design of a dual-mode tracking device for online dose ... mechanics. the type of processes permitted...
TRANSCRIPT
Sapienza Universita di Roma
Dipartimento di Ingegneria dell’Informazione, Elettronica e
Telecomunicazioni
Dottorato di ricerca in Elettromagnetismo XXVI ciclo
Design of a dual-mode tracking device for onlinedose monitoring in hadrontherapy
Luca Piersanti
Coordinatore:
Prof. Paolo Lampariello
Docente guida:
Prof. Luigi Palumbo
Tutor:
Prof. Adalberto Sciubba
Experience is what you get when you didn’t get what you wanted.
And experience is often the most valuable thing you have to offer.
– Randy Pausch, The Last Lecture
Contents
Introduction 5
1 Interaction of radiation with matter 91.1 Heavy charged particles . . . . . . . . . . . . . . . . . . . . . . . . 11
The Bethe-Bloch formula . . . . . . . . . . . . . . . . . . . . . . . 11
Energy dependence . . . . . . . . . . . . . . . . . . . . . . . . . . 12
Range . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 15
Nuclear fragmentation . . . . . . . . . . . . . . . . . . . . . . . . 16
1.2 Electrons and positrons . . . . . . . . . . . . . . . . . . . . . . . . 17
Collisional and radiative energy loss . . . . . . . . . . . . . . . . . 17
Multiple Coulomb scattering . . . . . . . . . . . . . . . . . . . . . 18
Backscattering of low energy electrons . . . . . . . . . . . . . . . . 20
Positron interactions . . . . . . . . . . . . . . . . . . . . . . . . . 21
1.3 Photons . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 22
Photoelectric absorption . . . . . . . . . . . . . . . . . . . . . . . 22
Compton scattering . . . . . . . . . . . . . . . . . . . . . . . . . . 23
Pair production . . . . . . . . . . . . . . . . . . . . . . . . . . . . 25
Attenuation coefficients . . . . . . . . . . . . . . . . . . . . . . . . 26
2 Radiotherapy and Hadrontherapy 272.1 Physical aspects of radiation therapy . . . . . . . . . . . . . . . . . 31
Absorbed dose . . . . . . . . . . . . . . . . . . . . . . . . . . . . 31
Energy deposition . . . . . . . . . . . . . . . . . . . . . . . . . . . 32
2
CONTENTS 3
Lateral beam spread . . . . . . . . . . . . . . . . . . . . . . . . . . 33
2.2 Biological aspects . . . . . . . . . . . . . . . . . . . . . . . . . . . 35
Ionization density . . . . . . . . . . . . . . . . . . . . . . . . . . . 35
Relative Biological Effectiveness and cell survival curves . . . . . . 37
Oxygen Enhancement Ratio . . . . . . . . . . . . . . . . . . . . . 39
2.3 Protons or 12C . . . . . . . . . . . . . . . . . . . . . . . . . . . . 39
2.4 Beam delivery techniques . . . . . . . . . . . . . . . . . . . . . . . 42
Gantries . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 45
Treatment of moving targets . . . . . . . . . . . . . . . . . . . . . 46
2.5 Dose monitoring . . . . . . . . . . . . . . . . . . . . . . . . . . . 48
PET . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 49
Prompt photons . . . . . . . . . . . . . . . . . . . . . . . . . . . . 50
Charged particles . . . . . . . . . . . . . . . . . . . . . . . . . . . 51
3 Measurement of secondary radiation 533.1 Experimental setup . . . . . . . . . . . . . . . . . . . . . . . . . . 54
Start Counter . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 55
Drift Chamber . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 56
Calorimeter . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 57
Data acquisition . . . . . . . . . . . . . . . . . . . . . . . . . . . . 58
Angle of detection . . . . . . . . . . . . . . . . . . . . . . . . . . . 59
3.2 Prompt photons . . . . . . . . . . . . . . . . . . . . . . . . . . . . 60
Calibration . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 60
Energy measurement . . . . . . . . . . . . . . . . . . . . . . . . . 61
Rate measurement . . . . . . . . . . . . . . . . . . . . . . . . . . . 63
Preliminary data with GSI 220 MeV 12C beam . . . . . . . . . . . 64
3.3 Charged secondary particles . . . . . . . . . . . . . . . . . . . . . 66
Particle identification and fluxes measurement . . . . . . . . . . . . 66
Bragg peak position monitoring . . . . . . . . . . . . . . . . . . . 69
Charged particles production region . . . . . . . . . . . . . . . . . 71
4 Dose Profiler optimization 774.1 Detector overview . . . . . . . . . . . . . . . . . . . . . . . . . . . 78
Neutral radiation operation mode . . . . . . . . . . . . . . . . . . . 79
Charged particles operation mode . . . . . . . . . . . . . . . . . . 81
4.2 Simulation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 82
Dose Profiler layout optimization . . . . . . . . . . . . . . . . . . . 83
4 CONTENTS
5 Event reconstruction and detector performance 935.1 Event reconstruction . . . . . . . . . . . . . . . . . . . . . . . . . 94
Prompt photons . . . . . . . . . . . . . . . . . . . . . . . . . . . . 94
Charged particles . . . . . . . . . . . . . . . . . . . . . . . . . . . 103
5.2 Detector performance evaluation . . . . . . . . . . . . . . . . . . . 104
Prompt photons . . . . . . . . . . . . . . . . . . . . . . . . . . . . 104
Charged particles . . . . . . . . . . . . . . . . . . . . . . . . . . . 105
Conclusions 109
Appendix A Kalman filter formalism 111
Bibliography 113
Introduction
Hadrontherapy, or ion therapy, is an emerging technique for cancer treatment
that exploits accelerated ions (mostly protons and carbons) instead of X-rays, as in
usual radiotherapy. This choice appears particularly advantageous when the pecu-
liar energy deposition in matter of charged particles is compared with the one of
photons. The former is characterized by an initial flat low deposition plateau, when
particle velocity is still high, and a very narrow peak (named after W. H. Bragg who
discovered it in 1903) that sharply rises at the end of particle’s range. The latter,
instead, shows a typical exponential trend that, after a broad peak centered in the first
2 ÷ 3 cm, depending on photon initial energy, decreases with increasing penetration
depth. Thus, the higher deposition selectivity of charged particles has made them
particularly appealing as projectile candidates for a new generation of therapy, since
a substantial sparing of the healthy tissues surrounding the tumor could be achieved
with respect to standard radiotherapy. Moreover, the possibility to dynamically shift
the Bragg peak depth, varying ions kinetic energy, opened the door to new and more
precise beam delivery techniques, named active scanning, that may allow to mini-
mize the number of passive devices on the beam line, reducing also patient clinical
complications induced by secondary radiation.
Unfortunately, radiotherapy’s standard monitoring techniques cannot be used in
this novel context. These rely exclusively on the fraction of primary beam transmit-
ted through the patient, that in a hadrontherapy treatment is negligible. For the afore-
mentioned reason, a great effort has been made in order to find the best approach to
monitor, possibly online, hadrontherapy treatment quality. Any novel approach must
rely solely on the secondary particles produced during the interaction of the beam
5
6 INTRODUCTION
with the patient.
The only technique that has been used in a clinical environment so far exploits the
collinear emission of photons due to positron annihilation. The e+ are due to the β+
decay of several radioactive isotopes (mostly 11C and 15O) produced by projectile
and target nuclear interaction. Back-to-back photons detection, usually performed
with detectors made of two “heads”, allows to estimate an activity map of the patient.
Then, an overall treatment quality assessment can be carried out once a correlation
between activity and deposited dose has been established (effectively and reliably
evaluated via Monte Carlo simulations). This approach, has been performed so far
only after the irradiation, placing the patient in a conventional PET scanner (this is
the reason why this technique is commonly referred to as PET). Such solution has
several drawbacks: (i) as time passes by, radioactive isotopes distribution is blurred
by patient metabolism (metabolic washout); (ii) the isotopes relatively short half-life
(11C ≈ 20 minutes and 15O ≈ 2 minutes) poses serious issues on the collectable
data sample, also considering the time needed to move the patient from the treatment
room to the PET ring; (iii) the isotopes low activity, if compared to standard PET,
requires longer acquisition times for data collection.
Other techniques have been recently proposed exploiting other secondary radia-
tion sources, such as prompt photons, emitted after nuclei de-excitation, or charged
particles, emitted after target or projectile fragmentation. Both approaches are un-
der study in order to assess their potential for online dose monitoring in a clinical
context. Prompt photons are emitted isotropically with an energy range that roughly
spans from 1 to 10 MeV. This makes their detection challenging, posing mechanical
issues (as collimators design and weight) in the detector development phase. Charged
particles, on the other hand, are easier to detect and to be tracked, but must exceed a
kinetic energy threshold needed to leave the patient body, that could actually limit the
available statistics and hence the spatial resolution on the dose release reconstruction.
The aim of this thesis is the description and the review of the preliminary mea-
surements on beam, the design, the subsequent optimization and the performance on
Monte Carlo data of a novel device able to detect both prompt photons and charged
particles for online dose monitoring in hadrontherapy applications. The reason of
such double functionality, never suggested so far, lies in the increased statistical sam-
ple that could be available with the detection of the two radiation sources. However,
this particular choice comes at a cost of a more complex design and mechanical re-
alization, as it will be pointed out in the following chapters. The design of this novel
detector has been included in the INSIDE project and it has been awarded a PRIN-
7
MIUR funding for the realization and the operation of such detector, together with
an online PET system, at CNAO treatment room.
In the first two sections a brief reminder of the main interaction mechanisms be-
tween radiation and matter (in the energy range of interest for medical applications)
and a state of the art of hadrontherapy and dose monitoring techniques will be pre-
sented. Then, the results of the feasibility measurements on beam (already published
on several peer reviewed journals) performed at LNS-INFN and GSI (Darmstadt)
laboratories, will be shown in the third chapter. The fourth and fifth will be devoted,
instead, to the project detailed description, and in particular to the layout optimiza-
tion, the reconstruction software development and the performance evaluation.
CHAPTER
1
Interaction of radiation withmatter
Content
1.1 Heavy charged particles . . . . . . . . . . . . . . . . . . . . . . . . 11
The Bethe-Bloch formula . . . . . . . . . . . . . . . . . . . . . . . 11
Energy dependence . . . . . . . . . . . . . . . . . . . . . . . . . . 12
Range . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 15
Nuclear fragmentation . . . . . . . . . . . . . . . . . . . . . . . . 16
1.2 Electrons and positrons . . . . . . . . . . . . . . . . . . . . . . . . 17
Collisional and radiative energy loss . . . . . . . . . . . . . . . . . 17
Multiple Coulomb scattering . . . . . . . . . . . . . . . . . . . . . 18
Backscattering of low energy electrons . . . . . . . . . . . . . . . . 20
Positron interactions . . . . . . . . . . . . . . . . . . . . . . . . . 21
1.3 Photons . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 22
Photoelectric absorption . . . . . . . . . . . . . . . . . . . . . . . 22
Compton scattering . . . . . . . . . . . . . . . . . . . . . . . . . . 23
Pair production . . . . . . . . . . . . . . . . . . . . . . . . . . . . 25
Attenuation coefficients . . . . . . . . . . . . . . . . . . . . . . . . 26
In this chapter the basic reactions, and their effects, which occur when radiation
encounters matter will be briefly reviewed. These processes are the basis for all cur-
rent particle detection devices and can characterize the sensitivity and the efficiency
of the detector itself. Moreover, these same reactions may also disturb and inter-
fere with a measurement: for instance, by causing energy information to be lost or
9
10 CHAPTER 1. INTERACTION OF RADIATION WITH MATTER
scattering a particle from its original trajectory. Finally these are also the processes
involved when living matter is exposed to radiation. Then, the knowledge of these
interactions is of enormous importance for experimental design of a new detector and
its optimization.
Radiation sees matter in terms of its basic constituents, depending on radiation
nature, energy and type of material, reactions may occur with the atoms, or the nu-
clei, or with their individual building blocks with a probability governed by quantum
mechanics. The type of processes permitted to each type of radiation define, among
other things, the penetrability through matter, the danger to biological organisms, the
difficulty or ease of detection and so on. The most common processes that character-
ize charged particles and photons passage trough matter are by far electromagnetic
interactions, in particular, inelastic collisions with the atomic electrons.
Because of their rather different nature is particularly useful to separate charged
particles in two classes: electrons and positrons and heavy particles (heavier than
electron, such as protons, α particles and other light ions). Because of additional
effects which arise, that are outside the aim of this thesis, heavy ions will be excluded
from this analysis.
Symbol Definition Units or value
T Kinetic energy MeVM Incident particle mass MeV/c2
mec2 Electron mass × c2 0.510998918(44) MeV
re Classical electron radius 2.817940325(28) × 10−15 mρ Density of absorbing material g/cm3
Z Atomic number of absorbing materialA Atomic weight of absorbing material g mol−1
e Elementary charge 1.602176565(35)×10−19 Cz Charge of incident particle in units of eNa Avogadro’s number 6.0221415(10)×1023 mol−1
h Planck’s constant 4.135667516(91)×10−15 eV sβ v/c of the incident particleγ Lorentz factor 1/
√1− β2
I Mean excitation potential eVν Average orbital frequencyδ Density correction to ionization energy lossC Shell correction to ionization energy loss
Wmax Maximum energy transfer in a single collision
Table 1.1: Summary of variables used in this chapter.
1.1. HEAVY CHARGED PARTICLES 11
1.1 Heavy charged particles
In general, there are two main features that identify the passage of charged parti-
cles through matter: energy loss and deflection from particle original direction. These
effects are primarily the result of two processes1: (i) inelastic collisions with the elec-
trons of the material; (ii) elastic scattering from nuclei. Of the two electromagnetic
processes, the inelastic collisions are almost entirely responsible for the energy loss
of heavy particles. Upon entering any absorber, the charged particle simultaneously
interacts with many electrons. For each interaction, the electron feels the attractive
impulse due to the Coulomb force as the particle passes nearby. Depending on the
closeness of such interaction, the impulse can be sufficient either to raise the electron
to a higher-lying shell within the atom (excitation) or to remove it completely from
the atom (ionization). Then the energy transfer comes at the expenses of charged
particle, and its velocity is therefore decreased. The maximum energy that can be
removed from a charged particle of mass M with kinetic energy T and given to an
electron of mass me in a single collision is 4Tme/M , hence very small. However,
the number of collisions per unit path length is so large, that a substantial cumulative
energy loss is observed even for thin layers of material. The net effect is a contin-
uous decrease of particle’s velocity until this is stopped or exits the medium where
it is traveling. Elastic scattering from nuclei also occurs frequently, not as often as
electron collisions though. Typically little energy is transferred in these collisions,
since the mass of the incident particle is usually very small compared to the nuclei
of most materials. Whether this is not the case, some energy is also lost trough this
mechanism. Even if inelastic collisions are statistical in nature, since their number
per unit path length is usually large, the fluctuations in the total energy loss are small
and it is then possible to work with the average energy loss per unit path length. This
quantity, often called the stopping power or commonly dE/dx, was first calculated
by Bohr using classical mechanics (a simplified version is due to Jackson [1]) and
later by Bethe et al. [2] using quantum mechanics.
The Bethe-Bloch formula
Bohr’s classical formula, as calculated by Jackson [1] is given by:
− dE
dx=
4πz2e4
mev2Ne ln
γ2mev3
ze2ν(1.1)
1emission of Cherenkov radiation, nuclear reactions (except nuclear fragmentation) andbremsstrahlung amongst others, will not be discussed in the following
12 CHAPTER 1. INTERACTION OF RADIATION WITH MATTER
whereNe is the electrons’ density. This formula gives a reasonable description of the
energy loss for heavy particles (from α to heavier nuclei). But for lighter particles,
e.g. protons, the equation (1.1) does not work properly because of quantum effects.
However, it contains all the fundamental characteristics of electronic collision loss
by charged particles. The complete quantum mechanical calculation, performed by
Bethe, Bloch and other authors, gives for the energy loss:
− dE
dx= 2πNar
2emec
2ρZ
A
z2
β2
[ln
(2meγ
2v2Wmax
I2
)− 2β2 − δ − 2
C
Z
](1.2)
The maximum energy transfer is possible with a head-on collision, given an incident
particle of mass M , kinematics indicates Wmax to be:
Wmax =2mec
2η2
1 + 2s√
1 + η2 + s2(1.3)
where s = me/M and η = βγ. If M me then
Wmax ' 2mec2η2
The mean excitation potential, I , is one of the most important terms of the Bethe-
Bloch formula and can be defined as the average orbital frequency from Bohr’s for-
mula times Planck’s constant, hν. Since this is a quantity very difficult to evaluate,
it is normally treated as an experimentally determined parameter for each element.
The quantities δ and C are density and shell corrections to the Bethe-Bloch formula
which become important at high and low energy respectively. In the first case, the
electric field of the traversing particle tends to polarize the atoms along its path.
Thus, electrons far from the particle trajectory will be shielded from the full electric
field intensity. Therefore, all the collisions with the outer electrons will contribute
less to the total energy loss than predicted by the equation (1.2). A comparison of
the Bethe-Bloch formula with and without density correction is shown in Figure 1.1.
The shell correction, instead, explains the effects which arise when the velocity of
the particle is comparable or even smaller than the orbital velocity of the atomic elec-
trons. In these cases, the hypothesis that the electron is stationary with respect to the
incident particle no longer holds and the equation (1.2) breaks down. The correction
is generally small, as it can be seen in Figure 1.1.
Energy dependence
An example of stopping power energy dependence is shown in Figure 1.2 which
reports the Bethe-Bloch formula as a function of kinetic energy for different particles.
1.1. HEAVY CHARGED PARTICLES 13
Figure 1.1: Comparison of the Bethe-Bloch formula with (solid) and without (dashed) shelland density corrections for copper. As can be noticed in the low energy region, shell correc-tion contribution is very small, on the contrary, a clear effect can be seen for energies greaterthan 1 GeV. From Leo [3].
The term within square brackets of the equation (1.2) varies slowly with particle
energy. Thus, the general behavior of energy loss can be inferred from the behavior
of the multiplicative factor. For a non relativistic particle, dE/dx therefore varies as
1/β2, or inversely with particle energy, and decreases with increasing velocity2 until
a minimum is reached when v ' 0.96 c. At this point, particles are usually referred to
as minimum ionizing or MIP. As the energy increases beyond this point the term 1/β2
becomes almost constant and dE/dx rises again for the logarithmic dependence of
(1.2). However, this relativistic rise is compensated by the density correction.
When comparing different charged projectiles of the same velocity, the only fac-
tor that may change outside the logarithmic term in equation (1.2) is z2. Therefore
particles with greater charge will have larger specific energy loss. Analyzing differ-
ent materials as absorbers, instead, dE/dx depends mainly on the electron density
of the medium. High atomic number, high-density materials will, therefore, result in
the largest energy loss. Indeed, more energy per unit length will be deposited towards
the end of its path rather than at its beginning.
From Figure 1.2, it is clear that during the slowing down process also particle’s
kinetic energy changes. This effect is shown in Figure 1.3, which shows the amount
2this behavior can be heuristically justified observing that since the particle spends a longer timein the vicinity of any given electron whether its velocity is low, the impulse felt by the electron, and sothe energy transferred, will be larger.
14 CHAPTER 1. INTERACTION OF RADIATION WITH MATTER
Figure 1.2: The stopping power, dE/dx, as function of energy for different particles. FromLeo [3].
Figure 1.3: A typical Bragg curve showing the variation of dE/dx as a function of penetra-tion depth of 5.49 MeV α particles in air (from 241Am radioactive decay). It is clear how theparticles are more ionizing towards the end of their path.
of energy deposited by a heavy particle as a function of its penetration depth inside
the absorbing medium. This is known as a Bragg curve, and, as can be noted, most of
the energy is deposited near the end of the path. At the very end, however, the particle
begins to pick up electrons for part of the time, this lowers the effective charge of the
projectile and thus the dE/dx drops.
1.1. HEAVY CHARGED PARTICLES 15
Range
The penetration depth of a particle traversing a medium before it loses all of its
energy is called range. Since the energy loss of charged particles can be safely as-
sumed to be continuous, this distance must be a well defined quantity, the same for
identical particles with the same energy traveling in the same material. Experimen-
tally, the range can be determined by passing a collimated source of particles at a
fixed energy through the medium under test varying its thickness and measuring the
ratio of transmitted to incident particles. A typical plot of this curve versus absorber
thickness is shown in Figure 1.4. As it can be seen, for small thicknesses, almost
all the particles survive. As the thickness approaches the range value, surprisingly,
this ratio does not drop immediately to zero, as expected of a well defined variable.
The curve slopes down, instead, over a certain range of thicknesses. This fact can be
Figure 1.4: Range (R) vs absorber thickness (t) plot for an alpha particle collimated source.I0 is the incident intensity, I is the transmitted intensity.
only explained assuming the energy loss mechanism not continuous, but statistical in
nature. Thus, in general, two identical particles with the same initial energy traveling
in the same material, will not undergo the same number of collisions and hence the
same energy loss. This phenomenon is known as range straggling. In a first ap-
proximation, the distribution of ranges for a set of identical particles has a gaussian
shape. Its mean value (R) is called the mean range and corresponds to the midpoint
on the descending slope of Figure 1.4. This is the thickness at which roughly half of
the particles pass through the material. In the practice, another quantity of interest is
the thickness at which all the particles are absorbed, this point is usually evaluated
by taking the tangent to the curve at the midpoint and extrapolating to the x axis
intercept (R0). This parameter is usually referred to as the extrapolated range.
16 CHAPTER 1. INTERACTION OF RADIATION WITH MATTER
Nuclear fragmentation
The possibility of accelerating heavy ions to energies of the order of several hun-
dreds of MeV per atomic mass unit, pointed out another process (the nature of which
is not electromagnetic) that needs to be taken into account when an accurate descrip-
tion of the radiation field produced after a collision of such particles with a target is
needed. This process is nuclear fragmentation. Two qualitatively different types of
collisions can be clearly distinguished by experimental observation:
1. the central (or near central) collisions, which comprise about 10% of all cases,
are characterized by an almost complete destruction of both the projectile and
target nuclei. These violent processes are high multiplicity events and a large
number of particles come out over a wide range of angles. In such a collision,
practically all nucleons in both colliding partners are participants;
2. the peripheral collisions, where, by contrast, the momentum and energy trans-
fers are relatively small. Only a few nucleons in the overlap zone effectively
interact during the collision and the number of participants nucleons is small.
