PET

POSITRON EMISSION TOMOGRAPHY

PET: POSITRON EMISSION TOMOGRAPHY

• The goal of PET is to generate images of the distribution of

positron emitters in vivo.

• PET systems rely on the detection of annihilation gamma

rays that follow positron decay.

• The gamma rays are detected in coincidence by detectors

that surround the patient.
PET HISTORY

• PET has been in existence since the 1970s due in large

part to the pioneering work of Michael Phelps, PhD,
Michel Ter-Pogossian, PhD, and others in the fields of
medical physics and nuclear medicine.

• Its clinical utilization began with myocardial perfusion in

1995 and rose dramatically in 1998, with the evaluation of
solitary pulmonary nodules and initial staging of lung
cancer.
PET HISTORY
• 1998 saw the creation of the first PET/computed
tomography (CT) hybrid system, and in 2001, such
systems became commercially available.

• The evolution of PET from its beginnings as an instrument

of research to its present day wide and growing use in
cancer, cardiac, and neurological imaging has resulted in
instrumentation that is making a major impact in clinical
care.
Fundamentals of PET Imaging
• PET is based on the physical properties of certain radioactive isotopes
known as positron emitters.

• A positron is the antimatter conjugate of an electron and

has the same mass as an electron but positive charge .

• Although simplified, it is convenient and sufficient to think of a β +

particle as a positively charged electron.

• These radionuclides emit positrons rather than gamma photons when

they undergo radioactive decay.
Fundamentals of PET Imaging

• FYI: A negatively charged beta particle, sometimes

called a negatron, or β− particle, is identical to an
electron except that its origin is the nucleus rather than
the electron cloud surrounding the nucleus.

• β− particles do not play a role in PET.

Positron/Beta Plus Decay

• Radioactive decay via positron emission is at

the heart of the PET image formation
process.

• Positron decay is a type of beta decay in which a

positively charged particle, known as a beta+ particle
(denoted β+), is emitted from a proton-rich nucleus as
that nucleus attempts to become more stable.
Positron/Beta Plus Decay
• In beta plus (β+) decay, a proton-rich parent nucleus P transforms a
proton into a neutron and ejects a positron e + and an electronic
neutrino ve.

• Thus, in β+ decay, the atomic number of the daughter decreases by

one, i.e. ZD = ZP – 1, the atomic mass number, just as in β– decay,
remains constant, i.e. AD = AP, and the general relationship for β+ decay
is written as:
Positron/Beta Plus Decay
• Radionuclides undergoing β+ decay are often called positron emitters
and are used in medicine for functional imaging with the special
imaging technique PET.

• The most common tracer for PET studies is fluorodeoxyglucose

(FDG) labelled with 18F which serves as a good example of β+ decay:
Positron/Beta Plus Decay
• As with β decay, positrons are emitted from the
nucleus with different energies.

• These energies have a continuous spectrum and

a specific maximum value.

• These energies have a continuous spectrum and

a specific maximum value that is characteristic
of the parent isotope.
Gamma Rays
• When a nuclear reaction or spontaneous nuclear decay occurs, the process may leave
the product (daughter) nucleus in an excited state.

• The nucleus can then make a transition to a more stable state by emitting a γ ray
photon and the process is referred to as γ decay.

• The energy of the photon emitted in γ decay is characteristic of the nuclear energy
transition, but the recoil of the emitting atom produces a spectrum centred on the
characteristic energy.

• Gamma rays typically have energies above 100 keV and wavelengths less than 0.1 Å.
Emission of Gamma Rays in PET
• Once emitted from the nucleus, the positron propagates through the
surrounding human tissue and undergoes scattering interactions, changing
its direction and losing kinetic energy principally by Coulomb interaction.
• When the positrons reach thermal energies, they interact with electrons by
the formation of a hydrogen-like orbiting pair called positronium.
• It reaches these thermal energies within a short distance which is
dependent on the energy of the positron, of which distance is a function of
the parent isotope and is typically on the order of a millimetre.

Physics of positron decay and annihilation, which results in two 511-keV

gamma rays.
Emission of Gamma Rays in PET
• Positronium (the combination of a positron and an electron) is
unstable and eventually decays, via annihilation, into a pair of
anti-parallel 511-keV photons, equivalent to the rest masses of the
two original particles.

• Conservation of momentum, which is close to zero immediately

before annihilation, ensures both photons are emitted almost
exactly 180° apart.

• These characteristic photon emissions (known as annihilation

radiation) — always 511 keV, always emitted simultaneously and
almost exactly 180° apart — form the basis of PET and result in
distinct advantages over single photon imaging in terms of
defining the LOR.
• Positrons emitted from a
radioactive nucleus propagate
through the surrounding material
before eventually coming to rest
a short distance from their site of
emission.

• At this point, the positron

annihilates with an electron,
creating two 511 keV photons
that are emitted approximately
180° apart.

• The perpendicular distance from

the line defined by the two
photons to the site of positron
emission places a limit on the
spatial resolution that can be
Question?
• So do PET scanners image positrons?
• The answer is no.
• Unlike conventional nuclear medicine imaging with gamma-emitting
radionuclides, the photons imaged in PET do not directly come from the nuclei
that are undergoing decay.

• Unlike conventional nuclear medicine imaging with gamma-

emitting radionuclides, the photons imaged in PET do not directly
come from the nuclei that are undergoing decay.
• Nor are the positrons being imaged. Because positrons are
particles that carry a positive charge, they travel only a very
short distance, usually no more than a millimetre or two, before
encountering a negatively charged electron.
Emission of Gamma Rays in PET
• When a positron and
electron collide, the
particles are
annihilated, and
according to the
conservation of matter
and energy, the
annihilation of the
electron and positron
results in the creation
of two high energy
gamma photons that
travel approximately
Emission of Gamma Rays in PET
• These high energy coincident annihilation photons produced by the annihilation
reaction is approximately 511keV, which is much greater than the less energetic
140 keV photons emitted by 99mTc.

• They are detected efficiently by a conventional gamma camera, and a specialized

ring of detectors is used.

• Their simultaneous detection using short timing intervals is called coincident

detection.

• Photon pairs that do not arrive at opposite points along the PET detector ring at
the same time (within a few nanoseconds) are ignored by the PET scanner.

• This action is called discrimination and helps improve localization of true

coincident events.
Attenuation, Absorption, and Scatter
• The images generated by PET scanners are accurate representations of the objects
being analyzed, but there are several factors that can degrade image quality.

• Absorption and scatter are two such factors.

• In nuclear medicine, attenuation refers to the decrease in intensity of a photon

signal as it passes through matter either by absorption or by scatter.

• Attenuation effects are directly proportional to the density and thickness of the
various tissues through which photons travel; that is, the more dense and thick a
tissue is, the more it will attenuate.

• If the matter through which a photon is traveling stops the photon completely, it is
called absorption.
Scatter
• Scatter refers to the alteration in the direction of a
photon’s path due to its interaction with matter (e.g.
tissues) along that path.

• Attenuation, Absorption and Scatter effects are related,

and both give rise to image reconstruction errors that can
adversely affect the accuracy of a PET scan.
Positron Emission Tomography Radioisotopes and
Radiopharmaceuticals

These radionuclides have fewer neutrons than their

nonradioactive, stable, counterparts; that is, stable carbon
has 12 nucleons, stable nitrogen has 14, stable oxygen has
16, and stable fluorine has 19 nucleons.
• The relative paucity of neutrons within these radionuclides results in
protons that are closer together, and repel one another making their
nuclei unstable.

