Synopsis

This talk will familiarize MRI engineers and clinicians with some of the principal limitations on MRI scanner hardware performance, along with methods for characterizing and mitigating these imperfections. We will also discuss issues arising from the interaction of the scanner's static and radiofrequency fields with the human body.

Highlights

• Better understand how each function subsystem of the MRI scanner works and how its performance is limited and non-ideal in practice
• Appreciate which artifacts can be corrected in post-processing (e.g., mild gradient nonlinearity) and the limitations of these software corrections
  
• Understand how the body interacts with both background and RF fields, how this impacts image quality, and what methods are available to address these interactions
• Review conventional techniques for measuring scanner imperfections such as B1 and B0 mapping

Target audience

Outcome/Objectives

Introduction

MRI scanner components have interdependent requirements and must work in harmony. It is therefore becoming rare to design magnet, gradient and RF subsystems separately and expect the system to work well when bolted together. The interaction of the gradient coil with the magnet is an important example, on which there is little in the literature. Also, it is important to match the gradient linearity, RF coverage region and homogeneous magnet region to prevent detection of unencoded spins from folding into the FOV will be discussed.

The system imperfections discussed in this talk are briefly outlined below. We divide the imperfections into two categories: (1) those arising from limitations on the scanner hardware itself, and (2) issues arising from the interaction of scanner electromagnetic fields and the human body.

Scanner hardware limitations

• Magnet homogeneity: Modern clinical scanners typically have very good static field homogeneity over the volume of the head (<1ppm). However, while image artifacts caused by magnet inhomogeneity are uncommon on modern scanners, they do occur. For example, fat saturation performance can be compromised in large subjects toward the edge of the bore, since the field can deviate by 3ppm in this region. Since the fat-water separation is approximately 3ppm, magnet inhomogeneity could thus lead to incomplete fat suppression (or other artifacts) on the surface of the abdomen for supine subjects.

• Magnet drift: While a magnet at rest may be quite stable, users should be alert for global drift of the magnet field caused by disturbances such as prolonged gradient slewing. While modern gradient coils are actively shielded, some residual eddy currents are induced in conducting structures such as the magnet’s vacuum shield and shim iron (where present). Small changes in the temperature of these structures leads to magnet drift. This can cause a variety of artifacts. For instance, in EPI time series acquisitions, this drift causes the image to slowly drift in the phase encode direction. Magnet drift can be corrected either during image reconstruction by correcting each line of k-space or by dynamically updating the RF carrier frequency during the acquisition. While both remedies are generally effective, adjusting the carrier frequency is more reliable since it ensures that the RF pulse performance is not compromised by the drift (e.g., fat and water suppression pulses remain on-resonance).

• Gradient nonlinearity: The spatial profile of the gradient field varies significantly between vendors and specific gradient coil models (e.g., whole-body versus head-only coils), but the field is never perfectly linear across the entire bore. It is especially important to consider gradient nonlinearity effects in multi-site imaging studies that use multiple gradient coils or vendor platforms [1], where image warping could be a confound. For mild gradient nonlinearity, image warping can usually be corrected in post-processing with minimal loss of resolution. Toward the edge of the bore, gradient nonlinearity grows more severe; field roll-off creates an encoding singularity that is difficult to correct in post-processing. This leads to warping of the arms in whole-body imaging, which in turn causes problems for positron emission tomography (PET) attenuation correction in combined PET-MRI acquisitions. Gradient nonlinearity has also been implicated as a confound in tumor malignancy diagnosis; apparent diffusion coefficient measurements have been shown to depend on patient positioning inside the scanner FOV [2]. For head gradient inserts, severe gradient nonlinearity in the neck and shoulder region can cause signal to wrap into brain images if the RF transmit and receive coils are sensitive in the neck/shoulders.

• Acoustic resonances: Every gradient coil has acoustic resonances that depend on its specific geometry and the way it is mounted inside the bore [3]. In addition to being an important consideration for subject safety and comfort, acoustic resonances can lead to image artifacts and, in extreme cases, damage to the gradient coil or even a magnet quench. For example, acoustic resonances are well-known to cause ghosting in EPI sequences. For these reasons, to ensure safety and good image quality, manufacturers typically block out acoustic resonances on the console user interface. This sometimes constrains imaging performance, particularly for slew-rate limited sequences such as EPI.

