Three-dimensional MRI in a homogenous 27 cm diameter bore Halbach array magnet

https://doi.org/10.1016/j.jmr.2019.106578Get rights and content

Highlights

  • A 23 ring Halbach array of permanent magnets has been designed to operate at 2.15 MHz.

  • Addition of shimming rings gives 2400 ppm homogeneity over a 20 cm diameter.

  • Three-dimensional images with a resolution of 3.5 × 3.5 × 3.5 mm have been acquired.

Abstract

Modern clinical MRI systems utilise very high magnetic fields strengths to produce high resolution images of the human body. The high up-front and maintenance cost of these systems means that much of the world lacks access to this technology. In this paper we propose a low cost, head-only, homogenous Halbach magnet array with the potential for paediatric neuroimaging in low-resource settings. The homogeneity of the Halbach array is improved by allowing the diameter of the Halbach array to vary along its length, and also adding smaller internal shim magnets. The constructed magnet has a bore diameter of 27 cm, mean B0 field strength of 50.4 mT and a homogeneity of 2400 ppm over a 20 cm diameter spherical volume. The level of homogeneity of the system means that coil-based gradients can be used for spatial encoding which greatly increases the flexibility in image acquisition. 3D images of a “brain phantom” were acquired over a 22 × 22 × 22 cm field of view with a 3.5 mm isotropic resolution using a spin-echo sequence. Future development of a low-cost gradient amplifier and an open-source spectrometer has the potential of offering a fully open-source, low-cost MRI system for paediatric neuroimaging in low-resource settings.

Graphical abstract

Photograph of a Halbach array containing 2948 individual magnets, operating at Larmor frequency of 2.15 MHz, with a homogeneity of 2400 ppm over a 20 cm diameter of spherical volume.

  1. Download : Download high-res image (98KB)
  2. Download : Download full-size image

Introduction

The first MRI scanners produced in the 1980s had field strengths on the order of 0.1–0.5 Tesla. With improved magnet design and technology, human-sized superconducting magnets of 1.5 T and 3 T became commercially available, and most clinical research is performed at these field strengths. Typical purchase costs of a commercial MRI scanner are 1 million euros per Tesla [1], with annual service contracts of hundreds of thousands of euros, and a high level of expertise required for operation and repair, placing such systems completely out of reach for many communities [2]. Eliminating the superconducting magnet from the MRI system allows for a significant reduction in cost but comes with a large reduction in available magnetic field strength. The major problem with such low-field MRI systems is simply signal-to-noise ratio (SNR) which, in the low field limit, is proportional to the 7/4th power of B0 [3], and while the this ratio tends towards linearity with increasing magnetic field strength reducing the magnetic field strength from a typical clinical strength of 1.5 T to ∼50 mT (the field strength of the magnet described in this paper) comes with a several hundred fold reduction in SNR. However, there are also some distinct advantages of low-field MRI [4] including the ability to scan patients with implants and the fact that the power deposited is much lower, meaning that the specific absorption rate (SAR) which is determined by federal law, is in practice never reached.

The pioneering attempts at performing human MRI at low-fields used the concept of pre-polarized MRI, in which an inhomogeneous pulsed magnetic field could be used to polarize the nuclei, with the signal being read out in a more homogeneous lower magnetic field. In the 1990s the Stanford group [5] showed that this principle could be used for hand and wrist imaging. Many academic groups have developed unconventional detectors such as atomic magnetometers [6], [7] and superconducting quantum devices (SQUIDs) [6], [8], [9], [10], [11] but these have remained in the academic arena. More recently, larger bore magnets have been produced with the aim of brain imaging. One example has been the Helmholz coil-based system used to acquire images of the human brain at 6.5 mT: images have been acquired at a spatial resolution of 2.5 × 3.5 × 8.5 mm3 using balanced steady-state free-precession techniques [12]. This system has produced high quality human brain images at very low field. The only disadvantage is that its large size reduces the portability of the system, a facet which would prove useful in developing countries.

The pioneering work of Blümler and associates [13], [14], [15], [16], [17], [18] as well as Perlo, Casanova and Blümlich [19], [20], [21], [22] highlighted the potential of a discretized version of a Halbach magnet, referred to as a Mandhala, to produce the main B0-field using arrays of permanent magnets. Through a combination of very high length-to-bore ratios and the ability to finely tune the positions of the individual magnets in the Halbach arrays field homogeneities needed to perform high-resolution NMR were obtained. The application of Halbach arrays to in-vivo has been limited due to the increased inhomogeneity that arises when the length-to-diameter ratio decreases [23] which is a practical requirement when building systems with a bore size suitable for human imaging. Designs that build in a B0-gradient in Halbach arrays as a spatial encoding method have been shown [24], [25], [26] but face technical challenges due to the high gradient strength that is required to overcome imperfections in the encoding field and may be slightly less flexible than systems that use conventional gradient based encoding since the gradient is a fixed feature of the magnet design.

