The development and widespread clinical utilization of MR scanners during the 1980s represented a technical tour de force. Engineering advances were combined with newly developed scientific and clinical understanding and with large financial investments to produce a completely new modality for the field of diagnostic imaging. It was possible to control the data acquisition process and to compute the resulting images in a reasonable time because of tremendous advances in computer technology. To avoid confusion with the field of nuclear medicine and also to avoid the negative connotations often associated with the word nuclear, the phrases nuclear magnetic resonance and NMR, which are commonplace in physics and chemistry, were not taken over into clinical applications. Instead, the presumably more agreeable phrases magnetic resonance imaging and MRI became universally accepted names for this clinical modality. Despite the inherent complexity of NMR, the technology was rapidly disseminated into the practice of medicine once human-size systems became generally available. Although precise numbers are not available, marketing studies suggest that, as of 2014, more than 65,000,000 clinical examinations are being performed worldwide each year. This indicates that more than 850,000,000 diagnostic MR studies were performed between the introduction of clinical MRI in the early 1980s and the end of 2014.

The impact of MRI on clinical medicine is often cited as an example of the benefits to the society of basic research in physics and chemistry. Table 1.1 lists important scientific and technical advances that were crucial to the eventual development of MRI scanners. Also listed are the names of some of the investigators and approximate dates for these developments. It is, of course, not possible to acknowledge in a single table all of the contributors to the development of MRI. The purpose of the table is to provide orientation concerning the large number of basic science advances that were required before MRI could be developed and to emphasize how recently much of this basic understanding has been achieved. It is interesting to note that in the 1980s many elderly patients who were scanned by MRI had been born before the magnetic properties, or even the existence, of atomic nuclei were known.

In 2001, a group of 225 physicians were surveyed and selected MRI, along with computed tomography, as the most important medical innovation in terms of advancing patient care in the previous 25 years (39). These results were particularly significant because (i) the surveyed physicians were not radiologists, who would logically be expected to be impressed by techniques that had changed their practices in major ways, but, rather, general internists, who ranked the innovations based on their impact on patient management, and (ii) the 25 years over which the innovations were evaluated had seen the introduction of many revolutionary medical technologies such as coronary artery bypass surgery, new treatments for depression, and so on.

The early history of MRI has generated a degree of controversy regarding the priority of various discoveries and inventions and the ultimate capabilities of the technique (40,41). In 2003, many years after their original publications, the Nobel Prize in Physiology or Medicine was jointly awarded to Paul Lauterbur and Sir Peter Mansfield “for their discoveries concerning magnetic resonance imaging” (42,43,44,45,46,47).

The official metric unit for magnetic field strength in the units of the Système International (SI) is the tesla, and the official abbreviation for this unit is capital “T” (48,49). In addition to being officially sanctioned, the tesla is of a convenient size to specify the strengths of main magnetic fields used in MRI. However, another unit of magnetic field strength, the gauss (abbreviated “G”), although not an official SI unit, has had a long historical usage and is still widely used by physicists and chemists; it is of a convenient size to describe the smaller fields (e.g., gradient fields) associated with MR scanners. The two units have an easy-to-remember relationship: 1 T is equivalent to 10,000 G. In this chapter, we use whichever of these units are more convenient for the magnetic field under discussion.

Magnetism was originally noticed more than 2,500 years ago as a property of certain rocks and minerals such as lodestone, which is now known to be the iron oxide (Fe₃O₄) and is called magnetite by geologists. Based on experiments with lodestones performed around the year 1600, the English physician William Gilbert proposed that Earth is a huge magnet and that it produces what would now be described as a dipole field. Although they are generally unaware of it, everyone on Earth is continually immersed in this field.

In 1820, the Danish physicist and chemist Hans Christian Oersted discovered that magnetic fields are also produced by electric currents flowing in wires. By the 1920s, modern atomic theory and quantum mechanics were developed to the point that the magnetic properties of lodestone and other magnetic materials could be described in terms of the motion of electrons within them. To achieve agreement with experiment, it was necessary to consider in addition to the orbital and translational motion of the electrons a further degree of freedom known as electron spin, which can be roughly visualized as a spinning motion of the electrons around their axes. In the presence of a steady magnetic field, the electron spins undergo a motion called precession that is analogous to the motion of a spinning top in a gravitational field. The spin axis precesses around the direction of the field at a fixed rate that is proportional to the strength of the applied field. This precessional motion repeats at a rate known as the Larmor frequency. All of the magnetic properties of materials and electric circuits can thus be explained in terms of the motion of electrons.

