With good funding for a start-up company and a good board, NMR Specialties appeared to do well for several years, although always, according to internal memos, fighting for profitability.15 Paul was certain that Yajko did not set out to be dishonest and deceive people. At the time bad practices started, Paul believed, Yajko was deceiving himself as much as he was deceiving others: “A little embezzlement because the horse didn’t come in and the debt could surely be made up on the next horse; lies upon lies until the whole thing got out of control.”
Luckily, there were board members who understood business and finance, one of whom, fulfilling his oversight responsibilities, found something wrong. This man was a lawyer who worked for Wells Fargo and who was brought in for his financial expertise. Going over the accounts receivable, he noticed an amount credited for an order that he knew had not been placed. He found that the company’s financial condition was not what the board, the bank, and the investors and customers had been led to believe. Things began to unravel. Yajko was crediting orders to accounts receivable for which he might reasonably expect a signed order soon, but had none in hand. In time, he was also crediting orders that had not even been discussed with the supposed customer. More: while he credited his company for these phony orders, he did not charge the accounts with the price of the raw materials with which to build these mythical machines. As if this were not enough, the board found that though the company was withholding federal taxes from the employees, it was paying no tax to the Internal Revenue Service.
When sales of ancillary NMR devices proved insufficient to support the company, NMR Specialties made the decision to begin building NMR spectrometers. There was a much larger market for spectrometers, but now NMR Specialties was in competition with real businesses, such as Westinghouse and Varian. Yajko dangled the possibility of a lucrative consulting fee for advice on building a high-field magnet to a Westinghouse engineer. Generally, companies have nondisclosure contracts with their employees, and this man may have been breaking the law by talking to NMR Specialties. In any case, the Westinghouse employee was danced to glory as Yajko tried to learn the plans for the Westinghouse spectrometer. The executives of NMR Specialties took him out to lunch, informal, briefcases left behind. While they were gone, Yajko had one of his own employees break into the visitor’s briefcase and copy the Westinghouse plan.
This was to be NMR Specialties high-field spectrometer. With the Westinghouse plan in hand, Yajko pretended to have a machine already built. In advertising his new line, Yajko handed around spectra that he said were test spectra from the new NMR Specialties machine. One copy was sent to Aksel Bothner-By at the Mellon Institute in Pittsburgh. Axel recognized the spectra; they had been taken from Rex Richards’s laboratory at Oxford University in England. Apparently Yajko didn’t realize he could be tripped up in this way. Aksel Bothner-By never forgave Paul Yajko, but it seems no action was taken against him.
Later it was learned that Yajko was lying to other customers as well, even to those on his board. Paul Lauterbur had ordered a spectrometer from NMR Specialties, one for biopolymer work, including a plan to study enzymes by replacing calcium, which had only a weak NMR signal, with cadmium, which had a very strong signal. When things began to hit the fan, employees admitted to Paul that they had been instructed to lie about the progress of its construction. This machine was never built. It turned out there was little match between pricing for the goods sold by NMR Specialties and their production price. In some cases, Yajko was signing contracts to deliver products at prices below the cost of the materials needed.
In May 1971, when NMR Specialties was about ten years old and bringing in about $500,000 a year, an extraordinary board meeting was hastily called, with the company’s banker, Herman J. Israel, invited as a guest. It was revealed that the company was effectively bankrupt. There were only two decisions that could be made: close the company down immediately and declare it bankrupt and dissolved or find someone willing to try to save it. The banker threatened to close the company that very day unless someone he trusted could be persuaded to take over as president, chairman of the board, and chief executive officer. Paul was the only academic on the board, the spring semester had just ended, and he had no summer salary. So he took the job.
Paul hadn’t a clue what he was getting into. He did his best to figure out what had gone on and to help steer the company to a soft landing. He said, “It was like trying to fly an airplane whose engines had stalled, a wing had come off and the fuselage was cracking up.” Yajko had raised money by selling stock to board members and relatives of his employees, so Paul and other members of the board hoped to salvage some of this for the innocent investors. Israel had a loan outstanding to NMR Specialties. He may have been embarrassed that he had promoted the company to the board of his bank and didn’t want to admit that he had been Yajko’s patsy. As more and more problems were revealed, Don Vickers, an NMR Specialties employee and Lauterbur’s lifelong friend, repeated, “You ain’t reached the bottom yet.” He was right.
One of the first things Paul did was to cancel a contract with Raymond Damadian on the grounds that the small, ailing company simply could not afford to build the magnet at so great a loss. Damadian was angry and thought that Paul was singling him out for bad treatment. He went to Israel to complain, asking that Paul be removed as CEO for incompetence. Israel, now knowing the truth of the state of NMR Specialties, showed little sympathy to Damadian. While Israel backed Paul on this one, another of Paul’s decisions infuriated the banker. Paul paid a bill from Westinghouse. The bill was legitimate, and Paul thought it was the right thing to do. “Israel said he knew some people in Akron who could do something about my knees.”
