Monte Carlo Photon Transport Simulations Research Articles

Framework for denoising Monte Carlo photon transport simulations using deep learning.

.SignificanceThe Monte Carlo (MC) method is widely used as the gold-standard for modeling light propagation inside turbid media, such as human tissues, but combating its inherent stochastic noise requires one to simulate a large number photons, resulting in high computational burdens.AimWe aim to develop an effective image denoising technique using deep learning (DL) to dramatically improve the low-photon MC simulation result quality, equivalently bringing further acceleration to the MC method.ApproachWe developed a cascade-network combining DnCNN with UNet, while extending a range of established image denoising neural-network architectures, including DnCNN, UNet, DRUNet, and deep residual-learning for denoising MC renderings (ResMCNet), in handling three-dimensional MC data and compared their performances against model-based denoising algorithms. We also developed a simple yet effective approach to creating synthetic datasets that can be used to train DL-based MC denoisers.ResultsOverall, DL-based image denoising algorithms exhibit significantly higher image quality improvements over traditional model-based denoising algorithms. Among the tested DL denoisers, our cascade network yields a 14 to 19 dB improvement in signal-to-noise ratio, which is equivalent to simulating to more photons. Other DL-based methods yielded similar results, with our method performing noticeably better with low-photon inputs and ResMCNet along with DRUNet performing better with high-photon inputs. Our cascade network achieved the highest quality when denoising complex domains, including brain and mouse atlases.ConclusionsIncorporating state-of-the-art DL denoising techniques can equivalently reduce the computation time of MC simulations by one to two orders of magnitude. Our open-source MC denoising codes and data can be freely accessed at http://mcx.space/.

The Impulse Approximation Scattering Function and Its Use in Monte Carlo Photon Transport Simulations

The viability of using the impulse approximation scattering function in Monte Carlo photon transport simulations is explored. This scattering function can be constructed from the double differential incoherent scattering cross section developed by Ribberfors and Berggren [Phys. Rev. A., Vol 26, p. 3325 (1982)]. A commonly used method for modeling photon Doppler broadening, which is referred to as the hybrid Doppler broadening method, can also be derived from this cross section. A new photon Doppler broadening method, called the consistent Doppler broadening method, is derived and discussed. This method eliminates some of the commonly employed approximations in the hybrid Doppler broadening method, in part, by using the impulse approximation scattering function. Integrated incoherent cross sections generated using the impulse approximation scattering function and the widely used Waller-Hartree scattering function are in good agreement above 20 keV. Below 20 keV, differences as high as 70% are observed, which differs from the roughly 5% differences observed by Ribberfors [Phys. Rev. A., Vol. 27, p. 3061 (1983)] for some of the materials. Integral and spectral quantities for two problems are also generated using the Monte Carlo photon transport capabilities of the Framework for Research in Nuclear Science and Engineering. Due to the small, relative result differences observed when using the impulse approximation scattering function, it is considered a viable alternative to the Waller-Hartree scattering function. In addition, some small, but expected, differences in spectral fluxes at low energies can be avoided by adopting the consistent Doppler broadening method.

Implementation and benchmarking of the local weight window generation function for OpenMC

OpenMC is a community-driven open-source Monte Carlo neutron and photon transport simulation code. The Weight Window Mesh (WWM) function and an automatic Global Variance Reduction (GVR) method was recently developed and implemented in a developmental branch of OpenMC. This WWM function and GVR method broaden OpenMC's usage in general purposes deep penetration shielding calculations. However, the Local Variance Reduction (LVR) method, which suits the source-detector problem, is still missing in OpenMC. In this work, the Weight Window Generator (WWG) function has been developed and benchmarked for the same branch. This WWG function allows OpenMC to generate the WWM for the source-detector problem on its own. Single-material cases with varying shielding and sources were used to benchmark the WWG function and investigate how to set up the particle histories utilized in WWG-run and WWM-run. Results show that there is a maximum improvement of WWM generated by WWG. Based on the above results, instructions on determining the particle histories utilized in WWG-run and WWM-run for optimal computation efficiency are given and tested with a few multi-material cases. These benchmarks demonstrate the ability of the OpenMC WWG function and the above instructions for the source-detector problem. This developmental branch will be released and merged into the main distribution in the future.