Also, in the 1920s, certain subtle effects in atomic spectra were noted that could be explained by associating a spin degree of freedom with some atomic nuclei. This led to the development of the field of nuclear magnetism (50,51). The behavior of nuclear spins is similar to that of electron spins, but the magnitudes of their effects are very much smaller than those of electrons, and it was initially doubted whether direct macroscopic effects of nuclear magnetism would ever be observed. As mentioned previously, in 1936, Gorter proposed a resonance method as a possible means of detecting nuclear magnetic effects. His protocol, called NMR, consisted in exposing nuclear spins to a strong magnetic field for a time long enough to align many of them with this field and then exciting them by use of an RF field at right angles to the initial field. This methodology was eventually shown to work and is the basic working principle for modern MRI devices.

Nuclei of many of the chemical elements can be studied using NMR, but far and away the most important for medical applications is the simplest atomic nucleus, the proton, which is the nucleus of atomic hydrogen. The magnetic moment of the proton is 658 times smaller than that of the electron, and this largely accounts for the difficulty in demonstrating macroscopic proton nuclear magnetism without the use of specialized equipment. The Larmor frequency for electrons and for all nuclear spins is linearly proportional to the applied magnetic field strength. For unshielded protons at 1 T, it is 42.577482 MHz. Whole-body clinical MRI studies have now been reported over at least the range from 0.02 to 9.4 T (200 to 94,000 G). This corresponds to a variation in Larmor frequency from 0.852 to 400 MHz. Therefore, the range of field strengths useful for medical imaging has been demonstrated to vary by a factor of at least 470. By using small magnets, much stronger fields can be obtained, and MRI studies of small animals and pathology specimens have been reported up to fields of at least 21.1 T, corresponding to a frequency of almost 900 MHz.

The human body contains an enormous number of hydrogen atoms, particularly in its water and lipid components. The proton spins can be manipulated by applied magnetic fields generated by electric currents in coils located outside the body, and signals produced by the motion (precession) of these spins can be detected using receiver coils that are also located outside the body. It turns out that, at the levels required for MRI, these applied electromagnetic fields have very little measurable influence on human tissues other than those on the nuclear spins. Some exceptions that occur at very powerful field levels or in the presence of magnetic foreign bodies will be discussed later. The remarkable fact that the elaborate electromagnetic activity associated with MRI couples strongly to the proton spins throughout the body, while the other components of human tissues are to a high degree transparent to these fields is the physical basis of the highly noninvasive character of this imaging modality.

The strength of Earth’s magnetic field is roughly 1/2 G, or 0.00005 T, and the proton Larmor frequency in Earth’s magnetic field is only about 2.1 kHz. The degree of nuclear spin magnetization in such a feeble field is far too small to result in an observable NMR signal without the use of special preparation techniques, and, as a result, Earth’s field is not useful in clinical MRI. The NMR signal is proportional to both the degree of spin magnetization and the rate of Larmor precession. In turn, these factors are both directly proportional to the magnetic field strength, B_o. Therefore, the NMR signal strength increases as the square of B_o. However, patient-generated electrical noise increases approximately linearly with frequency. This results in a signal-to-noise ratio (SNR) that increases, in strong fields, roughly in direct proportion to B_o, and this gives an SNR advantage to higher field scanners. This advantage must be balanced, of course, with other factors, such as cost and size, in deciding on the B_o field to use for a given scanner.

Superconductivity was discovered as an exotic property of certain metals at very low temperatures in 1911 and subsequently became a major research topic of theoretical and experimental solid-state scientists. The most remarkable property of these materials is that they have absolutely no electrical resistance when cooled below their superconducting transition temperatures. However, this temperature is so low that liquid helium is required to cool the metals to the point where superconductivity appears. It is natural to try to build powerful magnets from any material with zero electrical resistance. Unfortunately, for the superconductors known before about 1950 (the type I superconductors such as lead, tin, and mercury), once the current density exceeds a rather low critical value, the superconductivity vanishes and the resistance returns. Magnets made from these materials are, therefore, limited to producing disappointingly weak magnetic fields—a small fraction of a tesla at most. This left superconductivity as a fascinating scientific phenomenon but with very little commercial significance. However, the discovery of a new class of high-field superconducting alloys (type II superconductors) in the 1950s opened the way to producing much more powerful magnets. After a great deal of metallurgical experimentation on hundreds or thousands of materials carried out in many laboratories, Berlincourt and Hake announced in 1962 (52) that an alloy of niobium and titanium could carry superconducting currents at densities 30 times the current densities that were feasible in copper-wound resistive magnets. Moreover, it turned out that the Nb–Ti alloys had the very desirable metallurgical property of high ductility and could be readily formed into superconducting wires thousands of meters long. These wires could then be wound into superconducting coils capable of producing magnetic fields of several tesla. It has been emphasized that, although many materials with superior superconducting properties (superconducting transition temperature and critical current density) than Nb–Ti have been found, these materials have generally been brittle, expensive, and hard to fabricate into coils (53). Thus, the Nb–Ti alloys have dominated the superconducting magnet business to the present time because of its ductility which provides ease of fabrication and suitability for bulk production of reliable, cost-effective superconducting wire. However, except for customized magnets designed for research purposes, there was no large-scale commercial application of superconductivity, including these type II materials, prior to the advent of whole-body MRI in the early 1980s.

