Search

Purpose: This paper provides a comparison between a fast, commercial, in-patient Monte Carlodose calculation algorithm (GPUMCD) and geant4. It also evaluates the dosimetric impact of the application of an external 1.5 T magnetic field. Methods: A stand-alone version of the Elekta™ GPUMCD algorithm, to be used within the Monaco treatment planning system to model dose for the Elekta™ magnetic resonance imaging(MRI)Linac, was compared against geant4 (v10.1). This was done in the presence or absence of a 1.5 T static magnetic field directed orthogonally to the radiation beam axis. Phantoms with material compositions of water, ICRU lung, ICRU compact-bone, and titanium were used for this purpose. Beams with 2 MeV monoenergetic photons as well as a 7 MV histogrammed spectrum representing the MRILinac spectrum were emitted from a point source using a nominal source-to-surface distance of 142.5 cm. Field sizes ranged from 1.5 × 1.5 to 10 × 10 cm2. Dose scoring was performed using a 3D grid comprising 1 mm3 voxels. The production thresholds were equivalent for both codes. Results were analyzed based upon a voxel by voxel dose difference between the two codes and also using a volumetric gamma analysis. Results: Comparisons were drawn from central axis depth doses, cross beam profiles, and isodose contours. Both in the presence and absence of a 1.5 T static magnetic field the relative differences in doses scored along the beam central axis were less than 1% for the homogeneous water phantom and all results matched within a maximum of ±2% for heterogeneous phantoms. Volumetric gamma analysis indicated that more than 99% of the examined volume passed gamma criteria of 2%—2 mm (dose difference and distance to agreement, respectively). These criteria were chosen because the minimum primary statistical uncertainty in dose scoring voxels was 0.5%. The presence of the magnetic field affects the dose at the interface depending upon the density of the material on either sides of the interface. This effect varies with the field size. For example, at the water-lung interface a 33.94% increase in dose was observed (relative to the Dmax), by both GPUMCD and geant4 for the field size of 2 × 2 cm2 (compared to no B-field case), which increased to 47.83% for the field size of 5 × 5 cm2 in the presence of the magnetic field. Similarly, at the lung-water interface, the dose decreased by 19.21% (relative to Dmax) for a field size of 2 × 2 cm2 and by 30.01% for 5 × 5 cm2field size. For more complex combinations of materials the dose deposition also becomes more complex. Conclusions: The GPUMCD algorithm showed good agreement against geant4 both in the presence and absence of a 1.5 T external magnetic field. The application of 1.5 T magnetic field significantly alters the dose at the interfaces by either increasing or decreasing the dose depending upon the density of the material on either side of the interfaces.

Purpose: To establish the accuracy and speed ofbGPUMCD, a GPU-oriented Monte Carlo code used for high dose rate brachytherapydose calculations. The first objective is to evaluate the time required for dose calculation when full Monte Carlo generated dose distribution kernels are used for plan optimization. The second objective is to assess the accuracy and speed when recalculating pre-optimized plans, consisting of many dwell positions. Methods: bGPUMCD is tested with three clinical treatment plans : one prostate case, one breast case, and one rectum case with a shielded applicator. Reference distributions, generated with GEANT4, are used as a basis of comparison. Calculations of full dose distributions of pre-optimized treatment plans as well as single dwell dosimetry are performed. Single source dosimetry, based on TG-43 parameters reproduction, is also presented for the microSelectron V2 (Nucletron, Veenendaal, The Netherlands). Results: In timing experiments, the computation of single dwell position dose kernels takes between 0.25 and 0.5 s.bGPUMCD can compute full dose distributions of previously optimized plans in ∼2 s. bGPUMCD is capable of computing pre-optimized brachytherapy plans within 1% for the prostate case and 2% for the breast and shielded applicator cases, when comparing the dosimetric parameters D90 and V100 of the reference (GEANT4) and bGPUMCDdistributions. For all voxels within the target, an absolute average difference of approximately 1% is found for the prostate case, less than 2% for the breast case and less than 2% for the rectum case with shielded applicator. Larger point differences (>5%) are found within bony regions in the prostate case, where bGPUMCD underdoses compared to GEANT4. Single source dosimetry results are mostly within 2% for the radial function and within 1%–4% for the anisotropic function. Conclusions: bGPUMCD has the potential to allow for fast MCdose calculation in a clinical setting for all phases of HDR treatment planning, from dose kernel calculations for plan optimization to plan recalculation.

