Medical Imaging: The Inside Story

AS RECENTLY AS 1970 THE ONLY WAY doctors could see inside a patient’s body without surgery was with X rays. This technology, which dated back to the horse-and-buggy era and had seen few improvements since, amounted to a form of two-dimensional photography, complete with negatives and film. During the 1970s and 1980s, however, three separate new methods combined to greatly expand a physician’s ability to find and diagnose potentially dangerous conditions.

The first of these was computerized axial tomography (CAT), which still used X rays but replaced the film with computer-generated images. The images appeared as cross sections of the head or body; tomography comes from the Greek tomos , a cut or slice. Next came magnetic resonance imaging (MRI), which used computers in similar fashion while relying on new physical principles that could see through bones to display internal organs vividly. The third technique, positron emission tomography (PET), was effective at disclosing regions of strong metabolic activity—an important aid in showing the brain at work.

Advanced mathematical methods were central to these developments, and some of the most important of them sprang from the work of the physicist Allan Cormack. In 1955 he was a lecturer at the University of Cape Town, South Africa, when the staff physicist at nearby Groote Schuur Hospital resigned. Therapists at that hospital were using radioactive isotopes in their treatment, and South African law required a qualified scientist to supervise their use. “I was the only nuclear physicist in Cape Town,” Cormack later recalled, and he accepted an invitation to spend part of each week at Groote Schuur (which in 1967 would be the site of the world’s first heart transplant). “I was horrified by what I saw, even though it was as good as anything in the world,” he said. Radiotherapy treatments were planned on the assumption that all organs absorbed radiation in a uniform way, taking no account of the differences among bone, muscle, lung, and other tissues. To Cormack it looked like an open invitation to radiation overdoses. “It struck me that what was needed was a set of maps of absorption coefficients for many sections of the body.”

Cormack saw that he could map the radiation absorption of different “slices” of the body by directing a beam of radiation through each slice at many angles and directions and measuring the attenuation, or absorption, of the beam each time. The data, transformed by use of appropriate equations, would yield an absorptivity map of each slice.

It was like having a closed box of chocolates with some of its contents missing. If you could cut strips through the box in different directions and weigh them, magically reassembling the box each time, eventually you would be able to map what was where inside the box.

“This seemed like a problem which would have been solved before, probably in the nineteenth century,” Cormack said. In fact it had: The Austrian mathematician Johann Radon had given the solution in 1917, but Cormack did not learn of this until about 1970, so he studied the problem using his own derivations.

He conducted an initial experiment in 1957, using an aluminum cylinder surrounded by a ring of wood as his test object. After measuring its gamma-ray absorption from different angles, he drew curves on graph paper, plotted data points by hand, and found that part of the data did not fit his predictions. He asked questions at the machine shop and learned that the aluminum cylinder held a central peg with a different absorption coefficient. This encouraged him, for his data had disclosed the existence of the peg even though he had not known it was there. It took no great mental leap to imagine that his method might be able to disclose undiscovered tumors.

Cormack later married an American physicist, moved with her to Boston, and joined the faculty of Tufts University in nearby Medford, where he eventually became chairman of the physics department. In 1963 he began studying a more complex test object that mimicked a human head. An outer ring of aluminum represented the skull, Lucite within the ring stood in for the brain, and two aluminum disks, set within the Lucite, simulated tumors. For about a hundred dollars Cormack built a scanner with a collimated gamma-ray source (that is, one that created a beam of parallel rays), a detector, and a small platform on which he could hold, move, and rotate the test object within the fixed beam. This time the calculated results were good enough to publish, and Cormack wrote two papers around them. In time this work would win him a share of a Nobel Prize, but in 1964 almost no one noticed.