This brought to the foreground suspicions about Herman J. Israel’s business connections. As a part of winding down NMR Specialties, Paul was in discussion with “the local mafia” about selling the NMR Specialties building for a golf course and bar. He worked hard on the sale, but the buyers would not accept his price. Later, Israel was somehow able to sell the building to the same buyers at a much higher price.
It later turned out that Herman J. Israel, prominent banker and business associate of the governor, had used sham loans to buy bank stock and land in Ohio. In sentencing him to prison in 1977, the judge compared Israel’s activities to bank robbery.
And what about Paul Yajko? He tried his hand at another business, Larkton Scientific, and, apparently having learned nothing, did some time in jail for his shady practices. He was last seen selling apples at a roadside stand near his parents’ farm.
Paul Lauterbur was finally disentangled from NMR Specialties. He had more important things to do.
6
The First Fruitful Weeks
Science, if it is to flourish, must have no practical end in view.
—Albert Einstein
Life is so strange. It was because of the tortured history of NMR Specialties that Paul happened to be on hand to witness the experiments that raised in his mind the possibility of magnetic resonance imaging. Paul was always squeamish about everything medical and biological, everything that had to do with blood and other tissues. He was loath to go to doctors and totally intimidated by the idea that he might have to have an injection or to have blood drawn. So finding a way to do NMR studies noninvasively took on a special meaning for him.
But how? The NMR signal is governed by the simple Larmor equation, which holds that the frequency of the signal is proportional to the strength of the applied magnetic field. Paul’s initial insight into NMR imaging was one brilliant flash of an idea: gradients of magnetic field applied across the sample would localize any spatial position. This was the key. It is of some interest that this concept was simply waiting to be discovered since the days when Bloch and Purcell first found the phenomenon of magnetic resonance, and that the reaction of many in the NMR community, especially among physicists, was that something so simple could not be true. But the new idea, MRI, born on that bite of a Big Boy hamburger, c
hanged the direction of Paul’s research and of diagnostic medicine. The “thirty-year detour,” he called MRI, was “apparently a wise decision.”
Paul began to explore the potential of his new insight, or “moment—which didn’t feel like a really big moment—of realizing there was a principle that could be built on to this. After that it was a matter of thinking through each of the things that would have to be done if it were to be a practical technique, and that was spread over a period of at least several weeks.”1
Three conditions had to be met before MRI could become a useful diagnostic tool.
1. Once you have labeled separate regions using magnetic field gradients, how do you separate the signals into an image? “I realized that there was a trick that one could play that could make it possible.”
2. Would NMR be able to pick up small volumes of tissue with useful sensitivity? This was not at all obvious; excellent scientists in the field remained skeptical for years to come.
3. Could a large enough magnet be built? The magnets that chemists used to study 5 mm samples weighed a couple of tons. What would a system for studying humans look like?
Having satisfied himself that the answer was yes on all three counts, Paul reoriented almost all of his research in that direction. Paul had discovered that MR images could be made by placing a gradient of magnetic field across a sample. So now, as he said, “I deliberately turned the controls on the spectrometer in the wrong direction, so instead of making the field more uniform, I made it variable from point to point.” But he soon realized that conversion of the data thus obtained into an image, while possible, was very inefficient. He simplified the method by applying a linear gradient of magnetic field across the sample. Because the nuclei in the sample produce signals at different frequencies along this gradient, the value of each frequency serves as an address, a Zip Code, indicating where in the sample each signal has originated.
By the time Paul returned to Stony Brook for the fall 1971 teaching semester, he was sure that his new idea was practical. “And so I filled in everything. There was some mathematics that could be used to make the pictures, there was going to be enough signal to do the studies, you could build a big enough apparatus to actually put people in, and so that seemed like a useful thing to try to do.”2 During the autumn semester, he produced a cross-sectional image of an assembly of tubes. The results were a major breakthrough, but by now they were exactly what Paul expected. “I knew what I would see by that time,” he said. “It was not a matter of saying ‘Eureka! There’s an image.’ It was more like, Well, I must have done things right, because I got what I expected to see.”
Why had he been so certain?
Then I thought that what I didn’t know was whether the kind of radio signals that one could get from tissues from inside a person or an animal could possibly be turned into a picture, where you could say that this signal comes from here, this from here, this from here—this side you have your arm, leg, stomach, whatever. So I got to work thinking about that and using something I’d learned in a graduate course at Pitt—I thought of a variation of mathematics that might make it possible.3
(Paul grumbled that courses are useless but he sometimes learned something anyway.) The graduate course that had inspired Paul was one in quantum chemistry with the title “How to Solve Schrödinger’s Equations When You Don’t Know How to Solve Them.”