The reminder of the paragraph is devoted to the description of this latter case. Ac-
cording to a proposition of Serber [4], inelastic nuclear reactions at relativistic en-
ergies can be described in two steps which occur in two different time scales. The
first interaction may modify the composition of the reaction partners and introduces
a certain amount of excitation energy. The characteristic time of this reaction step
is of the order of 10−23 s. In the second step the system reorganizes, that means it
thermalizes and de-excites by evaporation of neutrons, protons and light nuclei as
well as by fission and emission of gamma rays. According to the statistical model,
the characteristic time for particle emission varies between 10−16 s for an excitation
energy of 10 MeV and 10−21 s at 200 MeV. The abrasion-ablation model, introduced
by Bowman, Swiatecki and Tsang [5] and schematically shown in Figure 1.5, de-
scribes nuclear fragmentation in terms of the two aforementioned stages. Nucleons
in the overlapping zone of the interacting projectile and target nuclei are abraded and
form the hot reaction zone (fireball), whereas the outer nucleons (spectators) are only
slightly affected by the collision. In the second step (ablation), the remaining projec-
tile and target fragments as well as the fireball de-excite by evaporating nucleons and
light clusters. Those emitted from the projectile fragments appear forward peaked in
the laboratory frame, due to the high velocity of the projectile. The projectile-like
fragments continue to travel with nearly the same velocity and direction. Neutrons
1.2. ELECTRONS AND POSITRONS 17
Figure 1.5: A simplified sketch of the abrasion-ablation model of the nuclear fragmentationdue to peripheral collisions of projectile and target nucleus as described by Serber [4].
and clusters from target-like fragments are emitted isotropically and with much lower
velocities. The particles ablated from the fireball cover the range between the projec-
tile and target emission.
1.2 Electrons and positrons
Collisional and radiative energy loss
Similarly to heavy charged particles, also electrons and positrons suffer a colli-
sional energy loss when they traverse matter. But, while the basic mechanism of colli-
sion loss described for heavy charged particles holds true for electrons and positrons,
the Bethe-Bloch formula must be slightly modified for two reasons. Firstly for their
small mass: the assumption made for heavy charged particles that the incident par-
ticle remains undeflected, in fact, invalid. Large deviations in the electron path are
now possible, because its mass is equal to that of the orbital electrons with which it
is interacting. Secondly, since the collisions are between identical particles, a much
larger fraction of energy can be lost in a single encounter. Several terms must be then
changed in the equation (1.2), in particular, the maximum allowable energy transfer
becomes Wmax = Te/2, where Te is the kinetic energy of the incident electron (or
positron).
Furthermore, because of their small mass an additional loss mechanism has to
be considered: the emission of electromagnetic radiation caused by scattering in the
electric field of a nucleus (bremsstrahlung). From classical theory, any charge must
radiate energy when accelerated, and the deflection of the electron (or the positron)
in its interactions with the absorber’s nuclei corresponds to such acceleration. The
18 CHAPTER 1. INTERACTION OF RADIATION WITH MATTER
linear specific energy loss through this radiative process is given by:
−(dE
dx
)rad
=NTeZ (Z + 1) e4
137 (mec2)2
(4 ln
2E
mec2− 4
3
)(1.4)
where N is the atomic number density. This radiation component exists also for
heavy charged particles, but it is generally negligible (at least in the majority of the
practical cases) as suggested by the presence of the mass squared factor in the de-
nominator of the (1.4). The total energy loss of positrons and electrons, therefore,
can be expressed as the sum of two terms:
−(dE
dx
)tot
= −(dE
dx
)rad
−(dE
dx
)coll
(1.5)
The ratio of the two specific energy losses is given approximately by:
(dE/dx)rad(dE/dx)coll
∼=TeZ
700(1.6)
where Te is in units of MeV.
For example, secondary electrons produced from prompt photons interactions
during a hadrontherapy treatment (that are of interest in this work) have energies
lower than 10 MeV. Radiative losses are, therefore, always a small fraction of the
energy losses due to ionization and are significant only in materials with high atomic
number. However, as the energy is increased, the probability of bremsstrahlung
rapidly rises, so that at a few tens of MeV, radiation loss is comparable to or even
greater than the collision-ionization loss. Then, it is possible to define for each ma-
terial a critical energy Ec at which the two radiations equal each other. Above this
energy radiation component will dominate over collision loss and vice-versa below
Ec. An approximate formula due to Bethe and Heitler [2] to estimate Ec is:
Ec ∼=1600mec
2
Z(1.7)
Table 1.2 gives a short list of critical energies for various materials commonly used
in experimental physics applications.
Multiple Coulomb scattering
In addition to inelastic collisions with atomic electrons, charged particles travers-
ing matter also suffer repeated elastic Coulomb scatterings from nuclei even if with a
1.2. ELECTRONS AND POSITRONS 19
Material Critical energy (MeV) Material Critical energy (MeV)
Pb 9.51 Lucite 100
Al 51 Polystyrene 109
Fe 27.4 NaI 17.4
Cu 24.8 Anthracene 105
Air (STP) 102 H2O 92
Table 1.2: Critical energies of some commonly used materials.
rather smaller probability. The cross section of these collisions can be described (as
a first approximation) by the Rutherford formula:
dσ
dΩ= Z2r2
e
mc/βp
4 sin4(θ/2)(1.8)
where θ is deflection angle. Because of its sin4(θ/2) dependence, the majority of
these collisions result in a small angular deflection of the particle, assuming that the
nuclei are much more massive than the incident particle (the energy transfer to the
nucleus is thus negligible). Then the particle follows a random zigzag path as it
traverses the material. The cumulative effect is, however, a net deflection from orig-
inal particle direction. According to the number of interactions Ncoll, three different
situations can occur:
1. single scattering. If the absorber is very thin, such that the probability of more
than one Coulomb scattering is small, then the angular distribution is given by
the simple Rutherford formula in (1.8);
2. plural scattering. If the average number of scatteringsNcoll < 20 then we have
plural scattering. This is the most difficult case to treat since neither Rutherford
formula nor statistical methods can be simply applied;
3. multiple scattering. If the average number of collisions is Ncoll > 20 and
energy loss is small or negligible, the problem can be treated statistically to
obtain probability distribution for the net deflection angle as a function of the
thickness of the material traversed.
The third case is the most frequently encountered in common applications, and the
remainder of this subsection is devoted to this topic. In general, rigorous calcula-
tions of multiple scattering are extremely complicated and several formulations exist.
20 CHAPTER 1. INTERACTION OF RADIATION WITH MATTER
Among the most used are the small angle approximations by Moliere and by Sny-
der and Scott. Their formulations have been demonstrated to be generally valid for
all particles up to angles of θ ' 30. Ignoring the small probability of large-angle
single scattering, a good idea of the effect of multiple scattering in a given mate-
rial can be obtained by considering the distribution resulting from the small angle
(θ < 10) single scattering only. In such case the probability distribution is approxi-
mately Gaussian:
P (θ)dΩ ' 2θ
〈θ2〉exp
(−θ2
〈θ2〉
)dθ (1.9)
The parameter 〈θ2〉 represents the mean squared scattering angle, as can be obtained
integrating: ∫θ2P (θ)dΩ
from θ = 0 to ∞. The square root√〈θ2〉 is known as the RMS scattering angle
and should be equal to the RMS scattering angle of the full multiple scattering angle
distribution. However, since the Moliere distribution has a long tail, the true value is
slightly larger. A better estimate is obtained by using an empirical formula proposed
by Highland [6] which is valid to within 5% for Z > 20 and for target thicknesses
10−3Lrad < x < 10Lrad and gives the sigma of the Gaussian function:
σθ [rad] =14.1 [MeV ]
pβcZ
√x
Lrad
(1 +
1
9log10
x
Lrad
)(1.10)
with Lrad: radiation length of material; x: thickness of material; p momentum of
particle. For low velocities and heavy elements somewhat larger errors (10÷20)%
are obtained.
Backscattering of low energy electrons
Because of their small mass, electrons are particularly susceptible to large angle
deflections by scattering from nuclei. An electron entering one surface of an absorber
may undergo sufficient deflection so that it re-emerges from the surface through
which it entered. This phenomenon is called backscattering. These backscattered
electrons do not deposit all their energy in the absorbing medium and therefore can
have a significant effect on the response of detectors designed to measure the energy
of incident electrons. This phenomenon is more pronounced for electrons with low
energy and absorbers with high atomic number (for non-collimated electrons on high
Z material such as NaI, for example, as much as 80% may be reflected back). Figure
1.6 shows the measured fraction of monoenergetic electrons that are backscattered
1.2. ELECTRONS AND POSITRONS 21
(backscattering coefficient or albedo) when normally incident on the surface of var-
ious media, additional data for materials commonly used as electron detectors are
given in Table1.3.
Figure 1.6: Fraction η of normally incident electrons that are back scattered from thick slabsof various materials, as a function of incident energy E. From Tabata et al [7].
Electron energy (MeV)
Scintillator 0.25 0.50 0.75 1.0 1.25
Plastic 0.08 ± 0.02 0.053 ± 0.010 0.040 ± 0.007 0.032 ± 0.003 0.030 ± 0.005
Anthracene 0.09 ± 0.02 0.051 ± 0.010 0.038 ± 0.004 0.029 ± 0.003 0.026 ± 0.004
NaI (Ti) 0.450 ± 0.045 0.410 ± 0.010 0.391 ± 0.014 0.375 ± 0.008 0.364 ± 0.007
CsI (Ti) 0.49 ± 0.06 0.455 ± 0.023 0.430 ± 0.013 0.419 ± 0.018 0.404 ± 0.016
Table 1.3: Fraction of normally incident electrons backscattered from various detector sur-faces. From Titus [8].
Positron interactions
The Coulomb forces that constitute the major mechanism of energy loss for
both electrons and heavy charged particles are present for either positive or nega-
tive charge on the particle. Whether the interaction involves a repulsive or attractive
force between the incident particle and orbital electron, the impulse and energy trans-
fer for particles of equal mass are about the same. Therefore, the tracks of positrons
in an absorber are similar to those of electrons, and their specific energy loss and
range are about the same for equal initial energies. Positrons differ significantly in
the annihilation radiation (of two collinear 0.511 MeV photons) that is generated at
22 CHAPTER 1. INTERACTION OF RADIATION WITH MATTER
the end of positron track. Because these 0.511 MeV photons are very penetrating
compared with the range of the positron, they can lead to the deposition of energy far
from the original positron track. This unique feature of positrons must be seriously
taken into account in the project and the design of a gamma-ray detector, especially
if gamma-ray energies are in the order of tens of MeV where, as will appear in the
following section, the probability of pair production rapidly rises.
1.3 Photons
Although a large number of possible interaction mechanisms are known for gamma
rays in matter, only three major types play an important role in radiation measure-
ments: photoelectric absorption, Compton scattering and pair production. All these
processes lead to the partial or complete transfer of the photon energy to electron
energy. Then the photon either disappears completely or is scattered through a sig-
nificant angle. This behavior is in evident contrast to the one of charged particles
discussed earlier in this chapter, which slow down gradually through continuous si-
multaneous interactions with many absorber atoms.
Photoelectric absorption
The photoelectric effect involves the absorption of a photon by an atomic electron
with the subsequent ejection of the electron (usually referred to as photoelectron)
from one of the atom’s bound shells. The interaction is with the atom as a whole and
cannot take place with free electrons, since a free electron cannot absorb a photon
and also conserve momentum. For bounded electrons the nucleus absorbs the recoil
momentum instead. The energy of the photoelectron is then:
E = hν −B.E. (1.11)
where B.E. is the binding energy of the electron. For gamma-ray energies of more
than a few hundred keV, the photoelectron carries off the majority of the original
photon energy. Figure 1.7 shows a typical total cross section plot as a function of the
gamma-ray energy, where all the cross section contributes are emphasized. Consider-
ing the photoelectric component, the edge lying highest in energy corresponds to the
binding energy of the K-shell electron. For gamma-ray energies slightly above the
edge, the photon energy is sufficient to undergo a photoelectric interaction in which
a K-electron is ejected from the atom. Below the edge, the cross section drops dras-
tically since the K-electrons are no longer available. Theoretically, the photoelectric
1.3. PHOTONS 23
Figure 1.7: Gamma-ray total cross section for carbon, where: σp.e., σRayleigh, σCompton,κnuc, κe are the photoelectric effect, coherent scattering, incoherent scattering, pair produc-tion in nuclear field and pair production in electron field components respectively.
effect is difficult to treat rigorously, but it is interesting to note the dependence of the
cross section on the atomic number Z. This varies depending on the photon energy,
however, in the MeV range, it goes as Z to the 4th or 5th power. Hence, higher
Z materials are the most favored for photoelectric absorption and, as will be dis-
cussed in later chapters, are an important resource when choosing the best material
for gamma-ray detectors.
Compton scattering
The interaction process of Compton scattering arises when an incoming photon
is scattered on electrons. In matter, of course, the electrons are bound but, if the
photon energy is high, with respect to the binding energy, this latter component can
be neglected and the electrons can be considered as basically free. Hence, the pho-
ton is deflected through an angle θ with respect to its original direction (shown in
Figure 1.8). The photon transfers a portion of its energy to the electron (considered
initially at rest) which is then referred to as recoil electron. Since all angles are
possible, the energy transfer can vary from zero to a large fraction of the primary
energy. The expression of the scattered photon energy as a function of the scattering
24 CHAPTER 1. INTERACTION OF RADIATION WITH MATTER
angle, shown in equation (1.12), and the expression of the photon scattering angle,
as a function of the incoming photon energy, shown in equation (1.13), can simply
be derived by writing the equations for the conservation of energy and momentum.
Using the symbols defined in Figure 1.8:
Figure 1.8: Kinematics of Compton scattering.
hν ′ =hν
1 + γ(1− cos θ)(1.12)
cos θ = 1− 2
(1 + γ)2 tan2φ+ 1(1.13)
where γ = hν/mec2 is the relativistic factor, hν and hν ′ the photon energies before
and after the scattering respectively. The electron kinematics and its scattering angle
can be derived applying energy and momentum conservation as well:
T = hν − hν ′ = hνγ(1− cos θ)
1 + γ(1− cos θ)(1.14)
cotφ = (1 + γ) tanθ
2(1.15)
The probability of Compton scattering per atom of the absorber depends on the num-
ber of electrons available as scattering targets and therefore increases linearly with
Z. The dependance on gamma-ray energy is illustrated in Figure 1.7 for the case of
carbon and generally falls off gradually with increasing energy. The cross section
for Compton scattering was one of the first to be calculated using quantum elec-
trodynamics and is known as the Klein-Nishina formula, that predicts the angular
distribution of scattered photons:
1.3. PHOTONS 25
dσ
dΩ=Zr2
e
(1
1 + γ(1− cos θ)
)2(1 + cos2 θ
2
)·
·(
1 +γ2(1− cos θ)2)
(1 + cos2 θ)[1 + γ(1− cos θ)]
) (1.16)
The distribution is shown graphically in Figure 1.9 and underlines the tendency for
forward scattering at high values of the photon energy.
Figure 1.9: Polar plot of the number of photons scattered into a unit solid angle at thescattering angle θ for several photon initial energies.
Pair production
If the photon energy exceeds twice the electron rest-mass energy (1.022 MeV)
the process of pair production is energetically possible. This involves the transfor-
mation of a photon into an electron-positron pair. In order to conserve momentum,
this can only occur in the presence of a third body, usually a nucleus (sometimes
pair production can occur in the field of an electron, this process is called triplet pro-
duction and has an energy threshold of 4·(mec2) instead of 2·(mec
2) that directly
arises from energy and momentum conservation). As a practical matter, the prob-
ability of this interaction remains very low until the gamma-ray energy approaches
several MeV, therefore pair production is generally confined to high energy gamma-
rays. All the excess energy carried in by the photon above the threshold goes into
kinetic energy shared by the positron and the electron. Since the positron will most
26 CHAPTER 1. INTERACTION OF RADIATION WITH MATTER
likely subsequently annihilate after slowing down in the absorber (depending on the
medium within the process occurs), two annihilation photons are usually produced
as secondary products of the interaction. No simple expression exists for the pair
production cross section, but this varies approximately as the square of the absorber
atomic number Z and, as shown in Figure 1.7, rises sharply with energy.
Attenuation coefficients
The total probability for a photon interaction in matter can be expressed as the
sum of the individual cross sections outlined above. Expressing the cross section per
atom, following the notation introduced in Figure 1.7, this yields:
σtot = σp.e. + ZσCompton + κpair (1.17)
The Compton cross section has been multiplied by Z to take into account all the Z
electrons per atom of the absorber. This is shown (dots) in Figure 1.7 for carbon.
Thus the probability per unit length for an interaction can be obtained multiplying
σtot by atoms density N :
µ = Nσtot = σtot (Naρ/A) (1.18)
This is more commonly known as total absorption coefficient and it is simply the
inverse of the mean free path of the photon. Then the fraction of photons surviving a
depth x can be expressed as:
I/I0 = exp(−µx) (1.19)
where I0 is the incident photons intensity.
CHAPTER
2
Radiotherapy andHadrontherapy
Content
2.1 Physical aspects of radiation therapy . . . . . . . . . . . . . . . . . 31
Absorbed dose . . . . . . . . . . . . . . . . . . . . . . . . . . . . 31
Energy deposition . . . . . . . . . . . . . . . . . . . . . . . . . . . 32
Lateral beam spread . . . . . . . . . . . . . . . . . . . . . . . . . . 33
2.2 Biological aspects . . . . . . . . . . . . . . . . . . . . . . . . . . . 35
Ionization density . . . . . . . . . . . . . . . . . . . . . . . . . . . 35
Relative Biological Effectiveness and cell survival curves . . . . . . 37
Oxygen Enhancement Ratio . . . . . . . . . . . . . . . . . . . . . 39
2.3 Protons or 12C . . . . . . . . . . . . . . . . . . . . . . . . . . . . 39
2.4 Beam delivery techniques . . . . . . . . . . . . . . . . . . . . . . . 42
Gantries . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 45
Treatment of moving targets . . . . . . . . . . . . . . . . . . . . . 46
2.5 Dose monitoring . . . . . . . . . . . . . . . . . . . . . . . . . . . 48
PET . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 49
Prompt photons . . . . . . . . . . . . . . . . . . . . . . . . . . . . 50
Charged particles . . . . . . . . . . . . . . . . . . . . . . . . . . . 51
Radiotherapy is an essential instrument for cancer treatment, used both for cura-
tive and palliative management of patients. Currently more than 50% of all patients
with localized malignant tumors are treated with radiations as part of their initial
therapy, either alone or in combination with chemotherapy or surgery.
27
28 CHAPTER 2. RADIOTHERAPY AND HADRONTHERAPY
Ionizing radiation represents a very effective tool for human cells killing, suffi-
ciently high radiation doses, in fact, can control nearly 100% of the tumors. This
means that, in principle, any tumor can be sterilized providing that a sufficiently high
radiation dose is delivered. Unfortunately, also the healthy tissues surrounding the
target volume are unavoidably exposed to radiations and this limits the amount of
dose that can be actually delivered. Normal tissue complication is, in fact, one of
the crucial parameters that needs to be minimized in a typical treatment, maximizing
at the same time the Tumor Control Probability (TCP). Since human cells have the
capability to repair themselves when exposed to radiation, fractionation of the total
dose represents the strategy normally followed in clinical routine to reduce healthy
tissue complication. A typical radiation regime consists of 60-70 Gy delivered in
30-35 daily fractions of 2 Gy each. Technological improvements in radiotherapy de-
livery in the past years have been focused to widen the therapeutic window (that is
the difference between the TCP and the normal tissue complication probability as a
function of the dose), and in some cases, such as prostate cancer, it has been possible
to double the dose to the target.
In conventional radiotherapy high energy X-rays (up to 25 MeV) are employed
and these are normally produced by electron linacs. The first electron linac was built
for research purposes at Stanford University by W. Hansen and his collaborators in
the 1950s. Soon thereafter, this new tool took place of all other electron-photon
sources. Today about 10,000 linacs are installed and operate in hospitals all over
the world and radiotherapy is used every year to treat about 20,000 patients on a
population of 10 millions. Such enormous development has been possible thanks
to the advancements made in computer assisted treatment systems and in imaging
techniques, such as: Computed Tomography (CT), Magnetic Resonance Imaging
(MRI) and Positron Emission Tomography (PET) scans. Nowadays, the cutting edge
technology in conventional radiotherapy goes by the name of Image Guided Radi-
ation Therapy (IGRT) and it comprises the so called Intensity Modulated Radiation
Therapy (IMRT). This technique takes advantage of at least six (up to nine, in normal
practice) non coplanar and non uniform X-ray fields combined with multi-leaves col-
limator and CT imaging. IMRT allows to reach an extraordinary dose conformation
around the target volume, sparing at most the Organs At Risk (OAR) in its vicinity.
While the capability of X-rays for cancer treatment was perceived shortly after
their discovery in 1895, it was only in 1946 that the potential of hadrontherapy1 was
1Hadrontherapy is a collective word commonly used to indicate all forms of radiation therapy thatmake use of beams of particles made of quarks: protons, neutrons, pions and also heavier nuclei such
29
foreseen by Robert R. Wilson. In fact, he had measured at the Berkeley cyclotron
charged particles depth profiles recognizing a significant increase in dose at the end
of particle range, the so called Bragg peak (see Figure 2.1), which was observed fifty
years before in the tracks of alpha particles by W. Bragg. It is very interesting to
note that in his original paper, that mainly refers to protons, Wilson cites also alpha
particles and carbon ions:
“The intense specific ionization of alpha particles [...] will probably make them
the most desirable therapeutically when such large alpha particle energies are at-
tained. For a given range, the straggling and the angular spread of alpha particles
will be one half as much as for protons. Heavier nuclei, such as very energetic carbon
atoms, may eventually become therapeutically practical”, from Wilson [10].
Figure 2.1: Dose deposition curves for different radiations: typical radiography X-rays(cyan), gamma rays from 60Co radioactive decay (blue), standard 18 MeV radiotherapy pho-tons (green), 135 MeV protons (black) and 250 MeV/u carbon ions (red).
In the following years, researchers at Lawrence Berkeley Laboratory (LBL) con-
ducted extensive tests on protons, pursuing the intuitions of Wilson and in 1954 the
first patient was treated with hydrogen, followed by helium in 1957 and neon ions
in 1975. The first treatments on humans were focused on breast cancer metastatic
patients. In particular, they were aimed at pituitary gland inhibition from producing
hormones that would stimulate the cancer cells to grow. Moreover, this choice was
as helium ions (alpha particles), lithium, boron, carbon, oxygen ions etc. Amaldi [9]
30 CHAPTER 2. RADIOTHERAPY AND HADRONTHERAPY
particularly favorable since the gland exact position was easily localized on standard
X-ray films. At LBL between 1954 and 1975 about 1,000 patients were treated with
protons. Three years after Berkeley, in 1957, the first tumor was irradiated with pro-
tons in Europe at Uppsala cyclotron by B. Larsson. As mentioned before, cancer
Country Facilities in operation Facilities planned Patients treated(or under construction) (operative facilities)
Protons Heavy ions Protons Heavy ions Protons Heavy ions
USA 11 0 9 1 38429Europe 10 2 8 4 19768 983Japan 5 3 1 2 10607 9139Russia 3 0 1 1 6608China 1 1 1 2 1078 194South Korea 1 0 1 1 1041South Africa 1 0 0 0 521Canada 1 0 0 0 170Taiwan 0 0 2 0Saudi Arabia 0 0 1 0South America 0 0 1 0Australia 0 0 1 0
Table 2.1: Geographical distribution of particle therapy facilities. Data from Particle Ther-apy Co-Operative Group 2012 survey [11] and Loeffler et al.[12].
therapy with heavy ions started in the 1970s at LBL, where Cornelius A. Tobias sug-
gested, following Wilson’s intuition, that particles heavier than protons could give
additional advantages (Tobias et al. [13]) and since then is constantly increasing.