His discoveries nevertheless were far-reaching, for they pointed toward an entirely new approach to medical diagnostics. Moving from gamma-ray treatment to X-ray imaging, Cormack envisioned a scanning instrument with 100 collimated X-ray sources set side by side around an arc of a circle. Each source would produce a narrow beam that would strike its own detector to measure how much of it had been absorbed. The platform then would turn through a small angle, perhaps one degree, and the procedure would repeat at one-degree intervals until a complete semicircle had been covered. The result would be a set of 100 x 180 measured values. This array of data, processed using Cormack’s equations, then would show an X-ray image of the patient’s chest in cross section.

Cormack lacked the means to pursue such a vision beyond the conceptual stage, but the next pioneer, Godfrey N. Hounsfield, was more fortunate. He was a staff scientist at the British firm of EMI, Ltd., which had grown rich by recording the music of the Beatles. Hounsfield had built a strong professional reputation by designing Britain’s first all-transistor computer, and management let him pursue topics of personal interest that might lead to new business opportunities.

Hounsfield had worked with radar during the war and was familiar with how it scanned its surroundings by illuminating them with a central rotating beam. He thought: What if we were to turn radar inside out? Rather than create an image by scanning from the center, what if we were to study the internal structure of an object through observations made entirely at the periphery? Since he was not familiar with Cormack’s work, he developed his own mathematical method to process the image, an iterative procedure using linear algebra instead of Cormack’s definite integrals.

Early on, Hounsfield realized that this approach offered considerable improvement over conventional X rays. X rays yielded sharp images of bones, but they worked poorly on soft tissues, which all had nearly the same absorbing power. With almost no contrast between them, there was no way to make individual organs stand out. Cancerous tumors thus were hard to detect, and imaging the brain was also difficult, for it lay encased in bone. But Hounsfield appreciated that in reconstructing images, a computer—unlike X-ray film—could greatly amplify the small contrasts contained within the data.

Like Cormack, he began by experimenting with gamma rays, which were easy to generate, collimate, and detect. He launched his studies with observations of a plastic box filled with water and holding pieces of metal and plastic. He took 28,800 readings over nine days. A computer digested the data and produced an image on a video screen, which was then recorded on film. It was fuzzy, but the original shapes and density variations were clearly discernible.

During 1969 Hounsfield improved his technique by replacing the gamma rays with an X-ray source, which cut the scan time to nine hours, and by working with animal brains obtained from slaughterhouses. Studies of human and animal tissues continued during 1970, and soon it was appropriate to think of clinical trials. By now Hounsfield had reduced the scanning time to several minutes, but the use of living patients created a new problem. As had been true in the early days of photography, a subject would have to stay completely still the whole time or else the image would blur. Hounsfield decided to emphasize examination of the head, which could readily be immobilized and was hard to study with conventional X rays.

EMI proceeded to build a prototype CAT scanner and had it installed in a small London hospital, far from the eyes of potential competitors. The first patient to be examined was a woman in her early forties with symptoms that suggested a brain tumor. The picture clearly pinpointed the location of a tumor in her left frontal lobe. A staff neurosurgeon, James Ambrose, operated and removed it, later saying that the tomogram “caused Hounsfield and me to jump up and down like football players who had just scored a winning goal.”

A few months later, in April 1972, EMI held a press conference. To a world that had known only flat X rays, the cranial CAT scanner was a wonder. It could discriminate between congealed and liquid blood, allowing a diagnostician to decide between prescribing a coagulant or an anticoagulant. It showed tumors and other lesions and gave promise of distinguishing malignant from benign tumors. The resulting surge of interest caught EMI unprepared, for the firm had planned to build only five prototypes. But soon a new company division, EMI Medical, began to manufacture and market CAT scanners. In 1975 EMI cut the scanning time to 20 seconds. Most patients could hold their breath that long, permitting unblurred images of the chest, and EMI promptly introduced a whole-body scanner. By 1977 some 1,130 scanners were in use around the world, more than 700 of them built by EMI. In 1979 Hounsfield and Cormack shared the Nobel Prize in Physiology or Medicine—an unusual choice in that the two had never met before the award ceremony, neither had formally studied physiology or medicine, and neither one had a doctoral degree in any subject.

