Positron Emission Tomography (PET)

A radiological technique for functional imaging

Please note that this exercise takes place at the Stockholm Centre for Physics,
Astronomy and Biotechniques (Alba Nova).
Address Roslagstullsbacken 21
You will find the PET-lab equipment in room E2:1021
The lab assistant (Anton Khaplanov) can be found in his office: C3:3001, tel 08-5537 8199 or by
MAIL [email protected]
For reference see S. Webb: The Physics of Medical Imaging or
R.F. Farr and P.J.Allisy-Roberts

Updated:
Jan 2007

1
1 Introduction (Webb 6.3.5)
Positron Emission Tomography (PET) is a technique used in clinical medicine and biomedical
research to create images that show anatomical structure as well as how certain tissues are
performing their physiological functions. However, the main emphasis in PET is not on
anatomical imaging, as e.g. X-ray or MR imaging. Functional imaging is the major modality.
Radioactive nuclei are introduced to the body as labels on tracer molecules designed to probe
physiological processes. These radioactive nuclei emit positrons that annihilate with electrons in
the tissue. An annihilation event mostly result in two gamma photons (in some cases also three
gamma photons, depending on the angular momentum coupling of the electron and positron)
being emitted in almost exactly 180 degrees and with an energy of 511 keV each. The gamma
photons are detected in coincidence in a detector ring (or several detector rings) so that two
detected gamma photons represent a straight line along which the event took place. An assembly
of such lines is gathered and processed with the aid of image processing tools to finally produce
an image of the activity and thereby of the functionality.

An extension of PET is TOFPET (Time-Of-Flight Positron Emission Tomography) (c. f. p. 159)

in which the time difference between the arrival of coincident photons is measured. In PET this
information is ignored and the annihilation is equally probable to have occurred along the full
extension of the line between the detectors. Incorporating this information gives more weight to
the more probable locations of the emission point for each event. The localisation of the emission
point is made to within 4-10 cm, depending on the time resolution of the system. Including these
data reduces statistical uncertainty in the reconstructed image and thus one obtains better images.
PET isotopes (the label nuclei) are produced on-site by small cyclotrons or linear accelerators
and are synthesised into highly specific chemical agents by automated systems. Anatomical
imaging such as MRI (Magnetic Resonance Imaging) or CT (Computed Tomography) is often
used to anatomically map the PET images to specific tissue. PET plays an increasingly important
role in the diagnosis of cancer, heart disease, and illnesses of the brain such as epilepsy, stroke
and the dementia. It is used as a basic research tool in studying the functional anatomy of the
brain. It is also used by the pharmaceutical industry to test new drugs.
The aim of this laboratory exercise is to demonstrate the principles of PET and TOFPET by
recording of annihilation radiation from radioactive positron emitters. The radiation is detected
by a PET system consisting of 48 detectors and accompanying electronics. Images are

2
constructed by software and reveal the positions of the sources. TOFPET is not yet used
commercially.

2. Beta Decay
Beta particles are fast electrons or positrons produced in the weak interaction decay of neutrons
or protons in neutron- or proton-rich nuclei. In a neutron-rich nucleus a neutron can transform
into a proton via the process

n → p + e− +ν e

where an electron and an antineutrino are emitted. This is also how a free neutron decays with a
half-life of 10.25 minutes. The daughter nucleus now contains one extra proton so that its atomic
number Z has increased by one unit. This can be written as

(Z , A) → (Z + 1, A) + e − + ν e
In proton-rich nuclei a positron and neutrino are emitted in the process

p → n + e+ +ν e

The corresponding decay is written as

(Z , A) → (Z − 1, A) + e + + ν e
The daughter nucleus now contains one proton less; therefore the atomic number has decreased
by one unit. There is also a third process called electron capture. In this process an atomic
electron that is "close" to the nucleus is captured by the nucleus

p + e− → n +ν e

Fig. 2.1. Experimental β- spectra obtained from decaying 64Cu. Explain the difference between the two
energy distributions.

3
A basic characteristic of the β-decay process is the continuous energy spectrum of the β -
particles. This is because the available energy in the decay is shared between the β - particle and
the neutrino or antineutrino. Typical energy spectra are shown in Fig. 2.1. The positrons emitted
in β+ - decay combine with electrons and annihilate resulting in emission of gamma rays which
are detected in the PET detector system (see more details in following chapters).

• Find (if possible) other β+ spectra of positron emitters that are used in PET applications:
15
O, 11C, 13N etc.
• Read about production of positron emitters (6.4.2).
• How does the β+-spectrum influence the resolution of PET?

