ABSTRACT. Monte Carlo simulations (MCS) are the gold standard for radiation transport calculations. Typically, MCS were a tool accessible only to programming specialists and required hours of computational time to calculate a patient dose distribution for a single field with adequate statistical accuracy. However, over the past decade, MCS have become more widely available to the general medical physics community through developments such as the TOPAS application, a more intuitive interface to the Geant4 general purpose Monte Carlo toolkit. At the same time, specialized GPU-based Monte Carlo systems have been developed that are able to calculate dose distributions in seconds, rivaling the speed of analytical algorithms. Consequently, several vendors now offer MCS within their treatment planning software for proton and photon therapy.

With MCS on the verge of becoming a standard tool in clinical practice, development efforts have shifted from providing accurate dose distributions in patients to investigating new imaging modalities (e.g. prompt-gamma) and understanding radiation therapy at a more fundamental level. The latter relates microscopic energy depositions to a macroscopic biologically observed effect. These simulations can involve simple extrapolations from lineal energy transfer distributions (the microscopic pendant to the macroscopic linear energy transfer) up to complex estimations of outcome, correlating energy depositions on DNA strands to the induction of double strand breaks (DSBs) followed by mechanistic modeling of cell repair that is then integrated over treatment volumes. Currently, such simulations typically still require programming expertise and hours of calculation time.

Here I will discuss current use of Monte Carlo simulations for patient treatment simulations, efforts to simulate the microscopic scale of radiation damage and the potential future translation of these efforts in treatment planning.

ABSTRACT. Microdosimetry is an effective approach to determine the effect of a mixed radiation field at cellular level. Monte Carlo codes, modelling particle interactions with matter, are extensively used in the scientific community to perform in-silico microdosimetric calculations for radiation protection in space and aviation and for hadron therapy Quality Assurance. This talk will show how to use one of these Monte Carlo codes, Geant4 [1], in this domain.

In particular the talk will be focused on the use of Geant4 to support the development of novel Silicon-On-Insulator (SOI) microdosimeters, performed at the Centre for Medical Radiation (CMRP), University of Wollongong, as alternative technology to address the shortcomings of traditional Tissue Equivalent Proportional Counters (e.g. [2-6]). It will be shown how the Monte Carlo simulations aided the improvement of the design of the device, in addition to the definition of a methodology to convert microdosimetric measurements in silicon to tissue [3],[6].

Comparisons of Geant4-based microdosimetric spectra against experimental measurements performed at the Heavy Ion Medical Accelerator in Chiba (HIMAC, Japan), using CMRP devices, quantified the reliability of this Monte Carlo code for microdosimetry. Geant4 was also used successfully to determine the RBE10 in the case of proton and carbon ion therapy, with good agreement with experimental measurements [2][5]. Finally guidelines will be provided to start to use Geant4 for microdosimetry.

Bibliography

[1] Allison, J., et al (2016) Recent developments in Geant4, NIM A, 835, pp. 186-225.

[2] Debrot, et al (2018) SOI microdosimetry and modified MKM for evaluation of relative biological effectiveness for a passive proton therapy radiation field, Physics in Medicine and Biology, 63 (23), art. no. 235007.

[3] Bolst, D., et al (2018) Optimisation of the design of SOI microdosimeters for hadron therapy quality assurance, Physics in Medicine and Biology, 63 (21), art. no. 215007.

[4] Bolst, et al, (2017) RBE study using solid state microdosimetry in heavy ion therapy, Radiation Measurements, 106, pp. 512-518.

[5] Tran, L.T. et al (2017) Characterization of proton pencil beam scanning and passive beam using a high spatial resolution solid-state microdosimeter, Medical Physics, 44 (11), pp. 6085-6095.

[6] Bolst, D., et al (2017) Correction factors to convert microdosimetry measurements in silicon to tissue in 12C ion therapy, Physics in Medicine and Biology, 62 (6), pp. 2055-2069.

ABSTRACT. Radiobiological models play a pivotal role in radiotherapy. These models can help to understand the underlying biology of cancer response to radiation treatment. They can also approximate the treatment environment and be used to develop decision support tools for oncologists that can provide guidance for treatment planning or design of future clinical trials. These models can be divided into top-down approaches (statistical data analytics) and bottom-up approaches, which start from first principles of physics, chemistry, and biology to model cellular damage temporally and spatially up to the observed clinical phenomena (Monte-Carlo [MC])*. Here, we will provide an overview on how MC methods can be employed to estimate the molecular spectrum of radiation damage in clustered and non-clustered DNA lesions (Gbp-1 Gy-1). We will briefly discuss the temporal and spatial evolution of the effects of ionizing radiation across the three phases (physical, chemical, and biological) and contrast the available MC codes that aim to emulate these phases along the molecular and cellular scales to varying extents. Will highlight future directions and prospects in radiotherapy.

*A Guide to Outcome Modeling in Radiotherapy and Oncology: Listening to the Data. CRC Press, 2018.

ABSTRACT. CATANA (Centro di AdroTerapia ed Applicazioni Nucleari Avanzate) was the first Italian protontherapy facility dedicated to the treatment of ocular neoplastic pathologies. Since 2002, it has been in operation at the LNS of INFN (INFN-LNS),successfully treating 400 patients to date. Nowadays, a slightly increased biological effectiveness (with respect to reference low-LET radiation) is considered in clinical proton treatment planning by assuming a fixed RBE of 1.1 for the whole radiation field. However, data emerging suggest how variations in RBE, which are currently neglected, might actually result in deposition of significant doses in healthy organs. In order to evaluate how the RBE value changes along a typical SOBP, Monte Carlo simulations and experimental measures were performed at seven positions along the CATANA 62-MeV SOBP. For this purpose four different microdosimetric detectors (silicon detectors and mini-TEPC) were adopted and uveal melanoma cells were irradiated. The whole experimental set-up and the physical characteristics were simulated using Geant4 toolkit. In particular, new codes were developed and implemented in order to reproduce LET track and LET dose distribution taking into account the secondary particles contribution. The spectra generated by these simulations were used as the physical input for the radiobiological simulations based on the MKM, LEM model and mixed-field approach, from which an estimation of the RBE along the SOBP was evaluated and compared with the experimental data.

ABSTRACT. The number of hadrontherapy facilities have worldwide increased noticeably during the last decades. For this kind of treatments, the beam properties within the patient must be known with high accuracy. Therefore, it is crucial to develop specific QA tools dedicated to this technique. Recently the capability of the ALPIDE chip to work as a silicon dosimeter for QA studies was investigated by the proton CT (pCT) group in Bergen. The ALPIDE sensor, originally developed for the Inner Tracking System of the ALICE detector, is a CMOS Monolithic Active Pixel Sensor (MAPS) and has a size of 30 mm × 15 mm × 0.05 mm, consisting of about half-million pixels, each with a size of 29.24 μm × 26.88 μm. For the purpose of this study, a Monte Carlo simulation of the ALPIDE chip was implemented using the Geant4-wrapping TOPAS simulation tool, and ion (carbon and helium) beams were shot along the longitudinal dimension (30 mm) of the chip. The ion beam energies selected were 140 MeV/u and 75 MeV/u, respectively for carbon and helium, corresponding to a range in silicon of ~25 mm. Due to multiple Coulomb scattering and the small thickness of the ALPIDE, few tracks stayed inside the chip for their entire paths for carbon ions and less for helium. To account for this scattering effect, up to 11 chips were stacked so that the ions could be tracked until they reached the Bragg peak region. As a preliminary experiment, a single ALPIDE chip was irradiated at the Heidelberg Ion-Beam Therapy Center (HIT) in Germany with carbon ions. Also in the experiment, it was possible to track the complete path of few carbon histories. These promising initial results showed the opportunity to further investigate this application with ALPIDE chip, to build a silicon QA system with micrometric resolution.

ABSTRACT. The results of Monte Carlo simulations of DNA damage are highly sensitive to the adopted simulation parameters and these can vary significantly across studies. In this work, we investigate the impact of changing several of the typically assumed parameters on the number of simulated DNA breaks. Here we focus on parameters in the chemistry stage of the DNA damage simulation using a human fibroblast nucleus model with a DNA content of 6×109 base pairs (6 Gbp) built in TOPAS-nBio. Three parameters were selected in the simulation to investigate their impact on the predicted DNA damage and set to values that are typically used in the literature, namely, the chemistry model, chemical stage length, and the hydroxyl (OH·) damage probability. The default Geant4-DNA chemistry model was compared to the default TOPAS-nBio chemistry model (TsEmDNAChemistry) and we found differences of up to 25% in the indirect strand break (SB) yield. Generally, the chemical stage is set to several nanoseconds to mimic scavenging properties of the cell. An investigation into different timescales of 1 ns, 2.5 ns or 10 ns found a difference of up to 60% in the indirect SB yield. The hydroxyl damage probability was defined as the probability of interactions between hydroxyl and DNA backbones or the whole DNA target to cause a strand break. It was set to 0.11, 0.4 or 0.65 in the simulations and the result suggested that the difference in damage probability can cause up to 84% difference in the indirect SB yield. This work investigated the effect of using different parameter values for the simulation of the chemical stage when simulating DNA damage and could provide important reference information for estimating uncertainties in the interpretation of future studies.

ABSTRACT. Radioembolization is one of the newest, non-invasive methods of cancer treatment. It consists in placing radioactive, spherical sources of microscopic size directly into the artery supplying the tumors with blood. The experimental setup is composed of two beta-radioactive sources Sr-90 placed in a such a way to obtain an uniform distribution of the dose in the Petri dishes with prostate cells. In this study the biophysical Monte Carlo code PARTRAC has been used. It is a tool for calculating track structures of a variety of ionizing radiation acts and their biological effects. The number of single and double DNA strand breaks was simulated for prostate cancer cells exposed to beta radiation. At the same time the radiobiological experiments were conducted for human prostate cancer and normal cell line to perform the clonogenic cell survival assay. The ionisation and excitation pattern allowed to estimate the number of single strand breaks (SSB) and double strand breaks (DSB) affecting the survival of cancer cells. Additionally, the size of colonies has been investigated with a countPHICS software. Further analysis of the colonies by measuring their size gives additional information, not incorporated in the clonogenic survival assay. The reduced colony size was observed for higher doses for both cell lines. Modeling the relationship between the DSB numbers and the energy deposits of the beta emitters within the cell and cell survival measurements provides primary results necessary to establish irradiation procedures in prostate radioembolization. The analysis may be exploit to create new cancer treatment method, which uses microspheric beta emitters.

ABSTRACT. Introduction

Hadrontherapy treatments rely on the estimation of the relative biological effectiveness (RBE). Biophysical models, such as the Microdosimetric Kinetic Model (MKM) [1] and the Nanodosimetry Oxidative stress model (NanOx) [2], evaluate the RBE using the calculation of specific energy spectra at micrometric scale (MKM, Nanox) and/or nanometric scale (LEM, Nanox). For the NanOx formalism, the chemical energy based on free radical production during irradiation is also considered. This study aims at benchmarking GEANT4-DNA and LQD/PHYCHEML/CHEM for the simulation of specific and lineal energy spectra and radiolysis species. Physical and chemical data are estimated for electron [1-100 keV], proton [1-250 MeV] and carbone ion [1 MeV/n- 400 MeV/n] monoenergetic beams.

