Dose Computation Model Research Articles

Commissioning dose computation model for proton source in pencil beam scanning therapy by convolution neural networks

Objective. Proton source model commissioning (PSMC) is critical for ensuring accurate dose calculation in pencil beam scanning (PBS) proton therapy using Monte Carlo (MC) simulations. PSMC aims to match the calculated dose to the delivered dose. However, commissioning the ‘nominal energy’ and ‘energy spread’ parameters in PSMC can be challenging, as these parameters cannot be directly obtained from solving equations. To efficiently and accurately commission the nominal energy and energy spread in a proton source model, we developed a convolution neural network (CNN) named ‘PSMC-Net.’ Methods. The PSMC-Net was trained separately for 33 energies (E, 70–225 MeV with a step of 5 MeV plus 226.09 MeV). For each E, a dataset was generated consisting of 150 source model parameters (15 nominal energies ∈ [E, E + 1.5 MeV], ten spreads ∈ [0, 1]) and the corresponding 150 MC integrated depth doses (IDDs). Of these 150 data pairs, 130 were used for training the network, 10 for validation, and 10 for testing. Results. The source model, built by 33 measured IDDs and 33 PSMC-Nets (cost 0.01 s), was used to compute the MC IDDs. The gamma passing rate (GPRs, 1 mm/1%) between MC and measured IDDs was 99.91 ± 0.12%. However, when no commissioning was made, the corresponding GPR was reduced to 54.11 ± 22.36%, highlighting the tremendous significance of our CNN commissioning method. Furthermore, the MC doses of a spread-out Bragg peak and 20 patient PBS plans were also calculated, and average 3D GPRs (2 mm/2% with a 10% threshold) were 99.89% and 99.96 ± 0.06%, respectively. Significance. We proposed a nova commissioning method of the proton source model using CNNs, which made the PSMC process easy, efficient, and accurate.

SU‐F‐T‐138: Commissioning and Evaluating Dose Computation Models for a Dedicated Proton Line Scanning Beam Nozzle in Eclipse Treatment Planning System

Purpose:In this study, we present an effective method to derive low dose envelope of the proton in‐air spot fluence at beam positions other than the isocenter to reduce amount of measurements required for planning commission. Also, we demonstrate commissioning and validation results of this method to the Eclipse treatment planning system (version 13.0.29) for a Sumitomo dedicated proton line scanning beam nozzle.Methods:The in‐air spot profiles at five beam‐axis positions (±200, ±100 and 0 mm) were obtained in trigger mode using a MP3 Water tank (PTW‐Freiburg) and a pinpoint ionization chamber (model 31014, PTW‐Freiburg). Low dose envelope (below 1% of the center dose) of the spot profile at isocenter was obtained by repeated point measurements to minimize dosimetry uncertainty. The double Gaussian (DG) model was used to fit and obtain optimal σ1, σ2 and their corresponding weightings through our in‐house MATLAB (Mathworks) program. σ1, σ2 were assumed to expand linearly along the beam axis from a virtual source position calculated by back projecting fitted sigmas from the single Gaussian (SG) model. Absolute doses in water were validated using an Advanced Markus chamber at the depth of 2cm with Pristine Peak (BP) R90d ranging from 5–32 cm for 10×10 cm2 scanned fields. The field size factors were verified with square fields from 2 to 20 cm at 2cm and before BP depth.Results:The absolute dose outputs were found to be within ±3%. For field size factor, the agreement between calculated and measurement were within ±2% at 2cm and ±3% before BP, except for the field size below 2×2 cm2.Conclusion:The double Gaussian model was found to be sufficient for characterizing the Sumitomo dedicated proton line scanning nozzle. With our effective double Gaussian fitting method, we are able to save significant proton beam time with acceptable output accuracy.

SU‐D‐BRC‐05: Effects of Motion and Variable RBE in a Lung Patient Treated with Passively Scattered Protons

Purpose:To evaluate effects of motion and variable relative biological effectiveness (RBE) in a lung cancer patient treated with passively scattered proton therapy using dose volume histograms associated with patient dose computed using three different methods.Methods:A proton treatment plan of a lung cancer patient optimized using clinical treatment planning system (TPS) was used to construct a detailed Monte Carlo (MC) model of the beam delivery system and the patient specific aperture and compensator. A phase space file containing all particles transported through the beam line was collected at the distal surface of the range compensator and subsequently transported through two different patient models. The first model was based on the average CT used by the TPS and the second model included all 10 phases of the corresponding 4DCT. The physical dose and proton linear energy transfer (LET) were computed in each voxel of two models and used to compute constant and variable RBE MC dose on average CT and 4D CT. The MC computed doses were compared to the TPS dose using dose volume histograms for relevant structures.Results:The results show significant differences in doses to the target and critical structures suggesting the need for more accurate proton dose computation methods. In particular, the 4D dose shows reduced coverage of the target and higher dose to the spinal cord, while variable RBE dose shows higher lung dose.Conclusion:The methodology developed in this pilot study is currently used for the analysis of a cohort of ∼90 lung patients from a clinical trial comparing proton and photon therapy for lung cancer. The results from this study will help us in determining the clinical significance of more accurate dose computation models in proton therapy.

