Investigations on multilateration of ionoacoustic signals for localisation of the bragg peak in pre-clinical research

Kalunga, Ronaldo (2025): Investigations on multilateration of ionoacoustic signals for localisation of the bragg peak in pre-clinical research. Dissertation, LMU München: Medizinische Fakultät

Abstract

Radiation therapy is one of the most typically used treatments in cancer care, with around 60% of patients undergoing this form of treatment. While X-rays and gamma rays (photon therapy) are the standard approach, proton therapy has emerged as a valuable alternative. Proton therapy is renowned for its ability to provide a more conformal dose delivery. Proton therapy’s superiority over photon therapy is due to protons depositing their maximum energy directly within the tumour while sparing surrounding healthy tissues. However, proton therapy is highly sensitive to range uncertainties. Range uncertainties in proton therapy arise primarily because we cannot precisely determine where the proton beam will stop, leading to the risk of overshooting or undershooting the target. Thus, there is a need for in vivo range verification methods to reduce range uncertainties. The two methods nearing routine clinical use are positron emission tomography (PET) and prompt gamma imaging (PGI). Range verification relies on monitoring nuclear reaction products along proton beams for these methods. However, PET and PGI methods do not directly correlate the measurable signal, beam range, or Bragg peak (BP) position. Additionally, their equipment is bulky and not cost-effective. Therefore, the research conducted during this work proposes a range verification method that is both cost-effective and establishes a direct correlation between the proton beam and ionoacoustic (IA) signals. At present, only two commercial platforms support small animal photon radiotherapy, though their imaging systems can be adapted for research beamlines. Proton therapy offers distinct advantages over photon therapy, which led to the development of the Small Animal Proton Irradiator for Research in Molecular Image-guided Radiation-Oncology (SIRMIO) project. It was led by Prof. Dr. Katia Parodi at Ludwig Maximilians-Universit¨at (LMU) Munich and funded by the European Research Council (ERC) under grant agreement 725539. SIRMIO aimed to create the first portable, imageguided research platform for small animal proton therapy. As part of this effort, different range verification methods are investigated. One of these methods is the one studied in this thesis, which is based on localising the BP using IA signals. The research presented here investigates BP localisation using IA signals, aiming to determine the BP position in both two-dimensional (2D) and three-dimensional (3D) space. The localisation was performed in homogenous and heterogenous media via time-of-flight (ToF) estimation from different sensor spatial locations. The localisation of the BP was assessed using a technique called multilateration. The initial studies were performed in-silico, using ideal point sources that emulated the BP position and evaluated the robustness of two numerical optimisation algorithms: Nelder-Mead Simplex and Levenberg Marquardt. Secondly, the robustness of the multilateration technique was assessed for two localisation methods: time-of-arrival (TOA) and time-difference-of-arrival (TDOA). By modelling random and systematic uncertainties in the geometrical ToF, the robustness of both TOA and TDOA was evaluated. Random uncertainties aimed to model the speed of sound variations, inaccurate knowledge of the sensor spatial location and errors on the ToF. On the other hand, the objective of modelling systematic uncertainties was to simulate the inaccurate knowledge of the measurement starting time from a proton beam accelerator. After fully understanding the numerical optimisation methods and the impact of uncertainties on TOA and TDOA, the localisation focus was addressed to a realistic simulation case using a pre-clinical beam with an energy of 20 MeV. The multilateration of the BP position was performed with a sensor network of 843 ideal point sensors arranged in a semi-circular configuration with a diameter of 60 mm. Similarly, the impact of different ToF extraction methods on BP localisation was evaluated. Moreover, the studies were further expanded to investigate the impact of the number of sensors on the ToF estimation and, consequently, their impact on the accuracy of the BP localisation. Experimental campaigns were conducted to benchmark the localisation of the BP using pre-knowledge gained from the simulation studies. These experimental studies retrieved the BP position in the Tandem accelerator with two different beam energies (20 and 22 MeV). The first experimental campaign aimed to localise the BP using 3 transducers. Furthermore, two different techniques were implemented to localise the spatial location of the transducers. The second experimental campaign aimed to localise the BP using 5 transducers. Moreover, the spatial locations of the transducers were estimated experimental using a single approach based on the measurement performed with an optical tracking system. For the SIRMIO case, a dedicated localisation setup with a 50 MeV beam energy was considered. This setup aimed to localise the BP under various conditions, including different proton beam time profiles, beam spatial locations, and numbers of sensors. The first step involved studying the error in ToF as a function of the proton time profiles and then assessing multilateration accuracy based on thesame proton time profiles. After identifying the optimal proton time profile, the BP was localised by keeping the proton time profile constant while varying the number of sensors. For the numerical methods, the Levenberg-Marquardt method demonstrated greater robustness compared to the Nelder-Mead Simplex method, with failure rates (FR) of 0.22% and 0% when localising the emulated BP positions with TOA and 1.12% and 4.85% when localising the source with TDOA, respectively. Considering ideal point sources, both localisation methods were equivalent in 2D. A mean error in localisation of 7.4×10^−4 mm and 7.8×10^−4 mm for TOA and TDOA was obtained. In 3D, the localisation error varied from 7.8×10^−4 mm and 1.0×10^−3 mm for TOA and TDOA. The speed of sound varies in vivo depending on the tissue type, which is expected to reduce the BP localisation accuracy. With a conservative assumption of a 5% error in the average speed of sound along the acoustic path (modelled by random uncertainties), it was observed that the localisation error after multilateration increased by around 2 mm for the examined geometry. The lowest error on the ToF estimation is obtained for the maximum-envelope extraction method when considering IA signals. Therefore, through optimal sensor positioning to minimise ToF errors, the BP could be localised in-silico with an accuracy exceeding 90 μm (equivalent to a 2% error). The BP was localised for the first experimental setup with errors ranging from 0.43 mm to 0.48 mm, depending on the sensor arrangement. The localisation was performed with a total dose of 1.69 Gy with a single shot. In the second experimental setup, the localisation was performed with 50 IA signals and a total dose of 29 Gy, achieving a localisation error of 1 mm. For both setups, the primary sources of localisation errors were inaccuracies in sensor positioning and low signal-to-noise ratio (SNR) due to the weak and directional nature of the IA emissions. The studies conducted for the SIRMIO beamline demonstrated that the proton time profile significantly impacts the ToF estimation, influencing the accuracy of BP localisation. The optimal localisation accuracy was achieved with proton time profiles ranging from 1 μs to 4 μs. In this setup, the BP was localised for different beam offsets along the x,y, and z axes. When applying offsets along the beam axis (x-axis), the maximum error was found to be 0.48 mm. Conversely, a maximum error of 1.23 mm was obtained for a transverse beam offset (z-axis). In conclusion, this work introduces a range verification method using IA signals within the framework of the SIRMIO project. Additionally, further discussions explore the potential for transitioning the studies presented in this thesis toward real-time range verification applications.

• Dokumententyp: Dissertationen (Dissertation, LMU München)
• Keywords: ionoacoustics, thermoacoustics, range verification, proton therapy
• Themengebiete: 600 Technik, Medizin, angewandte Wissenschaften ; 600 Technik, Medizin, angewandte Wissenschaften > 610 Medizin und Gesundheit
• Fakultäten: Medizinische Fakultät
• Sprache der Hochschulschrift: Englisch
• Datum der mündlichen Prüfung: 30. April 2025
• 1. Bericht­erstatter:in: Parodi, Katia