Material and Methods.

Geant4-DNA physics lists are tested with the different options available. These options are recommended for discrete particle interactions and use different electron elastic and inelastic models. LQD [3] uses CDW-EIS calculation for ionizations induced by ions. Three-dimensional energy transfer points are calculated in thin liquid water slices. Transfer points are analyzed with the TED code to calculate, in micrometric and nanometric targets, the specific and the lineal energy. Radiochemical species (e-aq, OH., H2O2.,O2.), are calculated with the Geant4-DNA chemistry module and with PHYCHEML/CHEM modules associated to LQD. Results are then extrapolated to SOBP by superposing appropriate monoenergetic ion tracks.

Results.

The comparison of the results obtained with Geant4-DNA and LQD will be presented for monoenergetic beams as well as for SOBP. This study is preliminary to a full implementation of MKM and NanOx biophysical models into the GATE platform and the associated physical and chemical databases.

Bibliography

1. Hawkins RB. Med Phys. 1998;25: 1157–1170. 2. Cunha M, et al. PMB, 2017;62:1248-1268. 3. Gervais B, et al. RPC, 2005 ;75: 493-513.

ABSTRACT. Mixed beams consist of at least two types of radiation which are characterized by the varying density of the ions produced as a function of the particle path in the absorbent. Research is carried out to characterize and verify the configuration in which cells can be simultaneously exposed to high and low LET in a controlled temperature environment. It is important to determine whether the response to mixed radiation is greater (synergy) than the cumulative effect (additivity) of X-rays and alpha particles and to explain the possible mechanisms behind the observed reactions. Biological effects of mixed compounds are one of the areas of radiation research. The risk of exposure to mixed compounds is not clearly defined, and the experimental results are ambiguous. In this work, the biological effects of X-rays, alpha particles and mixed beams were examined Monte Carlo simulations with the PARTRAC tools. The PARTRAC has been developed at the Institute of Radiation Protection in Neuherberg, Germany. It contains a set of related modules to perform calculations of ionizing particles track structures, overlaid with models of nuclear DNA, to assess the DNA damage and subsequent damage response processes in cells. Scoring of DNA end rejoining from different chromosomes provides simulated data on chromosomal aberrations (CA) and their types, such as dicentric chromosomes and translocations. Calculations have been performed for X-rays, alpha particles and for mixed beams that irradiated human peripheral blood lymphocytes. The results obtained from the PARTRAC code will be presented at the meeting in comparison to the experimental data from the experiment carried out at the University of Stockholm in Sweden. The results of the conducted experiment and Monte Carlo simulations indicate the synergy. Exposure to mixed beams leads to a higher than expected frequency of CA.

ABSTRACT. Nuclear reactions induced during high-energy radiotherapy (>8 MV) produce secondary neutrons that, due to their carcinogenic potential, constitute an important risk for the development of iatrogenic cancer. However, this risk cannot be properly assessed without a full description of the biophysics underlying neutron dose deposition in human tissue. To address this, a branch of the ANDANTE project sought to ascertain a model of neutron relative biological effectiveness based on phenomenological microdosimetric studies. We are attempting to reproduce and build on these studies by constructing a pipeline to explicitly calculate the dose-mean lineal energy (yD) for a range of neutron energies that consists of (i) the open source Monte Carlo (MC) toolkit Geant4, (ii) its radiobiological extension Geant4-DNA, and (iii) a track-sampling algorithm first reported in 2017 by Famulari et al. More specifically, macroscopic Geant4 simulations were performed in which the ICRU 33 sphere was irradiated by mono-energetic neutrons and the energy spectra of the secondary charged particle species were scored at various depths. These spectra are currently being sampled as the input to Geant4-DNA track-structure (TS) simulations. Upon the completion of each, the resulting tracks are sampled via the aforementioned algorithm and yD is calculated for each secondary/depth/initial neutron energy combination. Lastly, a weighted sum of these results will be obtained for each depth/neutron energy pair in which the weights for each particle are based on their relative dose contributions. In contrast, the ANDANTE group employed the semi-empirical microdosimetric functions implemented in the MC toolkit PHITS to perform their calculations. Consequently, the current work is more computationally expensive; however, this cost should be compensated for by an increase in accuracy. A comparison between this work and the results of the ANDANTE group will be presented alongside an extension of this analysis to nanometre-scale scoring volumes.

ABSTRACT. Gold Nanoparticles (GNPs) have attracted a lot of attention due to their potential benefit for improving the efficacy of X-ray radiotherapy. Owing to their high atomic number, the GNPs are able to absorb higher quantities of incident radiation with respect to the surrounding tissue, increasing the local energy deposition in the region surrounding the GNP, due to the increase of the production of photoelectrons, low energy Auger electrons and fluorescence X-rays. Monte Carlo simulations play a key role in the investigation of GNP radio-enhancement and it is widely recognised that track structure physics models are the state of the art for nano-scale studies.

In 2016, we have developed track structure physics models for electron transport for microscopic bulk gold (Geant4_DNA_AU_2016) and we have recently improved them in the low energy domain, especially in terms of stopping power (Geant4_DNA_AU_2018). Here we report the benchmarking of this newly developed physics models when calculating the physical dose around a GNP and the Dose Enhancement Factor (DEF).

It is demonstrated that Geant4_DNA_AU_2018 models give similar azimuthal property of two-dimensional absorbed dose around a single GNP, but results in larger absorbed dose and DEF than Geant4_DNA_AU_2016 models. In parallel, we investigated the performance of a newly developed multiple scattering model in Geant4 based on the Goudsmit-Saunderson (GS) model, when used together with the electromagnetic physics models based on the Geant4 Livermore approach. Our results show that the GS model does not affect the results of the simulations when studying GNP radio-enhancement with a condensed-history approach.

ABSTRACT. Introduction The aim of this study was to demonstrate a proof of concept that given a scan of a two-dimensional pathology slide, information regarding patient specific microdosimetric quantities can be computed using image segmentation software and Monte Carlo based dose calculation engines. Variation in microdosimetric quantities between pathology slides taken from the same patient, different patients, and across radiation qualities including various brachytherapy sources both novel and conventional will be investigated.

Methods Pathology slides taken from several non small cell lung carcinoma patients were analyzed using a difference of gaussian (DoG) based segmentation algorithm implemented in Scikit-Image. An intensity threshold between nucleus and extra-nuclear material in the cell is automatically determined. This method returns accurate results when scrutinized under visual examination. Malignant and healthy tissues were contoured manually by a pathologist.

The cellular radii distributions generated previously are then randomly sampled in a Large-scale Atomic/Molecular Massively Parallel Simulator (LAMMPS) pouring simulation.

Electron spectra and microdosimetric parameters were calculated in separate simulations using Geant4 version 10.3 Patch 3 with Livermore condensed history cross sections. The patient-specific cellular model was imported into a Geant simulation and placed at the center of a water volume. A previously generated electron spectrum was sampled throughout and around the model. Both cells and nuclei were modeled as water. The energy of each microdosimetric energy event was recorded.

Results Cell and nuclei size distributions were derived from patient pathology slides. Specific energy per single event, f_1 (z), and lineal energy distributions, y_f, in a model derived from a healthy cell size distribution were computed for a 20 KeV photon source.

Discussion and Conclusions A proof of concept of a method to derive patient specific microdosimetric distributions is presented. With the use of appropriate microdosimetric kinetic models computation of patient specific relative biological efficiencies is now possible.

ABSTRACT. The spatial distribution of energy deposition events is an essential aspect in the determination of the radiobiological effects of ionizing radiation at the cellular level. Microdosimetry provides a theoretical framework for the description of these events. Recently, several studies have turned to the formalism of microdosimetry to address problems such as the characterization of Linear Energy Transfer (LET) and Relative Biological Effectiveness (RBE) of ion beams for particle therapy applications. Microdosimetry quantities and their distributions can be obtained by means of Monte Carlo simulations. In this work, we present a track structure Monte Carlo application, developed for the computation of microdosimetric distributions of protons in liquid water, based on Geant4-DNA [1-2]. This application provides two sampling methods “uniform” and “weighted”, for the scoring of the quantities of interest in spherical sites, with diameters ranging from 1 to 10 μm. Besides the distribution of energy deposited per event, the code computes also the distribution of energy imparted to the site per electronic collision of the proton, which has never been included in other microdosimetry examples in the official release of Geant4. With this distribution, in particular, the weighted average of the energy imparted per collision of the traversing proton can be computed and used to obtain the macroscopic dose averaged LET as proposed by Kellerer [3]. Finally, results obtained with our application are in good agreement with PARTRAC, which has been taken as the reference track-structure Monte Carlo tool for comparison [4].

[1] S. Incerti et al., Med. Phys. 37 (2010) 4692-4708. [2] M. A. Bernal et al., Phys. Med. 31 (2015) 861-874. [3] A. M. Kellerer, “Fundamentals of microdosimetry”, in: K. R. Kase et al (Eds.), The Dosimetry of Ionizing Radiation, vol. 1, ch. 2 (1985), Academic Press INC. [4] W. Friedland, P. Kundrát, Mutat. Res. 756 (2013) 213-223.

ABSTRACT. Radiation therapy is a commonly prescribed treatment for tumors of the central nervous system. In both adult and pediatric patients, brain irradiation is commonly accompanied with debilitating cognitive dysfunction. The underlying mechanism responsible for the deficit in learning and memory is however not known. Recent experiments have shown that irradiated rodents with cognitive impairment had altered dendritic spine morphology as well as an activation of microglia, which suggest that nuclear DNA damage is not the only radiation target in the neuron. Here, we investigate alternative extra-nuclear radiation targets in a small neuron network using an in silico study. In this study, we model a small network of neurons and their support glial cells with the radiobiology toolkit TOPAS-nBio, an extension to the Geant4 platform TOPAS. TOPAS-nBio utilizes the physical and chemical processes of Geant4-DNA. Hippocampal pyramidal neuron and supporting astrocyte morphologies were taken from the neuromorpho database and rendered into TOPAS-nBio. The cells were arranged to mimic our in vitro irradiation studies to allow comparison to experiments. The energy deposited in each cell morphology from incident x-rays and protons was calculated. The energy deposited in different compartments of the neuron, including the nucleus and dendrite organelles (spines and mitochondria) was also calculated. Despite the smaller volume of the glial cells (~600 um3) compared to the neurons (~ 104 um3), glial cells may receive a significant amount of energy deposition. The model included three spine morphologies (stubby, mushroom and thin) as well as dendritic organelles such as ribosomes and mitochondria, vital for synaptic function. Results indicated that mushroom spines received the largest energy deposition, while thin spines received the smallest amount. This suggests that mushroom spines may be most susceptible to radiation damage, these results will be used to design future neuron irradiation experiments.

