The history
A century from a mathematician's formula to scanners that see inside you in seconds.
In 1917 a mathematician in Vienna showed how to rebuild a shape from its shadows, but nobody could do the sums. Fifty years later, a physicist in South Africa and an engineer at a British record and electronics company worked it out again, separately, and the engineer built a machine. The first patient was scanned in London in 1971. Since then CT has gone from nine days per picture to a whole body in seconds, and from counting light to counting single X-rays.
1917
Mathematics of reconstruction
Johann Radon, Austria
1 Oct 1971
CT scan of a patient
Hounsfield and Ambrose, London, UK
1971
Convolution (Ram-Lak) filter
Ramachandran and Lakshminarayanan, Bangalore, India
1974
Whole-body CT scanner
Robert Ledley, USA
1978
CT scanner in India (widely cited)
AIIMS, New Delhi
1989
Clinical spiral CT
Willi Kalender and colleagues, Germany
2005
Dual-source CT
Siemens, Germany
2021
Photon-counting CT cleared by the FDA
Siemens Healthineers, Germany
1895 – 1966
X-rays are found, a mathematician solves the problem in theory, and a few people build slice machines with film, toy trains and hand calculations.
1895
8 November 1895
Röntgen noticed a glowing screen across his lab while working with a cathode-ray tube, and found invisible rays that passed through card, wood and flesh but not bone. Within weeks doctors were taking X-ray pictures around the world. (How X-rays are made and stopped is the story of XRayClear.)
Why it mattered. Every CT scanner still starts with the same kind of ray, just measured far more carefully.
1917
Radon, a mathematician at the Technical University of Vienna, showed how a flat shape can be worked out exactly from all of its sums along straight lines, and gave the formula for doing it. Nobody had any use for it in medicine for another fifty years.
Why it mattered. This is the Radon transform: CT's projections are exactly these line sums, and reconstruction runs it backwards.
1930
1930–1931
Vallebona built a machine that moved the X-ray tube and film in step, so only one layer of the body stayed sharp; he called it stratigraphy. A year later the Dutch radiologist Ziedses des Plantes described his own version, planigraphy. Several people in different countries claimed the idea around this time.
Why it mattered. Doctors could now pick out a single slice, but the other layers still smeared across the picture.
1956
1956–1957
Working part-time as a hospital physicist, Cormack wondered how radiotherapy could be planned without knowing how X-rays were weakened inside each part of the body. He began working out how the inside could be calculated from measurements taken outside.
Why it mattered. It was the first time someone set out to compute a map of the body's attenuation.
1961
Oldendorf, a neurologist at UCLA, built a working model from his son's toy train, a record-player turntable and an alarm-clock motor. It picked out a nail hidden inside a ring of other nails, like something deep inside a skull. He was granted a US patent in 1963, but companies were not interested.
Why it mattered. It showed cross-sectional imaging could work, years before computers made it practical.
1963
1963 and 1964
Now at Tufts, Cormack published two papers in the Journal of Applied Physics showing how to rebuild a slice from line measurements, and tested the idea with gamma rays on simple phantoms of aluminium and wood. The papers got almost no attention at the time.
Why it mattered. The theory of CT was complete before anyone built a scanner, even though Cormack did not know about Radon's work.
1967 – 1979
Godfrey Hounsfield builds the first CT scanner, the first patient is scanned, and CT spreads from London to the world, including India, ending with a Nobel Prize.
1968
1967–1968
Hounsfield, an engineer who had built one of Britain's first all-transistor computers, had the idea of scanning a body from many angles and letting a computer rebuild it. His lab prototype, with a gamma-ray source, took up to nine days to collect the data for one slice and over two hours of computing to rebuild it. He tested it on preserved brains, then on cattle brains.
Why it mattered. It proved that a computer could turn shadows into a slice with enough detail to tell grey matter from white.
1969
c. 1968–1972
A popular story says Beatles record sales paid for the CT scanner. Historians doubt it: a 2012 study estimated that EMI spent about £100,000 on the project (not counting salaries), while the UK Department of Health and Social Security spent about £606,000 and agreed early to buy scanners. EMI's music profits may have helped the company afford research, but the exact link is disputed.
Why it mattered. It's a reminder that big medical inventions often rest on public money as well as private.
1971
1 October 1971
A woman with a suspected brain tumour became the first person scanned. The scan took about 30 minutes; the tapes were then carried across London for about two and a half hours of computing, and the picture was photographed off a screen. It showed a cyst-like mass about the size of a plum in her left frontal lobe.
Why it mattered. For the first time, doctors could see soft tissue inside a living skull without surgery or painful older tests.
1971
September 1971
Working on 3D images from X-ray and electron-microscope pictures, the two scientists published a way to rebuild slices by convolution: sharpen each projection with a simple filter, then back-project. It was about 30 times faster than the Fourier method. Their filter is still called Ram-Lak.
Why it mattered. Filtered back-projection became the standard way CT images were made for decades.
1972
20 April 1972
Ambrose and Hounsfield gave the opening paper at the British Institute of Radiology's annual congress, showing brain scans made without contrast agents. The Times reported it the next day, and hospitals around the world began to ask for scanners.
Why it mattered. The announcement set off a race among manufacturers to build CT scanners.
1973
June 1973
The first EMI scanners in North America were installed at the Mayo Clinic, where the first scan was done on 19 June 1973, and at Massachusetts General Hospital a few weeks later. Each head scan still took minutes per slice.
Why it mattered. Within a few years, dozens of companies were making scanners.
