1979 Nobel Prize in Physiology or Medicine: Cormack and Hounsfield Reconstruct the Body in 3D with X-rays and Computers

What You Will Learn in This Article

This article explains how Allan Cormack, a physicist with only a master's degree, and Godfrey Hounsfield, a company researcher with no bachelor's degree, created computed tomography (CT), which fundamentally changed 20th-century medical diagnosis. It also discusses why this discovery is a prime example of solving an inverse problem, and how it became the foundation for today's MRI, PET, and ultrasound and all other cross-sectional imaging diagnostic techniques.

A Story Different from Common Knowledge: Seeing the Inside of the Body in 3D Without Cutting It Open

Let's set out the background of this discovery. X-rays were initially used to diagnose fractures, tuberculosis, and pneumonia, and gained widespread recognition when they proved decisive in removing bullets and shrapnel from the body during World War I. Even so, it took many years before X-rays could be applied to medicine on a much wider scale.

Cormack was originally a physicist studying the interactions of the elementary particles that make up atoms, but he also took on the task of laying the physical and mathematical foundations required to build an X-ray tomography scanner. From the mid-1960s onward, he became interested in X-ray images of tissues with different densities, and eventually established the mathematical and physical principles of the Computer Assisted Tomography (CAT / CT) that combines X-rays with computers.

The Key Problem: X-rays are absorbed to different degrees as they pass through the body. However, X-ray imaging is a projection of a 3D body onto a 2D plane—the signals from multiple layers of tissue are combined into a single image. This results in blurry images, as the details of structures like the heart behind the bone and the lungs behind the heart overlap.

The Solution: The body is X-rayed from multiple angles, and the computer uses this multi-angle data to calculate and reconstruct the density at each location. The fundamental principle of CT is to reconstruct the original 3D information from multiple 2D projections.

In the language of computer science, this problem is a classic example of an inverse problem. The forward problem is "3D object → 2D projection," which is natural, while the inverse problem is the computational reconstruction of "multiple 2D projections → 3D object." It involves reversing the flow of information.

A Pivotal Moment in the Late 20th Century

1979 was a year in which several pivotal moments in the late 20th century converged.

In world history, the Iranian Pahlavi dynasty collapsed in January, followed by the completion of the Iranian Revolution with the return of Khomeini in February, marking a major turning point in Middle Eastern politics. In December, the Soviet Union invaded Afghanistan, marking the beginning of the decade-long Afghan War and a major contributing factor to the collapse of the Soviet Union. In May, Margaret Thatcher took office as Prime Minister of the United Kingdom, becoming the first female prime minister and ushering in the era of neoliberalism. In March, the Three Mile Island nuclear accident occurred, the worst nuclear accident in U.S. history, severely damaging public trust in the nuclear industry.

In Korean history, the assassination of President Park Chung-hee on October 26 (10.26) occurred, with the head of the Korean Central Intelligence Agency, Kim Jae-kyu, assassinating the president at his private residence. This marked the end of the Yushin regime. On December 12, the 12.12 incident occurred, a coup d'état led by the new military faction led by Chun Doo-hwan and Roh Tae-woo, arresting Army Chief of Staff Jeong Seung-hwa. This marked the beginning of the new military regime.

It was a year in which Korea's political landscape changed twice. And in that year, the Nobel Committee recognized CT—perhaps the principle that reality can only be understood when observed from multiple angles was also in demand in the political climate of the time.

Cormack: A University Professor with Only a Master's Degree

Allan M. Cormack (1924-1998) was an American physicist (born in South Africa). He earned a master's degree from the University of Cape Town in 1944 — and it is worth pausing here to note something distinctive.

Most Nobel laureates finish their degrees at elite universities and then extend their research careers from there. But Cormack held a professorship at a university with only a master's degree as his highest qualification, while Hounsfield spent his entire career doing research at a company without even a bachelor's degree from a traditional university and still won a Nobel Prize. The very trajectories of these two men are a distinctive feature of this award.