By then the CAT scan faced competition. It came from imaging systems that drew on a different physical phenomenon, nuclear magnetic resonance (NMR). NMR works on any atomic nucleus that has an odd number of protons or neutrons. (Fortunately, hydrogen, which is found in water and most organic compounds, meets this criterion, since its nucleus is a single proton. Potassium and other elements can also be used in NMR studies, but hydrogen is by far the most common subject.) Nuclei of this type have a property called spin—a basic attribute of atoms, like mass or electric charge. Spin can be thought of conceptually as the rotation of a nucleus about its axis. The nucleus does not actually rotate, but the physics works as if it did.

If an external magnetic field is applied to such a nucleus, its spin vector (which may be thought of as the axis of rotation) will line up along the direction of the field. This spin vector will point either “up” (in the same direction as the magnetic field) or “down” (in the opposite direction). The key fact that makes NMR possible is that a small but measurable energy gap exists between these two states—a gap proportional to the strength of the magnetic field.

NMR apparatus uses this gap to reveal clues about the sample being examined. The sample is placed within a powerful magnetic field, which aligns all the nuclei along its axis, either up or down. Then radio-frequency radiation is applied perpendicular to the magnetic field. This radiation “flips” nuclei from the lower-energy to the higher-energy spin state, and as the flipped nuclei decay back to the lower state, they emit radiation. The process is characterized by three values: the frequency of the emitted (or absorbed) radiation, its intensity, and the relaxation time for the sample to return to its normal state.

After its discovery in 1946 NMR spectroscopy saw its first widespread use by organic chemists. For their purposes, the first two of these quantities—the frequency and intensity of absorbed radiation—were of supreme importance, while the third—relaxation time—was mostly a curiosity. An organic chemist’s NMR spectrum is a jagged line with dozens or hundreds of peaks, which vary from a reference value by amounts measured in parts per million. By examining the number, intensity, and shift of these peaks, a chemist can deduce valuable information about the structure of a molecule. Since the quantities being measured are so small and must be measured so precisely, chemists take great care to make sure that the sample is pure and homogeneous and that the applied magnetic field is absolutely constant.

Beginning in the late 1950s, however, medical and biological researchers also started using NMR. Their needs were different from those of organic chemists. First of all, the samples they examined were actual bits of tissue—lumpy, hopelessly complicated mixtures that would have horrified a chemist. And while chemists used NMR to compare the behavior of different compounds under a standard set of conditions, medical researchers used it to compare the behavior of a single compound—water, in most cases—in widely varying conditions, such as in different sorts of tissue. They concentrated on measuring relaxation times rather than absorption intensity or emission frequencies.

Both these methods used NMR to measure the bulk properties of matter. They made no attempt to map the interior of a sample; indeed, chemical NMR apparatus keeps a thin glass tube or the sample spinning at high speed to suppress even microscopic internal variations. Medical researchers, less rigidly constrained, were able to use NMR on living creatures, measuring blood flow in a mouse, for example. Still, there was no way to map what lay inside a sample, so while NMR became an indispensable tool for chemists, it did not at first find much use in medicine.

Raymond Damadian had expected to work within these limitations when he first became acquainted with NMR. Damadian was a researcher at Downstate Medical Center in Brooklyn, New York. In 1969 he became interested in whether NMR might offer new insights into cancer. At the time, pathologists identified carcinomas by removing a tissue sample and examining its cells under a microscope. Damadian thought the water found in malignant cells might behave differently from that found in normal cells. If so, chemical analysis could detect cancer with quantitative measurements instead of visual inspection.

That April, attending a conference, he discussed his idea with a colleague, Freeman Cope of the U.S. Naval Air Development Center in Warminster, Pennsylvania. Cope had been using NMR to study sodium in brain tissue and now hoped to measure potassium, whose signal would be much weaker. He needed specialized cells that contained a great deal of this element. When Damadian said he knew how to obtain potassium-rich bacteria that grew in the Dead Sea, Cope proposed that they collaborate. The two men arranged to work with equipment at NMR Specialties, a small firm in New Kensington, Pennsylvania, near Pittsburgh.

