An increasing number of cancer patients is treated with proton beams or other light ion beams which allow to deliver dose precisely to the tumor. However, the depth dose distribution of these particles, which enables this precision, is sensitive to deviations from the treatment plan, as e.g. anatomical changes. Thus, to assure the quality of the treatment, a non-invasive in-vivo dose verification is highly desired. This monitoring of particle therapy relies on the detection of secondary radiation which is produced by interactions between the beam particles and the nuclei of the patient’s tissue. Up to now, the only clinically applied method for in-vivo dosimetry is Positron Emission Tomography which makes use of the beta+-activity produced during the irradiation (PT-PET). Since from a PT-PET measurement the applied dose cannot be directly deduced, the simulated distribution of beta+-emitting nuclei is used as a basis for the analysis of the measured PT-PET data. Therefore, the reliable modeling of the production rates and the spatial distribution of the beta+-emitters is required. PT-PET applied during instead of after the treatment is referred to as in-beam PET. A challenge concerning in-beam PET is the design of the PET camera, because a standard full-ring scanner is not feasible. For instance, a double-head PET camera is applicable, but low count rates and the limited solid angle coverage can compromise the image quality. For this reason, a detector system which provides a time resolution allowing the incorporation of time-of-flight information (TOF) into the iterative reconstruction algorithm is desired to improve the quality of the reconstructed images. Secondly, Prompt Gamma Imaging (PGI), a technique based on the detection of prompt gamma-rays, is currently pursued. Concerning the emissions of prompt gamma-rays during particle irradiation, experimental data is not sufficiently available, making simulations necessary. Compton cameras are based on the detection of incoherently scattered photons and are investigated with respect to PGI. Monte Carlo simulations serve for the optimization of the camera design and the evaluation of criteria for the selection of measured events. Thus, for in-beam PET and PGI dedicated detection systems and, moreover, profound knowledge about the corresponding radiation fields are required. Using various simulation codes, this thesis contributes to the modelling of the beta+-emitters and photons produced during particle irradiation, as well as to the evaluation and optimization of hardware for both techniques. Concerning the modeling of the production of the relevant beta+-emitters, the abilities of the Monte Carlo simulation code PHITS and of the deterministic, one-dimensional code HIBRAC were assessed. The Monte Carlo tool GEANT4 was applied for an additional comparison. For irradiations with protons, helium, lithium, and carbon, the depth-dependent yields of the simulated beta+-emitters were compared to experimental data. In general, PHITS underestimated the yields of the considered beta+-emitters in contrast to GEANT4 which provided acceptable values. HIBRAC was substantially extended to enable the modeling of the depth-dependent yields of specific nuclides. For proton beams and carbon ion beams HIBRAC can compete with GEANT4 for this application. Since HIBRAC is fast, compact, and easy to modify, it could be a basis for the simulations of the beta+-emitters in clinical application. PHITS was also applied to the modeling of prompt gamma-rays during proton irradiation following an experimental setup. From this study, it can be concluded that PHITS could be an alternative to GEANT4 in this context. Another aim was the optimization of Compton camera prototypes. GEANT4 simulations were carried out with the focus on detection probabilities and the rate of valid events. Based on the results, the feasibility of a Compton camera setup consisting of a CZT detector and an LSO or BGO detector was confirmed. Several recommendations concerning the design and arrangement of the Compton camera prototype were derived. Furthermore, several promising event selection strategies were evaluated. The GEANT4 simulations were validated by comparing simulated to measured energy depositions in the detector layers. This comparison also led to the reconsideration of the efficiency of the prototype. A further study evaluated if electron-positron pairs resulting from pair productions could be detected with the existing prototype in addition to Compton events. Regarding the efficiency and the achievable angular resolution, the successful application of the considered prototype as pair production camera to the monitoring of particle therapy is questionable. Finally, the application of a PET camera consisting of Resistive Plate Chambers (RPCs) providing a good time resolution to in-beam PET was discussed. A scintillator-based PET camera based on a commercially available scanner was used as reference. This evaluation included simulations of the detector response, the image reconstructions using various procedures, and the analysis of image quality. Realistic activity distributions based on real treatment plans for carbon ion therapy were used. The low efficiency of the RPC-based PET camera led to images of poor quality. Neither visually nor with the semi-automatic tool YaPET a reliable detectability of range deviations was possible. The incorporation of TOF into the iterative reconstruction algorithm was especially advantageous for the considered RPC-based PET camera in terms of convergence and artifacts. The application of the real-time capable back projection method Direct TOF for the RPCbased PET camera resulted in an image quality comparable to the one achieved with the iterative algorihms. In total, this study does not indicate the further investigation of RPC-based PET cameras with similar efficiency for in-beam PET application. To sum up, simulation studies were performed aimed at the progress of in-vivo dosimetry. Regarding the modeling of the beta+-emitter production and prompt gamma-ray emissions, different simulation codes were evaluated. HIBRAC could be a basis for clinical PT-PET simulations, however, a detailed validation of the underlying cross section models is required. Several recommendations for the optimization of a Compton Camera prototype resulted from systematic variations of the setup. Nevertheless, the definite evaluation of the feasibility of a Compton camera for PGI can only be performed by further experiments. For PT-PET, the efficiency of the detector system is the crucial factor. Due to the obtained results for the considered RPC-based PET camera, the focus should be kept to scintillator-based PET cameras for this purpose.

