Ryohei Tansho

Clinical implementation of a GPU-based simplified Monte Carlo method for a treatment planning system of proton beam therapy

We implemented the simplified Monte Carlo (SMC) method on graphics processing unit (GPU) architecture under the computer-unified device architecture platform developed by NVIDIA. The GPU-based SMC was clinically applied for four patients with head and neck, lung, or prostate cancer. The results were compared to those obtained by a traditional CPU-based SMC with respect to the computation time and discrepancy. In the CPU- and GPU-based SMC calculations, the estimated mean statistical errors of the calculated doses in the planning target volume region were within 0.5% rms. The dose distributions calculated by the GPU- and CPU-based SMCs were similar, within statistical errors. The GPU-based SMC showed 12.30-16.00 times faster performance than the CPU-based SMC. The computation time per beam arrangement using the GPU-based SMC for the clinical cases ranged 9-67 s. The results demonstrate the successful application of the GPU-based SMC to a clinical proton treatment planning.

Improved dose-calculation accuracy in proton treatment planning using a simplified Monte Carlo method verified with three-dimensional measurements in an anthropomorphic phantom

Treatment planning for proton tumor therapy requires a fast and accurate dose-calculation method. We have implemented a simplified Monte Carlo (SMC) method in the treatment planning system of the National Cancer Center Hospital East for the double-scattering beam delivery scheme. The SMC method takes into account the scattering effect in materials more accurately than the pencil beam algorithm by tracking individual proton paths. We confirmed that the SMC method reproduced measured dose distributions in a heterogeneous slab phantom better than the pencil beam method. When applied to a complex anthropomorphic phantom, the SMC method reproduced the measured dose distribution well, satisfying an accuracy tolerance of 3 mm and 3% in the gamma index analysis. The SMC method required approximately 30 min to complete the calculation over a target volume of 500 cc, much less than the time required for the full Monte Carlo calculation. The SMC method is a candidate for a practical calculation technique with sufficient accuracy for clinical application.

Improvement of spread-out Bragg peak flatness for a carbon-ion beam by the use of a ridge filter with a ripple filter

We have developed a novel design method of ridge filters for carbon-ion therapy using a broad-beam delivery system to improve the flatness of a biologically effective dose in the spread-out Bragg peak (SOBP). So far, the flatness of the SOBP is limited to about ±5% for carbon beams since the weight control of component Bragg curves composing the SOBP is difficult. This difficulty arises from using a large number of ridge-bar steps (e.g. about 100 for a SOBP width of 60 mm) required to form the SOBP for the pristine Bragg curve with an extremely sharp distal falloff. Instead of using a single ridge filter, we introduce a ripple filter to broaden the Bragg peak so that the number of ridge-bar steps can be reduced to about 30 for SOBP with of 60 mm for the ridge filter designed for the broadened Bragg peak. Thus we can manufacture the ridge filter more accurately and then attain a better flatness of the SOBP due to well-controlled weights of the component Bragg curves. We placed the ripple filter on the same frame of the ridge filter and arranged the direction of the ripple-filter-bar array perpendicular to that of the ridge-filter-bar array. We applied this method to a 290 MeV u(-1) carbon-ion beam in Heavy Ion Medical Accelerator in Chiba and verified the effectiveness by measurements.

Evaluation of monitor unit calculation based on measurement and calculation with a simplified Monte Carlo method for passive beam delivery system in proton beam therapy