• This repulsion and instability in a proton-rich nucleus is the basis for

positron decay, in which a positively charged particle leaves the nucleus
and a proton becomes a neutron.

• Another characteristic is the short half-life (t1/2) of these positron emitters.

• Today, the most widely used PET radiopharmaceutical is 18F-2-fluoro-2-

deoxy-D-glucose, also known as fluorodeoxyglucose (FDG).

• FDG has very similar structure and biochemical behaviour to glucose.

• Note the structural similarity between the FDG and glucose
molecules illustrated in Figure 1.3.
• The dissimilarity, although subtle, between these molecules
allows for very powerful information to be obtained from an FDG-
PET scan.
More Examples of Radiopharmaceuticals used in PET
Patient Preparation and Radiation
Safety: FDG
• There are a few important points to consider in preparing patients for an FDG-PET scan.
• Because FDG is a glucose analogue, it is necessary that patients avoid any caloric intake for
at least 4–6 hours prior to the study.
• Typically, patients are asked to fast overnight for morning appointments, and have only a
light breakfast for afternoon appointments.
• Serum glucose is routinely measured prior to FDG injection, and fasting levels are typically
70–110 ng/dl, which are ideal for an FDG-PET scan.
• Serum glucose levels greater than 200 ng/dl may result in significant changes in FDG
distribution, and patients with such levels are usually not scanned until better control is
obtained, because hyperglycaemia leads to competitive inhibition of FDG uptake into cells.
• Hyperinsulinemia is also a problem because it results in increased FDG uptake into skeletal
muscle.
• Fasting results in low (basal) insulin levels.
• Diabetic patients should not have regular insulin administered subcutaneously within 4 hours
of having FDG administered.
• After administration of FDG, patients must wait a period of at least 40–45 minutes
prior to scanning.

• This period is referred to as the uptake phase and is the necessary amount of time
for the FDG to be adequately biodistributed and transported into the patient’s cells.

• Patients are asked to rest in a quiet room, devoid of distractions, and they are also
asked to keep their movements, including talking, at an absolute minimum.

• This minimizes physiologic uptake of FDG into skeletal muscle, which can confound
interpretation of the scan.

• Patients should be comfortable and relaxed.

• The important consideration with respect to diabetics is that they refrain from insulin
administration for at least 4–6 hours for reasons described previously.
• In most cases, venous access can be obtained without difficulty by trained
technologists or venipuncturists.

• However, venous access can be quite difficult to obtain in small children, obese
patients, the elderly, and patients being treated with chemotherapy, etc.

• Although not ideal, and to be avoided, rather than cancel an FDG-PET scan for
lack of venous access, FDG can be administered orally in liquid form followed by
water.

• If this situation should arise, it is recommended that you discuss options regarding
oral FDG administration with your nuclear medicine or radiology specialist
• Patients are administered 140 μCi/kg of FDG with a minimum of 10mCi and a maximum of 20 mCi.

• Fixed doses of 15 mCi or 20 mCi can also be used.

• Although the photons created following a positron–electron annihilation are very high energy, the radiation dose from an
FDG-PET scan is less than one might expect, for two major reasons.

• First, the physical half-life for 18F is short at only 110 minutes.

• Second, the biological half-life of FDG is also relatively short, and it is excreted rapidly by the kidneys and eliminated in the
urine.

• Roughly 50% of the administered dose is present in the urine of those with normally functioning kidneys after about 2 hours.

• The combination of these two factors results in a relatively low effective t1/2.
For estimated radiation exposure following intravenous administration of FDG
• FDG-PET can have a significant impact on the treatment plan of a
patient with malignancy, and for this reason the relative radiation risk of
obtaining the scan is considered negligible.
• However, pregnant women should avoid undergoing an FDG-PET scan.
• FDG does cross the placenta, and will be distributed within the fetal
brain and be excreted by the fetal kidneys.
• The mother will also excrete the FDG into her bladder, which increases
the radiation dose to the nearby fetus.

• It is recommended that clinicians consult with their radiologists and

nuclear medicine specialists to determine whether another imaging
modality may be employed to answer the clinical question in a woman
who is pregnant.
• An FDG-PET scan can certainly be obtained postpartum if necessary.
• Breast-feeding is not recommended for 10 hours after administration of
this FDG
• Some PET centers recommend the use of muscle relaxants, cleansing
bowel preparations, and the placement of a Foley catheter in some
patients undergoing Man FDG-PET scan.
• Muscle relaxants and anxiolytics (e.g. diazepam) are thought to
minimize patient anxiety and lessen potential interference caused by
skeletal muscle uptake.
• Bowel preparations have been employed in the hope of lessening or
eliminating physiologic bowel activity which is often seen.
• Foley catheters, diuretics (e.g. furosemide), and intravenous fluids have
been used in various protocols to minimize the interference from
excreted activity in the genitourinary tract which could possibly obscure
adjacent disease.
• These steps have met with mixed results, are often unwelcome by
patients, and are controversial in some circles.
• Therefore, many other PET experts believe in a non-invasive, “keep it
simple” approach to PET imaging where none of these adjuncts is used.
Standardized Uptake Value

• Standardized uptake values (SUVs), are sometimes used in FDG-PET reports

in order to impart some semi-quantitative measurement of the degree of
FDG accumulation to areas of suspicion.

• The SUV is a unitless ratio that can be understood as the concentration of

FDG within a lesion divided by the concentration of radiotracer distributed
throughout the body.

• Mathematically, it can be expressed as follows:

• SUV = C (T)/(dose injected/body weight)

where C is the tissue concentration of FDG at time T.

• An SUV is a simplified index of FDG uptake and provides a relative
indication of the degree of metabolism within the lesion being
evaluated.

• The SUV measurement is directly proportional to metabolic activity.

• SUV can be notated as the maximum value within a lesion

(SUVmax), or the average value within a region of interest drawn
around a lesion (SUVavg).

• The SUVmax is more robust because it is more reproducible, being

less affected by the size and placement in the region of interest.
• Because the SUV is affected by multiple factors and is subject to
error, it should be used with caution.
• These factors include extravasation of radiotracer which alters whole-
body distribution, patient obesity (some advocate using lean body
mass rather than body weight in SUV determination for this reason),
time interval between FDG administration and scanning, size of the
region of interest used to make the SUV calculation, and serum
glucose level.

• SUV is used most frequently when evaluating a solitary pulmonary

nodule, where an SUV of 2.5 or greater within that nodule is
considered suspicious for malignancy, and an SUV less than 2.5
favours a benign, usually inflammatory condition.
• Another major use of SUV is in the follow up of cancer after therapy.
• The SUV provides a semi-quantitative index for determining the effect
of therapy.
PET IMAGE RESOLUTION
Positron Range Variation
• Although the radial distribution of annihilation events is
sharply peaked at the site of positron creation, a
calculation of the radius that includes 75% of all
annihilation events gives a realistic comparison of the
impact of the maximum positron energy on the spatial
resolution of PET imaging.
PET IMAGE RESOLUTION
Positron Momentum Variation
• Normally, the annihilation of the gamma rays is
expected to be anti-parallel, however the variation in
momentum of the positron results in an angular
uncertainty in the direction of the 511-keV photons that
is approximately 4 mrad (0.23°). This is referred to as
non-colinearity.
PET IMAGE RESOLUTION
DETECTOR POINT SPREAD FUNCTION
• For a detector composed of small discrete crystals, all interactions are
assumed to occur at the center of individual crystals for the purpose of
back projection and image reconstruction.
• As a result, the PSF for such detectors is similar to a step function with
a total width equal to the size of a crystal. The coincident PSF is,
therefore, a triangular function whose base width is again equal to a
crystal size.
• Thus, the FWHM of the coincident detector PSF is one-half the crystal
size.
PET IMAGE RESOLUTION
Parallax Error
• This results from the uncertainty of the depth of interaction (DOI) of
the gamma rays in the crystal.

