There has long been a push for higher field MRI systems to increase magnetic polarization and therefore SNR. However, the development of MRI systems at low-field (< 0.1T) and ultra-low field (< 10mT) could lead to an overall increase in the accessibility of MRI in terms of cost and portability. An overview of low-field MRI system design and imaging aspects will be presented, and the advantages and disadvantages of low-field MRI compared to high-field MRI will be discussed.
Advantages of low-field MRI
The major draw to work in low-field MRI is cost and portability. While soft-tissue image quality of conventional MRI scanner is unsurpassed, the cost and siting requirements of high-field human scanners can be prohibitive. The cost associated with the superconducting full body magnet and associated infrastructure is a significant portion of the total cost of siting a scanner. Eliminating these components, not only decreases the cost, but also opens the door to more mobile/portable operation. Most portable MRI endeavors have also been low-field because superconducting magnets are heavy and have stringent siting requirements related to the cooling (quench pipe), and the safety exclusion zone of HF magnets prohibits use in most environments. While it is unlikely that low-field scanner will ever compete with the diagnostic performance of high field scanners in terms of SNR (and therefore shorter imaging times and resolution), there are many scenarios in which the increased accessibility (in terms of cost and portability) of the device is more valuable and potentially life-saving. One compelling application for portable MR diagnostic devices is point-of-care imaging. For example, in hospitals when patients cannot safely be transported to the high-field scanner, a truly portable low-field scanner could potentially be transported to the patient. In these types of cases, where the patient is in critical condition, the source of the ailment is likely known (for example post-operative hemorrhaging). With some knowledge of the condition or trauma, low-field, portable MR devices could answer critical questions in a timely fashion, which could improve the prognosis. In addition to point-of-care devices, there is a more general need for MR imagers that are lower-cost and easier to site than conventional HF scanners. There are some obvious unconventional environments that could benefit from easy-to-site low-field MR diagnostics, including battle field hospitals, rural clinics, and sports arenas. Low-field scanners are far more compatible for these types of non-traditional environments where safety exclusion zones are difficult to enforce. In addition to cost and portability benefits, low-field MR has some unique imaging functionality. It is well-known that T1 relaxation times decrease and T1 contrast improves at lower field strengths [2], [3]. This increased contrast could enable delineation of certain tissue types better than HF, and enable tumor detection without contrast agents [4], [5]. Field cycling at low-field has also been used to vary the signal detection field and study T1 dispersion patterns in tissue [4], [6], [7]. This can be used for protein contrast imaging by taking advantage of the nitrogen dip in the T1 dispersion curve. This functionality is especially synergistic with the Stanford low-cost extremity scanner for diseases that affect muscle viability [6]. Low field MRI is also highly preferred to HF for imaging in the presence of metal implants, or penetrating trauma with metallic material (e.g. shrapnel, bullets). Not only is it safer for the patient, but low-field MRI is much less vulnerable to susceptibility effects. This immunity to susceptibility artifacts has been convincingly demonstrated with low-field scanners at MGH [8], Stanford [9], and Berkeley [4].Disadvantages of low-field MRI
The obvious draw-back of operation in low-field is the degraded signal-to-noise ratios (SNR). With traditional MRI hardware, the MR signal scales with B02. One factor is directly linked to the Boltzmann polarization of the sample from the applied B0 field. The second factor is related to the traditional detection method using inductive receiver coil. According to Faraday’s law, the voltage induced in the coil scales with dB/dt, and therefore the frequency of the MR signal factors into the measured signal level. With traditional hardware, the only way to approach the SNR in HF scanner is with signal averaging, which ideally is proportional to the square root of the number of measurements. For this reason, non-traditional methods like pre-polarization and SQUID magnetometers are often used to obtain acceptable SNR. Traditional image encoding at low-field is also challenging due to significant concomitant gradient fields [10]–[12]. Maxwell’s equations predict that any coil generating the desired gradients in Bz will also generate unwanted transverse components with the same field strength as the Bz gradient. At HF, these concomitant fields are insignificant compared to B0, but at ULF concomitant gradient coils can lead to significant phase and amplitude distortion. Therefore, low gradient strengths (compared to HF) are often used ~10-4T/m for ULF, to reduce the effect of the concomitant fields leading to longer readout times compared to HF for the same resolution target. The longer readout times due to weaker gradients is compounded by the large number of averages needed for sufficient SNR, leading to significantly longer imaging time compared to HF. However, promising work on been conducted for correcting for distortions caused by the concomitant fields at ultra-low field [10], [13], [14].Review of low-field MRI systems
To achieve sufficient SNR for imaging, many low-field MRI scanners use pre-polarization or field-cycling to boost the Boltzmann nuclear magnetism prior to signal detection at a lower field. This requires 2 B0 fields: 1) The stronger pulsed polarizing field, Bp, 2) The weaker homogeneous signal detection (readout) field, Bm. If Bp is applied orthogonally to Bm, and turned off rapidly (non-adiabatically), then the magnetization, M, is left aligned Bp, causing the spins to precess about Bm without the need for an excitation pulse. If Bp is turned off adiabatically (dBp/dt << γBm2), then the final M is left aligned Bm, and must be excited with a transmit coil. Non-adiabatic Bp pulsing has challenges associated with quickly dissipating stored energy in the Bp coil and Bp induced eddy currents. On the other hand, adiabatic Bp pulses take longer, resulting in T1 signal loss, and requires a spin-flip pulse. Since signal scales with B2 in MRI, ideal Bp determines one factor of B in these pre-polarized systems, leading to increased SNR.
