Paul Lauterbur and the Invention of MRI

by M. Joan Dawson

  For Chuck Dulcey, nineteen years old, it was an entrance into the world of adult responsibility and of science. He recalls, “I had a little office in the computing center, and everyone came to me from all over the department for help with their computations.” He was given the opportunity to attend the groundbreaking Brookhaven meeting on image reconstruction: “It was active—a lot of debate. There was a sense that it was a pivotal moment. Maybe I wasn’t able to catch everything, but I could catch the excitement.” Paul presented their paper on MR image reconstruction using projections.20

  Other early students to appear were also undergraduates, Cliff Weisel and Mike Feiler. Feiler, who came to the lab late in 1973, was an interesting personality. Those who knew him remembered him as a nerdy, overweight recluse in high school. He decided to remake himself, began a regimen of diet and exercise, and got to the point where, they say, he couldn’t go to the mailbox and back without a pretty girl hanging on each arm. He had the gift of gab and was the social center of the lab. Chuck Dulcey enthused, “We went out to lunch, just to be around him.” In the MRI laboratory, Mike focused mainly on mice. He was always injecting mice, growing tumors in mice, imaging mice.

  Waylon House arrived early in 1974 and did almost nothing but work: “If you ever go to the lab, there’s a spot on the floor where the pattern is worn off the linoleum. That’s where I stood for months on end making mechanical adjustments on that magnet.” He ate all his meals at a Pancake House. He drank coffee by the gallon, and his colleagues forever chided him for leaving his cup behind. (Years later, when he finished his fellowship, he was presented with a coffee cup with a chain on it.) Waylon was responsible for much of the technical work: “I was kind of the executor.” He explored a technique called slice selection for two-dimensional images that were not too thick; various methods of slice selection remain important in MRI today. He and Paul used a rotating magnet, and a standing field gradient to select the thin slices.21

  When Paul conceived of MR image reconstruction, he realized immediately that he had found a broadly applicable principle, one not limited to NMR. Reginald Dias took responsibility for showing experimentally that MRI has wider applicability. He made the first electron spin resonance (ESR) image in 1973 or 1974. He cut out filter paper in the shape of an R, soaked it in a solution of Fremey’s salt,22 and obtained a nice image, thereby claiming the work forever as his own. Typically, Paul and Reginald did not have access to the best equipment of the day, and the work was done on an old used magnet (the Varian E4). The cooling coils were clogged with mineral deposits, which they didn’t dare try to remove for fear of puncturing the coils. The magnet could therefore not be brought up to the high magnetic fields all MR scientists love, because the attendant heat could not be dissipated. But because ESR is much more sensitive than NMR, they were able to detect signals when the magnet was used at the low fields it could sustain. The image was exceedingly clear and left no doubt that ESR imaging was not just plausible but possible.

  Unfortunately, the work was never published, and the names Dias and Lauterbur are not much associated with ESR imaging. But Paul did talk about this and other ESR work at scientific meetings, and other scientists took up the challenge. In particular, Jim Hyde, first of Varian and then of the National EPR Center in Wisconsin, Larry Berliner at Ohio, and Hal Swartz at Wisconsin saw the possibilities, and independently they advanced biological ESR a long way.

  Other workers in Paul’s lab included David Kramer, a graduate student who went on to a career with Toshiba America in San Francisco. Ching-Nien Chen arrived in 1975, a graduate student who had earned a master’s degree in crystallography before coming to the United States to obtain a PhD in chemistry with Paul as her mentor. Ching-Nien worked slowly and methodically and was a little older than the typical graduate student during her tenure there. She went on to have a career at the National Institutes of Health (and was the first of Paul’s former students to retire before him).

  Joe Frank, who like Ching-Nien remained a lifelong personal friend of Paul’s, put himself through college on a limited budget and was determined to complete his studies in three years to limit the financial strain. He had spent his first year at Union College, in Schenectady, New York, and then transferred to Stony Brook. While recording the work of the group photographically, Joe got interested in the work scientifically and officially joined the group as an undergraduate research assistant.

  When he worked with Paul, Joe concentrated on the biological experiments and their image processing. At that time the images were calculated by hand-feeding the raw data into a mainframe computer using magnetic cards or tapes, the system Chuck Dulcey had set up. The calculated image would come out as an array of pixels, each with homogeneous calculated signal intensity. To turn this output into an image, the students reentered the data into the computer, using symbols to represent various intensities. To make the image, contours were drawn by hand around the biggest numbers. Regions of low intensity were represented by blank spaces, those with slightly more intensity by periods, and so on, going on to semicolons and then to various letters of the alphabet, with “i,” for example, representing less intensity than “w” or “m.” “W” and “m” are the blackest letters in the alphabet, so from these letters the students went to overprinting; using a combination of four letters they could get complete filling of the pixel area.

  The students debated a lot about which letters to use on this eccentric gray scale. Chuck Dulcey, of course, developed a program for this, which he named PICTG, for picture generator. PICTG produced an image, but one with unsatisfying sharp demarcations from one intensity region to another. Joe, as photographer, had the job of optically blurring the image so it would look more like an ordinary photograph. This he did using a special optical diffusion filter: used pantyhose.

