A number of technical, historical, and policy-analytical accounts of the development of MRI have been written.[41] We do not need to replicate these accounts nor go into extensive detail for the purposes of this case. The following chronology is a convenient means of identifying the key milestones, the contributors, and the nature of their contribution. The chronology is the final piece of the background necessary for our discussion of NSF's role in MRI, organized around the intrinsic technologies that comprise the innovation.
By the 1960s, researchers in a number of diverse fields were using NMR spectroscopy in their work. Some of them directed the instrument at biological tissues. In fact, Felix Bloch stuck his finger into his apparatus and observed a strong signal because of his finger's high water content, but he did not pursue the observation. Most NMR users were analytical chemists interested in the composition of physical, rather than biological, materials. Early investigations of human tissue using NMR were summarized in 1956 by Odeblad, a Swedish clinician (Odeblad, 1956), whose team had examined human milk, eye tissues and fluids, and cervical mucus during the menstrual cycle (Chen and Hoult, 1989: 37). Apparently reflecting the work being done by Kudravcev, an NIH researcher, a 1960 progress report from the National Heart Institute noted that experiments with NMR might lead to "a scanning technique [that] will give coarse pictures (in intensity variation) of muscle, arteries and other structures." Kudravcev performed experiments in 1961 demonstrating that magnetic field gradients could be used to localize a particular element of interest, but these results were deemed a "curiosity" and never published (Chen and Hoult, 1989: 38). In 1968, Jackson and Langham obtained good signals from a whole rat that distinguished between fat and water (Jackson and Langham, 1968: 385).
By 1970, the way was paved for two key experiments that would lead to the realization of NMR imaging and shape the course of its commercial development. The first of these was Raymond Damadian's comparison, in 1970, between NMR signals from cancerous and normal tissues from six rats. His results suggested that cancerous tissue has a "lower degree of organization and less water structure than normal tissue." Damadian saw NMR as a means of detecting cancer in a noninvasive way; he published the results in Science in 1971 (Damadian, 1971). Subsequent efforts to replicate Damadian's work showed that the differences Damadian observed were not unique to cancerous tissue; similar differences in NMR readings were observed for noncancerous but diseased tissue, compared with normal tissue (Blume, 1992: 193). Nonetheless, his results sparked intense interest in the application of NMR to clinical medicine. (Damadian was at the time a member of the faculty of the State University of New York's Downstate Medical Center in Brooklyn.)
The credit for actually visualizing NMR output as an image generally is ascribed to Paul Lauterbur. In the early 1970s, Lauterbur was a member of the chemistry department at SUNY Stony Brook. Lauterbur, interested in ways of using NMR to investigate animal tissue in a noninvasive manner, recognized that by intentionally varying fields generated by so-called shim coils, normally used to compensate for variations in the primary NMR static field, he could introduce locational information into the NMR signal. He further realized that he could generate a two-dimensional image from a series of one-dimensional projections generated by rotating the orientation of the magnetic gradient relative to the sample. The results, depicting the image of two tubes of water, were published in Nature in 1973 (Lauterbur, 1973). Lauterbur anticipated in his paper the use of NMR for medical imaging and for studying malignant tumors.
John Mallard, a professor of medical physics at the University of Aberdeen, was also interested in possible links between cancer and the structure of cell water. He had been using electron spin resonance (ESR) rather than NMR in his experiments, but after reading Damadian's paper, he and colleagues built an NMR spectrometer and used it measure relaxation times from tissues (Blume, 1992: 193-194). At the same time, Peter Mansfield and a group in the physics department at the University of Nottingham published a paper indicating how NMR could be used to generate one-dimensional projections of proton density (Mansfield and Grannell, 1973).
The idea of medical imaging using NMR was met with skepticism by radiologists, who by this time were getting good images using CT scanning, and with indifference by most NMR researchers. Nonetheless, a small number of physicists and chemists, initially intrigued and desiring better image quality, and medical school faculty, driven by the prospect of a new way of detecting cancer, pursued the paths laid down. At Nottingham, physicists Raymond Andrew and Waldo Hinshaw (an American postdoc) heard Lauterbur speak and set out to improve on Lauterbur's imaging technique. By 1974, Hinshaw introduced time variation into the strength and direction of the gradient field as a way of avoiding the complex data processing that Lauterbur's method required. By using three intersecting gradient coils and varying their fields, Hinshaw could create a single point of zero gradient field strength that could be moved about in a sample electronically, and a filter used to detect the signal from that point only (Hinshaw, 1974).