ABSTRACT. In hadrontherapy treatment planning, one critical step consists of the evaluation of the radiation biological effectiveness, often performed by biophysical models. The main models currently applied in clinics, LEM and MKM, are also interfaced with radiation transport codes, like FLUKA, to support and validate the clinical activities. Recently, also the BIANCA (BIophysical ANalysis of Cell death and chromosome Aberrations) model has been interfaced with FLUKA. BIANCA is a two-parameter, mechanistic model based on assumptions that link the radiation-induced critical DNA lesions (“Cluster Lesions” or CL) to the production of chromosome aberrations, and eventually to cell death. The CL yield is one of the two model parameters, and it is tuned as a function of particle type and energy.

Following validation against experimental dose-response curves (for both cell survival and chromosome aberrations), in the present work we produced a radiobiological database describing the effectiveness of different particles (including protons, He- and C-ions) as a function of energy, for a low- and a high- α/β ratio reference cell line. Starting from this database, we developed a method to perform predictions for other cell lines, basing on their photon response. Such database can be read by a transport code and/or a treatment planning system.

Herein, we will show examples of applications where the BIANCA tables were read by FLUKA, allowing to calculate cell death and relative biological effectiveness along the whole dose profile, as well as specific chromosome aberrations, which are indicators of normal tissue damage. Several irradiation conditions for protons, He- and C-ions were investigated, for spread out Bragg peaks with different depths and extensions (produced with an in-house developed analytical method), for different cell lines. This approach may be useful not only to further benchmark the models currently applied in clinics, but also to provide additional information about normal tissue damage.

ABSTRACT. Purpose Under most cell and small animal irradiations, the backscatter material present is insufficient for full backscatter condition. TG61 and other dosimetry protocols lack guidance in dosimetry under those conditions. The purpose of this study is to build Monte Carlo (MC) models of several orthovoltage irradiators on the market to establish correction factors under those conditions. Materials & Methods MC models for Xstrahl SARRP and Precision X-RAD irradiators were built with PENELOPE based on the X-ray spectra provided by the manufacturer. Mu_en/rho and full backscatter factor were simulated and compared with TG61 tables to validate the MC model. Depth doses and backscatter factors for different thickness water phantom were simulated for each irradiator. The MC model was applied to an actual cell irradiation experiment. MC simulated dose was compared with hand calculated dose using correction tables generated earlier. Results It is demonstrated that as much as 50% dose difference exists for different backscatter conditions at the beam qualities studied. Dose to the cell plate is dramatically different from TG61 calibration result due to the insufficient backscatter condition. The depth dose tables for insufficient backscatter thicknesses estimated dose to cell plates that is less than 5% from direct MC simulation. Conclusions MC models for several orthovoltage irradiators were successfully built. Irradiator specific correction factor tables were generated to handle the dosimetry under insufficient backscatter condition. These tables can greatly improve the dosimetry accuracy in radiobiology research.

ABSTRACT. DNA double-strand breaks (DSBs) are critical DNA lesion and they may lead to gross chromosomal rearrangements, mutations, and cell death. During DNA repair histone alterations, nucleosome repositioning and changes in higher-order folding of the chromatin fiber cause significant accumulation of proteins in the lesions areas. These protein clusters form nuclear foci which may be visible by light microscopy. In the experiment, 53BP1 foci induced in U2OS cells were tracked during the live-cell observation in order to study their time and spatial evolution which affect the probability of the successful lesion repair. Cells were irradiated with alpha particles, X-rays and mixed radiation. Features like size, intensity, and mobility were analyzed from recorded videos, captured with 1 frame per minute. Obtained results show a synergistic manner of mixed radiation. Models created based on analysis may be implemented in any tool using Monte Carlo techniques for DNA damage simulation.

ABSTRACT. Purpose: Knowledge of the imaging doses delivered to patients and accurate dosimetry of the radiation to organs from various imaging procedures is becoming increasingly important for clinicians. The purposes of this study were to develop kV X-ray imaging dose calculation system for image guided radiation therapy and to evaluate the impact of kV X-ray imaging doses on treatment doses.

Methods: The Vero4DRT was equipped with gantry-mounted orthogonal kV X-ray imaging subsystems, consisting of two sets of X-ray tubes and flat-panel detectors. The EGSnrc/BEAMnrc and EGSnrc/DOSXYZnrc packages were used to simulate kV X-ray imaging dose distributions of Vero4DRT. Then, the kV X-ray imaging dose distributions such as 4D-CBCT, 3D-CBCT, correlation modeling and monitoring doses were calculated for 9 lung cancer patients based on the planning CT images with dose calculation grid size of 2.5×2.5×2.5 mm3. Finally, the imaging dose distributions derived via 4D-CBCT, 3D-CBCT, correlation modeling and monitoring, of planning target volume (PTV), the skin and the bone, were evaluated by examining dose-volume histograms (DVHs).

Results: Based on 4D-CBCT and 3D-CBCT, the doses covering 2-cc volumes (D2cc) were maximally 6.0, 10.5 and 58.1 cGy for the PTV, the skin and the bone. Then, the maximum D2cc of correlation modeling and monitoring imaging were 6.0, 9.3 and 48.4 cGy for the PTV, the skin and the bone.

Conclusions: We have developed kV X-ray imaging dose calculation system for image guided radiation therapy using Vero4DRT and evaluated the impact of kV X-ray imaging dose on treatment dose.

ABSTRACT. Introduction: Survival and linear-quadratic model fitting parameters implemented in treatment planning for targeted radionuclide therapy depend on accurate cellular dosimetry. Purpose: We have developed cellular polygonal mesh (PM) models from representative confocal microscopic images to calculate absorbed dose rate S-values for different subcellular radioactivity distributions of 177Lu, and compared them with MIRDcell and a representative constructive solid geometry (CSG). Methods: Immunofluorescence staining was used to detect cytoplasm (Cy), nucleus (N) and Golgi (G) of 9 human osteosarcoma cells in confocal microscopy Z-stack images. Images were analyzed with ImageJ and used as reference for developing the PMs of cell surface (CS), Cy, N and G in 3Ds Max. Each cell structure was converted into GDML format. Geant4 Monte Carlo toolkit was used to calculate the absorbed energy fractions from CS, CY and G to N. “Penelope 2008” low energy EM models were employed to simulate transport of 177Lu β-spectrum emitted from CS, Cy, or G. Results & discussion: S-values for N←Cy and N←CS decreased on average with 39±3% and 13±10%, respectively, when moving from simple spheres (MIRDcell) to a more representative CSG. These average differences further decreased to 60±6% and 37±5%, when comparing MIRDcell to the PM geometries. The PMs also allowed to take into account the alleged translocation of 177Lu to the Golgi (S(N←G)), resulting in an increased absorbed dose to the N of +64% and up to +149% when compared to a homogeneous distribution of the activity in the cytoplasm (S(N←CY)). Our results indicate that specific cell line features and dimensions largely affect the absorbed dose to the main cellular targets. Furthermore, including the modeling of the Golgi resulted to be crucial to perform accurate dosimetry on cellular level. Conclusion: These results suggest dose overestimation when planning therapy based on generalized dosimetry models.

ABSTRACT. Purpose: Monte Carlo (MC) simulations of electronic portal imaging devices (EPIDs) have been hindered by the intensive computation time required. A novel method, FastEPID, based on pre-calculated photon energy deposition efficiency (η) and optical photon spread functions (OSFs), has been introduced for fast image simulation of EPID.

Methods: OSFs and η are pre-calculated utilizing a validated EPID MC model. During FastEPID simulation, a dummy air slab replaced the EPID model. For each x-ray photon incident on the air slab the corresponding η and OSF were assigned based on the photon energy. A random number (RN) was generated for photon detection. If the RN was smaller than or equal to η, the incident photon was “detected”, otherwise it was discarded. For each detected photon, the corresponding OSF was added to the final image with the center aligned to the incident position. FastEPID simulated slanted slit images and planar images of a Las Vegas (LV) phantom were validated against experimentally acquired and conventionally simulated results in terms of modulation transfer function (MTF), signal-to-noise ratio (SNR), contrast-to-noise ratio (CNR), and contrast. Improvement of the simulation time utilizing the FastEPID method was quantified by comparison with conventional simulation.

Results: The FastEPID simulation provides an accurate prediction of MTF at low/middle/high spatial frequencies (difference less than 0.07), indicating satisfactory performance for imager resolution. For LV phantom simulation, FastEPID results were similar to the conventional method, when compared to measurement, for SNR, CNR, and contrast. The FastEPID method reduces the simulation time by a factor of approximately 150. Simulation of MV images can now be performed in a matter of hours, rather than weeks.

Conclusion: The FastEPID method has significantly shortened simulation time without compromising image quality. Multiple applications, such as imager design optimization will benefit from this novel simulation method.

ABSTRACT. There is a growing interest in using high-Z nanoparticles (NP) to improve the efficacy of challenging radiotherapy (RT) treatments, particularly for radio resistant tumours. Their properties (size, distribution, concentration, etc.) have an impact on cellular damage coupling with RT. In order to understand or predict the biological outcome, we need to retrieve more precise radiation tracks data at the microscale and even at the nanoscale. Indeed, these information (track structure, lineal energy, imparted energy, etc.) can be linked to the radiation damage. The package Geant4-DNA provides models to describe event-by-event electromagnetic interactions of particles in water down to electrons energies of a few eV. So Geant4-DNA is an ideal candidate to simulate the use of NP in RT. By using this code, we decided to study the dose enhancement caused by gold, iron and gadolinium NP as extensive in vitro data is available from synchrotron radiation therapy programs. The method consists in the simulations of monoenergetic X-Ray beams (from a few dozen keV to a few hundred keV) in a water medium. The beams interact in the first case only with water to get microdosimetric information in a 1 μm-diameter sphere, the detector, at different localizations from a reference location of the medium. In the second case a NP is introduced at the reference location of the medium and electron spectrum and microdosimetric information are registered in the same microsphere. We show a variation of the secondary electrons spectrum and consequently of the microdosimetric quantities in the presence of NP. We will also compare the results obtained with the different models proposed with Geant4-DNA. Significant and systematic differences are expected between the microdosimetric quantities when comparing the situation with and without NP and between the different models.

