Introduction
According to the latest report by the World Health Organization, lung cancer stands out as a leading cause of cancer-related fatalities. Reference Siegel, Miller, Fuchs and Jemal1 Proton therapy, a specialised form of radiation treatment, offers a distinctive advantage by precisely delivering proton particles to the intended target, utilising the unique property of the Bragg peak to administer doses effectively. Nonetheless, this characteristic poses a potential risk as inaccuracies in proton delivery may exacerbate toxicity in surrounding normal tissues. Reference Khan and Gibbons2
Proton therapy holds promise in lung cancer management, yet it encounters technical hurdles such as respiratory motion and tissue density heterogeneity. The limited tolerance of adjacent normal tissues to radiation necessitates meticulous dose control to mitigate potential harm. Additionally, the respiratory-induced motion of both normal tissues and the tumour can affect the coverage of the tumour dose during treatment. Studies indicate that a significant percentage of lung tumours exhibit substantial motion during respiration, underscoring the need for imaging techniques to provide essential information for target visualisation, proton range estimation, patient alignment and treatment assessment.
In the context of non-small cell lung cancer (NSCLC) treatment, proton therapy shows potential in minimising damage to crucial healthy lung and cardiac structures, improving survival rates in locally advanced cases, reducing adverse effects in early-stage disease and demonstrating efficacy with decreased toxicity in early-stage NSCLC patients. Reference Brooks, Ning, Verma, Zhu and Chang3,Reference Vyfhuis, Onyeuku and Diwanji4
The escalation of proton dose concentration necessitates precise range and dose distribution estimates through in-vivo monitoring to uphold treatment plan compliance and preserve normal tissues and organs at risk. Reference McGowan, Burnet and Lomax5 While conventional imaging techniques furnish vital information, they may fall short in reducing range uncertainty and verifying treatment accuracy. Thus, the exploration of in-vivo monitoring methods capable of assessing proton range, dose distribution and non-invasiveness is essential. Recent research has predominantly focused on positron emission tomography (PET) and prompt gamma imaging (PGI) techniques for proton range verification, leveraging the detection of secondary particles emitted from non-elastic proton-nucleus interactions in the patient’s body. Prompt gamma rays offer a significant advantage for real-time monitoring due to their rapid diffusion after nuclear interactions, unlike PET imaging, which suffers from delayed signals and biological washout effects. Additionally, prompt gamma beams have a broad energy spectrum of about 15 MeV, originating from excited nuclear states unique to each element, providing valuable insights into tissue composition and element density for spectral analysis. Reference Min, Kim, Youn and Kim6,Reference Mackin, Peterson, Beddar and Polf7
In the realm of PGI, prompt gamma particles are emitted promptly after proton-nucleus interactions, offering real-time tracking of the Bragg peak during radiation delivery without washout effects, a distinct advantage over PET imaging.
Some investigations have centred on optimising the use of prompt gamma rays for determining beam range and dose distribution in various targets, with findings underscoring the dependency of prompt gamma particle spread on target tissue composition. Additionally, studies have highlighted the inverse relationship between prompt gamma emission intensity and oxygen concentration in the target tissue. Reference Polf, Panthi, Mackin, McCleskey, Saastamoinen, Roeder and Beddar8
Previous studies have demonstrated the verification of proton beam range by analysing prompt gamma distributions in proton therapy assessments. This has been done using various imaging cameras on both uniform and varied targets, through both simulations and experimental methods. Additionally, research has confirmed the range for head, neck and prostate patients using PGI. However, there remains a lack of thorough research on predicting the range of lung cancer using PGI. Reference Kormoll, Fiedler, Schöne, Wüstemann, Zuber and Enghardt9–Reference Peterson, Robertson and Polf11
This study aims to explore the correlation between dose deposition profiles following proton beam irradiation of a lung tumour and emitted scattered gamma particles. By fine-tuning proton beam parameters to align the Bragg peak with the tumour’s central region and efficient monitoring of proton beam ranges during treatment, the feasibility of estimating tumour position based on prompt gamma emission scattering profiles is investigated. Furthermore, the impact of tumour size on scattered gamma emission intensity is also scrutinised.