2.1 Annihilation Radiation

Fig 2.2 Gamma spectrum of 22Na recorded with a BaF2 detector. How do you explain the "peaks" in the
spectrum?

The annihilation of positrons with electrons can according to quantum theory produce high-
energy photons, gamma quanta. Annihilation might take place directly or via the formation of
positronium, a state in which an electron and a positron form a “light hydrogen atom” bound by
4
Coulomb attraction. In the lowest energy states the total spin of the system (the vector sum of the
two intrinsic spins) is 0 or 1 ( h ) depending on whether the spins are anti-parallel or parallel. If
the annihilation occurs in the spin 0 state, the process is very fast (< 100 ps) and two gamma
photons are created and emitted at 180 degrees, each with an energy corresponding to the mass
of the electron, 511 keV. The emission directions are due to the conservation of momentum.
Before annihilation both the total linear momentum and angular momentum are zero (if the β -
particles can be considered to be at rest) so the total linear momentum and angular momentum
after annihilation must both be zero. This is accomplished by emission of two spin 1 photons in
opposite directions. At annihilation in the spin 1 state three gamma photons are created and the
available energy is shared between the three photons. Fortunately, this is a less probable process
and of no interest in PET applications, since the three photons cannot easily be used to
reconstruct the annihilation point.

22
Fig 2.3 Level scheme showing the decay of Na.

3 Detectors
3.1 Interactions of Gamma Radiation with Matter
The radiation to be detected in PET is gamma photons and these interact with matter by three
basic interactions: the photoelectric effect, Compton scattering and pair production. The energy
of the gamma photons in PET (511 keV) is too low for pair production to occur, so the
interactions are mainly the photoelectric effect and Compton scattering. (read more about these
processes in the course literature or in instruction to lab. 1). These basic processes are the same
for any type of electromagnetic radiation and therefore occur also in other medical imaging
modalities like traditional X-ray, CT and gamma camera. A general problem with the Compton
scattering process is that it deteriorates the resolution of images.

5
3.2 Scintillation Detectors (Webb 6.2.2 and 6.3.5)
The scintillation detector was developed to detect nuclear radiation. A scintillator emits light
when it is hit by nuclear radiation. The light is emitted from the ultra violet to the infrared range
of wavelength (100 - 800 nm). There are several scintillating materials in use today: organic
crystals, organic liquids, plastics, inorganic crystals, liquids, gases and glasses. The advantage of
inorganic scintillators lies in their greater stopping power due to their higher density and higher
atomic number Z. They also have some of the highest light outputs (number of photons emitted
in the visible range of wavelength). High light output results in better energy resolution, since the
broadening of the energy spectra is basically statistical, i. e. proportional to the square rooth of
the number of photons. This makes them extremely suitable for the detection of gamma rays.
The scintillator most often used in PET applications is Bi4Ge3O12 (Bismuth Germanate or BGO).
BGO has a much greater stopping power than BaF2 (Barium Fluoride) and this gives BGO an
advantage in terms of efficiency. On the other hand BaF2 has a much faster time response, which
makes it more suitable for applications where good time resolution is needed. As a result, BGO
is used in ordinary PET systems where there is no need for extremely good time resolution, but
where the high stopping power of BGO makes it more efficient and cost effective. In TOFPET
applications, the time resolution is of crucial importance so here the choice is BaF2, the currently
fastest known scintillator suitable for detection of 511 keV gamma rays. In table 3.1 the
properties of BGO and BaF2 are listed and compared with those of NaI(Tl).

Table 3.1 Data on common scintillation materials used in PET and TOFPET applications

Scintillator Density Effective Z Relative Decay Wavelength

(g cm-3) light yield constant of emission
(ns) (nm)
Sodium 3.67 50 100 230 410
Iodide
(NaI)
Bismuth 7.13 74 12 300 480
Germanate
(BGO)
Barium 4.89 54 5 0.6 - 0.8 220 (195)
Fluoride 15 630 310
(BaF2)

The emission of light after excitation and ionisation in the scintillating crystal can often, as a first
approximation, be expressed as

No ⎛−t ⎞
N (t ) = exp⎜⎜ ⎟⎟
τd ⎝τd ⎠

N is here the number of photons emitted in the visible spectrum at time t, No is the total amount
of photons emitted, and τd the decay constant of the emission. The rise-time has been
approximated to be negligible. How does this function look? Make an outline!