ABSTRACT. Patient dose evaluation has been recommended as part of a quality assurance program in diagnostic radiology. However it is difficult to make dose calculations for specific patients in time frames accessible to a clinic. Most of the programs can only make dose estimations which are not patient specific. In this study Monte Carlo (MC) simulations were used to perform dose calculations in lungs based on the information from a helical computed tomography (CT) scan belonging to an anonymous patient from the oncology Hospital SOLCA-Quito, Quito-Ecuador. The scan was obtained with the Philips Brilliance Big Bore tomograph and the MC simulations were performed with the opensource software GATE. The computer model of the scanner was developed and validated previously in our laboratory. For each lung the average dose were calculated in the central slice of the phantom under symetrical irradiation with irradiation times ranging from 0.3 to 12 s. The latter irradiates the whole phantom and provides the actual dose in the central slice. The dose calculated in the central slice for each irradiation time is representative of the dose in the whole organ provided that it is homogeneous. Our results sugest that it is possible to determine the dose corresponding to the irradiation of the whole phantom using the dose calculated with a fraction of the irradiation time. Consequently, we propose a procedure to estimate the average dose in lungs with minimal computation time.GATE, Monte Carlo Simulations, absorbed dose, computed tomography

ABSTRACT. The spectrum of X-rays produced by the interaction of monoenergetic electrons of 130 keV was estimated by striking tungsten, molybdenum and rhodium targets, in order to determine their energetic characteristics at 50 cm from the focal point. The study was done using Monte Carlo calculations with the code MCNP5 where the X-ray tube of a Siemens SOMATOM Perspective tomograph was modeled. The calculated spectrum showed a continuous and a discrete component; for the case of tungsten the continuous part was compared with the spectrum calculated with the SpekCalc code, with an Root Mean Squared (RMS) of 0.042, while the discrete part for the three targets was compared with the spectrum of the National Institute of Standards and Technology (NIST). In the calculations, 108 stories were used and a relative uncertainty less than 0.1% was obtained. The tomograph was modeled according to the characteristics for the thoracic tomography scan; with isocenter of 53.5 cm, 70 cm of opening of the gantry and a cylindrical source of beam width of 3 cm. The eye lens and thyroid were simulated on the Bottle Mannequin Absorber (BOMAB). As the source term, the X-ray spectrum calculated for the tungsten target was used. To calculate the absorbed dose, a diagnostic protocol was considered with 70 mAs, rotation time 0.6 s, total time of 5.71 s and pitch equal to 1. It was found that the absorbed dose that reaches the eye lens is 7.80 ± 0.08 mGy, corresponding to an effective dose of 0.936 ± 0.010 mSv, while for the thyroid a value of 76.63 ± 0.78 mGy, 3.07 ± 0.03 mSv, was obtained.

ABSTRACT. Introduction: Inorganic nanoparticles (NPs) based on high atomic number as Au arouse special interest for clinical application thanks to their potential application on both, diagnostic and therapeutic areas. In this work we assess the NPs clinical implementation possibilities, by imposing realistic values for the parameters involved in diagnostic and therapeutic stages.

Material and methods: In order to determine the AuNPs’ mass concentrations needed for diagnosis, CT imaging of a serial of AuNPs concentrations was performed with clinical scan parameters and simulating the inside-the-body nanoparticles distribution by using an experimental setup with a CT calibration phantom. AuNPs diameter was setup to 50nm and PEG3000 coated. Potential dose enhancement effect was evaluated by using Monte Carlo method to simulate a RT treatment for an actual Head&Neck case. AuNPs distribution in CT image was assumed as efficient as the one provided by a PET/CT study. Two different beams were evaluated: the conventional 6MV beam delivered by an ONCOR linac of SIEMENS and a FFF-1MV beam implemented in this linac for CBCT.

Results: Sigma Aldrich AuNPs with 2.23mg/mL concentration presented a difference of 50 Hounsfield Units regarding water, being this value stablished as the minimum able to allow a contrast in CT image. 6MV beam effect over dose enhancement was negligible even using the highest evaluated AuNPs concentrations (40mg/mL). Unlike this conventional therapeutic beam, the FFF-1 MV beam for CBCT showed significant dose enhancement even for the 2.23mg/mL concentration stablished as the threshold for diagnosis purpose. This dose enhancement around 10% was similar in both, 40mg/mL and 2.24 mg/mL for FFF-1MV beam.

Conclusions: Hopeful results encourage the use of AuNP as theragnostic agent in RT, since concentrations low enough for not generating toxicity were contrasted in CT image and also dose enhancement was achieved in combination with lower nominal energy beams than usual.

ABSTRACT. Introduction Scattered radiations from brain can degrade the image quality in brain SPECT imaging. Presence of SPECT collimator can inherently eliminate some of these unfavorable scattered radiations. One of the SPECT collimator types which is widely used for functional brain imaging is the Fan-bam collimator. The aim of this study is to quantitatively evaluate the effect of scattered radiation on image quality in planar brain imaging with Fan-beam collimator. Materials & Methods A commercial gamma camera in conjunction with Fan-beam collimator were simulated by MCNPX code. Effects of radiation scattering on image quality were evaluated through employing a Snyder phantom and comparing the system response to an isotropic 99mTc point source in both spatial and frequency domains. The FWHM and FWTM of obtained PSFs at different distances of 2, 4, 6, 8 and 10 cm were studied in spatial domain, while the spatial frequency at 0.9 (SF0.9) and 0.1 (SF0.1) of maximum MTF at mentioned source to collimator distances were considered at frequency domain. Results Maximum difference between obtained FWHM in presence and absence of phantom was about 5%, while this difference was equal to 14% for FWTM. The analysis of system response in frequency domain demonstrated a great difference of 43% between obtained SF0.9 in presence and absence of phantom. On the other hand, this difference was small and about 2% for obtained SF0.1. Discussion & Conclusions Radiation scattering mainly degrade the image contrast resolution and has no considerable effect on spatial resolution of acquired images by Fan-beam collimator. The impact of radiation scattering on image quality were more obvious in frequency domain and SF0.9 can be considered as an operational parameter for quantitative assessment of radiation scattering effects on image quality in frequency domain.

ABSTRACT. The present work is to generate the neutron fluence response matrix of a passive Nested Neutron Spectrometer (NNS) system with gold activation foils as thermal neutron detectors using Geant4 (version 10.4. patch-2). The NNS, which is useful for measuring neutron spectra around radiotherapy accelerators, consists of seven concentric cylindrically-shaped moderator shells around a thermal neutron detector. Simulations were performed using 52 incident neutron energies that spanned the entire energy range of interest for medical linear accelerators; from 1 meV to 16 MeV. In our model, the high precision QGSP_BIC_HP physics list is used for all interactions except neutron elastic scattering below 4 eV, for which the ‘G4NeutronHPThermalScattering’ model is used instead. The gold foil is placed at the centre of the moderator shell configuration and the neutrons are made to have random incidence on the detector to achieve a near-isotropic fluence distribution. In order to reduce computational time and minimize statistical uncertainty, two simple variance reduction techniques were adopted; geometrical splitting and the Russian roulette. In Geant4, these techniques are implemented together as the Importance Biasing method. The geometrical splitting was performed in a "parallel geometry" so that the biasing could be performed on a different geometry than the actual NNS geometry. The neutron fluence response matrix of the passive NNS system with gold activation foil is obtainable by modelling all seven moderator configurations and the bare gold foil for the 52 energy bins.

ABSTRACT. Introduction

In diffusing alpha-emitter radiation therapy (DaRT) (Alpha TAU, Israel) seeds impregnated with small activity of 224Ra are placed inside the tumor. 224Ra releases short-lived alpha-particle emitting atoms, beta-particles and gamma-rays in its decay process. Although the range of alpha-particles is a couple of cell diameters, these atoms diffuse inside the tumor contributing to a high-dose region up to five mm around the seed. To void cold spots, seeds should be placed five mm or less apart. The purpose of this study was to through Monte Carlo (MC) simulations investigate use of a gamma camera to verify the seed position.

Materials & Methods

MC simulations were performed with GATE. Diffusion was not modeled. A stainless-steel seed (outer-diameter 0.7 mm, height 20 mm) with 20uCi of 224Ra homogeneously distributed on its surface was placed inside a 20x20x20 cm3 water phantom. 224Ra decay products and dose around the seed was scored. A gamma camera was simulated with a high-energy lead collimator. Two seeds were placed inside the water phantom, five mm apart. Two different energy-windows (235-245 and 505-515 keV) were chosen. Photon fluence was scored before and after the collimator.

Results

The decay chain of 224Ra results in alpha-particle emission (5.45 to 8.78 MeV), beta-particles (maximum energy 2.2 MeV) and gamma-rays (maximum energy 2.61 MeV). In addition, bremsstrahlung is produced in the water. The scored photon fluence before the collimator has three distinct peaks (0.240,0.583 and 2.61 MeV). After the collimator low-energy photons are filtered out. However, high-energy part of the spectrum leads to a noisy image regardless of the energy-window.

Discussions and Conclusions

Due to the low resolution of the gamma camera and complex DaRT photon spectrum, it is not possible to verify the placement of the DaRT seeds. Compton camera would be a better imaging modality for this application

ABSTRACT. In proton therapy, accurate determination of a relative stopping power (RSP) map of the patient is required for treatment planning and treatment plan adaptation. Proton computed tomography (pCT) could achieve low dose and high accuracy RSP images at the treatment site. Previous studies on pCT accuracy indicate the potential of imaging with protons and heavier ions.

An important tool for understanding the potential and limitations of pCT scanners is their detailed simulation. We investigated the phase II prototype pCT scanner of Loma Linda University and U.C. Santa Cruz. To reproduce and understand image artifacts, we extended an existing Geant4 simulation platform to make use of a beam model directly created from experimental tracking data. Furthermore, the simulation was amended to reproduce non-linearities due to quenching and subtle geometrical detector effects. Artifacts, originating from inaccuracies in the look-up table-based calibration, were identified by gradually increasing the degree of realism in the simulation.

We quantified the mean RSP accuracy of the simulation and compared it to measurements. Experimental and simulated scans of a water phantom and several cylindrical phantoms with plastic tissue equivalent inserts were investigated. We identified distorted intervals of the water-equivalent path length (WEPL) and verified that these correspond to the artifacts in the image by reconstructing the distorted WEPL fraction of every voxel in the volume. Regions with a high fraction corresponded to regions in the image with low RSP accuracy. Mean RSP accuracy was 0.9% both in measured and simulated scans.

In summary, we present a detailed simulation of the phase II pCT prototype scanner. The extended simulation platform creates realistic raw detector data which agree with experimental data. The RSP accuracy of experimental scans can be reproduced and detector related image artifacts could be identified.

ABSTRACT. A medical linear accelerator generates therapeutic x-rays with megavoltage (MV) energies, delivering the prescribed energy pattern to the patient. An MV imager (EPID) can capture the therapy beam exiting the patient, which can be used to verify the complex intensity patterns delivered. However, the fluence entering the EPID is contaminated with patient-induced scattered photons, which reduces image contrast and reduces the accuracy of in vivo patient dose calculations derived from the EPID imaging. If the patient-induced scattered photon fluence can be estimated, it can be removed from the EPID images, improving image contrast and improving accuracy of in vivo patient dose calculations. Our group investigated a hybrid algorithm combining analytical (AnA) and Monte Carlo (MC) techniques to provide accurate and fast estimates of the unwanted patient photon scatter fluence entering the EPID. The Monte Carlo simulation was modified so the energy fluence entering the scoring plane was based on the cross section probability for the discrete direction exiting the second or higher order interaction site to each pixel of the detector and accounted for the attenuation along each individual path way. The calculated, multiply-scattered photon fluence was normalized to incident energy fluence and compared with an EGSnrc-based validation tool for two simple geometric phantoms using a 1.5MeV monoenergetic, divergent x-ray beam with a 10x10 cm2 field, source-imager-distance 150 cm, and 40x40 cm2 imaging plane. For the phantoms tested here, the mean percentage differences between the hybrid and EGSnrc predictions for the multiply scattered photons were less than 0.7% and standard deviation of the mean less than 1.7 %. This approach can accurately predict multiply-scattered energy fluence entering the imager. In future, this method will be verified in more clinically realistic situations (i.e. human anatomy and complex beam fluences).

