Content from this work may be used under the terms of the Creative Commons Attribution 4.0 licence. Any further distribution
of this work must maintain attribution to the author(s) and the title of the work, journal citation and DOI.
Published under licence by IOP Publishing Ltd
HAPP Centre: 10th Anniversary Commemorative Volume
Journal of Physics: Conference Series 2877 (2024) 012113
IOP Publishing
doi:10.1088/1742-6596/2877/1/012113
1
Shadows and signals: a brief history of medical imaging
Martin J. Graves
Department of Radiology, University of Cambridge, Addenbrooke’s Hospital,
Cambridge Biomedical Campus, Hills Road, Cambridge CB2 0QQ, UK
E-mail: mjg40@cam.ac.uk
Abstract. Medical imaging is a revolutionary field that has transformed healthcare, providing
clinicians with the ability to view the inside of the body without invasive procedures. This
overview traces the evolution of the major medical imaging modalities from their inception to
the advanced technologies used today. Starting with the serendipitous discovery of X-rays by
Wilhelm Conrad Röntgen in 1895 the subsequent developments in various imaging modalities
including fluoroscopy, computed tomography (CT), magnetic resonance imaging (MRI), nuclear
medicine and ultrasound imaging are chronicled. The technological breakthroughs, pioneering
scientists and clinicians involved, and the significant impact of these advancements on diagnostic
medicine are discussed. Additionally the integration of digital imaging, artificial intelligence
(AI) and machine learning technologies into medical imaging is considered together with their
potential to revolutionise diagnostic procedures, enhance patient care and streamline healthcare
delivery.
1. X-rays: The dawn of medical imaging
The history of medical imaging can be traced back to 1895 when Wilhelm Conrad Röntgen, a German
Professor of Physics at the University of Würzburg accidentally discovered X-rays. Whilst performing
experiments with a cathode ray tube covered by a shield of black cardboard he found that a piece of
barium platinocyanide paper several feet away fluoresced. Further experimentation showed that there
was a form of invisible light coming from the cathode ray tube. Since Röntgen could not explain this
new type of rays he termed them ‘X-rays’, although colleagues subsequently named them Röntgen rays,
a name that is still used in many languages. On Friday 8th November 1895 he performed the first in
vivo X-ray of his wife’s hand, showing that these X-rays could pass through soft tissues but were stopped
by bones and metal, creating a shadow on a photographic plate (figure 1) [1].
The attenuation of X-ray beams as they pass through a material occurs due to the absorption of X-
ray photons by the atoms in the material. The degree of attenuation depends on the material's thickness,
density, atomic number and the X-ray's energy. Different tissues in the body attenuate X-rays differently,
allowing for the creation of contrast in the X-ray images. X-rays were originally captured and viewed
directly on photographic film, offering high spatial resolution but requiring chemical processing to
develop and fix the film. The images were then viewed on lightboxes. With the development of digital
imaging, and picture archiving and communications systems (PACS) there was a drive to eliminate film
from the process. Computed radiography (CR) was invented by the Japanese company Fuji in the early
1980s. CR uses a cassette-based system with photostimulable phosphor plates that store the latent image,
which is then scanned and digitised. Subsequently digital radiography (DR) was developed which
captures images directly onto a digital detector, providing immediate access to the image, leading to a
https://creativecommons.org/licenses/by/4.0/
HAPP Centre: 10th Anniversary Commemorative Volume
Journal of Physics: Conference Series 2877 (2024) 012113
IOP Publishing
doi:10.1088/1742-6596/2877/1/012113
2
faster and more efficient workflow compared to CR. A typical DR equipped X-ray room is shown in
figure 2.
The discovery of Röntgen/X- rays earned Röntgen the first Nobel Prize in Physics in 1901 with the
citation “in recognition of the extraordinary services he has rendered by the discovery of the remarkable
rays subsequently named after him”. Within a month surgeons in Europe and the US were using
radiographs to guide their operations and within six months radiographs were being used in the
battlefield to locate bullets in wounded soldiers. Although Röntgen refused to take out patents on his
discovery, many others claimed patents on improving the process and practicalities of creating X-rays.
Early commercial entities included Siemens & Halske, a German engineering company that
subsequently became part of Siemens.
Figure 1. First medical X-ray by Wilhelm Röntgen of his
wife Anna Bertha Ludwig's hand with a ring. By Wilhelm
Röntgen - Source 1: [1]. Source 2: [2], Public Domain,
https://commons.wikimedia.org/w/index.php?curid=123547
09 .
Figure 2. A modern DR X-ray room showing the X-ray tube
and a digital detector for a standing patient. Ptrump16
(https://commons.wikimedia.org/wiki/File:Dedicated_ches
t_x-ray_room.jpg), Added annotation by Martin Graves,
https://creativecommons.org/licenses/by-sa/4.0/legalcode.
https://commons.wikimedia.org/w/index.php?curid=12354709
https://commons.wikimedia.org/w/index.php?curid=12354709
HAPP Centre: 10th Anniversary Commemorative Volume
Journal of Physics: Conference Series 2877 (2024) 012113
IOP Publishing
doi:10.1088/1742-6596/2877/1/012113
3
2. Fluoroscopy: Watching the body in motion
On 5th February 1896 both Italian scientist Enrico Salvioni (with his "cryptoscope")[2] and William
Francis Magie of Princeton University (with his "Skiascope")[3] independently developed live X-ray
imaging apparatuses employing barium platinocyanide screens. Concurrently American inventor
Thomas Edison determined calcium tungstate to be the most efficient fluorescent material. By May 1896
he had produced the first commercially viable device, known as the "Vitascope," later renamed the
fluoroscope[4]. X-ray fluoroscopy allowed real-time imaging of the body's internal structures, making
it possible to view movements and processes such as the digestive system in action. However, early
fluoroscopy exposed patients and operators to high doses of X-ray radiation, necessitating the
development of safer practices and equipment.
Modern real-time X-ray imaging is called radiographic fluoroscopy (RF). Many RF techniques
involve the administration of an iodine-based contrast media into the body that absorbs X-rays. For
example, the imaging of blood vessels known as angiography requires a small catheter to be inserted
into an arterial blood vessel and manipulated to the anatomy of interest, e.g. the origin of the aorta to
visualise the coronary arteries. The X-ray contrast media is then injected through this catheter and the
physician can monitor the movement of the contrast media through the vessels and identify any
blockages, stenosis or abnormalities in real-time (figure 3). RF can also be used to guide catheters or
other instruments through the vessels for diagnostic or therapeutic purposes. A barium swallow is a
dedicated test of the pharynx, oesophagus and proximal stomach. In this case the patient drinks a thick
white, chalky liquid containing barium sulphate that also absorbs X-rays. Using RF the clinician can
view the movement of the barium through the pharynx and oesophagus as the patient swallows.