Commissioning dose computation models for spot scanning proton beams in water for a commercially available treatment planning system

To present our method and experience in commissioning dose models in water for spot scanning proton therapy in a commercial treatment planning system (TPS). The input data required by the TPS included in-air transverse profiles and integral depth doses (IDDs). All input data were obtained from Monte Carlo (MC) simulations that had been validated by measurements. MC-generated IDDs were converted to units of Gy mm(2)/MU using the measured IDDs at a depth of 2 cm employing the largest commercially available parallel-plate ionization chamber. The sensitive area of the chamber was insufficient to fully encompass the entire lateral dose deposited at depth by a pencil beam (spot). To correct for the detector size, correction factors as a function of proton energy were defined and determined using MC. The fluence of individual spots was initially modeled as a single Gaussian (SG) function and later as a double Gaussian (DG) function. The DG fluence model was introduced to account for the spot fluence due to contributions of large angle scattering from the devices within the scanning nozzle, especially from the spot profile monitor. To validate the DG fluence model, we compared calculations and measurements, including doses at the center of spread out Bragg peaks (SOBPs) as a function of nominal field size, range, and SOBP width, lateral dose profiles, and depth doses for different widths of SOBP. Dose models were validated extensively with patient treatment field-specific measurements. We demonstrated that the DG fluence model is necessary for predicting the field size dependence of dose distributions. With this model, the calculated doses at the center of SOBPs as a function of nominal field size, range, and SOBP width, lateral dose profiles and depth doses for rectangular target volumes agreed well with respective measured values. With the DG fluence model for our scanning proton beam line, we successfully treated more than 500 patients from March 2010 through June 2012 with acceptable agreement between TPS calculated and measured dose distributions. However, the current dose model still has limitations in predicting field size dependence of doses at some intermediate depths of proton beams with high energies. We have commissioned a DG fluence model for clinical use. It is demonstrated that the DG fluence model is significantly more accurate than the SG fluence model. However, some deficiencies in modeling the low-dose envelope in the current dose algorithm still exist. Further improvements to the current dose algorithm are needed. The method presented here should be useful for commissioning pencil beam dose algorithms in new versions of TPS in the future.

SU-E-T-468: Gamma Knife Perfexion Dosimetry: A Monte Carlo Model of One Sector.

We have implemented a Monte Carlo (MC) based dose computation model of one sector of the Gamma Knife Perfexion (GK PFX) using the Penelope MC dosimetry codes. The single sector simulation was rotated about the z-axis to model all eight GK sectors. GK dosimetric aspects examined include: 1) output factors (OF) for each of the three GK collimator sizes (4, 8, 16 mm), 2) OFs for each source row and collimator size, and 3) dose distribution profiles along the x- and z-axes, compared to film measurements and dose calculations from the Leksell GammaPlan (LGP) workstation. We defined the internal GK PFX geometry in Penelope with the aid of vendor-supplied proprietary information. A single source per row was modeled for five rows for each of the 3 collimators (15 beams modeled). MC simulations were carried out on a Linux cluster. Phase space files (PSFs) were collected for the 15 modeled collimators then rotated about the z-axis to model the sector of 24 sources per collimator. 3D dose distributions from the MC model, film, and LGP DICOM-RT dose exports were analyzed using Matlab. For OF calculations, a 16 cm diameter dosimetry sphere was modeled with a virtual detector volume at its center. Good agreement is found for row- and total-output factors (greatest deviation of any type < 4%) compared to reference values. Off-axis factors closely follow LGP predicted dose distributions along the x-axis and differ on the inferior side of the z-axis. Detailed geometric representations (radiation source, device components) of the GK PFX are required for high fidelity MC simulations. Calculated GK PFX OF values depend on the simulated detector volume size (4 mm OF most dependent). Our model shows strong agreement for the GK PFX OFs and dose profile curves compared to reference values. Non-disclosure agreement for proprietary information with Elekta AB. No financial contribution.