• Eddy currents: Despite advances in active shielding, modern gradient coils still generate significant eddy currents in other conducting structures in the bore. The eddy currents generate nuisance magnetic fields that cause a host of image artifacts, for example translation, contraction, and sharing for eddy currents generated by slice, phase encode, and readout gradients, respectively [4]. The transient fields generated by the eddy currents typically include several spatial harmonics in addition to the linear fields being switched, including B0 offsets and higher-order spherical harmonic fields. Typically switched coils generate modes that share similar geometry and symmetry (even vs. odd). The eddy currents and fields decay with time constants that depends on the material and temperature; typically eddy currents in cold structures of the cryostat decay slowly (with time constants in the 100s of ms or longer). Eddy currents cause the magnetic field generated in the bore to “lag” behind changes in current flowing in the coil. Thus, modern scanner gradients use pre-emphasis overshoot or undershoot on the gradient pulse waveforms to compensate for the eddy currents and faithfully apply the desired field evolution (and k-space trajectory). However, it is difficult to completely eliminate the eddy currents, particularly for the large gradient lobes used in diffusion imaging [5]. Eddy current mapping: One method for assessing the time constant of eddy currents is to switch one gradient coil field on for a long pulse, then switch it off immediately before RF excitation in a conventional gradient echo sequence. If the delay between the test gradient pulse and the RF excitation is varied, the spatial distribution of the eddy current-induced fields can be mapped [6]. The fields can then be decomposed into spatial and temporal components. Another way to measure the full spatiotemporal field dynamics is using an array of ~8-16 NMR field probes distributed on a spherical or cylindrical surface [7,8]. Field probes have been used to fully characterize all gradient and shim channels by measuring each coil’s impulse response [9]. This approach captures all eddy current behavior as well as coupling between different gradient and shim coils in the bore. Field probe calibration to be discussed in detail later in this session.

Interaction of scanner fields with the body

In addition to the above imperfections of the MRI scanner hardware itself, the introduction of the patient’s body to the bore causes important changes in both the static and radiofrequency fields generated by the scanner. Because these patient-specific perturbations lead to a large variety of image artifacts, it is important to understand how to measure these imperfections and what approaches are available to remedy them.

B0 INHOMOGENEITY

• Localized, sharply-varying ΔB₀ fields in the brain are caused by tissue susceptibility interfaces in the sinuses and oral cavity. These ΔB₀ fields cause a host of image artifacts in the prefrontal cortex, medial temporal cortex, and brainstem, all of which are critically important targets for both clinical and scientific neuroimaging. For example, residual ΔB₀ fields make it difficult to measure microscopic T₂^* in a voxel due to the overlaying macroscopic B₀ variations across the voxel. ΔB₀ also causes signal voids in gradient-echo sequences, particularly for thick slices shows geometric distortion and blurring near ΔB₀ hotspots in echo train sequences such as echo planar imaging (EPI), which is frequently used for functional and diffusion MRI [10]. Poor B₀ shimming also leads to metabolite line broadening and degraded water and lipid suppression performance in spectroscopic imaging [11,12]. Furthermore, ΔB₀ has undesired effects on RF excitation, inversion, and VERSE pulses [13], and it complicates the design of pulses for parallel transmission applications [14,15]. Finally, spatiotemporal ΔB₀ fluctuations caused by physiological processes such as respiration lead to ghosting in structural images and time-series instability in functional scans, especially in the deep brain and spine [16]. All of these problems are especially severe for ultra-high field MRI (UHF, 7 Tesla and above) [17].

• B0 field mapping: Spatial maps of ΔB0 are usually acquired with a multi-echo gradient echo sequence [18]. The phase images for each echo time are unwrapped and a line is fit to the phase evolution in each voxel to estimate the frequency. Using at least 3 echo times is recommended to improve the robustness of the frequency estimate and hence the contrast-to-noise ratio of the resulting ΔB0 maps. PRELUDE [19] is often held as the “gold standard” phase unwrapping method for accuracy but it suffers from long processing times, particularly for later echo times with multiple wraps. Fast new methods have recently been proposed for unwrapping in the spatial [20] and temporal domain [21].