ABSTRACT. Monte Carlo simulations are the researcher’s allies in medical imaging for the optimization of various imaging detectors, the improvement of resolution and detection sensitivity, or the development of new imaging systems such as multi-spectral CT, phase-contrast CT, proton imaging, Compton camera, in vivo range monitoring in ion therapy, etc. In order to propose computationally efficient simulations, researchers are proposing hardware acceleration solutions but also variance reduction techniques as hybrid modeling approaches combining either Monte Carlo and analytical simulations or using advanced machine learning processes such like artificial neural networks (ANN) or Generative Adversarial Networks (GAN). We will illustrate some last advances in the field on Monte Carlo simulations for medical imaging and dosimetry applications with the GATE Monte Carlo simulation platform. We will consider some developments related to theranostic approaches accounting for the simulation of dynamic processes and different resolution scales. We will show how the modelling of light transport in scintillation detectors plays an increasingly important role in detector design, we will illustrate how ANN can be used to model photon tracking in SPECT imaging or how GAN are used to replace phase space files usually produced to collect particles emerging from a voxelised patient geometry to simulate an imaging process. All presented methods are available in the open-source and collaborative GATE platform based on Geant4.

ABSTRACT. The Multi-layer Imager (MLI) is a novel megavoltage electronic portal imaging design composed of two or more layers of scintillators for increased detection and improved noise performance. Monte Carlo simulations suitable for imaging studies with the MLI are severely hindered by particle generation and tracking within each MLI layer. The aim of this work was to develop and validate a technique to facilitate rapid image generation for MLI based on the Monte Carlo method. The MLI prototype used in this study is composed of four layers. Each layer has a copper (Cu) build-up, a gadolinium oxysulfide (GOS) phosphor and an amorphous silicon (a-Si) detector for optical photon detection. The Geant4 Application for Tomographic Emission (GATE) was used as the base Monte Carlo tool for modelling and simulation of the MLI prototype. A specifically developed plugin within the GATE framework (“Actor”), was utilized to improve image generation time. The plugin utilized previously collected data that characterize the MLI, to form images at each MLI layer. The technique was validated against measured Modulated Transfer Function (MTF) curves. MV projections were simulated using Varian 6MV beam phase-space and the cone-beam 3D volume was reconstructed. The HU values of the reconstructed volume were also compared with measurements. There was good agreement with measured MTF and HU values. The average RMSE between simulation and measurement was 0.017 for MTF and 74 for HU. The estimated time reduction to generate projection images using the rapid technique compared with conventional Monte Carlo simulation was on the order of 100. A rapid technique suitable for MV image simulation based on the Monte Carlo method was validated against MTF and HU measurements for a novel MLI design with four layers. The technique can be applied to any multi-layer imager model with any possible combination of layers and materials.

ABSTRACT. Phase and dark-field images acquired with a grating interferometer using conventional x-ray tubes provide complementary information compared to absorption imaging alone. Therefore, x-ray grating interferometry (GI) has great potential for medical applications, such as GI breast CT (GI-BCT), for which a prototype is currently under construction at Paul Scherrer Institute. In the development process of GI based medical devices, simulations are a useful tool in design optimization concerning image quality, scattering contribution and dose deposition. Accurate modelling of photon transport in GI setups requires the combination of scattering and interference phenomena in one algorithm. Numerous Monte Carlo (MC) algorithms implementing different models for diffraction for the simulation of GI x-ray imaging systems have been presented. However, most of those algorithms suffer from long computation times or cannot provide consistent physical models implemented for combined scattering and interference processes. In this work and based on fundamental quantum mechanical concepts, including Feynman’s path integral approach to quantum mechanics, a new algorithm was developed relying on a more consistent physical description of photon transport in GI setups. This sum over paths algorithm is implemented as a photon transport extension library for EGSnrc, a well-established photon MC transport code. First validations by comparing results with a wave propagation algorithm suggest that the newly developed algorithm is capable to model interference effects occurring at gratings. In addition, it was possible to retrieve phase and absorption images from simulated detector signals in academic test cases. Dependent on parameter choices the sum over path approach has proven to be about 2.5 to 25 times faster than an in-house reimplementation of a previously published MC algorithm. Future development will include further performance increase, which is necessary to reach the cm-sized simulation volumes required for clinical systems. Furthermore, the algorithm has the potential to be extended for dose estimations.

ABSTRACT. Purpose: A major limitation for Monte Carlo (MC) simulations of electronic portal imaging devices (EPIDs) has been the long computation time required. Electrons contribute insignificantly to EPID images due to interactions within the patient and the shielding copper layer on top of the EPID. Therefore, terminating electron transport within phantoms is proposed as a method to shorten the simulation time without compromising image quality.

Methods: The Varian AS1200 EPID was modeled utilizing the Geant4 Application for Tomographic Emission (GATE), with both radiative and optical photon transport included. Varian TrueBeam 6 MV phase space was used as the source. Planar images of two phantoms, Las Vegas (LV) phantom and anthropomorphic pelvis phantom (left-to-right (LR) orientation and anterior-to-posterior (AP) orientation), were simulated conventionally using the AS1200 model. The simulations were repeated with electrons terminated within the phantom. This was accomplished by attaching the GATE actor, “Kill Actor”, to the phantom volume and selecting “electron” as the target particle. Phantom images were compared with and without electron termination in terms of signal-to-noise ratio (SNR), contrast-to-noise ratio (CNR), contrast, and gamma analysis. Improvement of the simulation time utilizing electrons termination was quantified by comparison with conventional simulation.

Results: Simulation time was reduced by 30%-45%, depending on the phantom thickness along the beam axis. Planar image quality wasn’t compromised. The electron termination technique provided similar SNR, CNR, and contrast compared to the conventional simulation. With 3%/0mm gamma criteria, the overall passing rate was 100% for the LV phantom image, 95% for the pelvis phantom image in AP and LR orientations.

Conclusion: By terminating electron transport within the phantom, EPID image simulation can be easily sped up by up to 45% without compromising image quality. Multiple applications, such as imager design optimization and volumetric image simulation will benefit from this technique.

ABSTRACT. Introduction: Monte Carlo (MC) simulations are a powerful tool for improving image quality in x-ray imaging modalities. An accurate x-ray source model is essential to a complete MC model for CBCT but can be difficult to implement on a GPU while maintaining efficiency and memory limitations. A statistical analysis of the photon distribution from a MC x-ray tube simulation is completed in hopes of building a compact source model. Materials & Methods: MC simulations of an x-ray tube were carried out using a modified version of BEAMnrc. A phase-space file was collected at the exit of the tube and photons sorted into three categories: primary, tube scatter, or off-focal radiation. The covariance and correlation of the components [energy (E), position (x,y), and direction (u,v)] were computed for each photon category. PCA and ZCA whitening transforms were calculated along with estimates of the resulting correlation matrices after the transform. Results: The correlation matrices showed all photon groupings had a relationship between x and u, as well as y and v. This correlation was largest in the primary data as the location of the photon at the exit plane of tube is directly related to its angular direction. There was also a minor correlation between E, x, and u in the data. Both PCA and ZCA whitening reduced the correlation between the new components allowing them to be treated as independent variables. Conclusion: Whitening transforms provide a potential method of reducing the memory required to accurately simulate a x-ray source in a GPU MC system. Instead of loading a pre-computed phase-space or multivariate quantile function a set of 15 quantile functions (5 per photon group) can loaded and sampled using MC methods. This method could reduce memory requirements by almost 6 orders of magnitude, going from gigabytes of data to kilobytes.

ABSTRACT. The use of antiscatter grid is a well-known method for deterioration reduction in the image quality caused by scattering. However, this method requires a higher beam intensity. Therefore, increasing the dose imparted in the patient. Pediatric patients are more radiosensitive and have longer life expectancy than adults. Thus, antiscatter grid use must be carefully analyzed. The objective of this work was determining the optimal antiscatter grid for pediatric patients via Monte Carlo simulation. The simulation geometry consists of a homogeneous acrylic phantom with 30x30 cm² area and thickness between 5 and 15 cm. A computed radiography detector with 40x40 cm² area and 300 µm thick. An antiscatter grid defined by grid ratio (r); strip density (N); interspace material, width (D) and height (H). Grid configurations were varied in the simulation to determine which one produces the best performance. Polyenergetic beams with tube potential from 40 to 130 kV without added filtration were used. Grid performance was evaluated by quantifying the signal difference-to-noise ratio improvement factor (SIF). Results show that the interspace material, N and H are the most important grid performance factors. Cotton fibre and aluminum interspace material yield the best and worst performances, respectively. Smaller H values are indicated for lower tube potentials while thicker interspaces are indicated for higher tube potentials values. A performance decrease arises from N increase. The use of antiscatter grids is not indicated for phantoms with thickness smaller than 10 cm. The antiscatter grid with the overall best performance was r = 10, D = 300 µm, H = 3 mm, N = 28 and cotton fibre interspace material with 49.1(0.7)% and 28.6(0.3)% image quality increase for the 10 and 15 cm phantom thickness, respectively. Next studies should focus in use other phantom materials, and analyze anthropomorphic phantoms.

ABSTRACT. We developed a Monte Carlo simulation platform for imaging and dosimetry in computed tomography dedicated to the breast (BCT), digital breast tomosynthesis (DBT) and 2D digital mammography. The multithread application is based on Geant4 toolkit vers. 10.00 and uses the physics list Option4; it runs on AMD Ryzen 7 2700 8-core processors (3.70 GHz) with 16 virtual cores, producing projection images with 7*10^4 photon histories/s for an average breast. The software permits to evaluate the mean glandular dose and the 3D dose distribution as well as to compute breast projected images, then reconstructed with either analytical or iterative algorithms. The breast models were taken from a dataset comprising 100 digital 3D breast phantoms from segmented images of healthy breasts acquired with the clinical BCT scanner at UC Davis at 80 kV. The voxels of the breast images were classified as air, skin, adipose or glandular tissue via a semi-automatic algorithm with voxel size in the range 0.25-0.35 mm. A software compression algorithm was applied to the 3D phantoms of the database in order to produce compressed breast digital phantoms for virtual mammography and virtual DBT imaging, with geometries and specifications of clinical scanners. Both monoenergetic (from 8 to 100 keV) and polyenergetic spectra were produced. Patient-specific MGD estimates, as well as simulated BCT 3D images were compared with clinical BCT scans. For image quality assessment, we simulated primary and scatter projections for: scatter influence studies and scatter correction techniques; X-ray spectrum optimization in terms of detail SNR and CNR analysis; embedded tumor phantoms; contrast-enhanced digital breast phantoms (with iodinated or gold nanoparticles agents). This platform will be adopted for the optimization of the imaging chain and for comparing BCT images to those produced with DBT and mammography scanners via virtual clinical trials.