TH‐C‐213CD‐01: Accuracy of Proton Dose Computation Algorithms and Need for Improvements

There has been a tacit assumption that semi‐empirical (SE) dose calculation algorithms used in the current practice of passively scattered proton therapy (PSPT) and intensity modulated proton therapy (IMPT) are sufficiently accurate. However, the sharp distal fall‐off of protons, scattering from apertures and other beam line components, interactions of secondary particles produced from beam line components and the patient present significant challenges for the current algorithms. The resulting inaccuracies and their potential clinical impact on outcomes have not been fully appreciated, especially for complex geometries such as those encountered in lung and head and neck. Some of the factors affecting accuracy for PSPT and IMPT are similar while others are quite different. For example, PSPT accuracy is affected by the angular and energy spread of the incident beam and its interactions with compensators and apertures; whereas, scattering from beam‐line components and the transport of secondary particles are more important for IMPT. Complex heterogeneities are of relevance for both PSPT and IMPT. Furthermore, distal edge of protons, assumed to have a sharp fall‐off, may be degraded significantly after passage through heterogeneities. While IMPT may be able to compensate for such degradation, semi‐empirical models are unable to accurately predict this effect.The Monte Carlo (MC) is the most accurate method for calculating proton dose distributions. A Monte Carlo system must be appropriately commissioned by fine tuning the characteristics of the beam entering the nozzle and calibrated using measured data. Once appropriately experimentally validated, Monte Carlo may be assumed to be the “gold standard” for situations where measurements where are difficult, impossible or highly time consuming, e.g., dose distributions in patients. While efforts are being made to accelerate Monte Carlo systems through parallelization and algorithmic improvements, they are likely to be still too slow for routine clinical use especially for applications such as IMPT, robustness analyses, robust optimization, etc. Intermediate solutions, that are simplified variations of Monte Carlo, are being developed. Examples include Macro Monte Carlo, Virtual Monte Carlo and Track Repeating Algorithm for Particles. For lack of a better term, we will call the family of such solutions “abridged Monte Carlo” (aMC). They offer considerable improvement in speed with nearly the same accuracy as Monte Carlo. Nevertheless, greater speed offered by SE models would continue to be required for many of the applications mentioned. A hybrid combination of a SE model and a MC or aMC method may then be applied for certain demanding applications. For instance, in IMPT optimization, most of the iterations may be carried out using a fast SE model interspersed with a small number of MC or aMC iterations.A short introduction giving background and explaining the current state of the art and differences between PSPT and IMPT algorithms will be followed by four presentations. The first one will discuss semi‐empirical dose calculation models for PSPT, their limitations and dosimetric and potential clinical consequences, factors underlying the accuracy limitations and possible improvements. The second presentation will address similar issues for semi‐empirical algorithms for scanning beams and IMPT. The third presentation will discuss the use of Monte Carlo techniques to model proton treatment machine heads for passive scattering and patients and to model scanned beams. It will also talk about calibration and validation of MC systems, inherent uncertainties in MC methods and the efforts to make MC systems easy to use. The fourth presentation will focus on abridged‐Monte Carlo techniques. It will give an overview of acceleration of dose calculations using such techniques and describe one or more algorithms in detail. It will include such topics as transport of particles, validation and calibration, implementation of aMC into a treatment planning systems and applications for IMPT optimization.Learning Objectives:1. Understanding the physics principles underlying current semi‐empirical proton dose computation models, their assumptions and approximations and their effect on accuracy.2. Appreciating the potential impact of approximations on clinical outcomes.3. Understanding the role of Monte Carlo techniques in identifying the causes of differences between results of semi‐empirical models, measurements and Monte Carlo and in the development of more accurate algorithms.4. Learning about the current status of Monte Carlo and abridged‐Monte Carlo techniques for clinical use.Conflicts of interest: Mohan ‐ Research grant from Philips. Soukup ‐ Elekta employee.