“I remember the first time I saw a potassium signal,” Damadian recalled. “This huge blip filled the oscilloscope screen. I had never seen an NMR machine, and it had a profound effect on me. I mean, wow! In a few seconds we were taking a measurement that would usually take me weeks and sometimes months to do accurately.” At this point Damadian and Cope were using NMR only on extracted samples, but Damadian foresaw a much more valuable role for the technology.

In September he applied to New York City’s Department of Health for a grant of $89,000. In his letter of application he wrote, “I am very much interested in the potential of NMR spectroscopy for early non-destructive detection of internal malignancies.” He hoped “to establish that all tumors can be recognized” by use of NMR and to “proceed with the development of instrumentation and probes that can be used to scan the human body externally.” He didn’t get the grant, and there was good reason: Damadian was way ahead of himself. Before funding a program to develop external NMR apparatus, the council needed proof that NMR could actually distinguish cancerous from noncancerous cells; otherwise the apparatus would be useless. Damadian set out to secure such data.

In June 1970 he returned to NMR Specialties, the trunk of his car holding cages filled with rats. They had tumors of sarcoma—cancer of the muscle and connective tissues. “I put a tumor sample in a test tube and measured the relaxation time,” he recollected. “Then I cut some normal tissue from some rats and measured that to see if the relaxation time was different. To my amazement, it wasn’t just different, it was considerably different. The tumorous samples had longer relaxation times.… I could now distinguish cancerous tissue from normal tissue by the NMR signal. No one had ever suggested this before, and I just knew I could do this on the whole human body.” Without cutting it open, that is.

Damadian returned to NMR Specialties again in July, to repeat his measurements with new rats while broadening them to include studies of hepatoma (liver cancer). As expected, hepatoma tissue differed sharply from normal liver tissue. Damadian also found that healthy tissues from various organs—brain, liver, kidney, intestine—showed markedly different relaxation times from one another. He wrote up his findings in a paper that appeared the following March in the prestigious journal Science . Although he did not discuss the formation of images, the article became a cornerstone of what would later be known as magnetic resonance imaging.

In March 1972 he filed a patent application titled “Apparatus and Method for Detecting Cancer in Tissue.” The accompanying diagram showed a patient standing within a large magnetic coil. A probe, passed over the chest, would stimulate a succession of small internal volumes, and the instrument would read off the relaxation times. The resulting data would be assembled like a mosaic to form a map of the body. The doctor would recognize the heart, the liver, the stomach—and, if the patient was unlucky, a cancerous tumor.

Damadian bought some NMR equipment and continued his research at Downstate. Other researchers read his Science paper and launched their own investigations. In the course of this work, Damadian acquired a competitor who would soon become a bitter rival: Paul C. Lauterbur, an experienced NMR specialist at the nearby State University of New York at Stony Brook.

Lauterbur came into the picture in 1971, when NMR Specialties was headed for bankruptcy. Hoping to save the company, an executive asked him to take over as president and chairman of the board of directors. That summer, while working at NMR Specialties, Lauterbur talked with a visiting researcher who was pursuing Damadian’s line of study. Lauterbur had not known of this work, but when he read Damadian’s paper in Science , he saw that a new era in medicine was at hand. He also saw that NMR would be vastly more effective if he could use it to create three-dimensional images. He began puzzling over how this might be done.

Over dinner at a restaurant one night in early September, Lauterbur suddenly came up with the solution. As he ate his Big Boy double-deck hamburger, Lauterbur remembered that the energy gap between a nucleus’s up and down states—and thus the frequency needed to flip its spin—is proportional to the applied magnetic field. Normally, to keep a spectrum’s sharp peaks from blurring, this field is held absolutely constant throughout the sample.

