Skip to content

Medical Imaging: CT and MRI

Abstract

An X-ray film is a shadow. A CT or MRI picture is a computation: the machine measures something that no eye could interpret and a program reconstructs the slice. Godfrey Hounsfield, a computer engineer at the British record and electronics company EMI, built the first CT scanner and saw its first patient scanned on 1 October 1971. Magnetic resonance imaging followed within the decade, out of a 1973 paper by Paul Lauterbur that Nature at first rejected, and brought with it a priority dispute that reached the newspaper advertising pages in 2003. Both machines are computers with an unusual input device, and neither could have existed before the minicomputer.

The Reconstruction Problem

A conventional X-ray adds up everything along the beam’s path. Bone, brain, and tumor are superimposed, and soft tissues that differ by a few percent in density cannot be told apart. The way out is to measure the same slice of the body from many angles and work backwards to the density at each point. The Austrian mathematician Johann Radon had shown in 1917 that a function can be recovered from its line integrals, in a paper with no application in view.

Several people saw the medical use before the hardware could deliver it. The American neurologist William Oldendorf proposed a pencil-beam scanner in 1961 and received a patent in 1963. Allan Cormack, a physicist working in the radiology department of the University of Cape Town, published the mathematics and a small experiment with a cobalt-60 source and a Geiger counter in the Journal of Applied Physics in 1963 and 1964. Nobody responded, and Cormack did not take the problem up again until he heard of Hounsfield’s work in 1972.

Hounsfield at EMI

Hounsfield was born on 28 August 1919 on a farm at Sutton-on-Trent in Nottinghamshire, learned electronics and radar as a volunteer reservist in the Royal Air Force, and took a diploma at Faraday House Electrical Engineering College in London. He joined EMI in October 1949. In 1958 he helped design the EMIDEC 1100, the first all-transistor computer sold commercially in Britain.

Around 1967, after the EMIDEC work, EMI let him choose his own research project. By his account the idea came on a country outing: one could work out what was inside a box by taking X-ray readings through it from all angles. His laboratory prototype used an americium gamma source and a single detector. The source was so weak that collecting the 28,000 measurements for one picture took up to nine days, and a large computer needed more than two hours to process them. He then moved to an X-ray tube and more sensitive detectors. An early test object, a preserved human brain from a hospital museum, gave misleading readings because the formalin had enhanced them, so he turned to fresh cattle brains. The neuroradiologist James Ambrose of Atkinson Morley Hospital suggested kosher animals, which are not stunned before slaughter, because stunning filled the brain with blood.

EMI did not make medical equipment and would not pay for the whole project. Hounsfield obtained a grant from the British Department of Health and Social Security, which the IEEE’s account puts at about 40,000 dollars.

Myth: The Beatles paid for the CT scanner

EMI owned the Beatles’ label, its profits rose sharply in the 1960s, and the story that record sales financed Hounsfield’s scanner is told in radiology lectures and museum captions. Zeev Maizlin and Patrick Vos went through the funding in the Journal of Computer Assisted Tomography in 2012. Leaving aside the salaries of Hounsfield and his team, EMI spent about £100,000 on the development; the Department of Health and Social Security spent £606,000. See Myths and Misconceptions.

Atkinson Morley, 1 October 1971

The first clinical machine was installed at Atkinson Morley Hospital in Wimbledon, where Ambrose worked. On 1 October 1971 it scanned a woman with a suspected brain tumor. The X-ray tube and detector moved across the head, turned one degree, and moved across again, through 180 degrees. The picture, 80 by 80 points, showed a dark cystic mass about the size of a plum in the left frontal lobe.

Ambrose presented the images at the annual congress of the British Institute of Radiology in 1972. The production machine, the EMI Mark I, took its 180 projections in about five minutes and reconstructed the slice on a Data General Nova 820 minicomputer with 32 kilobytes of memory. Radiologists who had spent their careers inferring brain lesions from displaced blood vessels could now see them. EMI received a Queen’s Award for Technological Innovation in 1973.