• Remedies for ΔB₀-induced EPI distortion: Several software tools are available for correcting EPI geometric distortion in post-processing using knowledge of the B₀ field map [22],[23]. However, these methods have limited efficacy when several voxel pile-up occurs. Another common way to reduce distortion is to use parallel imaging methods [24] to reduce the effective EPI echo spacing [25]. But for high acceleration factors, image quality often suffers from reduced SNR and residual aliasing artifacts. Finally, powerful head gradient insert coils with enhanced slew rates enable shorter echo spacings in EPI readouts, thus reducing distortion; however, dedicated head gradient coils are not available on most research or clinical MRI scanners. High spatial order B₀ shimming: EPI distortion and other artifacts can be eliminated at their source by improving the MRI scanner’s B₀ shimming capability beyond standard 2nd order shimming. Example hardware approaches include dedicated high-order spherical harmonic insert coils [26] and local multi-coil shim arrays [27]. With both approaches, shim performance can be further improved for 2D acquisitions by dynamically switching shim currents to be optimized for each acquired slice, rather than attempting to shim the whole head globally.

• Time-varying B₀ offsets: Spatiotemporal B₀ variations caused by physiological processes such as respiration cause ghosting in structural images and time-series instability (time-varying distortion) in functional EPI scans. These effects can be severe at ultra-high field and in body regions closer to the lungs, such as the cervical spine. Methods have been developed to track these changes either using rapidly-acquired single slice ΔB₀ field maps [28] or NMR field probes data correlated with respiratory bellows signals [29]. Accurate knowledge of these fields can be used to drive real-time shim updating tracking the respiratory cycle or to correct k-space data in post-processing to reduce artifacts.

B1 INHOMOGENEITY

• The interaction of the transmit radiofrequency (RF) fields (B₁⁺) with the body is a major limitation on MRI performance in the brain at 7T and above, and in the body at 3T and above. B₁⁺ inhomogeneity arises due to interactions of RF fields with body tissues having relatively high permittivity (~80 in the brain) in addition to significant conductivity (~1.8 Siemens/meter for cerebral spinal fluid). Since the RF wavelength scales with the square root of the permittivity, the wavelength at 7T (~297-300 MHz) is smaller than the diameter of the head, leading to dielectric wavelength effects and regions of constructive and destructive interference [30]. For a quadrature birdcage transmit coil at 7T, this typically leads to enhanced B₁⁺ in the center of the head and reduced B₁⁺ in the temporal lobes and inferior structures such as the cerebellum. In practice, the impact of B₁⁺ variations on image quality are too severe to correct in post-processing, especially for spin echo sequences, where B₁⁺ inhomogeneity dramatically degrades refocusing pulse performance.

• B₁⁺ mapping: The gold standard method for B₁⁺ mapping is to apply multiple nominal flip angles between 0 and 90 deg. (ideally with a long TR to avoid T₁ effects) and then fit the image intensity in each voxel to a sine curve. But this method is time-consuming to implement in practice. A faster method called “actual flip angle imaging” uses the ratio of signal intensity for gradient echo readouts acquired using two different short TRs. This method is insensitive to T₁ and allows 3D B₁⁺ maps to be rapidly acquired. An alternative approach uses the Bloch-Siegert (B-S) shift created by an off-resonance transmit RF pulse to compute B₁⁺ with high fidelity [31]. The drawback of B-S B₁⁺ mapping is the high specific absorption rate (SAR) incurred by the off-resonance pulses. Recent improvements in pulse design have achieved reduced SAR and improved robustness to B₀ inhomogeneity [32].

• Dealing with B₁⁺ inhomogeneity at 7T: Adiabatic RF pulses that are robust to B₀ and B₁⁺ variations have been used to homogenize image contrast over the whole brain [33], however these pulses typically incur a SAR penalty and/or long pulse duration. Thin pads made from high permittivity materials (e.g. barium titanate) placed on the sides of the head have been used to successfully “lens” the RF transmit field to improve performance in problem areas such as the temporal lobes [34]. Finally, perhaps the most powerful and versatile B₁⁺ mitigation tactic is to use an array of transmit coils to provide more degrees of freedom for shaping the RF field inside the body. In the simplest realization, known as “RF shimming”, the amplitude and phase of the transmit pulse on each channel is adjusted to achieve a more uniform B₁⁺ distribution. In the most general case, known as “parallel transmit” or “pTx”, each channel is driven independently by its own RF power amplifier [35]. This enables a unique RF pulse to be used for each channel, providing full flexibility to control the resulting B₁⁺ field produced by the array inside the body. In this approach, the RF pulses and gradient waveforms can be jointly optimized to improve performance. While a powerful tool, pTx typically requires accurate knowledge of each transmit channel’s subject-specific B₁⁺ field map. Recent advances have improved our ability to estimate these B₁⁺ maps [36] and to design “universal” pTx pulses that work across a wide variety of subjects’ heads [37].