Meanwhile, Andrew and Hinshaw's colleague at Nottingham, Mansfield, was using a different approach. He irradiated the entire sample with RF energy to destroy the net magnetization created by the static field, but to leave one element of the sample unaffected. The NMR-generating RF pulse was then applied, so that the signal received came from the selected element only. The theory and results were described in a 1976 paper (Mansfield, Maudsley, and Baines, 1976). According to Blume (1992: 199), Mansfield was interested primarily in cancer detection, whereas Andrew and Hinshaw saw MRI as a more general imaging technology. Nonetheless, both groups applied, successfully, to the Medical Research Council for support. Shortly thereafter, Andrew and Hinshaw developed a way to use two intersecting gradient coils to define a "sensitive line" rather than a point, thereby reducing the scan time and allowing for larger samples. By 1977, both groups of Nottingham physicists were producing striking images of human body parts: the cross-section of a wrist (Hinshaw, Bottomley, and Holland, 1977) and an image of a finger (Mansfield and Maudsley, 1977).
In the United States, Lauterbur's work focused on improving the quality of the image, whereas Damadian and others pursued cancer detection and medical imaging with NMR. Shortly after publishing his paper in Science, Damadian applied for a patent on an "Apparatus and Method for Detecting Cancer in Tissue," which was granted in 1974.[42] Damadian called his technique FONAR (Field Focusing Nuclear Magnetic Resonance) because, as described in the patent application, it focused both magnetic and RF fields on a single element in the sample. In 1976, Damadian published a second paper that showed an image through the thorax of a live mouse (Damadian, Minkoff, Goldsmith, Stanford, and Koutcher, 1976). At this point, Damadian and his colleagues, eager to move to whole-body scanning and lacking a major source of funding, built their own superconducting magnet large enough to accommodate a human body. In 1977, they used the machine to produce the image of a chest by moving the subject through 64 positions, requiring nearly 5 hours.[43] The results were published in the same year (Damadian, Goldsmith, and Minkoff, 1977).
Across the Atlantic in the United Kingdom, the Nottingham physicists were working toward improved image quality and toward whole-body scanning. In 1978, Hinshaw produced a good image of a rabbit's head, which took between 11 and 18 minutes to produce, depending on the resolution (Hinshaw et al., 1978). Mansfield did a cross section of his own abdomen, which required 40 minutes to produce (Mansfield et al., 1978). At about this time, the quality of the images being produced stimulated industrial as well as medical interest.
The link between NMR and MRI is exemplified by the development in 1975 by Kumar, Ernst, and colleagues at the Federal Institute of Technology in Zurich of the two-dimensional Fourier transform method (Kumar, Welti, and Ernst, 1975). This development was successfully implemented in imaging applications by members of the Aberdeen group (Edelstein, Hutchison, Johnson, and Redpath, 1980). This "spin-warp" technique, pioneered by Edelstein, proved to be the imaging technique embodied in most commercial MRI devices (Chen and Hoult, 1989: 40). Ernst, in his 1991 Nobel lecture, succinctly explains the technique and its origins:
The basic procedure for recording a 2D or 3D NMR image of an object is attributed to Paul Lauterbur. A magnetic field gradient, applied along different directions in space in a sequence of experiments, produces projections of the nuclear spin density of the object onto the direction of the gradient. From a sufficiently large set of such projections it is possible to reconstruct an image of the object, for example, by filtered backprojection in analogy to X-ray tomography. A different approach is directly related to 2D and 3D FT spectroscopy. Frequency encoding of the three spatial dimensions is achieved by a linear magnetic field gradient applied successively along three orthogonal directions for the durations t1, t2, and t3, respectively, in a pulse FT experiment. In full analogy to 3D spectroscopy, the time parameters t1 and t2 are incremented in regular intervals from experiment to experiment. The recorded signal ... is Fourier-transformed in three dimensions to produce a function ... which is equivalent to a 3D spatial image when the spatial information is decoded. ... In a further refinement, proposed by Edelstein et al., the time variables t1 and t2 are replaced by variable field gradient strengths gx and gy .... (Ernst, 1992)
As the CT scanner market began to stabilize around 1977 and the leading companies emerged, these same companies recognized the potential for MRI in the same market. EMI was one of the first; it began to develop an NMR scanner in 1976 using Lauterbur's projection method and a resistive magnet. GE[44], Philips, and Siemens began to develop their own MRI machines, and they had the resources to do it from scratch, although not without risk, since at this early stage there were substantial technological and market uncertainties (Blume, 1992: 209). Lauterbur visited Philips in 1977, and in 1979 the company assembled a team to develop a whole-body scanner. By 1981, they had succeeded (Blume, 1992: 214-215).