The Particle Therapy Co-Operative Group (PTCOG) is an international institution
that regularly monitors the hadrontherapy centers all around the world. Its last sur-
vey, published on March 2013, reports that 93,895 patients have been treated with
protons, more than 10,000 with carbon ions and about 3,500 with other ions. As
can be seen from Table 2.1, the number of hadrontherapy facilities, planned or un-
der construction, is rapidly increasing insomuch as the total could double in the next
decade.
In Europe the development of cancer therapy with protons took place at the Paul
Scherrer Institute (PSI) in Villigen (Switzerland) Pedroni et al. [14], while carbon
ion therapy was pioneered at the GSI Helmholtz Centre for Heavy Ion Research in
Darmstadt Haberer et al. [15], from which the HIT (Heidelberg Ion Therapy center)
Haberer et al. [16] was born. Italy’s first hadrontherapy center, CATANA [17] was
2.1. PHYSICAL ASPECTS OF RADIATION THERAPY 31
built at Laboratori Nazionali del Sud (INFN) in Catania and is aimed to choroidal
and iris melanoma therapy. The first patient was treated in 2002 and, according to
PTCOG statistics [11], 293 patients have been irradiated so far with protons accel-
erated by a super-conducting cyclotron. At the end of 2011 the first patient course
of treatment with protons was completed at Centro Nazionale di Adroterapia Onco-
logica (CNAO) [18] in Pavia. A year later the first carbon ion treatment started, and
nowadays the center normally operates with both particles on a daily basis. Together
with Ion Beam Therapy center (HIT) in Heidelberg, CNAO is the only facility in
Europe that allows both protons and carbon ions therapy.
2.1 Physical aspects of radiation therapy
Absorbed dose
In radiation therapy absorbed dose is one of the fundamental quantities to which
the radiobiological and clinical effects are directly related, regardless of the type of
radiation and the nature of the biological effect. It is defined (ICRU report [19])
as the mean energy deposited (E) by ionizing radiation per unit mass (m) and it is
expressed in gray (1 Gy = 1 J/kg):
D =dE
dm(2.1)
Considering a parallel particle beam with fluence F (dN particles traversing a surface
dS), the dose deposited in a thin slice of absorber material with mass density ρ can
be calculated as follows:
D [Gy] = 1.6× 10−9 × dE
dx
[keV
µm
]× F
[cm−2
]× 1
ρ
[cm3
g
](2.2)
where dE/dx is the additive inverse of the energy loss per unit path length (also re-
ferred to as stopping power) defined in equation (1.2). A quantity closely related to
the stopping power is the Linear Energy Transfer (LET) which refers to the energy
deposited in the target medium by the slowing-down particle and it is measured in
keV/µm. When a charged particle traverses matter several secondary electrons are
produced as a consequence of the process of ionization (usually called delta rays if
their energy is, in turn, high enough to ionize). In some practical cases, one might be
focused only on the energy deposited in the track vicinity, and exclude interactions
that produce delta rays with an energy larger than a certain threshold. This limit is
meant to exclude secondary electrons that carry energy far away from the original
32 CHAPTER 2. RADIOTHERAPY AND HADRONTHERAPY
track. This usually refers to restricted linear energy transfer. If the threshold tends
to infinity, the quantity is called unrestricted linear energy transfer and it equals the
electronic component of the stopping power. Even though LET depends on particle
energy and species, normally photons and protons are referred to as low-LET radia-
tion for their typical sparse ionization density, while carbon ions are called high-LET
particles due to their larger ionization density as shown in Figure 2.2.
Figure 2.2: Comparison of the microscopic structure of proton and carbon tracks at differentenergies with a simplified depiction of a DNA molecule. Ionization and consequently thedamage to the DNA is low at high energies but, for carbons, increases significantly decreasingthe energy. This yields to a clustered damage that is more difficult to repair. For protons,even at low energies, the ionization density is rather sparse, resulting in a lower LET andRBE values. From Fokas [20]
Energy deposition
As already mentioned in the previous chapter and shown in Figure 2.1, the energy
released by photons decreases exponentially with the penetration depth, showing a
peak between 1 and 2÷3 cm depending on the photon initial energy. This shift is
mostly due to buildup processes that arise when the gamma energy grows. On the
contrary, heavy charged particles exhibit a peculiar dose deposition curve, that fol-
lows the Bethe-Bloch equation (1.2). This is characterized by a small amount of
2.1. PHYSICAL ASPECTS OF RADIATION THERAPY 33
energy lost when the particle velocity is high (entry channel), while most of it is re-
leased in a very narrow portion of the path, close to the end of particle range (the so
called Bragg peak). Moreover, being the range a function of the energy, the depth of
the Bragg peak inside the patient can be varied and adjusted by changing the energy
of the beam (as shown in Figure 2.3). Thus, in a radiotherapeutic context, this sharp
Figure 2.3: Measured depth-dose curves in water for carbon ions with different beam ener-gies. From Schardt et al. [21].
and very precise deposition could lead to a better conformation to the target volume
and it could be extremely useful for treatment of deep seated tumors (where photon
irradiation becomes very uneffective), or tumors near OAR. Furthermore, given the
relatively low energy lost along the entry channel, the overall dose delivered to the
healthy tissues surrounding the tumor is lower, as shown in Figure 2.4 and Figure 2.5,
being constant the dose deposited on the tumor.
Lateral beam spread
As described in the previous chapter, the passage of a particle or, in our case
of interest, a particle beam through matter will lead to a generalized diffusion of
the beam itself with respect to its original direction. The beam spread is mainly
caused by elastic Coulomb interactions with the target nuclei (multiple scattering),
while scattering due to electronic interactions can be neglected. For small angles the
angular distribution can be approximated by a Gaussian function whose sigma can
34 CHAPTER 2. RADIOTHERAPY AND HADRONTHERAPY
Figure 2.4: Comparison between desired dose profile (a), photon therapy with a singlefield (b), proton therapy (c) and carbon ion therapy (d) for a given tumor volume (pink) inproximity to an OAR (yellow). A higher conformity to target volume can be achieved withprotons or carbon ions and, at the same time, the OAR receives a much lower dose withrespect to photon therapy.
be obtained from equation (1.10). Hence, targets with heavy elements will cause a
larger angular spread than light elements with the same thickness. In general the
angular spread of heavy charged particles is small for thin targets, but as the energy
decreases it becomes more significant due to the βpc term in the denominator of
(1.10). Considering two different beams with the same range (e.g. 150 MeV protons
and 285 MeV/u carbon ions with R = 15.6 cm) a lateral spread three times larger
can be observed for protons. In general, two different contributions to the overall de-
flection can be distinguished: the scattering from the materials in front of the patient
(beam pipe exit window, external beam monitors, collimators, compensators and air),
and the scattering inside patient tissues, between the entry channel and the stopping
depth. While the former is dominant at low energies, where even a small angular
spread translates in a significant deflection (considering the typical traveling distance
of 0.5÷1.0 m), the latter dominates at high energies, where the penetration depth in
the patient increases. For all the aforementioned reasons, and especially for protons,
the material in the beam path in front of the patient should be minimized. Examples
of Monte Carlo calculations of lateral beam spread for protons and carbon ions are
reported in Figure 2.6 and Figure 2.7.
2.2. BIOLOGICAL ASPECTS 35
Figure 2.5: Comparison of treatment plans for a target volume sited in the skull base: twofields with carbon ions (left) and nine fields with IMRT (photons). Even though a comparabledose conformation can be achieved with both techniques, the use of carbon ions will lead toa dramatic reduction in the integral dose to the surrounding healthy tissues and the sparing ofOAR. From Durante et al. [22].
2.2 Biological aspects
Ionization density
The main difference between photon and heavy ion irradiation is in their micro-
scopic spatial energy distribution. The probability of a ionization event by a photon
within the volume of a single cell is, in fact, very small. This means that a large
number of photons is needed in order to deposit a relevant dose but, since photons
interaction points are randomly distributed, the net effect is that the ionization den-
sity can be assumed to be homogeneous. On the contrary, heavy ions energy spatial
distribution is completely different. It is, in fact, localized and can be divided in
two stages: (a) the emission of secondary electrons (often referred to as δ rays), as
a consequence of Coulomb interaction between projectile and target, (b) δ electrons
scattering inside the medium and their consequent energy loss. The mean free path
for δ rays results of the order of few nanometers, this implies a higher probability
(with respect to photons) for a double ionization to occur on each of the two opposite
DNA strands (whose separation is 2 nm), therefore inducing a more severe damage
to the cell. Moreover, since cells repair capability is reduced if the DNA damage is
36 CHAPTER 2. RADIOTHERAPY AND HADRONTHERAPY
Figure 2.6: Calculated beam spread for carbon ions (red) and protons (blue) in a typicaltreatment beam line. A parallel particle beam (5 mm FWHM) that passes through a nozzle(including a thin vacuum window and beam monitors) and enters a water target placed at1 m distance from nozzle exit has been simulated. At low energies the beam width is mainlydetermined by scattering in the nozzle, while at higher energies the scattering in the targetdominates. Carbon ions show a much smaller spread than protons for the same penetrationdepth. From Schardt [23].
Figure 2.7: Monte Carlo simulation of a 230 MeV proton pencil beam traversing a waterphantom. Picture courtesy of K. Zink.
2.2. BIOLOGICAL ASPECTS 37
complex, radiation damage from heavy ions is larger than photons’. An example of
microscopic dose deposition distributions for X-rays and 15 MeV/u carbon ions is
shown in Figure 2.8.
Figure 2.8: Illustration of the different microscopic dose distribution by X-rays and15 MeV/u carbon ions. In both cases the macroscopic dose is 2 Gy. From Scholz [24]
Relative Biological Effectiveness and cell survival curves
In order to estimate correctly the effectiveness of heavy ions as projectiles, the
definition of Relative Biological Effectiveness (RBE) must be introduced. The RBE
is a very powerful and versatile concept that takes into account and, to some extent
summarizes, several treatment specific parameters, such as: radiation quality, tissue
specific response, biological endpoint (e.g. TCP and normal tissue complication) and
the dose. RBE is defined as the ratio of the dose of a reference radiation (typically60Co γ-rays) and the dose of the radiation under test needed to produce the same
biological effect (this is usually referred to as iso-effect condition).
RBEiso =Dref
Dtest(2.3)
It is of fundamental importance to note that not only is the RBE different in each
biological tissue, but it can be different for every location of the treatment, even
within the same tumor volume. This feature must therefore be taken into account
whenever the treatment is being planned, e.g. developing more sophisticated models
for the determination of RBE values.
A very powerful tool commonly used in radiobiology to compare the different
effects of different radiation types are cell survival curves. These curves illustrate the
relationship between the fraction of cells that maintain their reproductive integrity
and the absorbed dose. Conventionally the surviving fraction (S, defined as the ratio
38 CHAPTER 2. RADIOTHERAPY AND HADRONTHERAPY
of survivor cells and seed cells) is depicted on a logarithmic scale on the ordinate
against dose on a linear scale on the abscissa. The radiation type is characterized by
a different contour shape. Densely ionizing radiations show an almost exponential
relationship between survival and dose, represented as a straight line on the semi-
log plot. On the contrary, sparsely ionizing radiation curves show an initial slope
followed by a shoulder region and then an almost straight line for high values of the
dose. The most common way to parametrize the survival is by means of the Linear
Quadratic (LQ) model developed by Hall [25]:
S(D) =Nsurv
Nseed= e−(αD+βD2) (2.4)
where D is the absorbed dose and α [Gy−1] and β [Gy−2] are two experimental
parameters (that depend on tissue and tumor type) that characterize the initial slope
of the curve and its bending respectively; the ratio α/β defines the shoulder of the
curve and represents the amount of dose for which linear term contribution equals
the quadratic term. Moreover, from these plots it is possible to graphically determine
RBE values of a certain radiation, fixing a determined survival level. An example is
shown in Figure 2.9.
Figure 2.9: Cell survival curves and RBE determination for 10% and 1% survival level fora typical heavy ion (red, dashed) and photon (black, solid) irradiation. Confronting the twocurves for a certain survival rate, it can be noted that RBE is not constant with D even for thesame radiation.
Summarizing, RBE is a very powerful benchmark to describe radiation efficacy
in tumor cells killing. But it must be reminded that it is not a constant value for a
given radiation2: fast moving heavy ions have low LET and hence RBE is approx-
imately one (i.e. close to that of X-rays), slow heavy ions have high LET and then2this holds true for heavy ions. For treatment planning calculations with protons a constant value
of RBE = 1.1 is typically used.
2.3. PROTONS OR 12C 39
are more effective than photons in killing human cells. This can represent an advan-
tage for tumor therapy, since in the entrance channel (where ion velocity is high) the
killing efficiency must be as low as possible, while in the Bragg peak region (where
the ion is about to stop) it will be enhanced.
Oxygen Enhancement Ratio
When tumors are growing in size new vessels need to be generated to supply oxy-
gen to the cells in the tumor core. For various reasons, e.g. vessels are not generated
fast enough or their quality is not good, this can result in hypoxic regions (regions
with lower oxygen level than normal cells) and this poses a demanding challenge in
tumor therapy. Hypoxic regions, in fact, occur frequently in the center of the cancer
mass and are characterized by a larger radio-resistance. This effect is still not well
understood but can be quantified defining the Oxygen Enhancement Ratio (OER):
OER =Dhypoxic
Daerobic(2.5)
where Dhypoxic and Daerobic are the doses with reduced and normal oxygen supply
respectively resulting in the same clinical effect. Typically is around 3 for conven-
tional radiation, while is somewhat lower for heavy ions. In Figure 2.10 are reported
the cell survival studies carried out at LBL laboratories as a preparation for heavy
ion treatments. As can be observed, the difference between hypoxic and normal cells
is reduced for high-LET radiation as their curves tend to converge. Moreover, the
OER decreases as particle energy decreases, as one could expect from highly ioniz-
ing radiation. A consistent behavior has been observed for a wide variety of ions and
cell lineages Barendsen et al. [26], Bewley [27], Furusawa et al. [28] and Staab et
al. [29], where minimum OER values have been found for heavier ions, such as neon
or carbon with respect to light ions (e.g. helium). These lower values are probably
due to the higher radiation damage caused by ion direct hits, that is less sensitive to
the presence of oxygen, compared to the indirect hits induced by free-radicals (typ-
ical of X-rays). In hypoxic regions, in fact, the amount of free-radicals that can be
produced is lower than in normoxic cells, keeping dose constant.
2.3 Protons or 12C
As far as it has been discussed, hadrontherapy seems a promising alternative to
conventional radiotherapy for those applications where the use of photons or surgery
40 CHAPTER 2. RADIOTHERAPY AND HADRONTHERAPY
Figure 2.10: Influence of oxygen level on cell survival of human kidney cells for carbonions at different energies and hence different LET: 33 keV/µm (blue) and 118 keV/µm (red)compared to X-rays (black). Curves based on experimental data by Blakely et al. [30]. FromSchardt et al. [23].
is particularly discouraged (e.g. treatment deep seated tumors or malignancies near
organs at risk). In order to give a wider overview and a deeper insight on the two main
“competing technologies” in hadrontherapy, in this section the main advantages and
disadvantages of protons and carbon ions will be briefly discussed.
As introduced in Chapter 1, heavy ions undergo nuclear fragmentation as a con-
sequence of their interaction with the target (fragmentation in air has a much smaller
impact). This produces a certain amount of low Z fragments that lead, from a treat-
ment point of view, to several drawbacks: fragments have longer range, different
directions and different RBE with respect to primary particles. An overall mitigation
of the beam occurs and this is the reason why there is a visible tail in the Bragg peak
curve of carbon ions in Figure 2.1. On the contrary, protons nuclear fragmentation is
a negligible effect and the relative depth-dose curves show a sharp falloff.
A second physical aspect that must be considered involves multiple scattering.
From equation (1.10) is clear that the deflection is inversely proportional to particle
mass. This is the reason why carbon ions suffer much less lateral beam spread than
protons, as it is shown in Figure 2.11 and Figure 2.12. Therapeutically speaking, a
lower beam spread translates in a more definite dose deposition and hence in a more
precise tumor conformation.
Another key feature of carbon beam is a higher value of RBE with respect to
2.3. PROTONS OR 12C 41
Figure 2.11: X-ray film images of a collimated carbon ion (top) and proton (bottom) beamin water as a function of depth. The blurring effect visible from 7.5 cm bottom film is a clearindication of the higher multiple scattering undergone by protons.
(a) (b)
Figure 2.12: Treatment planning comparison for carbon ions (a) and protons (b). A bet-ter tumor conformation and normal tissue sparing due to lower multiple scattering can behighlighted for carbon ions. Pictures courtesy of GSI (a) and iThemba labs, Cape Town (b).
protons (especially in the Bragg peak region), that makes heavy ions even more ef-
fective in tumor killing and can be easily understood from Figure 2.2, where proton
and carbon ionization tracks are reported for different particle energy. The ionization
of carbon ions is, in fact, so dense that the probability of a double ionization on both
DNA strands is much higher when compared to sparsely ionizing proton tracks. This
leads to a more complex cell damage that is harder to be repaired. Furthermore, the
effects produced with carbon ion irradiation are less sensitive to cells oxygenation,
42 CHAPTER 2. RADIOTHERAPY AND HADRONTHERAPY
being the OER in the Bragg peak region close to 1.
Finally, an undeniable advantage of proton therapy lies in the lower cost of a
proton facility with respect to a carbon ions accelerator. Protons, in fact, can be
accelerated to hadrontherapy’s typical energies by means of cyclotrons (normal or
super conducting) that can easily fit into a hospital environment. Carbon ions, on
the other hand, need a much bigger facility, being their mass twelve times the one
of protons and their magnetic rigidity (defined as R = Bρ = p/q) two times larger
for the same particle velocity. This is the reason why all the existing centers use
a synchrotron (with a diameter of tens of meters) to accelerate ions heavier than
protons. This implies that bigger, more complex and more expensive facilities must
be operated.
2.4 Beam delivery techniques
In order to cover the entire tumor region, the Bragg peak must be spread out
overlapping several beams with different energies. The resulting Spread Out Bragg
Peak (SOBP) aims to provide a constant biological effect within the target volume.
Thus, the Treatment Planning System (TPS) must take into account the variation of
the RBE as a function of the penetration depth and of the beam type. For instance,
the distal part of the volume, that is irradiated only with highly effective ions, will
receive a higher dose with respect to the more proximal regions, for which the to-
tal dose deposited is the sum of the contribution of all the low-RBE traversing ions.
This leads to the flat profile shown in Figure 2.13. In order to homogeneously dis-
tribute the dose on the target area, as planned in the TPS, two main strategies have
been followed in the various ion therapy facilities all over the world: passive beam
modulation and active beam scanning. As the name suggests, passive systems adapt
in three dimensions the beam to the target volume only using passive field shaping
elements (schematically shown in Figure 2.14). The initially narrow beam delivered
by the accelerator is broadened by a scatterer, then the monoenergetic particles are
spread out with a range modulator, in order to cover the whole volume depth. At
this point an additional range shifting can be performed and then a collimator and
a compensator (tailored specifically for each patient) adapt the beam shape to the
target. One of the major limitations of this technique is SOBP fixed width. This, in
fact, can lead to a significant dose deposition outside the target volume (especially in
the proximal part, since the particle range is adjusted to match the distal contours),
as shown in Figure 2.14. Even though this problem could be partially overcome (di-
2.4. BEAM DELIVERY TECHNIQUES 43
Figure 2.13: Spread out Bragg peaks with carbon ions (red) and protons (green) comparedto the dose deposited by photons (blue). From Durante et al. [22]
Figure 2.14: Scheme of a fully passive modulation delivery system. All the principal ele-ments are outlined: the scattering system that broadens the beam, the range modulator forenergy modulation and the range shifter to spread out the Bragg peak. Healthy tissues areshielded by a collimator, while the adaptation to the distal contour of the tumor is performedwith a compensator. The net result is a non negligible dose to the normal tissues in theproximal part of the tumor (double hatched area). From Schardt et al. [23]
viding the tumor volume in more sub-volumes which are irradiated consecutively),
another limitation arise from the presence of several centimeters of material directly
on the beam path: the dose from secondary particles (especially neutrons).
In the second approach, instead, the volume is divided in several iso-energetic
slices and each slice is sub-divided in a grid of elementary volumes (voxels). Each
44 CHAPTER 2. RADIOTHERAPY AND HADRONTHERAPY
voxel is then sequentially irradiated by the scanning beam by means of two pairs of
deflecting magnets. The scan path follows a zigzag line connecting all the voxels
in the grid. When an entire slice has been irradiated, the beam extraction is inter-
rupted, extraction energy is then changed and the irradiation of the next slice can
begin. A sketch of this second technique is shown in Figure 2.15. The active scan-
Figure 2.15: Left: GSI active scanning system working principle. The target volume isirradiated by moving a pencil beam with fast scanning magnets, beam parameters are sup-plied synchronously to each pulse by control system. Right: the entire tumor is divided inseveral iso-energetic slices, (the slice being irradiated is magnified). During the irradiationeach voxel (white dot) receives the planned dose, the green dots represent pencil beam arrivalpoint. From Schardt et al. [23]
ning has several advantages: no patient specific hardware is needed for treatment
(except for immobilization); any irregular volume can, theoretically, be homoge-
neously irradiated; dose can be varied for each voxel (this allows to compensate for
pre-irradiation of proximal regions); the material in the beam line can be minimized,
reducing beam attenuation and fragmentation. On the other hand, more demanding
control and safety systems are required together with remarkable accelerator perfor-
mances on stability and reproducibility of beam position. However, active scanning
allows a much more flexible capability to tailor the dose distribution than passive de-
livery systems. For this reason the term Intensity Modulated Particle Therapy (IMPT)
has been introduced, in analogy to the IMRT techniques in photon therapy, to address
such delivery system.
