Latent uncertainties of the precalculated track Monte Carlo method

Robust mixed electron-photon radiation therapy planning for soft tissue sarcoma

BACKGROUND

Mixed electron-photon beam radiation therapy (MBRT) is an emerging technique in which external electron and photon beams are simultaneously optimized into a single treatment plan. MBRT exploits the steep dose falloff and high surface dose of electrons while maintaining target conformity by leveraging the sharp penumbra of photons.

PURPOSE

This study investigates the dosimetric benefits of MBRT for soft tissue sarcoma (STS) patients.

MATERIAL AND METHODS

A retrospective cohort of 22 STS of the lower extremity treated with conventional photon-based Volumetric Modulated Arc Therapy (VMAT) were replanned with MBRT. Both VMAT and MBRT treatments were planned on the Varian TrueBeam linac using the Millenium multi-leaf collimator. No electron applicator, cutout or additional collimating devices were used for electron beams of MBRT plans. MBRT plans were optimized to use a combination of 6 MV photons and five electron energies (6, 9, 12, 16, 20 MeV) by a robust column generation algorithm. Electron beams in this study were planned at standard 100 cm source-axis distance (SAD). The dose to the clinical target volume (CTV), bone, normal tissue strip and other organs-at-risk (OARs) were compared using a Wilcoxon signed-rank test.

RESULTS

As part of the original VMAT treatment, tissue-equivalent bolus was required in 10 of the 22 patients. MBRT plans did not require bolus by virtue of the higher electron entrance dose. CTV coverage by the prescription dose was found to be clinically equivalent between plans of either modality:V 50Gy $V_{\text{50Gy}}$ (MBRT) = 97.9 ± 0.2% versusV 50Gy $V_{\text{50Gy}}$ (VMAT) = 98.1 ± 0.6% (p=0.34). Evaluating the absolute paired difference between doses to OARs in MBRT and VMAT plans, we observed lowerV 20Gy $V_{\text{20Gy}}$ to normal tissue in MBRT plans by 14.9 ± 3.2% (p < 10 - 6 $p<10^{-6}$ ). Similarly,V 50Gy $V_{\text{50Gy}}$ to bone was found to be decreased by 8.2 ± 4.0% (p < 10 - 3 $p<10^{-3}$ ) of the bone volume.

CONCLUSION

For STS with subcutaneous involvement, MBRT offers statistically significant sparing of OARs without sacrificing target coverage when compared to VMAT. MBRT plans are deliverable on conventional linacs without the use of electron applicators, shortened source-to-surface distance (SSD) or bolus. This study shows that MBRT is a logistically feasible technique with clear dosimetric benefits.

Development of a stand-alone precalculated Monte Carlo code to calculate the dose by alpha and beta emitters from the Ra-224 decay chain

BACKGROUND

Recent developments in alpha and beta emitting radionuclide therapy highlight the importance of developing efficient methods for patient-specific dosimetry. Traditional tabulated methods such as Medical Internal Radiation Dose (MIRD) estimate the dose at the organ level while more recent numerical methods based on Monte Carlo (MC) simulations are able to calculate dose at the voxel level. A precalculated MC (PMC) approach was developed in this work as an alternative to time-consuming fully simulated MC. Once the spatial distribution of alpha and beta emitters is determined using imaging and/or numerical methods, the PMC code can be used to achieve an accurate voxelized 3D distribution of the deposited energy without relying on full MC calculations.

PURPOSE

To implement the PMC method to calculate energy deposited by alpha and beta particles emitted from the Ra-224 decay chain.

METHODS

The GEANT4 (version 10.7) MC toolkit was used to generate databases of precalculated tracks to be integrated in the PMC code as well as to benchmark its output. In this regard, energy spectra of alpha and beta particles emitted by the Ra-224 decay chain were generated using GAMOS (version 6.2.0) and imported into GEANT4 macro files. Either alpha or beta emitting sources were defined at the center of a homogeneous phantom filled with various materials such as soft tissue, bone, and lung where particles were emitted either mono-directionally (for database generation) or isotropically (for benchmarking). Two heterogeneous phantoms were used to demonstrate PMC code compatibility with boundary crossing events. Each precalculated database was generated step-by-step by storing particle track information from GEANT4 simulations followed by its integration in a PMC code developed in MATLAB. For a user-defined number of histories, one of the tracks in a given database was selected randomly and rotated randomly to reflect an isotropic emission. Afterward, deposited energy was divided between voxels based on step length in each voxel using a ray-tracing approach. The radial distribution of deposited energy was benchmarked against fully simulated MC calculations using GEANT4. The effect of the GEANT4 parameter StepMax on the accuracy and speed of the code was also investigated.