But what if you intentionally used a varying magnetic field? For instance, what if you put a patient inside a field that increased uniformly from front to back? Because of the increasing field strength, the energy gap would be greater at the back than at the front. And since the frequency of the emitted radiation is proportional to the energy gap, you would be able to tell where (along the front-to-back axis) a sample of emitted radiation had come from by looking at its frequency.

This was Lauterbur’s vitally important realization. Medical NMR researchers had concentrated on measuring the relaxation time of emitted radiation while essentially ignoring its frequency; with Lauterbur’s setup, researchers would make use of both quantities. The relaxation data, as before, would reveal what sort of tissue was present, while the frequency data would reveal where the tissue was. A single pulse of radio waves through the body, suitably processed by a computer, would yield a one-dimensional map. A semicircle around the body would yield a two-dimensional cross section, resembling a CAT scan.

Lauterbur started with the two-dimensional case and published his results in Nature in 1973. His paper included the first crude example of an image produced by NMR—a two-dimensional projection of a pair of capillary tubes in water. Damadian, meanwhile, was proceeding with his own approach to NMR imaging, wherein his probe would obtain data, point by point, for display as a single horizontal cross section through the body. He missed the Nature paper and heard of Lauterbur’s work only as rumors. Then Lauterbur gave a talk at Brooklyn College late in 1973, and a friend of Damadian told him about it. Lauterbur had spoken of scanning the human body with NMR—and he was leading people to believe that he had invented this concept.

Damadian read the Nature paper and saw no reference to his own article in Science . He blew up: “Here I was talking about medical scanning and getting ridiculed and here was this guy standing up and saying that he had invented it.” His anger deepened during subsequent years as Lauterbur continued to publish and persisted in not citing Damadian’s work.

Yet Lauterbur could quite properly claim to have invented the concept of NMR imaging, for his approach to image reconstruction was both feasible and original. True, Damadian’s Science paper had proposed “an external probe for the detection of internal cancer,” but neither this paper nor his subsequent ones specifically discussed how actual images might be formed. In 1973, though, both men were still far from being able to create a human image. Damadian had set this as a goal, and as he remembered, “I would have died before I let him beat me. There was no way I was going to let Lauterbur get that image before me.”

He stepped up the pace when Lauterbur showed him an NMR image of the chest of a mouse. Early in 1976 Damadian built a workable probe, which meant that he could create NMR images using his point-by-point technique. One such image showed a tumor in a mouse. It was not a photo or even a detailed mosaic; it was more of an abstract-looking pattern formed from several hundred brightly colored squares. Even so, Science put it on the cover of its December 24 issue. An editor later explained that he wanted something appropriate for Christmas, and Damadian’s image was bright with seasonal-looking colors.

There was little joy in Damadian’s laboratory that Christmas, though, because funds were tight. This raised serious problems, especially since he was preparing to leap from mice to men by Grafting an instrument big enough to hold a patient. He wanted a magnet of 5,000 gauss (the earth’s field has a strength of 0.7 gauss). He didn’t have the money to buy one, so he and his assistants would have to build it from scratch. Damadian came up with a design that had two superconducting coils, each with a diameter of 53 inches. If successful, it would be the ninth largest superconducting magnet in the world.

Through good fortune he was able to purchase some surplus superconducting wire at 10 cents on the dollar. A computer program from nearby Brookhaven National Laboratory calculated detailed specifications for the magnet. His assistants found machine tools that were dirt cheap in secondhand stores on Manhattan’s Canal Street. Damadian’s assistant Larry Minkoff learned welding by reading articles in Popular Science .

As time pressed and funds remained short, Damadian abandoned his plans for a second coil and decided to use a single superconducting magnet, even though it would be less effective. He also needed an antenna to pick up the weak radio signals that his NMR apparatus would elicit from atomic nuclei, lhe group built the antenna as a coil four feet m diameter. When they installed it within the magnet and tried it out, it gave no response.

