Why existing signals were not yet images
By the mid-twentieth century, nuclear magnetic resonance had entered physics and chemistry laboratories. Researchers placed a sample in a magnetic field, excited atomic nuclei with radio waves of an appropriate frequency, and measured their response. Hydrogen nuclei were particularly useful for observing water-containing substances. Signal frequency, intensity and recovery behaviour could reveal a sample's chemical environment. A spectrum, however, described what properties a sample had; a new method was needed to put those properties back into their respective locations.
Medical research also created a need for spatial localization. In 1971, Raymond Damadian compared normal rat tissues with two malignant tumours and found marked differences in the measured tumours' magnetic relaxation times. This drew attention to the nuclear magnetic resonance properties of tissue water, but measuring differences in sampled tissue and locating abnormal regions within an intact body remained two different experimental tasks.
Gradients give frequency a position
Paul Lauterbur, a chemist at the State University of New York at Stony Brook, turned magnetic-field inhomogeneity, normally an obstacle to precise spectroscopy, into a means of localization. He deliberately made field strength vary gradually along one direction. Nuclei of the same kind at different positions encountered different fields, so their resonance frequencies varied with position. The received signal therefore carried spatial information.
Data acquired along one direction amounted to compressing the sample into a projection. Changing the gradient's direction relative to the sample and repeating the measurement produced multiple projections. Combining them through reconstruction made it possible to infer how the signal was distributed across a plane. Image detail depended jointly on gradients, signal measurement and reconstruction, so imaging did not have to rely on a very short radiation wavelength to resolve small structures, as ordinary optical imaging did.
References: [1]
The small-sample experiment published in 1973
On 16 March 1973, Nature published Lauterbur's two-page paper, Image Formation by Induced Local Interactions: Examples Employing Nuclear Magnetic Resonance. The date marks formal publication. Experimental objects included two narrow tubes filled with ordinary water and surrounded by heavy water. In the corresponding hydrogen-nucleus signal image, the two regions containing ordinary water could be distinguished separately.
This simple arrangement made it easier to test whether position encoding worked. The shapes and relative positions of the two regions were known beforehand and could be compared with the reconstruction. The paper made a material property's spatial distribution visible, and researchers subsequently extended the approach to biological samples. Scaling such small-sample experiments up to the human body still required solutions for body-sized magnets, signal quality, scanning time and motion interference.
Early image-making also bore the marks of manual laboratory work. Lauterbur's team initially recorded data on an oscilloscope and drew the images, later experimenting with lighter and darker typewriter characters to represent contrast. Students helped with photography and data-processing programs. As samples expanded from narrow tubes to small animals and plants, organizing many measurements into reliable images became work as important as building magnets and radio-frequency circuits.
Different advances in Nottingham and Zurich
In Nottingham, Britain, Peter Mansfield and Peter Grannell began from solid-state nuclear magnetic resonance. Their 1973 paper combined multipulse techniques with magnetic-field gradients to investigate diffraction-like signals. Using models made of equally spaced camphor sheets, they recorded signals and applied Fourier transforms to distinguish arrangements of three and five layers. The experiment clearly demonstrated a mathematical relationship between time signals and spatial structure.
In 1975, Anil Kumar, Dieter Welti and Richard Ernst published a Fourier imaging method: pulsed gradients applied across a series of measurements produced images through two- or three-dimensional Fourier transformation. Mansfield subsequently continued to advance rapid imaging. Different teams addressed spatial encoding, reconstruction efficiency and acquisition speed, gradually turning a feasible experiment into a usable examination technique.
From experimental signals to medical examinations
Bringing the imaging principle into hospitals required engineers, physicists and doctors to work together. Magnets had to supply a stable main field, gradient coils had to alter the field under programmed control, radio-frequency systems had to excite and receive weak signals, and computers had to reconstruct images. Breathing and other body movements changed acquisition conditions, making shorter measurements directly relevant to whether images could be used to observe abnormalities.
By the early 1980s, medical magnetic resonance equipment was entering clinical use. Magnetic fields and radio waves provided tissue information and new ways of observing the brain, spinal cord, joints and other soft tissues. Doctors could compare tissue contrast under different scanning conditions, identify the location of abnormalities and inform treatment. The two-page paper of 1973 became a key milestone because it added location to existing nuclear magnetic resonance measurements, ultimately allowing signals to be matched to human anatomy.