The particular aim of our project is to construct a platform to image pediatric hydrocephalus in developing countries. As discussed extensively by Obungoloch et al. [27] and references therein hydrocephalus is one of the most common pediatric conditions which requires both neuroimaging and neurosurgery. The image spatial resolution can be much coarser than used for conventional neuro-MRI, with voxels on the order of 3 × 3 × 10 mm sufficient for identifying fluid compartments for fenestration or drainage. Critical requirements for such a system include low upfront cost, realized by the earlier described reduction in magnet cost and lower demands on the RF and gradient amplifiers due to the relatively small size of the system, a reduced operational and maintenance cost by removing the need for cryogen cooling and using electronic components that can easily be replaced or repaired, (ideally) portability which will allow greater access to MR technology in remote regions as well as making citing of the system in fixed locations easier, and very simple data acquisition, e.g. a simple three-dimensional spin-echo pulse sequence requiring minimal planning in cases where highly trained radiologists are not available. One can think of other applications of this technology if these requirements are met; deployment to disaster regions and field hospitals and integration in to ambulances to reduce the time needed to acquire (initial) diagnostic images. Additionally, a size and cost reduction of the MRI system opens up non-human applications such as quality control on food products and water quality monitoring.

In this paper we set out to design a head-only, homogenous Halbach magnet array for imaging hydrocephalus in young children using conventional-gradient-based image reconstruction. The homogeneity of the magnet is improved by optimizing the radius of the Halbach cylinder along the length of the magnet. The choice was made to use smaller magnets than other Halbach designs with the aim of averaging out the inevitable manufacturing imperfections of each individual magnet: this also reduces the demands on the magnet enclosure in terms of structural strength and weight, and increases safety during the construction process.

Section snippets

Magnet design

The minimum bore size is determined by the width of the shoulders that need to be accommodated; in this magnet design we chose a bore diameter of 27 cm which is sufficiently large to accommodate the majority of the target population of paediatric patients up to the age of 8.

The main trade-offs to consider in the magnet design are the magnetic field homogeneity, the absolute value of the magnetic field, and the weight and the cost of the total assembly. As discussed in several previous

Results

Fig. 3 shows the simulated magnetic field distribution of the optimised Halbach ring configuration generated by the genetic algorithm. The B0 field at the centre of the magnet was simulated to be 50.67 mT with a B0 homogeneity over a 20 cm DSV (marked by the white dotted line) of 440 ppm.

Fig. 4(a) shows the magnetic field measured with the Hall probe and the three-dimensional robotic positioner with the same field-of-view as the simulations shown in Fig. 3. The homogeneity over the same 20 cm

Discussion

There are several potential improvements to the system which are being investigated.

One of the disadvantages of using permanent magnet material is the temperature dependence of the magnetic field. In the case of N48 NdBFe magnets, the dependence is −0.12%/°C, corresponding to a field drift of −2.5 kHz per degree temperature increase. Heat produced by the gradient system has the potential to cause a drift in the magnetic field strength. In order to see if this shift is significant in terms of

Conclusion

This work has shown the feasibility of performing three-dimensional MRI using a custom-designed low-cost, high-homogeneity Halbach magnet. Images have been acquired at a spatial resolution of a few millimeters within a data acquisition time of tens of minutes with a SNR of ∼35 after matched filtering has been applied. Many of the components of this system have been custom-built, and future work will aim to replace the remaining commercial components by custom-designed ones to further the aim of

Acknowledgements

This work was supported by Horizon 2020 European Research Grant FET-OPEN 737180 Histo MRI, Horizon 2020 ERC Advanced NOMA-MRI 670629, Simon Stevin Meester Prize and NWO WOTRO Joint SDG Research Programme W 07.303.101. We are grateful to Drs. Martin van Gijzen and Rob Remis at the TU Delft for collaborative discussions, and Danny de Gans at the TU Delft for construction of the RF amplifier.

References (44)

  • M. Nakagomi et al.

    Development of a small car-mounted magnetic resonance imaging system for human elbows using a 0.2 T permanent magnet

    J. Magn. Reson.

    (2019)
  • L. Glover, Why Does an MRI Cost So Darn Much?, Money, n.d. http://money.com/money/2995166/why-does-mri-cost-so-much/...
  • WHO, Global Maps for Diagnostic Imaging, WHO, n.d....
  • W.A. Edelstein et al.

    The intrinsic signal-to-noise ratio in NMR imaging

    Magn. Reson. Med.

    (1986)
  • J.P. Marques et al.

    Low-field MRI: An MR physics perspective

    J. Magn. Reson. Imag.

    (2019)
  • A. Macovski et al.

    Novel approaches to low-cost MRI

    Magn. Reson. Med.

    (1993)
  • I. Savukov et al.

    Magnetic-resonance imaging of the human brain with an atomic magnetometer

    Appl. Phys. Lett.

    (2013)
  • A.N. Matlashov et al.

    SQUIDs vs. induction coils for ultra-low field nuclear magnetic resonance: experimental and simulation comparison

    IEEE Trans. Appl. Supercond.

    (2011)
  • R. McDermott et al.

    Microtesla MRI with a superconducting quantum interference device

    PNAS

    (2004)
  • V.S. Zotev et al.

    Multi-channel SQUID system for MEG and ultra-low-field MRI

    IEEE Trans. Appl. Supercond.

    (2007)
  • V.S. Zotev et al.

    SQUID-based instrumentation for ultralow-field MRI

    Supercond. Sci. Technol.

    (2007)
  • M. Sarracanie et al.

    Low-cost high-performance MRI

    Sci. Rep.

    (2015)
  • Cited by (83)

    View all citing articles on Scopus
    View full text