Material and method
Monte Carlo simulations
The simulations were done using Gate, an open-source Monte Carlo (MC) code. A Monte Carlo simulation was executed to simulate proton beam radiation and an imaging system, meeting the necessary requirements for studying PGI in proton therapy. The QGSP_BIC_HP_EMY physics list was utilised, a choice validated in prior studies, Reference Grevillot, Bertrand, Dessy, Freud and Sarrut12–Reference Santin, Strul and Lazaro14 encompassing both electromagnetic and hadronic processes and demonstrating good agreement with prompt gamma measurements. Reference Robert, Dedes and Battistoni15 QGSP (Quark-Gluon String Precompound) model is utilised for high-energy hadron collisions and BIC (Binary Light Ion Cascade) for cascade modelling and tracking of low-energy particles, HP to describe the cross sections of neutrons with energy of 20 MeV and less, and finally, EMY provides a range of electromagnetic processes with accurate simulation of gamma particles and charged particle transfer. To ensure only particles with a momentum exceeding 100 eV/c can escape the phantom, a minimum kinetic energy of 100 eV for gamma rays and 5.34 μeV for neutrons was set. The range cuts were established at 0.1 mm for all particles, and variance reduction techniques were not utilised.
Verification of the Gate simulation code involved comparing the range obtained from simulating a 60 MeV single pencil beam, consisting of 108 protons representing the most distal pencil beam in a typical treatment plan, with the range obtained from the cubic water phantom and the continuous slowing down approximation (CSDA) range from The National Institute of Standards and Technology (NIST) library data. The pencil beam exhibited a lateral full-width-at-half-maximum of 9 mm, aligning with current Proton Beam Scanning system standards. Reference Smeets, Roellinghoff and Prieels16,Reference Parodi, Mairani and Brons17 The statistical t-student test was employed to compare the results of the Gate simulation with the NIST library data, ensuring the accuracy and reliability of the simulation outcomes.
Imaging system
The 2D imaging system, developed and refined by Min et al. in 2011, Reference Min, Lee and Kim18 was utilised for the effective measurement of prompt gamma particle distribution. This detection system incorporates a multi-slit collimation system, CsI (Tl) scintillator arrays and a multichannel data acquisition system to examine the relationship between the detection profile and proton range. Each individual detector within the imaging region is arranged in a two-dimensional layout of 50 (longitudinal) × 42 (lateral) detectors with a 6-mm pitch, including a collimation hole measuring 150 mm (length) × 4 mm (width) × 4 mm (height) and a CsI (Tl) scintillator measuring 50 mm (length) × 4 mm (width) × 4 mm (height). These components are enclosed in a 10-cm thick tungsten shield to minimise background interference, as illustrated in Figure 1.
Figure 1. Dimensions of 2-D prompt gamma detection system.
Case of study
The anthropomorphic phantom Non-uniform rational B-spline (NURBS) cardiac and torso (NCAT), created by Segars et al. in 2001, Reference Segars19 was utilised to examine lung tumours. To account for tissue heterogeneity, attenuation coefficients were adjusted using NIST databases. For dosimetry calculations, the phantom lung area measuring 256 × 256 × 300 mm3 with voxel sizes of 1 × 1 × 1 mm3 was taken into consideration.
Matrix-generated spherical tumours of varying radii were positioned on the cross-sectional boundary of the lung wall. The tumours increased in size from the lung wall boundary towards the interior of the lung, with the tumour centre coordinates calculated as the sum of the tumour radius and lung wall coordinates.
The tumour sizes, classified into 5 T groups based on tumour diameter, were derived from the NMT (Tumour, Node, Metastasis) system analysis using data from the International Association for the Study of Lung Cancer (IASLC) database. Lung tumour classification based on tumour diameter information from the IASLC database is presented in Table 1. Figure 2 is a schematic image of the dosimetry matrix of the lung area in the axial view, showing how various tumours are located in the lung.
Table 1. Classification of lung tumours based on the diameter of the tumour
Figure 2. Placement and position of various tumours in the NCAT phantom from the axial view.
Studies by Striping et al. in 2013 Reference Sterpin, Sorriaux and Vynckier20 assumed human body tissues are predominantly composed of 1H, 12C, 14N and 16O. The 1H weight fraction in most tissues falls within 10% of the gamma spectrum, influencing proton flux through elastic proton-proton interaction without impacting the gamma spectrum. Reference Sterpin, Janssens and Smeets21,22 To validate the simulated phantom, a 60 MeV proton beam was irradiated to the NCAT phantom. The gamma ray spectrum emitted within the 2–8 MeV energy window was compared with values described by Kozlowski et al. in 2001, Reference Kozlovsky, Murphy and Ramaty23 sourced from gamma emission line cross-section data for these elements.