Purpose: To validate GPUMCD, a new package for fast Monte Carlo dose calculations based on the GPU (graphics processing unit), as a tool for low-energy single seed brachytherapy dosimetry for specific seed models. As the currently accepted method of dose calculation in low-energy brachytherapy computations relies on severe approximations, a Monte Carlo based approach would result in more accurate dose calculations, taking in to consideration the patient anatomy as well as interseed attenuation. The first step is to evaluate the capability of GPUMCD to reproduce low-energy, single source, brachytherapy calculations which could ultimately result in fast and accurate, Monte Carlo based, brachytherapy dose calculations for routine planning. Methods: A mixed geometry engine was integrated to GPUMCD capable of handling parametric as well as voxelized geometries. In order to evaluate GPUMCD for brachytherapy calculations, several dosimetry parameters were computed and compared to values found in the literature. These parameters, defined by the AAPM Task-Group No. 43, are the radial dose function, the 2D anisotropy function, and the dose rate constant. These three parameters were computed for two different brachytherapy sources: the Amersham OncoSeed 6711 and the Imagyn IsoStar IS-12501. Results: GPUMCD was shown to yield dosimetric parameters similar to those foundmore » in the literature. It reproduces radial dose functions to within 1.25% for both sources in the 0.5< r

Purpose: Graphic processing units (GPUs) are increasingly used for scientific applications, where their parallel architecture and unprecedented computing power density can be exploited to accelerate calculations. In this paper, a new GPU implementation of a convolution/superposition (CS) algorithm is presented. Methods: This new GPU implementation has been designed from the ground-up to use the graphics card’s strengths and to avoid its weaknesses. The CS GPU algorithm takes into account beam hardening, off-axis softening, kernel tilting, and relies heavily on raytracing through patient imaging data. Implementation details are reported as well as a multi-GPU solution. Results: An overall single-GPU acceleration factor of 908× was achieved when compared to a nonoptimized version of the CS algorithm implemented in PlanUNC in single threaded central processing unit (CPU) mode, resulting in approximatively 2.8 s per beam for a 3D dose computation on a 0.4 cm grid. A comparison to an established commercial system leads to an acceleration factor of approximately 29× or 0.58 versus 16.6 s per beam in single threaded mode. An acceleration factor of 46× has been obtained for the total energy released per mass (TERMA) calculation and a 943× acceleration factor for the CS calculation compared to PlanUNC. Dose distributions also have been obtained for a simple water-lung phantom to verify that the implementation gives accurate results. Conclusions: These results suggest that GPUs are an attractive solution for radiation therapy applications and that careful design, taking the GPU architecture into account, is critical in obtaining significant acceleration factors. These results potentially can have a significant impact on complex dose delivery techniques requiring intensive dose calculations such as intensity-modulated radiation therapy(IMRT) and arc therapy. They also are relevant for adaptive radiation therapy where dose results must be obtained rapidly.

The numerical calculation of dose is central to treatment planning in radiation therapy and is at the core of optimization strategies for modern delivery techniques. In a clinical environment, dose calculation algorithms are required to be accurate and fast. The accuracy is typically achieved through the integration of patient-specific data and extensive beam modeling, which generally results in slower algorithms. In order to alleviate execution speed problems, the authors have implemented a modern dose calculation algorithm on a massively parallel hardware architecture. More specifically, they have implemented a convolution-superposition photon beam dose calculation algorithm on a commodity graphics processing unit (GPU). They have investigated a simple porting scenario as well as slightly more complex GPU optimization strategies. They have achieved speed improvement factors ranging from 10 to 20 times with GPU implementations compared to central processing unit (CPU) implementations, with higher values corresponding to larger kernel and calculation grid sizes. In all cases, they preserved the numerical accuracy of the GPU calculations with respect to the CPU calculations. These results show that streaming architectures such as GPUs can significantly accelerate dose calculation algorithms and let envision benefits for numerically intensive processes such as optimizing strategies, in particular, for complex delivery techniques such as IMRT and are therapy.

Purpose: An MRI-linear accelerator is currently being developed by the vendor Elekta™. The treatment planning system that will be used to model dose for this unit uses a Monte Carlo dose calculation algorithm, GPUMCD, that allows for the application of a magnetic field. We tested this radiation transport code against an independent Monte-Carlo toolkit Geant4 (v.4.10.01) both with and without the magnetic field applied. Methods: The setup comprised a 6 MeV mono-energetic photon beam emerging from a point source impinging on a homogeneous water phantom at 100 cm SSD. The comparisons were drawn from the percentage depth doses (PDD) for three different field sizes (1.5 x 1.5 cm2, 5 x 5 cm2, 10 x 10 cm2) and dose profiles at various depths. A 1.5 T magnetic field was applied perpendicular to the direction of the beam. The transport thresholds were kept the same for both codes. Results: All of the normalized PDDs and profiles agreed within ± 1 %. In the presence of the magnetic field, PDDs rise more quickly reducing the depth of maximum dose. Near the beam exit point in the phantom a hot spot is created due to the electron return effect. This effect is more pronounced for the larger field sizes. Profiles selected parallel to the external field show no effect, however, the ones selected perpendicular to the direction of the applied magnetic field are shifted towards the direction of the Lorentz force applied by the magnetic field on the secondary electrons. It is observed that these profiles are not symmetric which indicates a lateral build up of the dose. Conclusion: There is a good general agreement between the PDDs/profiles calculated by both algorithms thus far. We are proceeding towards clinically relevant comparisons in a heterogeneous phantom for polyenergetic beams. Funding for this work has been provided by Elekta.