3.3 Detectors for our PET system

6
The scintillators used in our PET system are made of BaF2. BaF2 has two light components (see
table 3.1). Therefore a more correct description of the light emission is

⎛−t ⎞ ⎛−t⎞
N (t ) = A ⋅ exp⎜ ⎟ + B ⋅ exp⎜⎜ ⎟⎟
⎜τ ⎟
⎝ f ⎠ ⎝ τs ⎠

Here τf and τs are the decay constants of the fast and slow components respectively. It is of
course the fast component that is used to obtain the fast timing, whereas the total signal should
be used for the low energy discrimination.

Measurements of time differences between events can be made with higher accuracy if the
detector is fast. A fast detector also accepts higher counting rates since the dead time is reduced.
This last feature makes it possible to reduce the dose given to a patient during an examination.

A typical scintillation detector consists of a scintillating crystal coupled to a photomultiplier tube

(PMT, see fig. 3.1), a resistor chain that distributes the voltages to the dynodes, and all this
housed in a metallic shield. For details see T. Bäck, et al: A TOF-PET system for educational
purposes, Nuclear Instruments and Methods A477 (2002) 82
When a 511 keV photon enters the detector it will deliver all (photo effect) or a fraction
(Compton scattering) of its energy to an electron. The electron will ionise and excite the atoms in
the crystal, causing the crystal to scintillate and emit a small flash of light. When the emitted
light photons reach the photo cathode of the PMT, photo electrons will be emitted. These
electrons are accelerated towards the first dynode with a voltage (difference) of ≈100 V. At the
dynode more electrons are released. This multiplication process continues via 8 - 14 dynodes
(depends on type of PMT) until the growing electron cloud finally reaches the anode, where the
signal is collected. The PMT thus converts the light into an amplified electronic pulse. (See
figure 3.1). Standard electronic devices can further process this pulse and information on the
impinging radiation can be extracted. (See next chapter: 4. Electronics)

The PMT (see fig. 3.1) will give an electric pulse proportional to the number of scintillation light
quanta that reach the photo cathode (i. e. deposited energy). Thus energy spectroscopy is
possible with a scintillation detector. In our PET-system, one can by energy discrimination,
suppress Compton scattered gamma rays that otherwise would produce noise in the image (see
next chapter).

7
 Fig 3.1 Principle of operation of a photomultiplier tube (PMT). In our application the light source is a
scintillation crystal. Together the BaF2 crystal and PMT form a detector for gamma rays.

4 Electronics
4.1 Discrimination of pulses with low amplitude or originating from noise

As mentioned in chapter 2 Compton scattered gamma photons give a worsened resolution in PET
applications. Therefore it is an advantage if one can discriminate PM pulses originating from
events caused by the Compton effect. Since only a fraction of the energy of the original gamma
photon is given to the electron in a Compton event, the pulses from the PM tube with low
amplitude contain Compton events and also noise pulses. Such pulses are not wanted and can be
taken away by a discriminator. In the discriminator a voltage level is set. Only pulses that are
bigger than this level are accepted and give an output pulse from the discriminator. In our
TOFPET application we use a special type of discriminator called a "constant fraction
discriminator" (CFD). The CFD gives a good definition of the time of arrival of the electric pulse
from a PM. Simultaneously the CFD performs an energy discrimination of pulses of low
amplitude. For those who are interested a brief description of a CFD is given below.

8
4.2 Constant Fraction Discriminator

Fig. 4.1. The formation of the constant-fraction signal

In order to measure time intervals precisely, the arrival times of the different events must be
derived exactly in order to achieve optimum time resolution. This is the function of the Constant
Fraction Discriminator (CFD). The output pulse, from the anode of the PMT, is fed to the input
of the CFD. In figure 4.1 the principle of operation of a CFD is illustrated.

The CFD is designed to trigger on a certain optimum fraction of the pulse height, thus making
the performance of the CFD independent of pulse amplitude. The input signal to the CFD is split
into two parts. One part is attenuated a fraction ƒ of the original amplitude V, the other part is
delayed and inverted, see Fig. 4.1. These two signals are subsequently added to form the
constant-fraction-timing signal. The delay is chosen to make the optimum fraction point on the
leading edge of the delayed pulse line up with the peak amplitude of the attenuated pulse. These
two signals are subsequently added to form a bipolar pulse. The zero crossing occurs at a time
after the arrival of the pulse that is independent of amplitude. The constant-fraction discriminator
incorporates a timing discriminator that triggers on the zero crossing and produces an output
logic pulse that serves as the time marker. In addition a leading-edge discriminator provides
energy selection. This energy selection constitutes the energy threshold used to suppress
Compton scattered gamma photons in the PET-system. No events with energy below the
threshold will give rise to a signal from the CFD and thus will be excluded. The thresholds are
set individually for all detectors by software.

