglese, MD, PhD, is assistant professor, Department of Radiology, New York University Medical Center, New York.

References:

1. Vaughan JT. Technological challenges and solutions for high field in vivo magnetic resonance. In: Proceedings of the 10th Scientific Meeting of the International Society for Magnetic Resonance in Medicine, May 18-24, 2002, Honolulu. 2002.
2. Danielson ER, Ross BD. Magnetic Resonance Spectroscopy of Neurological Diseases. New York: Marcel Dekker Inc; 1998.
3. Ross B, Michaelis T. Clinical applications of magnetic resonance spectroscopy. Magn Reson Q. 1994;10(4):191-247.
4. Hetherington HP, Pan JW, Chu WJ, et al. Biological and clinical MRS at ultra-high field. NMR Biomed. 1997;10:360-371.
5. Hattori N, Abe K, Sakoda S, Sawada T. Proton MR spectroscopic study at 3 Tesla on glutamate/glutamine in Alzheimer’s disease. Neuroreport. 2002;13(1):183-186.
6. Li BS, Regal J, Gonen O. SNR versus resolution in 3D 1H MRS of the human brain at high magnetic fields. Magn Reson Med. 2001;46:1049-1053.
7. Hoult DI, Lauterbur PC. The sensitivity of the zeugmatographic experiment involving human samples. J Magn Reson. 1979;34:425-433.
8. Edelstein WA, Glover GH, Hardy CJ, Redington RW. The intrinsic signal-to-noise ratio in NMR imaging. Magn Reson Med. 1986;3:604-618.
9. Barker PB, Hearshen DO, Boska MD. Single-voxel proton MRS of the human brain at 1.5T and 3.0T. Magn Reson Med. 2001;45:765-769.
10. Gonen O, Gruber S, Li BS, et al. Multivoxel 3D proton spectroscopy in the brain at 1.5 versus 3.0 T: signal-to-noise ratio and resolution comparison. AJNR Am J Neuroradiol. 2001;22:1727-1731.
11. Vaughan JT, Hetherington HP, Otu JO, et al. High frequency volume coils for clinical NMR imaging and spectroscopy. Magn Reson Med. 1994;32(2):206-218.
12. Gruetter R, Weisdorf SA, Rajanayagan V, et al. Resolution improvements in in vivo 1H NMR spectra with increased magnetic field strength. J Magn Reson. 1998;135(1):260-264.
13. Bartha R, Drost DJ, Menon RS, Williamson PC. Comparison of the quantification precision of human short echo time (1)H spectroscopy at 1.5 and 4.0 Tesla. Magn Reson Med. 2000;44(2):185-192.
14. Law M, Cha S, Knopp EA, et al. Glioma grading with multi-slice, multi-voxel, multi-TE spectroscopic MRI and multi-slice perfusion MRI. In: Proceedings of the International Scientific Meeting of the International Society for Magnetic Resonance in Medicine, May 18-24, 2002, Honolulu. 2002:308.
15. Kreis R, Ernst T, Ross B. Absolute concentrations of water and metabolites in the human brain. II. Metabolic concentrations. J Magn Reson Series B. 1993;102:9-19.
16. Pfeuffer J, Tkac I, Provencher SW, Gruetter R. Toward an in vivo neurochemical profile: quantification of 18 metabolites in short-echo-time (1)H NMR spectra of the rat brain. J Magn Reson. 1999;141(1):104-120.
17. Klose U. In vivo proton spectroscopy in presence of eddy currents. Magn Reson Med. 1990;14(1):26-30.
18. de Graaf AA, van Dijk JE, Bovee WM. QUALITY: quantification improvement by converting lineshapes to the Lorentzian type. Magn Reson Med. 1990;13(3):343-357.
19. Brown TR. Practical applications of chemical shift imaging. NMR Biomed. 1992;5:238-243.
20. Vaughan JT, Garwood M, Collins CM, et al. 7T vs 4T: RF power, homogeneity, and signal-to-noise comparison in head images. Magn Reson Med. 2001;46(1):24-30.
21. Mlynarik V, Gruber S, Moser E. Proton T (1) and T (2) relaxation times of human brain metabolites at 3 Tesla. NMR Biomed. 2001;14(5):325-331.
22. Pan JW. Clinical and physiological studies with high field human spectroscopy and spectroscopic imaging. In: Proceedings of the 10th Scientific Meeting of the International Society for Magnetic Resonance in Medicine, May 18-24, 2002, Honolulu. 2002.
23. Law M, Hamburger N, Johnson G, et al. Differentiating surgical from non-surgical lesions using perfusion MR imaging and proton MR spectroscopic imaging. Technol Cancer Res Treat. 2004;3:557-65.