Development and benchmarking of the Weight Window Mesh function for OpenMC

OpenMC is a community-driven open-source Monte Carlo neutron and photon transport simulation code. It supports geometry modeling using constructive solid geometry, and is capable of performing fixed source particle transport simulation based on continuous-energy and group-wise nuclear cross-section data. The lack of variance reduction technique is one of the main drawbacks of this code, which prevent its use for deep penetration shielding calculations. In this work, a commonly used variance reduction technique – using the Weight Window Mesh (WWM) function has been developed and benchmarked for OpenMC. This enables OpenMC to use any WWM file produced for the MCNP code. For benchmarking, two deep penetration shielding cases, a sphere simulation model and the thickest shielding cases in the TIARA high energy neutron shielding experiment has been employed, using the WWM produced by the ADVANTG code. In addition, the OpenMC WWM function has also been applied on the IFMIF-DONES Test Cell global neutron flux distribution calculations. The successful benchmarking and application demonstrate ability of the newly developed OpenMC WWM function for applications in general purposes deep penetration shielding calculations.

Benchmarking and verification of the OpenMC code for accelerator-based neutron source analyses

OpenMC is a community-developed Monte Carlo neutron and photon transport simulation code. It is capable of performing fixed source particle transport simulation based on continuous-energy nuclear cross-section data. In this work, OpenMC has been benchmarked and verified for the application to accelerator-based neutron sources, e.g. the International Fusion Material Irradiation Facility- DEMO Oriented NEutron Source facility (IFMIF-DONES). To this end, the McDeLicious code, which is an extension of the MCNP Monte Carlo code with the capability to simulate the deuterium-lithium neutron source on the basis of evaluated d + 6,7Li cross section data, has been migrated to the OpenMC code. The OpenMC-McDeLicious has been tested for the calculation of uncollided neutron and photon flux and spectra by comparing to results obtained with MCNP6-McDeLicious. SINBAD experimental benchmarks and d-Li experiments have been used for validating OpenMC. These tests and verifications show good agreement between OpenMC and MCNP code, as well as the experimental data. In addition, OpenMC-McDeLicious is used for obtaining the nuclear responses, e.g. Displacement Per Atom (DPA), gas production, nuclear heating, in the irradiation test module of the IFMIF-DONES neutron source. Most OpenMC-McDeLicious results show good agreement with MCNP6-McDeLcious within 5%. This indicates the successful benchmarking and validation of OpenMC code for the applications to accelerator-based neutron source facilities.

Parallelized Monte Carlo Photon Transport Simulations for Arbitrary Multi-Angle Wide-Field Illumination in Optoacoustic Imaging

Optoacoustic imaging (OAI) or photoacoustic imaging can resolve the distribution of tissue chromophores and optical contrast agents deep inside tissue from the optoacoustic detection with multi-spectral illumination. The development of a fast and accurate modeling method for the photon transport in OAI is necessary to quantitatively evaluate the tissue optical parameters. This paper presents a parallelized Monte Carlo modeling method especially for OAI (MCOAI) to simulate photon transport in bio-tissues with arbitrary multi-angle wide-field illumination. The performance of the MCOAI method is verified by comparison with the graphics processing unit (GPU)-accelerated MCX method in the typical cases with the pencil beam and ring source illumination. The simulation results demonstrate the GPU-based MCOAI method has equivalent accuracy and significantly improved computation efficiency, compared to the MCX method. The simulations with cylindrical and hemispherical source illumination further illustrate that the MCOAI method can effectively implement three-dimensional photon transport simulation for typical illumination geometries of OAI systems. A cross-section of Digimouse is selected as a realistic heterogeneous phantom illuminated with six different light sources usually employed in OAI, in order to prove the necessity of establishing photon transport modeling in OAI for the quantitative visualization in deep tissues. The MCOAI method can provide a powerful tool to efficiently establish photon transport modeling with arbitrary illumination modes for OAI applications.