9
 Stop

Fig 4.2 Block diagram of detector arrangement and electronic modules

4.3 The PET Data Acquisition System

In figure 4.2 the different parts of the PET system of detectors and electronics are illustrated.
More details are found in figure 4.3. The Time to Digital Converter (TDC) converts a small time
difference (ps - ns) to a digital value that can be handled numerically.

The TOF-PET system contains 48 cylindrical BaF2 crystals with a diameter of 15 mm and a
length of 20 mm. These scintillators are optically coupled to the PMT with help of a silicon
grease that transmits light down into ultra violet (UV) wave lengths. The PMTs are Hamamatsu
R2076 and are equipped with 19 mm diameter synthetic silica windows. These windows are also
transparent to the fast UV component of BaF2.

10
 Fig 4.3 Detailed block diagram of the different electronic parts of the PET system.

The PET Logic Unit (PET LU) is an interface between the 48 outputs of the CFDs and the
START and 8 STOP inputs of the TDC. The PET Logic Unit (PLU) generates the common start
signal to the TDC. The PET LU is located on a VME card, specially designed for the read out of
our TOF-PET system.
Six detectors are ORed together to form a group (G1 – G8). The 48 detectors make eight groups.
As can be seen any of the 48 detectors that give a pulse can start the TDC. A true coincidence
event cannot occur within a group of six detectors, but can occur with any detector in all the
other groups.
A hit in any of the 48 detectors must have a chance to start the measurement. The only condition
is that the signal should be above the threshold of the CFD. Therefore all of them are ORed in
the PET LU. This gives the common START to the TDC. It is the first signal that arrives from
the CFDs to the PET LU, (after that the system has opened for a new event), that will start the
TDC. The "first" signal will also stop itself in the corresponding stop input to the TDC. In this
way the detector that started the new event will be identified by its delay in its own stop channel.
(See figure 4.3 and the description below).
As already mentioned the incoming signals are Ored in groups of 6, providing 8 STOP signals to
the stop inputs of the TDC. The full time range of the TDC is 90 ns and the resolution is 25 ps
according to the manufacturer. The detector group, that produced the STOP signal, is identified
by its TDC stop input number (G1-G8). Within a group an individual detector is identified by its
11
specific time delay: 10, 20, 30, 40, 50 or 60 ns. For example, if the first detector in a group
stopped the TDC, the result of the time to digital conversion is 10 ns plus the flight time. For the
second detector the result is 20 ns plus flight time and so on. Since the flight time is ≤ 2 ns each
detector is easily identified within a time window of 10 ns. As seen in figure 4.3 the different
delays are fixed and included in the PET LU electronics. They cannot be altered by software.
With a pair of BaF2 detectors a best time resolution of 340 ± 10 ps has been achieved. Real
coincidence events between two detectors are called doubles. A detector that starts the TDC but
does not get a stop from any other detector group gives an event called a single.

The software stores the time and channel number of all TDCs that were stopped before the Full
Scale Time Range was reached.

4.4 Event Building and Software

Fig 4.4 Block diagram illustrating soft- and hardware of the data acquisition system.

A dedicated C-program in the VME CPU reads the data from the TDC and builds the raw events.
The 68030 VME CPU runs the operating system OS9000 suitable for real time event processing.
The Data are sent to a Digital Unix workstation via a dual port RAM card. All electronics are
VME based. The workstation receives the data via a PCI-based interface, and processes the raw
events with a C-program. This program is linked to a MATLAB user interface, where the user
can control the acquisition system. The user interface also includes a picture of the PET detector
array and provides a feedback by drawing every coincidence line between the two individual
coincident detectors, event by event. The event stream is also stored in a list mode file for off
line processing.

12
5 The Principles of PET and TOFPET
The first step in creating a PET image is to inject radioactively labelled tracer molecules into the
body. Different metabolic processes and organs may be examined, depending on the tracer. The
positrons emitted in the following β -decay travel a short distance (depending on the kinetic
energy of the positron) before it annihilates with an atomic electron in the tissue. As mentioned
above two gamma photons are then emitted in 180o directions. Some pairs of gamma-rays are
detected in coincidence by two detectors in the ring. The figure in the introduction part shows
examples of annihilation in different radioactive regions of a "body" of radioactivity and the
subsequent detection in one ring of detectors.

Fig. 5.1 Two detectors and a connecting “tube” in which the annihilation event might have taken place.
As is seen the size of the detectors contributes to the spatial resolution of the detector system.