1974
June 1974
To compare ways of rebuilding images, Shepp and Logan invented a simple head made of ellipses, with a skull, ventricles and small tumours. The Shepp–Logan phantom is still the standard test image for CT maths.
Why it mattered. It gave everyone the same test, so methods could be compared fairly.
1975
November 1975
After tests at Frankfurt's university hospital from 1974, Siemens began series production of its SIRETOM head scanner in November 1975. In 1977 its SOMATOM body scanner took about four seconds per slice.
Why it mattered. Scan times fell from minutes to seconds in just a few years.
1978
According to the Indian Society of Neuroradiology, India's first CT scanner, an EMI 1010, was set up at AIIMS in 1978, and by the mid-1980s scanners were being installed across the country. Other hospitals also claim early scanners, so exact firsts are debated.
Why it mattered. CT soon replaced older, painful brain tests such as pneumoencephalography in India too.
1979
October 1979
The Nobel Prize in Physiology or Medicine went to a physicist and an engineer for the development of computer-assisted tomography. Neither was a doctor, and Cormack had no PhD. Oldendorf, who had shared a major prize with Hounsfield in 1975, was left out.
Why it mattered. It recognised that the theory and the machine had been invented separately, on two continents.
1977 – 2004
Slip rings, spiral scanning and more and more detector rows cut scan times from minutes to a fraction of a second.
1987
Early gantries were connected by cables, so they had to stop and turn back after every slice. Slip rings pass power and data through sliding contacts, so the ring can spin continuously. Siemens' SOMATOM Plus turned once a second, then the fastest CT in the world.
Why it mattered. Continuous spinning made the next big idea, the spiral scan, possible.
1989
November 1989
Kalender, Seissler and Vock showed the first clinical spiral CT at the RSNA meeting in Chicago: the table moved steadily while the tube spun, and software rebuilt slices at any position. A whole chest could be scanned in one breath-hold.
Why it mattered. Every modern CT scan is a spiral scan.
1998
The first multi-slice scanners had four rows of detectors and turned in half a second, collecting four slices per turn. That was about eight times the performance of one-second, single-slice scanning.
Why it mattered. More rows meant thinner slices, faster scans and the first good 3D pictures.
2004
Announced at RSNA in 2003, the first 64-slice scanners were installed in 2004. With 64 rows of 0.625 mm detectors, a heart could be scanned in about 5 to 8 seconds, making CT pictures of the coronary arteries practical.
Why it mattered. CT angiography of the heart became a routine test.
By the numbers
From nine days on Hounsfield's lab prototype to a quarter of a second on today's fastest scanners.
2005 – today
Two tubes, wide detectors, smarter reconstruction and photon counting freeze the heart, tell materials apart and lower the dose.
2005
The SOMATOM Definition carried two X-ray tubes and two detectors on the same ring. At 0.33 s a turn, it could collect a slice of the heart in just 83 milliseconds, and the two tubes could run at different voltages for dual-energy scans.
Why it mattered. It froze the beating heart and made dual-energy CT practical.
2007
November 2007
Toshiba's Aquilion ONE had 320 rows of detectors covering 16 cm, enough to image a whole heart or brain in a single rotation without moving the table. It began shipping in 2008.
Why it mattered. Wide detectors let CT watch blood flow through a whole organ over time.
2008
Iterative reconstruction methods such as GE's ASiR replaced part of filtered back-projection with repeated guess-and-correct steps that model noise. Studies suggested similar image quality at about half the dose. Other makers followed, and today some use deep learning.
Why it mattered. Dose fell a long way while pictures stayed sharp.
2021
30 September 2021
The US FDA cleared the NAEOTOM Alpha, the first photon-counting CT scanner. Its cadmium telluride detectors turn each X-ray straight into an electrical pulse and sort it by energy, instead of adding up flashes of light.
Why it mattered. It is the biggest change to CT detectors since the 1970s.
Did you know?
The first CT image of a patient took about 30 minutes to scan and two and a half hours to compute, and the data tapes were carried across London.
Allan Cormack won a Nobel Prize for medicine without a PhD, and neither he nor Hounsfield was a doctor.
William Oldendorf's 1961 imaging model used his son's toy train, a record-player turntable and an alarm-clock motor.
A 2012 study found the UK government spent about six times as much on developing the first CT scanner as EMI did, which casts doubt on the story that the Beatles paid for it.
The Ram-Lak filter, used in CT for decades, is named after two scientists working in Bangalore in 1971.
The people
1887 – 1956 · Mathematician · Austria-Hungary (now Czech Republic)
Worked out in 1917 how to rebuild a shape from its line sums.
1924 – 1998 · Physicist · South Africa and USA
Worked out CT's theory in Cape Town and at Tufts, and shared the 1979 Nobel Prize without having a PhD.
1919 – 2004 · Electrical engineer · England
Built the first CT scanner at EMI; the Hounsfield unit is named after him.
1923 – 2006 · Neuroradiologist · South Africa and England
Scanned the first patient with Hounsfield in 1971 and showed doctors what CT could do.
1922 – 2001 · Biophysicist · India
With A. V. Lakshminarayanan, published the convolution back-projection method in 1971.
1925 – 1992 · Neurologist · USA
Built an early cross-sectional imaging model in 1961 from household parts.
1926 – 2012 · Dentist, physicist and computer pioneer · USA
Built ACTA, the first whole-body CT scanner, in 1974.
1949 – 2024 · Medical physicist · Germany
Led the development of spiral CT, first shown in 1989.
Where it happened
Sources
Dates marked “c.” are approximate, and historians sometimes disagree about who was first. If you spot a mistake, tell us.
↑↓ move↵ open/ or ⌘K search anywhere