3. Computed tomography (CT): A new dimension
The major limitation of conventional X-ray imaging is that the images are two-dimensional (2D)
projections (shadows) of the tissues through which the beam passes. In 1959 William Oldendorf, an
American neurologist conceived the idea of “scanning a head through a transmitted beam of X-rays and
being able to reconstruct the radiodensity patterns of a plane through the head”. Oldendorf had the idea
by watching an engineer who was working on an automated apparatus to reject frostbitten fruit by
detecting dehydrated portions. By 1961 he had completed a working prototype, using materials found
in his home such as his son's toy train, a phonograph turntable and a spring motor from an alarm clock.
He used a narrow beam of 𝛾-ray radiation from a 131I source that was detected by a sodium iodide (NaI)
Figure 3. A single frame from a real-time X-ray
coronary angiography procedure showing a partial
occlusion (arrowed) of the left circumflex
coronary artery. Todt T, Maret E, Alfredsson J,
Janzon M, Engvall J, Swahn E
(https://commons.wikimedia.org/wiki/File:Coron
ary_angiography_of_a_STEMI_patient,_showing
_partial_occlusion_of_left_circumflex_coronary_
artery.jpg), „Coronary angiography of a STEMI
patient, showing partial occlusion of left
circumflex coronary artery“, https://creative
commons.org/licenses/by/2.0/legal code .
HAPP Centre: 10th Anniversary Commemorative Volume
Journal of Physics: Conference Series 2877 (2024) 012113
IOP Publishing
doi:10.1088/1742-6596/2877/1/012113
4
crystal-photomultiplier with the scintillations counted by a ratemeter [5]. His test object comprised a
block of plastic into which two concentric but irregularly spaced, rings of nails were inserted. The inner
ring mimicked the brain and the outer ring the human skull. Oldendorf demonstrated that when the test
object was rotated at 16 rpm and the centre of rotation moved at approximately 80 mm/hr, the acquired
projection data showed that the nails positioned in the centre of the object were clearly discernible from
the outer ring of nails. Oldendorf was successfully granted a patent for his idea on the 8th October 1963
[6]. Oldendorf spent three frustrating years trying to gain industrial interest in his original CT work [7].
He quit trying in 1964 when an X-ray equipment manufacturer told him that "we cannot imagine a
significant market for such an expensive apparatus which would do nothing but make a radiographic
cross-section of a head" [8].
Oldendorf’s 1961 paper contained no mathematical detail of his method, but around the same time
Allan MacLeod Cormack, a South African/American nuclear physicist was working on a side project to
develop the mathematical underpinnings of a method to determine how a detailed picture of the internal
structure of the human body could be obtained from projections created by X-ray or 𝛾-ray radiation
passing through the body. He published his work in two landmark papers in 1963 [9] and 1964 [10].
During the period 1967-19368 Godfrey Hounsfield, an electrical engineer working in the Central
Research Laboratories (CRL) of the UK company EMI Ltd, also conceived of an idea to determine the
contents of a three-dimensional (3D) box from a set of random readings taken through the box.
Hounsfield simplified the reconstruction problem by considering the 3D object in the box as a stack of
2D slices rather than a 3D volume. Based upon his theoretical work Hounsfield generated an initial
project proposal in 1968, however, there was no evidence that his concept for a 3D X-ray scanner would
be useful clinically or even viable commercially. Furthermore EMI CRL was not prepared to provide
any funding unless Hounsfield and his team could convince the Department of Health and Social
Security (DHSS) to also provide some funding. On the basis that the DHSS required a preliminary
technical report to consider funding Hounsfield wrote a new proposal in August 1968 requesting internal
funding of £20,000 to perform the necessary work. The proposal was rejected but at the end of October
1968 Hounsfield was given £5,000 by EMI based on a verbal agreement from the DHSS that they would
also contribute to the work. With the limited funding available Hounsfield like Oldendorf used a 95Am
𝛾-ray source and a NaI crystal-photomultiplier detector but set up his test rig using an old lathe bed. The
various test objects were translated across the 𝛾-ray beam and then rotated by 1°. This translate-rotate
process was then repeated until the test object had been rotated 180°. The measured data was punched
onto paper tape and fed into a large mainframe computer to perform a simple iterative reconstruction.
The first test objects were initially simple Perspex and aluminium objects, followed by specimens of
cows’ brains and pigs’ bodies obtained from an abattoir and ultimately a preserved section of a human
brain. In July 1969 with a promise of further funding from the DHSS EMI increased the budget to
£17,393 with the DHSS providing £5,000 in December 1996. These initial results were very encouraging
and in January 1970 a decision was made to build a prototype head-only machine to acquire clinical
images. The DHSS provided a further £7,419 in March 1970. There was, however, insufficient funding
to build a prototype and EMI refused not only to fund any further work, but also demanded that their
money to date be repaid. The issue was resolved when Hounsfield and his manager William (“Bill”)
Ingham persuaded Gordon Higson of the DHSS to place an order for four yet unbuilt scanners. The
prototype was to be installed at Atkinson Morley’s Hospital in Wimbledon (figure 4) and the three
clinical systems were to be placed at the Hospital for Neurology and Neurosurgery in London,
Manchester and Glasgow. The plan was that income from these machines would then fund a further
system for Hounsfield and his team to continue their development work. The DHSS would also fund
half of the remaining research costs in exchange for a small royalty on sales [11]. The DHSS probably
made over £1 million in royalties.
The first clinical images were obtained using the prototype system on the 1st October 1971, although
the images were not seen until the following day as the data had to be taken away and reconstructed on
a mainframe computer back at the EMI CRL. Fortunately the minicomputer was being developed around
the same time and a Data General Nova 820 minicomputer with 32K of memory, a 2.5-MB two-sided
HAPP Centre: 10th Anniversary Commemorative Volume
Journal of Physics: Conference Series 2877 (2024) 012113
IOP Publishing
doi:10.1088/1742-6596/2877/1/012113
5
hard drive, a reel-to-reel magnetic tape storage device and a printer was added to the system.