Tumour localisation
As outlined in the ICRU63 report, Reference Malmer24 it is crucial to ensure that 98% of the tumour volume receives at least 98% of the prescribed dose (D98 ≥ 98%). Therefore, it is essential to characterise the beam’s physical properties concerning beam energy and tumour size within a phantom. To achieve this, considering the Gaussian distribution of the proton pencil beam, each tumour phantom was exposed to irradiation from a new source at three randomly selected energies within the treatment range. The energy emission characteristics and spatial beam distribution were adjusted based on the tumour size, and the appropriate energy was chosen by aligning the Bragg peak position of each energy with the tumour centre.
To facilitate in-vivo monitoring of distal dose and tumour location, 109 protons to mitigate statistical effects were directed towards the lung tumour in the body depth direction. Subsequently, the distribution profile of prompt gamma rays was captured using a camera system with a 2–8 MeV window positioned on the right side of the NCAT phantom.
Results
Figure 3 displays the emitted gamma energy spectrum following 60 MeV proton pencil irradiation of the NCAT phantom within the 2–8 MeV energy window. By comparing the gamma spectrum emitted by the phantom with data derived from cross-section gamma emission lines resulting from nuclear interactions between protons and 12C, 14N and 16O targets at energies ranging from 2 to 8 MeV in Table 2, it is observed that peak intensities were noted at energies of 2.31, 3.8, 4.44, 5.27 and 6.13 MeV, corresponding to gammas emitted from interactions of protons with the nuclei of 12C, 14N and 16O tissue elements. This validation confirmed the suitability of the phantom for this study.
Figure 3. Gamma spectrum flounce emitted by irradiated 60 MeV proton beam to NCAT phantom.
Table 2. Available experimental cross-section data of gamma emission lines from nuclear interactions with 12C, 14N and 16O targets at energies of 2–8 MeV Reference Kozlovsky, Murphy and Ramaty23 (g.s = ground state)
The parameters of the proton pencil beam source were tailored individually to accommodate 98% volume coverage of lung tumours of varying sizes. Additionally, the radiation energy was randomly adjusted at three levels. Figure 4 showcases longitudinal profiles of dose distribution after irradiation at three distinct energies to phantoms harbouring different tumour sizes.
Figure 4. Longitudinal profiles of dose distribution after proton beam irradiation at three estimated and selected energies in the range of clinical energy for lung cancer treatment to tumour phantom with radius a: 5 mm, b: 12 mm, c: 20 mm, d: 30 mm and e: 40 mm.
Table 3 outlines the alignment of the lung tumour centre and the Bragg peak across the estimated energies. Based on the findings, energies of 77, 79.5, 83, 87.5 and 91.5 MeV were identified as optimal choices for covering tumour volumes ranging from 1 to 5, respectively.
Table 3. The position of the centre of the lung tumour and the Bragg peak in the estimated energies
Figure 5 illustrates the prompt gamma distribution measured using a 2-dimensional detection system across various tumour groups, presented in axial (top-left) and sagittal (bottom-left) views. Furthermore, the image showcases the intensity profile of dose distribution laterally at the distal dose edge position (top-right) and longitudinally along the path of the proton beam across various tumour phantom groups (bottom-right). In the longitudinal profile of prompt gamma distribution intensity, the position of the Bragg peak of the irradiated proton beam is indicated by a dotted line.
Figure 5. The prompt gamma distribution measured using a 2-dimensional detection system in axial (top-left) and sagittal (bottom-left) views, as well as the intensity profile of the dose distribution laterally at the distal dose edge position (top-right) and longitudinally along the path of the proton beam from tumour phantoms (bottom-right). The tumour phantoms with radiuses: a: 5 mm, b: 12 mm, c: 20 mm, d: 30 mm and e: 40 mm.
Discussion
The unique properties of proton therapy, including the Bragg peak effect in dose deposition, require careful consideration in treatment design and delivery compared to traditional radiation therapy. As a result, the need for online monitoring of dose distribution is crucial. One approach that has been explored for this purpose is gamma imaging, which allows for real-time monitoring during proton therapy. Previous research has investigated various gamma imaging methods for estimating proton range through Monte Carlo simulations and experimental approaches. Reference Smeets, Roellinghoff and Prieels16,Reference Moteabbed, Espana and Paganetti25–Reference Paganetti27
In this study, we sought to assess the feasibility of using 2D gamma imaging with a multi-slit collimator scintillation camera for online monitoring of dose deposition during proton therapy for lung cancer. Through Monte Carlo simulations, we aimed to evaluate the potential of this method for accurately estimating proton range in a clinical setting. This research adds to the growing body of knowledge on the use of gamma imaging for monitoring proton therapy treatments and highlights its potential utility in enhancing treatment precision and efficacy.