This idea was demonstrated in 1954 by Packard and Varian using the earth’s magnetic field (~50uT) for NMR signal detection [15]. In this system, a pulsed 10mT electromagnet orientated perpendicular to the earth’s field was used to pre-polarize the spins. The electromagnetic polarizing field is rapidly turned off (non-adiabatically), causing the spins to precess about the earth’s field. This idea was extended to Nuclear Magnetic Well-logging (NML) [16], the study of liquids [17], and eventually extended to imaging with the addition of gradient coils [18].
Pre-polarized ultra-low field (ULF) MRI scanners using Superconducting Quantum Interference Devices (SQUIDs) have been developed at Los Alamos (Bm = 46uT) [19], [20] and Berkeley (Bm = 130uT) [4], [21]. In these cases, the Larmor frequency is f0 < 10kHz, a regime in which tuned induction coils are relatively insensitive. In these systems, the induction receive coils are replaced with SQUID magnetometers, whose sensitivity is decoupled from the precession frequency. SQUID magnetometers are challenging to use even at ULF because of their small dynamic range, requiring Pb or Nb shielding of the SQUID and current-limiting cryo-switches or Josephson junctions for the pick-up coils. However, these types of detectors are up to 2 orders of magnitude more sensitive (1fT/Hz) in this ULF regime compared to inductive receiver coils. At Los Alamos, the use of SQUID arrays has also enabled combined MEG/ULF MRI measurements. This combines the relative resolution strengths of the 2 modalities (spatial for MRI and temporal for MEG), allowing accurate registration of the 2 modalities and enabling increased accuracy of reconstruction of the MEG data (an ill-posed inverse problem) [19], [20]. Both the Berkeley system and Los Alamos system have demonstrated human brain images.
In 1993, Macovski and Conolly introduced a novel low-cost MRI design using a 0.42T pulsed pre-polarization field and a 52mT bias field [22]. Instead of using a single homogeneous high-field magnet, a (relatively) high-field pulsed field is used to polarizing the spins with a low homogeneity requirement; and a homogeneous low-field field is used for the readout. Although, this requires two magnets instead of one, the specifications are much easier to attain at a low-cost. This pre-polarized MRI (PMRI) scanner has been used to produce impressive T1 and T2 weighted wrist images with a 3D RARE sequence [23]. The PMRI scanner was also used to produce T1 dispersion images by using a variable “evolution field” between the pre-polarization pulse and the readout. Protein contrast images were generated by subtracting images acquired with evolution fields at the top and bottom of the nitrogen dip in the T1 dispersion curve [6].
This format of using a polarization (Bp), evolution (Be), and detection (Bd) B0 field is analogous to the fast field-cycling (FFC) MRI method used by Lurie et al. in Aberdeen [7], [24]. Instead of using a separate polarization and readout magnet, the Aberdeen scanner uses a primary static magnet and an additional field-offset coil that enables an evolution field to be selected. Two systems have been described using 1) a 59mT Halbach magnet and 2) a 450mT superconducting magnet.
At MGH, a 6.5mT LF scanner has been developed without using pre-polarization or cryogenics. The scanner uses a planar water-cooled coils to create a very open bore, and has been used for hyperpolarized lung imaging in upright subjects [25] and human brain imaging [8]. Sufficient SNR is obtained with the use of fast imaging techniques that enable significant signal averaging in a reasonable time. This is done with under-sampled 3D fully balance SSFP, which dynamically refocuses spin magnetization following the measurement, eliminating the T1 recovery delay [26]. 3D brain images have been demonstrated in 6min, with distinguishable: hemispheres, cerebellum, cortical tissue vs. white matter tissue. The resulting contrast with b-SSFP is related to T2/T1, which is close to unity at ULF, leading to contrast similar to proton-density.
