A new prompt gamma spectroscopy-based approach for range verification in proton therapy

Proton therapy is a well-established technology in radiotherapy, whose benefits stem from both physical and biological properties. Ions deposit the maximum dose, i.e. the ratio between the energy absorbed by the tissue and its mass (𝐺𝑦 = 𝐽/𝑘 𝑔), in a localized region close to the end of their range (called the Bragg Peak BP). The combination of the favorable depth-dose profile with advanced delivery techniques translates into a high dose conformality in the tumor, as well as into a superior sparing of normal tissue compared to conventional radiotherapy with photons. Today, there are 107 proton therapy and 14 carbon ion centers operating worldwide, and many new ones are under construction. In Italy, the Trento proton therapy center and the proton and carbon ion center - Centro Nazionale per l’Adronterapia Oncologica - CANO in Pavia are already operational, while a third one is under construction at the Istituto Europeo Oncologico (IEO) in Milan and will be in operation in 2023. Although clinical results have been encouraging, numerous treatment uncertainties remain major obstacles to the full exploitation of proton therapy. One of the crucial challenges is monitoring the dose delivered during the treatment, both in terms of absolute value and spatial distribution inside the body. Ideally, the actual beam range in the patient should be equal to the value prescribed by the Treatment Planning System (TPS). However, there are sizeable uncertainties at the time of irradiation due to anatomical modifications, patient alignment, beam delivery, and dose calculation. Treatment plans are optimized to be conformal in terms of target coverage, healthy tissue spearing, and robust towards uncertainties. For this reason, the irradiation target is defined as a geometrical volume (Planning Target Volume PTV) corresponding to the physical tumor volume, to which safety margins of a few millimeters are added isotropically. Range errors determine the selection of the safety margins applied to the tumor volume, whose values depend on clinical protocols as well as on the treated area. For example, the Massachusetts General Hospital (MGH) prescribes safety margins equal to 3.5% of the nominal range +1 mm, while the University of Florida proton therapy center considers 2.5% of the nominal range + 1.5 mm. Decreasing the range uncertainties would reduce the safety margins, and hence the dose delivered to the normal tissue surrounding the tumor. In addition, a reduction of the proton range uncertainty could lead to the use of novel beam arrangements making greater use of the distal beam edge. Therefore it would be possible to maintain target coverage while reducing OAR and healthy tissue doses when the range uncertainty is low. Monitoring the proton range in vivo is a key tool to achieve this goal, and thus to improve the overall treatment effectiveness. Several techniques have been proposed to address the fundamental issue of in vivo proton verification, most of which exploit secondary particles produced by the interaction of protons and target nuclei, and are detectable outside the patient. Using these techniques, pre-clinical and clinical tests have obtained promising results in terms of absolute proton estimation. However, none of the investigated techniques are currently employed in the daily clinical workflow. A method already tested on patients is based on PET (Positron Emission Tomography) photon detection. The amount and emission distribution of PET photons depend on the target activity induced by the beam, as well by the delivered dose. Although this method has been clinically tested on patients, it has several limitations. The yield of annihilation photons produced during treatment depends on several factors, including the activity produced by the beam, which is fairly limited (up to two orders of magnitude lower than the diagnostic PET), the metabolic biological washout, and the background due to prompt radiation originated from other reaction channels. These issues have been partly resolved by the use of in-beam PET scanners, which measure annihilation photons during the treatment. One of the most advanced versions is the INSIDE (INnovative Solution for In-beam Dosimetry in hadronthErapy) PET scanner installed at CNAO (Centro Nazionale di Adronterapia Oncologica) in Pavia, Italy. Currently, it is part of a clinical trial and has acquired in-beam PET data during the treatment of various patients. Although encouraging results were obtained, still some limitations in its clinical applicability remain. In-beam PET is designed to work with low-duty-cycle accelerators, and so far it has only been installed in a fixed beam line. The other promising approach for in-vivo range monitoring is based on the prompt gammas (PGs) detection from nuclear de-excitation due to beam interactions in the tissue. The adjective prompt reflects the fact that they are emitted just a few pico-nano seconds after the impact of the proton on the target nucleus. The PGs escaping the patient have energy up to approximately 8 MeV, and their production is spatially correlated to the proton range. The feasibility of using an in vivo prompt gamma-based range verification for proton therapy has been demonstrated by numerous experimental and Monte Carlo studies, as well as by its recent application to the clinical practice for inter-fractional range variations. The current accuracy achieved on patients for retrieving the range of a single pencil beam is 2-3 mm. A major limitation identified by all studies that prevent the full exploitation of any prompt-gamma based approach for single spot range verification is the low statistics of the events produced. This issue is caused