French
English
Abstract:
Monitoring the ion range is of the utmost importance to fully exploit the ballistic properties of hadron
therapy. Several nuclear imaging techniques are being studied around the world, including prompt
gamma (PG) detection techniques to address this. In this field, Monte Carlo simulation tools are
essential for designing, studying performance, and feeding artificial intelligence-based reconstruction
algorithms. Among these simulation tools, GATE is a reference tool in which a module has been
developed to accelerate the calculation of PG emission in proton therapy. The aim of the thesis is to
extend this module to all ions of interest in hadron therapy and to improve its accuracy in order to be
compatible with the energy and time resolutions of the best current and future gamma ray detectors.
The module will ultimately be tested by comparing simulated data with experimental data acquired
using the two PG detection techniques with which the thesis supervision team is partnered: PGTI and
PGEI techniques.
Scientific context
Uncertainties regarding the path of ions remain a major challenge in hadron therapy. They require
large margins to be taken around tumors to ensure that the entire tumor volume is irradiated, and
they prevent the use of irradiation fields that may be optimal from a dosimetric point of view but are
too risky if organs at risk are located downstream of the tumor. To take full advantage of the ballistic
properties of ions, since the 1990s, numerous research teams have been developing techniques to
monitor the path of ions, based mainly on the detection of radiation generated during nuclear
reactions undergone by a fraction of the incident ions: positron emission tomography, which exploits
the production of beta+ emitting nuclei during these nuclear reactions, and prompt gamma ray (PG)
detection techniques [Krimmer 2018].
In general, Monte Carlo simulation tools are essential for designing these detection systems and
studying in detail the performance that can be expected depending on the type of treatment or the
type of accelerator used to deliver the treatment. Furthermore, in the case of PG systems, lightweight
systems are increasingly being studied because they can be more easily integrated into the clinical
environment while offering performance compatible with the needs of clinicians. However, the
information provided by these systems on the path of the ions is indirect in the sense that the
detectors do not directly measure a spatial distribution of PG emission points but rather the
characteristics of the PGs related to the path of the ions: the energy spectrum of the PGs (Prompt
Gamma Spectroscopy) and the time of flight (TOF) corresponds to the time between the arrival of the
ions at the patient and the detection of a PG. To efficiently link these indirect quantities to the ion
trajectory, the use of artificial intelligence is becoming mandatory. This is the case, for example, with
the PGTI (Prompt Gamma Time Imaging) technique, which aims to reconstruct PG emission points
solely from a very precise TOF measurement on the scale of a few picoseconds. As this
reconstruction is difficult to perform using standard image reconstruction techniques, the use of
reconstruction algorithms based on artificial intelligence is currently being considered. For all these
reasons, it is important to develop Monte Carlo simulation tools that are accurate and fast.
For the past ten years, CREATIS and IP2I have been collaborating on a variance reduction method
that is particularly well suited to the emission of PG rays, which is a relatively rare event. This
involves the track length estimator (TLE) technique, initially proposed in the 1980s for evaluating
photon dose using the Kerma approximation. The calculation of the PG source using the track length
estimator (vpgTLE), which produces a 3D map of the PG energy yield (without temporal information),
was first proposed in 2015 by the CREATIS-IP2I collaboration and implemented in the MC GATE
platform version 9 [Kanawati2015]. A gain of approximately 1000 in relative uncertainty was reported
for a proton therapy treatment plan for a patient. In 2024, this GATE vpgTLE module was extended
to PG emission time as part of Oreste Allegrini's doctoral work [Létang2024].
Recently, two successive Master's 2 internships enabled the teams to make two major advances: (i)
porting the vpgTLE module to the latest Python-based version of GATE, namely version 10 released
in the fall of 2024, and (ii) integrating the energy and time distribution of PGs from secondary
inelastic neutron nuclear processes.
References
- [Kanawati2015] El Kanawati, W., et al. (2015). Monte Carlo simulation of prompt γ -ray emission in
proton therapy using a specific track length estimator. Physics in Medicine and Biology, 60(20), 8067.
https://doi.org/10.1088/0031-9155/60/20/8067 - [Krimmer2018] Krimmer, J., et al. (2018). Prompt-gamma monitoring in hadrontherapy: A review.
