A question of “first”
There is no single first computer-based medical image. Several groups in different countries worked on related ideas during the 1960s and 1970s, and the technologies that resulted built on much older mathematics and physics. A fair way to describe the story is that computed tomography (CT) emerged in the early 1970s, magnetic resonance imaging (MRI) followed over the 1970s and 1980s, and both would be impractical without computers.
What makes these methods different from an ordinary X-ray photograph is that the scanner never records a finished picture. It records measurements, and a computer calculates the picture from them.
Why a computer is needed
A conventional X-ray flattens the body onto one image, so structures at different depths overlap. Tomography, from a Greek root meaning “slice,” aims to show a single cross-section instead. Earlier film-based methods blurred out everything except one plane. Computed tomography took a different approach.
In CT, an X-ray source and detectors rotate around the patient. At each angle the detectors record how much of the beam has passed through. No single reading says what lies along the path, only how much total absorption occurred. With many readings from many angles, though, it becomes possible to work backward to a map of how much each small region absorbed. That is a large set of simultaneous calculations, and doing it by hand would be unrealistic. The mathematical idea of recovering a function from its line integrals goes back to work by the Austrian mathematician Johann Radon in the early twentieth century, long before any scanner existed.
CT: engineering and theory
Two names are most often linked with CT. Godfrey Hounsfield, an engineer at the British company EMI, built an early working scanner in the early 1970s, and the South African-born physicist Allan Cormack had worked independently on the underlying theory and calculations in the 1950s and 1960s. The two shared the Nobel Prize in Physiology or Medicine in 1979. Other researchers explored similar reconstruction ideas in different fields, including astronomy and electron microscopy, and a more complete account would name them too.
Early scanners were slow by today’s standards. The reconstruction took substantial computing time, and the first systems imaged the head only. Over the following years, designs improved, scanning times fell and whole-body scanners became common in hospitals. The details of those improvements varied by manufacturer, and this article does not try to rank them.
MRI: signals, magnets and the Fourier transform
MRI does not use X-rays. It places the body in a strong magnetic field and uses radio waves to make certain atomic nuclei, mostly hydrogen in water and fat, emit signals. The signals are then spatially encoded using changing magnetic fields.
The raw data collected is not an image either. A computer applies a mathematical method, typically a form of the Fourier transform, to convert the recorded signals into a picture. Efficient algorithms for the Fourier transform were developed in the 1960s, and these made such calculations practical on available machines.
Paul Lauterbur and Peter Mansfield received the Nobel Prize in Physiology or Medicine in 2003 for discoveries that made MRI possible. Many researchers contributed to nuclear magnetic resonance and its medical use, and the question of who deserves credit has been argued over, notably in discussion of work by Raymond Damadian. The honest summary is that the field was built by many hands.
Other computer-assisted methods
CT and MRI are the best-known examples, but they are not the only ones.
- Ultrasound uses sound echoes, and modern scanners use computing to form and enhance images in real time.
- Positron emission tomography (PET) detects signals from a tracer in the body and reconstructs where it is concentrated, using similar principles to CT.
- Digital radiography replaced film with electronic detectors, so ordinary X-rays can be viewed, adjusted and sent to colleagues on a screen.
In each case, the key shift was the same: a sensor collects data, and software decides how to show it.
Storing and sharing images
Digital images created a new problem: where to keep them. Hospitals developed picture archiving and communication systems, usually called PACS, to store scans and deliver them to doctors’ screens. A common standard called DICOM helps equipment from different makers exchange images and related information. These tools depend on advances in data storage and networking, and they helped end the era of film libraries and lightbox walls, though many places continue to use a mix of old and new methods.
Software, judgment and limits
Because the image is calculated, it is shaped by choices made in software, including filters, contrast settings and the way slices are stacked or displayed. Images can contain artifacts, meaning features that come from the process rather than from the body, such as streaks near metal implants or blurring from patient movement. Trained radiologists learn to recognize these.
More recently, researchers have applied pattern-recognition methods, including neural networks, to help find abnormalities in scans or speed up reconstruction. This is an active area, and results vary with the data used to train the tools. Such systems are generally presented as aids to clinicians rather than replacements.
Imaging is also not risk-free or always appropriate. CT uses ionizing radiation, so doctors weigh benefits against exposure, while MRI involves strong magnets that can be hazardous around some implants. Scans can also reveal incidental findings that cause worry without improving health. Whether a given scan is right for a patient is a medical decision, not something this history can settle.
A lasting pattern
The path from CT to MRI shows a general feature of computing history: a hardware idea often waits for the arithmetic power to make it usable. The mathematics was sometimes decades older than the machines. When processors became fast enough and memory large enough, the concept moved from theory to hospital routine.