Calibrating the dose per monitor unit (DMU) for individual patients is important to deliver the prescribed dose in radiation therapy. We have developed a DMU calculation method combining measurement data and calculation with a simplified Monte Carlo method for the double scattering system in proton beam therapy at the National Cancer Center Hospital East in Japan. The DMU calculation method determines the clinical DMU by the multiplication of three factors: a beam spreading device factor FBSD, a patient‐specific device factor FPSD, and a field‐size correction factor FFS(A). We compared the calculated and the measured DMU for 75 dose fields in clinical cases. The calculated DMUs were in agreement with measurements in ±1.1% for all of 25 fields in prostate cancer cases, and in ±3% for 94% of 50 fields in head and neck (H&N) and lung cancer cases, including irregular shape fields and small fields. Although the FBSD in the DMU calculations is dominant as expected, we found that the patient‐specific device factor and field‐size correction also contribute significantly to the calculated DMU. This DMU calculation method will be able to substitute the conventional DMU measurement for the majority of clinical cases with a reasonable calculation time required for clinical use. PACS number: 87.55.khCalibrating the dose per monitor unit (DMU) for individual patients is important to deliver the prescribed dose in radiation therapy. We have developed a DMU calculation method combining measurement data and calculation with a simplified Monte Carlo method for the double scattering system in proton beam therapy at the National Cancer Center Hospital East in Japan. The DMU calculation method determines the clinical DMU by the multiplication of three factors: a beam spreading device factor FBSD, a patient-specific device factor FPSD, and a field-size correction factor FFS(A). We compared the calculated and the measured DMU for 75 dose fields in clinical cases. The calculated DMUs were in agreement with measurements in ±1.1% for all of 25 fields in prostate cancer cases, and in ±3% for 94% of 50 fields in head and neck (H&N) and lung cancer cases, including irregular shape fields and small fields. Although the FBSD in the DMU calculations is dominant as expected, we found that the patient-specific device factor and field-size correction also contribute significantly to the calculated DMU. This DMU calculation method will be able to substitute the conventional DMU measurement for the majority of clinical cases with a reasonable calculation time required for clinical use. PACS number: 87.55.kh.

Application of dose kernel calculation using a simplified Monte Carlo method to treatment plan for scanned proton beams

Full Monte Carlo (FMC) calculation of dose distribution has been recognized to have superior accuracy, compared with the pencil beam algorithm (PBA). However, since the FMC methods require long calculation time, it is difficult to apply them to routine treatment planning at present. In order to improve the situation, a simplified Monte Carlo (SMC) method has been introduced to the dose kernel calculation applicable to dose optimization procedure for the proton pencil beam scanning. We have evaluated accuracy of the SMC calculation by comparing a result of the dose kernel calculation using the SMC method with that using the FMC method in an inhomogeneous phantom. The dose distribution obtained by the SMC method was in good agreement with that obtained by the FMC method. To assess the usefulness of SMC calculation in clinical situations, we have compared results of the dose calculation using the SMC with those using the PBA method for three clinical cases of tumor treatment. The dose distributions calculated with the PBA dose kernels appear to be homogeneous in the planning target volumes (PTVs). In practice, the dose distributions calculated with the SMC dose kernels with the spot weights optimized with the PBA method show largely inhomogeneous dose distributions in the PTVs, while those with the spot weights optimized with the SMC method have moderately homogeneous distributions in the PTVs. Calculation using the SMC method is faster than that using the GEANT4 by three orders of magnitude. In addition, the graphic processing unit (GPU) boosts the calculation speed by 13 times for the treatment planning using the SMC method. Thence, the SMC method will be applicable to routine clinical treatment planning for reproduction of the complex dose distribution more accurately than the PBA method in a reasonably short time by use of the GPU-based calculation engine. PACS number(s): 87.55.Gh.Full Monte Carlo (FMC) calculation of dose distribution has been recognized to have superior accuracy, compared with the pencil beam algorithm (PBA). However, since the FMC methods require long calculation time, it is difficult to apply them to routine treatment planning at present. In order to improve the situation, a simplified Monte Carlo (SMC) method has been introduced to the dose kernel calculation applicable to dose optimization procedure for the proton pencil beam scanning. We have evaluated accuracy of the SMC calculation by comparing a result of the dose kernel calculation using the SMC method with that using the FMC method in an inhomogeneous phantom. The dose distribution obtained by the SMC method was in good agreement with that obtained by the FMC method. To assess the usefulness of SMC calculation in clinical situations, we have compared results of the dose calculation using the SMC with those using the PBA method for three clinical cases of tumor treatment. The dose distributions calculated with the PBA dose kernels appear to be homogeneous in the planning target volumes (PTVs). In practice, the dose distributions calculated with the SMC dose kernels with the spot weights optimized with the PBA method show largely inhomogeneous dose distributions in the PTVs, while those with the spot weights optimized with the SMC method have moderately homogeneous distributions in the PTVs. Calculation using the SMC method is faster than that using the GEANT4 by three orders of magnitude. In addition, the graphic processing unit (GPU) boosts the calculation speed by 13 times for the treatment planning using the SMC method. Thence, the SMC method will be applicable to routine clinical treatment planning for reproduction of the complex dose distribution more accurately than the PBA method in a reasonably short time by use of the GPU‐based calculation engine. PACS number(s): 87.55.Gh

