MRI: A Systems Overview
Ralf Berthold Loeffler1

1Diagnostic Imaging, St. Jude Children's Research Hospital, Memphis, TN, United States

Synopsis

MRI systems consist of 3 main components plus computer systems for user interaction, measurement control and signal processing. The 3 components are dedicated to static and gradient magnetic field generation, as well as the RF system for RF transmission and reception. While the purpose of the components has not changed over time, actual implementation has due to technological advances as well as demands by new MRI techniques. This talk will present the basic designs for the different components and discuss current implementations and potential future developments.

Introduction

Forty-five years after its invention1, magnetic resonance imaging (MRI) is now a mature technology. MRI magnets have 3 main components for generating the MRI signal via static and gradient magnetic fields and an RF transmit/receive system.

Static Magnet

The most voluminous part of an MRI creates the main static field B0 to generate 2 states with different energies for the nuclear spin system needed to detect spin transitions. Energy needed/expended by this transition is linear to magnetic field strength:

$$\Delta E\sim B\qquad\qquad\qquad (1).$$

Most MRI magnets have superconducting coils. The basic design includes a pair of solenoid Helmholtz coils (HCs) that create a homogeneous magnetic field. These coils are partially submerged in a liquid-helium tank in a cryostat.

As HCs have a large stray field on the outside, magnets are “self-shielded” by adding another, opposed set of HCs larger than the field-generating coils such that both fields cancel on the outside.

As primary HCs cannot produce fields with the desired homogeneity of <1ppm, small metal shims in trays around the patient bore and additional “shim” coils with higher spatial symmetry are used for field homogenization2. Shims and basic settings for shim coils are calibrated during magnet tune-up, and patient- or location-specific fine tuning of shim coils occurs after patient positioning or before specific measurements if high homogeneity is required.

Magnet development is driven by the trend towards higher field strengths3 and by cost considerations. Latest human magnets use only 20l of liquid helium, and animal magnets generate a field of 3T without using liquid helium.

Magnet Gradients

Magnetic field gradients4 are used to encode spatial information onto the MRI signal. If the magnetic field varies spatially, so does the energy of transition (Eq.1). Therefore, if the spatial variance of the generating field is known, emitted photons can be assigned a spatial location by their energy. As MRI requires strong spatially and temporally varying magnetic fields, it has gradient coils.

Gradient coils generate a field with a vector component parallel to the main field that is linear and depend only on 1 spatial coordinate (x, y, or z). Gradient coils are located outside the cylindrical patient bore. Gradient fields are usually created with modified HCs for the z-field, where current through one side of the coil is reversed to produce a linear field. The x- and y-fields are created by Golay coils, each comprising 2 parallel arches connected with linear conductors. The x- and y-components are created with a set of independent coils relatively rotated by 90°.

Ideal gradients follow a prescribed temporal shape and are completely linear in one and flat in other dimensions, which require more elaborate coil designs (e.g. “fingerprint coils”). The changing magnetic fields induce eddy currents in the cryostat, which can distort the desired field. This can be fixed by building actively shielded gradient coils, which have a counteracting outer coil that reduces the outside field.

Switching gradients creates noise up to 130dBA: when gradients build up a magnetic field inside the static field, a force is exerted on wires that attracts or repels them. This force creates slight movement of the coil and coil mount. As gradient waveforms play out in acoustic-range frequencies, this is perceived as sound, which can be mitigated by building more rigid coils and casting gradients into solid fiberglass as well as use of acoustic isolation. Also, favorable waveforms can be used—sinusoidal waves create less noise than trapezoidal waves.

Although active shielding can yield the desired gradient shape, residual eddy currents and coil inductivity can distort it. This is circumvented by pre-emphasis, wherein the input signal is predistorted so that additional distortions create the desired final signal. In modern magnets, pre-emphasis is done by distorting the digital shape of the gradient sent to gradient amplifiers.