  The group produced color images as well. Paul favored these because a gray scale alone can’t display all the image characteristics. The intensity of an MR image depends both on the concentration of protons giving rise to the NMR signal and on the decay rate, the rate at which the signal disappears. In acquiring an image, the decay rate was combined with the concentration data to maximize contrast. So in a gray-scale image a dark spot could be due either to a large concentration of protons or to protons with particularly slow signal decay. Paul always found this vexing. So he decided to produce images in which intensity was coded by gray scale and signal decay by color—red for fast decay and blue for slow decay, for example. These were recorded on transparencies in red or blue, overlaid on the gray-scale image, and photographed. This was a low-tech and inexpensive but time-consuming way of solving the problem. Unfortunately, only one of these primitive artworks was ever published.23 Many journals could not accept color images, and those that could charged the authors so much to print them that Paul couldn’t afford to do it.

  Marcellino Bernardo was a later recruit to the group, joining in the 1980s. Marcellino, Paul remembered clearly, had a rapport with dicey machinery. At one point the tape drive, on which all data collection depended, went down. Paul was sure it was beyond repair, and there was general consternation in the lab about how much time (and money) would be lost finding a replacement. When the manufacturers were called about a repair, they simply gave up. Marcellino started working on it in the evening. He worked through the night, and by morning it was functioning perfectly. “It was like making the dead rise up and walk,” Paul said.

  Ed Heidelberger signed on as an undergraduate researcher and became so excited that he recruited his sister, Ruth. Paul particularly admired Ed because, not once but twice, he succeeded in completing experiments that more experienced graduate and postdoctoral associates were not able to pull off. Ed was first author of a paper titled “Aspects of Cardiac Diagnosis Using Synchronized NMR Imaging.”24 A particular problem with MRI of the heart and lungs is that breathing and heartbeat cause the structures of the thorax to move periodically, making the images fuzzy. Ed’s paper showed a way to surmount part
of the problem by timing the collection of the NMR data to the heartbeat. This was awfully important, to pin the heart image down to a particular spot in its cycle. Paul said of Ed, “Not knowing it couldn’t be done, he went ahead and did it.” The method was expanded to include the motions of breathing, and its descendants are still used today.

  Renaissance Man

  As Paul began, so he continued. He was chemist, physicist, mathematician, computer programmer, and tinkerer. He thought of MRI, brought that thought to reality, and developed many of its applications. Between the years 1972, with his seminal concept of MRI, and 2000, Paul was author of 278 publications on MRI. He pioneered nearly all aspects of MRI, aspects that have now developed into independent fields of study, to the extent that practitioners in one subfield of MRI often can’t understand the language of practitioners in another. Even now, and I suspect well into the future, people who undertake the development of certain new techniques find that the germ of these ideas was explored by Paul Lauterbur many years ago. (Or they may not find it, and so don’t acknowledge the earlier work.) People sometimes said that the wide scope of his pioneering work was discouraging. They couldn’t do anything truly new; almost everywhere, Paul had gone before.

  As Paul continued his research, he invited everyone to visit his laboratory for observation and study. And people came, an unusual number of them—from industry, academia, and government laboratories, foreign and domestic. The students were somewhat dismayed by all the interruption these visitors caused. One visiting scientist managed to break the instrumentation. In 1972, Paul began supplying a bibliography of MRI papers to all, and helped organize meetings to compare methods and results. The bibliographies continued until 1975, when MRI simply got too big for this project (and Paul was overwhelmingly busy!). As Paul had hoped, other laboratories began making more and more contributions to MRI. As the depth and breadth of applications grew, both large and small companies began to see opportunities, and within less than ten years commercial diagnostic instruments came onto the market. Competitive pressures among physicians, hospitals, and industrial interests helped spur its explosive growth.

  Paul’s demonstrations of the new technique’s application to specific medical problems were interspersed with his further theoretical, chemical, and physical studies elucidating its characteristics. Paul was, for example, already considering problems associated with head imaging in 1972. Imaging pathophysiology of the brain was a serious problem at the time. The best available method, CT scanning, suffered severely from distortions due to the skull. Theoretically, since the skull is simply opaque to MR signals, MRI could do a better job. On February 12, 1972, Paul wrote out on white lined paper a possible sequence of “magnetographic” studies in the diagnosis of a brain tumor, and sketched what these images would look like. (This was the only time he used the term “magnetography” because the name was already in use for instruments used to map magnetic fields of the Earth or Solar System.)

  The period of research from 1971, when Paul first thought of imaging, until 1985, when he left Stony Brook, was an exciting time and one of the most productive in Paul’s career. But not everyone found it as exhilarating as Paul and his students did. Waylon House had some negative things to say to a reporter. “Lauterbur was the kind of guy given to lots of strokes of genius—about one out of ten would work.” And “He certainly had an ego problem. . . . He was a man who cared about reputation. He wanted accolades.” Randy Lauffer interviewed with Paul for a postdoctoral position and had this to say: “I visited Prof. Lauterbur in 1982 or 1983 regarding a post-doctoral position at SUNY–Stony Brook. I had read the early work with manganese and it appeared that a new class of pharmaceutical was about to be developed. Prof. Lauterbur was enthusiastic, engaging and brilliant. But I came away from my visit very disappointed. It was difficult to discern any plan from him regarding what might be done or should be done to develop useful agents. It was all concepts. Most of the time in the lab was spent dealing with the vagaries of home-built imaging devices that were frequently inoperable. In addition there were no colleagues available to discuss physiological models or pharmaceutical development. By contrast, the Massachusetts General Hospital in Boston had entire teams and countless collaborators devoted to every separate aspect of imaging development and nascent imaging agents.”25 So Randy took Paul’s vision to MIT.