Turning Signals into Images
MRI requires big, complicated machinery. Many people have wondered what kind of computer was powerful enough to produce Paul’s first MR images; this was, after all, in the days when computers were not terribly smart. The answer surprises: Paul used no computer at all to make his first images. He attached a resistor to a wire and attached another wire to a capacitor, with a vacuum tube in between. Numbers were read and penciled onto a grid. (“In those days, digitizing really meant using your fingers,” Paul remarked.) The numbers were translated into pictures by hand—a sort of paint-by-numbers exercise. So much for the high-tech needs of MRI! Paul once told me, “Just as with lies, misconceptions can never be tracked down; they multiply. It’s like thinking Bach needed an iPod to compose his music.”
Now, how could he get there? Paul explained, “A much simpler idea came to me during the next several days. Sets of linear gradients oriented in different directions could uniquely encode each of a finite number of points representing the object, and I thought that an iterative comparison of the ‘projections’ thus generated with those from images, progressively refined to minimize the differences, could converge on a correct solution.”4 In iterative techniques a preliminary solution is obtained, and then the parameters are automatically changed so that differences between the observed and computed points are minimized, and a new solution (in MRI a new image) is computed.5
Paul asked around among mathematicians and computer scientists whether his “projection reconstruction” technique could really be made to work. About half of the people he asked said of course, it was simple, and the other half said no, it was impossible. One colleague said he didn’t even need a back-of-the-envelope calculation to know it was impossible. (Paul’s comment: “Things might go better for him if he’d had another envelope.”)
With so much conflicting advice, Paul went ahead and tried it by hand calculation and found that at least in simple cases, it worked. It was just simulated data—writing numbers in squares on graph paper and calculating the results, and then trying to get back the numbers he had put in. He wrote down small arrays of numbers representing the one-dimensional data that would be generated by linear magnetic field gradients and added them along the horizontal and vertical directions. The results were very encouraging. He then added data at 45 degrees and 135 degrees that could be generated similarly, and produced even better results.
About this time he talked to Dan Tycko, a computer scientist at Stony Brook, who reached into a pile of papers and pulled out a recent journal with a seminal publication by Richard Gordon and Gabor T. Herman, pioneers in image computation, containing an essentially identical algorithm for “reconstruction from projections.”6 “Is this what you are trying to do?” Dan asked Paul. “Damn! It was not only my idea but my algorithm.” It turned out that Paul had been one of several people who had reinvented this mathematical technique. He was both dismayed that his own idea was not original after all and relieved that this challenge to the development of MRI had already been overcome.
In fact, many researchers, in many different areas, were beginning to think about similar mathematical techniques for producing images. Projection reconstruction was just coming into all kinds of imaging technologies, but the programs were not yet fully developed. Ronald Bracewell, a highly regarded professor of engineering at Stanford, was doing something similar in radio astronomy; Gabor Herman and Richard Gordon were pursuing it in electron microscopy; and others were exploring it in different fields. The Atomic Energy Commission was using projection reconstruction to image underground explosions, using data from underground radiation detectors. Allan Cormack and Godfrey Hounsfield would receive the Nobel Prize in Physiology or Medicine in 1979 for introducing and making practical usefulness of “computer assisted tomography” (now known simply as computed tomography, or CT) using projection reconstruction.
The Brookhaven Conference: A Picture of Judy
Later, Paul was helped greatly by a highly successful conference at Brookhaven National Laboratories in July 1974 on techniques of three-dimensional reconstruction.7 The conference was, with Paul’s help, organized by Bob Marr, a Brookhaven staff member, who worked on the mathematics for reconstructing positron emission tomography (PET) images; Bob edited the book of proceedings. The conference brought together scientists and mathematicians from many different fields; all were working on the mathematics of image formation. Ronald Bracewell spoke about his mathematical approach. Aaron Klug described his work on reconstructing electron microscopic images of viruses. Gabor Herman and Richard Gordon, who had recently developed
a back-projection imaging technique called ART, were there. The latter, then at SUNY–Buffalo, was an especially loquacious presence. These two scientists demonstrated reconstruction from projections of a photograph of a woman named Judy. Judy was a great favorite among image processors at that time. She became a standard. Everyone could compare the quality of their techniques by comparing their images of Judy.
Three future Nobel laureates—Allan Cormack, Aaron Klug, and Paul Lauterbur—were in attendance. Paul’s paper was titled “Reconstruction in Zeugmatography: The Spatial Resolution of Magnetic Resonance Signals.” As examples, he showed an image of a bundle of twenty-one parallel glass tubes (so that the physical scientists and mathematicians in attendance could compare the image with the real specimen) and a section through the thorax of a live mouse to show that biological imaging is possible. He pointed out the ways in which the use of projection reconstruction of MRI is similar to that for other techniques, and the special problems singularly associated with using NMR for image formation. The journal Science titled its news article about the conference “Is ART Science?”
Paul described the conference as being like the six blind men and the elephant; they were all talking about apparently unrelated things, but by the end of the conference—like one blind man describing the trunk and the other describing the tail—they knew they had all recognized the elephant together. At the same time, however, Kevin Smith, a mathematician, gave a talk showing that projection reconstruction imaging wasn’t possible!
Paul Lauterbur and the Invention of MRI Page 10