For a long time the only two facilities that pioneered this latter approach were
PSI (Switzerland) and GSI (Germany) for protons and carbon ions irradiation respec-
tively. Their research on the active scanning technique, albeit with some variations,
proceeded in parallel and the acquired experience represents the basis for comparison
for all the other treatment facilities worldwide and also for industrial solutions.
2.4. BEAM DELIVERY TECHNIQUES 45
Recently, the National Centre of Oncology (CNAO) was created in Pavia (through
the collaboration of the Istituto Nazionale di Fisica Nucleare (INFN), CERN (Switzer-
land), GSI (Germany), LPSC (France) and of the University of Pavia (Italy)) with
the goal of treating tumors by using both protons and carbon ions. It operates a
synchrotron capable of accelerating protons up to 250 MeV and carbon ions up to
480 MeV/u with an active scanning technology. At the end of 2011 the first patient
course of treatment with protons was completed and a year later the first carbon ion
treatment started. Nowadays the center normally operates with both particles on a
daily basis.
Gantries
In conventional radiotherapy, as well as for other imaging techniques like MRI,
CT or PET, the patient is treated in supine position, in order to minimize unwanted
organ movements. The electron linac is mounted, in fact, on a rotational support
(gantry) that, in combination with the routable patient couch, allows to choose the
most favorable angles for the treatment. Every commercial radiotherapy system in-
cludes a 360 rotating gantry and there is not any limitation on the angles that can be
used for therapy.
During particle therapy’s early stages, i.e. when treatments were performed in-
side research laboratories with large accelerators designed for nuclear physics re-
search, the beam was typically transported horizontally. The scenery changed when
the first proton therapy facilities were planned. In order to demonstrate the superi-
ority of particle therapy, a full exploitation of the more favorable depth-dose profile
was required. The main technical issue is the high magnetic rigidity of the beam that
implies a bending radius of the order of 1 m. For heavy ions the situation is much
worse, because an even higher bending power is required (this is the reason why
carbon ion facilities need larger accelerators). The magnetic rigidity of 380 MeV/u
carbon ions with a range of 25 cm in water is, in fact, about three times the one of
protons with the same range. Furthermore, a high precision on the rotating move-
ment is required. For all the aforementioned reasons, a gantry system for ion therapy
represents an expensive and a very challenging work of engineering. The first ro-
tating isocentric gantry system for heavy ions was built at HIT center (Heidelberg,
sketched in Figure 2.16) and is in operation for both protons and carbon ions since
late 2012.
46 CHAPTER 2. RADIOTHERAPY AND HADRONTHERAPY
(a) (b)
Figure 2.16: (a) HIT gantry treatment room 3D drawing. (b) Gantry view from the accel-erator room. With its 13 m diameter, 25 m length and 670 tons of weight (compared to theusual 100÷120 tons of proton gantries) this is the largest gantry ever built. Figure (a) FromHIT website [31], Figure (b) courtesy of University Hospital Heidelberg.
Treatment of moving targets
So far organ irradiation with scanned beams has been carried out only in areas
that could be immobilized by external aids. In these cases the target can be assumed
to be still and the uncertainties due to patient motion (e.g. by breathing) are negli-
gible. The patient needs, however, to be immobilized with masks, belts or special
frames, in order to take advantage of the highly conformal dose deposition. This
procedure can be problematic if the target organ is in the abdomen or in the thorax,
where breathing motion or pressure related problems (e.g. bladder) are unavoidable.
Moreover motion patterns are in general complex even though in the upper abdom-
inal region are mostly translational. This can lead to variation in the radiological
path length of the target voxels, that for hadrontherapy result in a high impact on
the quality of the treatment, since Bragg peak is shifted accordingly to such varia-
tion, as shown in Figure 2.17 and Figure 2.18. Conventional radiotherapy does not
suffer from this complication, being the depth-dose variations negligible. Irradiation
of moving targets is a very active field of research and several options to take into
account patient motion have been suggested so far:
1. Planned target volume expansion. In this way the moving target results com-
pletely covered at any time. This has the clear disadvantage that the dose on
2.4. BEAM DELIVERY TECHNIQUES 47
(a) (b)
(c) (d)
Figure 2.17: Calculated dose deposition for a lung tumor without (a) and with motion (b).As it is clearly visible, the presence of motion leads to severe overdosage or underdosage inthe target volume. Carbon ions range modification during the two breathing phases: inhale(c) and exhale (d). Iso-range curves are shown in blue (2 cm), green (4 cm), yellow (6 cm),orange (8 cm) and red (10 cm). Figures courtesy of C. Bert, private communication.
normal tissues limits the overall dose that can be given to the target volume.
2. Rescanning. This strategy is based on a statistical assumption: if the scanning
is repeated N times, the variance of the average dose decreases with a factor
of 1/√N , if target motion and beam motion are considered uncorrelated. This
technique has the disadvantage of prolonging the irradiation time and, since the
dose per scan has to be lowered, it can cause problems to the beam monitoring
ionization chambers, that are not sensitive to low currents.
3. Gating. In contrast to the previous options this one requires the monitoring
of breathing cycle. Observing the time evolution of the target motion a flat
minimum region can be found at the end of the exhale phase. If the irradia-
tion is restricted to this time frame, uncertainties due to target motion can be
reduced to less than 10% of the free breathing case. The only drawback is the
48 CHAPTER 2. RADIOTHERAPY AND HADRONTHERAPY
Figure 2.18: As a consequence of breathing induced motion, the same beam could traveltrough different beam paths causing a shift of Bragg peak position. While this effect is almostnegligible for photons, it has a higher impact on carbon ions where a large amount of energycould be deposited outside the tumor volume (red vertical lines).
prolonged treatment time needed to keep constant the delivered dose.
4. Tracking. This strategy requires a synchronous three dimensional online mo-
tion compensation. The beam, in fact, must follow target movements at any
time and ideally this approach should lead to the same result of the static case.
Motion tracking technique is still under evaluation, but detailed simulations
have already demonstrated the potential of 3D motion compensation. How-
ever, some critical technology issues arise such as: the availability of a dy-
namic treatment planning and a beam delivery system permitting lateral tracing
and fast range adaptation (in order to properly shift the Bragg peak depth).
Recently, the combination of two amongst the aforementioned movement mitigation
strategies has been proposed, for instance gating and rescanning are planned to be
used together at the gantry2 at PSI Gottschalk et al. [32].
2.5 Dose monitoring
Hadrontherapy’s higher precision in tumor irradiation urgently demands the de-
velopment of brand new dose release monitoring techniques. This potential clinical
benefit requires, in fact, the conformal dose delivery to be monitored in-situ and non
invasively. A reliable treatment feedback is, in fact, extremely needed since heavy
ions dose release is much more sensitive to morphological variations or patient mis-
placements Karger et al. [33]. Monitoring in-vivo (or online) the energy deposition
2.5. DOSE MONITORING 49
during tumor treatment will surely boost therapy effectiveness, allowing for a fast
quality control within and around target volume. Then, a system capable of mea-
suring the delivered dose and able to verify the conformity of the actually irradiated
volume with the treatment planning is highly needed Pedroni et al. [34].
Unfortunately, X-rays standard methodologies for patient positioning cannot be
used, since these require that a non-negligible fraction of the treatment beam would
be transmitted through the patient. A simple reconstruction of patient position is in
fact performed on body anatomy or fiducial markers. Such techniques are clearly not
applicable in hadrontherapy, where the deposited energy sharply decreases behind
the target and almost no exit dose (for protons) or only a small fraction of it (for
carbon ions) is available after the Bragg peak. In the following, a brief state of the art
of the existing techniques, as well as the one still under evaluation from the scientific
community, for dose monitoring in hadrontherapy will be presented.
PET
Historically, the first method that has been proposed is the Positron Emission To-
mography (PET), which exploits the back to back photon production of β+ emitters
(mostly from 11C and 15O radioactive decay) generated as a consequence of the ir-
radiation after nuclear fragmentation of the target and the projectile Paans et al. [35]
and Parodi et al. [36]. Considering the geometrical constraints of a treatment room,
it is not possible to install a standard PET scanner to monitor the patient during the
irradiation. Only limited geometries are, in fact, allowed (so-called double head),
but these are limited by a low angular acceptance and by the presence of artifacts
that limit the quality of the reconstructed image Pawelke et al.[37] and Enghardt et
al. [38]. The availability of Time of Flight (ToF hereafter) techniques, that exploit
also the time information carried by the two collinear photons, has reduced the back-
ground noise and the artifacts of standard PET, leaving untouched the acceptance
issues though. Another limitation of this approach is its intrinsic off-line nature,
given the rather long 11C half-life (≈ 20 minutes). This does not allow to collect
sufficient data within treatment duration, causing a spatial resolution worsening. The
aforementioned limitations can be avoided performing an off-line PET, i.e. placing
the patient inside a conventional PET scanner just after the treatment. On the other
hand, with this latter solution, metabolic processes (e.g. blood circulation) interact
with the radioactive nuclides. The overall image quality and the spatial relation be-
tween dose deposited and activity will then appear deteriorated. Fortunately, this
latter effect is well taken into account by Monte Carlo simulations, it is then possible
50 CHAPTER 2. RADIOTHERAPY AND HADRONTHERAPY
to reconstruct the dose deposition pattern from β+ activity even though this do not
show a clear correlation with the expected dose profile, as shown in Figure 2.19
Figure 2.19: Top: Treatment plan (left) and Monte Carlo calculation (right) of the dose re-ceived by a patient with pituitary adenoma treated with two orthogonal fields (lateral followedby posterior-anterior) at 0.9 GyE/field. Bottom: Measured (left) and Monte Carlo calculatedβ+ activity. Range of color wash display is from blue (minimum) to red (maximum). FromParodi et al. [39]
However, this opportunity has been tested at Massachusetts General Hospital
and at Heidelberg Ion-Beam Therapy Center (HIT) for several clinical cases Parodi
et al. [40] [39].
Prompt photons
All the issues of the PET approach have called the scientific community upon
the development of a novel monitoring technique that uses the proved correlation be-
tween Bragg peak position and prompt photons emission region Min et al. [41] [42]
Testa et al. [43] and Agodi et al. [44]. Since prompt radiation occurs within few ns as
a result of target and projectile nuclear de-excitation, this method is metabolism in-
2.5. DOSE MONITORING 51
dependent. In order to detect prompt radiation, a Single Photon Emission Computed
Tomography (SPECT) approach can be followed. Moreover, the geometrical con-
straints are less stringent with respect to PET, since the detectors are not requested
to be placed one in front of the other. Unfortunately a standard SPECT solution
cannot be used in practice, given prompt photons rather wide energy spectrum (that
it has been measured by Agodi et al. [44] and is comprised between 1÷10 MeV),
that would require too thick collimators leading to an insufficient statistics. An in-
novative approach, with respect to SPECT, would be to realize a Compton camera
(as proposed by Kabuki et al. [45]) in order to track the prompt radiation. In its
simplest version only two position detectors with a good energy resolution can be
employed (a scatterer and an absorber, that must detect and contain Compton elec-
tron and scattered gamma respectively). The information about the photon direction
is then obtained via software.
More refined solutions can be found in astrophysics applications that could be
tailored to this specific context: e.g. detectors able to reconstruct both electron and
gamma ray trajectories after Compton scattering inside the detector itself (Kormoll
et al. [46]). These latter systems are currently under evaluation by the scientific com-
munity, but they may suffer from low statistics issues, since the amount of secondary
photons produced during a standard treatment is limited. In order to enhance the
available statistics, the angular acceptance could be increased, for instance widening
the dimensions of the detector or reducing its distance from the patient. Unfortu-
nately, this solution is not always compatible with the available space in treatment
rooms. Moreover, the neutron background, always present in treatment rooms, could
represent an unwanted and unavoidable source of additional noise for such a detector.
The use of prompt photons as instruments to evaluate the dose deposition during a
hadrontherapy treatment is still in its prototyping phase, this means that it has never
been used in a clinical environment so far.
Charged particles
A recent proposal, based on novel measurements of secondary charged parti-
cles (mostly composed by hydrogen and its isotopes) produced after the interaction
between target and projectile Braunn et al. [47], Agodi et al. [48], suggests to ex-
ploit such radiation for Bragg peak monitoring. The kinetic energy of such particles
is comprised between 10÷150 MeV, while their production region spans the whole
beam path inside the patient. The energy spectrum varies with the emission angle,
being high energetic protons production more favored by forward emission (with re-
52 CHAPTER 2. RADIOTHERAPY AND HADRONTHERAPY
spect to the beam direction). Since secondary protons production follows the beam
path, its location can be correlated with Bragg peak position. Additional measure-
ments of secondary charged particles produced as a consequence of phantom irradia-
tion have been recently performed at GSI laboratories Piersanti et al. [49] and at HIT
Gwosch et al. [50]. In both cases the results seem to encourage the exploitation of
secondary protons to monitor the dose release. Additional data from phantom irradi-
ation with a therapeutical proton beam have been collected at CNAO clinical facility
and are about to be published.
What has been observed so far is a clear correlation between charged particle
emission point and Bragg peak position (as it will be shown in the next chapter).
Then, it has been proposed that the shape of the emission profile could be used to
identify the Bragg peak position during each irradiation. Moreover, this profile has a
very steep rise (in correspondence to the patient entry channel) and hence could be
also potentially used to verify patient correct positioning online.
A future detector, capable of exploiting secondary charged radiation, has to be
made of a tracker and a calorimeter in order to point back each trajectory and ap-
propriately weight it according to its energy. Given the actual know-how, charged
particles tracking and energy measurement (in the MeV range) does not represent a
technology issue. However, especially for proton beams where the target is the only
actor in the fragmentation process, some statistics issues could arise for high angle
measurement (90 or 60 with respect to the beam direction). This could be a possi-
ble limitation of this novel approach that, in order to represent a valid alternative to
PET, will need nevertheless extensive testing and solid results.
Summarizing, at the state of the art there is not any monitoring system for hadron-
therapy currently used in medical routine, while systems that exploits PET or SPECT
technologies are now under evaluation by the international community and have been
already tested for some clinical cases. These latter systems, however, don’t seem to
guarantee a concrete feasibility for online monitoring given prompt photons wide
energy spectrum. The opportunity to use Compton cameras or secondary charged
particles, instead, is a very recent discovery that could pave the way for new interest-
ing possibilities.
CHAPTER
3
Measurement of secondaryradiation
Content
3.1 Experimental setup . . . . . . . . . . . . . . . . . . . . . . . . . . 54
Start Counter . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 55
Drift Chamber . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 56
Calorimeter . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 57
Data acquisition . . . . . . . . . . . . . . . . . . . . . . . . . . . . 58
Angle of detection . . . . . . . . . . . . . . . . . . . . . . . . . . . 59
3.2 Prompt photons . . . . . . . . . . . . . . . . . . . . . . . . . . . . 60
Calibration . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 60
Energy measurement . . . . . . . . . . . . . . . . . . . . . . . . . 61
Rate measurement . . . . . . . . . . . . . . . . . . . . . . . . . . . 63
Preliminary data with GSI 220 MeV 12C beam . . . . . . . . . . . 64
3.3 Charged secondary particles . . . . . . . . . . . . . . . . . . . . . 66
Particle identification and fluxes measurement . . . . . . . . . . . . 66
Bragg peak position monitoring . . . . . . . . . . . . . . . . . . . 69
Charged particles production region . . . . . . . . . . . . . . . . . 71
In order to meet hadrontherapy’s improved capability in matching the dose re-
lease with cancer position, new dose monitoring techniques need to be developed
and introduced into clinical use. A careful study of charged and neutral particles,
produced as a consequence of nuclear fragmentation and de-excitation processes,
53
54 CHAPTER 3. MEASUREMENT OF SECONDARY RADIATION
plays a crucial role in the design of any dose monitoring device and it would be
extremely useful to tune Monte Carlo simulations.
During patient irradiation, in fact, several interaction mechanisms take place and
both neutral and charged secondary radiation is produced. The former is composed
mostly of prompt photons (emitted as a consequence of nuclear de-excitation of
the target) and back-to-back photons (produced after positron emission, and subse-
quent annihilation, by β+ emitters, like 15O and 11C, created during the irradiation).
Charged particles, on the other hand, are produced as a consequence of nuclear frag-
mentation of the projectile (mostly in forward direction) and of the target (almost
isotropic but with lower energies).
The uncertainty on energy release position can be due to several factors: cal-
ibration of the CT images, possible patient morphological changes or patient mis-
positioning on the treatment couch. Hence, the design of a novel imaging detector,
that can exploit the secondary radiation produced during patient irradiation, is eagerly
needed.
However, the key features of such secondary radiation (energy spectrum, fluxes,
cross sections and so on) for this kind of interactions in this energy range are still
rather unexplored and must be better understood.
Two main products of the target-projectile interaction will be described in the fol-
lowing: prompt photons and secondary charged particles production from 12C beam
impinging on a poly-methyl-metacrylate (PMMA) target. Three different on beam
measurements will be briefly reviewed in this chapter: prompt photons (Agodi et
al. [44]) and charged particles production (Agodi et al. [48]) from 80 MeV/u fully
stripped 12C on PMMA performed at LNS-INFN Laboratories (Catania, Italy) and
charged particles production from 220 MeV/u fully stripped 12C on PMMA (Pier-
santi et al. [49]) performed at GSI Laboratories (Darmstadt, Germany).
3.1 Experimental setup
All the aforementioned measurements shared the same experimental setup, with
some minor changes due to the different beam energies and data taking conditions
though. A fast plastic start scintillator (Start Counter), a PMMA target, a charged
particle tracker (Drift Chamber) and a calorimeter (LYSO), used to detect both neu-
tral and charged radiation, represent the common detectors used. For GSI data taking
an additional Start Counter and a Veto plastic scintillator (2 mm thick), placed in
front of the Drift Chamber in order to stop low energy electrons (E < 0.6 MeV),
3.1. EXPERIMENTAL SETUP 55
have been employed. A sketch of the experimental setup of both measurements is
schematically shown in Figure 3.1.
Figure 3.1: Schematic view of the experimental setup for the two experiments; in both casesthe DAQ is triggered with the coincidence of the Start Counter and LYSO. The Veto plasticscintillator has been included to stop low energy electrons before entering the Drift Chamber.
A PMMA cube (4×4×4 cm3) has been chosen as target for LNS measurement
(since the range of 80 MeV/u 12C is of the order of 2 cm), while the ≈ 10 cm range
of GSI 220 MeV/u ions demanded a longer target (5×5×20 cm3).
Start Counter
The Start Counter is made of a 1.1 mm thick plastic scintillator (BC-404) read out
by two photomultiplier tubes (Hamamatsu H10580) and it is placed directly on the
beam line. This detector has been designed for timing and triggering purposes and
holds a role of fundamental importance. It aims, in fact, at secondary particles Time
of Flight (ToF hereafter) measurement as well as incident carbon ions measurement.
In order to reduce the background noise and to increase the time resolution, the time
coincidence of the two PMTs has been taken as trigger reference signal and in the
following the term Start Counter will refer to this quantity.
56 CHAPTER 3. MEASUREMENT OF SECONDARY RADIATION
Drift Chamber
A Drift Chamber, made of twelve planes (each one composed of three cells) dis-
tributed on two views, has been used as charged particle tracker and is shown in
Figure 3.2. Some dedicated studies on Drift Chamber efficiency and spatial resolu-
tion have been performed in order to find its best working point. Efficiency plot as
a function of high voltage and spatial resolution as a function of the distance from
sense wires have been carried out and are shown in Figure 3.3 and Figure 3.4
(a)
(b)
Figure 3.2: (a) Drift Chamber mechanical drawing. (b) Drift Chamber lateral layout, all thesense wires of the lateral view are shown (red dots).
It has been operated with an Ar/CO2 80%-20% gas mixture and its performances
(single cell spatial resolution σDCH ≤ 200 µm, single cell efficiency εDCH = (93± 3)%
have been reviewed by Abou Haidar et al. [51], since this detector has been used as
beam monitor in the FIRST experiment at GSI (Pleskac et al [52]). Apart from
3.1. EXPERIMENTAL SETUP 57
Figure 3.3: Drift Chamber efficiency study performed at LNS with different gas mixturesand ionizing particles. 80 MeV/u carbon ions with P10 (blue) and ArCO2 80%-20% (black).80 MeV protons with ArCO2 80%-20% (red).
Figure 3.4: Drift Chamber single cell spatial resolution as a function of the track distancefrom sense wire. Data from LNS 80 MeV protons dataset.
charged particles tracking, the Drift Chamber plays also the role of charged particles
veto, neutral events, in fact, are selected when no signal comes out of it.
Calorimeter
Finally, the calorimeter is made of a 2×2 matrix of 1.5×1.5×12 cm3 cerium-
doped Lutetium Yttrium ortho-Silicate (LYSO) crystals, whose scintillation light is
read out by one EMI 9814B photomultiplier. The reason of this choice lies in LYSO’s
very fast response (that is of crucial importance when measuring ToF), high density
and high light output. This particular type of crystals find a wide application in
medical imaging although in smaller sizes (e.g. PET imaging) and an exhaustive set
58 CHAPTER 3. MEASUREMENT OF SECONDARY RADIATION
of LYSO features is reported in Table 3.1.
LYSO characteristics
Effective atomic number 66Density (g/cm3) 7.4Decay constant (ns) 40÷44Peak emission (nm) 428Light yield (% NaI) 75Refractive index 1.82
Table 3.1: LYSO optical characteristics.
Data acquisition
The Data acquisition system (DAQ hereafter) was triggered by Start Counter
and LYSO coincidence and both charge and arrival time of the secondary particles
have been recorded (by an ADC and a TDC module respectively). All the required
analogue signals need to be delayed in order to properly take into account signal
transit time in the various logic modules. For the Drift Chamber analysis, instead,
only the arrival time of the ionization electrons on the sense wires has been used. A
sketch of the DAQ signal flow chart that has been developed for the two experiments1
is shown in Figure 3.5 and a series of pictures of the experimental setup at GSI is
shown in Figure 3.6
Figure 3.5: Data acquisition flow chart. The trigger is defined as the time coincidencebetween Start Counter and LYSO detectors. Once the trigger signal has been created it goesto the dead time logic (DT) and the data acquisition can begin.
1A similar logic scheme has been developed together with a custom VME DAQ system for theradiation physics group at GSI, as reported by Piersanti et al. [53]
3.1. EXPERIMENTAL SETUP 59
Figure 3.6: Top: Experimental setup mounting phase at GSI Laboratories. Bottom: DAQelectronics and drift chamber detail.