RESULTS

In the case of alpha decay, primary alpha particles show the highest contribution (>99%) in deposited energy compared to their secondary particles. In most cases, protons act as the main secondary particles in the deposition of energy. However, for a lung phantom, using a range cutoff parameter of 10 µm on primary alpha particles yields a higher contribution of secondary electrons than protons. Differences between deposited energy calculated by PMC and fully simulated MC are within 2% for all alpha and beta emitters in homogeneous and heterogeneous phantoms. Additionally, statistical uncertainties are less than 1% for voxels with doses higher than 5% of the maximum dose. Moreover, optimization of the parameter StepMax is necessary to achieve the best tradeoff between code accuracy and speed.

CONCLUSIONS

The PMC code shows good performance for dose calculations deposited by alpha and beta emitters. As a stand-alone algorithm, it is suitable to be integrated into clinical treatment planning systems.

Development of a GPU-superposition Monte Carlo code for fast dose calculation in magnetic fields

Objective. To develop and validate a graphics processing unit (GPU) based superposition Monte Carlo (SMC) code for efficient and accurate dose calculation in magnetic fields. Approach. A series of mono-energy photons ranging from 25 keV to 7.7 MeV were simulated with EGSnrc in a water phantom to generate particle tracks database. SMC physics was extended with charged particle transport in magnetic fields and subsequently programmed on GPU as gSMC. Optimized simulation scheme was designed by combining variance reduction techniques to relieve the thread divergence issue in general GPU-MC codes and improve the calculation efficiency. The gSMC code’s dose calculation accuracy and efficiency were assessed through both phantoms and patient cases. Main results. gSMC accurately calculated the dose in various phantoms for both B = 0 T and B = 1.5 T, and it matched EGSnrc well with a root mean square error of less than 1.0% for the entire depth dose region. Patient cases validation also showed a high dose agreement with EGSnrc with 3D gamma passing rate (2%/2 mm) large than 97% for all tested tumor sites. Combined with photon splitting and particle track repeating techniques, gSMC resolved the thread divergence issue and showed an efficiency gain of 186–304 relative to EGSnrc with 10 CPU threads. Significance. A GPU-superposition Monte Carlo code called gSMC was developed and validated for dose calculation in magnetic fields. The developed code’s high calculation accuracy and efficiency make it suitable for dose calculation tasks in online adaptive radiotherapy with MR-LINAC.

XIORT-MC: A real-time MC-based dose computation tool for low- energy X-rays intraoperative radiation therapy

PURPOSE

The INTRABEAM system is a miniature accelerator for low-energy X-ray Intra-Operative Radiation Therapy (IORT), and it could benefit from a fast and accurate dose computation tool. With regards to accuracy, dose computed with Monte Carlo (MC) simulations are the gold standard, however, they require a large computational effort and consequently they are not suitable for real-time dose planning. This work presents a comparison of the implementation on Graphics Processing Unit (GPU) of two different dose calculation algorithms based on MC phase-space (PHSP) information to compute dose distributions for the INTRABEAM device within seconds and with the accuracy of realistic MC simulations.

METHODS

The MC-based algorithms we present incorporate photoelectric, Compton and Rayleigh effects for the interaction of low-energy X-rays. XIORT-MC (X-ray Intra-Operative Radiation Therapy Monte Carlo) includes two dose calculation algorithms; a Woodcock-based MC algorithm (WC-MC) and a Hybrid MC algorithm (HMC), and it is implemented in CPU and in GPU. Detailed MC simulations have been generated to validate our tool in homogeneous and heterogeneous conditions with all INTRABEAM applicators, including three clinically realistic CT-based simulations. A performance study has been done to determine the acceleration reached with the code, in both CPU and GPU implementations.

RESULTS

Dose distributions were obtained with the HMC and the WC-MC and compared to standard reference MC simulations with more than 95% voxels fulfilling a 7%-0.5 mm gamma evaluation in all the cases considered. The CPU-HMC is 100 times more efficient than the reference MC, and the CPU-WC-MC is about 50 times more efficient. With the GPU implementation, the particle tracking of the WC-MC is faster than the HMC, with the extraction of the particle's information from the PHSP file taking a major part of the time. However, thanks to the variance reduction techniques implemented in the HMC, up to 400 times less particles are needed in the HMC to reach the same level of noise than the WC-MC. Therefore, in our implementation for INTRABEAM energies, the HMC is about 1.3 times more efficient than the WC-MC in an NVIDIA GeForce GTX 1080 Ti card and about 5.5 times more efficient in an NVIDIA GeForce RTX 3090. Dose with noise below 5% has been obtained in realistic situations in less than 5 s with the WC-MC and in less than 0.5 s with the HMC.

CONCLUSIONS

The XIORT-MC is a dose computation tool designed to take full advantage of modern GPUs, making possible to obtain MC-grade accurate dose distributions within seconds. Its high speed allows a real-time dose calculation that includes the realistic effects of the beam in voxelized geometries of patients. It can be used as a dose-planning tool in the operating room during a XIORT treatment with any INTRABEAM device.

Adaptive step size algorithm to increase efficiency of proton macro Monte Carlo dose calculation

Purpose

To provide fast and accurate dose calculation in voxelized geometries for proton radiation therapy by implementing an adaptive step size algorithm in the proton macro Monte Carlo (pMMC) method.