Dose calculation is a critical component for radiation therapy. To ensure its accuracy, especially in complicated cases with large tissue heterogeneity, modern dose‐calculation algorithms contain a lot of details in their models, which demand a high computation power to achieve an acceptable level of efficiency. Especially for Monte Carlo simulation, the most accurate dose‐calculation method, its high computational burden hinders the clinical applications in many contexts. Recently, a number of research have been devoted to accelerating dose calculations for both photon and proton radiotherapy on computer graphics processing units (GPU), specialized processors with a highly parallel structure originally designed for manipulating computer graphics. The achieved speed‐up factors and accuracy, as well as the low cost in hardware setup and maintenance, indicate a great potential for the clinical applications of GPU‐based dose calculations. Yet, there are still significant challenges. For instance, there exist conflicts between the GPU's data‐parallel processing structure and the randomness of Monte Carlo simulations and hence, it is challenging to design GPU‐friendly parallelization schemes for this method. It is also desirable to develop suitable physics models that are simple for GPU implementations but attain sufficient accuracy. The proposed symposium aims to bring together experts to exchange the most recent advances in this topic, including pencil‐beam algorithms, superposition/convolution algorithms, and both photon and proton Monte Carlo simulations. Challenges in these methods and potential solutions will also be discussed. Learning Objectives: 1. Understand how GPU is used to accelerate dose calculations in radiotherapy, including pencil‐beam algorithm, superposition/ convolution, and Monte Carlo methods. 2. Understand the current status of these algorithms. 3. Understand the challenges for GPU‐based dose calculations and the potential solutions. Dr. Sami Hissoiny is employed by Elekta. The GPU code from Dr. Todd McNutt was licensed to Elekta and Gulmay.

Purpose: To compare a fast GPU‐based dose calculation algorithm to Monte Carlo(MC) simulations and TG‐43 results in low dose rate brachytherapy, in terms of accuracy and relative execution speed. Methods: Dose calculations were performed in a voxelized numerical phantom comprising bone, air and gold inserts. The source consisted in a single seed of 125I (SelectSeed, Nucletron, The Netherlands). Dose distributions were obtained from calculations based on the TG‐43 formalism, from MC simulations (GEANT4 v.9.3) and from a GPU‐based version of the TG‐43 formalism capable of handling heterogeneities. This feature was implemented in the GPU algorithm by computing the equivalent water length travelled through each voxel between the emission site and the dose calculation point. Results: Dose profiles were plotted along heterogeneities to visualize the behavior of each method, with MC simulation as the gold standard. Unsurprisingly, the TG‐43 method overestimated the dose behind high‐density/high‐Z regions with errors in excess of 100% in some cases. The modified TG‐43 algorithm implemented on the GPU algorithm was able to better reproduce MC results, with errors in the range −9.3% to 10.5% behind the same regions. Larger differences occurred for backscattering effects, which are only modeled in the MC method, but only on a limited region (1–2 mm). The TG‐43 formalism provides dose distributions almost instantaneously while MC simulations typically required up to four hours for statistically significant results. The modified TG‐43 GPU algorithm was able to reproduce MC results in approximately four seconds. Conclusions: For complex geometries, the GPU‐based TG‐43 algorithm was shown to provide dose distributions that are closer to those obtained with MC simulations while being significantly faster. This improvement can potentially improve the planning in LDR brachytherapy, where inter‐source attenuation and tissue heterogeneities were shown to influence dosimetric parameters.