Angle of detection
As it can be noted from Figure 3.1, a detection angle of θ = 90 has been adopted
(this angle has been varied in Piersanti et al. [49] to measure charged particles pro-
duction also at θ = 60 and θ = 120). This choice has been made for two distinct
reasons: avoid DAQ dead time saturation and reconstruction issues due to beam-spot
size. In fact, if the emission spectrum of prompt photons is isotropic, charged parti-
cles’ is not. Secondary fragments are produced preferably with forward angles with
respect to the beam direction (θ < 90 in Figure 3.1). Hence, choosing a smaller
measurement angle will result in a higher number of charged fragments within de-
tector’s acceptance. This will lead to a higher DAQ rate and hence more likely to
dead time saturation. The second problem arise for every angle θ 6= 90. When any
charged particle track is pointed backwards to the PMMA, the emission shape spatial
resolution worsens as (sin θ)−1, due to the projection along the beam line, and this
effect could become dominant for small detection angles. Furthermore, if θ 6= 90,
the emission shape is convoluted with the size of the primary beam spot projected
on the beam line (see Figure 3.7), adding a term ∝ σbeam · cotg(θ) to the emission
profile. Thus, as the two aforementioned factors increases for smaller angles, the
tracking accuracy, from a geometrical point of view, improves with larger detection
angles.
60 CHAPTER 3. MEASUREMENT OF SECONDARY RADIATION
Figure 3.7: Sketch of the beam spot size (grey cylinder) and its contribution to the recon-struction of the fragments emission region. The experimental setup is supposed to be at anangle θ with respect to the primary beam direction.
3.2 Prompt photons
In this section the main results obtained from the interaction of a 80 MeV/u12C beam on PMMA will be reviewed. Moreover, some preliminary data (yet un-
published) regarding prompt photon production from 220 MeV/u 12C GSI beam on
PMMA will be presented as well.
Calibration
Together with the standard 22Na (0.511 MeV) and 60Co (1.17 and 1.33 MeV)
radioactive sources, the LYSO energy calibration has been carried out exploiting a
source which produces photons of energy well above 1 MeV. Unfortunately it was
not possible to use one amongst the standard isotopes, given their usual low lifetime,
and an indirect production mechanism has been exploited instead. Thus, an AmBe
neutron source (2.5×106 neutrons/s) hosted inside a 5 cm thick paraffin (Cn H2n+2)
container has been used, as reported in Bellini et al. [54]. The container allowed both
to moderate the neutron flux (that otherwise would have saturated the detector) and
to produce two gamma lines: the first at 2.22 MeV from deuteron formation and the
second at 4.44 MeV from 12C∗ de-excitation. The measured spectrum is shown in
Figure 3.8. The two photon lines are clearly visible, together with the so-called single
escape lines (that is a phenomenon that occurs when one of the two collinear photons
produced by the annihilation of a positron escapes the detector without interacting).
Interposing a 2 mm thick nichel rod between the source and the detector it is possible
to generate a set of high energy lines centered around a mean value of 8.8 MeV. The
calibration curve is hence derived and is shown in Figure 3.9.
3.2. PROMPT PHOTONS 61
Figure 3.8: LYSO energy spectrum with AmBe source moderated with paraffin.
Figure 3.9: Linear calibration curve (black dashed) for LYSO detector obtained combining22Na, 60Co with AmBe data (red circles).
Energy measurement
In order to select a prompt photon event, the time difference (∆T ) between the
energy deposition in the LYSO (TLYSO) and carbon ion arrival time on the Start
Counter (TSC) is considered. The correlation between reconstructed photon energy
E and measured ∆T is shown in Figure 3.10a. Here four distinct regions can be
highlighted: prompt photons main population (green); a faster component (red) due
to prompt production inside the Start Counter; LYSO flat background noise (blue -
E < 2 MeV) and a diffuse cloud mostly due to neutrons whose arrival time is not
correlated with prompt radiation (magenta). From Figure 3.1 is evident how prompt
photons produced inside the Start Counter traverse a shorter path and hence are faster
than photons coming from the target. The shape of prompt photons population is
62 CHAPTER 3. MEASUREMENT OF SECONDARY RADIATION
not vertical, as one would expect. This is due to a time slewing effect induced by
front-end electronics fixed voltage threshold. This artifact can be adjusted fitting the
distribution of ∆T in bins of energy and extracting the correction function C(E)
shown in Figure 3.10b. The energy spectrum as a function of the corrected time
∆Tcorr = ∆T − C(E) is reported in Figure 3.10c together with the time resolution
(σ∆Tcorr) of the detector as a function of the energy, Figure 3.10d. For E > 3 MeV
σ∆Tcorr ≈ 300 ps has been obtained.
(a) (b)
(c) (d)
Figure 3.10: (a) Calibrated LYSO energy as a function of ∆T , four major components areunderlined: (i) prompt photons produced inside the target (green - slower population, theirpath towards the detector is longer), (ii) prompt photons produced inside the start counter(red - faster component), (iii) a flat background with E < 2 MeV due to LYSO intrinsicnoise (blue), (iv) a diffused cloud mainly due to neutrons (magenta). (b) Estimated timeslewing correction. (c) Energy versus ∆T corrected spectrum. (d) Time resolution σ∆Tcorr asa function of the measured energy, for E >3 MeV a resolution of the order of 300 ps hasbeen achieved.
In order to estimate prompt radiation energy spectrum, the number of photons for
each energy bin has been calculated fitting the ∆Tcorr distribution with a superimpo-
sition of a Gaussian function (signal), centered in zero with a sigma fixed at σ∆Tcorr ,
3.2. PROMPT PHOTONS 63
and a polynomial function (background) and evaluating the area under the Gaussian.
This value has been corrected for the dead time fraction (εDT) of the DAQ and nor-
malized to the number of incident ions (for further details see Agodi et al. [44]). The
measured energy spectrum is shown in Figure 3.11 and it has been compared to the
one predicted by Monte Carlo code FLUKA (Ferrari et al. [55] and Battistoni et
al. [56]), obtained within the detector acceptance and folded with detectors response
(LYSO and Start Counter efficiency, resolution and acceptance). Data-Monte Carlo
agreement on the shape and on the normalization is not perfect, this is due to the
lack of experimental cross sections for these interactions in this energy range. For
this reason, yield and energy measurements, for both neutral and charged secondary
radiation, are eagerly needed by Monte Carlo community to improve the analytical
models currently employed to describe such reactions.
0
0.05
0.1
0.15
0.2
0.25
0.3
2 3 4 5 6 7 8 9 10
Cou
nts/
prim
ary
(x10
-6)
Energy [MeV]
Exp. DataSimulation
Figure 3.11: Data (black dots) - Monte Carlo (FLUKA, red solid line) comparison of promptphotons energy spectrum from a 80 MeV/u 12C beam impinging on PMMA target. Bothspectra are normalized to the number of incident ions.
Rate measurement
In order to design a detector that exploits prompt radiation to monitor the dose
deposition, the measurement of photons differential rate is of crucial importance to
assess the available statistics in a typical treatment. As first step, the fraction of
observed photons with E > 2 MeV (this energy threshold is needed to reject LYSO
intrinsic noise) has been evaluated as the ratio of measured prompt rate and carbon
ions rate. This fraction has been calculated and it averages to:
Fprompt =Rprompt
RC= (3.04± 0.01stat ± 0.20sys)× 10−6 (3.1)
64 CHAPTER 3. MEASUREMENT OF SECONDARY RADIATION
as shown in Figure 3.12. An uncertainty contribution from systematics is required to
justify the data dispersion that is well above statistical fluctuations.
Figure 3.12: Fraction of prompt photons as a function of carbon ions beam rate. The redline is the linear fit to the data and the red band accounts for both systematic and statisticalerrors.
The double differential rate for prompt photons has been evaluated to be:
d2Nγ
dNC dΩ(θ) =
1
NmeasC /εSC
[Nmeasγ
εDT εSC εLYSO ΩLYSO
]90
(3.2)
where NmeasC is the number of carbon ions measured with the Start Counter, Nmeas
γ
is the number of prompt photons detected by LYSO, εDT is the dead time fraction of
the DAQ, εSC is the Start Counter detection efficiency εSC = (96 ± 1)% (as reviewed
in Agodi et al. [44]), ΩLYSO and εLYSO are the angular acceptance and scintillation
efficiency of LYSO detector. These last two quantities have been evaluated by using
FLUKA for all the isotopes and detection angles to be: ΩLYSO ' 1.6 × 10−4 sr and
εLYSO = (81.3 ± 2.5)%. For the aforementioned setup a flux of:
dN2γ
dNCdΩ(θ = 90) = (2.323± 0.007stat ± 0.151sys)× 10−3 sr−1 (3.3)
has been obtained.
Preliminary data with GSI 220 MeV 12C beam
Additionally, some preliminary data (yet unpublished) from 220 MeV 12C beam
impinging on a 5 × 5 × 20 cm3 PMMA target collected at GSI will be briefly re-
viewed in the following. The experimental setup is conceptually equivalent to the
3.2. PROMPT PHOTONS 65
one at LNS, as it can be noted comparing the upper sketch with the bottom one in
Figure 3.1. However, the main difference at GSI is due to the “therapeutic-like”
energy of the primary particles, if compared to the 80 MeV/u of LNS data taking.
Thus, the experiment outcome will be characterized by a more realistic (closer to the
clinical reality) energy spectrum and flux of prompt photons. The same data analysis
technique, that has been described in the previous paragraphs, has been used also
with this dataset, and prompt photons energy spectrum for different detection angles
is shown in Figure 3.13.
Figure 3.13: Energy spectra of prompt photons detected by LYSO scintillator at θ = 60,θ = 90 and θ = 120. These spectra have been normalized to the number of incident car-bon ions and have been corrected for dead time efficiency, detector efficiency and detectorgeometrical acceptance.
The fluxes for all the three detection angles have been estimated as well:
dN2γ
dNCdΩ(θ = 60) = (5.25± 0.04stat ± 0.25sys)× 10−3 sr−1 (3.4)
dN2γ
dNCdΩ(θ = 90) = (5.68± 0.04stat ± 0.13sys)× 10−3 sr−1 (3.5)
dN2γ
dNCdΩ(θ = 120) = (4.04± 0.05stat ± 0.47sys)× 10−3 sr−1 (3.6)
In the systematic error, the only contribution comes from the systematic error on the
dead time efficiency εDT. The rate at (θ = 90) can be compared to the one measured
66 CHAPTER 3. MEASUREMENT OF SECONDARY RADIATION
by Agodi et al. [44] reported in equation (3.3). Since the path of the carbon ion
in the target is related with the number of prompt photons emitted, a different rate
for different energy beams is thus expected (since the depth is proportional to the
incident beam energy). As confirmation of this assumption, the beam with higher
energy produces a higher prompt photon flux than the lower one.
3.3 Charged secondary particles
Charged fragments can represent an innovative tool to monitor Bragg peak po-
sition during a hadrontherapy treatment. Their eventual use has been proposed only
recently and extensive data taking campaigns must be foreseen in order to assess
their real potential. In the following the main results of 80 MeV/u 12C on PMMA
(referred to as LNS) and 220 MeV/u 12C on PMMA (referred to as GSI) experiments
will be reviewed.
LNS measurement was aimed at protons flux evaluation and Bragg peak position
monitoring. Even though the beam energy is not in hadrontherapy typical range, the
capability to follow the Bragg peak shift, when changing the penetration depth of the
beam, and the estimation of protons flux are fundamental questions that need to be
addressed before this new technique could be even proposed. Then, with GSI mea-
surement some more refined aspects have been evaluated: the possibility to recon-
struct the emission shape of charged particles and a quantitative method to link some
geometrical parameters of this shape with the actual Bragg peak position. Moreover,
during this latter experiment different detection angles have been studied (θ = 60,
90 and 120).
Particle identification and fluxes measurement
As discussed in section 3.1 the Drift Chamber is used also as veto detector, to dis-
criminate between neutral and charged events. In fact, considering the experimental
setup geometry, an ion traveling from the target towards LYSO calorimeter will most
likely hit all the twelve tracking planes of the Drift Chamber. Hence, an event selec-
tion based on the number of wires fired by the ion can be performed. The distribution
of hit cells (Nhits) measured at GSI is shown in Figure 3.14, where data is compared
to Monte Carlo predictions for each hydrogen isotope. A clear peak at Nhits = 12
is observable, this indicates that charged particles tend to cross all the planes hitting
only one cell per plane. Then, each event with Nhits > 8 has been flagged as charged
particle. Isotopes discrimination can be performed exploiting ToF information car-
3.3. CHARGED SECONDARY PARTICLES 67
ried by charged particles together with the energy deposited inside LYSO crystal.
The ToF, as it has been recorded by front end electronics, is the sum of two con-
Figure 3.14: Comparison data (circles) and Monte Carlo (solid line) of the number of hitcells (Nhits) in the Drift Chamber when an event is recorded in the LYSO with E > 1 MeVat 90. In the simulation each isotope contribution has been underlined: protons (upwardtriangles), deuterons (squares) and tritons (downward triangles). The clear underestimationin the simulated data for Nhits < 7 could be due to several contributions: Monte Carlo lowerbackground (with respect to GSI experimental hall) and absence of electronic cross talk.
tributions: the time needed by a carbon ion to travel from the Start Counter to its
fragmentation point (x) in the target (Tch(x) − T12C) and the time needed to travel
from the fragmentation point to LYSO detector (TLY − Tch(x)). An estimate of the
time needed by the carbons to reach their interaction point has been evaluated with
FLUKA. Using this information the ToF associated to each fragment can be finally
calculated. An example of particle identification performed on GSI data (θ = 90) is
reported in Figure 3.15a, a Monte Carlo comparison is shown in Figure 3.15b, where
a very good agreement between data and simulation can be noticed. Three popu-
lations are clearly visible in the plots, which correspond to protons, deuterons and
tritons. For θ = 120, see Figure 3.15d, there is a significant drop in statistics, for
this reason all the measurements and the results that will be reviewed in the following
will only refer to the more abundant data samples at 60 and 90.
The double differential production rate of each secondary particle isotope nor-
malized to the number of incoming carbon ions (NC) has been estimated exploiting
the same definition reported in equation (3.2) with an additional term due to Drift
68 CHAPTER 3. MEASUREMENT OF SECONDARY RADIATION
(a) (b)
(c) (d)
Figure 3.15: Energy versus ToF distributions. (a) Data θ = 90. (b) Monte Carlo θ = 90,single isotope contributions have been separated: protons (black), deuterons (red), tritons(green), electrons (blue). (c) Data θ = 60. (d) Data θ = 120. The lines used to separateprotons, deuterons and tritons in (a) are superimposed.
Chamber tracking efficiency (εtrack):
d2Np,d,t
dNCdΩ(θ) =
1
NC/ εSC
[Np,d,t
εDT εSC εtrack εLYSO ΩLYSO
]60,90
(3.7)
The Monte Carlo simulation of the experiment produced: εtrack = (93 ± 3)% and
96%<εLYSO<97% for all the isotopes and the angle configurations. The main source
of uncertainty on particles flux measurement is due to particle identification and dead
time estimation. The former has been evaluated by moving the boundary lines shown
in Figure 3.15a, obtaining a maximum relative error up to 20%, depending on the data
sample.
The yield of all isotopes normalized to the number of incoming carbon ions and
to the solid angle covered by the LYSO detector were calculated for the data sets at
3.3. CHARGED SECONDARY PARTICLES 69
60 and 90:
dN2p
dNCdΩ(θ = 60) = (11.03± 0.09stat ± 0.80sys)× 10−2 sr−1
dN2d
dNCdΩ(θ = 60) = (4.72± 0.04stat ± 0.47sys)× 10−2 sr−1 (3.8)
dN2t
dNCdΩ(θ = 60) = (1.15± 0.02stat ± 0.27sys)× 10−2 sr−1
dN2p
dNCdΩ(θ = 90) = (2.30± 0.03stat ± 0.18sys)× 10−2 sr−1
dN2d
dNCdΩ(θ = 90) = (1.00± 0.02stat ± 0.11sys)× 10−2 sr−1 (3.9)
dN2t
dNCdΩ(θ = 90) = (0.17± 0.01stat ± 0.04sys)× 10−2 sr−1
Bragg peak position monitoring
In order to change the penetration depth of the beam, the target has been placed on
a single axis movement stage, that allows to perform a position scan (along x axis, i.e.
beam direction) with an accuracy of 0.2 mm. Hence, shifting the target (to simulate
a Bragg peak displacement with respect to the experimental setup) would allow to
evaluate wether a correlation between the positions of Bragg peak and fragments
production peak does exist. The configuration with the centers of PMMA, Drift
Chamber and LYSO aligned along z has been taken as reference and the position of
the target in the stage reference frame has been named as “0”. From this point the
target has been moved up to 19 mm backward and 13 mm forward with respect to
x axis positive direction, as schematically shown in Figure 3.16.
From the Monte Carlo simulation of the experiment, Bragg peak expected po-
sition has been estimated to be xBragg = (11.0 ± 0.5) mm from the front face of the
PMMA, this has been also visually confirmed by direct observation of the target dete-
rioration after data taking, visible as a yellow band shown in Agodi et al. [57]. In the
current frame of reference (the front face of PMMA is at x = 2 cm), that corresponds
to xBragg = (9.0± 0.5) mm. Then, for each run with different target position, the pro-
duction point of the protons has been monitored using the mean of the Gaussian fit
to the distributions of the fragment emission coordinates xPMMA and yPMMA. Since
yPMMA is the coordinate of proton production in the vertical plane, it should not be
affected by the position scan. Thus, its behavior has been used to estimate the tech-
nique’s systematic uncertainty. The relationship between xPMMA, yPMMA and xBragg
70 CHAPTER 3. MEASUREMENT OF SECONDARY RADIATION
Figure 3.16: Schematic view of Bragg peak position scanning performed during LNS ex-periment. The Bragg peak position relative to the “0” configuration has been determined withFLUKA to be: xBragg = (9.0 ± 0.5) mm. The picture is not to scale.
for proton energy Ekin > 60 MeV is shown in Figure 3.17. A linear behavior can
be clearly observed, meaning that charged secondary particle emission point follows
accurately Bragg peak movement. As expected, no dependency on yPMMA can be
noted, since the shift has been performed along x axis only.
Figure 3.17: Correlation between charged particles production coordinates mean values(xPMMA black dots, yPMMA white dots) and Bragg peak coordinate expected position (xBragg).Bragg peak position has been shifted moving the target along the beam direction (x coordi-nate in the current frame of reference). A clear linear dependency between xPMMA and xBraggcan be observed, while yPMMA shows no dependency at all.
3.3. CHARGED SECONDARY PARTICLES 71
Charged particles production region
As previously mentioned, once the tracks have been reconstructed in the Drift
Chamber, these are then extrapolated backwards to the PMMA position, in order to
identify charged fragment production point (xPMMA, yPMMA) along the beam path.
xPMMA distributions for the two experiments are shown in Figure 3.18 together with
the simulated dose deposition. A bi-dimensional reconstruction of the interaction
point has been carried out with GSI data, leading to the impressive shape reported
in Figure 3.19, where the front part of the target (up to Bragg peak position) can be
clearly identified from the extrapolation of Drift Chamber tracks.
(a) (b)
Figure 3.18: Simulated depth-dose distribution (hatched) superimposed on the longitudinalprofile (solid line) of secondary charged particles emission point as a function of the targetthickness (x) at 90. (a) LNS experiment: the beam enters the target in x = 2 cm and movestowards left. (b) GSI experiment: the beam travels from left to right and PMMA front face isin x = -(6.45 ± 0.02) cm.
Given the rather different beam energies a comparison between Figure 3.18a and
Figure 3.18b is not possible, however starting from GSI data (that have been obtained
on a therapeutical beam) a strategy to link the shape of the xPMMA distribution in
Figure 3.18b to the Bragg peak position has been developed. First of all, there is
a clear correlation between the beam entrance position and the sharp rising edge of
the emission profile. Moreover, since xPMMA distribution is well described by the
empirical function:
f(x) = p01
1 + exp(x−p1p2
) 1
1 + exp(−x−p3
p4
) + p5 (3.10)
it has been possible to link Bragg peak position to two specific parameters ∆40 and
δ40 of such function. These represent the width of f(x) at 40% of its maximum
72 CHAPTER 3. MEASUREMENT OF SECONDARY RADIATION
Figure 3.19: GSI charged fragments emission point (xPMMA, yPMMA) spatial distribution at90. The vertical dashed line on the left indicates the coordinate of the beam entrance in thePMMA at xBE = -(6.45 ± 0.02) cm. The dashed line on the right indicates the theoreticalcalculation of the Bragg peak position xBragg = (1.80 ± 0.02) cm, in the current frame ofreference.
(∆40 = Xright - Xleft, being Xleft and Xright the x coordinates corresponding to the
rising and the falling edges respectively), and the distance between Xleft and the
x intercept of the tangent to f(x) at x =Xright, as shown in Figure 3.20. It is important
to note that the emission profiles of all isotopes relative to both datasets at 60 and
90 are well described by f(x).
The accuracy of this method strongly depends on several factors: the statistics
of collected data, multiple scattering suffered by the particle along the path inside
the patient and intrinsic fluctuation of nuclear fragmentation process, due to its sta-
tistical nature. While statistics of collected data strongly depends on the dose that
is being delivered and the geometrical acceptance of the detector, in the reminder of
this section the other two sources of uncertainty will be analyzed.
The contribution of multiple scattering to the global resolution can be evaluated
from the distribution of yPMMA. This, in fact, represents the vertical component of
the beam profile and should be, theoretically, comparable with the beam spot size
of the accelerator. This is nominally a Gaussian centered in y = 0 with a standard
deviation of σbeam ≈ 1 cm. Then, the multiple scattering worsening can be obtained
from:
σ2yPMMA
= σ2beam + σ2
MS + σ2DCH (3.11)
3.3. CHARGED SECONDARY PARTICLES 73
Figure 3.20: Longitudinal profile of charged particles emission point as a function of thetarget thickness described with the function proposed in equation 3.10. All the interestingquantities are represented as well as a graphical definition of Xleft and Xright.
where σ2yPMMA
can be derived from data (see Figure 3.21), while the quadrature con-
tribution of Drift Chamber tracking resolution (σDCH ≈ 200 µm) has to be swum
over 40 cm from the Drift Chamber to the beam line. In a worst case scenario this
will lead to:
σPMMADCH ≈ l · σDCH/d√
Nhits≈ 400 mm · 2 · 0.2 mm/80 mm√
4≈ 1 mm (3.12)
where l is the distance from the chamber to the target, d is the distance between
the first and the last plane of the chamber (80 mm) and a safe value of hit cells per
view Nhits = 4 has been chosen. Hence, for this particular setup, the uncertainty
due to multiple scattering has been inferred to be σMS ≈ 6 mm. This contribution
could increase of a factor 2÷3 during a real treatment, where the amount of material
traversed by the ions is somewhat larger. Multiple scattering can pollute the spatial
information carried by the particle and its minimization is of vital importance in
order to properly reconstruct the emission profile. However, considering the energy
spectra measured at GSI reported in Figure 3.22, selecting a particular subset of data
(e.g. protons with kinetic energy larger than 100 MeV), the deflection of particle
trajectories due to multiple scattering would be of minor importance.