Work on portable MRI systems has generally been done at low-field due to reduced safety risks and the fact that low-field magnets can be more lightweight. The size of the magnet can also generally be reduced by relaxing the homogeneity constraint of the magnet. There have been several portable NMR/MRI devices that use an inhomogeneous magnet for polarization and image encoding. In these cases, T2* is very short and the magnetization must be refocused with RF pulses, limiting acquisition to spin-echo based sequences. The oil well-logging industry was the first to explore the idea of “external sample” or “inside out” NMR for measuring fluid in rock formation [16]. The NMR MOUSE (Mobile Universal Surface Explorer) is a handheld single-sided device that exploits the field fall off as an encoding field for 1D imaging [27], [28]. The idea of using the natural field variation as an encoding field gradient was also used for 1D in the Magnevu portable extremity scanner [29].
At MGH, the idea of image encoding with the natural field variation of an inhomogeneous magnet was extended to a portable MRI scanner for brain imaging using a Halbach magnet [30]. Instead of using a separate B0 magnet and gradient system, a single permanent structure was used for polarization and image encoding. The inhomogeneity of the lightweight, low-field (70mT) magnet is exploited for image encoding in the 2D axial plane. The encoding field is modulated by physical rotating the magnet around the sample. To encode along the axis of the Halbach cylinder, Transmit Array Spatial Encoding (TRASE) is used [31], [32]. TRASE is an RF encoding technique that uses long spin-echo chains, and benefits from operation at low-field due to decreased susceptibility to RF heating.
Conclusion
Specialized low-field, low-cost, and/or portable MRI scanners could be used as complementary diagnostic or monitoring tools for to answering specific time-critical questions. Although, low-field MRI systems suffer from low SNR, potential gains in accessibility are continually driving factor for new research.[1] A. Macovski, “MRI: a charmed past and an exciting future,” J. Magn. Reson. Imaging JMRI, vol. 30, no. 5, pp. 919–923, Nov. 2009.
[2] S. H. Koenig, R. D. Brown, M. Spiller, and N. Lundbom, “Relaxometry of brain: Why white matter appears bright in MRI,” Magn. Reson. Med., vol. 14, no. 3, pp. 482–495, Jun. 1990.
[3] R. D. Brown and S. H. Koenig, “1 /T1p and Low-Field 1 / T1 of Tissue Water Protons Arise from Magnetization Transfer to Macromolecular Solid-State Broadened Lines,” Magn. Reson. Med., vol. 28, no. 1, pp. 145–152, Nov. 1992.
[4] J. Clarke, M. Hatridge, and M. Mößle, “SQUID-Detected Magnetic Resonance Imaging in Microtesla Fields,” Annu. Rev. Biomed. Eng., vol. 9, no. 1, pp. 389–413, 2007.
[5] S.-H. Liao et al., “Characterization of tumors using high-Tc superconducting quantum interference device-detected nuclear magnetic resonance and imaging,” Appl. Phys. Lett., vol. 97, no. 26, p. 263701, Dec. 2010.
[6] S. E. Ungersma et al., “Magnetic resonance imaging with T1 dispersion contrast,” Magn. Reson. Med., vol. 55, no. 6, pp. 1362–1371, Jun. 2006.
[7] D. J. Lurie et al., “Fast field-cycling magnetic resonance imaging,” Comptes Rendus Phys., vol. 11, no. 2, pp. 136–148, Mar. 2010.
[8] M. Sarracanie, C. D. LaPierre, N. Salameh, D. E. J. Waddington, T. Witzel, and M. S. Rosen, “Low-Cost High-Performance MRI,” Sci. Rep., vol. 5, p. 15177, Oct. 2015.
[9] R. D. Venook et al., “Prepolarized magnetic resonance imaging around metal orthopedic implants,” Magn. Reson. Med., vol. 56, no. 1, pp. 177–186, Jul. 2006.
[10] P. L. Volegov, J. C. Mosher, M. A. Espy, and R. H. Kraus Jr., “On concomitant gradients in low-field MRI,” J. Magn. Reson., vol. 175, no. 1, pp. 103–113, Jul. 2005.
[11] D. G. Norris and J. M. S. Hutchison, “Concomitant magnetic field gradients and their effects on imaging at low magnetic field strengths,” Magn. Reson. Imaging, vol. 8, no. 1, pp. 33–37, Jan. 1990.
[12] D. A. Yablonskiy, A. L. Sukstanskii, and J. J. H. Ackerman, “Image artifacts in very low magnetic field MRI: The role of concomitant gradients,” J. Magn. Reson., vol. 174, no. 2, pp. 279–286, Jun. 2005.