• Gamma rays travel some (unknown) distance in the crystal (or

adjacent crystals) before being completely absorbed.

• As a result, if the gamma ray enters the crystal at an oblique angle, the
location of the interaction will not be the same as the point of entry
into the crystal the crystal of the interaction may not even be the same
as the one first entered.
PET IMAGE RESOLUTION
Parallax Error

• Thus, unless the DOI within a crystal can be accurately determined, an

incorrect line of response (LOR) will be assigned to this interaction
because the LOR is normally assigned to a position at the front of the
crystal of interaction.

• The parallax effect worsens as the source position moves radially

away from the center of the scanner because a larger fraction of the
gamma rays enter the crystals at oblique angles.
Parallax Error
Parallax Error.
• The gamma ray (solid
line) interacts in a crystal
after penetrating one or
more adjacent crystals in
the detector ring.

• Without depth-of-
interaction information,
the detection electronics
will incorrectly assign the
line of response (the
dotted line) based on the
front of the interaction
Key Terms

• Coincidence efficiency is related to count sensitivity and

should be as close to 100% as possible.

• Energy resolution is related to scatter rejection and the

closer it is to 0%, the better it is at rejecting scatter.

• Decay time is a measure of how long the scintillation light

persists; longer times limit count rate capability
Coincidence Events Measured in a PET Scanner

Three kinds of coincidence events that the tomograph

accepts:
[Link] Coincidences are those in which one or both
gamma rays scatter within the patient.
Coincidence Events Measured in a PET Scanner

2. Random/Accidental Coincidences are those in which

two separate decays result in the detection of only one
gamma ray from each one and the two events are close
enough in time to be in coincidence.
Coincidence Events Measured in a PET Scanner

3. True coincidences are those in which gamma rays are

detected from a single decay that have not scattered in
the patient.
Coincidence Events Measured in a PET Scanner

True coincidences

Block diagram of a basic PET scanner with illustration of events in

coincidence.
Coincidence Events Measured in a PET Scanner
• The goal is to measure and reconstruct the distribution of true coincidences while
minimizing the scattered and random coincidences and correcting for the bias
(but not necessarily the noise) associated with scattered and random
coincidences.

• True and scattered events are referred to as prompt events because they come
from the decay of a single nucleus and, thus, the gamma rays are detected
almost simultaneously.

• The prompt rate (trues + scatters) is related linearly to the activity in the patient.

• However, the randoms rate increases as the square of the activity in the patient
and becomes more dominant at higher activity levels.
Coincidence Events Measured in a PET Scanner

• Increasing the number of true coincidences leads to

less noise in the image and allows one to reconstruct
the data with high spatial resolution of the distribution
of decay events, given the physical limitations already
discussed.
Options to Increase Detected Coincidence Events in a PET study
Increase the patient dose
• Increasing the patient dose is only advantageous for 2D systems but
not a practical approach because higher doses would result in higher
radiation exposures to the patient.
Use more efficient scintillators/detectors
• This would improve sensitivity through high light output and fast decay.
Use more of the energy spectrum
• This is done by using more of the energy spectrum by accepting events
with energy less than the photopeak
Increase solid angle
• This is done through techniques like the removal of all axial collimation
(3D volume imaging), reducing ring diameters, and extending the axial
dimension of the detector array.
PET SYSTEM GEOMETRY
• The amplifiers
integrate the detector
signals.
• The pulse-height
analyzer (PHA)
selects the energy
range of events to be
accepted.
The basic components of the electronics system for•a The timing
tomograph.
discriminators
generate the timing
• The signals from the PMTs are integrated.

• In a BGO block system, which typically has many PMTs and readout channels,
the major contribution to dead time is from these integrating amplifiers.

• The amplified signals are routed to a pulse-height analyzer (PHA) to select

the energy range for the events to be processed.

• The signals also go to timing discriminators to generate timing signals for

the coincidence system.

• Normally, the timing discriminators include lower-energy discriminators to

reject events that have too low an energy to be passed by the PHA.

• Thus, the amplifiers see the full detector count rate, which is generally much
higher than the output of the timing discriminators.
• The actual electronics in a full tomograph are more
complicated than indicated above.

• There are modules to decode the event position, to

correct for the variance in photopeak position in various
detector elements, to collect data for dead-time
correction, and to sort the data and store it on disk.

• Major aspects of the system design are the details of

the implementation of the coincidence electronics and
the method to perform randoms correction.
Data Acquisition
1. Coincidence Processing
• The basis of coincidence detection is that pairs of related 511 keV annihilation
photons can be associated together by the detector system based upon their times
of measurement.

• Two photons detected within a short time interval are assumed to have arisen from
the same positron–electron annihilation and a coincidence event is recorded.

• The time interval determining when events are considered to be coincident is

denoted 2τ and is a system parameter that is not usually adjustable by the user.
• In order to minimize random coincidences, this interval should be kept as short
as possible.

• For typical BGO based systems, 2τ may be around 12 ns.

Shorter time windows are made possible by detector materials
such as LSO that have faster scintillation decay times and,
thus, better time resolution.

• Further reductions in the coincidence window are limited by

differences in the arrival times of the two photons.
• When a photon is incident upon a PET detector, an electrical
pulse is generated.
• A constant fraction discriminator then produces a digital logic
pulse when the detector signal reaches a fixed fraction of the
peak pulse height.
• This digital logic pulse is defined to have a duration τ and is fed
into a coincidence module that determines whether it forms a
coincidence event with any of the signals from other detectors in
the ring.
• A coincidence event is indicated if there is an overlap in time
between separate logic pulses from two different detectors.
• Coincidence circuits allow two
photon detection events to be
associated with each other
based upon their arrival times.
• Photons detected at A and B
produce signals that are
amplified and analysed to
determine whether they meet
the energy acceptance
criteria.
• Those signals that fall within
the energy acceptance window
produce a logic pulse (width τ)
that is passed to the
coincidence processor.
• A coincidence event is
• Coincidence detection assumes that only two photons were detected, but
with multiple disintegrations occurring concurrently, it is possible for
three or more photons to be recorded by separate detectors within the
coincidence time window.

• When this occurs, it is unclear which pair of detectors corresponds to a

legitimate coincidence and multiple events of this sort are often
discarded.

• This circumstance is most likely to occur when there is a large amount of

activity in or around the FOV, and it contributes to count loss at high
count rates.

• Another possible scenario is that only one photon is detected within the
coincidence time window and no coincidence event will be recorded
Data Acquisition Geometries
• The data acquisition geometry refers to the
arrangement of detector pairs that are permitted
to form coincidence events and, in practice,
involves the presence or absence of interplane
septa.

• Data acquisition with septa in place is referred to

as 2‑D mode; data acquisition without any
interplane septa is referred to as 3‑D mode.