by: i) the short duration of a single spot delivery, ii) the immense gamma-ray production rate during delivery, iii) the finite rate capability of detectors, iv) the electronic throughput limits and v) the signal-to-background ratio. A particular PG range verification technique is prompt gamma spectroscopy (PGS). It relies on the analysis of the prompt gamma energy spectrum, which is characterized by specific energy lines corresponding to the reaction channels of the irradiated protons with the elements of the human body. The most common reactions are those with Oxygen and Carbon atoms, which become excited and eventually emit prompt gamma rays up to 8 MeV. Different studies on simplified geometries demonstrated that, by using the PGS technique, it is possible to estimate not only proton range variations, but also differences in the elemental composition of tissues. In this study, we present a novel approach for in vivo range verification via prompt gamma spectroscopy, based on creating signature gammas emitted only when protons traverse the tumor, and whose yield is directly related to the beam range. We propose to achieve this goal by loading the tumor with a drug-delivered stable element, that emits characteristic de-excitation PG following nuclear interactions with the primary protons. The use of tumor marker elements is not new in clinics: an example is a diagnostic PET which employs 𝛽+ emitter isotopes linked to a drug carrier, that is uptaken by the tumor allowing its diagnosis via PET scans (e.g. 18-FDG). In our approach, the radioisotope is substituted by a stable element, which decays via PG emission only when the proton interacts with it. By detecting signature gamma lines emitted by the tumor marker element, it is possible to assess if the beam has interacted or not with the tumor and increase the accuracy of the proton range estimation. Selection and characterization of candidate tumor markers The first part of this work focused on the identification of potential candidate elements following three criteria: i) emission of signature gamma energy lines following the proton irradiation, different from the characteristic emission of 12-Carbon and 16-Oxygen; ii) it should not be toxic for the patient iii) selection of an element whose carrier maximizes the tumor selectivity. While (i) is a purely physical constraint, and was deeply investigated in this work, points ii) and iii) depend also on several biological parameters, such as the achievable element concentration in the tumor, molecular carrier, tumor physiology, etc. To fulfill these criteria, we looked at elements that are already employed in medicine, either for diagnostic or therapeutic purposes and for which a drug carrier already exists. This allowed the applicability of our methodology in the clinic. Combining these criteria with simulations from the code TALYS, we identified three candidate tumor markers: 31-Phosphorous, 63-Copper and 89-Yttrium.We employed TALYS to characterize the elements in terms of the energy spectrum and gamma production cross section, and compared the results to Carbon and Oxygen, which are the two most abundant elements in the body. TALYS indicates that the three candidate elements produce signature gamma lines between 1 and 2 MeV, while Carbon and Oxygen signatures are between 3 and 8 MeV. Furthermore, the gamma yield per incident proton generated by the labeling elements is on average one order of magnitude higher than Carbon and Oxygen. To verify TALYS theoretical calculations, we designed an experimental campaign of prompt gamma spectroscopy measurements to characterize the emission of these elements when irradiated with a therapeutic proton beam. We irradiated two types of targets: solids made of 99.99% of candidate elements, and water-based solutions containing the label elements. While solids were used to characterize the PG energy spectrum emitted by the elements without background, the liquid targets were used to study the methodology in a setup closer to the clinical scenario, i.e by investigating the gamma emission of a compound material with a well-defined concentration of the marker element. Furthermore, using water-based solutions we were able to characterize the PG spectrum emitted by different element concentrations (from 2 M to 0.1 M), and evaluate the minimum value that provides a detectable signature. We characterized the elements by irradiating the different targets by using monoenergetic proton beam at 25 MeV and 70 MeV. Due to the thickness of the target, the beam looses all its energy inside the target, thus, these energies can be representative of a proton beam stopping in the first 5 mm of the tumor and after 4 cm depth, respectively. The 70 MeV proton beam was available at the experimental room of the Trento proton therapy center (Italy), while the Cyrcé cyclotron (Institut Pluridisciplinaire Hubert CURIEN-IPHC) in Strasbourg (France) accelerates protons up to 25 MeV. In the experiments performed in Trento and Strasbourg, we employed a LaBr3:Ce gamma-ray detector, which is suitable for our measurements as it is characterized by a fast detection response and high energy resolution. The data confirmed that all candidates emit signature PGs different from water (here used as a proxy for normal tissue), and that the gamma yield is directly proportional to the element concentration in the solution. We detected four specific gamma lines for 31P (1.14, 1.26, 1.78 and 2.23 MeV) and 63Cu (0.96, 1.17, 1.24, 1.326 MeV), while only one for 89Y (1.06 MeV). We compared all experiments with TOPAS MC. It is one of the leading toolkits for simulating particle interaction in the matter for medical physics applications. The comparison between simulations and experiments suggested that TOPAS is able to