[13] J. O. Nieminen and R. J. Ilmoniemi, “Solving the problem of concomitant gradients in ultra-low-field MRI,” J. Magn. Reson., vol. 207, no. 2, pp. 213–219, Dec. 2010.
[14] Y.-C. Hsu et al., “Efficient concomitant and remanence field artifact reduction in ultra-low-field MRI using a frequency-space formulation,” Magn. Reson. Med., vol. 71, no. 3, pp. 955–965, Mar. 2014.
[15] M. Packard and Varian, R, “Free Nuclear Induction in the Earth’s Magnetic Field,” Phys. Rev., vol. 93, p. 941, 1954.
[16] R. L. Kleinberg and J. A. Jackson, “An introduction to the history of NMR well logging,” Concepts Magn. Reson., vol. 13, no. 6, pp. 340–342, Jan. 2001.
[17] G. J. Béné, “Nuclear magnetism of liquid systems in the earth field range,” Phys. Rep., vol. 58, no. 4, pp. 213–267, Feb. 1980.
[18] J. Stepišnik, V. Eržen, and M. Kos, “NMR imaging in the earth’s magnetic field,” Magn. Reson. Med., vol. 15, no. 3, pp. 386–391, Sep. 1990.
[19] V. S. Zotev et al., “Microtesla MRI of the human brain combined with MEG,” J. Magn. Reson. San Diego Calif 1997, vol. 194, no. 1, pp. 115–120, Sep. 2008.
[20] L. Parkkonen, R. J. Ilmoniemi, F.-H. Lin, and M. Espy, “Ultra-Low-Field MRI and Its Combination with MEG,” in Magnetoencephalography, S. Supek and C. J. Aine, Eds. Springer Berlin Heidelberg, 2014, pp. 941–972.
[21] B. Inglis, K. Buckenmaier, P. Sangiorgio, A. F. Pedersen, M. A. Nichols, and J. Clarke, “MRI of the human brain at 130 microtesla,” Proc. Natl. Acad. Sci. U. S. A., vol. 110, no. 48, pp. 19194–19201, Nov. 2013.
[22] A. Macovski and S. Conolly, “Novel approaches to low-cost MRI,” Magn. Reson. Med., vol. 30, no. 2, pp. 221–230, Aug. 1993.
[23] N. I. Matter et al., “Three-dimensional prepolarized magnetic resonance imaging using rapid acquisition with relaxation enhancement,” Magn. Reson. Med., vol. 56, no. 5, pp. 1085–1095, Nov. 2006.
[24] D. J. Lurie, M. A. Foster, D. Yeung, and J. M. S. Hutchison, “Design, construction and use of a large-sample field-cycled PEDRI imager,” Phys. Med. Biol., vol. 43, no. 7, p. 1877, 1998.
[25] L. L. Tsai, R. W. Mair, M. S. Rosen, S. Patz, and R. L. Walsworth, “An Open-Access, Very-Low-Field MRI System for Posture-Dependent 3He Human Lung Imaging,” J. Magn. Reson. San Diego Calif 1997, vol. 193, no. 2, pp. 274–285, Aug. 2008.
[26] K. Scheffler and S. Lehnhardt, “Principles and applications of balanced SSFP techniques,” Eur. Radiol., vol. 13, no. 11, pp. 2409–2418, Nov. 2003.
[27] R. S. G. Eidmann, “The NMR MOUSE, a mobile universal surface explorer,” J Magn Reson Ser. A, vol. 122, pp. 104–109, 1996.
[28] F. Casanova, J. Perlo, and B. Blümich, Single-Sided NMR. Springer, 2011.
[29] F. H. R. Jr, W. Cornelius, and C. W. Crowley, “Mri system with peripheral access and inhomogeneous field,” CA2239803 A1, 27-Dec-1998.
[30] C. Z. Cooley et al., “Two-dimensional imaging in a lightweight portable MRI scanner without gradient coils,” Magn. Reson. Med., vol. 73, no. 2, pp. 872–883, Feb. 2015.
[31] J. P. Stockmann, C. Z. Cooley, B. Guerin, M. S. Rosen, and L. L. Wald, “Transmit Array Spatial Encoding (TRASE) using broadband WURST pulses for RF spatial encoding in inhomogeneous B0 fields,” J. Magn. Reson. San Diego Calif 1997, vol. 268, pp. 36–48, Jul. 2016.
[32] J. C. Sharp and S. B. King, “MRI using radiofrequency magnetic field phase gradients,” Magn Reson Med, vol. 63, no. 1, pp. 151–161, 2010.