The Gate simulation code was initially validated by comparing the range obtained from simulating a 60 MeV single pencil beam with 108 protons to that of a cubic water phantom and the CSDA range from the NIST library data. The results revealed that the simulated range was within 1% of the CSDA range from the NIST library data, confirming the accuracy of the simulation code. The statistical t-test was employed to compare the results, providing further validation of the simulation code’s accuracy.
To assess the consistency between the range values generated by the Gate simulation code and those from the NIST library data, a t-test was conducted, resulting in a p-value of 0.997. This indicates that there is no significant difference between the range values obtained from the Gate simulation code and the NIST library data, aligning with previous findings by Zarifi et al. [Reference Zarifi, Taleshi Ahangari and Jia28] and reinforcing the accuracy of the simulation code.
Subsequently, the NCAT phantom was exposed to a 60 MeV proton beam, and the gamma ray spectrum within the 2–8 MeV energy range was compared to data from Kozlowski et al. [Reference Kozlovsky, Murphy and Ramaty23]. The peak intensities in the spectrum corresponded to gamma emissions resulting from proton interactions with 12C, 14N and 16O tissue elements, confirming the precision of the simulated phantom in replicating human body properties and proton-tissue interactions. This underscores the significance of utilising detailed and precise phantoms in the advancement of proton therapy treatments, as demonstrated by the NCAT phantom’s ability to accurately model physical interactions essential for effective treatments.
A study by Lopes et al. [Reference Lopes, Crespo and Simões29] presented a similar investigation into patient variations’ impact on the prompt gamma distribution index using the NCAT phantom with multi-slat collimation. Evaluation of the gamma energy spectrum obtained from the phantom revealed that gamma particles with energies of 2.31, 3.8, 4.44, 5.27 and 6.13 MeV exhibited the highest flux intensity, emanating from interactions between protons and target elements 12C, 14N and 16O. Validation of the simulated NCAT phantom’s capability in characterising prompt gamma distribution via gamma spectroscopy affirmed the accuracy of the findings.
The study aimed to characterise the physical properties of the proton beam concerning beam energy and tumour size in a phantom to ensure that 98% of the tumour volume receives 98% of the prescribed dose, as recommended by the ICRU63 report. Each tumour phantom was irradiated with a proton pencil beam source featuring a Gaussian distribution, adjusted for three different energy levels based on tumour size. The appropriate energy level was selected by aligning the Bragg peak location of each energy with the centre of the tumour. Table 3 outlines the lung tumour centre position and corresponding Bragg peak location for the selected energies in each tumour group.
For the T1a group with a tumour centre at 97 mm, the 77 MeV energy was chosen, aligning with the Bragg peak at 97 mm. Similarly, for the T1b group with a tumour centre at 104 mm, the 79.5 MeV energy was selected, matching the Bragg peak at 104 mm. This process was repeated for the T2a group with 83 MeV (112 mm tumour centre), T2b group with 87.5 MeV (122 mm tumour centre) and T3 group with 91.5 MeV (132 mm tumour centre). Figure 4 presents the prompt gamma distribution measured using a 2D detection system for the different tumour groups, showcasing axial and sagittal views of the prompt gamma distribution along with the intensity profile of the dose distribution laterally at the distal dose edge position and longitudinally along the proton beam path.
By optimising the proton beam energy and spatial distribution based on tumour size and selecting the appropriate energy to align the Bragg peak with the tumour centre, the study aimed to ensure precise dose delivery to 98% of the tumour volume while minimising radiation exposure to surrounding healthy tissues. The PGI system was utilised for monitoring the distal dose and tumour localisation during the irradiation process.