Graphics processing unit-accelerated mesh-based Monte Carlo photon transport simulations.

.The mesh-based Monte Carlo (MMC) algorithm is increasingly used as the gold-standard for developing new biophotonics modeling techniques in 3-D complex tissues, including both diffusion-based and various Monte Carlo (MC)-based methods. Compared to multilayered and voxel-based MCs, MMC can utilize tetrahedral meshes to gain improved anatomical accuracy but also results in higher computational and memory demands. Previous attempts of accelerating MMC using graphics processing units (GPUs) have yielded limited performance improvement and are not publicly available. We report a highly efficient MMC—MMCL—using the OpenCL heterogeneous computing framework and demonstrate a speedup ratio up to 420× compared to state-of-the-art single-threaded CPU simulations. The MMCL simulator supports almost all advanced features found in our widely disseminated MMC software, such as support for a dozen of complex source forms, wide-field detectors, boundary reflection, photon replay, and storing a rich set of detected photon information. Furthermore, this tool supports a wide range of GPUs/CPUs across vendors and is freely available with full source codes and benchmark suites at http://mcx.space/#mmc.

Machine learning approach for rapid and accurate estimation of optical properties using spatial frequency domain imaging.

.Fast estimation of optical properties from reflectance measurements at two spatial frequencies could pave way for real-time, wide-field and quantitative mapping of vital signs of tissues. We present a machine learning-based approach for estimating optical properties in the spatial frequency domain, where a random forest regression algorithm is trained over data obtained from Monte-Carlo photon transport simulations. The algorithm learns the nonlinear mapping between diffuse reflectance at two spatial frequencies, and the absorption and reduced scattering coefficient of the tissue under consideration. Using this method, absorption and reduced scattering properties could be obtained over a 1 megapixel image in 450 ms with errors as low as 0.556% in absorption and 0.126% in reduced scattering.

Graphics processing units-accelerated adaptive nonlocal means filter for denoising three-dimensional Monte Carlo photon transport simulations.

.The Monte Carlo (MC) method is widely recognized as the gold standard for modeling light propagation inside turbid media. Due to the stochastic nature of this method, MC simulations suffer from inherent stochastic noise. Launching large numbers of photons can reduce noise but results in significantly greater computation times, even with graphics processing units (GPU)-based acceleration. We develop a GPU-accelerated adaptive nonlocal means (ANLM) filter to denoise MC simulation outputs. This filter can effectively suppress the spatially varying stochastic noise present in low-photon MC simulations and improve the image signal-to-noise ratio (SNR) by over 5 dB. This is equivalent to the SNR improvement of running nearly more photons. We validate this denoising approach using both homogeneous and heterogeneous domains at various photon counts. The ability to preserve rapid optical fluence changes is also demonstrated using domains with inclusions. We demonstrate that this GPU-ANLM filter can shorten simulation runtimes in most photon counts and domain settings even combined with our highly accelerated GPU MC simulations. We also compare this GPU-ANLM filter with the CPU version and report a threefold to fourfold speedup. The developed GPU-ANLM filter not only can enhance three-dimensional MC photon simulation results but also be a valuable tool for noise reduction in other volumetric images such as MRI and CT scans.

Scalable and massively parallel Monte Carlo photon transport simulations for heterogeneous computing platforms.

.We present a highly scalable Monte Carlo (MC) three-dimensional photon transport simulation platform designed for heterogeneous computing systems. Through the development of a massively parallel MC algorithm using the Open Computing Language framework, this research extends our existing graphics processing unit (GPU)-accelerated MC technique to a highly scalable vendor-independent heterogeneous computing environment, achieving significantly improved performance and software portability. A number of parallel computing techniques are investigated to achieve portable performance over a wide range of computing hardware. Furthermore, multiple thread-level and device-level load-balancing strategies are developed to obtain efficient simulations using multiple central processing units and GPUs.

Analyzing the relationship between decorrelation time and tissue thickness in acute rat brain slices using multispeckle diffusing wave spectroscopy.