Die Protonentherapie zeichnet sich durch eine konformale und fokussierte Tumorbestrahlung aus, die es ermöglicht, gesundes Gewebe besser zu schonen als bei der konventionellen Strahlentherapie. Dieses Potential wird jedoch durch Unsicherheiten bei der Vorhersage der Protonenreichweite im Gewebe oder durch anatomische Veränderungen über den Verlauf der Therapie eingeschränkt. In der vorliegenden Arbeit wurde daher der klinische Nutzen eines Reichweiteverifikationssystems auf Grundlage von Prompt-Gamma-Imaging (PGI) zur Behandlungskontrolle untersucht. Dafür wurden Messungen mit einem PGI-System während Prostata- und Kopf-Hals-Tumor-Bestrahlungen durchgeführt und retrospektiv ausgewertet. Einerseits konnte dabei mittels PGI die Genauigkeit verschiedener Methoden zur Reichweitevorhersage überprüft werden. Es zeigte sich, dass die 2019 klinisch eingeführte Methode zur Reichweitevorhersage (DirectSPR) nicht von der mit PGI gemessenen Protonenreichweite in Prostata-Tumor-Bestrahlungen abweicht, wodurch die Reduktion der auf DirectSPR basierenden Reichweiteunsicherheiten unabhängig bestätigt werden konnte. Andererseits konnte die Detektionsfähigkeit von PGI bei der Erkennung relevanter und nicht relevanter anatomischer Veränderungen in applizierten Bestrahlungsfeldern nachgewiesen werden. Insbesondere wurde für die feldweise Klassifizierung der Prostata-Bestrahlungen eine Sensitivität und Spezifität von 74% bzw. 79% festgestellt. Damit konnte in dieser Dissertation erstmals systematisch das klinische Anwendungspotential eines Systems zur PGI-Reichweiteverifikation gezeigt werden. Als zusätzliche Untersuchung wurde in einer Kollaboration mit dem Massachusetts General Hospital zum ersten Mal ein Vergleich zwischen zwei verschiedenen, auf prompter Gammastrahlung basierenden Systemen zur Reichweiteverifikation durchgeführt. Dazu wurde ein standardisiertes Studienprotokoll etabliert, welches die Vergleichbarkeit und die klinische Implementierung von Reichweiteverifikationssystemen generell unterstützen könnte.:1 Einleitung 2 Strahlentherapie mit Protonen 2.1 Physikalische Grundlagen der Protonentherapie 2.2 Behandlungsablauf in der Protonentherapie 2.2.1 Bildgebung zur Therapieplanung 2.2.2 Bestrahlungsplanung 2.2.3 Strahlapplikation 2.3 Genauigkeit in der Protonentherapie 2.3.1 Ursachen für Behandlungs- und Reichweiteunsicherheiten 2.3.2 Aktueller Stand der Behandlungs- und Reichweiteverifikation 3 Methodik der Reichweiteverifikation mittels Prompt-Gamma-Bildgebung (PGI) 3.1 Funktionsprinzip der PGI-Schlitzkamera 3.2 Datenaufnahme und -verarbeitung 3.2.1 Detektoraufbau und Signalaufnahme 3.2.2 PGI-Simulation und Bestimmung der Reichweiteabweichung 3.3 Charakterisierung des PGI-Prototyps 3.3.1 Kalibrierung des Systems 3.3.2 Positionierungspräzision 3.4 Überblick zur PRIMA-Studie 3.5 Experimentelle Studien zur PGI-Simulationsgenauigkeit 3.5.1 Abhängigkeit vom PGI-Sichtfeld und der Protonenenergie 3.5.2 Validierung der erweiterten Simulationssoftware 3.5.3 Abhängigkeit von der Tumorentität 3.5.4 Schlussfolgerungen 4 Validierung der CT-basierten Reichweitevorhersage mittels PGI 4.1 Konzept der Validierung 4.2 Gesamtabschätzung der Validierungsunsicherheit 4.3 Ergebnisse der Validierung 4.4 Diskussion 5 Detektionsfähigkeit anatomischer Veränderungen mittels PGI 5.1 Prinzipieller Aufbau der Studie 5.2 Grundwahrheit auf Basis von CT- und Dosisinformationen 5.2.1 Manuelle Klassifizierung 5.2.2 Klassifizierung auf Grundlage von integrierten Tiefendosisprofilen 5.2.3 Ergebnis der Etablierung einer CT-basierten Grundwahrheit 5.3 Etablierung einer Klassifikation auf Basis von PGI-Daten 5.3.1 Verarbeitung der PGI-Daten mittels Cluster-Algorithmus 5.3.2 Definition von spot- oder