• The 2‑D/3‑D designation refers to the acquisition

geometry rather than the resulting images as both
modes produce similar volumetric images.
• In 2‑D acquisition mode, an array of septa is inserted between the
detector rings.
• These septa are annular and are typically made of tungsten.
• The purpose of the septa is to physically absorb photons incident at large
oblique angles relative to the transverse plane, allowing only those
photons incident approximately orthogonal to the z axis of the scanner.
• These septa differ significantly from gamma camera parallel-hole
collimators as in the PET case, no collimation is provided within the
transverse planes.
• By physically rejecting almost all oblique photons from reaching the
detectors, the count rate is substantially reduced, resulting in a low rate
of random coincidences and low detector dead time.
• In addition, 2‑D acquisition is associated with a low rate of scattered
coincidence events since only photons emitted in and scattering within a
transverse plane can pass through the septa.
• In 3‑D acquisition mode, the septa are entirely removed from the
FOV and there is no longer any physical collimation restricting the
photons that are incident upon the detectors.

• Coincidence events can be recorded between detectors in different

rings and potentially between all possible ring combinations.

• Photons emitted at oblique angles with respect to the transverse

plane are no longer prevented from reaching the detectors, and
system sensitivity is substantially increased compared to 2‑D
acquisition.

• Sensitivity gains by a factor of around five are typical, although the

exact value depends on the scanner configuration and the source
distribution
• In 2‑D mode, sensitivity varies slightly between adjacent slices
but does not change greatly over the AFOV.
• In 3‑D mode, the sensitivity variation in the axial direction is
much greater and has a triangular profile with a peak at the
central slice.
• The advantage of 3‑D acquisition is its large increase in
sensitivity compared to 2‑D acquisition.
• As a consequence of the substantial sensitivity increase, 3‑D
acquisition is associated with higher detector count rates,
leading to more randoms and greater dead time than
corresponding acquisitions in 2‑D mode.
• Furthermore, 3‑D mode cannot take advantage of the scatter
rejection afforded by interplane septa and, as a result, records
a greatly increased proportion of scattered coincidence events.
Data Acquisition
• The data recorded during a conventional PET acquisition are the total number of
coincidence events measured between the various detector pairs.
• These data are typically binned into 2‑D matrixes known as sinograms
• Full ring PET scanners simultaneously measure multiple
projections at different angles ϕ with respect to the patient.
• An example showing the orientation of two parallel
projections is shown in (a).
• Projection data of this sort are typically stored in sinograms;
an example is shown in (b).
• In a sinogram, each row represents a projection at a different
angle ϕ.
• Each projection is made up of discrete elements that are
indexed by s and contain the number of coincidence counts
recorded along individual lines of response.
• The two example projections shown in (a) are also
highlighted in sinogram (b).
• If a 2‑D acquisition geometry is considered, each row of the sinogram
represents a projection of the radionuclide distribution at a particular angle
around the patient.

• These projections consist of coincidence events recorded between pairs of

detectors, where each detector pair forms LORs that are approximately
parallel to each other.

• The number of counts in each element of the projection is proportional to a

line integral of the radionuclide distribution within the limitations imposed
by the various physical effects such as scatter and attenuation.

• The sinogram is indexed along the y axis by angle and the x axis by distance.
• The prior discussion of 2‑D acquisition mode only
considered coincidence events between detectors in a
single ring, referred to as a direct plane.

• However, in practice, the interplane septa do not

completely eliminate the possibility of coincidence
events being detected between different nearby rings.

• Inclusion of these slightly oblique events is

advantageous as it increases sensitivity.
• Coincidence events between detectors in immediately adjacent
rings are combined into a sinogram that is considered to have
been measured in a plane located between the two detector
rings.

• This plane is referred to as a cross plane and is considered to be

parallel to the direct planes, despite the fact that the contributing
LORs are slightly oblique to these planes.

• As well as increasing sensitivity, inclusion of these cross planes

increases axial sampling by producing 2N – 1 slices from an N
ring scanner.

• The total number of ring combinations contributing to a direct

plane plus those contributing to a cross plane is sometimes
• When the septa are removed, as is the case in 3‑D acquisition mode,
there is no longer any physical restriction on the detector rings that can
be used to measure coincidence events.
• An N ring scanner could have a maximum ring difference of N – 1,
resulting in up to N2 possible sinograms.
• In 2‑D mode, such a system would have a total of 2N – 1 sinograms,
so it can be seen that the total volume of data is substantially higher in
3‑D mode.
• It should be noted that 3‑D acquisition mode results in data that are
redundant in the sense that only a subset of the sinograms (those in the
transverse planes) are required for tomographic image reconstruction.
• The purpose of acquiring the additional oblique data is to increase
sensitivity and reduce statistical noise in the resulting images.
Side elevation of an eight ring PET
scanner in 2‑D ((a) and (b)) and 3‑D
(c) acquisition modes.
(a)Lines of response joining
opposing detectors in the same
ring forming direct planes.
(b) Lines of response between
detectors in adjacent rings.
• These lines of response are
averaged to form cross planes
(dotted line) that are assumed to
be located at the mid-point
between adjacent detectors.
• Both direct and cross planes are
simultaneously acquired during
2‑D acquisition.
(c) 3‑D acquisition in which each
ring is permitted to form
coincidence events with all other
Time of Flight
• Detectors operating in coincidence mode provide spatial
information related to individual positron–electron annihilations
but this information is not sufficient to determine the exact
location of each event.

• A line joining the two detectors can be assumed to intersect the

site of the annihilation but the exact position along this line
cannot be determined.
•
• For this reason, PET systems measure signals from multiple
events and the resulting projections are used to reconstruct
images using computed tomography.
• The difference in the detection times of the two annihilation photons
provides a mechanism for precisely localizing the site of individual
positron–electron annihilations.

• Given that photons travel at the speed of light, essentially irrespective

of the composition of the material through which they pass, the
difference in the arrival times of the two photons can potentially be
used to localize their original point of emission.

• This is clearly attractive because it means that each coincidence

measurement provides significantly more information, promising
substantial improvements in image statistical quality.

• Incorporating information derived from differences in the photon

arrival times has been referred to as TOF mode and a number of PET
systems have been developed that exploit this approach.
(a) A coincidence event detected along a line of
response between detectors A and B.
• The average time difference between the two
detectors is given by (x + Δx)/c – (x – Δx)/c =
2Δx/c, where c is the speed of light.

(b) With conventional PET, no information is

available about the location of the annihilation
event along the line of response.
• During reconstruction, the event is assigned
with equal weight to all pixels between A and
B.

• (c) With time of flight PET, the time difference

between the signals recorded at detectors A
and B is used to estimate the position of the
annihilation event along the line of response.
• During reconstruction, events are weighted
according to the detector time difference and a
function that reflects the limited time
Data Corrections

Normalization

• Normalization refers to a software correction that is applied to

the measured projection data in order to compensate for
variations in the sensitivity of different LORs.

• Without such a correction, images display systematic

variations in uniformity and pronounced artefacts that include
spike and ring artefacts at the centre of the FOV
• Sinograms corresponding to
a centrally located uniform
cylinder before normalization
(a) and after normalization
(b).
• Transverse images are shown
for reconstructions without
normalization (c) and with
normalization (d).
• The artefacts and non-
uniformity seen in the image
without normalization should
be noted.
• Normalization files are experimentally determined
correction factors that are applied as multiplicative
terms for each LOR.

• They are periodically updated to reflect the current state

of the detector system and are applied to all
subsequently acquired data.

• In order not to degrade the statistical quality of the

patient data, normalization coefficients need to be
measured with low statistical noise, and a variety of
methods have been developed to achieve this.
Randoms Correction
• Randoms make up a potentially large component of all measured coincidence events
and, if left uncorrected, will contribute to a loss of image contrast and quantitative
accuracy.