Let's fill in a bit more of his background. Cormack grew up in Johannesburg, South Africa, born in 1924 into a family that loved music. In high school he enjoyed music, theater, and sports, and was particularly captivated by astronomy. To dig deeper into that beloved astronomy, he taught himself mathematics and physics on the side, and that self-taught mathematics turned out to be decisive to his later research. He eventually set astronomy aside and took a bachelor's and master's degree in practical electrical engineering at the University of Cape Town, before falling in love with physics and going on to study cutting-edge nuclear physics at the Cavendish Laboratory at Cambridge University.

The mathematics he studied independently became the mathematical foundation for CT half a century later, an example of how early interests can bear fruit in unexpected ways.

A Decisive Moment. His interest in X-ray imaging of tissues with different densities and of soft tissues did not come from pure academic curiosity. What kindled it was a stretch of time when he was working as a part-time physicist in the radiology department of a hospital in Cape Town.

His part-time job in a hospital radiology department was the catalyst. It was a real-world problem in a clinical setting, rather than an academic research project.

Even after moving to a university position in the United States, computer-assisted tomography was not his main research focus, and he only worked on it intermittently when he had time. He published two papers on CAT in 1963 and 1964, but the academic response was minimal, so he suspended the work for a while. Only later, between 1970 and 1972, did he resume the effort and formalize the mathematical theory of CAT. Interestingly, one fact stands out — Hounsfield's first CAT scanner was not built on the basis of Cormack's mathematical theory. Furthermore, the Nobel Prize was not the crown of a lifelong obsession for Cormack. To underline this, after receiving the Nobel Prize in 1979, he completely retired from research and spent the rest of his life on his hobbies.

A unique case where research at the level of a side project led to a Nobel Prize. A trajectory different from many other Nobel laureates.

Hounsfield: A Company Researcher Without a Bachelor's Degree

Godfrey N. Hounsfield (1919-2004) was a British electrical engineer. His career can be summarized as follows — service in the British Royal Air Force from 1939 to 1945, a bachelor's degree from the Faraday House Industrial College in 1946, and then a stint as a researcher at the Thorn/EMI research laboratory from 1951 to 1984.

He only had a bachelor's degree from an industrial college, not a traditional university, and he got a job at EMI (Electric and Musical Industries). EMI was originally known for its music industry—the company is well-known as the Beatles' record label. Hounsfield performed computer-related development at the EMI research institute. It is a well-known anecdote that EMI supported Hounsfield's CT research with profits from the Beatles (although the actual cause and effect is debated, it is a symbolic story).

Around 1968, Hounsfield successfully developed a method of scanning objects inside the body with X-rays from multiple angles and reconstructing them in 3D using a computer. He approached the problem independently of Cormack's mathematics. In 1971, the world's first clinical CT scanner (the EMI scanner) scanned its first patient at a hospital in London. It was immediately used to diagnose brain tumors.

He is an exceptional case of a company researcher with no bachelor's degree from a traditional university winning the Nobel Prize in Physiology or Medicine. This is precisely the contrast noted earlier — a man who spent his entire career doing research at a company and won the Nobel Prize on that basis.

Radon Transform: The Power of Mathematics

The mathematical foundation of CT is the Radon transform, proposed by Johann Radon in 1917. Radon created this transform as a pure mathematics problem, and 60 years later, it became a key tool in medical imaging.

The idea behind the Radon transform:

• Forward: A 2D function f(x, y) is the density at each location. Integrating this function over lines at various angles yields "projections" at each angle.
• Inverse: Reconstruct the original function f(x, y) from the projection data at multiple angles.

CT is exactly this inverse transform:

• Forward: X-rays are passed through the body at various angles to measure the attenuation (integrating the density function of the body at various angles).
• Inverse: The density function of the body is calculated from this multi-angle data (reconstructing the cross-sectional image of the body).

The filtered back projection algorithm is the standard algorithm used in practical CT. Each projection at each angle is filtered according to the angle, back-projected, and then combined to reconstruct the image.

It is impossible without computers. Even a single CT slice requires processing data from thousands of angles, and with multiple slices, the amount of computation is enormous. The emergence of minicomputers in the early 1970s was a critical condition for the practical application of CT. The essence of this discovery is the convergence of mathematics, medicine, and computing.

Solving the Inverse Problem: A CS Framework

Now let's summarize the principles of CT in the language of computer science.