Purpose: The line‐of‐response OSEM (LOR‐OSEM) algorithm allows a PETimage reconstruction from sinograms without any data compression(span=1, mashing=1). The main objective of this work is to accelerate the computation of this algorithm for modern PETscanners such as the Philips Gemini GXL by its implementation on modern GPU devices.Methods: We implemented the LOR‐OSEM algorithm on the NVIDIA Tesla 2050 GPU. The implementation incorporates the attenuation and normalization correction in the sensitivity matrix as weight factors (ANW‐LOR‐OSEM algorithm). The system matrices are built on‐the‐fly by using the multi‐ray Siddon algorithm. We used 3 rays per detector pair in the tangential direction and 2 rays in the axial direction. To reduce this computation time, the symmetries of the scanner were exploited. This implementation was validated using Monte Carlo simulated data with the GATE package.Results: The reconstruction was computed for a 188×188×57 array (FOV=376 mm, 2×2×3.15 mm3 voxel size) and for a 144×144×57 array (FOV=576 mm, 4×4×3.15 mm3 voxel size). If the sinograms are pre‐corrected for attenuation and detector efficiency, and if the projection data matrix which depends only of the scanner geometry is pre‐calculated, the time to compute the LOR‐OSEM algorithm for 10 subsets, 1 iteration and 112 million coincidences is 30.5 seconds for the 188×188×57 array and 29.4 seconds for the 144×144×57 array. This time is 73.4 seconds for the 188×188×57 array and 72.7 seconds for the 144×144×57 array for the ANW‐LOR‐OSEM algorithm Conclusions: The LOR‐OSEM algorithm was successfully implemented on a Tesla C2050 GPU, including the calculation of the sensitivity matrix, for a PET system that has 85 million LORs. The reported reconstruction times are compatible with a clinical use. The NVIDIA Tesla GPU appears to be a low‐cost, high‐ performance solution for advanced PET reconstruction such as real time 4D gated reconstruction.

Purpose: To develop a fast dose calculation algorithm for permanent implant brachytherapy in heterogeneous media based on raytracing. Method and Materials: We have implemented a modified version of the TG‐43 dosimetry protocol based on an incremental version of Siddon's raytracing algorithm that accounts for tissue composition and interseed attenuation. Raytracing along straight lines from sources to dose calculation points is used to evaluate the length traveled in each voxel, for which the density and composition are known. The water‐equivalent distance corresponding to this radiological length is then used to compute the dose with the TG‐43 formalism. The raytracing being numerically intensive, we have implemented the algorithm on a Graphics Processing Unit (GPU) in order to exploit its parallel processing features and therefore reduce the overall dose computation time. For this purpose we used a NVIDIA 8800 GT GPU and the CUDA programming language. The algorithm was tested with a simple voxelized phantom and the results were compared to regular TG‐43 dose calculations. Results: Tissue composition and interseed attenuation were shown to impact significantly the dose calculation in a simple scenario. The GPU version of the modified TG‐43 algorithm was up to 34× faster than its Central Processing Unit (CPU) version. The accuracy of the results was the same on the GPU as on the CPU. Conclusion:Brachytherapydose calculations can potentially be more accurate by accounting for heterogeneities with raytracing‐based algorithms. These algorithms are numerically intensive but can exploit the parallel architecture of stream processors such as GPUs for acceleration. The GPU implementation presented here allows for execution times that are more acceptable for a clinical use. Future work will include the evaluation of the accuracy of the algorithm with Monte Carlo simulations.Conflict of Interest: Research sponsored by Varian Medical Systems Inc.

Purpose: To study the impact in terms of execution time and accuracy of using graphics hardware for calculating the dose in a treatment planning system. The architecture of Graphics Processing Units (GPU) is well suited for numerical tasks that are intrinsically parallel, such as dose calculations. Method and Materials: This work was made within the framework of PlanUNC, or PLUNC, a treatment planning system developed and maintained by the Department of Radiation Oncology of the University of North Carolina at Chapel Hill for research and development purposes. The objective was to transparently integrate a GPU dose calculation engine to PLUNC. The CUDA platform from NVIDIA was used for the GPU implementation. A convolution/superposition (CS) dose calculation algorithm was ported by developing programs (called kernels) that are executed on the GPU. Firstly, the CS engine of PLUNC was directly ported to the GPU, with the original code preserved as much as possible. Secondly, parts of the original algorithm were redesigned to better exploit the massively parallel architecture of GPUs. The numerical experiments were conducted with a NVIDIA GeForce GTX280 and an Intel Q6600 CPU. Results: Acceleration factors of 10× to 20× were achieved with the GPU implementation relative to the CPU version with the direct port of the CS algorithm. The numerical accuracy of the results was preserved with the GPU implementation. A 40× acceleration factor was obtained for the TERMA calculation subroutine, which was rewritten with the GPU architecture in mind. These acceleration factors were sufficient to significantly improve the responsiveness of the PLUNC graphical interface. Conclusion: This work demonstrates the potential of graphics hardware for dose calculation in treatment planning systems. This could in turn have a significant impact on optimization strategies for complex delivery techniques such as IMRT. Research sponsored by Varian Medical Systems, Inc.