• Random coincidences are generally smoothly distributed across the FOV but the
magnitude of the randoms component depends on the count rate encountered during
data acquisition.

• 3‑D acquisition mode or studies involving large amounts of activity in or near the
FOV are usually associated with high randoms fractions.

• Randoms correction is essential for all quantitative studies and is routinely

implemented on almost all scanner systems.
• One widely adopted correction method involves estimating the
number of randoms contributing to the prompts (trues + scatter +
randoms) using an additional coincidence circuit.

• This secondary coincidence circuit is acquired simultaneously with

the prompt measurement but is only sensitive to random events.

• Preferential selection of randoms is achieved by delaying the logic

pulse from one of the detectors such that it cannot form a true
coincidence event between corresponding annihilation photons
• Diagram illustrating the concept of how a
delayed coincidence circuit can be used to
estimate the number of random events in
the prompt circuit.
• Detection events from two opposing
detectors, indicating three coincidence
events in the prompt circuit.
• (b) Data from detector 2 delayed with
respect to detector 1 and indicating one
coincidence event in this delayed circuit.
• The temporal delay prevents true
coincidence events from being recorded in
the delayed circuit, but random coincidence
events still occur with the same frequency
as in the prompt circuit.
• If data are acquired with sufficient statistical
quality, the total number of delayed
coincidence events provides an estimate of
the total number of randoms in the prompt
circuit.
Attenuation correction

• Despite their high energy, only a small fraction of the emitted 511 keV
photon pairs escape the body without undergoing some form of
interaction.

• Compton interaction is the most likely mechanism and, depending on the

energy and direction of the scattered photon, may result in the detection
of a scattered coincidence event.

• However, it is more likely that the scattered photon will not result in a
coincidence event for a variety of reasons.

• The scattered photon may emerge from the Compton interaction along a
path that is not incident upon the detectors.
• Alternatively, the scattered photon may undergo further Compton
interactions, resulting in a lower energy and a greater likelihood
of photoelectric absorption.

• Even if the scattered photon does reach the detectors, it may

have lost so much energy that it does not meet the energy
acceptance criteria of the scanner and will be rejected.

• Thus, as well as creating scattered coincidence events, Compton

interactions lead to a much greater loss of true coincidence
events.

• This underestimation of true counts is referred to as attenuation.

• One of the advantages of PET over SPECT is the ease with which
• A particular advantage of PET imaging is that
attenuation correction factors do not depend on the
location of a source along a given LOR.
• Consider a figure below. For an event to be registered,
both gamma rays from the decay must be detected
Thus, the probability that a gamma ray will
reach detector 1 is:
• The probability that a gamma ray will reach to detector
2 is

The total probability that the pair of gamma rays is

detected is then
• Thus, the probability of detection (of not being attenuated) is
independent of the position X′ of the decay site along the LOR.

• If a source is placed outside of the patient, then the attenuation

measured along any LOR with the source will be the proper
correction for the emission data (the transmission scan).

• The correction is simply measured by taking the natural

logarithm of the ratio between the transmission scan and a
blank scan (a scan with the external source without any objects
in the scanner’s FOV).
CT Based Attenuation Correction
• Another approach to the measurement of
transmission data is to combine a CT scanner
with a PET system.

• The resulting data are taken at an energy that is

very different from 511 keV, so the data must be
scaled to use for attenuation correction of the
emission data.
• CT based attenuation correction has proved to be very effective although a number of
potential problems require consideration.

• Patient motion, commonly motion of the arms or head, can cause the CT and PET
images to be misregistered, leading to incorrect attenuation correction factors, which in
turn cause image artefacts and quantitative error.

• Respiratory motion can also lead to similar problems as the CT and PET data are
acquired over very different time intervals.

• CT data acquisition is extremely short and usually captures a particular phase in the
respiratory cycle.

• In contrast, PET data are acquired over multiple breathing cycles and the resulting
images represent an average position that will be somewhat blurred in regions where
respiratory motion is significant.
Scatter Correction
• Scatter correction techniques include integral transform,
function fitting, energy-based subtraction, and analytic
calculation Scatter correction is required because the
limited energy resolution of PET systems means that
scattered photons can only be partially rejected by
energy discrimination.

• Uncorrected scatter forms a background in reconstructed

images that reduces lesion contrast and degrades
quantitative accuracy.

• This scatter background is a complex function of both the

emission and attenuation distributions and is non-uniform
across the FOV.
• In 2‑D mode, physical collimation ensures that the scatter
contribution is relatively low compared to 3‑D mode and
approximate corrections, based on scatter deconvolution,
have been widely used.
• The form of the scatter distribution function can be measured
experimentally using line sources at different positions in a
water phantom.
• Analytical expressions derived from this scatter function can
be determined and convolved with the projection data from
individual patient studies to estimate the scatter distribution.
• This method assumes a uniform scattering medium and has
limited accuracy in areas such as the thorax.
• It also cannot account for scatter between adjacent planes,
which is significant for 3‑D acquisition mode.
• An alternative algorithm that has been applied to 3‑D
brain studies involves a tail fitting approach.
• In brain studies, the LORs that do not pass through the
head are comprised of scatter that can be modelled by
fitting a Gaussian function to the tails of each projection.

• This function can be interpolated to the projections that

do pass through the head and used as an estimate of the
scatter contribution along these LORs.
• This method provides a first order correction for scatter
and has limited accuracy in areas of non-uniform
attenuation or any area where the tails of the projections
cannot be accurately measured.
• More accurate scatter correction can be achieved in 3‑D using a
model based approach.
• This method makes use of the physics of Compton scattering to
model the distribution of coincidence events for which one of
the two photons experienced a single scattering interaction.
• The assumption that the scatter component in the measured
data is dominated by these single scatter events has been
shown to be reasonable.

• An initial estimate of the radionuclide distribution is first

obtained from a reconstruction that does not include any
scatter correction.
• This image is then used, in conjunction with an attenuation map
derived from CT or transmission data, to estimate the scatter
contribution to each LOR.
• Different implementations have been developed, but each makes use of
the Klein–Nishina formula to determine the probability of a photon
scattering through a certain angle and being detected by a certain detector.
• Determining these probabilities for all possible scattering positions and
LORs is computationally demanding.

• However, good scatter estimates can be obtained by interpolating data

estimated using a coarse grid of scattering locations and a subset of
LORs.
• These data can then be interpolated for all LORs, resulting in an estimate
of the scatter distribution that has been found to be highly accurate over a
range of anatomical locations.
Dead Time Correction
• Detector count rates vary between patients and can also
vary within studies performed on the same patient. In order
to achieve quantitatively accurate images, the true count
rate should ideally increase linearly with increasing activity
in the FOV.
• Although this is usually the case at low count rates, the PET
scanner’s response becomes increasingly non-linear at high
count rates.

• Individual detector modules within a scanner require a finite

period of time to process each detected photon.
• If a second photon is incident upon a detector while an
earlier photon is still being processed, the secondary photon
may be lost.
• Other sources of dead time arise during coincidence
event processing.

• When more than two events occur within the

coincidence time window, it is impossible to
determine the correct coincidence pair.

• In this circumstance, all detection events may be

discarded, contributing to dead time losses, or
alternatively, all possible coincidence events can be
included, increasing the randoms component.
• Dead time correction compensates for this loss of sensitivity.

• Corrections are usually based upon experimental measurements of

the scanner’s response to a decaying source of activity.