As the great clinical potential of MRI became appreciated during the early 1980s reservations regarding the technical and cost challenges associated with whole-body superconducting magnets rapidly gave way to enthusiasm for the advantages they offered. Superconducting magnets are designed to achieve a prescribed maximum field strength and, in principle, can be operated to produce any field strength up to this design value. In practice, however, they are almost always operated at a fixed, constant field, near the upper limit of their capabilities. Because of the superconducting properties of the magnet coils, once the desired field strength is reached and the superconducting circuit is closed, the power supply can be disconnected, and the magnet
will operate in the persistent mode. That is, the full field strength can be maintained, without additional power input, for months or years. Typical performance specifications for these magnets are that, at the magnet center, the field is spatially uniform (i.e., homogeneous) to better than 10 parts per million (ppm) over a sphere 40 to 50 cm in diameter and that the drift of the field (i.e., temporal stability) is less than 0.1 ppm per hour. Generally speaking, the cost and manufacturing difficulty of building these magnets increase rapidly with the prescribed field strength and the diameter of the warm bore (the central opening where the patient is located during imaging). Although largely out of sight, the interior components of a superconducting whole-body magnet are remarkable high-technology devices operating with very tight tolerances, at extremely low temperatures and supporting enormous magnetic forces (54,55,56).

As mentioned above, whole-body superconducting magnets have been used in MRI clinical applications and research over a wide range of field strengths. However, for technical and historical reasons, the majority of superconducting whole-body magnets placed into clinical practice between the mid-1980s and the present operate at 1.5 T. Since approximately the year 2000 there has been a significant increase in the clinical use of whole-body scanners operating at 3 T. This is done to take advantage of the improved SNRs and improved resolution available at higher field strengths. Still more recently, whole-body systems operating at 7 T have been placed in a number of research centers. At the present time, there are roughly 25,000 1.5-T and 5,000 whole-body 3-T systems in place worldwide. 1.5-T magnets still account for about three-fourths of the new installations but the fraction of 3-T systems in new installations is increasing with time.

Figure 1.2 is a cutaway drawing illustrating many of the features of modern MRI superconducting magnets. In this example, the field is produced by an intense current flowing through six coils that are connected in series. The structure that contains the cryogenically cooled coils is called the cryostat. A typical superconducting magnet has a total of about 5 to 7 main coils on coil forms of about 0.65 m radius surrounded by two shielding counter coils on a larger radius. The total length of the superconducting wire is on the order of 120 km or 75 miles. This total length of wire must be constructed without any interruption of its superconducting properties in order to achieve persistent operation and adequate temporal stability but the total wire length wire is made up of 10 to 20 shorter segments. The technical problem of joining shorter segments of wire while maintaining the superconducting properties at the linkages was one of the major challenges in the development of superconducting magnets. The total weight of the mass at helium temperature of a modern 3-T magnet (superconducting wire and its supporting structures) is on the order of 9,000 lb. The warm bore opening of the main magnet is normally in the range from 85 to 100 cm. After allowance is made to accommodate the gradient coils and RF transmit/receive coils, the diameter of the space available for the patient and the patient support system is on the order of 60 to 70 cm.

The superconducting coils are immersed in a bath of liquid helium at 4.2 K or somewhat colder to cool the niobium–titanium alloy well below its superconducting transition temperature, which is about 9.2 K. In the original superconducting magnet designs, the transition from liquid helium temperatures to room temperature (298 K) was made through an intermediate liquid nitrogen stage at 77 K. With improved technology such as the use of cryogenic refrigerators, it is now no longer necessary to include an intermediate stage at liquid nitrogen temperature. Nonetheless, there is an inexorable boiling of the cryogenic liquids due to thermal energy
entering the helium bath from the outside. This necessitates periodic refilling of the cryogenic fluids. These refills were originally required every few weeks. Now it is possible to go for several months, and even longer in many cases, between helium refills. Figures 1.3 and 1.4 show, respectively, the appearance of a 3-T and a 7-T magnet.