clusterbasierten Klassifikationsmodellen 5.4 Ergebnisse der PGI-Detektionsfähigkeit 5.4.1 Auswertung für Patienten mit Prostata-Tumor 5.4.2 Auswertung für Patienten mit Tumoren im Kopf-Hals-Bereich 5.5 Diskussion 6 Genauigkeit zweier Reichweiteverifikationsmethoden – bizentrischer Vergleich 6.1 Material und Methoden 6.1.1 Bildgebung 6.1.2 Bestrahlungsplanung 6.1.3 Durchführung und Auswertung 6.2 Ergebnisse 6.3 Diskussion 7 Zusammenfassung 8 Summary / Proton therapy is a conformal and focused irradiation of the tumor, which allows for a better sparing of healthy tissue than with conventional radiotherapy. However, this potential is limited by uncertainties from the proton range prediction in the patient or anatomical changes over the course of the treatment. Therefore, in this work, the clinical benefit of a range verification system based on the prompt-gamma-imaging (PGI) method for treatment verification was investigated. For this purpose, measurements were carried out with a PGI system during prostate and head and neck cancer irradiations and evaluated retrospectively. On the one hand, PGI was used to review the accuracy of several range prediction methods. The results showed that a specific method for range prediction (DirectSPR), which was clinically introduced in 2019, does not deviate from the PGI-measured proton range in prostate cancer irradiations. This means that the reduction of the range uncertainties with DirectSPR could be independently confirmed. On the other hand, the detection capability of PGI in identifying relevant and non-relevant anatomical changes in delivered treatment fields was demonstrated. In particular, for the fieldwise classification of prostate irradiations a sensitivity and specificity of 74% and 79% was determined, respectively. Thus, the clinical potential of a PGI range verification system was for the first time systematically demonstrated in this thesis. Furthermore, in a collaboration with the Massachusetts General Hospital a first-time comparison of two different range verification systems based on prompt gamma radiation was conducted. Therefore, a standardized study protocol was established, which could generally foster the comparability and clinical implementation of range verification systems.:1 Einleitung 2 Strahlentherapie mit Protonen 2.1 Physikalische Grundlagen der Protonentherapie 2.2 Behandlungsablauf in der Protonentherapie 2.2.1 Bildgebung zur Therapieplanung 2.2.2 Bestrahlungsplanung 2.2.3 Strahlapplikation 2.3 Genauigkeit in der Protonentherapie 2.3.1 Ursachen für Behandlungs- und Reichweiteunsicherheiten 2.3.2 Aktueller Stand der Behandlungs- und Reichweiteverifikation 3 Methodik der Reichweiteverifikation mittels Prompt-Gamma-Bildgebung (PGI) 3.1 Funktionsprinzip der PGI-Schlitzkamera 3.2 Datenaufnahme und -verarbeitung 3.2.1 Detektoraufbau und Signalaufnahme 3.2.2 PGI-Simulation und Bestimmung der Reichweiteabweichung 3.3 Charakterisierung des PGI-Prototyps 3.3.1 Kalibrierung des Systems 3.3.2 Positionierungspräzision 3.4 Überblick zur PRIMA-Studie 3.5 Experimentelle Studien zur PGI-Simulationsgenauigkeit 3.5.1 Abhängigkeit vom PGI-Sichtfeld und der Protonenenergie 3.5.2 Validierung der erweiterten Simulationssoftware 3.5.3 Abhängigkeit von der Tumorentität 3.5.4 Schlussfolgerungen 4 Validierung der CT-basierten Reichweitevorhersage mittels PGI 4.1 Konzept der Validierung 4.2 Gesamtabschätzung der Validierungsunsicherheit 4.3 Ergebnisse der Validierung 4.4 Diskussion 5 Detektionsfähigkeit anatomischer Veränderungen mittels PGI 5.1 Prinzipieller Aufbau der Studie 5.2 Grundwahrheit auf Basis von CT- und Dosisinformationen 5.2.1 Manuelle Klassifizierung 5.2.2 Klassifizierung auf Grundlage von integrierten Tiefendosisprofilen 