• After randoms correction, residual non-linearity in the scanner’s

response can be attributed to dead time.

• An analytical model of the scanner’s count rate response can be

determined from these experimental data and used for dead time
correction of subsequent patient data.

• A global correction factor can be applied for a particular acquisition,

assuming that dead time effects are similar for all detectors in the
ring.
• Alternatively, different corrections can be applied to each
detector block or group of blocks.

• Corrections can be determined based upon an estimate of the

fraction of the acquisition period that each detector was busy
processing events and unable to process other photons.

• Alternatively, the single photon rate at a particular detector can

be used as input for a model of the scanner’s dead time
performance to estimate the magnitude of the dead time effect.

• Dead time correction only compensates for count losses and

does not compensate for the event mis-positioning that can
occur as a result of pulse pile-up
Image Calibration

• The above corrections substantially eliminate the image artefacts

and quantitative errors caused by the various physical effects that
degrade PET data.
• As a result, the reconstructed images reflect the activity distribution
within the FOV, within the limitations imposed by the system’s
limited spatial resolution.

• Furthermore, these reconstructed images can be used to quantify

the in vivo activity concentration in a particular organ or tissue.
• Although this capability is not always fully exploited, the potential to
accurately quantify images in terms of absolute activity
concentration facilitates a range of potential applications.
• After image reconstruction, including the application of the
various physical corrections, PET images have arbitrary
units, typically counts per voxel per second.

• Quantitative data can be extracted from the relevant parts

of the image using region of interest techniques but cannot
be readily compared with other related data such as
measurements made with a radioactivity calibrator (‘dose’
calibrator).

• In order to convert the PET images into units of absolute

activity concentration such as becquerels per millilitre, a
calibration factor is required.
• This calibration factor is experimentally determined,
usually using a uniform cylinder phantom.

• The cylinder is filled with a known volume of water,

to which a known amount of radioactivity is added.

• After ensuring the radioactivity is uniformly

distributed within the phantom, a fully corrected
PET image is acquired.
• The calibration factor CF can be determined using:

• Where A/V is the known activity concentration (Bq/mL) within the phantom;

• C is the mean voxel data (counts · voxel–1 · s–1) from a large region well within
the cylinder part of the image;

• and p is the positron fraction of the radionuclide used in the calibration

experiment (typically 18F, positron fraction 0.97).
• The positron fraction is a property of the radionuclide and is the fraction
of all disintegrations that give rise to the emission of a positron.
• The above calibration assumes that the true activity within the
phantom is accurately known.
• This can usually be achieved to an acceptable level of tolerance using
an activity calibrator that has been calibrated for the isotope of interest
using a long lived standard source that is traceable to a national
metrology institute.

• In principle, a single calibration factor can be applied to subsequent

studies performed with different isotopes as long as the positron
fraction is known.
• Calibrated PET images can, thus, be determined by multiplying the raw
image data by the calibration factor and dividing by the positron
fraction for the particular isotope of interest.
Design considerations for PET systems
1. Spatial Resolution
• High spatial resolution is clearly an important design objective for PET
imaging systems.
• As such, the trend in modern scanner systems has been to decrease the
width of individual detectors and to increase the total number of
detector elements surrounding the patient.
• The increased concentration of detector elements decreases the
sampling interval and generally improves spatial resolution.
• Although modern designs involve detectors that are only a few
millimetres wide, the need for high sensitivity means that they are often
a few centimetres long.
• Problems can occur when photons are incident on one detector but
penetrate through to an adjacent detector.
• The location within the detector (depth of interaction) is
typically not measured and the detection event is
assigned to a location at the face of the detector.

• This problem frequently occurs when detectors are

arranged in a ring (or similar) configuration and gives
rise to a loss of resolution at more peripheral locations.

• This resolution loss generally occurs in the radial

direction as opposed to the tangential direction due to
the angle of incidence of the photons on the detectors.
• Photon penetration between
adjacent detectors in a ring
based system leads to mis-
positioning of events.

• This primarily affects the

radial component of spatial
resolution which degrades
with distance from the
centre of the field of view.
• Another factor that influences spatial resolution is the distance between opposing
detectors.
• This distance is relevant because of a small uncertainty in the relative angle of the 511
keV annihilation radiation.

• Residual momentum of the positron

and electron immediately before
annihilation causes the two 511 keV
photons to deviate slightly from the
expected 180° angle.

• As a result, a line joining detection

events does not intersect the exact
point of annihilation.

• The extent of this non-collinearity is

greatly exaggerated in the figure, but
it does contribute to a loss of spatial
resolution, especially for large
diameter PET systems.
• Although the basic assumption of coincidence detection is that
annihilation radiation is emitted 180° apart, this is not strictly true.

• Positrons frequently annihilate before they have lost all

momentum, and this residual momentum translates to a small
deviation of about ±0.25° from the expected back to back
emissions.

• This effect is referred to as non-collinearity and tends to degrade

spatial resolution as detector separation increases.

• For PET systems with opposing detectors separated by only a few

centimetres, such as those optimized for specific organs such as
the brain or breast, this is not a major issue.
• The distance travelled by a positron between its point of
emission and annihilation is an additional factor that degrades
the spatial resolution that can be achieved by PET systems.

• As previously discussed, this distance, or positron range, is

dependent on the energy of the positron and also the type of
material through which the positron is passing; a greater
range is expected in tissue such as lung compared to soft
tissue.

• It should be noted that software corrections have been

implemented that model the effect of positron range and can
potentially reduce the loss of resolution in reconstructed
images.
Sensitivity
• The best possible spatial resolution that can be obtained by a PET system is not
always achieved in clinical practice due to statistical noise in the measured
data.

• In order to suppress this noise, clinical protocols generally employ low-pass

filters or other similar image reconstruction methods, but the consequence is
invariably a loss of spatial resolution.

• Improving the statistical quality of the measured coincidence data not only
reduces image noise but also allows the opportunity to reduce image smoothing
and improve spatial resolution.

• The need to optimize this trade-off between statistical noise and spatial
resolution influences both image reconstruction development and scanner
design
• Noise in PET images is influenced by a number of factors, including
the sensitivity of the detector system, the amount of radioactive
tracer administered to the patient and the amount of time the patient
can remain motionless for an imaging procedure.

• Limitations on the latter two factors mean that high sensitivity is an

important objective for scanner design. Sensitivity is determined by
the geometry of the detector arrangement and the absorption
efficiency of the detectors themselves.

• Reducing the distance between opposing detectors increases the solid

angle of acceptance and increases sensitivity.

• However, the requirement to accommodate all regions of the body

imposes a minimum ring diameter for whole body systems
• For such systems, extending the axial field of view (AFOV) of
the detector system provides a mechanism for increasing
sensitivity.

• Cost constraints have prevented the construction of PET

systems that cover the entire length of the body.

• However, extended axial coverage can be achieved using

systems with much smaller AFOVs by scanning sections of the
body in a sequential fashion.

• For such systems, extending the AFOV of the scanner not only
increases sensitivity but also reduces the number of bed
translations required for whole body coverage.
• In addition to the geometry of the detector system,
sensitivity is also determined by the absorption
efficiency of the detectors.

• A high absorption efficiency for 511 keV photons is

desirable in order to make best use of those photons
that are incident upon the detectors.

• Absorption efficiency or stopping power of the detector

material is, therefore, an important consideration for
PET system design.
Quantitative accuracy
• One of the strengths of PET is its capability to quantify physiological
processes in vivo.