Novel techniques in the field of wavefront shaping have enabled light to be focused deep inside or through scattering media such as biological tissue. However, most of these demonstrations have been limited to thin, static samples since these techniques are very sensitive to changes in the arrangement of the scatterers within. As the samples of interest get thicker, the influence of the dynamic nature of the sample becomes even more pronounced and the window of time in which the wavefront solutions remain valid shrinks further. In this paper, we examine the time scales upon which this decorrelation happens in acute rat brain slices via multispeckle diffusing wave spectroscopy and investigate the relationship between this decorrelation time and the thickness of the sample using diffusing wave spectroscopy theory and Monte Carlo photon transport simulation.

SU‐E‐T‐558: Monte Carlo Photon Transport Simulations On GPU with Quadric Geometry

Purpose:Monte Carlo simulation on GPU has experienced rapid advancements over the past a few years and tremendous accelerations have been achieved. Yet existing packages were developed only in voxelized geometry. In some applications, e.g. radioactive seed modeling, simulations in more complicated geometry are needed. This abstract reports our initial efforts towards developing a quadric geometry module aiming at expanding the application scope of GPU‐based MC simulations.Methods:We defined the simulation geometry consisting of a number of homogeneous bodies, each specified by its material composition and limiting surfaces characterized by quadric functions. A tree data structure was utilized to define geometric relationship between different bodies. We modified our GPU‐based photon MC transport package to incorporate this geometry. Specifically, geometry parameters were loaded into GPU's shared memory for fast access. Geometry functions were rewritten to enable the identification of the body that contains the current particle location via a fast searching algorithm based on the tree data structure.Results:We tested our package in an example problem of HDR‐brachytherapy dose calculation for shielded cylinder. The dose under the quadric geometry and that under the voxelized geometry agreed in 94.2% of total voxels within 20% isodose line based on a statistical t‐test (95% confidence level), where the reference dose was defined to be the one at 0.5cm away from the cylinder surface. It took 243sec to transport 100million source photons under this quadric geometry on an NVidia Titan GPU card. Compared with simulation time of 99.6sec in the voxelized geometry, including quadric geometry reduced efficiency due to the complicated geometry‐related computations.Conclusion:Our GPU‐based MC package has been extended to support photon transport simulation in quadric geometry. Satisfactory accuracy was observed with a reduced efficiency. Developments for charged particle transport in this geometry are currently in progress.

Neural network modelling of dose distribution and dose uniformity in the Tunisian Gamma Irradiator

In this paper an approach to model dose distributions, isodose curves and dose uniformity in the Tunisian Gamma Irradiation Facility using artificial neural networks (ANNs) are described. For this purpose, measurements were carried out at different points in the irradiation cell using polymethyl methacrylate dosemeters. The calculated and experimental results are compared and good agreement is observed showing that ANNs can be used as an efficient tool for modelling dose distribution in the gamma irradiation facility. Monte Carlo (MC) photon-transport simulation techniques have been used to evaluate the spatial dose distribution for extensive benchmarking. ANN approach appears to be a significant advance over the time-consuming MC or the less accurate regression methods for dose mapping. As a second application, a detailed dose mapping using two different product densities was carried out. The minimum and maximum dose locations and dose uniformity as a function of the irradiated volume for each product density were determined. Good agreement between ANN modelling and experimental results was achieved.

Accelerating mesh-based Monte Carlo method on modern CPU architectures

In this report, we discuss the use of contemporary ray-tracing techniques to accelerate 3D mesh-based Monte Carlo photon transport simulations. Single Instruction Multiple Data (SIMD) based computation and branch-less design are exploited to accelerate ray-tetrahedron intersection tests and yield a 2-fold speed-up for ray-tracing calculations on a multi-core CPU. As part of this work, we have also studied SIMD-accelerated random number generators and math functions. The combination of these techniques achieved an overall improvement of 22% in simulation speed as compared to using a non-SIMD implementation. We applied this new method to analyze a complex numerical phantom and both the phantom data and the improved code are available as open-source software at http://mcx.sourceforge.net/mmc/.