SU‐D‐BRB‐02: Monte Carlo Modeling of the Gamma Knife Perfexion

Purpose: For dosimetric and research irradiation studies, we have implemented a Monte Carlo (MC) dose computation model based on the physical and radiological characteristics of the Gamma Knife Perfexion (GK PFX) using the Penelope MC dosimetry codes. GK dosimetric aspects examined include: 1) output factors (OF) for each of the three GK collimator sizes (4, 8, 16 mm), 2) OFs for each source row and collimator size, and 3) dose distribution profiles. Methods: Vendor proprietary information facilitated our modeling of the GK PFX irradiation geometry, which was mathematically defined within Penelope. MC simulations were carried out on a Linux cluster. 3D dose distributions were analyzed using Matlab. A 16 cm diameter dosimetry sphere was modeled with a virtual detector volume at its center. Detector volume varied from 33 to 590 mm̂3 to study detector volume effects. A single source per row was modeled for five rows for each collimator (15 beams modeled). Single‐source dose distributions were rotated about the z‐axis of the axially symmetric geometry and summed to simulate all 192 sources. Results: Good agreement is found for row‐ and total‐output factors (greatest deviation &lt;2% for the 4 mm collimator) compared to reference values. Simulated and measured full‐width at half‐ max values of 3D dose distribution profiles show sub‐millimeter differences (0.4 mm, 4 and 8 mm collimators; 0.9 mm, 16 mm collimator). There is excellent agreement for integrated profile shapes. Conclusions: Detailed geometric representations (radiation source, device components) of the GK PFX are required for high fidelity MC simulations. Calculated GK PFX OF values are dependent on the simulated detector volume size (4 mm OF most dependent). Our model shows strong agreement for the GK PFX OFs and dose profile shapes compared to reference values.Acknowledgement: Non‐disclosure agreement for proprietary information with Elekta AB. No financial contribution.

Radiosurgery of functioning pituitary adenomas: Comparison of different treatment techniques including dynamic and conformal arcs, shaped beams, and IMRT

Purpose: Evaluation of different techniques including intensity-modulated radiotherapy (IMRT) for stereotactic radiosurgery (SRS) of pituitary adenoma (PA). Methods and Materials: Between January 2003 and February 2005, 152 SRS procedures were performed. Ten patients with PA were compared: conformal vs. dynamic arc treatment with micromultileaf collimator (mMLC) vs. circular collimators vs. 8–10 conformal static mMLC beams with and without IMRT. Prescribed total dose: 18 Gy (90%). Constraints: D max optic chiasm 10Gy temporal lobe 10Gy temporal lobe. Results: For the end point "improvement in coverage," an advantage with IMRT was noted for 5 of 10 patients as compared with the dynamic arc approach. Volume treated >18 Gy outside the planning target volume was lowest in 9 of 10 patients after IMRT; 1 patient achieved better conformity with circular collimators. As for Vol 10Gy temporal lobe, an advantage was depicted for 1 of 10 patients with IMRT, the other techniques appearing equally effective in shielding the temporal lobe. With all techniques Vol 10 Gy temporal lobe was max optic chiasm Conclusions: Novalis-based radiosurgery using dynamic arc treatment with mMLC is considered a safe and appropriate approach for SRS of PA.

Conformal therapy using maximum entropy optimization

AbstractConformal therapy offers considerable advantages over conventional radiotherapy treatment, because it has the potential for matching almost exactly the delivered dose distribution to the prescribed dose distribution. Associated (inverse) treatment planning methods address a constrained linear optimization problem. In this article, a method based on maximizing the total entropy of the beam profiles is developed. Maximum entropy optimization constrains the computed dose to be within well‐defined tolerances of the prescribed dose, and has advantages of robustness, fast convergence, and high accuracy. For the work reported here, it is assessed using clinically prescribed irregular target dose volumes based on magnetic resonance imaging and computed tomography images. Results are shown for a two‐dimensional, homogeneous absorption, primary dose computation model, to illustrate the feasibility of the approach; however, the method may be extended to accommodate a more general three‐dimensional model, including inhomogeneities and scatter dose contributions. Optimization of beam offset for a regular angular displacement of beams is also considered, with particular regard to implications on total beam energy, entropy, and computation time.

A comparison of three inverse treatment planning algorithms

Three published inverse treatment planning algorithms for physical optimization of external beam radiotherapy are compared. All three algorithms attempt to minimize a quadratic objective function of the dose distribution. It is shown that the algorithms are based on the common framework of Newton's method of multidimensional function minimization. The approximations used within this framework to obtain the different algorithms are described. The use of these algorithms requires that the number of weights of elemental dose distributions be equal to the number of sample points taken in the dose volume. The primary factor in determining how the algorithms are implemented is the dose computation model. Two of the algorithms use pencil beam dose models and therefore directly optimize individual pencil beam weights, whereas the third algorithm is implemented to optimize groups of pencil beams, each group converging upon a common point. All dose computation models assume that the irradiated medium is homogeneous. It is shown that the two different implementations produce similar results for the simple optimization problem of conforming dose to a convex target shape. Complex optimization problems consisting of nonconvex target shapes and dose limiting structures are shown to require a pencil beam optimization method.

A finite-size pencil beam model for photon dose calculations in three dimensions.