5.2.3 Ergebnis der Etablierung einer CT-basierten Grundwahrheit 5.3 Etablierung einer Klassifikation auf Basis von PGI-Daten 5.3.1 Verarbeitung der PGI-Daten mittels Cluster-Algorithmus 5.3.2 Definition von spot- oder clusterbasierten Klassifikationsmodellen 5.4 Ergebnisse der PGI-Detektionsfähigkeit 5.4.1 Auswertung für Patienten mit Prostata-Tumor 5.4.2 Auswertung für Patienten mit Tumoren im Kopf-Hals-Bereich 5.5 Diskussion 6 Genauigkeit zweier Reichweiteverifikationsmethoden – bizentrischer Vergleich 6.1 Material und Methoden 6.1.1 Bildgebung 6.1.2 Bestrahlungsplanung 6.1.3 Durchführung und Auswertung 6.2 Ergebnisse 6.3 Diskussion 7 Zusammenfassung 8 Summary

An increasing number of cancer patients is treated with proton beams or other light ion beams which allow to deliver dose precisely to the tumor. However, the depth dose distribution of these particles, which enables this precision, is sensitive to deviations from the treatment plan, as e.g. anatomical changes. Thus, to assure the quality of the treatment, a non-invasive in-vivo dose verification is highly desired. This monitoring of particle therapy relies on the detection of secondary radiation which is produced by interactions between the beam particles and the nuclei of the patient’s tissue. Up to now, the only clinically applied method for in-vivo dosimetry is Positron Emission Tomography which makes use of the beta+-activity produced during the irradiation (PT-PET). Since from a PT-PET measurement the applied dose cannot be directly deduced, the simulated distribution of beta+-emitting nuclei is used as a basis for the analysis of the measured PT-PET data. Therefore, the reliable modeling of the production rates and the spatial distribution of the beta+-emitters is required. PT-PET applied during instead of after the treatment is referred to as in-beam PET. A challenge concerning in-beam PET is the design of the PET camera, because a standard full-ring scanner is not feasible. For instance, a double-head PET camera is applicable, but low count rates and the limited solid angle coverage can compromise the image quality. For this reason, a detector system which provides a time resolution allowing the incorporation of time-of-flight information (TOF) into the iterative reconstruction algorithm is desired to improve the quality of the reconstructed images. Secondly, Prompt Gamma Imaging (PGI), a technique based on the detection of prompt gamma-rays, is currently pursued. Concerning the emissions of prompt gamma-rays during particle irradiation, experimental data is not sufficiently available, making simulations necessary. Compton cameras are based on the detection of incoherently scattered photons and are investigated with respect to PGI. Monte Carlo simulations serve for the optimization of the camera design and the evaluation of criteria for the selection of measured events. Thus, for in-beam PET and PGI dedicated detection systems and, moreover, profound knowledge about the corresponding radiation fields are required. Using various simulation codes, this thesis contributes to the modelling of the beta+-emitters and photons produced during particle irradiation, as well as to the evaluation and optimization of hardware for both techniques. Concerning the modeling of the production of the relevant beta+-emitters, the abilities of the Monte Carlo simulation code PHITS and of the deterministic, one-dimensional code HIBRAC were assessed. The Monte Carlo tool GEANT4 was applied for an additional comparison. For irradiations with protons, helium, lithium, and carbon, the depth-dependent yields of the simulated beta+-emitters were compared to experimental data. In general, PHITS underestimated the