ABSTRACT. The use and frequency of kV imaging in Radiotherapy has been growing rapidly, and is fully justified by the improved target localization that is provided. However, several studies published in recent years underlined that additional doses due to daily Cone-Beam Computed Tomography (kV-CBCT) imaging cannot be considered as negligible, in particular for populations with higher potential risks, such as children and young adults. The evaluation of doses delivered during CBCT imaging procedures has been conducted for long using either measurement or calculation-based methods, for various kV-CBCT systems and imaging protocols. However, there is currently no clinically available tools to estimate imaging doses in image-guided radiotherapy (IGRT). The following study presents a Monte Carlo (MC) based software that calculates imaging dose distributions in patients for four IGRT systems: OBI (kV-CBCT, Varian), XVI (kV-CBCT, Elekta), ExacTrac (2D kV-kV, BrainLab) and the Cyberknife 2D kV-kV imaging system (Accuray). The software was validated against OSL measurements in a pediatric anthropomorphic phantom (Grant, CIRS/ATOM). Measurements and calculations showed an agreement within a 20% range. This software is currently being tested in four French clinical centers to evaluate the impact of IGRT additional doses on the treatment dose volume histogram for real patients (pediatric and adults) treated for head, head-and-neck, lung and prostate cancers.

ABSTRACT. Triangle meshes are used in Monte Carlo Simulation (MCS) to model phantoms with fine elements and complex shapes. The main drawback of such geometry is the high number of triangles required to simulate a realistic surface. This sampling limitation may be critical for some medical applications (phase-contrast X-ray imaging, bioluminescence and fluorescence imaging, etc.). In this work, we propose to enhance the surface modeling of triangle meshes without increasing the number of triangles. During the particle collision calculation an interpolation mechanism was proposed to virtually simulate a continuous envelope surface that represent the mesh. The aim was to improve the surface definition without drastically increasing memory storage and computation time. We propose to adapt the position and the normal vector of the hit on the mesh surface by taking advantage of the local change of the phantom shape. This was done by parametrizing a quadratic polynomial surface according to curvature factors, normal vectors, and weights which depend on the hit position on the triangle face. This surface and the particle path were used to compute the new position and the new normal vector of the hit. For validation purposes, the MCS was simplified to consider only the geometrical aspect i.e. without physics effects. Results have shown a clear improvement on the surface modeling, more realistic, when the proposed method is used. The simulation time for a meshed thorax phantom was the same with or without the proposed method. Further work includes testing the proposed method in different configurations (objects, number of triangles, etc.) and considering different medical applications with physics effects in order to perform a complete evaluation study.

ABSTRACT. Monte Carlo (MC) simulations are ubiquitous in medical applications for modelling radiation transport and energy deposition across length scales from patients to subcellular structures to nanodevices.  With growing interest in, e.g., radiation response on subcellular length scales and developing new treatments targeting cellular structures, the range of applicability of multi-purpose MC codes is being pushed to lower energies and shorter length scales.  However, these developments present challenges as the transport of electrons is generally treated ~classically with radiation quanta modelled as point-like objects undergoing sequences of free-flight interrupted by discrete interaction sites. With the electron’s wavelength increasing with its decreasing kinetic energy, coupled with the nanometre-size of biological targets of interest, the applicability of the classical trajectory MC approach becomes questionable.  This presentation will focus on departures from traditional trajectory MC simulations of electron transport. We will consider the validity of trajectory MC simulations of electron transport in the context of quantum theory, and, on the basis of general arguments, describe the potential for quantum effects to emerge at sub-1 keV kinetic energies. As a full quantum theoretic treatment of electron transport in condensed (biological) media is too complex to be feasible at present, calculations within a simplified model consisting of a plane wave electron incident on a cluster of point scatterers representing a water droplet will be presented.  Within this model, quantum mechanical (QM) calculations will be contrasted with the corresponding trajectory MC results, considering varying water molecule (scatterer) elastic/inelastic cross sections, droplet structure (minimum separation between molecules in the droplet constrained by some threshold or random arrangement), as well as droplet size and shape. Differences between QM and MC results are generally larger when inelastic scatter and/or medium structure are included in simulations. Across the parameter space, relative errors on MC results are typically only sub-1% once electron energy exceeds 1 keV, and hence convergence of classical MC with QM calculations only occurs near 1 keV.  Ongoing research to develop more realistic models of low-energy electron transport in condensed media will be described.

ABSTRACT. Monte Carlo (MC) simulations play an important role in nuclear medicine (NM) research, in both single-photon emission computed tomography (SPECT) and positron emission tomography (PET). Simulations are routinely used in the design of new imaging systems, development and validation of image reconstruction algorithms as well as data processing and analysis methods. Further, applications of MC simulations include calculations of internal radiation doses absorbed as a result of diagnostic procedures and, more importantly, in radionuclide therapies. Finally, they are also used in the investigations of cyclotron production of medical radioisotopes. The research program of our Medical Imaging Research Group covers all these areas. In my talk I will discuss all these applications and present examples of MC simulation studies performed by us and others.

ABSTRACT. Given its excellent resolution versus sensitivity trade-off, multi-pinhole SPECT has become a powerful tool for clinical imaging of small sections of the body such as the brain. A next-generation, adaptive, multi-pinhole system, AdaptiSPECT-C, dedicated to brain imaging is currently under investigation by our research team. The pinhole geometry has proven to be an essential key to enhance image quality. Indeed, the use of keel edge pinhole has the potential to reduce the penetration of gamma rays through the edge of the pinhole aperture, as well as the amount of scatter within the system. In this study, we investigated the imaging performance using the XCAT brain perfusion phantom of two pinhole profiles for AdaptiSPECT-C, one based on a keel edge aperture and the other one based on a knife edge collimation. A range of aperture diameters from 4 to 7 mm for each design was studied. An approach developed by our group to modeling the system response using GATE simulation, was employed to compute accurately the system matrix for the knife and the keel edge designs. In addition to the geometrical response of the system, all the effects degrading reconstruction were incorporated into the matrix. Several acceleration techniques, such as forced detection, system symmetries, and heavy computational resource (cluster) were efficiently used in a way to obtain a low noise matrix within an acceptable time (~2hrs). We demonstrated through GATE simulations that the use of a keel edge profile for AdaptiSPECT-C leads to superior imaging performance compared to knife edge collimation in case of clinical 123-I brain imaging. We are currently working on performing a task-performance Channelized Hoteling Observer (CHO) study of defect detection in perfusion (strokes).

ABSTRACT. Compton Cameras are gamma-ray imaging devices, which consist of a stack of detector-layers working in time coincidence. They have been widely employed in astronomy and homeland security. In recent years, due to the developments in detector technology, there has been a growing interest in their application in the medical field. In order to design and optimize the performance of such a system, Monte Carlo simulations play an essential role. GATE is a medical oriented open-source simulation toolkit based on Geant4, which allows the simulation of a broad range of experimental imaging settings by using macros. In this work, we present an extension of GATE, which already contains modules for PET and SPECT imaging, with the aim of supporting Compton camera simulations and their analysis. The developed module is designed so that the interaction information of the particles in the specified detector volumes is stored and digitized to simulate their response. Several digitization processors have been implemented to reproduce the performance of the most commonly employed detectors such as strip detectors, pixelated and monolithic scintillator crystals. Additional tools have been included to facilitate the access to ground-truth information with the aim of recovering the ideal Compton kinematics and characterizing the possible sources of degradation in an experimental device. The module has been successfully validated against experimental data taken with a Compton Camera prototype based on LaBr3 continuous crystals coupled to SiPMs, built at IFIC-Valencia. On the other hand, the developed tools have proven to be helpful to identify and reduce the different sources of degradation found in the acquired data. The versatility of the module has been proven through a preliminary simulation study of the comparison between the performance of two different prototypes. The module will be available in the next GATE release.

ABSTRACT. Context and objectives The use of Monte Carlo (MC) techniques is highly desirable for personalized dose calculations in radiation oncology and also in emerging treatment modalities such as peptide receptor radionuclide therapy. Patient-specific, MC-based calculations can contribute to more accurate dose estimates and ultimately to a better understanding of dose-effect relationships in this relatively new field. MC codes must be fast and accurate in order to be integrated to clinical activities. The purpose of this study is to validate in standardized conditions irtGPUMCD, a fast GPU-based, in-house MC code dedicated to internal dosimetry. Methods The adult female phantom of the International Commission on Radiological Protection (ICRP) 110 report was used as a benchmarking geometry for this study focused on 177Lu. Fifteen anatomical structures in the trunk region were considered as targets, with the liver and the kidneys used as source organs. The S-values for cross and self-dose were evaluated with irtGPUMCD simulations of 108 photons, with gamma branching ratios of 177Lu taken from ICRP107. The self-dose from betas was calculated for source organs only, based on local deposition of dose in irtGPUMCD. The S-values relative errors were evaluated between irtGPUMCD and IDAC-DOSE2.1, an internal dosimetry program based on the ICRP specific absorbed fractions. Results and discussion The relative error between IDAC-DOSE2.1 and irtGPUMCD S-values was below 4% for gamma cross and self-dose and below 1% for beta self-dose. The maximum uncertainty was 0.2%. The time required on average for 108 photons was approximately 13 min on a single GPU, scaling down to 6.5 min in a dual-GPU configuration. These results are promising and let envision the use of irtGPUMCD for internal dosimetry in 177Lu treatments. However, a more thorough validation is required with a larger set of simulations involving different source-target organs for the female and male ICRP110 phantoms.

ABSTRACT. Introduction. Clinical dosimetry of 177Lu-labeled peptides used in Molecular RadioTherapy (MRT) is based on SPECT activity quantification. Monte Carlo (MC) modeling of SPECT imaging aims at improving SPECT quantification in complex geometries. So far, SPECT modeling using GATE (Geant4-based Monte Carlo code) cannot account for non-circular orbit acquisitions. In addition, to avoid volumes overlap in the voxelized phantom geometry, the detector heads are usually positioned at a larger distance from the phantom (representing the patient). The aim of this work is to improve SPECT modeling of 177Lu imaging in GATE, by implementing the auto-contouring acquisition mode. Materials & Methods. GATE was used to simulate SPECT images of a Jaszczak phantom with two 3D-printed organs (spleen and kidneys) included. To bring the detector closer to the phantom and to avoid volumes overlap, the voxelized phantom (contained in a bounding box) was replaced by tessellated meshes created from the phantom CT image. The Siemens Symbia T2 dual headed gamma camera (5/8” NaI(Tl) crystal and MELP collimator) was modeled. Detector head parameters (radial position and the acquisition angle) from the experimental DICOM images were extracted to model the exact non-circular motion of the camera. Sixty projections per head were generated using the 208 keV emission of 177Lu, with an energy window width set at 15%. Results and Conclusions. We developed the methodology to simulate auto-contouring acquisition in SPECT imaging to obtain realistic images using tessellated meshes within GATE. Profiles from experimental images (obtained within the EMPIR MRTDosimetry project) and simulated images were compared to validate our preliminary results. Additionally, the Structural Similarity Index (SSIM) was used which revealed 97% similarity between the measured and simulated images for this Siemens gamma camera model. This implementation can be extended to any gamma camera or phantom for SPECT imaging.