Improvements to the image reconstruction were also developed including replacing Hounsfield’s
iterative algebraic reconstruction method with a much faster filtered back-projection algorithm.
At the 1972 annual meeting of the Radiological Society of North America (RSNA) Hounsfield and
James Ambrose, the neuroradiologist at Atkinson Morley’s Hospital gave a presentation on
“Computerized Axial Tomography” immediately following the RSNA President’s address. The scanner
was also displayed in the commercial exhibition. The event was such a success that EMI took multiple
orders for the scanner, each with a $100,000 non-refundable deposit [12].
Hounsfield and EMI had filed a patent on 23rd August 1968 [13] based upon the lathe-bed work.
The US patent citing Oldendorf’s 1963 patent [6] was filed on 27th December 1971 [14], but not granted
until 11th December 1973 during which time the claims had been modified in light of Cormack’s 1963
work. Hounsfield and Ambrose first described the system at the Annual Congress of the British Institute
of Radiology in April 1972 [15] with two full articles published in 1973 with Hounsfield providing a
technical description of the scanner (Hounsfield 1973) and Ambrose the clinical applications (Ambrose
1973).
Hounsfield continued to develop the scanner through several subsequent generations including the
development of whole-body imaging and continual improvements in spatial resolution, acquisition time
and true 3D volumetric acquisitions. By mid-1975 EMI had installed one hundred and sixty machines
with an average cost of $400,000 [16].
Hounsfield received numerous awards, honorary degrees and decorations for his work. In 1967 he
was awarded a CBE and in 1975 became a Fellow of the Royal Society. In 1979 he and Allan Cormack
shared the 1979 Nobel Prize for Physiology or Medicine "for the development of computer assisted
tomography". In 1981 Hounsfield was knighted by the Queen. He retired from EMI in 1986 and passed
away on 12th August 2004 [17].
Throughout the 1980s and 1990s computerised tomography (CT) technology became more
sophisticated. Scanners with even higher resolutions and faster scanning times were developed. The
introduction of 'slip ring' technology allowed continuous rotation of the X-ray source and detectors
(figure 5), which led to the development of spiral or helical CT in the late 1980s. This was a major
development, allowing for entire organs to be imaged in a single breath-hold, vastly improving the
comfort for patients and the quality of the images. The turn of the century saw the introduction of
multislice CT scanners, capable of capturing multiple slices simultaneously, further reducing scan times
and improving spatial resolution (figure 6).
Figure 4. EMI CT brain scanner installed
at Atkinson Morley's Hospital,
Wimbledon, London, UK in 1971 (the first
used clinically) by EMI, Hayes, Middlesex,
1970-1971. Case courtesy of Raphael
Ambros,
Radiopaedia.org. From the case rID: 85113 .
HAPP Centre: 10th Anniversary Commemorative Volume
Journal of Physics: Conference Series 2877 (2024) 012113
IOP Publishing
doi:10.1088/1742-6596/2877/1/012113
6
Figure 5. Basic principles of X-ray computerised tomography. In (A) the X-ray tube and detector rotate
about the patient. At each rotation angle, e.g., 𝜃1, the X-ray projection (𝑟1) is sampled and is shown as
a line intensity in the sinogram (dotted blue line). As the CT rotates the sinogram (B) is built up from
the r profiles at each angle 𝜃.
Today CT technology continues to evolve [18]-[20]. Advances in detector technology, image
reconstruction algorithms and machine learning are making scans faster and safer (by reducing radiation
dose) and providing functional as well as morphological information. Three-dimensional
reconstructions and CT angiography are now commonplace, and the integration of PET (Positron
Emission Tomography) with CT has led to PET/CT scanners that can show both anatomic and metabolic
activity within the body.
Figure 6. CT scan of a patient who has
suffered a traumatic injury to their head. A)
shows an axial slice through the brain
demonstrating a severe intracranial bleed
(arrow). B) shows the same slice but with the
window level and width changed to
demonstrate the skull. A depressed fracture
of the bone can be seen (arrow). C) is a
reformat in the coronal plane of the axial
slices through the head. The fracture is seen
more clearly (arrow). D) is a 3D rendering of
the CT data producing a shaded surface
display of the skull where the 3D anatomy of
the skull fracture is demonstrated (arrow).
HAPP Centre: 10th Anniversary Commemorative Volume
Journal of Physics: Conference Series 2877 (2024) 012113
IOP Publishing
doi:10.1088/1742-6596/2877/1/012113
7
The most recent technological development is photon-counting CT where a new generation of highly
sensitive detectors based on cadmium telluride (CdTe), cadmium zinc telluride (CZT) or silicon (Si) can
count each individual photon that passes through the body and measure its energy [21]. This allows
further reduction in radiation dose, enhanced image contrast and the capability to perform material-
specific imaging. This latter ability enhances the diagnosis and characterisation of various pathologies
by allowing clinicians to differentiate between, for example, the constituents of atherosclerotic plaque
or the composition of kidney stones.
4. Magnetic resonance imaging (MRI): A magnetic leap forward
Magnetic resonance imaging (MRI) was developed in the late 1970s and early 1980s, and provided
superior soft tissue contrast compared to CT. The principles of MRI are completely different from any
other imaging modality. MRI is based upon the phenomenon of nuclear magnetic resonance (NMR) that
was discovered by Felix Bloch working at Stanford [22] and Edward Purcell working at Harvard [23]
almost simultaneously in 1946. They were subsequently awarded the 1952 Nobel Prize in Physics for
“the development of new methods for nuclear magnetic precision measurements and discoveries in
connection therewith”. The Boston Herald reported that Purcell’s discovery ‘wouldn’t revolutionize
industry or help the housewife’. Bloch, a Swiss-born Jew and friend of quantum physicist Werner
Heisenberg, quit his post in Leipzig in 1933 in disgust at the Nazis’ expulsion of German Jews (as a
Swiss citizen, Bloch himself was exempt). Bloch’s subsequent career at Stanford was crammed with
major contributions to physics and he has been called the father of solid-state physics. The initial concept
for the medical application of NMR originated with Raymond Damadian in 1971. He and his colleagues
at the State University of New York Health Science Center, who were starved of mainstream research
funding, even went so far as to design and build their own superconducting magnets operating in their
Brooklyn laboratories. The first human NMR image is attributed to them. NMR imaging of two tubes
of water using a magnetic field gradient was first demonstrated by Paul Lauterbur at the State University
of New York at Stony Brook in 1973. His seminal paper “Image formation by induced local interactions,
examples employing nuclear magnetic resonance” [24] was originally rejected. However, thirty years
later Nature placed this work in a book of the twenty one most influential scientific papers of the 20th
century.