In order to estimate the fluctuation of the emission mechanism, the available data
have been subdivided in reference samples of 103 events. These roughly correspond
to 2.3 × 108 impinging carbons at 90 and 4.7 × 107 at 60, according to the fluxes
reported in equation (3.8) and (3.9), given the reduced acceptance of the LYSO de-
tector. As a reference, in a standard treatment a single pencil beam aimed at the
74 CHAPTER 3. MEASUREMENT OF SECONDARY RADIATION
Figure 3.21: Comparison between GSI data (left) and Monte Carlo simulation (right) ofyPMMA distribution at 90.
Figure 3.22: GSI measured kinetic energy at 90 for protons (blue), deuterons (red) andtritons (magenta).
distal contour of the tumor (for which the detector monitoring capability is of crucial
importance) receives an ion density of the order of 108 particles per cm3. Then, for
every subset of 103 events ∆40 and δ40 have been evaluated. All data sets acquired
at 60 need a further pre-processing step. The measured shape results, in fact, in a
convolution of the beam spot size σbeam and the detection angle: σ = σbeam · cotgθ
3.3. CHARGED SECONDARY PARTICLES 75
(as shown in Figure 3.7). The results on a population of 13 subsets at 90 and 100
at 60 are listed in Table 3.2 where the precision of the measurement of ∆40, δ40 and
Xleft and their mean values are reported.
Angle (deg) σ∆40 (cm) σδ40 (cm) σXleft (cm) ∆40 (cm) δ40 (cm)
90 0.34 0.37 0.08 6.60 ± 0.09 9.40 ± 0.1060 0.31 0.28 0.09 6.83 ± 0.03 9.44 ± 0.03
Table 3.2: Mean values and dispersion of ∆40 and δ40 parameters.
An extensive calibration campaign with several beam energies (and hence sev-
eral penetration depths) is eagerly needed and will be crucial for this method to be
validated. Furthermore, the precision obtained onXleft could represent an interesting
added value, since it is strongly related to the beam entrance position in the patient
and it potentially could be used to spot eventual patient mis-positioning during the
treatment.
CHAPTER
4
Dose Profiler optimization
Content
4.1 Detector overview . . . . . . . . . . . . . . . . . . . . . . . . . . . 78
Neutral radiation operation mode . . . . . . . . . . . . . . . . . . . 79
Charged particles operation mode . . . . . . . . . . . . . . . . . . 81
4.2 Simulation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 82
Dose Profiler layout optimization . . . . . . . . . . . . . . . . . . . 83
In order to monitor the dose release during a hadrontherapy treatment, an innova-
tive detector able to exploit both prompt photons and charged fragments to identify
the Bragg peak position has been designed, and its project will be carefully described
from this chapter on. This twofold nature in a single detector is a unique feature,
never suggested so far, that will allow to maximize the information on dose depo-
sition that can be collected during a conventional treatment, considering its highly
restrictive time constraints. Lack of statistics, together with detector encumbrance
and clinical workflow necessities, is, in fact, one of the most challenging issues to
overcome when treatment-like conditions come into play. Detector spatial resolu-
tion and, therefore, reconstruction precision of secondary radiation emission, only
to mention a few, are strongly affected by the number of events actually available.
This is also one of the reasons why only small angles (with respect to beam direction)
have been considered for charged particles monitoring so far. The optimization of the
detection angle θ, as mentioned in the previous chapter, is a crucial task. At narrow
angles the measurement has the clear advantage that the emission flux is enhanced
77
78 CHAPTER 4. DOSE PROFILER OPTIMIZATION
and that charged particles energy spectrum is harder. Thus, the multiple scattering of
protons, deuterons and tritons (charged fragments or simply charged in the follow-
ing) inside the patient is minimized. On the other hand, due to the projection on the
beam line, the spatial resolution on the emission shape should worsen as sin(θ)−1
and this geometrical contribution could become dominant for small detection an-
gles. Furthermore, for measurement at θ 6= 90, the emission is convoluted with the
transverse beam spot size projected on the beam line. Both quoted effects get worse
as the detection angle shrinks. On the other hand, the spatial distribution of neu-
tral radiation is isotropic. This in principle does not pose any constraint for prompt
gamma detection, but in order to avoid DAQ saturation due to charged fragments for-
ward peaked emission distribution, and considering the limited space available for an
external monitoring device in a treatment room, detection angles larger than 30 are
strongly encouraged. Taking into account all the aforementioned issues, this proposal
has been included in the “INSIDE” project, that has been awarded a MIUR PRIN-
2010 funding. INSIDE foresees the development of a dual mode device that will be
installed and operated at CNAO treatment room, together with an online PET sys-
tem, at a detection angle θ ≥ 60 with respect to the beam direction. In Figure 4.1
is shown the 3D rendering of CNAO treatment room with the integration of the two
detectors.
4.1 Detector overview
The Dose Profiler has been designed to work as a charged particle detector and
a Compton camera simultaneously. It is composed of two main sub-detectors: a
tracker (made of six active planes) and a calorimeter (respectively TRK and CAL in
the following). Additionally, an electron absorber (ABS hereafter) has been placed
after the tracker, in order to prevent any recoil electron from reaching the calorimeter.
The reason of such choice will be discussed in the following paragraphs.
Such twofold nature, especially considering prompt gammas measured energy
spectrum (shown in Figure 3.11), implies a careful study and optimization of the ma-
terials that constitute the detector itself. Compactness, reliability, large geometrical
acceptance and high tracking efficiency represent, in fact, the minimum requirements
of such detector, considering its foreseen use into clinical workflow. A sketch of the
longitudinal section of the Dose Profiler and its 3D reconstruction taken during the
simulation phase is shown in Figure 4.2, where all the active elements, as well as the
building structures, are depicted.
4.1. DETECTOR OVERVIEW 79
Figure 4.1: 3D rendering of Dose Profiler integration (green) inside CNAO treatment room.(Top-left) parking position; (top-right) operation during treatment at +90; (bottom-left) op-eration during treatment at -90; (bottom-right) particular of Dose Profiler mechanical struc-ture.
Neutral radiation operation mode
As previously stated, prompt photons detection will be performed exploiting a
working principle similar to that of a Compton camera. Such device, in its classi-
cal realization, is made of two distinct detectors: a scatterer and an absorber. The
incoming photon, while traversing the scatterer medium, has a certain probability
to undergo Compton scattering, depending on thickness and atomic number of the
medium itself. This will produce a deflected photon (γ′), that is detected by the ab-
sorber (Eγ′ , ~rγ′) and a recoil electron (e′), whose energy and position are recorded
by the scatterer itself (Ee′ , xe′ , ye′ , ze′). From Compton equation (1.12) a cone of
acceptable directions is then derived. The original gamma ray source is determined
integrating different events (i.e. by the intersection of various cones).
A slightly different approach has been chosen for the Dose Profiler, that is com-
posed, in fact, by six scattering planes. These will provide additional information on
the electron, allowing to track also its direction. In order to close Compton kinemat-
ics, scattered photon direction must be detected as well. This can be done exploiting
80 CHAPTER 4. DOSE PROFILER OPTIMIZATION
(a) (b)
Figure 4.2: Dose Profiler simulated layout. (a) Dose Profiler longitudinal section. All theactive elements of the detector are present as well as the main support structures: trackingplanes (black - plastic scintillator), front-end electronics printed circuit boards (green - fiber-glass), electron absorber (cyan - plastic scintillator), calorimeter (red - LYSO) and detectorcase (gray - aluminum). (b) 3D view of the detector.
the high density scintillating crystal placed behind the ABS with a position sensitive
readout (e.g. using Multi-Anode-PhotoMultiplier Tubes). The choice of such a scin-
tillator is justified considering the final application of the detector, that must be as
compact as possible for its operation in a clinical environment. However, given the
compactness of the calorimeter and LYSO absorption length for the expected energy
range (≈ 16 ÷ 18 mm), the energy information of the Compton gamma would be
surely deteriorated because of poor photon containment or low interaction probabil-
ity. For this reason, a statistical approach for Compton kinematics reconstruction has
been proposed and it will be reviewed in the following chapter.
Additional issues arise wether both scattered photon and electron are detected
inside the CAL. In this situation at least two clusters of photoelectrons will reach
the photocathode, as shown in Figure 4.3. The localization of scattered photon is
thus more complicated (especially if clusters are close one to another) and generally
less precise. Moreover, the presence of a high Z material (such as LYSO, Zeff = 66)
results, for the electrons, in a higher probability of being back-scattered (Tabata et
al. [7] and Table 1.3). This will produce a backward traveling track inside the TRK,
that acts as background. For this reason a slab of plastic scintillator has been placed
in front of the calorimeter. Not only should the backscattering probability drastically
4.1. DETECTOR OVERVIEW 81
Figure 4.3: Simulation of a double cluster event in the calorimeter. In this particular casethe two populations are well separated and their boundaries are easily defined.
decrease (according to Tabata et al. [7] a reduction of a factor ten could be obtained)
but, if the thickness is properly adjusted, it should also absorb completely the electron
flux, reducing calorimeter clustering multiplicity and providing a very useful electron
energy measurement.
In the end, an additional help could come from the detection of e+ e− pairs,
produced after photon interaction with the TRK. This contribution is, of course, lim-
ited (its impact strongly depends on TRK building material) even though it could
complement Compton statistics only in the high energy range (i.e. Eγ ≥ 5 MeV).
Charged particles operation mode
As far as detector design is concerned, charged particles tracking in an energy
range of 10 ÷ 150 MeV does not represent a crucial issue from a technological
point of view. Dose Profiler spatial resolution, in fact, will be anyhow limited by
multiple scattering inside several centimeters of patient tissues. Since detector overall
resolution is given by the quadrature sum of the single uncertainty contributions, the
use of cutting-edge tracking devices with sub-millimetric precision would result in
an avoidable overshoot, when the uncertainty due to multiple scattering should be
comprised in the range of 8 ÷ 10 mm. For this reason, the tracker spatial resolution
has been designed to be of the order of few millimeters. In this way the overall
82 CHAPTER 4. DOSE PROFILER OPTIMIZATION
resolution should not remarkably worsen and, at the same time, the realization of
such detector will be not too critical in terms of costs and technology efforts.
Charged particles tracking will be performed by the six active layers that will pro-
vide a set of hit-points (xi, yi) and the same tracking algorithm of Compton electrons,
thoroughly described in the following chapter, will be used to reconstruct particle’s
trajectory. Moreover, the possibility to measure the energy of such radiation, even
with a poor resolution, would offer a valuable information on track quality. Charged
particles could be weighted, in fact, according to their kinetic energy, penalizing low
energy tracks that are more likely to have suffered larger deflection due to multiple
scattering inside the patient. For this reason also, a compact high density calorimeter
has been placed behind the TRK.
4.2 Simulation
The simulation of the detector has been focused on two main topics, resulting in
different levels of accuracy for the geometrical description of the simulated model. A
simplified geometry has been employed to define project foundations, eagerly needed
for detector dimensioning, such as: materials definition and size optimization, total
readout channels estimation and reconstruction software training. Then a more de-
tailed and refined simulation has been developed, in order to study, with the required
precision, the Dose Profiler behavior in different working conditions and to bench-
mark its performances. The reminder of this chapter will be devoted to describe
the optimization work while the following is devoted to present the reconstruction
software and the on beam simulated performances.
The Monte Carlo software used for the Dose Profiler simulation is FLUKA (re-
lease 2011.2). FLUKA is a general purpose tool for calculations of particle transport
and interactions with matter, covering a wide range of applications spanning from
proton and electron accelerator shielding to target design, calorimetry, activation,
dosimetry, detector design, cosmic rays, radiotherapy and so on. It can simulate the
interaction and propagation in matter of about 60 different particles including pho-
tons and electrons from 1 keV up to thousands of TeV. FLUKA can handle even very
complex geometries, using a combinatorial geometry package that allows to track
charged particles in presence of electric or magnetic fields. The software is fully
customizable via a set of user interface routines (written in Fortran 77), that allow to
control each step of the simulation, helping the user to meet any particular require-
ment. The simulation of the Dose Profiler demanded the development of several
4.2. SIMULATION 83
custom routines to better reproduce a realistic environment and to tailor the output
data to the foreseen DAQ data format. To this extent, the structure of the data for
each simulated event has been organized according to different abstraction levels. A
general track database has been created to store the information regarding all the
particles created for each event, such as: initial and final position (x, y, z)i,f , mo-
mentum (px, py, pz)i,f , particle type, particle parent, time of generation and so on.
Then, to each sub-detector has been associated another database containing all the
energy releases of the particles interacting with them. Quenching effect inside scin-
tillators has been taken into account according to Koba et al. [58]. Finally, a library
that interfaces FLUKA to ROOT (Brun et al. [59]), a commonly used data analysis
framework for high energy physics, has been built from scratch, in order to share the
same code both for the simulation analysis and the data analysis.
Dose Profiler layout optimization
All the simulations needed for the optimization of the layout used a customized
point-like source of 106 primary photons with an energy spectrum extracted from
the one measured by Agodi et al. [44] and reported in Figure 3.11. Prompt photons
expected statistics for a distal tumor slice of 2÷ 3 mm of thickness is about one order
of magnitude lower. Thus, simulating 106 primaries should reduce the statistical
fluctuations of the optimization study.
The photon emission has been chosen perpendicular to the TRK and aligned
with planes’ axis and a simplified detector geometry (planes, electron absorber and
calorimeter with a transverse section of 10 × 10 cm2) has been employed.
Tracker
Tracking planes design has been focused on two main aspects: the optimization
of the material constituting the TRK and its thickness. Given standard Compton
camera features, in fact, the choice of high Z materials (such as CMOS detectors) as
planes building blocks is rather appealing, because of their favorable Compton cross
section with respect to lighter materials. However, this option has to be seriously
pondered considering prompt photons measured energy spectrum (that is roughly
comprised between 1 ÷ 10 MeV, as shown in Figure 3.11). Scattered electrons ki-
netic energy, in fact, is expected to have a mean value of 2 ÷ 3 MeV and, since
multiple scattering is the main limiting factor of detector resolution, the higher the Z
of the planes the larger the electron’s angular straggling. Hence, planes dimension,
84 CHAPTER 4. DOSE PROFILER OPTIMIZATION
spacing, thickness and material have been severely tested and optimized, in order
to maximize the geometric acceptance and Compton cross section but, at the same
time, to control multiple scattering deflection of charged particles and electron tracks.
As already mentioned in the previous sections, the quest for TRK best material is a
crucial task, that can affect Dose Profiler global capability to work with the limited
amount of statistics foreseen in a typical hadrontherapy treatment. Hence, a compro-
mise between larger interaction probability and maximum deflection allowed must
be found. The two parameters that need to be optimized are clearly: (i) the number
of Compton events, and (ii) particle’s deflection angle due to multiple scattering (and
hence its contribution on the overall detector resolution).
The number of Compton events can be increased either enlarging the thickness
of the TRK planes or choosing a material with high atomic number (i.e. whose
Compton scattering cross section is enhanced). Unfortunately both these choices
have the undesirable effect of boosting the multiple scattering deflection. As rule of
thumb, recalling the equation 1.10, it would be preferable to increase the thickness of
the tracking planes, instead of using a high Z material, since in the former case the
standard deviation of multiple scattering angle increases as the square root of layer
thickness, while in the latter it grows linearly with the atomic number. A careful
analysis of the possible materials and thicknesses suitable for TRK has been made
considering multiple scattering contribution only on the first plane traversed (since
the electron track will be reconstructed relying mostly on the first two hits).
The choice has been restricted only to two options commonly used for charged
particle tracking: scintillating fibers (polystyrene) and CMOS sensors (silicon). While
the former solution needs two orthogonal planes to collect both the transverse inter-
action coordinates (xhit, yhit), with the latter a simultaneous readout can be performed
resulting in thinner planes. Squared scintillating fibers (0.25 ÷ 1 mm thick) and sili-
con strips (0.2÷ 0.3 mm thick) have been tested in order to find the best compromise
for TRK geometry.
The angular straggling suffered by the electrons inside the tracker has been esti-
mated as the angle between their production direction (pprode′ ) and the entry direction
in the following plane (i.e. the exit direction from the first plane pexite′ , being the
deflection in air negligible), as it is sketched in Figure 4.4.
In order to limit the data analysis only to interesting events, a realistic experimen-
tal “trigger condition” has been defined. This considers events with recoil electron
passing through at least 3 tracking planes and the scattered gamma ray interacting
inside the calorimeter. All the results presented in the following refer to this experi-
4.2. SIMULATION 85
Figure 4.4: Multiple scattering evaluation scheme. The angle between pprode′ and pexit
e′ hasbeen chosen as parameter to estimate the deflection suffered by Compton electron in the firstlayer.
mentally driven condition, unless differently stated.
Multiple scattering contribution to detector spatial resolution worsening has been
evaluated as follows:
1. once a Compton scattering matching the trigger condition occurs, the deflec-
tion angle of the electron track is calculated between the first two tracking
planes, as shown in Figure 4.4;
2. scattered photon direction and energy have been assumed to be known (ex-
tracted directly from Monte Carlo data);
3. the original photon direction is reconstructed and pointed backwards to its pro-
duction point, being the sources of uncertainty the deflection angle and the
Doppler correction of Compton scattering kinematics.
The photon source has been placed at a distance of 30 cm from the detector to
reproduce a realistic operating condition inside a treatment room. In Figure 4.5 the
distributions of the reconstructed source coordinates (xproj, yproj) for 2 × 1000 µm
scintillating fibers have been reported. The photon source was centered in (0,0,-30).
As it can be noted, these show a peculiar shape that results from the convolution of the
two aforementioned physical effects with the experimental cut introduced after the
trigger condition request. In order to find the best estimate for the multiple scattering
deflection two methods have been used: (i) a two Gaussians fit to the xproj, yproj
distributions, and (ii) the RMS extracted from the same dataset. In order to take
into account also the particles that undergo a severe scattering, this latter approach
has been used to estimate the single event resolution. The complete set of results,
86 CHAPTER 4. DOSE PROFILER OPTIMIZATION
relative to all the materials and the thickness under test is reported in Table 4.1 and
graphically summarized in Figure 4.6.
Figure 4.5: Distribution of the reconstructed source coordinates xproj (left) and yproj (right)obtained with planes of 2 × 1000 µm squared scintillating fibers. The global fit (red - solid)has been performed summing two Gaussian functions (black and blue - dotted) with the sameexpected value.
Material Thickness Compton evt. N. layers hit xproj yproj xproj/√N yproj/
√N
(µm) (trigger) (avg) (cm) (cm) (cm) (cm)
Polystyrene 2× 250 728 3.07 2.088 1.882 0.077 0.070Polystyrene 2× 500 1024 2.73 2.528 2.146 0.079 0.067Polystyrene 2× 750 1154 2.48 2.610 2.516 0.077 0.074Polystyrene 2× 1000 1033 2.29 2.622 2.592 0.082 0.081Silicon 200 612 3.04 2.859 2.394 0.116 0.097Silicon 300 789 2.87 3.154 3.282 0.112 0.117
Table 4.1: Multiple scattering study for tracker materials. Several scintillating fibers havebeen tested, as well as the two thinnest options on the market for silicon strips. The reso-lution has been evaluated projecting the reconstructed photon towards its production point(xproj, yproj). The RMS of xproj and yproj distributions has been used as evaluation parameterfor multiple scattering deflection. The total number of simulated primary particles is 106.
The aforementioned resolution, evaluated as the residual distance of the recon-
structed track from the photon source considering multiple scattering as the sole
source of uncertainty, gives an idea about the interplay of the two parameters that
have been optimized. For this reason it cannot be considered as a resolution in all
4.2. SIMULATION 87
Figure 4.6: (left) Single event resolution on xproj and yproj coordinates. (right) Overallresolution due to multiple scattering deflection. The single event resolution has been dividedby the square root of the number of events.
respects, since tracker efficiency, spatial resolution and tracking algorithm efficiency
have been neglected.
In order to better understand the dramatic impact of multiple scattering effect on
tracking accuracy, in Figure 4.7 have been reported the 2D plot of yproj versus xproj
when the effect is switched off (a) and on (b) directly in the simulation.
(a) (b)
Figure 4.7: Photon source reconstruction when (a) multiple scattering has been switched offfrom the simulation and only Doppler correction is present; (b) both effects are enabled inthe simulation. The source has been placed in (0,0,-30).
The evident tradeoff between the number of Compton events and the maximum
acceptable angular straggling of both the electron and, even if with minor conse-
quences, of the charged track has suggested to discard the silicon option (also con-
sidering its higher costs and technological issues for large geometries). Moreover,
from the results in Table 4.1 is clear that increasing the thickness of the planes, and
thus accepting to deal with a higher multiple scattering, can somewhat lead to a better
88 CHAPTER 4. DOSE PROFILER OPTIMIZATION
overall resolution (since the√N factor is enhanced).
Considering the results reported in Table 4.1 either 500 µm or 750 µm fibers
could be considered as the best compromise between number of Compton events,
angular straggling, detector’s efficiency and assembling simplicity. However, in or-
der to minimize the amount of material traversed by charged particles, 500×500 µm2
scintillating fibers have been chosen as tracking planes building blocks.
Electron absorber
The role of the ABS is to stop any recoil electron right after the TRK, in order
to minimize clustering effects inside the calorimeter. In fact, when a Compton scat-
tering occurs two secondary particles are produced. These, with a certain probability
distribution, could both reach the calorimeter interface producing two signals. Thus,
the CAL readout will show two clusters (one relative to the recoil electron and the
other to the scattered photon) that can worsen the detection accuracy and hence the
reconstruction of photon interaction point in the calorimeter. Moreover, interposing
a plastic scintillator before the CAL will also drastically reduce the number of back-
scattered electrons, thanks to its low effective atomic number (Zeff = 3.5) compared
to the one of LYSO (Zeff = 66). Then, the parameter that has been minimized is the
number of recoil electrons at the CAL boundary matching a trigger condition. ABS
thickness has been varied from 0.5 cm to 2.5 cm and the results of the optimization
are reported in Table 4.2 and in Figure 4.8.
Thickness e− on CAL e− on CAL e− stopped(cm) w/o ABS with ABS (%)
0 5785 5785 00.5 5686 1082 80.971 5578 506 90,93
1.5 5645 120 97.872 5579 26 99.53
2.5 5680 8 99.86
Table 4.2: Optimization of the electron absorber.
A thickness of 1.5 cm with over 97% of electrons absorption has been chosen as
final value.
4.2. SIMULATION 89
Figure 4.8: Stopping efficiency of the electron absorber as a function of its thickness.