Acknowledgements

References

1. Jovicich J, Czanner S, Greve D, Haley E, Van Der Kouwe A, Gollub R, Kennedy D, Schmitt F, Brown G, MacFall J, Fischl B, Dale A. Reliability in multi-site structural MRI studies: Effects of gradient non-linearity correction on phantom and human data. Neuroimage. 2006;30(2):436-443. PMID: 16300968.

2. Kornak J, Marinelli L, Hylton N. Gradient Nonlinearity Correction to Improve Apparent Diffusion Coefficient Accuracy and Standardization in the American College of Radiology Imaging Network 6698 Breast Cancer Trial. J Magn Reson Imag. 2016;42(4):908-919.

3. Hedeen RA, Edelstein WA. Characterization and Prediction of Gradient Acoustic. Magn Reson Med. 1997;37(1):7-10.

4. Le Bihan D, Poupon C, Amadon A, Lethimonnier F. Artifacts and pitfalls in diffusion MRI. J Magn Reson Imaging. 2006. PMID: 16897692.

5. Setsompop K, Kimmlingen R, Eberlein E, Witzel T, Cohen-Adad J, McNab JA, Keil B, Tisdall MD, Hoecht P, Dietz P, Cauley SF, Tountcheva V, Matschl V, Lenz VH, Heberlein K, Potthast A, Thein H, Van Horn J, Toga A, Schmitt F, Lehne D, Rosen BR, Wedeen V, Wald LL. Pushing the limits of in vivo diffusion MRI for the Human Connectome Project. Neuroimage. 2013;80:220-233. PMID: 23707579.

6. Welz A, Zaitsev M, Jia F, Liu Z, Korvink J, Schmidt H, Lehr H, Post H, Dewney A, Hennig J. Development of a non-shielded PatLoc gradient insert for human head imaging. Proc 17th Sci Meet Int Soc Magn Reson Med. 2009;Honolulu(2007):3073.

7. De Zanche N, Barmet C, Nordmeyer-Massner JA, Pruessmann KP. NMR Probes for measuring magnetic fields and field dynamics in MR systems. Magn Reson Med. 2008;60(1):176-186. PMID: 18581363.

8. Barmet C, De Zanche N, Pruessmann KP. Spatiotemporal magnetic field monitoring for MR. Magn Reson Med. 2008;60(1):187-197. PMID: 18581361.

9. Vannesjo SJ, Haeberlin M, Kasper L, Pavan M, Wilm BJ, Barmet C, Pruessmann KP. Gradient system characterization by impulse response measurements with a dynamic field camera. Magn Reson Med. 2013;69(2):583-593. PMID: 22499483.

10. Stockmann JP, Wald LL. In vivo B<inf>0</inf> field shimming methods for MRI at 7T. Neuroimage. 2017.

11. Rosen BR, Wedeen VJ, Brady TJ. Selective saturation in NMR. J Comp Assist Tomog. 1984;8(5):813-818.

12. De Graaf RA, Wiley InterScience (Online service). In Vivo NMR Spectroscopy : Principles and Techniques. John Wiley & Sons; 2007.

13. Hargreaves BA, Cunningham CH, Nishimura DG, Conolly SM. Variable-rate selective excitation for rapid MRI sequences. Magn Reson Med. 2004;52(3):590-597. PMID: 15334579.

14. Schneider R, Ritter D, Haueisen J, Pfeuffer J. B0-informed variable density trajectory design for enhanced correction of off-resonance effects in parallel transmission. Magn Reson Med. 2014;71(4):1381-1393. PMID: 23716347.

15. Davids M, Schad LR, Wald LL, Guerin B. Fast three-dimensional inner volume excitations using parallel transmission and optimized k-space trajectories. Magn Reson Med. 2016;76(4):1170-1182. PMID: 26527590.

16. Gore JC, Barry RL, Vannesjo SJ, Smith SA, By S. Spinal cord MRI at 7T. Neuroimage. 2017;168:437-451.

17. Stockmann JP, Wald LL. In vivo B0 field shimming methods for MRI at 7T. Neuroimage. 2017;(June):71-87. PMID: 28602943.