MRI primarily looks at the nucleus of hydrogen, i.e. a single proton. Since a typical adult is
approximately 60% water and 16% fat it is possible to create images which are sensitive to the protons
within water and fat molecules. MRI uses three magnetic fields to create images.
The first is a strong typically 1.5 T or 3 T static magnetic field (known as 𝐵0) that creates a weak net
nuclear magnetisation within the human body. Such strong magnetic fields are achieved using
superconducting magnets, although lower field strength magnets, e.g., ≲ 0.4 T, are available using rare-
earth permanent magnets. The net magnetisation precesses around 𝐵0 with a characteristic frequency
given by 𝜔0 = 𝛾𝐵0. This is known as the Larmor frequency after Joseph Larmor who in 1897 (long
before NMR was discovered) proposed that charged particles should precess about a magnetic field and
that the frequency of precession should be directly proportional to the field strength. In the case of NMR
the constant of proportionality is 𝛾, the gyromagnetic ratio, which is 42.47 MHz for the nucleus of 1H.
At 3 T, for example, the Larmor frequency is 128 MHz, i.e. in the radiofrequency (RF) part of the
electromagnetic spectrum.
Since the net magnetisation is aligned along the same direction as 𝐵0 (which shall be defined as the
z-direction here) it is necessary to rotate the magnetisation so that it is aligned orthogonal to 𝐵0. This is
achieved by applying a short duration ‘pulse’ of a second external magnetic field (known as 𝐵1) that is
also orthogonal to 𝐵0 and is alternating with a frequency that is equal to the Larmor frequency, i.e. it is
resonant with the precessional frequency of the net magnetisation. The 𝐵1 field is created using a
transmitting RF antenna, more usually called a transmit coil. For example, a simple solenoidal coil
position at right angles to 𝐵0 would create an alternating magnetic field orientated along the length of
the coil as described by Fleming’s right hand thumb rule. The desired rotation of the net magnetisation
about the axis of 𝐵1 is controlled by the amplitude and duration of this transmitted RF pulse. Once the
HAPP Centre: 10th Anniversary Commemorative Volume
Journal of Physics: Conference Series 2877 (2024) 012113
IOP Publishing
doi:10.1088/1742-6596/2877/1/012113
8
RF pulse has been switched off the precessing transverse magnetisation induces an electromotive force
(emf) in the solenoidal coil, which would now be considered a receive coil. This emf or signal is then
amplified and digitised to create the MRI raw data. Immediately after the RF excitation the net
magnetisation will return to thermal equilibrium via processes known as relaxation. The net transverse
magnetisation will decay exponentially with a time constant called T2, whilst the magnetisation will
recover along the z-direction with a time constant known as T1. It is the variation in T1 and T2 relaxation
times in different tissues that is exploited in MRI to create images with differential signal intensities
between tissues, e.g. the grey and white matter in the brain. In biological tissues the T1 relaxation time
is approximately an order of magnitude longer than the T2 relaxation time. Combinations of transmitted
RF pulses are used to weight the amplitude of the signal induced in the receive coil to highlight either
T1 or T2 relaxation processes or other contrast mechanisms (figure 7).
Figure 7. Some clinical MRI examples. A) is a whole-body MRI scan representing the distribution of
the protons in water and B) an image showing the distribution of the protons in fat. C) is a T1-weighted
image of a patient with a brain tumour (arrow) that is difficult to identify. D) is a T2-weighted image at
the same location showing a high signal from the tumour. E) is a single frame from a cine movie of the
heart as it beats. The blood in the right ventricle (RV) and the left ventricle (LV) is bright whilst the
surrounding heart muscle (myocardium) is darker. F) shows a sagittal view through a knee. The femur
is at the top and the tibia at the bottom with the meniscus between them. The arrow is pointing to a torn
meniscus.
The third and final magnetic field used in MRI is a spatially varying magnetic field superimposed on
𝐵0. To form an MR image it is necessary to have three linear magnetic field gradients aligned along the
physical x, y and z directions. These gradient magnetic fields are created using another set of coils inside
the bore of the magnet. These gradients are switched on and off to spatially encode the MR signal
induced in the receiver coil. It is the switching of these gradient pulses that gives rise to the characteristic
knocking noises heard during an MRI examination. There are three logical gradients used in a typical
MRI acquisition referred to as slice selection, phase encoding and frequency encoding. Each logical
gradient is mapped to the relevant physical axis depending upon the orientation of the acquisition plane,
HAPP Centre: 10th Anniversary Commemorative Volume
Journal of Physics: Conference Series 2877 (2024) 012113
IOP Publishing
doi:10.1088/1742-6596/2877/1/012113
9
i.e. axial, coronal and sagittal. Single and double oblique orientations may also be obtained by
appropriate rotation of the logical gradients. The standard method of acquiring MRI images known as
‘spin-warp’ uses gradient pulses to encode position into the frequency and phase of the MR signal. This
means that the images can be reconstructed by a straightforward Fourier transformation.
Most of the initial development work for MRI was performed in the UK with four main groups taking
delivery of whole-body air-cored resistive magnets around the same time in 1977. Three were academic
comprising John Mallard at the University of Aberdeen, and Peter Mansfield and Raymond Andrew
leading separate groups in Nottingham. The final system was for EMI led by Hugh Clow. The first in
vivo MR images started to appear in 1977. Damadian produced an image of the chest using a very
different approach that did not involve magnetic field gradients, known as the FONAR (Field fOcused
Nuclear mAgnetic Resonance) method [25]. Because the uniformity of Damadian’s magnetic field was
so poor the FONAR method utilised a small volume in the centre of the magnet where the field produced
the desired Larmor frequency. The patient was then physically translated through this volume to build
up the image. The academic groups initially created their images using a line scanning method, whilst
the EMI group implemented a projection reconstruction method as originally used in the EMI CT
scanner, producing their first successful image of the head in September 1978. The next major
development was the so-called ‘spin-warp’ method invented at the University of Aberdeen (William
Edelstein, James Hutchison et al. 1980). This technique rapidly became the basis for most MRI
acquisitions and remains so to this day.