Calorimeter
An array of LYSO scintillating crystals (widely used for PET imaging) has been
chosen as high density calorimeter and its role is, like the whole detector itself,
twofold. It is needed, in fact, for charged tracks weighting according to their own
kinetic energy (assigning a higher weight to harder particles, i.e. are more likely
to suffer less multiple scattering straggling) and for Compton photon direction re-
construction. Its readout will be performed via position sensitive photomultipliers
(Multi-Anode PhotoMultiplier Tubes, or MAPMT) made of an array of 8 × 8 sen-
sitive anodes. In light of what has been discussed thus far, it follows that a thicker
crystal would enhance scattered photon interaction probability and charged particles
energy measurement accuracy. It would also increase, however, the angular error
induced by the so-called Depth Of Interaction (DOI in the following) uncertainty. In
fact, the longitudinal interaction coordinate of Compton photon inside the CAL (zint)
cannot be directly determined with such detector, at least on first approximation. This
implies that the depth used for event reconstruction (zreco) has to be always set equal
to the calorimeter midpoint. Thus, the thicker the crystal the larger the uncertainty
on photon real direction (∆θ). Assuming ∆θ as the only source of uncertainty of
the detector, then the impact of such parameter on the overall tracking performance
has been evaluated. A simplified sketch of reconstruction uncertainty due to DOI is
shown in Figure 4.9, while the complete results of the optimization procedure are
reported in Table 4.3 and in Figure 4.10.
Since the contribution of DOI to the overall uncertainty is rather small, espe-
90 CHAPTER 4. DOSE PROFILER OPTIMIZATION
Figure 4.9: Sketch of DOI uncertainty on reconstructed photon direction. The prompt pho-ton undergoes a Compton scattering inside the tracker (black dot), scattered photon γ′ inter-acts inside the calorimeter at z = zint, while z = zreco (i.e. CAL half depth) is the coordinateused for event reconstruction. The uncertainty due to DOI can be estimated by means of theangle ∆θ. The sketch of the electron absorber has been omitted for clarity.
Thickness N. γ′ on CAL ∆θ Resolution uncertainty(cm) (trigger) (deg) (cm)
1 491 0.862 0.0271.5 648 1.140 0.0313 867 2.097 0.050
4.5 1044 2.605 0.0566 1157 3.031 0.062
Table 4.3: Calorimeter thickness optimization results considering only events matching thetrigger condition.
cially when it is compared to multiple scattering deflection, as it can be seen from
Table 4.3, only the interaction probability has been considered as optimization pa-
rameter. However, given the linearly growing costs and overall weight of thicker
crystals (that has to be somewhat contained for mechanical reasons), a thickness of
2 cm has been chosen as reference for the calorimeter.
In the following chapter the detailed simulation, the event reconstruction code,
that has been developed for the Dose Profiler, together with the global performances
on Monte Carlo data will be widely reviewed.
4.2. SIMULATION 91
Figure 4.10: Cyan - Number of Compton photons interacting inside the calorimeter. Red -Resolution worsening due to DOI uncertainty (arbitrary units).
CHAPTER
5
Event reconstruction anddetector performance
Content
5.1 Event reconstruction . . . . . . . . . . . . . . . . . . . . . . . . . 94
Prompt photons . . . . . . . . . . . . . . . . . . . . . . . . . . . . 94
Charged particles . . . . . . . . . . . . . . . . . . . . . . . . . . . 103
5.2 Detector performance evaluation . . . . . . . . . . . . . . . . . . . 104
Prompt photons . . . . . . . . . . . . . . . . . . . . . . . . . . . . 104
Charged particles . . . . . . . . . . . . . . . . . . . . . . . . . . . 105
In this chapter the event reconstruction working principle, its software implemen-
tation and the results on Monte Carlo data will be reviewed. As stated in the previous
chapter, a fully comprehensive simulation (scintillating fibers, front-end electronics
printed circuit boards and aluminum case included) has been developed in order to
have a more realistic and reliable dataset, with respect to the one used for layout
optimization. All the geometrical parameters and materials used for this simulation
have been reported in Table 5.1.
The event reconstruction will be focused mostly on neutral events, since charged
particle tracking does not represent a challenging issue for the development of a ded-
icated algorithm. In fact, as it will be clearer in the following, the direction of a
proton can be traced in the same way of a recoil electron produced as a consequence
of Compton scattering. Hence, a higher emphasis will be placed on Compton kine-
matics reconstruction.
93
94CHAPTER 5. EVENT RECONSTRUCTION AND DETECTOR
PERFORMANCE
Component Active area (cm) Material Density (g/cm3) ZeffDetector case 24.2 x 24.2 x 17.1 Aluminum 2.70 13Front-end frames 23.2 x 23.2 x 0.3 Fiberglass 1.85 –Tracker planes 19.2 x 19.2 x 0.05 Polystyrene 1.05 3.5Electron shield 20.0 x 20.0 x 1.5 Polystyrene 1.05 3.5Calorimeter 20.0 x 20.0 x 2.0 LYSO 7.40 66
Table 5.1: Dose Profiler simulated components.
5.1 Event reconstruction
The data analysis software that has been developed for the Dose Profiler has
been written in C/C++ and has been directly interfaced with ROOT, a commonly
used framework for particle physics data analysis.
Prompt photons
Event topology
For each event a pre-processing of Monte Carlo data has been performed in or-
der to classify the event itself according to photon interaction modality (Compton
scattering or pair production) and to the region where the first interaction took place.
Thus, five different possibilities are allowed: (i) air event, (ii) TRK event, (iii) ABS
event, (iv) CAL event and (v) no interaction. For each one of the aforementioned
possibilities either Compton or pair flag can be associated to the event (given the
photon energy range, photoelectric effect has been neglected). In order to have a
first glance on the event typology, a point-like source of isotropic prompt photons
has been simulated. Their energy has been extrapolated from LNS measured spec-
trum (see Figure 3.11). In order to reduce the statistical fluctuations, a sample of 106
primary photons has been employed. To estimate the average number of photons pro-
duced during the irradiation of a tumor slice with n Gy of carbon ions, the following
formula has been used:
Nexpγ = n [Gy] · dNC
slice · Gy·dN2
γ
dNCdΩ· dΩDP (5.1)
the second multiplicative factor, as suggested by Kramer et al. [60] for a 120 mm3
tumor divided in 39 slices with 3 mm pitch, can be assumed to be 1.8 × 107 carbon
ions per slice per Gray. The third factor is obtained from GSI measured flux, reported
in equation (3.5), while the solid angle covered by the Dose Profiler (dΩDP) can be
5.1. EVENT RECONSTRUCTION 95
evaluated supposing to place the detector at a distance of 30 cm from the patient:
dΩDP ≈SLYSO
d2=
20.8× 20.8 cm2
[(30 + 14.6) cm]2= 0.22 sr (5.2)
Thus, the simulated sample roughly corresponds to a flux ten times bigger than the
one expected for a single tumor slice irradiated with 4 Gy:
Nexpγ = 4 [Gy] · 1.8× 107
[12C/slice · Gy
]× 5.6 · 10−3
[Nγ/
12C · sr]× 0.22 [sr]
that leads to Nexpγ ≈ 90,000 photons/slice.
The results of the event typology scan have been reported in Table 5.2. As ex-
Event type Isotropic source
No interaction 60.69 %CAL event 32.22 %ABS event 4.65 %TRK event 2.25 %AIR event 0.19 %
Table 5.2: Event typology scan, the uncertainty on the reported values is of the order of0.1 %.
pected, the probability of a Compton event inside 500 µm of scintillating fiber is
rather low (of the order of 2 %). If the experimental trigger condition defined in the
previous chapter is requested (at least 3 planes traversed by the recoil electron and the
scattered photon detected in the CAL), this probability drops down to 0.1 %, hence
representing the main limitation to this technique. Its main advantage, on the other
hand, lies in prompt photons capability to exit the patient regardless of their energy,
as opposed to protons that, instead, clearly show an energy threshold behavior that
will be reviewed in the following.
Direction reconstruction: a Monte Carlo driven approach
In a standard Compton camera, to reconstruct the direction of the incoming pho-
ton, both the energy and the position of the scattered photon and the recoil electron
are needed. Considering the poor energy resolution of 500 µm scintillating fibers, a
statistical approach, driven by Monte Carlo simulation, has been proposed instead.
The kinematics of Compton scattering (assuming the atomic electron to be at rest) is
given by:
~pγ = ~pγ′ + ~pe− (5.3)
96CHAPTER 5. EVENT RECONSTRUCTION AND DETECTOR
PERFORMANCE
that can be reformulated as follows:
|pγ | pγ = |pγ′ | pγ′ + |pe− | pe− (5.4)
pγ = c1 pγ′ + c2 pe− (5.5)
where c1 and c2 are two positive coefficients defined by:
c1 =|pγ′ ||pγ |
; c2 =|pe− ||pγ |
(5.6)
that must satisfy the normalization condition:
c21 + c2
2 + 2c1c2 cos θ = 1 (5.7)
being θ the angle between scattered photon and recoil electron direction. In this way,
once the direction of the scattered particles is known, it is possible to reconstruct the
direction of the original gamma ray using, for instance, the mean values of c1 and
c2 distributions extracted from Monte Carlo simulation (cMC1 , cMC
2 ), that are shown
in Figure 5.1. Moreover, an important information could be derived if the electron
energy would be recorded by ABS. This, in fact, would represent a valuable weight
for cMC2 (and consequently for cMC
1 ) choice. However, an extensive calibration of this
method, for different beam energies and primary ions, is required in order to optimize
the coefficient selection technique.
Then, the reconstructed momentum unit vector (p recoγ ) has to be extrapolated
backwards to its generation point (~p recoγ ). An error vector ∆~r can be hence defined,
whose projections along x and y axes (∆y and ∆y respectively) have been considered
as transverse resolution estimators, as sketched in Figure 5.2.
Compton event reconstruction
Once a Compton event has been recognized, the reconstruction will proceed in
two steps: (i) recoil electron tracking and (ii) photon interaction point detection inside
the CAL. The two steps will be thoroughly described in the following paragraphs and
the flow chart of the reconstruction algorithm has been sketched in Figure 5.3.
Electron direction The reconstruction of the electron direction represents the most
challenging part of the tracking process. In the energy range of interest for hadron-
therapy applications, the electron trajectory inside the detector is far from being
straight, due to the multiple scattering contribution. To this extent, a simple linear
chi square fit would not achieve satisfactory results and a more complex algorithm
5.1. EVENT RECONSTRUCTION 97
c1 coef.Entries 2050Mean 0.3647RMS 0.2131
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 10
10
20
30
40
50
c1 coef.Entries 2050Mean 0.3647RMS 0.2131
(a)
c2 coef.Entries 2050Mean 0.7223RMS 0.192
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 10
10
20
30
40
50
c2 coef.Entries 2050Mean 0.7223RMS 0.192
(b)
Figure 5.1: Monte Carlo values of cMC1 (a) and cMC
2 (b) parameters obtained simulating anisotropic point-like source of photons with LNS energy spectrum.
has been developed. The tracking process is then divided in two separate subsequent
tasks: (i) track finding and (ii) track fitting.
The track finding algorithm has the role to find and gather together all the hits
of a single track. Its simplified working principle is shown in Figure 5.4 and can be
summarized as follows:
1. the track segments (or seeds) are built considering all the combinations of the
hits (black stars) in the first two adjacent planes (s1, s2, s3 in Figure 5.4);
2. each seed is linearly extrapolated to the next plane;
3. the hit that minimizes the distance from the projected seed is clustered to the
segment (currently each hit can be associated to only one segment);
98CHAPTER 5. EVENT RECONSTRUCTION AND DETECTOR
PERFORMANCE
Figure 5.2: Reconstructed prompt photon direction (p recoγ ) is extrapolated backwards to itsproduction point (~p recoγ ). An error vector (∆~r) and its projections along x and y axes havebeen considered as spatial resolution estimators.
Figure 5.3: Compton event decoding flowchart: the raw data collected by the DAQ electron-ics are pre-processed in order to cluster adjacent fiber signals to form the so-called hits. Thenthe event can be decoded (Compton or charged particle) according to the hits distribution inthe tracker: if a Compton event has been recognized the process splits in two sub-routinesto reconstruct both electron and photon direction. Once the track parameters of the twosecondary particles have been obtained, the statistical reconstruction approach described inthe previous section can be applied. The boxes are colored differently depending on whichsub-detector is involved: green - tracker, gray - calorimeter and blue - absorber.
4. once the last plane is reached, a list of candidate tracks can be built.
5.1. EVENT RECONSTRUCTION 99
Figure 5.4: Track finding algorithm working principle. All the possible combinations of theexperimental hits (black stars) on the first two planes are used as seeds (s1, s2 and s3). Theseare then linearly extrapolated to the following planes and the hit that minimizes its distancefrom the projection is clustered to the track candidate.
Once all the hits have been clustered, the track fitting algorithm takes place in
order to give a quantitative description of the track parameters. As first step, a simple
chi square linear fit is performed separately for each view:
χ2 =∑N
(mi − fi,j)2
σ2i
(5.8)
wheremi is the i-th particle position measurement, fi,j = ai,jz+bi,j is the measure-
ment expectation value (function of the j-th track) and σi is the uncertainty of the i-th
measurement. The track parameters obtained with the linear fit (ax,zi,j , bx,zi,j , ay,zi,j and
by,zi,j ) are then passed to a Kalman filter [61], [62], that represents the second step of
the fitting algorithm.
What goes under the name of “Kalman filtering” is really a two step process,
consisting of a “filter” and a “smoother”. The former begins at the first hit of the track
making a prediction for the location of the next hit. The prediction is then refined,
according to the actual measurement, updating the error matrices. This procedure
goes on until the end of the track is reached. When the filtering process is finished,
the estimation of the track at any given plane does not have any information in it
about the hits further down the track. For this reason a smoothing is required in
order to incorporate such information. It steps back up the track from the bottom,
refining the track parameters at each step using the information from the “last” hits.
As it will be explained in the Appendix A, where a wider description of the Kalman
100CHAPTER 5. EVENT RECONSTRUCTION AND DETECTOR
PERFORMANCE
formalism used in this thesis is available, the use of a Kalman filter has the advantage
of balancing the multiple scattering effect and the measurement error.
Photon direction The calorimeter readout will be performed by 4× 4 Multi Anode
PhotoMultiplier Tubes (MAPMT) matrix, each of them made of 8 × 8 active anodes
directly coupled to LYSO crystals. In order to investigate the possibility of improving
the resolution along z coordinate (dominated by DOI effect), a continuous crystal
has been simulated as well. The optical photons distribution at the photocathode has
been hence used to carry out a detailed analysis focused on the optimization of the
longitudinal resolution.
In order to find the best evaluation method of photon interaction position inside
CAL, two strategies have been tested: (i) center of gravity and (ii) 2D Gaussian
fit. With the former, the center of gravity of the optical photon distribution in the
transverse plane has been calculated according to:
CG =
∑n1 Ni,j · ri,j∑n
1 Ni,j(5.9)
where Ni,j is the number of optical photons on a single pixel, ri,j is the pixel position
vector.
The second approach, instead, performs for each event a bi-dimensional Gaussian
fit to the optical photons distribution with a function defined by:
f(x, y) = A · exp
[−(x− µx)2
σ2x
− (y − µy)2
σ2y
](5.10)
where A is a positive constant, µx and µy are the mean values of the Gaussian func-
tion (they also represent the x and y coordinates of the measured interaction point)
and σx and σy its standard deviations (that give the spatial resolutions for the two co-
ordinates). In order to benchmark the two aforementioned methods, the same Monte
Carlo dataset has been used to test the two techniques. The result of such test, in
terms of spatial resolution on the transverse coordinates, has been shown in Fig-
ure 5.5, where the advantage of using the 2D fit in spite of center of gravity is clearly
evident. The 2D Gaussian resolutions for x and y coordinates have been reported in
Figure 5.6.
Depth of interaction As already mentioned in the previous chapter, since the
calorimeter readout will be performed in the transverse plane, there is no direct way
to measure scattered photon Depth of Interaction (DOI) inside the scintillating crys-
tal. However, this could be estimated from the shape of the transverse optical photon
5.1. EVENT RECONSTRUCTION 101
Figure 5.5: Resolution on reconstructed photon interaction point with center of gravity (red)and 2D Gaussian fit (black) methods for x (left) and y (right) coordinate.
Figure 5.6: Resolution on photon interaction point x coordinate (left) and y coordinate(right) with 2D Gaussian fit. In both cases a resolution of the order of 400 µm has beenobtained.
distribution. In fact, a deeper interaction will result in a narrower photon distribution
and viceversa. For this reason a deterministic method has been proposed to guess the
DOI. The relationship between σx (and similarly σy) and the Monte Carlo true value
of zMCint has been evaluated from the calibration curve shown in Figure 5.7. Then, the
DOI (zestint) can be deduced from:
σx = AzMCint +B
zestint =
σxA−B (5.11)
The resolution on the depth of interaction has been computed and it is shown in
102CHAPTER 5. EVENT RECONSTRUCTION AND DETECTOR
PERFORMANCE
Figure 5.7: Calibration curve used for depth of interaction evaluation. On the y-axis the σx(a similar behavior has been obtained also with σy) of the 2D Gaussian fit has been plottedversus the true interaction depth (from Monte Carlo simulation).
Figure 5.8. In order to fully appreciate the benefits of this approach, it is interesting
to compare the σDOIz < 1 mm obtained with DOI calibration to the one that would
be obtained from a uniformly distributed variable along CAL thickness (20 mm)
σunifz = 20/
√12 ≈ 5.8 mm.
Figure 5.8: Resolution on z coordinate obtained with DOI calibration. In order to fullyappreciate the outcome of this approach, the final value of σz ≈ 800 µm must be comparedto the one that would be obtained from a uniformly distributed variable along calorimeterthickness (20 mm) σunif
z = 20/√
12 ≈ 5.8 mm.
5.1. EVENT RECONSTRUCTION 103
Charged particles
As already mentioned in the previous sections, the reconstruction of charged par-
ticles does not pose any further complication as far as the software development is
concerned. In fact, the same logic scheme of recoil electrons can be applied also to
protons (since they do undergo multiple scattering as well, even though with minor
impact given their greater mass). The flow chart of proton tracks decoding algorithm
is shown in Figure 5.9 and, as it can be clearly noted, it is fully inherited from the
one of the Compton case. In this case the only difference is given by CAL and ABS
Figure 5.9: Charged event decoding flowchart. The logical flux is the same of the oneof recoil electrons shown in Figure 5.3. The only difference lays in CAL and ABS trackweighting according to proton’s kinetic energy. This step is of fundament importance, sinceit balances and limits the multiple scattering worsening effect on detector resolution. Theboxes are colored differently depending on which sub-detector is involved: green - tracker,gray - calorimeter and blue - absorber.
weighting of the track according to its kinetic energy. In the decoding phase it is
useful, in fact, to assign a higher importance to those particles which are more likely
to have suffered less multiple scattering (i.e. high energy protons) and hence whose
spatial information is more accurate.
104CHAPTER 5. EVENT RECONSTRUCTION AND DETECTOR
PERFORMANCE
5.2 Detector performance evaluation
Prompt photons
Two different sets of simulations have been produced in order to evaluate de-
tector performance and to test reconstruction algorithm capabilities against prompt
photons. In a preliminary phase, in order to maximize the geometrical acceptance of
the detector, a mono-directional gamma source emitting photons towards the center
of the Dose Profiler front plane has been simulated. Then, to represent a more real-
istic situation, an isotropic gamma source has been chosen. Both sources have been
placed at 30 cm from detector front plane and have been aligned along detector lon-
gitudinal axis. The energy spectrum of the source has always been extracted from the
experimental one measured at LNS (see Figure 3.11). A sketch of the two simulated
geometries has been shown in Figure 5.10. For each photon source another fiber
Figure 5.10: Simulated setup for prompt photons performance evaluation. Left: mono-directional source emitting towards the detector center. Right: isotropic photon source.
thickness scan has been performed, in order to validate the results obtained with the
layout optimization taking into account also photon and electron tracking efficiency,
event selection efficiency and the different spatial resolution of the interaction point
coordinates. The complete results have been reported in Table 5.3, while an exam-
ple of resolution distribution for 2×500 µm scintillating fibers and isotropic photon
source has been shown in Figure 5.11. It is interesting to observe that while the
single event resolution does not show a significant change for the two configurations,
the number of triggering events is drastically reduced, as expected, with an isotropic
source. This affects the overall resolution, since it is obtained dividing single event
resolution by the square root of the event number.
However, given the originality and the prototypical nature of such detector, that
still needs a severe optimization and calibration of the reconstruction algorithm, an
overall resolution of (6.42 ± 0.16) mm represents certainly a very promising result.
5.2. DETECTOR PERFORMANCE EVALUATION 105
Source Fiber thickness N. trigger Single evt. resolution Overall resolution(µm) (mm) (mm)
mono-directional2 × 250 245 57.9 ± 0.7 3.70 ± 0.042 × 500 440 65.8 ± 0.6 3.14 ± 0.032 × 1000 736 76.5 ± 0.6 2.82 ± 0.02
isotropic2 × 250 66 61.3 ± 1.9 7.55 ± 0.232 × 500 120 70.3 ± 1.7 6.42 ± 0.16
2 × 1000 159 81.0 ± 1.2 6.42 ± 0.11
Table 5.3: Prompt photon performance with two different source typologies: mono-directional and isotropic gammas emitted on-axis at 30 cm from the Dose Profiler. Thenumber of trigger events has been evaluated for a realistic primary sample of 90,000 pho-tons/slice. The overall resolution has been obtained dividing single event resolution for thesquare root of N. trigger.
Figure 5.11: Single event resolution for an isotropic on-axis prompt photon source. Witha fiber thickness of 500 µm a single event spatial resolution of (70.3 ± 1.7) mm has beenobtained, dividing this number by the square root of the expected statistic sample an overallresolution of (6.42 ± 0.16) mm can be achieved.