• A prerequisite for this kind of quantitative analysis is that the images

accurately reflect the local activity concentration in the body.

• In order to ensure this kind of quantitative accuracy, it is important to

minimize effects that corrupt the data and to correct residual corruption
as necessary.

• Quantitative error can arise from many sources but is primarily due to
random coincidence events, photon scatter within the body, photon
attenuation within the body and detector dead time.
The figure below illustrates some of these situations

(a) A true coincidence event can occur

when both photons escape the body
without interacting.

(b) A random coincidence event occurs

when two photons from unrelated
annihilation events are detected at
approximately the same time.

(c) A scattered coincidence event can

occur when either photon is scattered
within the body but is still detected.

(d) No coincidence event is recorded

when one or both photons are
attenuated, typically due to scatter
• Software processing prior to (or during) image reconstruction can mitigate
the above effects, but the accuracy of these corrections may not be
reliable if the contamination overwhelms the signal from true coincidence
events.
• PET systems are, therefore, designed to minimize the contribution of the
various degrading factors described above.
• In terms of scanner design, very little can be done to reduce attenuation
as photons that are absorbed within the body do not reach the detectors.
• However, scattered photons can potentially be rejected by the detection
system if their energy falls outside a predetermined acceptance range.
• Annihilation radiation that is scattered within the body will emerge with
energies less than 511 keV.
• The exact energy of the scattered photon will depend on the number of
Compton scattering interactions that have occurred and the angle through
which the photon was scattered.
• Energy discrimination can, therefore, be used to reject photons that
have energies less than 511 keV and are assumed to have been
scattered.

• This approach relies on the detection system having high energy

resolution and, in practice, energy discrimination reduces but does not
eliminate scattered coincidence events.

• The limited energy resolution of current PET systems means that, in

order to avoid rejecting too many true (unscattered) coincidence
events, the energy acceptance window is usually set to accept quite a
broad range of energies around 511 keV.

• High energy resolution is, nevertheless, an important design objective,

particularly for those scanner systems that detect a large proportion of
scattered photons.
• Decreasing the coincidence timing window decreases the number of
random events as the shorter time interval reduces the likelihood of
a coincidence occurring by chance between two unrelated photons.
•
• There is, however, only limited scope for reducing the duration of
the coincidence time window as it is restricted by the timing
resolution of the detector system and the fact that off-centre
annihilations result in real time differences between detection
events.

• Optimization of the coincidence timing window for a particular

scanner represents a compromise between wanting to reduce the
number of random coincidence events without significantly
reducing the number of true coincidences.
• Detector systems that are able to measure photon
detection times with low variability (high timing
resolution) are, therefore, desirable from the perspective
of randoms reduction.

• Timing resolution also contributes to detector dead time

as a shorter coincidence timing window reduces the
likelihood of more than two photon detection events
occurring.

• Other contributions to dead time include the time

required by the detector to measure an individual photon
event and the time spent processing coincidence events.
Other Considerations
• Spatial resolution, sensitivity and quantitative accuracy are the
main physics issues influencing PET system design, but there
are various other factors that need to be considered.
• One obvious issue is the overall cost of the system.
• This significantly influences design decisions as the detectors
represent a significant fraction of the overall production cost.
• The choice of detector material, the thickness of the detectors,
the diameter of the detector ring and the axial extent of the
detectors all contribute to the total cost of the system.
• Another important design issue that affects the overall cost is
integration of the PET system with a second modality in a
combined scanner.
• In most cases, this means integration of the PET subsystem with a CT subsystem, although
combined PET/magnetic resonance (MR) scanners exist and present more significant
technical challenges.

• Combined PET/CT scanners have effectively replaced stand-alone PET for clinical
applications and the optimal CT configuration included in a combined system is limited
mainly by cost concerns.

• In practice, this means the number of slices to include in the multi-detector CT system.

• High performance multi-detector CT is important for systems intended for cardiac PET/CT
applications, whereas lower slice capability may be adequate for oncological PET/CT.

• Other relevant design issues are related to the computer workstation used for acquiring and
processing data.

• These include fast image reconstruction times, convenient integrated control of both PET
and CT subsystems, and seamless integration with other institutional information
technology systems such as clinical information systems and image archive systems.
PET Detector Systems

• PET detectors are specially engineered to handle high energy photons, and
because there is a ring of detectors, there is no need for the ring to rotate in
order to obtain a tomographic image.
• Instead of traditional sodium iodide (NaI) crystals, PET ring detector crystals
are composed of compounds such as bismuth germanate (BGO), lutetium
oxyorthosilicate (LSO), or gadolinium oxyorthosilicate (GSO).
• All of these compounds are used by different PET and PET/CT manufacturers,
and they have physical properties which make them well suited to PET
imaging
Radiation Detectors
• Although different radiation detector designs have been used in PET,
almost all current systems adopt an approach based on scintillation
detectors.

• Scintillation detectors are inorganic crystals that emit scintillation light

in the visible range when high energy photons are incident upon them.

• This light is converted to an electrical signal by means of a

photodetector, usually a PMT, coupled to the crystal material.

• Various scintillator materials have been used in PET, including thallium

doped sodium iodide (NaI(Tl)), bismuth germanate (BGO) and cerium
doped lutetium oxyorthosilicate (LSO).
 Some of the properties of the crystal materials
that are relevant for PET applications.
• The properties of an ideal crystal for PET would include a high stopping power for 511
keV photons (high linear attenuation coefficient); short scintillation light decay time to
reduce dead time and allow short coincidence time windows to reduce random
coincidences; and high light output.
• High light output enables good energy resolution, which gives rise to improved scatter
rejection.
• It also affords cost savings in the construction of a complete scanner system as the
number of photodetectors required to resolve a given number of crystal elements can
potentially be reduced.
NaI(Tl)

• Although NaI(Tl) is ideal for lower energy single photon imaging, its relatively low linear
attenuation coefficient for 511 keV photons makes it less attractive for PET applications.

• Sensitivity could potentially be increased by increasing the thickness of the crystals, which are
typically 1–3 cm thick.

• However, the scope for substantially increasing crystal thickness is limited as it results in a loss
of spatial resolution.

• This is because thicker crystals are prone to more significant depth of interaction problems as
the apparent width of the detector increases for sources located off-centre.

• Thin crystals composed of a material with a high stopping power for 511 keV photons are, thus,
desirable to ensure best possible sensitivity while maintaining spatial resolution.

• For this reason, BGO and, more recently, LSO have replaced NaI(Tl) as the scintillator of choice
for PET
BGO
• BGO has the advantage of a high stopping power for 511 keV photons and has become an
important scintillator for PET applications.

• However, it is not ideal in many respects as it has relatively poor energy resolution and a
long crystal decay time.

• The poor energy resolution translates into a limited ability to reject scatter via energy
discrimination.

• In addition, the long decay time translates into a greater dead time and increased number of
random coincidences at high count rates.

• As such, BGO is well suited for scanner designs that minimize scatter and count rate via
physical collimation, such as those with interplane septa.

• Attempts to increase sensitivity by removing the interplane septa typically result in a high
scatter, high count rate environment for which BGO is not ideal.
LSO

• Although LSO has a lower linear attenuation coefficient than BGO, its shorter crystal
decay time and slightly improved energy resolution convey significant advantages as a
PET scintillator.

• LSO has become the scintillator of choice for scanner designs that operate without
interplane septa because its short decay time makes it well suited for high count rate
applications.

• The fast decay time of LSO also enables a time of flight (TOF) data acquisition mode.