yields of the considered beta+-emitters in contrast to GEANT4 which provided acceptable values. HIBRAC was substantially extended to enable the modeling of the depth-dependent yields of specific nuclides. For proton beams and carbon ion beams HIBRAC can compete with GEANT4 for this application. Since HIBRAC is fast, compact, and easy to modify, it could be a basis for the simulations of the beta+-emitters in clinical application. PHITS was also applied to the modeling of prompt gamma-rays during proton irradiation following an experimental setup. From this study, it can be concluded that PHITS could be an alternative to GEANT4 in this context. Another aim was the optimization of Compton camera prototypes. GEANT4 simulations were carried out with the focus on detection probabilities and the rate of valid events. Based on the results, the feasibility of a Compton camera setup consisting of a CZT detector and an LSO or BGO detector was confirmed. Several recommendations concerning the design and arrangement of the Compton camera prototype were derived. Furthermore, several promising event selection strategies were evaluated. The GEANT4 simulations were validated by comparing simulated to measured energy depositions in the detector layers. This comparison also led to the reconsideration of the efficiency of the prototype. A further study evaluated if electron-positron pairs resulting from pair productions could be detected with the existing prototype in addition to Compton events. Regarding the efficiency and the achievable angular resolution, the successful application of the considered prototype as pair production camera to the monitoring of particle therapy is questionable. Finally, the application of a PET camera consisting of Resistive Plate Chambers (RPCs) providing a good time resolution to in-beam PET was discussed. A scintillator-based PET camera based on a commercially available scanner was used as reference. This evaluation included simulations of the detector response, the image reconstructions using various procedures, and the analysis of image quality. Realistic activity distributions based on real treatment plans for carbon ion therapy were used. The low efficiency of the RPC-based PET camera led to images of poor quality. Neither visually nor with the semi-automatic tool YaPET a reliable detectability of range deviations was possible. The incorporation of TOF into the iterative reconstruction algorithm was especially advantageous for the considered RPC-based PET camera in terms of convergence and artifacts. The application of the real-time capable back projection method Direct TOF for the RPCbased PET camera resulted in an image quality comparable to the one achieved with the iterative algorihms. In total, this study does not indicate the further investigation of RPC-based PET cameras with similar efficiency for in-beam PET application. To sum up, simulation studies were performed aimed at the progress of in-vivo dosimetry. Regarding the modeling of the beta+-emitter production and prompt gamma-ray emissions, different simulation codes were evaluated. HIBRAC could be a basis for clinical PT-PET simulations, however, a detailed validation of the underlying cross section models is required. Several recommendations for the optimization of a Compton Camera prototype resulted from systematic variations of the setup. Nevertheless, the definite evaluation of the feasibility of a Compton camera for PGI can only be performed by further experiments. For PT-PET, the efficiency of the detector system is the crucial factor. Due to the obtained results for the considered RPC-based PET camera, the focus should be kept to scintillator-based PET cameras for this purpose.