The Medical Research Council (MRC) initially funded both Nottingham groups and the Aberdeen
group, whilst EMI once again approached the DHSS and the insightful Gordon Higson. By late 1978
the DHSS had agreed to contribute £350,000 on a cost sharing basis with EMI to develop a system
around a 0.3 T superconducting magnet to be installed in the Royal Postgraduate Medical School
(RPMS) at Hammersmith Hospital after Atkinson Morley declined to take the system ostensibly due to
the number of visitors a new system would attract. The ‘Neptune’ system was installed in the RPMS in
January 1981 with images appearing shortly thereafter [26]. In October 1979 EMI merged with Thorn
Electrical Industries Ltd to form Thorn EMI. In 1980 Thorn EMI sold the CT business to the US General
Figure 8. The original Aberdeen Mark 1 MRI scanner.
Note how the patient is positioned between rather than
through the magnet windings. Image taken in the Suttie Art
Space in Aberdeen Royal Infirmary where the device has
been on display since February 2016. AndyGaskell, CC
BY-SA 4.0 , via Wikimedia Commons.
HAPP Centre: 10th Anniversary Commemorative Volume
Journal of Physics: Conference Series 2877 (2024) 012113
IOP Publishing
doi:10.1088/1742-6596/2877/1/012113
10
Electric (GE) company and the Neptune project to Picker International Ltd. Picker began in 1909 when
New York druggist James Picker first supplied Kodak X-ray plates and accessories to local hospitals. In
1958 the Picker family sold the business to CIT Financial Corporation who then sold Picker to the RCA
Corporation in 1980 who in turn shortly afterwards sold it to the UK General Electric Company (GEC)
(not to be confused with the US GE). GEC combined Picker with several other of its subsidiaries to
form Picker International. In 1999 the company changed its name to Marconi Medical Systems as GEC
attempted to rebrand itself under the Marconi name to differentiate itself from the US GE. In 2001 Royal
Philips Electronics of the Netherlands acquired Marconi Medical Systems. The other commercial
venture was a spin-out of the work conducted at the University of Aberdeen when John Mallard formed
a small company, M&D Technology Ltd, to sell a commercial 0.08 T version of the original 0.04 T G
air-cored resistive magnet [27] (figure 8). This system was unique in that the magnet was oriented so
that 𝐵0 was vertical and the patient was positioned between the two middle solenoid coils making up
the magnet. Three systems were sold, one to Edinburgh Royal Infirmary, one to a private clinic in
Geneva and one to St Bartholomew’s Hospital (Barts) in London (the system on which the author started
his MRI career in 1984). The Barts’ system is now in the Science Museum in London.
In 2003 the Nobel Prize for Physiology or Medicine was jointly awarded to Lauterbur and Mansfield
“for their discoveries concerning magnetic resonance imaging”. This was much to the chagrin of
Raymond Damadian who also felt that he should have been included.
MRI technology continues to rapidly evolve with substantial developments in key MR system
technologies. From a hardware perspective the standard MRI system field strength is 1.5 T with an
increasing adoption of 3 T MRI system in clinical practice and research, as well as the development and
regulatory approval of 7 T systems for enhanced signal-to-noise ratio (SNR) and image quality.
Advancements in semiconductor power devices have enabled improvements in MR gradient and RF
amplifier hardware, contributing to higher performance and lower energy consumption. New image
acquisition methods include fast imaging sequences and techniques to reduce image acquisition times
like parallel imaging (PI), compressed sensing (CS) and simultaneous multi-slice (SMS) imaging. There
are developments in image reconstruction methods incorporating artificial intelligence (AI) and deep
learning (DL) algorithms, workflow automation and advancements in methods to obtain quantitative
imaging-based biomarkers of disease. Like CT there are now hybrid systems such at PET/MR and MR
systems integrated with linear accelerators for image-guided radiotherapy. A comprehensive review of
the many developments in MRI can be found in the article by Kabasawa [28].
5. Nuclear medicine and positron emission tomography (PET)
Nuclear medicine uses small amounts of radioactive materials to diagnose and to treat various diseases.
These radioactive materials are introduced into the body by injection, swallowing or inhalation. The
radiopharmaceuticals are designed to target specific organs, bones or tissues where they emit 𝛾-rays that
can be detected by special types of cameras (such as gamma cameras or PET scanners). This process
allows the creation of detailed images that show the structure and function of organs and tissues. The
tissue specificity arises from the range of radiation-emitting radionuclides, which can be used to label
specific biomarkers, biochemicals and pharmaceuticals without disturbing their biological function. In
addition the radiation can be detected above the low natural radiation background. These two aspects
are fundamental to realising the tracer principle for which George de Hevesy received the Nobel Prize
in Chemistry in 1943 "for his work on the use of isotopes as tracers in the study of chemical processes".
Nuclear medicine is unique because it provides medical information that is often unavailable through
other imaging procedures and it can detect diseases at an earlier stage. It is used to diagnose and to
manage a wide range of conditions, including cancers, heart disease, gastrointestinal, endocrine,
neurological disorders and other abnormalities within the body.
5.1. Single photon emission computed tomography (SPECT)
The emitted 𝛾-rays are detected by a gamma camera that was invented in the 1950s by Hal O. Anger,
an American engineer and biophysicist. Anger developed the first gamma-camera or Anger-camera as
HAPP Centre: 10th Anniversary Commemorative Volume
Journal of Physics: Conference Series 2877 (2024) 012113
IOP Publishing
doi:10.1088/1742-6596/2877/1/012113
11
it was known then in 1957 [29]. The camera uses a large crystal scintillator usually made of sodium
iodide (NaI) doped with thallium, which scintillates (glows) when struck by 𝛾-rays. The size and
thickness of the crystal can affect the camera's sensitivity and resolution. Attached to the crystal are
multiple photomultiplier tubes (PMTs) that detect the very weak light flashes (scintillations) from the
crystal and convert them into electrical signals. The arrangement and number of PMTs can influence
the spatial resolution of the image. In front of the crystal a collimator made of lead or another dense
material with holes drilled through it is used to ensure that only gamma rays travelling in certain
directions reach the crystal. This improves image clarity and contrast. Modern gamma cameras, now
more commonly referred to as a single-photon emission computed tomography (SPECT) cameras,
utilise digital systems for signal processing and image reconstruction (figure 9). These systems can
handle high data rates, allowing for dynamic imaging sequences and enhanced image quality. Advanced
algorithms are used for image reconstruction, analysis and display. This includes methods for correcting
image artifacts, enhancing contrast and quantifying radiotracer uptake. A very common radioisotope
used in nuclear medicine imaging is technetium-99m (99mTc) that emits a 140 keV 𝛾-ray. There are
numerous different ligands that can be combined with 99mTc. The ligand is chosen to have an affinity
for the specific organ to be targeted. For example, the exametazime chelate of 99mTc can cross the blood–
brain barrier and flow through the vessels in the brain for cerebral blood flow imaging. Other ligands
include sestamibi for myocardial perfusion imaging and mercapto acetyl tri-glycine (MAG3) to evaluate
renal function.