ABSTRACT. Introduction Texture analysis (TA) is an effective technique for the evaluation of tumor spatial heterogeneity in oncologic positron emission tomography. The feasibility of TA for myocardial perfusion single-photon emission computed tomography (SPECT) study has remained to be clarified. In this work, we used SIMIND Monte Carlo code and a digital XCAT voxel phantom and evaluated TA in myocardial perfusion SPECT (MPS). Materials & Methods One normal and 16 abnormal cardiac XCAT phantoms with 256×256 matrix size, 256 slices and 4.1 mm3 of voxel size were generated. These male digital phantoms simulated 100, 20, and 10 activity/voxel of 99mTc radiopharmaceutical in the heart, liver, and lung, respectively. Using the SIMIND Monte Carlo code, projection datasets with 64×64 matrix size and 6.6 mm of pixel size were generated from the 17 XCAT phantoms. The LIFEx freeware (version 4.09) was used to perform TA with 31 textural features. The textural features were calculated based on 4 matrices: gray level co-occurrence matrix (GLCM), gray-level zone length matrix (GLZLM), gray-level run length matrix (GLRLM), and neighborhood gray-level different matrix (NGLDM). Results Normal and abnormal myocardial perfusion patterns were simulated using the SIMIND and XCAT phantoms (Figure 1a). Textural features were obtained from the normal and abnormal myocardial perfusion patterns and classified using two-way cluster analysis (Figure 1b). Significant differences of textural features between normal and abrnomal perfusion patterns were observed in homogeneity, contrast, correlation, and dissimilarity for GLCM (p < 0.0041); SRE, LRE, HGRE, SRHGE, LRHGE, GLNU, RLNU, and RP for GLRLM (p < 0.0235); contrast and busyness for NGLDM (p < 0.0189); and SZHGE, LZLGE, and GLNU for GLZLM (p < 0.0230). Conclusions TA is helpful for determining myocardial perfusion abnormality based on the textural features. The selection of optimal textual features should need to perform TA in clinical MPS study.

ABSTRACT. Intensity modulated proton therapy (IMPT) is an advanced form of radiation therapy. While the radiobiological effects of IMPT depend primarily on the physical dose distribution, studies have shown that linear energy transfer (LET) plays an important role as well. We developed a hybrid method to calculate 3D linear energy transfer (LET) in patient geometry for IMPT. First, we used a well-bench-marked Geant4 code to generate 3D LET distributions in water for the proton beamlets. Second, the LET kernels were fitted via 'error function' and incorporated in the Treatment Planning System (TPS) using ray-casting algorithm. Since the LET kernels were pre-calculated, the time required to compute LET in patient geometry was greatly reduced. It is found the LET distributions calculated by the hybrid method agreed well with those found using a full Monte Carlo code.These LET distributions computed using our hybrid method were used to evaluate potential clinical benefits and toxicities for various tumor sites including lung, head and neck, esophagus, and brain. The LET calculation code has also been used in IMPT treatment planning, allowing for radiobiological optimization by including LET-weighted constraints in the inverse treatment planning process.

ABSTRACT. In protontherapy, secondary particles are produced through primary beam interactions with the patient’s body. The fragments created in inelastic interactions of the beam with the target nuclei have low kinetic energy, high atomic number and high LET as compared to primary protons. Secondary particles produce an altered dose distribution, due to different ranges of fragments. The range of these fragments is of the order of 10-100 μm so they are in general confined within a single cell. They have high LET, locally leading to an increase of RBE for the same delivered dose. The energy dependence of the nuclear interaction cross section makes target fragmentation relevant mostly in the entrance region, therefore for normal healthy tissues. The inclusion of target fragmentation processes can be important for the accurate calculation of the dose in the treatment. Nowadays, target fragmentation is not implemented in commercial TPSs (Treatment Planning System). Furthermore, the production yield of fragments at therapeutic energy is still poorly measured. In this work Monte Carlo (MC) simulations were employed to estimate the effect of target fragmentation in Protontherapy. The production cross sections of fragments at therapeutic energies have been evaluated by means of FLUKA MC code. Starting from energy distribution of fragments, the range distribution has been derived. Furthermore, the fluence of target fragments at different depths has been calculated, in order to evaluate the contribution in the entrance channel and in the Bragg peak region. From these results, it emerges that main fragments are secondary proton, but a significant contribution to the dose distribution is also given by Helium fragments. In order to asses the biological impact of target fragmentation, LET distributions have been derived including the secondary particles, so to estimate RBE as a function of depth by means of different models.

ABSTRACT. Recently, the application of analytical dose calculation (ADC) algorithms for proton pencil beam scanning dose optimization has been questioned, and concerns regarding the accuracy of these algorithms, especially for fields with pre-absorber, oblique treatment angles and big air-gaps have been raised. We therefore use a previously validated Monte Carlo (MC, TOPAS/Geant4 based) setup of our proton Gantry to investigate two mathematical ADC approaches, namely the ray-casting (RC) and the pencil-beam deconvolution technique (PB). In this retrospective study, we investigate RC and PB for seven cranial patients (2 skullbase, 4 brain, 1 nasal cavity; 14 plans consisting of 45 fields) previously treated at our institute. These include fields without pre-absorber (air-gaps between 14 cm to 27 cm), fields with pre-absorber (air-gaps between 13 cm to 25 cm), and fields with an automatic pre-absorber for superficial spots only (air-gaps between 10 cm to 13 cm). For all patients, more than 97%/96% (minimum values, treatment in the nasal cavity) and on average 99%/98% (mean values) of the voxels agree within ±5% between MC and RC/PB. The agreement does neither depend on the pre-absorber setting nor on the air-gap. This favorable ADC performance when compared to previous studies on the PB algorithm with pre-absorber might be explained by the decomposition of pencil beams, which in our study is done in the patient only, and not upstream of the pre-absorber. The agreement to the MC simulation does not substantially differ between RC and PB, RC calculations are however more than a factor 15 faster than PB. In conclusion, MC techniques demonstrate that with ADC algorithms, good clinical agreement can be achieved for cranium patients, even when using a pre-absorber, oblique treatment angles and large air-gaps.

ABSTRACT. Introduction

The verification of a protontherapy plan by measuring the proton range with a PET detector needs a precise Monte Carlo simulation, to understand how much the measured PET activity distribution differs from the expected one, and how it translates to dose. In this work we quantify how much the uncertainties in the nuclear cross sections may affect the proton range verification and propose a tool to help in the path to diminish this problem.

Materials & Methods

We have used GAMOS_6.0.0, based on Geant4_10.04.p02. The 29 different nuclear physics lists can be summarized in: G4QGSP_BIC_HP, G4QGSP_BERT_HP, G4QGPS_BIC_AllHP and G4FTFP_INCLXX_HP. We have simulated an active scanning treatment on a patient head entering by the left side, close to the ear, so that different areas of tissue, air and bone are traversed.

Results & Discussion

Depending on the selected voxel line cross positions and the time patterns beam-on/beam-off, we found a wide dispersion of the distal end differences among the four physics lists. These differences can be as high as 5-10 mm, much bigger than the precision we obtained reconstructing the PET images, ~1 mm, similar to what other studies obtain.

These differences are due to the substantial differences in the cross sections used by the different physics lists. And none of the physics lists cross sections have a good agreement with experimental data. Our proposal lies on the direct use of the experimental cross sections in the Geant4 code. We have developed a tool that transforms a cross section file as those exported by EXFOR or JANIS, or a user-made one, and is used by the physics list we developed for Geant4, G4QGSP_BIC_AllHP. We also provide a new cross section database with the best match to the positron production experimental data.

ABSTRACT. Introduction: The use of titanium implants during surgery of brain malignancies may cause detrimental dose perturbations in the subsequent proton therapy treatment. In this study we compare the dose calculation by a commercial Treatment Planning Systems' (TPS) with Monte Carlo simulations and with film measurement for proton beams traversing a CranioFix titanium implant Materials and Methods: The resulting dose perturbation was investigated through EBT Gafchrome film measurements, TOPAS Monte Carlo simulation and a clinical proton convolution superposition algorithm (PCS 13.7.16 Eclipse). A single energy beam (80 MeV) was planned along the symmetry axis of the CranioFix for both TOPAS and Eclipse. Results: For the TOPAS simulations and Gafchrome measurements a dose dip along the central axis is visible. The TOPAS simulations further showed that this dip arises from the primary incoming protons and not the secondary particles. The PCS algorithm does not show the same dip rather the proton range along the central axis is markedly shortened compared to measurements and TOPAS simulations. The dip was 25% and 42% of the maximum dose for measurement and TOPAS, respectively. The central axis doseprofile shows large variation between the three techniques. The range difference with and without the titanium implants were 3.6mm, 1.3mm, 3.4mm for TOPAS, Gafchrome and Eclipse, respectively. Discussion and Conclusion: The dose perturbation caused by in-field presence of the titanium craniofix is significant.The Monte Carlo simulation in TOPAS was able to catch the dip in dose along the central axis, however the was a slight discrepancy in the depth of the dip compared to film measurements. This may be caused by a lack of simulated histories since the dose will increase while the dip stays the same. The cause of the range differences between measurement and simulations of around 2-3 mm is less obvious and will require further investigations.

ABSTRACT. Introduction Photon scattering contributes significantly to the imaging degrading effects in 3D PET imaging. It results in a loss of contrast and overall image quality which makes accurate tracer quantification challenging. The objective of this study is to develop a Monte Carlo (MC) based model to validate the calculated true and scatter distributions with measured estimates in order to develop a physics based scatter correction algorithm. To investigate scatter effects in PET imaging the MC simulation tool GATE 8.0 has been used. The physics modelling is based on Geant4 and the clinical scanner geometry is a pre-defined GE Discovery 610 PET/CT detector model.

Results Results have shown that the MC model can be used to model voxelized geometries and sources. A platform has been established to isolate the scatter distribution in projection and in image space. A validation workflow has been developed to calibrate the MC simulation data based on its input in image space. Preliminary results show an agreement between MC simulation and its input of 2.2% for a NEMA line source simulation.

Discussion & Conclusions Further MC simulations and measurements of the NEMA body phantom are need to validate the MC simulation with clinical scanner measurements. This will lead to further investigations of the MC isolated scatter distribution in image and projection space.

ABSTRACT. Computer simulations are used in many areas of research and development.

Specifically, Monte Carlo simulations allow the precise simulation of experimental conditions. 

A properly benchmarked Monte Carlo system can thus save beam time for experiments or create potential scenarios that are difficult to create experimentally. Computer simulations are particularly important in a field such as radiation therapy where it is important to evaluate the dose to a patient before the irradiation.

In particle therapy, in particular, Monte Carlo simulations are used for different tasks such as the evaluation of the dose delivered to patients, study the physics of proton/ion beams, design the beamlines, design of the treatment heads, quality assurance and finally, many other specific research issues. 

In this talk, I will try to illustrate the status of the art in the use of Monte Carlo simulations in particle therapy reporting experiences from different groups around the world.

ABSTRACT. Within the field of radiotherapy, tissue-equivalent materials allow easy setup as well as allow for the construction of anthropomorphic phantoms for dosimetry. These materials should, ideally, have absorption and scattering properties which are the same as the body tissue. A good range of tissue-equivalent materials have been developed for photons over the past 50 years. However, these materials have not been optimised for the newer modality of proton beam therapy. An analytical model and FLUKA simulations have been developed to test tissue-equivalent materials in comparison to human tissues for protons. So far research has been completed for three commercially available materials that have been designed to be bone-equivalent for photons, two epoxy resin materials (SB5 Hard Cortical and CIRS Cortical) and a 3D printable material called Accura Bluestone. The absorption and scattering properties (such as stopping power, range, scattering length and fluence correction factor) of these bone-equivalent materials have been compared against ICRP cortical bone. Results so far from analytical model and Monte Carlo simulations show that SB5 and CIRS are the most equivalent to cortical bone within 3% for all dosimetric parameters. Thus, further optimisation is required to improve these materials for dosimetry (within 1%) of proton beam therapy. Work is ongoing to optimise 3D printable materials so that they are suitable and viable for proton beam therapy. Further research is also ongoing into the simulation of other new compositions to create new formulations that are improved for proton therapy beams.