A Patient-Specific QA Procedure for Moving Target Irradiation in Scanned Ion Therapy

Three-dimensional (3D) pencil-beam scanning technique has been utilized since 2011 in NIRS-HIMAC. Beam delivery system and treatment planning software (TPS) require dosimetric patient-specific QA to check each individual plan. Any change in the scanned beams will result in a significant impact on the irradiation dose. Therefore, patient-specific QA for moving target irradiation requires additional procedure. In an additional QA for moving target irradiation, we placed 2D ionization chamber on the PMMA plate tilted with respect to the beam axis. The PMMA plate was set on the stage of the moving phantom. The moving phantom was moved according to patient data. We measured the dose distribution for both the static target and the moving target. We compared the results for the moving target with those for the static targets by means of a gamma index analysis. In the additional patient-specific QA, the gamma analysis between the moving and static targets showed the good agreement. We confirmed that this new technique was a beneficial QA procedure for moving target irradiation.

Development of QA System for the Rotating Gantry for Carbon Ion Therapy at NIRS

At the National Institute of Radiological Sciences (NIRS), we have been developing the rotating-gantry system for the carbon-ion radiotherapy. This system is equipped with a three-dimensional pencil beam scanning irradiation system. To ensure the treatment quality, calibration of the primary dose monitor, range check, dose rate check, machine safety check, and some mechanical tests should be performed efficiently. For this purpose, we have developed a measurement system dedicated for quality assurance (QA) of this gantry system. The ion beam’s dose output are calibrated by measurement using an ionization chamber. A Farmer type ionization chamber is inserted into the center of a water equivalent phantom. The thickness of the phantom could be changed so that employ both calibration of the output at entrance and output checking at center of the irradiation field. The ranges of beams are verified using a scintillator and a CCD camera system. From the taken images, maximum gradient points are determined by some image processing and compared with reference data. In this paper, we describe consideration of the daily QA for the rotating-gantry.

SU-E-T-488: Dose Calculation Model Using the Simplified Monte Carlo Method with an Initial Beam Model Adapted to a Beam-Wobbling System

PURPOSE We have developed an accurate dose calculation model based on a simplified Monte Carlo (SMC) method adapted to a beam-wobbling delivery system at National Cancer Center Hospital East (NCCHE). We used an initial beam model specific to the beam-wobbling system to reproduce accurately different dose distributions in two lateral directions (x- and y-directions) perpendicular to each other. METHODS The SMC calculates a dose distribution by tracking individual protons. The SMC starts tracking protons at an entrance of a range compensator. Protons are generated in an initial phase space adapted to the wobbler system. Since two wobbling-magnets are located at separate places with different distances from the iso-center, different dose distributions are formed in x- and y-directions. We derived an initial phase space distribution for the beam-wobbling system using an analytical method. We used the SMC method with the initial beam model to calculate dose distributions accurately. To verify accuracy of the calculation method, we measured the dose distribution in a homogeneous phantom formed by 235 MeV protons passing through a L-shaped range compensator. We used a 2D-array of parallel-plate ionization chambers (2D Array seven29®) to measure dose distributions with a sampling period of 5 mm. RESULTS The measured dose distribution in the x-direction was different from that in the y-direction. Our calculation model reproduces the measurement results well in both lateral directions. In addition, the calculation reproduced the dose increments in edge regions contributed by edge-scattered protons in collimator. It indicates the advantage of the SMC. CONCLUSIONS A dose calculation model has been developed based on the simplified Monte Carlo method applied to a beam-wobbling system. By adapting the initial beam model to the wobbling system, the SMC method is found to reproduce observed different dose distributions in x- and y-directions well.