Charged particles
In order to test the behavior of Dose Profiler against charged particles a some-
what different simulation has been implemented. A patient head, schematized by
two concentric spheres made of cortical bone (outer: rbo = 10 cm) and brain (inner:
rbr = 9 cm) ICRU certified materials, has been modeled and two proton sources have
106CHAPTER 5. EVENT RECONSTRUCTION AND DETECTOR
PERFORMANCE
been placed in the center of the sphere and at 5 cm from the center, as shown in Fig-
ure 5.12. In this way two different clinical scenarios can be reproduced: a worst case
where the tumor is located exactly in the center of patient head, and an intermediate
case where the tumor is closer to patient bone. This variation of the simulated setup
has been imposed by proton’s different interaction with matter, with respect to pho-
tons. Charged particles, in fact, lose energy in a continuous way and the presence of
10 cm or 5 cm of patient head will change dramatically the performance of the detec-
tor. Proton’s initial kinetic energy has been varied between 75 MeV and 250 MeV for
Figure 5.12: Simulated setup for protons performance evaluation. Two concentric sphereshave been used to model a human head, choosing ICRU certified materials for cortical boneand brain. A proton source placed at the center of the head (blue) and one placed at 5 cmfrom the skull (red) have been employed to schematize two different clinical scenarios.
the centered source and between 70 MeV and 140 MeV for the displaced source. The
results of the efficiency scan (number of tracked protons divided by the number of
incoming protons) have been reported in Figure 5.13, while the spatial resolution as a
function of the initial kinetic energy has been shown in Figure 5.14. A clear thresh-
old effect is present, as expected, for charged particles detection. Once the energy
threshold is exceeded, protons can escape from the patient head and can be detected
with a rapidly increasing efficiency. This holds true also for Dose Profiler spatial
resolution, that converges around 2 ÷ 3 mm. As mentioned above, such resolution
represents a very promising milestone for charged particle monitoring, considering
the fact that this technique has been proposed only in the last couple of years and it
needs extensive calibration and verification campaigns in order to validate its results.
Additionally, the observed threshold effect makes the possibility to use also prompt
radiation even more appealing, since it could complement and increase the statistic
data sample acquired during a therapeutic treatment.
5.2. DETECTOR PERFORMANCE EVALUATION 107
Figure 5.13: Dose Profiler efficiency versus proton’s initial kinetic energy. As expected,a threshold effect (Ekin ≈ 120 MeV for the centered source, and Ekin ≈ 80 MeV for thedisplaced source) is present. However, it is interesting to note that once the charged particlecomes out from the patient, this is detected by the Dose Profiler with a rapidly increasingefficiency.
Figure 5.14: Dose Profiler spatial resolution versus proton’s initial kinetic energy for thecentered (blue) and the displaced (red) sources. For both cases the effect of the energy thresh-old is clearly visible. When the kinetic energy is slightly larger than the threshold value themultiple scattering effect pollutes the spatial information of the track. However, increasingthe energy the resolution rapidly converges to 2 ÷ 3 mm.
Conclusions
Hadrontherapy is a fast growing reality in cancer treatment, which is rapidly
proving to be a valid alternative to radiotherapy for an increasing number of tumors.
More than ten facilities are currently operating in Europe (two of them in Italy alone)
and this number will duplicate in the next few years. One of the major hindering
factors that is actually limiting its diffusion is, without any doubts, the lack of mon-
itoring devices that could allow a reliable and precise treatment-quality evaluation.
The higher spatial selectivity of ions in their energy deposition poses, in fact, serious
issues and constraints both on the beam delivery and on the required precision of
dose monitoring for a fast and effective treatment-quality feedback.
As of today, a single technique (referred to as PET) has been used in a clinical
environment, although for research purposes only. It exploits the collinear emission
of photons due to positron annihilation induced by β+ emitters produced as a conse-
quence of nuclear fragmentation of the projectile and the target. Unfortunately, this
method has shown several drawbacks the most limiting of which are its offline nature
and β+ emitters low activity.
However, other secondary particles have been proposed as means to estimate the
dose deposited inside a patient such as: prompt photons and charged particles emitted
after nuclei de-excitation and nuclear fragmentation respectively. The possibility to
use such sources to monitor the treatment quality is still a matter of research currently
under evaluation of the scientific community. It has been shown by several authors,
however, that a clear correlation between prompt photons, charged particles emission
region and Bragg peak position does indeed exist and it could be usefully exploited
for dose monitoring purposes.
109
110 CONCLUSIONS
The work presented in this thesis was firstly aimed at the preliminary measure-
ments on beam needed to demonstrate such correlation and then at the design of a
novel device capable of monitoring the Bragg peak position exploiting both prompt
photons and charged particles detection. This dual-mode functionality is a unique
feature, never suggested so far, that will allow to increase the data sample collectable
during a typical treatment considering its highly restrictive time constraints. Lack of
statistics, together with detector encumbrance and clinical workflow necessities, is,
in fact, one of the most challenging issues to overcome when treatment-like condi-
tions come into play. However, this dual working modality comes at the expenses
of a more complex layout, that needs to be carefully optimized for several, often
conflicting, effects.
The device, called Dose Profiler, will work as a charged particles tracker and a
Compton camera and is made of three sub-detectors: (i) a scintillating fibers tracker,
(ii) a plastic scintillator absorber and (iii) a LYSO calorimeter. The first has the
twofold role of tracking protons and scatter photons, the absorber stops all the recoil
electrons measuring also their energy, while the calorimeter weights proton tracks
(according to their kinetic energy) and provides scattered photon direction by means
of a position sensitive readout. In order to close Compton kinematics avoiding scin-
tillating fibers poor energy measurement, a statistical approach, driven by Monte
Carlo simulations, has been proposed instead.
The Dose Profiler layout has been optimized limiting the multiple scattering ef-
fect in the tracker layers, increasing the absorber stopping efficiency and minimizing
the contribution due to depth of interaction uncertainty on photon position detec-
tion inside the calorimeter. Then, a dedicated software for the event reconstruction
and the performance evaluation has been developed and tested against two sets of
simulations. A point-like isotropic photon source and a simplified model of human
head with a proton source inside it have been used to assess the capabilities and
the criticality of such detector. The preliminary results, although obtained only on
Monte Carlo data, seem very promising showing a spatial resolution on the recon-
structed dose release position of the order of 5 ÷ 6 mm for photons and 2 ÷ 3 mm
for protons. Considering the originality of the proposed approach and the fact that
the detector is still in its early prototyping phase, this novel device showed great po-
tential for its application in a treatment room foreseen in 2016. It needs, however,
to be severely tested on beam, in order to confirm such encouraging performance,
as its reconstruction software still has to be optimized, thus leaving room for further
improvements.
APPENDIX
A
Kalman filter formalism
The Kalman filter formalism, that has been used for the electron reconstruction
algorithm, focuses on a p × 1 state vector (xk) that contains the state parameters to
be estimated, and on a model that propagates the state vector from layer to layer. The
state vector extrapolation model is given by:
xk = Fk−1 xk−1 + wk−1 (A.1)
where Fk−1 is the Jacobian matrix of xk, that propagates the state vector from point
k − 1 to point k, and wk−1 represents the noise that corrupts the information con-
tained in the state vector, that in our case of interest is due to multiple scattering.
The process noise is assumed to be unbiased and to have finite variance, its covari-
ance matrix is represented by Qk. However, the components of xk are not measured
directly. The m measurements mk at point k are linear functions of xk such that:
mk = Hk xk + εk (A.2)
where mk is a m × 1 vector, Hk is the observation model which maps the true state
space into the observed space (m× p matrix), and εk represents measurement noise
(in analogy to the process noise also εk is assumed unbiased and with finite variance,
its covariance matrix is Vk).
The working principle of the filter can be divided in two distinct phases: predic-
tion and update. In the prediction phase the estimate of x at time k given observations
up to and including at time k−1 (xk−1k ) is called “a priori” state estimate. The a priori
prediction is then updated with the actual observation to refine the state estimate, thus
obtaining the so called “a posteriori” state estimate (xkk). The two phases typically
111
112 APPENDIX A. KALMAN FILTER FORMALISM
alternate. In the prediction phase the predicted state estimate and estimate covariance
are given by:
xk−1k = Fk xk−1
k−1 (A.3)
Ck−1k = Fk C
k−1k−1 F
Tk +Qk (A.4)
After observation and actual measurement the information about the predicted state
can be complemented. Let rk andRk be the measurement residual and its covariance
matrix respectively, these are defined as:
rk = mk −Hk xk−1k (A.5)
Rk = Hk Ck−1k HT
k + Vk (A.6)
The Kalman optimal gain Kk is:
Kk = Ck−1k HT
k R−1k (A.7)
and finally the updated state estimate and the state covariance can be derived as:
xkk = xk−1k +Kk rk (A.8)
Ckk = (I −KkHk)Ck−1k (A.9)
In our case of interest a 4 × 1 state vector x0 has been defined:
x0 =
P1
P2
P3
P4
=
x0 − z0 · ax,0y0 − z0 · ay,0
ax,0
ay,0
(A.10)
where (P1,P2,0) is the intercept of the unscattered track with the reference x-y plane
at z = 0, P3 and P4 are the angular coefficients in the reference system (ax,0 and
ay,0). In this particular case, the multiple scattering covariance matrix Qk has been
evaluated by Wolin et al. [63] to be:
Qk =
z2
0 〈P3, P3〉 z20 〈P3, P4〉 −z0 〈P3, P3〉 −z0 〈P3, P4〉
z20 〈P3, P4〉 z2
0 〈P4, P4〉 −z0 〈P3, P4〉 −z0 〈P4, P4〉−z0 〈P3, P3〉 −z0 〈P3, P4〉 〈P3, P3〉 〈P3, P4〉−z0 〈P3, P4〉 −z0 〈P4, P4〉 〈P3, P4〉 〈P4, P4〉
(A.11)
Bibliography
[1] J. Jackson, Classical Electrodynamics, 2nd ed. New York: John Wiley & sons,
1975.
[2] H. Bethe and J. Ashkin, “Passage of radiations through matter,” Experimental
Nuclear Physics, vol. 1, 1953.
[3] W. Leo, Techniques for Nuclear and Particle Physics Experiments: A How-To
Approach, 2nd ed. New York, Berlin, Heidelberg: Springer–Verlag, 1994.
[4] R. Serber, “Nuclear reactions at high energies,” Physical Review, vol. 72, pp.
1114 – 1115, 1947.
[5] J. D. Bowman, W. J. Swiatecki et al., “Abrasion and ablation of heavy ions,”
(unpublished) LBL Report No LBL-2908 University of California, 1973.
[6] V. L. Highland, “Some practical remarks on multiple scattering,” Nuclear In-
struments and Methods, vol. 129, no. 2, pp. 497 – 499, 1975.
[7] T. Tabata, R. Ito et al., “An empirical equation for the backscattering coefficient
of electrons,” Nuclear Instruments and Methods, vol. 94, no. 3, pp. 509 – 513,
1971.
[8] F. Titus, “Measurements of the energy response functions of scintillators for
monoenergetic electrons,” Nuclear Instruments and Methods, vol. 89, no. 0, pp.
93 – 100, 1970.
113
114 BIBLIOGRAPHY
[9] U. Amaldi, “History of hadrontherapy in the world and italian developments,”
Rivista Medica, vol. 14, no. 1, pp. 7 – 22, 2008.
[10] R. R. Wilson, “Radiological Use of Fast Protons,” Radiology, vol. 47, pp. 487
– 491, 1946.
[11] Particle therapy co-operative group home. Available online: http://ptcog.web.
psi.ch
[12] J. S. Loeffler and M. Durante, “Charged particle therapy’ optimization, chal-
lenges and future directions,” Nature Reviews Clinical Oncology, vol. 10, no. 7,
pp. 411 – 424, 2013.
[13] C. Tobias, J. Lyman et al., “Radiological physics characteristics of the extracted
heavy ion beams of the bevatron,” Science, vol. 174, pp. 1131 – 1134, 1971.
[14] E. Pedroni, R. Bacher et al., “The 200-mev proton therapy project at the
paul scherrer institute: conceptual design and practical realization,” Medical
Physics, vol. 22, no. 1, pp. 37 – 53, 1995.
[15] T. Haberer, W. Becher et al., “Magnetic scanning system for heavy ion therapy,”
Nuclear Instruments and Methods in Physics Research Section A: Accelerators,
Spectrometers, Detectors and Associated Equipment, vol. 330, no. 1, pp. 296 –
305, 1993.
[16] T. Haberer, J. Debus et al., “The heidelberg ion therapy center,” Radiotherapy
and Oncology, vol. 73, no. 2, pp. S186 – S190, 2004.
[17] Centro di adroterapia e applicazioni nucleari avanzate. Available online:
http://www.lns.infn.it/CATANA/CATANA/
[18] Centro nazionale di adroterapia oncologica per il trattamento dei tumori.
Available online: http://www.cnao.it
[19] ICRU report 51, Quantities and units in radiation protection dosimetry.
Bethesda, Md: International Commission on Radiation Units and Measure-
ments, 1993.
[20] E. Fokas, G. Kraft et al., “Ion beam radiobiology and cancer: Time to update
ourselves,” Biochimica et Biophysica Acta (BBA) - Reviews on Cancer, vol.
1796, no. 2, pp. 216–229, 2009.
BIBLIOGRAPHY 115
[21] D. Schardt, P. Steidl et al., “Precision bragg–curve measurements for light–ion
beams in water,” 2008, GSI-Report 2008-1 (GSI Scientific Report 2007), p. 373
(unpublished).
[22] M. Durante and J. S. Loeffler, “Charged particles in radiation oncology,” Nature
Reviews Clinical Oncology, vol. 7, no. 1, pp. 37–43, 2010.
[23] D. Schardt, T. Elsasser et al., “Heavy-ion tumor therapy: physical and radiobi-
ological benefits,” Rev. Mod. Phys., vol. 82, pp. 383 – 425, 2010.
[24] M. Scholz, “Effects of ion radiation on cells and tissues,” in Radiation Effects
on Polymers for Biological Use, ser. Advances in Polymer Science. Springer
Berlin Heidelberg, 2003, vol. 162, pp. 95 – 155.
[25] E. Hall, Radiobiology for the Radiologist. Philadelphia: Lippincott Williams
& Wilkins, 2000.
[26] G. W. Barendsen, C. J. Koot et al., “The effect of oxygen on impairment of
the proliferative capacity of human cells in culture by ionizing radiations of
different let,” International Journal of Radiation Biology, vol. 10, no. 4, pp.
317 – 327, 1966.
[27] D. K. Bewley, “A comparison of the response of mammalian cells to fast neu-
trons and charged particle beams,” Radiation Research, vol. 34, no. 2, pp. 446
– 458, 1968.
[28] Y. Furusawa, K. Fukutsu et al., “Inactivation of aerobic and hypoxic cells from
three different cell lines by accelerated 3he-, 12c- and 20ne- ion beams,” Radi-
ation Research, vol. 154, no. 5, pp. 485 – 496, 2000.
[29] A. Staab, D. Zukowski et al., “Response of chinese hamster v79 multicellular
spheroids exposed to high-energy carbon ions,” Radiation Research, vol. 161,
no. 2, pp. 219 – 227, 2004.
[30] E. A. Blakely, C. A. Tobias et al., “Inactivation of human kidney cells by high-
energy monoenergetic heavy-ion beams,” Radiation Research, vol. 80, no. 1,
pp. 122 – 160, 1979.
[31] Heidelberg ion-beam therapy center (hit). Available online: http://www.
klinikum.uni-heidelberg.de
116 BIBLIOGRAPHY
[32] B. Gottschalk and E. Pedroni, “Treatment delivery systems,” in Proton and
Charged Particle Radiotherapy, T. F. DeLaney and H. M. Kooy, Eds. Philadel-
phia: Lippincott Williams and Wilkins, 2008, ch. 5, pp. 33 – 49.
[33] C. P. Karger, O. Jackel et al., “Clinical dosimetry for heavy ion therapy,”
Zeitschrift fur Medizinische Physik, vol. 169, pp. 159 – 169, 2002.
[34] E. Pedroni and U. Schneider, “Proton radiography as a tool for quality control
in proton therapy,” Medical Physics, vol. 22, no. 4, pp. 353 – 363, 1995.
[35] A. M. J. Paans and J. M. Schippers, “Proton therapy in combination with pet as
monitor: a feasibility study,” Nuclear Science, IEEE Transactions on, vol. 40,
no. 4, pp. 1041–1044, 1993.
[36] K. Parodi and W. Enghardt, “Potential application of pet in quality assurance of
proton therapy,” Physics in Medicine and Biology, vol. 45, no. 11, pp. N151 –
N156, 2000.
[37] J. Pawelke, W. Enghardt et al., “In-beam pet imaging for the control of heavy-
ion tumour therapy,” in Nuclear Science Symposium, 1996. Conference Record.,
IEEE, vol. 2, 1996, pp. 1099 – 1103.
[38] W. Enghardt, K. Parodi et al., “Dose quantification from in-beam positron emis-
sion tomography,” Radiotherapy and Oncology, vol. 73, Supplement 2, no. 0,
pp. S96 – S98, 2004.
[39] K. Parodi, H. Paganetti et al., “Patient study of in vivo verification of beam
delivery and range, using positron emission tomography and computed tomog-
raphy imaging after proton therapy,” International Journal of Radiation Oncol-
ogy, Biology, Physics, vol. 68, pp. 920 – 934, 2007.
[40] K. Parodi, W. Enghardt et al., “In-beam pet measurements of β+ radioactivity
induced by proton beams,” Physics in Medicine and Biology, vol. 47, no. 1, pp.
21 – 36, 2002.
[41] C.-H. Min, C. H. Kim et al., “Prompt gamma measurements for locating the
dose falloff region in the proton therapy,” Applied Physics Letters, vol. 89,
no. 18, p. 183517, 2006.
[42] C. Min, J. Kim et al., “Determination of distal dose edge location by measur-
ing right-angled prompt-gamma rays from a 38 mev proton beam,” Nuclear
BIBLIOGRAPHY 117
Instruments and Methods in Physics Research Section A: Accelerators, Spec-
trometers, Detectors and Associated Equipment, vol. 580, no. 1, pp. 562 – 565,
2007.
[43] E. Testa, M. Bajard et al., “Monitoring the bragg peak location of 73 mev/u
carbon ions by means of prompt gamma-ray measurements,” Applied Physics
Letters, vol. 93, no. 9, p. 093506, 2008.
[44] C. Agodi, F. Bellini et al., “Precise measurement of prompt photon emis-
sion from 80 MeV/u carbon ion beam irradiation,” Journal of Instrumentation,
vol. 7, p. P03001, 2012.
[45] S. Kabuki, K. Ueno et al., “Study on the use of electron-tracking compton
gamma-ray camera to monitor the therapeutic proton dose distribution in real
time,” in Nuclear Science Symposium Conference Record (NSS/MIC), 2009
IEEE, 2009, pp. 2437–2440.
[46] T. Kormoll, F. Fiedler et al., “A compton imager for in-vivo dosimetry of proton
beams’ a design study,” Nuclear Instruments and Methods in Physics Research
Section A: Accelerators, Spectrometers, Detectors and Associated Equipment,
vol. 626 – 627, no. 0, pp. 114–119, 2011.
[47] B. Braunn, M. Labalme et al., “Nuclear reaction measurements of 95 mev/u 12c
interactions on pmma for hadrontherapy,” Nuclear Instruments and Methods in
Physics Research Section B: Beam Interactions with Materials and Atoms, vol.
269, no. 22, pp. 2676 – 2684, 2011.
[48] C. Agodi, G. Battistoni et al., “Charged particle’s flux measurement from
PMMA irradiated by 80 MeV/u carbon ion beam,” Physics in Medicine and
Biology, vol. 57, no. 18, p. 5667, 2012.
[49] L. Piersanti, F. Bellini et al., “Measurement of charged particle’s yields from
PMMA irradiated by a 220 MeV/u 12C beam,” Physics in Medicine and Biol-
ogy, vol. 59, no. 7, p. 1857, 2014.
[50] K. Gwosch, B. Hartmann et al., “Non-invasive monitoring of therapeutic carbon
ion beams in a homogeneous phantom by tracking of secondary ions,” Physics
in Medicine and Biology, vol. 58, no. 11, pp. 3755 – 3773, 2013.
118 BIBLIOGRAPHY
[51] Z. Abou-Haidar, C. Agodi et al., “Performance of upstream interaction region
detectors for the FIRST experiment at GSI,” Journal of Instrumentation, vol. 7,
p. P02006, 2012.
[52] R. Pleskac, Z. Abou-Haidar et al., “The FIRST experiment at GSI,” Nuclear
Instruments & Methods in Physics Research. Section A, Accelerators, Spec-
trometers, Detectors and Associated Equipment, vol. 678, pp. 130 – 138, 2012.
[53] L. Piersanti, C. Schuy et al., Development of a VME data acquisition
system, ser. GSI Report 2013-1. Darmstadt: GSI Helmholtzzentrum fur
Schwerionenforschung, 2013, p. 479. Available online: http://repository.gsi.de/
record/52370
[54] F. Bellini, T. Bohlen et al., “Extended calibration range for prompt photon
emission in ion beam irradiation,” Nuclear Instruments & Methods in Physics
Research Section A: Accelerators, Spectrometers, Detectors and Associated
Equipment, vol. 745, no. 0, pp. 114 – 118, 2014.
[55] A. Ferrari, P. Sala et al., “FLUKA: a multi-particle transport code,” CERN-
2005-10, INFN/TC–05/11, SLAC–R–773, 2005.
[56] G. Battistoni, S. Muraro et al., “The FLUKA code: Description and bench-
marking,” Proceedings of the Hadronic Shower Simulation Workshop 2006,
Fermilab 6 – 8 September 2006, AIP Conference Proceeding 896, pp. 31 –
49, 2007.
[57] C. Agodi, F. Bellini et al., “Study of the time and space distribution of β+
emitters from 80 MeV/u carbon ion beam irradiation on PMMA,” Nuclear
Instruments & Methods in Physics Research. Section B, Beam Interaction with
Materials and Atoms, vol. 283, no. 0, pp. 1 – 8, 2012.
[58] Y. Koba, H. Iwamoto et al., “Scintillation Efficiency of Inorganic Scintillators
for Intermediate-Energy Charged Particles,” Progress in NUCLEAR SCIENCE
and TECHNOLOGY, vol. 1, pp. 218 – 221, 2011.
[59] R. Brun and F. Rademakers, “ROOT: An object oriented data analysis frame-
work,” Nuclear Instruments and Methods in Physics Research Section A: Ac-
celerators, Spectrometers, Detectors and Associated Equipment, vol. A389, pp.
81 – 86, 1997.
BIBLIOGRAPHY 119
[60] M. Kramer, O. Jakel et al., “Treatment planning for heavy-ion radiotherapy:
physical beam model and dose optimization,” Physics in Medicine and
Biology, vol. 45, no. 11, p. 3299, 2000. Available online: http:
//stacks.iop.org/0031-9155/45/i=11/a=313
[61] R. E. Kalman, “A new approach to linear filtering and prediction problems,”
Transactions of the ASME. Series D, Journal of Basic Engineering, vol. 82, pp.
35 – 45, 1960.
[62] R. E. Kalman and R. S. Bucy, “New results in linear filtering and prediction
theory,” Transactions of the ASME. Series D, Journal of Basic Engineering,
vol. 83, pp. 95 – 107, 1961.
[63] E. Wolin and L. Ho, “Covariance matrices for track fitting with the Kalman
filter,” Nuclear Instruments and Methods in Physics Research A, vol. 329, pp.
493–500, 1993.