The scale on which CT density is reported is named after him: air is −1000 Hounsfield units, water 0, and dense bone +1000 and more. Hounsfield and Cormack shared the 1979 Nobel Prize in Physiology or Medicine. Neither had a medical degree. Hounsfield was knighted in 1981 and died on 12 August 2004.

⚠️ Dead End: EMI in the Scanner Business

EMI had the market to itself for a short time. The head scanner invited the obvious next product, and in 1974 Robert Ledley at Georgetown University showed the ACTA, a whole-body scanner with a 160-by-160 image. By about 1977 no fewer than seventeen companies were selling CT machines, among them General Electric and Siemens, which already sold X-ray equipment to the same hospitals. EMI could not compete and stopped making scanners. In October 1979 it merged with Thorn Electrical Industries to form Thorn EMI.

The technology went on without it. In 1989 Willi Kalender at Siemens solved the artifact problem of scanning while the table moved continuously, which made spiral CT practical; detectors that recorded four slices at once arrived in 1998, 16 in 2001, and 64 in 2006.

Magnetic Resonance

Nuclear magnetic resonance was a tool of chemists. Hydrogen nuclei in a strong magnetic field absorb and re-emit radio waves at a frequency set by the field strength, and the signal says something about their chemical surroundings. In March 1971 the physician Raymond Damadian reported in Science that tumors and normal tissue gave different NMR signals. He filed a patent for a cancer-detecting apparatus, granted on 5 February 1974. It did not describe a way to make a picture.

Lauterbur, a chemist at the State University of New York at Stony Brook, supplied that. His notebook entry on the idea is dated September 1971, and he said it came to him over a meal at a Big Boy restaurant near Pittsburgh. If the magnetic field varies across the sample, the resonance frequency varies with position, and each frequency in the signal can be assigned to a place. Nature rejected the paper; Lauterbur asked for a second review, and “Image Formation by Induced Local Interactions” appeared in 1973. He said later that one could write the history of science of the past fifty years in terms of papers rejected by Science or Nature. His university declined to patent the method because it did not expect to recover the cost.

Peter Mansfield, a physicist at Nottingham, worked out the mathematics of the method and developed echo-planar imaging, which later cut scan times from hours to seconds; his group imaged a human finger in 1976. Damadian’s group, working with a home-built machine they called Indomitable, produced the first scan of a human body on 3 July 1977. It took almost five hours to make one image of the chest of his colleague Larry Minkoff. A team under John Mallard at Aberdeen obtained the first clinically useful image of a patient on 28 August 1980. Damadian’s company FONAR, founded in 1978, put a commercial scanner on sale in 1980 and found no buyers; General Electric’s research center built a 1.5-tesla whole-body machine in the early 1980s.

Like CT, MRI depends on computing at every step. The raw data are radio signals indexed by frequency and phase, and the image is their Fourier transform (see Cooley, Tukey, and the FFT). In 1990 Seiji Ogawa noticed that oxygen-poor blood disturbs the local magnetic field, which is the basis of functional MRI.

The Nobel Dispute

The 2003 Nobel Prize in Physiology or Medicine went to Lauterbur and Mansfield. Damadian was left out, and a group formed by FONAR, “The Friends of Raymond Damadian”, bought full-page newspaper advertisements under the headline “The Shameful Wrong That Must Be Righted”. Damadian’s patent had done better in court than his claim did in Stockholm: a judgment of 129 million dollars against General Electric for infringing it was upheld on appeal. Lauterbur died on 27 March 2007, Damadian on 3 August 2022.

Images as Data

Early scanners stored pictures in formats only their makers could read. In 1983 the American College of Radiology and the National Electrical Manufacturers Association formed a joint committee, and its first standard, ACR/NEMA 300, came out in 1985. The third version, in 1993, added network support and a new name, DICOM, Digital Imaging and Communications in Medicine. DICOM is the reason a scan made on one vendor’s machine can be stored in another vendor’s archive and read on a third’s workstation. The image archives built on it are also the raw material of machine-learning diagnosis (see AI in Clinical Practice).

📚 Sources