18. Hetherington HP, Chu WJ, Gonen O, Pan JW. Robust fully automated shimming of the human brain for high-field 1H spectroscopic imaging. Magn Reson Med. 2006;56(1):26-33. PMID: 16767750.

19. Jenkinson M, Beckmann CF, Behrens TEJ, Woolrich MW, Smith SM. Fsl. Neuroimage. 2012;62(2):782-790. PMID: 21979382.

20. Karsa A, Shmueli K. SEGUE: a Speedy rEgion-Growing algorithm for Unwrapping Estimated phase. IEEE Transactions on Medical Imaging. 2018.

21. Robinson S, Schödl H, Trattnig S. A method for unwrapping highly wrapped multi-echo phase images at very high field: UMPIRE. Magn Reson Med. 2014.

22. Jezzard P, Balaban RS. Correction for geometric distortion in echo planar images from B(o) field variations. Magn Reson Med. 1995;34(1):65-73. PMID: 7674900.

23. Andersson JLR, Skare S, Ashburner J. How to correct susceptibility distortions in spin-echo echo-planar images: Application to diffusion tensor imaging. Neuroimage. 2003;20(2):870-888. PMID: 14568458.

24. Griswold M a., Jakob PM, Heidemann RM, Nittka M, Jellus V, Wang J, Kiefer B, Haase A. Generalized autocalibrating partially parallel acquisitions (GRAPPA). Magn Reson Med. 2002;47(6):1202-1210. 25. Wald LL. The future of acquisition speed, coverage, sensitivity, and resolution. Neuroimage. 2012;62(2):1221-1229. PMID: 22421052.

26. Kim T, Lee Y, Zhao T, Hetherington HP, Pan JW. Gradient-echo EPI using a high-degree shim insert coil at 7 T: Implications for BOLD fMRI. Magn Reson Med. December 2016. PMID: 27910126.

27. Juchem C, Nixon TW, McIntyre S, Boer VO, Rothman DL, De Graaf RA. Dynamic multi-coil shimming of the human brain at 7 T. J Magn Reson. 2011;212(2):280-288. PMID: 21824794.

28. Vannesjo SJ, Miller KL, Clare S, Tracey I. Spatiotemporal analysis of breathing-induced fi elds in the cervical spinal cord at 7T. Proc Int Soc Magn Reson Med. 2017:0752.

29. Duerst Y, Wilm BJ, Dietrich BE, Vannesjo SJ, Barmet C, Schmid T, Brunner DO, Pruessmann KP. Real-time feedback for spatiotemporal field stabilization in MR systems. Magn Reson Med. 2015;73(2):884-893. PMID: 24634192.

30. Vaidya M V., Collins CM, Sodickson DK, Brown R, Wiggins GC, Lattanzi R. Dependence of B1- and B1+ field patterns of surface coils on the electrical properties of the sample and the MR operating frequency. Concepts Magn Reson Part B Magn Reson Eng. 2016.

31. Sacolick LI, Wiesinger F, Hancu I, Vogel MW. B1 mapping by Bloch-Siegert shift. Magn Reson Med. 2010;63(5):1315-1322. PMID: 20432302.

32. Khalighi MM, Rutt BK, Kerr AB. RF pulse optimization for Bloch-Siegert B+1 mapping. Magn Reson Med. 2012;68(3):857-862. PMID: 22144397.

33. Wrede KH, Johst S, Dammann P, Umutlu L, Schlamann MU, Sandalcioglu IE, Sure U, Ladd ME, Maderwald S. Caudal Image Contrast Inversion in MPRAGE at 7 Tesla. Problem and Solution. Acad Radiol. 2012;19(2):172-178. PMID: 22104286.

34. Teeuwisse WM, Brink WM, Webb AG. Quantitative assessment of the effects of high-permittivity pads in 7 Tesla MRI of the brain. Magn Reson Med. 2012. PMID: 21826732.

35. Katscher U, Boernert P. Parallel RF transmission in MRI. NMR Biomed. 2006;19(3):393-400. PMID: 16705630.

36. Cloos MA, Knoll F, Zhao T, Block KT, Bruno M, Wiggins GC, Sodickson DK. Multiparametric imaging with heterogeneous radiofrequency fields. Nat Commun. 2016.

37. Gras V, Vignaud A, Amadon A, Le Bihan D, Boulant N. Universal pulses: A new concept for calibration-free parallel transmission. Magn Reson Med. 2017;77(2):635-643. PMID: 26888654.