The equations they had used proved to work well for small antennas, 2 or 3 inches across, but were useless at larger sizes. Michael Goldsmith, another of Damadian’s assistants, grimly proceeded by trial and error, building antennas and seeing how large he could go, then adjusting the parameters by guesswork. On the fiftieth try he reached 14 inches in diameter, barely large enough to encircle a human chest. With this antenna, which was made of cardboard and copper-foil tape, the machine was complete. Damadian christened it Indomitable.

It looked like the entrance to a tunnel. The machine made measurements at a single, immovable point in space. Hence the human guinea pig had to build up his image by shifting position while this point remained fixed. He wore the antenna like a vest and sat atop a long wooden plank, going backward and forward as it was moved from side to side. At each position an NMR device scanned a small spot within the patient. When a large number of scans had been taken, they would be assembled into a two-dimensional section of the subject’s body.

The magnet proved faulty, delivering only 500 gauss instead of 5,000. In May 1977 Damadian gamely went ahead anyway. He donned the antenna himself and sat on the rail, with a cardiologist standing by in case the magnet interfered with his heart. But Damadian was overweight and the antenna failed to work. A thinner person was needed, and after some weeks of hesitation Minkoff volunteered. He had to sit in the machine for nearly five hours as Damadian built his image point by point. Still, it worked. They measured 106 data points, which was enough for a crude cross section of the chest. Damadian wrote in his lab notebook on July 3:

FANTASTIC SUCCESS!

4:45 A.M. First Human Image

Complete in Amazing Detail

Showing Heart

Lungs

Vertebra

Musculature

Damadian had beaten Lauterbur, who, as it turned out, was not even in the race. Lauterbur had made no attempt as yet to image a human, instead studying mice and other small creatures. But even as Damadian was producing his first crude images and Lauterbur was refining his own technique, CAT scanning was taking the world by storm. While Damadian had carried out a whole-body NMR scan, he was in no position to compete with CAT. The “amazing detail” existed only in his mind, for his 106 data points were far too few. CAT scanners were giving images with quality similar to that of television, and in seconds rather than hours. Lauterbur later described Damadian’s point-by-point technique as “an obvious dead end. It was slow and produced low resolution and poorly defined images.”

Knowing that his patched-together apparatus could be greatly improved, Damadian was less pessimistic. He left Downstate and started a company with financial support from family and friends. He continued to use his point-by-point approach, but with better circuitry that cut the scan time. The new design escaped the difficulties of superconductivity by working with permanent magnets made of magnetized alloys. Damadian also adopted new methods for image reconstruction that gave more speed and better detail, less like a mosaic and more like a Seurat painting.

Damadian’s company, Fonar Corporation, displayed a prototype of a commercial machine in April 1980 at a meeting of the American Roentgen Ray Society and began taking orders. Then in 1981 Damadian received a shock. Diasonics, a Silicon Valley start-up company, had built a scanner with a superconducting magnet of 3,500 gauss. Using a new technology relying on two-dimensional Fourier transformations, its images showed far more detail and resolution than Damadian had thought possible. Adopting the Diasonics technology, he fought back with a new model that had a 3,000-gauss permanent magnet and held his own against the competition. Fonar remains in business today. Lauterbur, meanwhile, never tried to market his invention. He continued as a research scientist, eventually moving to the University of Illinois as a laboratory director. Fortunately, there was business for all. By 1992 some 4,000 scanners were in clinical use. (Along the way, in the mid-1980s, the name of the technology changed. The N in NMR stood for the scary-sounding nuclear , and to make it less off-putting, the term magnetic resonance imaging was adopted.)

In some important respects MRI is more useful than CAT. CAT pushes X rays to the limit, overcoming the lack of contrast in soft tissue with clever computer processing. But MRI offers exceptional contrast right at the start, even without computer enhancement. Its magnetic fields have proved safe in clinical use, and since it uses no ionizing radiation, patients can be scanned repeatedly with no risk of an overdose. MRI responds to hydrogen nuclei in water, which is abundant in soft tissue but sparse in bone. Bones therefore drop out of MRI images —a reversal of traditional X-ray photos, which show bone but not soft tissue.