  Paul’s first images made a big impression on the NMR community, but it was a lemon from Raymond Andrew’s laboratory that raised consciousness in the wider scientific world. The lemon was simply beautiful, like a work of art. Nature used it as the cover image of a December 1977 issue. Like Paul, Waldo Hinshaw and others in Andrew’s lab were interested in the medical applications of MRI, and they also started with simple test objects, which yielded results that were easily interpreted. In the same paper with the lemon, they published an image of a human wrist.26 The wrist image was especially beautiful and showed in no uncertain terms that MRI had a place in medical diagnostics, bringing many, many scientists and physicians into the field. “They were producing the best images by far at the time,” Paul said.

  Figure 7.3

  One of the first clear MR images, the Nottingham group’s lemon, used as a December 1977 Nature cover image. Reproduced by permission.

  A note on human frailty: now decades later, scores of people think that Paul made that first spectacular wrist image. People mentioned to him their marvel at the early MR image of a human wrist. How beautiful it was. How important it was. Even when Paul told them to thank Waldo Hinshaw, they often said, “No—you made that image, I remember it clearly.” Waldo, who with his colleagues was the first after Paul to publish an MR image, took all of this with good grace.

  8

  Baby Grows Up

  To save a life is as if you saved the world.

  —Talmud

  The British, Raymond Andrew’s and Peter Mansfield’s laboratories at Nottingham and then John Mallard’s at Aberdeen, picked up first on Paul’s new ideas about imaging with NMR and were largely responsible for its development in the early years. Things happened more or less this way. Peter Mansfield entered the medical imaging business from an entirely different direction than Paul Lauterbur. He was interested in the atomic structure of solids. In 1972, while sitting in the Physics Department tearoom with two of his colleagues, Al Garroway and Peter Grannell, Mansfield realized that he could get atomic diffraction effects from which the atomic structure of crystals could be worked out if he applied a magnetic gradient across the sample. This is analogous to the visible light diffraction caused by regularly repeating objects. They did calculations, ran experiments, and repeated everything many times over. They wrote up the results for presentation at the First Specialized Colloque Ampère in Krakow, Poland, in September 1973. At the same time they submitted a more formal publication to the Journal of Physics C, which appeared in November 1973.

  When Peter presented his new work, John Waugh, with whom he had a sometimes stormy and acrimonious rivalry, asked how his work was related to the new work of Paul Lauterbur. Peter had never heard of Paul Lauterbur and certainly knew nothing about his new work on imaging of liquid specimens, which had been published recently in Nature. On returning to Nottingham he looked up Paul’s work and saw that although the goals were completely different, their use of magnetic field gradients was similar. He saw the importance of Paul’s work and saw that imaging in biology and medicine was a piece of cake in comparison to the work he was trying to do in solids. He decided to give up solids—much too hard, leave them for future studies—and be the first on the MRI bandwagon.

  Every MRI method, of course, has its advantages and disadvantages. The most important disadvantages of Paul’s projection reconstruction method at that time were speed and sampling uniformity; the plane of a two-dimensional image was also coarsely defined.1 Paul emphasized isotropic three-dimensional imaging and the use of back projection to deconvolute the raw data. This is the most efficient and natural way of making pi
ctures of our three-dimensional selves. For many years, Paul’s was the only practical method of making three-dimensional images. But it could take Paul several minutes to more than an hour to obtain an image. Patients will stay still for about fifteen minutes. Paul knew that soon enough, computers would become fast and powerful enough for the necessary imaging speed, and he didn’t see getting the imaging time down as the most important development to focus on at the time. Peter and other pioneers did. While in three-dimensional MR imaging, back projection has the time advantage, in two-dimensional imaging other techniques are faster. And the computers of the time were not big and fast enough to handle three-dimensional data sets.

  The problem of sampling uniformity is more complex. In back-projection reconstruction, the sampling is radial from the central or zero point. This means that the outer parts of the specimen are not sampled as well as the inner parts, and the zero point is not sampled at all, leaving a bright white spot in the middle of every image. There are various ways of dealing with this artifact in projection reconstruction, but there are also ways to do much more uniform sampling using other methods. Peter Mansfield introduced into the field a useful notation, familiarly known as k-space. K-space is a temporary virtual space enclosing the phase and frequency of imaging data, and it functions to simplify their conceptualization. K-space is covariant with actual physical space, so that k and physical space are interconvertible. The observed signals can be described in a much simpler way in k-space than in physical space, and this simplicity has aided the development of many alternative methods of sampling imaging data.