Nuclear Instruments and Methods in Physics Research Section A: Accelerators, Spectrometers,
Detectors and Associated Equipment. https://doi.org/10.1016/j.nima.2017.07.063 - [Létang2024] Létang, J. M., Allegrini, O., & Testa, E. (2024). Prompt-gamma track-length estimator
with time tagging from proton tracking. Physics in Medicine & Biology, 69(11), 115052.
https://doi.org/10.1088/1361-6560/AD4A01
Objectives of the thesis
The objective of the thesis is to continue developing this vpgTLE module to make it more accurate,
robust, and flexible, and to become the benchmark simulation tool for PG device simulations.
In terms of accuracy, certain assumptions need to be reconsidered, in particular the independence of
the energy and time distributions of GPs, which penalize accuracy on certain delayed GP lines with a
time resolution of a few hundred picoseconds. While this resolution is already barely sufficient given
the best current resolutions of gamma-ray detectors (on the order of a few hundred ps), it needs to
be significantly improved in order to anticipate the progress of future detectors. The objective of the
thesis is to tackle these issues of the vpgTLE module to make it more accurate, robust, and flexible,
i.e. the benchmark simulation tool for PG device simulations.
Furthermore, the module only works with proton beams. It is therefore important to extend the
module to heavier ions (e.g., carbon, helium, oxygen) to cover all ions currently considered in hadron
therapy. This will certainly involve revising the format of the PG database (which serves as input data
for the module) to optimize the size of the files associated with each patient element and each
primary or secondary projectile. In addition, this PG database, which must be rebuilt offline each time
the Geant4 physics is updated, is currently generated using an efficient low-level Geant4 executable,
but its configuration remains delicate. An alternative to this executable must therefore be developed
within GATE.
All these developments require, above all, expertise in advanced numerical methods in order to
develop a high-performance calculation tool open to the community. It is also important to have an
excellent understanding of detection systems and clinical irradiation conditions in order to best
determine the specifications for the vpgTLE module.
Methodology implemented
The main challenge associated with this thesis topic is the drastic optimization of data storage
methods in order to extend the vpgTLE module to all ions currently considered in hadron therapy and
improve the module's accuracy in terms of temporal resolution. More specifically, this will involve
optimizing the format of the PG database and the way in which PG emission is calculated in each
voxel, while maintaining the correlation between energy and PG emission time, at least for lines
associated with relatively long de-excitation times.
In the last part of the thesis, the module will be tested by comparing simulated data with experimental
data acquired using the two PG detection techniques with which the thesis supervision team is
partnered: PGTI and PGEI techniques. Finally, the module will be used to study the expected
performance of these techniques in clinical conditions. Particular care will be taken to document the
code for its integration into GATE and its use, particularly in the context of the PGTI project, which
plans to use it to train reconstruction algorithms based on artificial intelligence.
Our collaboration with MedAustron (Wiener Neustadt, Austria), which already delivers carbon ion
therapy, will provide essential expertise and data to validate the new heavy ion models.
Expected impact
This project aims to establish vpgTLE as the open-source reference simulation tool for PG detection
in hadron therapy, combining high calculation speed with physical accuracy. By extending the module
to heavy ions and generating massive, temporally accurate datasets, it will drive the development of
AI-based reconstruction and the next-generation of time-of-flight detectors. Ultimately, these
advances will reduce range uncertainties, enabling safer and more optimized clinical treatments.
Key-words: Variance Reduction techniques, Monte Carlo Simulation, Medical Physics
Profile:
- Engineering or MSc degree in physics, applied mathematics, computer science or related disciplines.
- Experience in image processing and programming in Python/C++ are required.
- Language: English required, French optional
Period: 3 years
Location: Centre Léon Bérard (Lyon), cancer care and research center + Collaboration with Étienne Testa from IP2I (Villeurbanne) a nuclear physics laboratory