The two detectors in fig. 5.1 define a line (tube) along which the annihilation has taken place.
Every detected pair of photons represents such a line and the assemblies of lines map in some
sense the injected activity. In regions where the density of lines is high, there is a high
probability of being activity. Now, the line density is not high only in regions of activity. There
will be regions of high line density not originating in activity. Those constitute a kind of noise
and the aim of the image processing is to suppress noise regions as far as possible. These regions
of "false" activity generally have much less intensity compared with regions of "true" activity.
The assembly of lines contains the image information. This assembly is treated with different
image processing tools to produce the final image. These principles are explained in detail in the
diploma work that can be found on the nuclear physics homepage. The simplest way of
processing is to draw all LORs (see fig. 5.1).The LORs will cross at regions of high activity,
other regions can be suppressed by a discriminating threshold.

5.1 TOPFET
With the information of the difference in time of flight for the two detected gamma quanta, one
can locate the point of highest activity to a point along the LOR. Usually the intensity along the
line is given as a gaussian distribution. This defines regions of activity which are less probable
than others. The time resolution is not good enough to sharpen the image, but good enough to
reduce the noise. This subject will be discussed and demonstrated further during the laboratory
exercise.

13
6. The Laboratory Exercise
In order to be prepared for the exercise, you should read about PET in this instruction, in your
textbook and on the internet, for example on the pages of Karolinska Sjukhuset (KS) in
Stockholm and at Akademiska sjukhuset (Petcenter) in Uppsala where PET is used in research
and diagnostics.

The exercise will start with a discussion about the physics behind the PET technique. We will
see how the detectors respond to the gamma-rays and how a spectrum may be created from that
response. We will also briefly discuss the electronics and acquisition system used to extract the
data from the PET system.

After this we will run the PET system and collect data, using 1 and 2 22Na point sources. We will
compare images reconstructed with TOF information to non-TOF images and compare the
response of the system depending on the positions of the sources.

6.1 The Graphical User Interface

The data acquisition (DAQ) and image reconstruction is done in a single graphical user interface
(GUI). Following are the main controls that will be necessary for the exercise.

Start and Stop

Whenever the DAQ is running, the data on the coincident events will be saved to a ASCII file.
The data saved is the numbers identifying the two detectors, the time difference between the
detectors (used for TOF images) and the positions of the ring and the source. Whenever the DAQ
is restarted, the current file will be appended, so make sure that you either change the file or
erase the file when a new measurement is started (otherwise your images will show a
superposition of different measurements).

Singles and Doubles monitors

Here you can see the number of detection events per second without the coincidence condition
(singles) and coincident events (doubles). Keep an eye on these when the source position is
changed and when new sources are added. These monitors show the average number since the
start of the acquisition.

Translation
The platform supporting the sources is driven by a step motor along the normal to the ring. There
is a narrow disk of space inside the detector ring defined by the diameter of the scintillator
crystals that can be imaged. It will be necessary to move the source into this area.

Rotation
The detectors are placed rather sparsely along the ring, this limits the number of possible
projections of the source. If the ring is rotated by a small angle, additional projections can be
obtained.

Reconstruction
Simply press on PET or TOFPET to reconstruct a back-projected image using the data from the
current file.

14
6.2 To do before the exercise in the laboratory
• Read about the physics of PET
• Find out something about different applications of PET
• You can search the Web for information: PET Center in Uppsala, Karolinska Inst in
Stockholm
• Try to answer the following questions (to be discussed at the beginning of the laboration):
• How can we make use of the beta decay in PET?
• How does a scintillation detector work?
• How does a photo multiplier tube (PMT) work?
• What is a constant fraction discriminator?
• What is a time-to-digital converter (TDC)?
• What are the advantages/disadvantages of PET in medical applications?
• What are the main differences between our PET system and a commercial PET
system?

6.3 The measurements to be done during the exercise

For all measurements save the reconstructed images with and without the TOF data.

Source in the centre

The assistant will place the source so that it is on a line along the axis of the ring. Use the
translation controls to put the source into the sensitive space, how thick is this space. Use the
event counters to make sure the source is in the optimal position.

Source off centre

The source will be moved to a new position. Compare what you see with the previous
measurement. Try rotating the ring. Collect data into the same file at several angular positions.
Explain the changes you see in the images.

Two sources
We will then put 2 sources in the plane of the ring. Again, see if the images can be improved by
rotating the ring.

7. The Report
The lab report should contain a description of the experiment performed and the principle of the
PET technique. You do not have to include every image taken during the exercise n the report,
but use enough to demonstrate all features and problems you found. In particular, a comparison
between images using time-of-flight information and those without must be included. Discuss
the advantages and disadvantages of using the TOF technique in medical PET systems. Discuss
whether the rotation of the ring has improved the images, are there ways to improve the images
further?

15