To maintain the superconducting state the conductor must be kept below its transition temperature. If even a small region of the wire is heated above this temperature, it begins to dissipate heat and the temperature increases still further. The result can be a self-propagating process leading to a magnet quench, wherein the entire stored energy in the magnetic field is converted into heat and the metal returns to its normal resistive state. This raises the temperature of the liquid helium above its boiling point and it quickly evaporates as the magnetic field collapses. Magnets are designed to sense the onset of a quench and to rapidly activate heaters that spread the energy deposition throughout the coils. Otherwise, the deposition of the total stored energy in a single location could melt the wire locally and destroy the magnet. In a controlled quench, the field will decrease from its initial value to nearly zero in about 1 minute. Provision must be made to permit the safe removal of the helium gas that is produced during a quench. This gas is very cold and direct contact with it should be avoided. Also by displacing room oxygen, a quench, if the gas is not properly vented, represents a suffocation risk to people nearby. All the wires in a superconducting magnet are subject to intense forces that can cause slight movement and local frictional heating that may be sufficient to initiate a quench. This possibility is minimized by a process called training. This is one of the final steps in magnet manufacture and involves cycling the magnet to fields somewhat above its planned operating field to assure stable operation at the rated field. Quenches are also possible if a cryostat vacuum fails or if necessary refills of the cryogenic liquids are not made on time. Superconducting magnets are supplied with an emergency switch system that permits a deliberate, but nondestructive, quench in the event that someone has become trapped against or inside the magnet by a ferromagnetic object, such as an oxygen bottle.

In recent years, both the cost and availability of liquid helium have become somewhat problematic. Helium is generally collected as part of natural gas extraction and is produced from radioactive decay of neutron-rich materials like uranium. Since helium is much lighter than nitrogen (the predominant gas in the Earth’s atmosphere) and is generally nonreactive chemically, once released as a gas it floats away and is lost. Three technologies are generally considered important candidates for reducing helium consumption. First, “zero boiloff” magnets incorporate recondensing compressors (often called “cold heads”) that rely on the Gifford–McMahon cycle to extract heat from the helium gas and recondense it to a liquid, maintaining the low temperature for the superconductor. Human-sized magnets with field strengths up to 7 T have been produced using recondensing compressors. Second, “cryogenless” magnets rely on a large mass of metal that, once cold, can maintain a low temperature. While these systems can operate with very little helium, care must be taken that, should a quench occur, heat will be extracted from the superconducting wire quickly enough to prevent damage to the wire. While few magnets to date have been built using this technology, it is likely to become an important technology particularly at clinical field strengths (1.5 and 3 T). Third, more exotic superconducting materials, such as MgB₂, are able to maintain a superconducting state at a higher temperature, permitting the use of more readily available cryogens. These materials are generally less ductile and far more expensive than Nb–Ti alloys, limiting their applicability for the foreseeable future.

There are inevitable imperfections in the coil locations within a manufactured magnet. These imperfections lead to the presence of error fields which reduce the uniformity of the field within the field of view (FOV) below what it would have been if the coils could have been wound and positioned exactly as in the ideal design.

Various shimming techniques are used to tweak the final magnetic field to correct for these inevitable manufacturing errors and for additional errors that result from the magnetization of magnetic materials, such as building girders, in the
vicinity of the magnet. Passive shimming, somewhat analogous to tuning a piano, is accomplished by positioning small ferromagnetic strips at specific locations within slots around the bore of the magnet (Fig. 1.2) to counteract field inhomogeneities once they have been mapped and characterized (57).

Active shimming is an additional method for improving magnet homogeneity. This technique makes use of sets of specially designed coils placed within the magnet each of which is designed with a specific symmetry to produce a field closely approximating an idealized field referred to mathematically as a spherical harmonic. If superconducting shim coils are used as shown in Figure 1.2, they are located, like the main coils, within the cryostat and are maintained at liquid helium temperature. To make it possible to change the current in superconducting shim coils, it is necessary to provide current leads to an external power supply. These leads can be switched from one shim coil to another and a set of currents can be built up in the array of coils to cancel out each of the harmonic fields for which a shim coil has been provided. Superconducting switches are used to restore each coil to the persistent mode once the power supply has brought the shim current to the desired value. An alternative or supplement to the use of superconducting shim coils is to provide a set of resistive shim coils which can be located within the room temperature bore of the magnet adjacent to the gradient coils (not shown in Figure 1.2). Each of these coils also is designed to produce primarily a single harmonic field but, in this case, each coil is permanently connected to its own power supply. This makes it relatively easy to reset the resistive shim coils as necessary.