predict the energy of all characteristic gammas detected in the experimental spectrum, while the yield is either underestimated or overestimated, depending on the gamma-ray energy and element. Previous works had already shown TOPAS limited accuracy in reproducing nuclear de-excitation gammas, even for the most common materials like 16-Oxygen and 12-Carbon, and suggested that this discrepancy stems from the nuclear reaction models implemented in the physics list. Our findings support the hypothesis that the nuclear reaction cross section models available in TOPAS MC predict results with limited accuracy also for 31P, 63Cu and 89Y. Prompt-gamma yield and proton range correlation The finding of the first part of this work indicated that loading the tumor with 31P, 63Cu and 89Y generates a signature PG energy spectrum when irradiated with protons at therapeutic energies. In the second part of the project, we experimentally showed how the PG yield correlates with the proton range. We designed a multilayer phantom to mimic the irradiation of a deep-seated tumor. The phantom was composed of 15.5 cm of solid water (proxy of normal tissue), followed by 5 cm of liquid target filled with water-based solutions containing the marker element (tumor region) and an additional 2 cm of solid water for protons stopping downstream of the tumor.We irradiated the phantom with protons of energy ranging from 154 MeV (16.3 cm range in water) to 184 MeV (22.5 cm range) in order to build an experimental curve of the PG yield of different gamma-ray lines versus the proton range. We also acquired a blind spectrum at an unknown proton energy and used the curve to predict the range. By using the de-excitation peaks of 6.12 MeV from 16O, 4.44 MeV from 12C and 1.26 MeV from 31P, we successfully predicted the proton range of the blind data within 2 mm from the nominal value. The same test was repeated using a 63-Copper target, but due to the signature gamma lower yield, we overestimated the proton range prediction of 5 mm. As already observed for the liquid targets, large discrepancies were found between the experimental data and the simulation. This confirmed that TOPAS MC is not an accurate tool for predicting the PG yield. Toward the clinical application In the last part of the thesis, we discuss the applicability of the presented approach to patients. All experimental measurements were performed in conditions not clinically realistic because they investigated the basic principles of the methodology and provided a proof-of-principle. Using the measurements acquired at 70 MeV with liquid targets, we evaluated the expected PGs produced during a proton therapy treatment if the tumor were irradiated with 109 protons, the elements were loaded with a concentration of 0.4 mM (possible value when a glucose-based carrier is used) and a detection system with a larger solid angle acceptance (5sr) than the one used in our experiments (0.13 sr). We also started a preliminary in-silico investigation of our methodology applied to a real patient geometry. All experimental and simulated results so far presented were obtained by irradiating only homogeneous phantoms without taking into account patient heterogeneity and complex elemental compositions of the different tissues. To reproduce the patient geometry, we used a Computed Tomography (CT) image (3D map of the patient’s anatomy and tissue densities). The tumor region was localized on the prostate organ and its elemental composition and was artificially modified to achieve a homogeneous 31-Phosphorus, 63-Copper and 89-Yttrium concentration at a 5% percentage mass fraction for speeding up the computational time. TOPAS MC was used to simulate the irradiated of the tumor region with a 174 MeV proton beam and we simulated different beam position shifts from the nominal plan of 0.2, 0.4, 0.7, 1.0, 1.2, 1.4, 1.7 and 2 cm. Following the approach of the Massachusetts General Hospital group for prompt gamma spectroscopy range verification, we estimated the voxel-based gamma-ray yield from the elemental composition of the patient (CT scan) and from the gamma-ray production cross sections. TOPAS MC was used only for the calculation of the proton kinetic energy in each voxel of the patient. This analysis highlighted that gammas generated by the label elements are strongly correlated to the elemental composition of tissues traversed by the beam. When the beam partially misses the tumor region, the number of signature PGs emitted by the marker element decreases. Several aspects of the methodology still require further investigation and optimization from a physical, engineering and biological point of view. in vitro and in vivo toxicity studies must be conducted to determine the best carrier molecule that maximizes the tumor’s element concentration. Furthermore, to increase the accuracy of proton range estimation a novel gamma spectroscopy detection system must be designed to be fully integrated with the gantry treatment room. In conclusion, in this work, we demonstrated that loading the tumor with a label element before proton treatment generates signature gammas that can be used to verify the beam range in vivo. We selected three candidate elements already used in the clinic as promising tumor markers. We successfully employed these elements to simulate a proton range verification methodology on a homogeneous phantom. We showed how the current nuclear reaction models for prompt gamma spectroscopy applications are not accurate in predicting the PG yield from all the elements investigated. Further work is necessary to investigate the effect of a non-homogeneous element uptake due to tumor physiology on the proton range accuracy, as well as the diffusion of the label element on the normal tissue surrounding the tumor.