Many tracers have very short half-lives, i.e. the time for the radioactivity to decay by a half, so a
local (on-site) method of creating the radioisotope is necessary, e.g. 𝛾-ray emitting radioisotopes are
produced in generators. Generators provide a local supply of a short-lived radioactive substance from
the decay of a longer-lived parent radionuclide; hence they can be transported over longer distances. For
example, 99mTc with a half-life of 6 hours can be locally extracted from the radioactive decay of
molybdenum-99 (99Mo) that has a half-life of 66 hours.
5.2. Positron emission tomography (PET)
PET is the most specific and sensitive means for imaging molecular interactions and pathways within
the human body. The specificity arises from the range of positron-emitting radionuclides, which can be
used to label specific biomarkers, biochemicals and pharmaceuticals without disturbing their biological
function. The most common PET tracer is 18F-FDG (Fluorodeoxyglucose) that acts as a glucose
analogue and is used to localise tissues with altered glucose metabolism. 18F decays by emitting a
Figure 9. A dual-headed single-photon
emission computed tomography
(SPECT) camera circa 2006.
User:Drahreg01 (https://commons.wik
imedia.org/wiki/File:SiemensEcamDu
et.JPG), „SiemensEcamDuet“,
https://creativecommons.org/licenses/
by-sa/3.0/legal code .
HAPP Centre: 10th Anniversary Commemorative Volume
Journal of Physics: Conference Series 2877 (2024) 012113
IOP Publishing
doi:10.1088/1742-6596/2877/1/012113
12
positron that annihilates with an electron, forming two 511 keV photons produced approximately 180°
apart. Other useful PET radioisotopes include gallium-68 (68Ga), rubidium-82 (82Rb) and carbon-11
(11C) for chelation with specific ligands.
Positron-emitting radionuclides are commonly produced in very expensive charged particle
accelerators, e.g. linear accelerators or cyclotrons [30]. The production of 11C with a half -life of 20.4
minutes will require a local cyclotron, whereas 18F has a half-life of 109.7 minutes and can be shipped
from a commercial supplier further afield. On-site production of tracers requires sophisticated
production facilities that use automated kits provided by the cyclotron companies with the tracers
synthesised in shielded containment chambers known as hot cells.
A PET scanner comprises a ring of lutetium (Lu)-based scintillation detectors and multiple PMTs or
solid-state detectors that detect the photon scintillations from the crystal and convert them into electrical
signals. The near simultaneous (typically within a timing window of 6 to 12 ns) detection of the two
photons means that it is possible to localise their source along a straight line of coincidence (also called
the line of response or LoR). In practice the LoR has a non-zero width as the emitted photons are not
exactly 180° apart. Pairs of photons that do not arrive within the timing window are ignored. If the
resolving time of the detectors is less than 500 ps it is possible to localise the event to a segment of a
chord whose length is determined by the detector timing resolution. As the timing resolution improves
the signal-to-noise ratio (SNR) of the image will improve, requiring fewer events to achieve the same
image quality.
PET images are reconstructed by coincidence events being grouped into projection images called
sinograms. The sinograms are sorted by the angle and for 3D images the tilt of the LoR. The sinogram
images are analogous to the transmission projections acquired by a CT scanner and can be reconstructed
in a similar way (figure 10). The main confounding factors are the requirements to correct for the effect
of scatter of the photons and the attenuation of photons by the tissue.
Figure 10. A) shows four coincidence detections in a PET scintillator and the lines of response from a
point source of a positron-emitting radionuclide. B) shows the associate sinogram as a function of the
angle 𝜃 and distance r from the centre of the PET ring. C) shows the results of back projecting the LoRs.
The resultant image of the point source is blurred because a filtered back projection algorithm was not
used.
Although the development of positron emitting radioisotopes detected initially by pairs of detectors
can be traced back to 1951 [31], the first PET system that could make quantitative measurements of the
regional tissue concentration of a PET tracer was built by Michel Ter-Pogossian, Michael Phelps,
Edward Hoffman and Nizar Mullani around 1974 at Washington University [32] with Department of
Energy (DoE) and National Institutes of Health (NIH) support.
HAPP Centre: 10th Anniversary Commemorative Volume
Journal of Physics: Conference Series 2877 (2024) 012113
IOP Publishing
doi:10.1088/1742-6596/2877/1/012113
13
Figure 11. Images from a 18F-FDG PET/CT examination. A) shows a 3D rendering of the bones from
the CT acquisition. B shows a projection of all the data in the PET acquisition using an inverted
greyscale. C) shows a single slice from the PET acquisition showing a large tumour in the patients’ liver.
There is high activity (dark ring around the tumour) around the margins of the tumour where it is rapidly
growing (blue arrow). The centre of the tumour is necrotic, i.e. dead, and appears as low activity (bright
centre). D) shows a fusion of the PET data (in shades of red) fused with the CT data (in grayscale)
clearly showing the regions of uptake. Note a high signal in the patient’s rib (red arrow).
Ter-Pogossian is recognised to have "led the research that turned the positron emission tomography
(PET) scanner from an intriguing concept to a medical tool used in hospitals and laboratories
everywhere." [33]. Phelps, who is often credited with inventing PET, received the 1998 Enrico Fermi
Presidential Award for his work. The first whole-body PET scanner appeared in 1977.
PET cameras are nowadays integrated with a CT scanner, creating a PET/CT. This has two main
advantages. Firstly, the relatively low spatial resolution PET image is fused with the higher resolution
CT image to show accurately where in the human body the PET tracer is localised. Secondly, the
information from the CT scan can be used to determine the attenuation of the photons as they pass
through the body, thereby improving the quantification of the tracer (figure 11).