Prompt-gamma imaging (PGI) was proposed in the 2000s as a promising in vivo range-verification method to maintain the physical advantage of proton beams by reducing unwanted range-uncertainties. Recently, PGI with a slit camera has been successfully implemented in clinical application. Despite its high accuracy and sensitivity to range deviation being shown in several studies, the clinical benefits of PGI have not yet been investigated. Hence, to fully exploit the advantages of PGI, this thesis aims to investigate the feasibility of PGI-based range verification for the automatic classification of treatment deviations and differentiation of relevant from non-relevant changes in the treatment of head-and-neck (H&N) tumors. In the first part of this thesis, the four most common types of treatment deviations in proton therapy (PT) were investigated regarding their PGI signature and by considering clinically relevant and non-relevant scenarios. A heuristic decision tree (DT) model was iteratively developed. To gain understanding of the specific signature of the error sources, different levels of geometrical complexities were explored, from simple to complex. At the simplest level, a phantom with homogeneous density was used to distinguish range-prediction and setup errors. Next, in the intermediate complexity level, a phantom with heterogeneous density was used to inspect the additional error scenarios of anatomical changes. Finally, real patient CT scans were used to investigate the relevance of changes based on clinical constraints. In the final model, a five-step filtering approach was used during pre-processing to select reliable pencil-beam-scanning spots for range verification. In this study, five features extracted from the filtered PGI data were used to classify the treatment deviation. The model is able distinguish four introduced scenarios into six classes as follows: (1) overestimation of range prediction, (2) underestimation of range prediction, (3) setup error with larger air gap, (4) setup error with smaller air gap, (5) anatomical change, and (6) non-relevant change. To ensure the application was effective, independent patient CT datasets were used to test the model. The results yielded an excellent performance of the DT classifier, with high accuracy, sensitivity, and specificity of 96%, 100%, and 85.7%, respectively. According to these findings, this model can sensitively detect treatment deviations in PT based on simulated PGI data. In the second part of this work, an alternative approach based on machine learning (ML) was taken to automatically classify the error sources. In the first stage, the two approaches were compared, using the same features as well as the same training and test datasets. The results show that the ML approach was slightly better than the heuristic DT approach in terms of accuracy. However, the performance of both approaches was excellent for the individual scenarios. Thus, these results confirm that the PGI-based data classification with five features can be applied to detect individual sources of treatment deviation in PT. In the second stage, there was an investigation of more complex and more realistic combinations of error scenarios, which was out of the scope of the DT approach. The results demonstrated that the performance of the ML-based classifiers declined in general. Furthermore, the additional features of the PG shift did not substantially improve the performance of the classifiers. As a consequence, these findings mark important issues for future research. Potentially, usage of the spatial information from the spot-based PGI data and more complex techniques such as deep learning may improve the performance of classifiers with respect to scenarios with multiple error sources. However, regardless of this, it is recommended that these findings be confirmed and validated in simulations under measurement-like conditions or with real PG measurements of H&N patients themselves. Moreover, this classification model could eventually be tested with other body sites and entities in order to assess its compatibility and adaptation requirements. In summary, this study yielded promising results regarding the automatic classification of treatment-deviation sources and the differentiation of relevant and non-relevant changes in H&N-tumor treatment in PT with PGI data. This simulation study marks an important step towards fully automated PGI-based proton-range verification, which could contribute to closing the treatment-workflow loop of adaptive therapy by supporting clinical decision-making and, ultimately, improving clinical PT.:1 Introduction 2 Background 2.1 Proton therapy 2.1.1 Rationale for proton therapy 2.1.2 Uncertainties and their mitigation 2.2 In vivo range-verification techniques 2.2.1 Range probing 2.2.2 Proton tomography 2.2.3 Magnetic resonance imaging 2.2.4 Ionoacoustic detection 2.2.5 Treatment-activated positron-emission tomography imaging 2.2.6 Prompt-gamma based detection 3 Prompt-gamma imaging with a knife-edged slit camera 3.1 Current state-of-the-art 3.2 Prompt-gamma camera system 3.3 Data acquisition and analysis 4 Error-source classification using heuristic decision tree approach 4.1 Study design 4.1.1 Case selection 4.1.2 Investigated scenarios 4.1.3 Prompt-gamma simulation and range shift determination 4.2 Development of the model 4.2.1 First-generation model 4.2.2 Second-generation model 4.2.3 Third-generation model 4.3 Model testing 4.4 Discussion: decision-tree model 5 Error-source classification using a machine-learning approach 5.1 Machine learning for classification 5.1.1 Support-vector-machine algorithm 5.1.2 Ensemble algorithm – random forest 5.1.3 Logistic-regression algorithm 5.2 Study design 5.2.1 Case selection 5.2.2 Feature selection 5.3 Model generation 5.4 Model testing 5.5 Discussion 6 Summary/ Zusammenfassung Bibliography Appendix List of Figures List of Tables List of Abbreviations