Methods

The in-house developed local-to-global MMC method for proton dose calculation is extended with an adaptive step size algorithm for efficient proton transport through a voxelized geometry by sampling transport parameters from a pre-simulated database. Adaptive choice of an adequate slab size in dependence of material interfaces in the proton’s longitudinal and lateral vicinity is investigated. The dose calculation algorithm is validated against the non-adaptive pMMC and full MC simulation for pencil and broad beams with various energies impinging on academic phantoms as well as a head and neck patient CT.

Results

For material interfaces perpendicular to a proton’s direction, choice of nearest neighbor slab thickness shows best trade-off between dosimetric accuracy and calculation efficiency. Adaptive reduction of chosen slab size is shown to be required for material interfaces closer than 0.5 mm in lateral direction. For the academic phantoms, dose differences of within 1% or 1 mm compared to full Geant4 MC simulation are found, while achieving an efficiency gain of up to a factor of 5.6 compared to the non-adaptive algorithm and 284 compared to Geant4. For the head and neck patient CT, dose differences are within 1% or 1 mm with an efficiency gain factor of up to 3.4 compared to the non-adaptive algorithm and 145 compared to Geant4.

Conclusion

An adaptive step size algorithm for proton macro Monte Carlo was implemented and evaluated. The dose calculation provides the accuracy of full MC simulations, while achieving an efficiency gain factor of three compared to the non-adaptive algorithm and two orders of magnitude compared to full MC for a complex patient CT.

Monte Carlo dose calculations for high-dose-rate brachytherapy using GPU-accelerated processing

PURPOSE

Current clinical brachytherapy dose calculations are typically based on the Association of American Physicists in Medicine Task Group report 43 (TG-43) guidelines, which approximate patient geometry as an infinitely large water phantom. This ignores patient and applicator geometries and heterogeneities, causing dosimetric errors. Although Monte Carlo (MC) dose calculation is commonly recognized as the most accurate method, its associated long computational time is a major bottleneck for routine clinical applications. This article presents our recent developments of a fast MC dose calculation package for high-dose-rate (HDR) brachytherapy, gBMC, built on a graphics processing unit (GPU) platform.

METHODS AND MATERIALS

gBMC-simulated photon transport in voxelized geometry with physics in (192)Ir HDR brachytherapy energy range considered. A phase-space file was used as a source model. GPU-based parallel computation was used to simultaneously transport multiple photons, one on a GPU thread. We validated gBMC by comparing the dose calculation results in water with that computed TG-43. We also studied heterogeneous phantom cases and a patient case and compared gBMC results with Acuros BV results.

RESULTS

Radial dose function in water calculated by gBMC showed <0.6% relative difference from that of the TG-43 data. Difference in anisotropy function was <1%. In two heterogeneous slab phantoms and one shielded cylinder applicator case, average dose discrepancy between gBMC and Acuros BV was <0.87%. For a tandem and ovoid patient case, good agreement between gBMC and Acruos BV results was observed in both isodose lines and dose-volume histograms. In terms of the efficiency, it took ∼47.5 seconds for gBMC to reach 0.15% statistical uncertainty within the 5% isodose line for the patient case.

CONCLUSIONS

The accuracy and efficiency of a new GPU-based MC dose calculation package, gBMC, for HDR brachytherapy make it attractive for clinical applications.

A new approach to integrate GPU-based Monte Carlo simulation into inverse treatment plan optimization for proton therapy

Monte Carlo (MC)-based spot dose calculation is highly desired for inverse treatment planning in proton therapy because of its accuracy. Recent studies on biological optimization have also indicated the use of MC methods to compute relevant quantities of interest, e.g. linear energy transfer. Although GPU-based MC engines have been developed to address inverse optimization problems, their efficiency still needs to be improved. Also, the use of a large number of GPUs in MC calculation is not favorable for clinical applications. The previously proposed adaptive particle sampling (APS) method can improve the efficiency of MC-based inverse optimization by using the computationally expensive MC simulation more effectively. This method is more efficient than the conventional approach that performs spot dose calculation and optimization in two sequential steps. In this paper, we propose a computational library to perform MC-based spot dose calculation on GPU with the APS scheme. The implemented APS method performs a non-uniform sampling of the particles from pencil beam spots during the optimization process, favoring those from the high intensity spots. The library also conducts two computationally intensive matrix-vector operations frequently used when solving an optimization problem. This library design allows a streamlined integration of the MC-based spot dose calculation into an existing proton therapy inverse planning process. We tested the developed library in a typical inverse optimization system with four patient cases. The library achieved the targeted functions by supporting inverse planning in various proton therapy schemes, e.g. single field uniform dose, 3D intensity modulated proton therapy, and distal edge tracking. The efficiency was 41.6  ±  15.3% higher than the use of a GPU-based MC package in a conventional calculation scheme. The total computation time ranged between 2 and 50 min on a single GPU card depending on the problem size.