By 1981, Andrew's group at Nottingham, seeking additional support so they could initiate clinical trials, entered into a relationship with Picker, a U.S. firm, which was being acquired by GEC.[45] Similarly, Mallard and his group at Aberdeen looked to industry for support in 1980. Because the Medical Research Council had provided early support, patent rights to the machine would be held by the British National Research Development Corporation, which would license it and share royalties with the inventors. Mallard had been supported through gifts of instruments by the U.S. company Technicare, which was interested in positioning itself in the new technology.[46] Technicare wanted to hire Mallard and his entire group and move them to the United States, which was not acceptable, but the company did manage to attract Hinshaw, the native American, from Andrew's group at Nottingham. Mallard traveled to Japan, where he attracted interest and financial support from Asahi, which allowed him to collaborate with any British company of his choosing. Eventually, Mallard ended up forming his own company, M&D Technology Ltd. (Blume, 1992: 212).
In 1978, Damadian formed the FONAR Corporation to build MRI machines. He decided to use a permanent magnet rather than a superconducting one to keep the selling price low, and by the end of the year had produced one with a 500-gauss field (Blume, 1992: 219). FONAR produced the first commercial MRI scanner in March 1980. In 1982, FONAR and Damadian sued Johnson & Johnson and Technicare for infringing on his patent.[47]
While Mallard and Damadian were forming their own companies, existing medical instrument makers tapped MRI expertise in the universities through hiring, consulting arrangements, and linkages with university medical schools. Technicare hired Hinshaw in 1979; GE hired Paul Bottomley, another member of Andrew's group at Nottingham, and Edelstein from Aberdeen, in 1980. Diasonics, a maker of computerized ultrasound devices, moved into the spot left at the University of California, San Francisco's radiology department when Pfizer moved out.[48]
By 1983, eight companies had completed prototypes, had multiple clinical placements outside the factory, and had machines available for clinical placement:
General Electric, M&D Technology Ltd., and Toshiba had completed engineering models but had only limited clinical placements. Each of these 11 companies had by this time established a close collaborative relationship with a university or medical school. Eight other companies were in early stages of development (U.S. Congress, Office of Technology Assessment, 1984).
Medicare began paying for MRI scans in 1985, and in subsequent
years MRI became the fastest-growing medical imaging technology.
Although MRI and CT scanners are not strictly comparable in terms
of their applications, MRI was gaining market share at the expense
of CT scanners. By 1988, when our case ends, MRI was an established
technique in major hospitals worldwide. The approximately 1,300
units sold had gone to the major university medical schools and
larger hospitals. Expectations were that the next markets would
be smaller hospitals, which were less likely to pay the price
of large machines with superconducting magnets. Worldwide market
shares of the major suppliers were as follows:
Market Share (%) | |
| General Electric | 32 |
| Siemens | 20 |
| Diasonics | 14 |
| Picker | 9 |
| Philips | 8 |
| Other | 17 |
General Electric had bought Technicare from Johnson & Johnson in 1986. Technicare had not been profitable during the time it was owned by Johnson & Johnson; most of its customers replaced their machines with GE units within the next 2 years (Walter Robb, personal communication, October 8, 1996).
To examine possible linkages in technology research areas, another small-scale experiment using the ISI Research Front Database was performed for this case study. In this example, citation linkages between the individual specialty clusters comprising the database were used to produce a "map" that shows in graphic form the link between research on NMR/MRI imaging and NMR spectroscopy research. The results are shown in Figure 5.
In Figure 5, each small circle represents one of the specialty clusters in the 1990 Research Front Database, which were identified as representing research related to this case largely by searching on names of some of the prominent researchers that had been identified. The lines linking the specialties represent relationships defined by citation patterns in the papers indexed by ISI. The dashed lines represent the weakest links, the fine lines links of intermediate strength, and the heavier line joining the two networks of clusters a strong link. A mathematical program was used to reduce the multidimensional matrix of links to the two-dimensional map shown in the figure. In principle, the distance between the specialties is inversely related to the strength of the links between them, but the two-dimensional projection distorts some of these relationships.
The map resolves into two sets of interrelated specialties, showing a group of interacting research areas concentrated in imaging at the top and NMR spectroscopy
research areas at the bottom. There is a strong link between
one of the specialties at the top, which deals primarily with
blood-flow imaging of the brain and includes title words like
MR ANGIOGRAPHY, CEREBRAL BLOOD FLOW VELOCITY, and INVIVO NMR
DIFFUSION
SPECTROSCOPY, and one of the central specialties in the lower
network, which deals with C-13 NMR-SPECTRA and ISOTOPICALLY ENRICHED
PROTEINS. Although the link is strong in terms of citation patterns,
the absence of other links between the two networks has "pushed"
them apart in the mapping program. In terms of researchers, Ernst's
team in Zurich is a major factor in the lower network, while Lauterbur
and the Nottingham group are major players in the upper network.
The strong link between the two areas of research confirmed the
findings derived from interviews. One tentative conclusion from
this experiment was that the bibliometric mapping technique represents
a potentially helpful tool for bounding the research themes relevant
to an innovation.