• LSO has proved to be a successful crystal for PET detector applications despite the fact
that the material contains around 2.6% 176Lu, which is itself radioactive.

• A component of the emissions from 176Lu is detected within the energy acceptance
window and the dominant effect is to contribute random coincidences.
• In practice, the increased randoms rate is not a major
problem for clinical studies, and the naturally occurring
radiation has even been used for quality assurance.

• Lutetium-176 has a half-life of 3.8 × 1010 a and, thus,

provides a long lived source of radiation that can be used
to check consistency of detector response without the need
for external sources.

• It should be noted that, for commercial reasons, some PET

systems employ cerium doped lutetium yttrium
oxyorthosilicate (LYSO (Ce)) which has substantially similar
properties to LSO.
Detector Arrangements
• As discussed above, the interaction of 511 keV annihilation radiation with
the scintillation crystals gives rise to optical light that can be detected by
a suitable photodetector.
• A photodetector is a device that produces an electrical signal when
stimulated by light of the sort emitted by a scintillation detector.
• For most PET applications, PMTs have been the preferred photodetector
because their high gain results in an electrical output with a good signal
to noise ratio.
• In addition, PMT output is proportional to the intensity of the incident light
and, thus, proportional to the energy deposited in the crystal.
• This provides a mechanism for selective acceptance of only those
detection events with energies within a specific range and can be used to
reject scattered photons.
• In addition, PMTs provide high amplification with little degradation in the
timing information that is essential for electronic collimation.
• Although PMTs are by far the most widely used photodetector for PET applications, they are
somewhat bulky and highly sensitive to magnetic fields.

• For these reasons, they are often not used in combined PET/MR systems where space is
limited and operation in high magnetic fields is a requirement.

• In these and some other applications, semiconductor based photodiodes are an alternative
to PMTs.

• It should be noted that in these applications, the semiconductor device is used in

conjunction with a scintillation detector and is used to detect scintillation light, not the
annihilation photons.

• Avalanche photodiodes have much lower gains than PMTs but can be very small and have
been shown to be effective in high magnetic field environments.

• Their low gain requires very low noise electronics and they are also sensitive to small
temperature variations.
• Space and cost constraints mean that individual scintillation crystals are not usually
coupled directly to individual photodetectors in a one to one fashion.
• Instead, the most common arrangement is a block detector in which a group of crystal
elements share a smaller number of PMTs as shown below.

(a) A PET detector block

consisting of scintillator material
coupled to an array of
photomultiplier tubes.
• The scintillator is cut into an
array of individual crystal
elements.
• Four photomultiplier tubes are
used to read out the signal
from the 8 × 8 array of crystal
elements.
(b) The x and y position of each
photon is determined from the
signal measured by each of the
four photomultiplier tubes
labelled A–D, using the
• The design of each block varies between manufacturers
and scanner models but usually involves a matrix of
crystal elements, a light guide and four PMTs.

• An example configuration might be an 8 × 8 array of

closely packed 4.4 mm × 4.0 mm × 30 mm crystal
elements, where the longest dimension is in the radial
direction to maximize detection efficiency.

• The light guide allows light to be shared between four

circular PMTs and the relative light distribution depends
on the location of the crystal in which the photon
interacted.
• The (x, y) position of the detection event is calculated from the outputs of the four PMTs
using a weighted centroid algorithm, similar to the Anger logic of a gamma camera.

• Although individual crystals can be identified in this way, the response is not linear
throughout the block due to differences in the locations of the different crystal elements
relative to the PMTs.

• Experimentally determined look-up tables are used to relate the measured (x, y) position to
a corresponding detector element, effectively performing a form of linearity correction.

• In this way, only four PMTs are needed to localize signals from a much greater number of
crystal elements.

• The number of crystal elements divided by the number of PMTs in a PET system has been
referred to as the encoding ratio.

• A high encoding ratio implies lower production costs and is, therefore, desirable
• One of the advantages of the design described above is that each block
operates independently of its surrounding blocks.

• This leads to good count rate performance as light is not transferred between
blocks and the PMTs of one block are unaffected by detection events in an
adjacent block.

• An alternative arrangement, referred to as quadrant sharing, increases the

encoding ratio by locating the PMTs at the corners of adjacent blocks.

• This arrangement differs from the conventional block design in that each PMT
can now be exposed to light from up to four different blocks.

• This can result in better spatial resolution and a higher encoding ratio but is
also susceptible to greater dead time problems at high count rates.
• Another alternative to the block design adopts an approach similar to that used in conventional gamma
cameras.

• These Anger-logic designs involve detector modules that have a much larger surface area compared to
conventional block detectors, e.g. 92 mm × 176 mm.

• Each module is comprised of many small crystal elements which are coupled, via a light guide, to an array of
multiple PMTs.

• Light is spread over a larger area than in the block design and positional information is obtained using Anger-
logic in the same way as a gamma camera.

• The PMTs used in this design are typically larger than those used in block detectors, increasing the encoding
ratio.
• The larger area detector modules encourage more uniform light collection compared to block designs, which
leads to more uniform energy resolution.

• However, a disadvantage of this design is that the broad light spread among many PMTs can lead to dead
time problems at high count rates.
Scanner Configurations

• The detectors described above form the building blocks used to construct complete
scanner systems.
• Various scanner configurations have been developed, although the dominant design
consists of a ring of detectors that completely surrounds the patient (or research
subject) in one plane as shown in the figure below.
(a) Full ring PET system shown in the transverse plane,
indicating how each detector can form coincidence
events with a specific number of detectors on the
opposite side of the ring.
• For clarity, this fan-like arrangement of lines of
response is shown for only eight detectors.
• The dashed line indicates how the imaging field of
view is necessarily smaller than the detector ring
diameter.

(b) PET system shown in side elevation, indicating the

limited detector coverage in the z direction.
• The shaded area indicates the coincidence field of
view.
• The dashed lines indicate the singles field of view.
• End shields reduce, but do not eliminate, detection of
single photons from outside of the coincidence field of
• As with other scanner systems, this plane is referred to as the
transverse or transaxial plane and the direction perpendicular to
this plane is referred to as the axial or z direction.

• Several rings of detectors are arranged in a cylindrical geometry,

allowing multiple transverse slices to be simultaneously acquired.

• As coincidence detection requires two opposing detectors, a full

ring system of this sort allows coincidence data to be acquired at
all angles around 180°.

• Although complete angular coverage is achieved in the transverse

plane, there is much more limited coverage in the axial direction.
• Cost constraints and, to some extent, limited patient tolerance of
extended tunnels mean that the detector rings usually extend for only
a few centimetres in the axial direction.

• Human whole body systems typically have an AFOV design has been
to increase the AFOV, thus increasing both sensitivity and the number
of transverse slices that can be simultaneously acquired. of around
15–20 cm, although the trend in scanner.

• The diameter of the detector ring varies considerably between

designs and this dimension reflects the intended research or clinical
application.

• Small animal systems may have ring diameters of around 15 cm,

brain oriented human systems around 47 cm and whole body human
• Systems with ring diameters that can accommodate the
whole body are clearly more flexible in terms of the range
of studies that can be performed.

• However, smaller ring diameters have advantages in terms

of increased sensitivity, owing to a greater solid angle of
acceptance, and potentially better spatial resolution, owing
to reduced photon non-collinearity effects.

• It should be noted that the spatial resolution advantage is

complicated by greater depth of interaction problems as
the detector ring diameter decreases and shorter crystals
or depth of interaction measurement capability may be
required.