Design of a novel class of open MRI devices with nonuniform Bo, field cycling, and RF spatial encoding
R. Todd Constable1, Charles Rogers III2, Baosong Wu1, Kartiga Selvaganesan1, and Gigi Galiana2

1Radiology and Biomedical Imaging, Yale School of Medicine, New Haven, CT, United States, 2Radiology and Biomedical Imaging, Yale School of Medicine, New, CT, United States


This work describes a novel approach to the design of MRI systems that combines 4 existing developments in order to create a new class of MRI devices. The developments to be described include combining parallel RF receive, Bloch-Siegert phase encoding, nonlinear spatial encoding, and field cycling. Together these methods allow for open magnets with nonuniform main fields greatly increasing design opportunities for small magnets with specific applications. The relatively low cost of this class of MR devices will allow for the placement of MRIs in doctor’s offices where they could be built into an examination table.


In this work we describe a new approach to MRI aimed at developing a new low cost highly accessible MRI device that could be used in doctor’s offices much like ultrasound is today. Four relatively recent developments make such a system feasible. These developments include: (1) parallel receiver technology for sharing some of the burden of spatial encoding; (2) the use of RF for spatial encoding using the Bloch-Siegert shift, eliminating the need for spatial encoding gradients; (3) developments in imaging in nonuniform fields, enabling efficient Bloch-Siegert encoding and eliminating the usual homogeneous field requirements; and (4) field cycling, providing high field for spin polarization and low field for readout reducing sensitivity to field inhomogeneities. We combine these elements in the design of an open MRI system with depth of penetration comparable to that of ultrasound (up to 20cm or more). This system can be built into a patient examination table, mounted on a wall for weighted spine imaging, and potentially made portable to increase the accessibility of MRI (Figure 1).


1. The main field: The main magnet uses a resistive electromagnet that produces a nonuniform Bo that can be ramped up and down rapidly much like a gradient coil. One configuration is shown in figure 1, both alone and in potential applications. This electromagnet produces a maximum field of 0.35T for polarization, which is dropped to 24mT for readout. The field is dropped for readout to minimize the effects of nonuniform Bo (Figure 2). Currently, an electromagnet is required for the main field in order to provide this cycling flexibility. This field cycling approach unfortunately requires multiple polarization periods but this also introduces novel contrast mechanisms not available with fixed field MRIs.

2. Slice selection: Slices are defined by ramping the field down in steps such that each nonplanar slice is at ~24mT at some time for readout, as shown in Figure 2. At this low field the nonuniform Bo is manageable, and the polarization and T1 recovery balance to produce acceptable Mz across the volume. All RF Tx/Rx can be tuned to a fixed readout frequency in this case 1MHz for protons at 24mT. While we currently step the field down, sequences could be envisioned wherein the 24mT resonant frequency plane is swept though the imaging volume in a continuous fashion.

3. RF Spatial Encoding: The system uses no gradient fields for spatial encoding, other than the nonlinear Bo field for slice selection, but instead takes advantage of the Bloch-Siegert shift for spatial encoding. This allows the device to be open, simplifies construction criteria and yields a silent imaging device. Planar array surface coils can be frequency-switched such than any pair of elements from a coil array can be used to generate nonuniform Bloch-Siegert encoding patterns. Different pairs of coils with different relative phases are used to generate B1 fields (Figure 3) that generate a set of nonlinear phasors sufficient for image reconstruction. In addition the BS-pulses can be played during readout since they are off-resonant, and thus they function in a manner similar to a nonuniform readout gradient.

4. Conventional pulse sequences can be easily translated to the proposed system, replacing readouts with nonlinear RF spatial encoding (Figure 4). The usual excitation RF pulses are at the resonance frequency of 1MHz for ~24mT field while the Bloch-Siegert encoding pulses are off-resonant at 873kHz. Frequency-switched parallel coil elements are used for conventional Tx/Rx as well as for Bloch-Siegert encoding. Most pulse sequences are available with this approach with the added flexibility of field-cycling to generate novel contrast mechanisms.

Discussion and Conclusions

In this work we outline the general design of a novel approach to MRI. We reveal a scheme for performing MRI in the presence of a non-uniform Bo field, without gradient coils. The use low field for imaging is necessary to manage the nonuniform Bo field of this system, but using an electromagnet allows field cycling such that we can apply high field for polarization to yield maximal signal and then drop the field for imaging each slice. This approach opens up a new broad class of MRI devices covering a wide range of geometries with numerous design possibilities and clinical applications. The device will be relatively low cost (compared to modern superconducting MRIs) and make MRI much more widely accessible. Such devices could lead to having MRIs in doctor’s offices or specialty imaging centers.


The authors would like to thank Terry Nixon for tremendous help in the design and building of this device.


No reference found.


Figure 1: Field cycling nonuniform magnet for open, low cost MRI. The patient would lay on top of the magnet (it can slide up and down the patient table in the same manner as an x-ray flat panel detector). Conventional contrasts and image resolutions are available and dedicated applications such as MR mammography or weighted spine imaging (with the device mounted on a wall) could be envisioned.

Figure 2: Longitudinal magnetization (top) as a function of field cycling (middle – black indicates magnet current levelt). Maximum field is used for polarization and then stepped down for slice selection. Lowering the field for imaging minimizes the impact of nonuniform Bo. Similar signal intensity is obtained for 10 slices covering up to 20cm depth away from the plane of the magnet. All image slices are nonuniform planes.

Figure 3: RF arrays are used for excitation, Bloch-Siegert spatial encoding, and parallel receive. Since the BS-pulses are off-resonant, it is possible to perform simultaneous BS-encoding and readout such that phasors are produced analogous to readout gradient. Pairs of coil elements are selected for BS pulses (generating sets of different encoding phasors (right panel)) while the remaining coil elements are used for parallel receive. Each element is dual-tuned and can be used for Tx/Rx.

Figure 4: Conventional pulse sequences are available with this device. The field inhomogeneity is always on during imaging but low because the Bo field is low. Bloch-Siegert pulses are played out much like a conventional read gradient though in this case they yield nonuniform phasors for spatial encoding.

Proc. Intl. Soc. Mag. Reson. Med. 27 (2019)