Neutron Activation Study of the GRR-1 Research Reactor Core Supporting Plate

A computational method for the radiological characterization of the Greek Research Reactor (GRR-1) core supporting grid plate is presented. It is based on three-dimensional Monte Carlo neutron and photon transport simulations, analytical radionuclide inventory calculations, and measured gamma dose rates. The spatial distribution of neutron fluxes and spectra were derived by an implicit MCNP reactor core model. The radionuclide inventory was estimated using the FISPACT code. The associated source term was included in an accurate MCNP model of the grid plate assembly deriving the resulting gamma dose rates. The dominant gamma dose-producing nuclide was 60Co generated by activation of cobalt impurity in the stainless steel parts. The cobalt impurity concentration in the stainless steel parts was determined on the basis of best agreement between gamma dose rate calculations and measurements. The specific activity of grid plate components was evaluated as a function of cooling time after reactor shutdown. The proposed methodology provides a useful tool for work planning, control of occupational exposure and waste management during reactor renovation, and maintenance or decommissioning activities.

Prospects for quantitative two‐dimensional radiochromic film dosimetry for low dose‐rate brachytherapy sources

Radiochromic film (RCF) has been shown to be a precise and accurate two-dimensional dosimeter for acute exposure radiation fields. However, "temporal history" mismatch between calibration and brachytherapy films due to RCF dose-rate effects could introduce potentially large uncertainties in low dose-rate (LDR) brachytherapy absolute dose measurement. This article presents a quantitative evaluation of the precision and accuracy of a laser scanner-based RCF-dosimetry system and the effect of the temporal history mismatch in LDR absolute dose measurement. MD-55-2 RCF was used to measure absolute dose for a low dose-rate 137Cs brachytherapy source using both single- and double-exposure techniques. Dose-measurement accuracy was evaluated by comparing RCF to Monte Carlo photon-transport simulation. The temporal history mismatch effect was investigated by examining dependence of RCF accuracy on irradiation-to-densitometry time interval. The predictions of the empirical cumulative dose superposition model (CDSM) were compared with measurements. For the double-exposure technique, the agreement between measurement and Monte Carlo simulation was better than 4% in the 3-60 Gy dose range with measurement precisions (coverage factor k = 1) of <2% and <6% for the doses greater or less than 3 Gy, respectively. The overall uncertainty (k = 1) of dose rate/air-kerma strength measurements achievable by this dosimetry system for a spatial resolution of 0.1 mm is less than 4% for doses greater than 5 Gy. The measured temporal history mismatch systematic error is about 1.8% for a 48 h postexposure time when using the double exposure technique and agrees with CDSM's prediction qualitatively. This work demonstrates that the model MD-55-2 RCF detector has the potential to support quantitative dose measurements about LDR brachytherapy sources with precision and accuracy better than that of previously described dosimeters. The impacts of this work on the future use of new type of RCF were also discussed.

SU-FF-T-314: Initial Monte Carlo Analysis of the Dosimetric Effects of Gold Nanoparticle Radiosensitizers

Purpose:Monte Carlo simulations have been undertaken to quantify the dosimetry of tumors that have absorbed nanoparticlegold as radiosensitizers and irradiated with various radiotherapy modalities. Method and Materials: EGSnrcMP and BEAMnrcMP, well known and documented Monte CarloPhotonTransport simulation codes are used for therapy simulations. Previous measurements of uptake ratio's provide a basis for generating new radiation interaction cross sections for use in radiation transport simulations. Cross sections generated by the simulation software are checked against cross sections generated using the NIST (National Institute of Standards and Technology) Standard Reference Database 8 (XGAM) and found to agree to within about 0.5%. 250 kVp, and 6 MV and Ir‐192 HDR spectra provided with the EGS package where used as sources. Tumors containing 0.5% to 5% Au by weight (the balance made up by water) and sized from 0.5 to 3 cm were simulated at various depths and compared to pure water. The effects of normal tissue absorption of gold in media surrounding the tumor were also investigated. Results: As expected, 250 kVp orthovoltage units showed the largest benefit from nanoparticles. A 50% absorption increase in tumors absorbing 1.5% Au by weight. Ir‐192 therapy beams available from High Dose Rate Brachytherapy units treating 1.5% and 5% nanoparticlegold by weight showed a 13% and 38% increase in dose absorption rate for tumors 1cm from the source. 6MV photon beams treating a 3cm tumor at 10cm depth showed a modest 5% improvement. Surrounding media absorption of goldnanoparticles shifts both target and surrounding tissue absorption rates higher, but does not significantly change their relative absorption rates. Conclusion: These simulations suggest goldnanoparticles in some modalities are worth investigating as radiosensitizers. Initial research should focus on HDR modalities first.