SU‐E‐T‐352: Comparison of Dose Distributions Between Two Arrangements of a Range Compensator and of An Aperture Collimator in a Passive Scattering Method for Proton Therapy

Purpose: Two arrangements of a range compensator (RC) and an aperture collimator (AC), RC placed in front of AC (ARGT‐1) or AC placed in front of RC (ARGT‐2), are currently used in proton therapy. We studied difference of the dose distributions for two arrangements.Methods: We calculated dose distributions in water formed by a 190‐MeV modulated proton beam (SOBP width 80 mm) passing through two arrangements of a simple geometrical RC and an AC using a Simplified Monte Carlo method. The air gap between the downstream device and water surface was set at 195 mm. Results: Calculated dose distributions were normalized in the same uniform‐dose region. We compared sharpness of the 80%–20% dose falloffs at a depth of 150 mm where protons passing through the thinnest part of RC mainly contribute the dose. The value for the ARGT‐1 is written first and that for the ARGT‐2 next. Dose falloffs were 8.5±0.2 mm / 7.9±0.2 mm near an inner RC edge. Dose falloffs were 7. 1±0.2 mm / 8.0±0.2 near the AC edge region. In addition, dose contributions of edge‐scattered protonsscattered in the AC at water surface were 19% / 10% Conclusions: We verified that ARGT‐1 is superior to the other in dose conformation near the lateral target boundary and that ARGT‐2 is superior to the other in that near the distal target boundary. We also noticed that the RC serves to absorb a part of scatteredprotons in the AC for ARGT‐2.

SU‐GG‐T‐466: A Retrospective Analysis in Head and Neck Cancer by Using the Simplified Monte Carlo Algorithm

Purpose: We have developed a simplified Monte Carlo calculation (SMC) algorithm of dose distribution in proton therapy to improve the accuracy of dose calculations yet with a reasonable calculation time. We had verified SMC by comparing with measurements. We retrospectively analyzed plans of head and neck cancer patients that occurred side‐effect radiation damage using SMC. Method and Materials: The SMC uses measured effective source parameters and a measured mono‐energetic proton depth dose curves, it is easy to adapt to facilities. Protons are traced their trajectories taking into account the multiple Coulomb scattering using the Highlands formula and range losses using the water‐equivalent model in materials. The relative dose deposit in a water voxel or in a patient voxel is obtained from the residual range of an incident proton in water and the water‐equivalent thickness of the voxel using a shifted measured Bragg curve in water. We retrospectively analyzed a plan of a treated patient with T4N0M0 adenoid cystic carcinoma.Results: SMC estimated higher dose delivery to right optic nerves and optic chiasma than that of the PBA. This means that there is greater risk of vision loss. The PBA underestimated deep‐region dose as compared with the SMC, and resulted in the difference of DVHs. Discussions and Conclusions: We retrospectively analyzed a plan of a treated head and neck cancer patient that occurred side‐effect radiation damage. The PBA underestimated the risk of untoward radiation damage by dose underestimation in deep region supported by phantom studies. The results of the SMC expected a high risk of the side‐effect radiation damage. Thus, for clinical analysis, the difference of DVHs between the SMC and the PBA suggests the effectiveness of Monte Carlodose calculation in treatment planning. We continue widely many retrospective analyses using the SMC.