A three-dimensional dose computation model employing a finite-size, diverging, pencil beam has been developed and is demonstrated for Cobalt-60 gamma rays. The square cross-section pencil beam is simulated in a semi-infinite water phantom by convolving the pencil beam photon fluence with the Monte Carlo point dose kernel for Cobalt-60. This finite-size pencil beam is calculated one time and becomes a new data base with which to build larger beams by two-dimensional superposition. The pencil beam fluence profile, angle correction for beam divergence, the Mayneord inverse square correction, radial and angular sampling rates, error propagation, and computation time have been investigated and are reported. Radial and angular sampling rates have a great effect on accuracy and their appropriate selection is important. Percent depth doses calculated by finite-size pencil beam superposition are within 1% of values calculated by full convolution and the agreement with values from the literature is within 6%. The latter disagreement is shown to be due to a low-energy photon component which is not modeled in other calculations. Computation time measurements show the pencil beam method to be faster than full convolution and one implementation of the differential-scatter-air-ratio (dSAR) method.

A dose computation model for 241Am vaginal applicators including the source‐to‐source shielding effects

A dose computation model has been developed for the determination of dose distributions around vaginal plaque applicators containing encapsulated 241Am sources. Encapsulated sources of 241Am emit primarily 60-keV photons which have a half-value layer thickness of 1/8 mm of lead. This makes possible highly effective in vivo shielding of normal tissues at risk, by placing thin lead shields at appropriate places on the applicator. However, self-absorption of photons in the source material itself is intense, requiring bulky sources of about 1 cm diameter. These sources also produce considerable source-to-source shielding which must be taken into account in dose calculations. Our dose computation model for a single source employs three-dimensional integration of dose contributions from volume elements of the source including the effects of absorption and scattering of photons in the source material, titanium encapsulation, and water. An empirical correction to Berger's data on buildup factors of point, isotropic sources is made to account for the effects of anisotropic photon emission by cylindrical 241Am sources. The second part of our dose computation model takes into account source-to-source shielding effects on both primary and scattered photons for the vaginal plaque geometry. The results of the model have been verified for accuracy by comparisons with extensive dosimetry measurements using lithium fluoride thermoluminescent dosimeters.

Dose distributions around cylindrical 241Am sources for a clinical intracavitary applicator.

Encapsulated, cylindrical sources containing 2, 5, and 8 Ci of 241Am have been designed and fabricated for intracavitary irradiation of uterine cancers. Exposure rates in air and dose rates in water around these sources have been measured using an ionization chamber and a lithium fluoride thermoluminescent dosimetry system. Dose rates in water at a distance of 2.5 cm from the source center along a direction transverse to the source axis were found to be 10.4, 24.3, and 23.3 cGy/h for the 2-, 5-, and 8-Ci sources, respectively, using an ionization chamber. Under the same conditions, the thermoluminescent dosimetry system yielded the values of 10.3, 23.1, and 22.3 cGy/h. It was observed that the ratio of dose-to-water and exposure in air is sensitive to the scattering geometry and source geometry in the case of 241Am photons. This ratio was found to increase substantially as conditions of full scattering were approached. A three-dimensional integration model was employed for the determination of dose distributions around these sources. Results of this dose computation model have been compared against the measured data and were found to be in good agreement with each other. Average deviations of calculated data from measured data were in the range of 0.2 to 0.5 cGy/h and larger deviations were observed in the paraxial region, where the effects of oblique filtration are more severe.(ABSTRACT TRUNCATED AT 250 WORDS)

Differential pencil beam dose computation model for photons

Differential pencil beam (DPB) is defined as the dose distribution relative to the position of the first collision, per unit collision density, for a monoenergetic pencil beam of photons in an infinite homogeneous medium of unit density. We have generated DPB dose distribution tables for a number of photon energies in water using the Monte Carlo method. The three-dimensional (3D) nature of the transport of photons and electrons is automatically incorporated in DPB dose distributions. Dose is computed by evaluating 3D integrals of DPB dose. The DPB dose computation model has been applied to calculate dose distributions for 60Co and accelerator beams. Calculations for the latter are performed using energy spectra generated with the Monte Carlo program. To predict dose distributions near the beam boundaries defined by the collimation system as well as blocks, we utilize the angular distribution of incident photons. Inhomogeneities are taken into account by attenuating the primary photon fluence exponentially utilizing the average total linear attenuation coefficient of intervening tissue, by multiplying photon fluence by the linear attenuation coefficient to yield the number of collisions in the scattering volume, and by scaling the path between the scattering volume element and the computation point by an effective density.