Currently PET scanners record around 1% of the coincident pairs of emitted photons. Extending the
length of the PET detector ring would substantially increase the system sensitivity. A commercial total-
body PET system is now available with an axial field-of-view (FOV) of 200 cm compared to a typical
current generation system with a FOV of 25 cm, improving sensitivity by a factor of 40 [34].
6. Ultrasound: Listening to the body
In the 1940s and 1950s ultrasound emerged as a new imaging modality. Ultrasound uses high-frequency
sound waves greater than 20 kHz to create images of the inside of the body such as a fetus during
pregnancy. The advantage of ultrasound is that it does not use ionising radiation, making it safer for
certain applications particularly in obstetrics. The first published work on medical ultrasonics was by
Karl Theodore Dussik in Austria in 1942 on transmission ultrasound investigation of the brain [35].
Much of the development of ultrasound in clinical practice was initiated by Ian Donald, Regius Professor
of Midwifery at the University of Glasgow who together with his colleagues began a series of studies
that would firmly establish a role for ultrasound in clinical diagnostics. Donald initially attempted to use
an ultrasound device that was used in the Glasgow shipyards for testing welds in large pressure vessels
HAPP Centre: 10th Anniversary Commemorative Volume
Journal of Physics: Conference Series 2877 (2024) 012113
IOP Publishing
doi:10.1088/1742-6596/2877/1/012113
14
to distinguish between fibroids and cysts. Tom Brown, a research and development engineer with Kelvin
& Hughes Ltd who made the ultrasound flaw detector, offered to help. Brown realised that some form
of pictorial imaging was needed and believed that it might be possible to make radar-like images of
internal organs. Brown conceived and designed the low-cost prototype which was to be the first direct
contact ultrasound scanner and had it built onto a borrowed hospital bed table in the firm's workshops.
The prototype was made available to Donald, assisted by John MacVicar, in early 1957. They quickly
realised its potential and began exploring its clinical applications, publishing their first paper in June
1958 [36].
Figure 12. A) shows the basic principles behind ultrasound. A pulse of ultrasound from the transducer
travels through the tissue at an approximate velocity (c) of 1540 ms-1. At a tissue interface a proportion
of the ultrasound energy (the echo) is reflected to the probe whilst the remainder continues to be
transmitted and will be proportionately reflected at the next tissue interface. The distance travelled by
the ultrasonic pulse is given by the velocity (c) multiplied by the time (t) between the pulse and the echo
divided by 2. B) shows an ultrasound image of the liver and right kidney. C) shows a radiologist using
an ultrasound machine. Since ultrasound is a real-time imaging method the radiologist views the images
whilst they move the probe over the patient’s body as shown in D.
An ultrasound machine has a transducer probe that is used to send and to receive pulses of sound
waves. When the probe is placed on the skin it emits pulses of high-frequency sound waves that
propagate through the body and when they encounter different tissues (such as organs and blood vessels)
some of the sound waves are reflected while others continue to travel further until they hit other
boundaries and are also reflected. These reflected signals or echoes are processed to create real-time
images or videos of the internal structures of the body. Image reconstruction involves algorithms that
calculate the distance from the probe to the tissue boundaries by considering the speed of sound in the
tissues and the time it took for the echoes to return. The amplitude of the returning echoes is highly
dependent on the density and composition of tissues that the sound waves have encountered, creating
contrast in the image, e.g. distinguishing between different types of tissues, fluids and other structures
(figure 12).
The dynamic nature of ultrasound images allows for the observation of the structure and movement
of the body's internal organs, as well as blood flowing through blood vessels. It can be used for a variety
of diagnostic purposes, including examining the heart, kidneys, liver, blood vessels, pregnant uterus and
other organs, guiding procedures such as biopsies and assessing fetal development during pregnancy.
Ultrasound is preferred for many conditions because it does not use ionising radiation, making it safer
than X-rays, particularly for pregnant women and the developing fetus. It is versatile, relatively low cost
and can be performed at the bedside.
HAPP Centre: 10th Anniversary Commemorative Volume
Journal of Physics: Conference Series 2877 (2024) 012113
IOP Publishing
doi:10.1088/1742-6596/2877/1/012113
15
Advancements in ultrasound technology include the development of compact ultrasound probes that
can connect to smartphones or tablets, effectively transforming them into portable ultrasound devices
(figure 13). This innovation allows for real-time imaging to be performed almost anywhere, making
diagnostic medical imaging more accessible, especially in remote areas or in situations where traditional
ultrasound machines are not available. These mobile-connected ultrasound devices are being
increasingly used in diverse medical fields for quick assessments, emergency diagnostics and in settings
with limited access to conventional imaging facilities.
7. Digital imaging and advances
The late 20th and early 21st centuries have seen significant advances in digital imaging techniques,
enhancing the quality and accessibility of medical images. Digital technology has also facilitated the
development of telemedicine, allowing for remote diagnostics and consultation. Artificial Intelligence
(AI) has started and will continue to play a significant role in modern medical imaging, revolutionising
how images are acquired, reconstructed and interpreted, leading to improvements in how diseases are
diagnosed and how patient care is delivered [37]. AI algorithms can analyse medical images more
quickly and accurately than traditional methods. Through training on huge volumes of annotated data,
they can detect patterns and anomalies that might be missed by the human eye and have shown high
accuracy in identifying various conditions, including cancers, neurological disorders and cardiovascular
diseases. They can help in reducing diagnostic errors, potentially resulting in earlier diagnosis and
improving patient outcomes. AI can streamline medical imaging workflows, reducing the time
radiologists spend on image analysis. Automated report generation and prioritisation of cases based on
urgency are examples where AI aids in managing the clinical workload more effectively. AI systems
are increasingly integrated with hospital electronic health records (EHRs) to provide a more
comprehensive view of a patient's medical history, improving diagnostic accuracy and facilitating
personalised medicine.
However, despite its potential the integration of AI into medical imaging faces challenges, including
regulatory hurdles and the need for large annotated datasets for training as well as concerns about data
privacy and security. Furthermore, ensuring AI models are transparent, explainable and unbiased
remains a priority to gain trust from healthcare professionals and patients alike.