An inverse problem is a problem of calculating the cause or original from observed results. Examples include:

• Sound separation: Separating individual sound sources from a mixture of sounds.
• Image restoration: Restoring a sharp original image from a blurry image.
• Inverse lighting: Re-illuminating an image with different lighting conditions.
• CT/MRI and other medical imaging: Reconstructing a 3D body from various signal data.

Common features: The forward problem is natural (density → projection, 3D → 2D, etc.), while the inverse problem is often ill-posed—there may be multiple solutions, or it may be extremely sensitive to noise. To solve these problems, regularization or prior information must be used.

How CT solves the inverse problem:

• Ensuring information redundancy with data from multiple angles: A single angle is insufficient, but multiple angles allow the original 3D information to be uniquely determined.
• Utilizing mathematical consistency: The theory of the Radon transform guarantees that the inverse transform has a unique solution under certain conditions.
• Reducing noise with filtering: The filter portion of the filtered back projection algorithm ensures practical accuracy.

Inverse problems in the age of deep learning: Today, many inverse problems are solved with deep learning. GANs are used for image restoration, and diffusion models are used for conditional generation. The recent discovery that inverse problems can be solved with learning is a descendant of CT. Although the principles are different, the category of problems is the same.

Limitations of the analogy: Of course, CT is a very different deterministic approach from the inverse problem-solving methods of the deep learning era. However, the fundamental structure of "calculating the original from observed data" is exactly the same.

A Legacy That Continues Today

The discovery of CT has led to the development of various subsequent imaging technologies.

• MRI (Magnetic Resonance Imaging): Similar to CT in that it creates cross-sectional images, but uses magnetic fields instead of X-rays. It provides better visualization of soft tissues. Developed in the 1970s and clinically adopted in the 1980s (Nobel Prize in 2003 for Lauterbur and Mansfield).
• PET (Positron Emission Tomography): Creates images of tissue metabolism using radioactive tracers. It can be used to identify metabolically active tissue — for example, to determine whether a tumor is present in a patient's neck area.
• 3D Ultrasound: Also creates cross-sectional images using ultrasound, with 3D reconstruction from scans taken at multiple angles. It is a standard procedure in obstetrics and gynecology.
• Photoacoustic Imaging: Combines light and sound to create images of tissue oxygen saturation.
• PET/CT, PET/MRI Hybrid: Fusion imaging that combines multiple modalities.
• Deep Learning-Based Reconstruction: Allows for accurate CT reconstruction with fewer angles, reducing radiation exposure.
• Artificial Intelligence-Based Diagnosis: Automatic detection of lung cancer and liver diseases from CT images.

Why is it important?

What the two achieved is an empirical demonstration of the principle that "the convergence of mathematics and engineering can fundamentally change clinical diagnosis."

Before CT, medical diagnosis relied heavily on the visual interpretation of X-ray images. After CT, diagnosis became more precise with quantitative and three-dimensional imaging. This change forms the basis for today's precision medicine, surgical planning, and radiation therapy. The transition from qualitative observation to quantitative images is one of the major trends in medicine in the latter half of the 20th century.

The story of a university professor with only a master's degree and a company researcher without a bachelor's degree who made a world-changing medical advancement — the uniqueness noted repeatedly above is what makes this award so special. It is an empirical demonstration that degrees and experience are not essential conditions for the ability to make discoveries. It is often cited as one of the reasons for the easing of degree requirements in many universities and companies today.

Cormack's trajectory, where research at a part-time level led to a Nobel Prize — it is an example of the principle that continuous interest can yield unexpected results. His own recollection that "it was not something he had devoted his life to" remains both interesting and quietly chilling.

1979 Cormack & Hounsfield Summary: Cormack established the mathematical and physical foundations of CT (1960s-1970s), and Hounsfield created the world's first clinical CT scanner at EMI (1971). They solved the inverse problem of reconstructing 3D data from X-ray projection data taken from multiple angles using computers. This is the theoretical root of today's MRI, PET, 3D ultrasound, and deep learning image reconstruction.

→ Previous: 1978 — Arber, Smith, and Nathans → Next: 1980 — Benacerraf, Snell, and Dausset