MRI demonstrated its value early. A young man was suffering from intractable seizures, and a CAT scan showed nothing wrong. MRI disclosed within his brain a mass of fibrous tissue, which surgeons removed. Another man had cancer that doctors could not find with X rays. MRI revealed a tumor in his brain, and he too underwent successful surgery. MRI systematically detected lesions that CAT missed while also finding hemorrhages and multiple sclerosis. By looking past the vertebrae, MRI gave clear images of the spinal cord.

As MRI and CAT were becoming established in clinical practice, another imaging technique, PET, was growing out of Cormack’s old field of nuclear medicine. After 1945 the Atomic Energy Commission promoted the clinical use of radioactive isotopes. Those that emitted positron radiation proved to be especially helpful in treating certain diseases. A positron is the same as an electron except that it has a positive charge. When a positron meets an ordinary electron, they annihilate each other, yielding two energetic gamma rays. These pass easily through the body, flying in opposite directions, and can be detected readily. Certain isotopes of carbon, oxygen, and nitrogen are particularly good positron emitters.

The main obstacle to clinical use of these isotopes has been their half-lives, which are measured in minutes. This means they must be produced on the spot, with a cyclotron. Cyclotrons are expensive and demanding, and the first hospital-based system did not enter service until 1964. The physicist Michel Ter-Pogossian installed it at Washington University in St. Louis.

In 1972 his group built the “lead chicken,” a positron imaging device that resembled a helmet spiked with 26 detectors. The device gave data, but researchers had to process it by hand, since they did not have computer algorithms that would allow them to construct images easily. Then in 1973 Godfrey Hounsfield published a description of his CAT scanner. Two of Ter-Pogossian’s colleagues, Michael Phelps and Edward Hoffman, realized that Hounsfield’s methods for image reconstruction would work for them as well. They dismantled the lead chicken and built a new instrument, dubbing their process PETT (positron emission transaxial tomography).

PET (as it became known after transaxial was dropped) was most useful in following radioactive tracers that flowed within the bloodstream and concentrated in areas of active metabolism. Thus, while CAT and MRI showed the structure of organs, PET revealed their workings. It proved capable of diagnosing Alzheimer’s disease, which had eluded other techniques and often had shown itself only at autopsy. PET also enabled researchers to map the working brain, watching specific regions light up as they became active and received more blood flow.

Unlike CAT and MRI, PET has never truly caught on for diagnostic use. The first five years of CAT saw more than 1,000 machines installed, but throughout the world only about 40 institutions set up PET centers in the five years that followed the initial paper by Phelps and Ter-Pogossian in 1975. The major firms of Siemens and General Electric have entered the field, but PET still lacks the broad applicability of CAT and MRI and is far more costly. It requires not only a scanner but a cyclotron and a laboratory for nuclear chemistry, along with all the necessary permits and licenses.

Amid these innovations, conventional X rays remain the most prevalent form of imaging. A 1997 review counted 4.3 million MRI scans in the United States per year, 21 million CAT scans, and 208 million uses of X rays (including dental X rays). X-ray equipment, inexpensive and easy to use, remains the method of choice for such common injuries as bone fractures. X rays can also detect cancer and tuberculosis and give good views of the gastrointestinal tract after a patient has swallowed a radiopaque dye. The technology has gone on to broaden its use, with mammography, the examination of women’s breasts for tumors, being an important recent application.

Still, while X rays continue to predominate, they are no longer the only option. CAT and MRI are probably no farther away than your local medical center, and most -Americans live within 200 miles of a PET scanner. Together these systems allow patients to receive internal examinations that are tailored to their specific needs. Before placing them under the knife, doctors can now extract volumes of information unimaginable to a surgeon just a generation ago.