8. Conclusion
From the discovery of X-rays to the advent of AI in imaging the field of medical imaging has undergone
remarkable transformations. Each advancement has opened new possibilities for diagnosis and
treatment, significantly impacting patient care. As technology continues to evolve the future of medical
Figure 13. A pocket-sized ultrasound
device. Images are displayed on a
smartphone via a device-generated
wireless connection. Under a Creative
Commons (CC) license permission to
publish this picture was obtained from
HEALCERION, Seoul, Korea (S2 File).
https://doi.org/10.1371/journal.pone.01
85031.g001 .
https://doi.org/10.1371/journal.pone.0185031.g001
https://doi.org/10.1371/journal.pone.0185031.g001
HAPP Centre: 10th Anniversary Commemorative Volume
Journal of Physics: Conference Series 2877 (2024) 012113
IOP Publishing
doi:10.1088/1742-6596/2877/1/012113
16
imaging promises even greater innovations, including the development of more sophisticated AI
algorithms that can provide real-time diagnostics, the integration of imaging data with genomic
information for personalised medicine, and the creation of less invasive imaging techniques that offer
higher resolution and clearer insights into the human body. The convergence of imaging technology
with other fields such as nanotechnology and biotechnology may lead to breakthroughs in early detection
and treatment of diseases at a molecular level. Moreover, the increasing accessibility of advanced
imaging technologies in remote areas through mobile and cloud-based solutions will be likely to
democratise healthcare, making high-quality diagnostics available to a broader population. Ultimately
the ongoing integration of AI and other innovative technologies in medical imaging holds the promise
of transforming healthcare delivery, making it more predictive, personalised and accurate.
References
[1] Röntgen W K 1895 Über eine neue Art von Strahlen : vorläufige Mitteilung Sitzungsber. Phys.
Med. Gesell.
[2] Rowland S and Hammond Smith H 1896 Report on the application of the new photography to
medicine and surgery The British Medical Journal 1 874
[3] Oestreich A E 1995 Professor William F. Magie and the American discovery of the fluoroscope,
1896 AJR Am. J. Roentgenol. 165 1060
[4] Anonymous 1896 The Edison X ray experiments, apparatus and fluoroscope Sci. Am. 74 219
[5] Oldendorf W 1961 Isolated flying spot detection of radiodensity discontinuities—displaying the
internal structural patterns of a complex object IRE Trans. Biomed. Electron. 8 68
[6] Oldendorf W 1963 US Patent No. 3,106,640
[7] Oldendorf W 1991 The continued quest for an image of brain J. Neuroimaging 1 46
[8] Oldendorf W H 1980 The quest for an image of brain (New York: Raven Press).
[9] Cormack A M 1963 Representation of a function by its line integrals, with some radiological
applications. I J. App. Phys. 34 2722
[10] Cormack A M 1964 Representation of a function by its line integrals, with some radiological
applications. II J. App. Phys. 35 2908
[11] Beckmann E C 2006 CT scanning the early days Br. J. Radiol. 79 5
[12] Schulz R A et al. 2021 How CT happened: the early development of medical computed
tomography Journal of Medical Imaging, 8
[13] Hounsfield G N 1968 UK Patent No. 1,283,915
[14] Hounsfield G N 1971 US Patent No. 3,778,614
[15] Ambrose J and Hounsfield G N 1973 Computed transverse tomography Br. J. Radiol. 46 148
[16] Strong A B and Hurst R A 1994 EMI patents on computed tomography: history of legal actions
Br. J. Radiol. 67 315
[17] Isherwood I 2004 Sir Godfrey Hounsfield – Nobel Laureate Eur. Radiol. 14 2152
[18] Booij R et al. 2020 Technological developments of X-ray computed tomography over half a
century: User's influence on protocol optimization Eur. J. Radiol. 131 109261
[19] Lell M M and Kachelriess M 2020 Recent and upcoming technological developments in
computed tomography: High speed, low dose, deep learning, multienergy Invest. Radiol. 55 8
[20] Hsieh J and Flohr T 2021 Computed tomography recent history and future perspectives J Med
Imaging (Bellingham) 8 052109
[21] Willemink M J et al. 2018 Photon-counting CT: Technical principles and clinical prospects
Radiology 289 293
[22] Bloch F et al. 1946 Nuclear induction Physical Review 69 127
[23] Purcell E M et al. 1946 Resonance absorption by nuclear magnetic moments in a solid Physical
Review 69 37
[24] Lauterbur P C 1973 Image formation by induced local interactions: examples employing nuclear
magnetic resonance Nature 242 190
[25] Damadian R et al. 1977 NMR in cancer: XVI. FONAR image of the live human body Physiol.
HAPP Centre: 10th Anniversary Commemorative Volume
Journal of Physics: Conference Series 2877 (2024) 012113
IOP Publishing
doi:10.1088/1742-6596/2877/1/012113
17
Chem. & Phys. 9 97
[26] Higson G R 1987 Seeing things more clearly Br. J. Radiol. 60 1049
[27] Mallard J R 2006 Magnetic resonance imaging-the Aberdeen perspective on developments in the
early years Phys. Med. Biol. 51 R45
[28] Kabasawa H 2022 MR imaging in the 21st Century: Technical innovation over the first two
decades Magn. Reson. Med. Sci. 21 71
[29] Anger H O 1958 Scintillation Camera Rev. Sci. Instrum. 29 27
[30] Saha G B et al. 1992 Cyclotrons and positron emission tomography radiopharmaceuticals for
clinical imaging Semin. Nucl. Med. 22 150
[31] Wrenn F R et al. 1951 The use of positron-emitting radioisotopes for the localization of brain
tumors Science 113 525
[32] Ter-Pogossian M M et al. 1975 A positron-emission transaxial tomograph for nuclear imaging
(PETT) Radiology 114 89
[33] Wackers F J T 2018 Michael M. Ter-Pogossian (1925-1996) J. Nucl. Cardiol. 25 1090
[34] Jones T and Townsend D 2017 History and future technical innovation in positron emission
tomography J Med Imaging (Bellingham) 4 011013
[35] Dussik K T 1942 On the possibility of using ultrasound waves as a diagnostic aid. Z Neurol
Psychiatr 174 153
[36] Donald I et al. 1958 Investigation of abdominal masses by pulsed ultrasound Lancet 1 1188
[37] Najjar R 2023 Redefining radiology: A review of artificial intelligence integration in medical
imaging Diagnostics (Basel) 13