Power stages of modern gradient amplifiers have outputs of several hundred amperes and voltage in the higher-hundreds range. Currently, gradient technology focuses on acoustic issues to increase patient comfort.

RF System

The RF system consists of 2 almost independent subsystems to transmit exciting RF pulses to the volume of interest and receive the resonant signal during the acquisition phase5. Excitation is achieved with a volume (body) coil integrated into the magnet bore. The body coil is connected to an RF power amplifier (output 8–15kW).

Modern receiver systems have multiple antennas/coils. The number of coils used for simultaneous reception has increased due to parallel imaging. Parallel imaging spatial localization is achieved by using gradients and information about the spatial arrangement of receiver coils. With more coils, more spatial localization can be shifted to the RF system.

Individual coils are connected to fixed gain preamplifiers located near coils to minimize transmission noise. Traditionally, the pre-amplified signal was transmitted to the electronics cabinet, where each coil channel is connected to an amplifier, the amplified signal is fed into an analog-to-digital converter, and the digitized signal is recorded. Modern clinical MR systems can have up to 128 individual receive channels6.

Future developments in transmit technology are expected mainly in parallel transmit technology, as ultra-high-field MRI gains traction. On the receive side, the aim is to shorten the signal’s analog path. Vendors digitize the signal close to the coil, as the digital signal is immune to additional analog noise on transmission lines and can be easily multiplexed and transmitted via a single line to the digital signal processor.

Computer System

MRI needs computer power for user interaction with the system (e.g. performing measurements and post-processing), generating measurement instructions and record the received signal in real time, and processing the signal to generate the final image. Even in modern systems, the operating system dedicated to scanner control cannot be off-the-shelf, as measurement control is high priority and data acquisition needs to be performed with minimal interruption. Modern scanners use GPU arrays for massive parallel computing.

Conclusion

In conclusion, MRI's have become more integrated over time, owing to hardware/software integration and automation. Actual implementations are more complex than presented here and involve design decisions varying by vendor. Future developments are expected to further the trend to more integrated and cost-efficient systems.

Acknowledgements

The author thanks Mark Brown, PhD, Arne Reykowski, PhD and Claudia Hillenbrand, PhD for providing material and information. The author also thanks Vani Shanker, PhD for scientific editing.

References

1. Lauterbur PC. Image Formation by Induced Local Interactions - Examples Employing Nuclear Magnetic-Resonance. Nature 1973;242(5394):190-191.

2. Romeo F, Hoult DI. Magnet field profiling: analysis and correcting coil design. Magn Reson Med 1984;1(1):44-65.

3. Lvovsky Y, Jarvis P. Superconducting Systems for MRI—Present Solutions and New Trends. IEEE Transactions on Appiled Superconductivity 2005;15(2):1317-1325.

4. Turner R. Gradient coil design: a review of methods. Magn Reson Imaging 1993;11(7):903-920.

5. Hoult DI. The NMR receiver: A description and analysis of design. Progress in Nuclear Magnetic Resonance Spectroscopy 1978;12(1):41-77.

6. Fujita H. New horizons in MR technology: RF coil designs and trends. Magn Reson Med Sci 2007;6(1):29-42.

7. Kathiravan S, Kanakaraj J. A review on potential issues and challenges in MR imaging. Scientific World Journal 2013;2013:783715.

8. Oppelt A. Imaging systems for medical diagnostics : fundamentals, technical solutions, and applications for systems applying ionizing radiation, nuclear magnetic resonance, and ultrasound. Erlangen, Germany; Berlin: Publicis Corporate Pub. ; Siemens; 2005. 996 p. p.

9. Wood ML, Griswold MA, Henkelman M, Hennig J. Inflection Points in Magnetic Resonance Imaging Technology-35 Years of Collaborative Research and Development. Invest Radiol 2015;50(9):645-656.

Proc. Intl. Soc. Mag. Reson. Med. 26 (2018)