Benchmarking NaI(Tl) electron energy resolution measurements

A technique for validating electron energy resolution /spl eta//sub e/ results measured using the modified Compton coincidence technique (MCCT) has been developed. This technique relies on comparing measured gamma-ray energy resolution /spl eta//sub /spl gamma// with calculated values that were determined using the measured /spl eta//sub e/ results. These /spl eta//sub /spl gamma// calculations were based on Monte Carlo photon transport simulations, the measured NaI(Tl) electron response, a simplified cascade sequence, and the measured /spl eta//sub e/ results. To demonstrate this technique, MCCT-measured NaI(Tl) /spl eta//sub e/ results were used along with measured /spl eta//sub /spl gamma// results from the same NaI(Tl) crystal. Agreement to within 5% was observed for all energies considered between the calculated and measured /spl eta//sub /spl gamma// results for the NaI(Tl) crystal characterized. The calculated /spl eta//sub /spl gamma// results were also compared with previously published /spl eta//sub /spl gamma// measurements with good agreement (< 10%). In addition to describing the validation technique that was developed in this study and the results, a brief review of the /spl eta//sub e/ measurements made using the MCCT is provided. Based on the results of this study, it is believed that the MCCT-measured /spl eta//sub e/ results are reliable. Thus, the MCCT and this validation technique can be used in the future to characterize the /spl eta//sub e/ of other scintillators and to determine NaI(Tl) intrinsic energy resolution.

Simulating the effects of multiple scattering on images of dense sprays and particle fields.

Most optical measurements in turbid media (including sprays, fogs, particulate and colloidal suspensions) assume single scattering of the detected photons. Multiple scattering introduces error, which has been quantified in very few systems. To quantify this error, we have written a flexible Monte Carlo photon transport simulation code capable of handling any three-dimensional geometry. Simulations of planar laser spray imaging with large, nonabsorbing particles show that up to 50% of the photons reaching the camera are multiply scattered. Because forward scattering dominates, the image is affected little. For particles with more absorption or with size closer to the wavelength of the light than those we have simulated, the effects are expected to be more serious.

Monte Carlo calculations of dosimetry parameters of the Urocor Prostaseed source

This report presents the results of Monte Carlo calculations of the dosimetric parameters of the Urocor ProstaSeed 1251 seed source. This source contains five spherical silver markers, of 0.5 mm diameter, with 1251 deposited on the spheres through ion exchange. The silver spheres are then encapsulated within a 0.8 mm in diameter and 4.5 mm long cylindrical shell of titanium, as is common to this type of sources. Physical dimensions of the source are confirmed by measurement. Four (4) geometric models of the source, based on different assumptions on the locations of the silver spheres within the seed, were used in dose rate calculations. Monte Carlo photon transport simulation was used to calculate the dosimetric parameters of dose rate constant, radial dose function, and anisotropy function using these geometric models of the source. The Monte Carlo calculated dose rate constant of the ProstaSeed source was found equal to 0.925 cGy/Uh with approximate uncertainties of 5%. Radial dose function values and anisotropy function values were derived from the relative dose distribution around the source calculated by the Monte Carlo code. The calculated values of these dosimetric parameters agree with previously published thermoluminescence-dosimeter-measured values for this source after consideration of measurement and calculation uncertainties.