ABSTRACT. There are a number of key uncertainties associated with proton beam therapy compared to photon therapy. Mitigation of these uncertainties is attempted through robust optimisation at the treatment planning stage. Anatomical changes, however, occurring over a fractionated course of treatment are difficult to predict and account for. Such anatomical changes can have serious clinical consequences when treating with protons, namely the delivery of high dose to healthy tissue. To help detect these anatomical changes in a non-invasive way, without extending the overall treatment time, a new gamma-ray detection system is being developed.

We report the first results of a new method to determine proton beam range in three dimensions, for pencil-beam scanning systems. The range is determined through the reconstruction of the origin of prompt gamma (PG) rays emitted from nuclear de-excitations following proton bombardment. The prototype system is comprised of 16 symmetrically-spaced LaBr3(Ce) detectors, in a symmetrical design. The position reconstruction capability of the detector system was initially investigated by means of Geant4 simulations. To determine the PG-rays emission positions in 3D, the information recorded by each detector is fed into an in-house developed reconstruction algorithm.

The Geant4 simulated data show that, with realistic detector characteristics, the reconstruction algorithm is capable of determining an isotropic point gamma-ray source with an uncertainty of less than 1 mm in 3D space. The reconstruction algorithm has been empirically validated and an in-silico investigation has now begun to ascertain the system performance with a spot scanning proton beam therapy system. The simulation results to-date along with the empirical validated will be shown and discussed.

ABSTRACT. Multiple Coulomb scattering (MCS) is the most frequent process when Monte Carlo simulating the transport of therapeutic protons (<300 MeV) through a medium. It leads to a succession of many small angle deflections and causes the protons to follow curved trajectories. As part of a work in the context of proton computed tomography, we developed a method to analytically sample proton trajectories under the influence of MCS. Currently, the formalism only works in media without material heterogeneities in the direction transversal to the proton beam. Discrete nuclear scattering events are not modeled. We think our work would be of interest to the community of Monte Carlo users and developers, e.g. to accelerate simulation of MCS.

Our method models the likelihood of a proton trajectory based on its integrated curvature weighted by an energy and thus depth dependent function which can be easily parametrized, either analytically or through a Monte Carlo simulation. The proton trajectories are approximated as polynomial functions of depth. In this way, sampling trajectories is reduced to sampling polynomial coefficients according to an N-dimensional Gaussian distribution, where N is the polynomial degree. We find that N=4 or higher is sufficiently accurate for medical physics applications where simulation grids are rarely finer the 1 mm. The spatial distributions of analytically sampled and Monte Carlo simulated trajectories agree to better than 5%. With the analytical method, millions of trajectories can be generated in a matter of seconds on a single CPU. The method could be a starting point for variance reduction techniques or useful as forward projector when simulating proton tomographic images.

ABSTRACT. The Monte Carlo (MC) simulations can serve as an independent dose calculation over the course of radiotherapy treatment, from the plan stage to quality assurance (QA) stage. To utilize a general-purpose MC code in a clinical workflow, an interface that converts the plan and treatment information in DICOM format into MC components such as geometries and beam source is a crucial element. For this purpose, a DICOM-RT Ion interface has been developed and integrated into TOPAS MC code to perform such conversions in real time. DICOM-RT objects supported by the interface include Ion Plan (RTIP), Ion Beams Treatment Record (RTIBTR), CT image, and Dose so that MC simulations can calculate a patient dose as well as dose in water phantom with a planned and/or a delivered fluence map.

The DICOM-RT Ion interface is commissioned for two treatment machines. It is successfully deployed within our automated job submission workflow. When a new RTIP or RTIBTR is stored, the MC runs automatically and creates simulated dose and LET distributions on the patient CT or water phantom. The simulated dose is used during the planning stage to evaluate the analytic dose calculation of the TPS when the target is near air cavities or highly heterogeneous medium. It is also used during the plan review and QA stage as an independent dose calculation and/or log file analysis of the delivered dose.

This study demonstrated the importance of DICOM-RT Ion interface to adopt a general-purpose MC code as an independent dose calculation and the deployment of MC as a part of routine clinical workflow may enhance the efficiency of quality assurance for proton pencil beam radiotherapy.

ABSTRACT. TOPAS wraps and extends the Geant4 Simulation Toolkit to make advanced Monte Carlo particle transport simulations easier to use for medical physicists and radiation biologists. While TOPAS was initially funded to focus on proton therapy research, it has since expanded into a tool that can model any treatment or imaging system. TOPAS can import patient geometry from any imaging modalities, score dose, fluence, etc., save and replay advanced phase space, provides advanced graphics, and is fully four-dimensional to handle variations in beam delivery and patient geometry during treatment.

Funded by the US NCI's initiative ‘Informatics Technology in Cancer Research’ (ITCR), TOPAS is free and well supported for all non-profit users in medical physics and radiation biology worldwide. As of February 15, 2019, TOPAS has 610 licensed users at 252 institutions in 37 countries.

We will show how TOPAS lets users create entirely novel setups from adjusting minor details to modeling whole new therapy and imaging machines with no need to learn any programming language. All aspects of simulation are controlled from a unique system of control files, the TOPAS Parameter Control System. An optional GUI makes setup even easier. The modular nature of the control system architecture is well suited for collaborative research.

While few users find it necessary, users who wish to write C++ code (to exploit Geant4's most obscure features) can use TOPAS' Extensions Mechanism to have their own code merged with the rest of TOPAS. We will show how user code still takes full advantage of the Parameter Control System, so new parts interact seamlessly with the rest. Users can easily share their extensions with others.

Further, we will demonstrate how TOPAS is being used in: • proton therapy • heavy ion therapy • electron and x-ray linacs • brachytherapy • imaging • microdosimetry • radiation biology

ABSTRACT. Purpose: Proton minibeam radiation therapy (pMBRT) is an innovative approach to minimize the side effects of radiotherapy [1,2] which has been implemented at the Orsay Proton Therapy Center [3,4]. pMBRT combines the normal tissue sparing provided by parallel submillimetric beams spatially fractionated with the benefits of proton therapy (finite range and Bragg peak depth-dose profile). We present here a dosimetric comparison of possible treatment plans in patients receiving pMBRT.

Method: Monte Carlo simulations (TOPAS v3.1.2) were used to evaluate the dose distribution delivered in patients developing high grade gliomas for different treatment plans at the Orsay facility. For each plan planar pMBRT (pencil beam scanning system with the use of multislit collimators [4]) and Single Field Uniform Dose (SFUD) delivery were used. A hybrid analytical and Monte Carlo intensity modulated dose calculation algorithm was developed to optimize the spots positions and relative weight of individual Bragg peaks composing the field. The dose-volume histograms of these plans were finally calculated and compared to the one given by a broad-beam TPS.

Results: For each plan, a good homogeneity (±5% variations in the dose) was reached in the targeted region while maintaining spatial fractionation at shallow depths in pMBRT. The optimal parameters (slits width, distance between iso-energy layers and spots positioning) were assessed. Finally, the Peak-to-Valley Dose Ratio (PVDR) and valley dose were evaluated for each configuration.

Conclusion: The pMBRT delivery methods evaluated in this work opens the door to an efficient treatment of radioresistant tumors, such as gliomas [5,6] and will support the design of future clinical research.

[1] Prezado and Fois, Med. Phys 2013 [2] Prezado et al. Scie. Reports 2017 [3] Peucelle et al. Med. Phys 2015 [4] L. De Marzi et al. Med. Phys. 2018 [5] Prezado et al, Scie Reports 2018 [6] Prezado et al, IJROBP in press

ABSTRACT. Monte Carlo (MC) simulations are essential for dose calculation in proton beam therapy (PBT) as they reliably address the complexity of the interactions of protons in heterogeneous media. Their use in clinical routine is not a standard yet due to the long computation times.

The MC code FRED is a GPU-accelerated, trimmed-down dose calculation engine developed at Sapienza University of Rome for PBT applications. The implementation of physics processes contributing to the beam modeling and the dose deposition, like nuclear interactions with target nuclei and the trajectory deflection via a multiple Coulomb scattering, are validated against general purpose MC codes and experimental data. The parallelization of particle tracking on GPU enables fast computation times whilst preserving a high accuracy on the dose calculation.

We will present the results of FRED commissioning and validation against the proton beam model used in CNAO (Pavia, Italy) and Krakow (Poland) facilities. We will report on the obtained accuracy and time performance of the system for quality assurance (QA) and patient treatment plans. The tracking rate of QA plans calculation in CNAO and Krakow was up to 10^7 protons/s. The γ-index pass rate for QA plans was 99.6% (analytical algorithm vs. FRED, 2mm/2%) and 97.9% (measurements vs. FRED, 2mm/3%) in CNAO and Krakow, respectively. We currently explore FRED for treatment planning studies to investigate the clinical impact of variable RBE definition and parametrization on biological dose calculation in patients.

Beside the current research results, the new developments and future application of FRED in radiation therapy will be presented. The implementation of scoring in multiple regions enables application of range shifter, multi-leaf collimator or detector development for range monitoring in PBT. The ongoing FRED kernel developments include implementation of photon interactions and nuclear models for light and heavy ions.

ABSTRACT. Boron Neutron Capture Therapy (BNCT) is a binary radiation therapy which is able to selectively destroy malignant cells while sparing the normal tissue. A 10B containing drug, studied to target neoplastic cells, is administered to the patient which is then irradiated with thermal neutrons that induce the 10B(n,α)7Li capture reaction. The particles produced by the capture reaction have high LET and range comparable to the cell diameter ensuring a cell level selectivity. Therefore BNCT effectiveness is strongly dependent on the ability to induce a high concentration of 10B inside the neoplastic cells. The correct evaluation of the dose deposited in the tumour is a key element to further the BNCT efficacy. Therefore the Pavia BNCT group has studied a BNCT-SPECT imaging system based on a CdZnTe solid state photon detector to be installed inside the treatment room of a clinical facility. SPECT imaging of the 478 keV photon emitted in the 94% of the cases due to the 10B thermal neutron capture reaction would allow for a direct and on line quantification of the dose delivered to the tumour. Since the SPECT system would be working in the BNCT treatment room it is important to study the mixed neutron and gamma background present during the patient irradiation. The Pavia BNCT group has studied a clinical facility based on an accelerator neutron source from a computational point of view. Moreover the clinical treatment room has been characterized by evaluating the activation of patient, walls and air. In the present work we studied the background inside the treatment room and simulated its interaction with the CdZnTe detector. Moreover we studied the ability of the CdZnTe detector to correctly identify the 478 keV peak due to the boron capture reaction when working in the mixed neutron and gamma background.

ABSTRACT. TBA