In additional simulations, the multi-slit gamma detector was incorporated alongside proton beam radiation for various tumour phantom groups. Figure 4 illustrates the longitudinal profiles of emitted gamma particles and proton pencil beam dose deposition in different tumour phantoms. Evaluation of the gamma particle distribution from the tumour phantoms revealed the highest emission along the path of the proton beam entering the body until reaching the lung wall, comprised of soft tissue and bone. Within the lung tumour region, a significant decrease in gamma particle emission intensity was observed. Previous studies exploring gamma particle emission dependence on target element structure, composition and various tissues such as soft tissue, bone and lung have consistently shown minimal gamma particle emission following proton interaction with lung tissue nuclei, validating the simulation results in this study. Reference Coruh, Demez and Ewell30
On one hand, Polf et al. [Reference Polf, Avery, Mackin and Beddar10] found that the emission of prompt gamma rays during patient radiotherapy was directly linked to the concentration of oxygen in the target tissue. They noted that an increase in tissue oxygen concentration led to a decrease in prompt gamma emission, aligning with the results of the current study. The findings suggest that higher oxygen concentration results in reduced prompt gamma particle emission.
Analysing the prompt gamma distribution intensity shown in Figure 4 reveals that larger lung tumours exhibit higher oxygen concentration, consequently leading to a further decrease in prompt gamma particle emission. This implies that as lung tumour size increases, estimating range based on prompt gamma particle emission patterns becomes increasingly challenging.
The intensity of simulated prompt gamma distribution and background neutron contribution significantly impacted the measurements and results recording. Adjusting the momentum to 100 eV/c in this study reduced background neutrons, but refining nuclear models, cross-section data using experimental data and extensive validations are necessary for more reliable conclusions. Modelling the entire treatment room can help reduce background neutrons, albeit at the cost of longer calculation times. High-speed Monte Carlo simulations can offer efficient and real-time monitoring for improved treatment practices.
Despite the potential benefits, utilising prompt gamma particles for range estimation in a clinical context continues to pose challenges. These challenges include prompt gamma detection limitations, indirect in-vivo dosimetry difficulties concerning prompt gamma particles, constraints related to the clinical environment and workflow, the duration and intensity of proton beam irradiation and limitations on irradiation intensity. Addressing these fundamental challenges is essential to enhance the clinical application of prompt gamma particles for range estimation during proton therapy.
Conclusion
This study aimed to assess the viability of estimating proton range through 2D gamma imaging with a multi-slit collimator scintillation camera for online dose deposition monitoring in lung cancer proton therapy. The research encompassed validating the Gate simulation code, scrutinising range variations, analysing gamma energy spectra, optimising physical attributes of proton beams, selecting appropriate treatment energies and integrating gamma detectors into proton beam radiation models across tumour phantom groups. Findings suggest that estimating proton range using emitted prompt gamma particles is achievable for smaller lung tumours but poses challenges for larger tumours due to elevated oxygen concentration leading to diminished prompt gamma emission.
The study underscores the significance of acknowledging uncertainties, background neutron impacts and the imperative for refining nuclear models and cross-section data to enhance range estimation accuracy. Despite the potential utility of prompt gamma particles for range estimation, obstacles persist in prompt gamma detection, in-vivo dosimetry, clinical workflow constraints and intricacies associated with proton beam irradiation. Overcoming these hurdles is essential for realising prompt gamma-based range estimation in clinical applications. While this study presents promising outcomes for range estimation with prompt gamma particles, further research and technological advancements are crucial to tackle existing challenges and elevate the precision and effectiveness of this technique in proton therapy for lung cancer treatment.
Acknowledgements
This article is an excerpt from the thesis research project approved at Semnan University of Medical Sciences with the code of ethics IR.SEMUMS.REC.1400.243. In this way, the authors express their gratitude and appreciation for the financial support of the research assistants of university.
Financial support
The authors gratefully acknowledge the research council of Semnan University of Medical Sciences for financial support.
Competing interests
Elham Rohollahpour, Hadi Taleshi Ahangari and Sajjad Raghavi declare that they have no conflict of interest.
Ethical approval
All procedures performed in this study were in accordance with the ethical standards of the institutional and/or national research committee and with the 1964 Helsinki Declaration and its later amendments or comparable ethical standards.
Informed consent
The approval of the Ethics Committee is not required for retrospective observational studies at the hospital where the present study was conducted.
Sanction declaration statement
We read the sanction declaration statement, sanctions law and regulations.
-
The authors are not employed by the Governments of Iran, Syria or Cuba
-
The authors are preparing articles in their ‘personal capacity’
-
The authors are employed at an academic or research institution where research or education is the primary function of the entity (Semnan University of Medical Sciences)
Open access
