Monte carlo simulation of electron beams from varian truebeam linear accelerator used in radiotherapy: estimation of initial beam parameters

In Radiotherapy (RT), electron beams are widely used to treat superficial tumours. The dose delivered to the patients was precisely estimated by Monte Carlo (MC) methods. MC codes need Initial Beam Parameter (IBP) [mean energy, energy spectrum, spot size] and structural details of the machines as input to predict the dose. This study determined the IBP of Varian TrueBeam (TB) Linear Accelerator using PRIMO and validated it using depth dose curves and profiles. The MC Code GEANT 4 (VirtuaLinac) and PENELOPE (PRIMO) were used. The energies used were 6,9,12 & 18 MeV, the applicator used was 10x10 cm2 and 25x25 cm2, and the source to surface distance value was set as 100cm. The MC model available in PRIMO for TB was validated. Initial mean energies were obtained of 6.81, 9.77, 13.15 and 19.79 MeV for 6, 9, 12 and 18 MeV nominal energies, respectively. Sigma of Gaussian distribution was estimated as 0.59, 0.62, 0.71 and 0.82, respectively. The average values of percentage dose difference between the measured and MC simulated PDD and profiles were within 5%. Found that PRIMO is a convenient and easy-to-operate software application with the potential to estimate dose distribution in water phantoms or CT Images precisely.


Introduction
High-energy radiation is used in radiotherapy as a medical procedure to destroy cancer cells or reduce tumour size.In order to kill or inhibit the growth of cancer cells, radiotherapy damages the DNA within the cancer cells.Normal cells nearby may also be harmed by the radiation, although they are frequently more capable of self-healing than cancer cells.Depending on the nature and location of the cancer, as well as the patient's health and treatment objectives, radiotherapy may be used alone or in conjunction with other modalities like surgery or chemotherapy.
Mainly there are two types of radiotherapy: (a) External Beam Radiation Therapy (EBRT) in which the radiation employed comes from a machine outside the body and (b) Brachytherapy in which a radioactive material is placed into the body close to the malignancy.[1] High Energy Linear Accelerators (LINAC) are needed for EBRT, and they are equipped with both photon and electron beams.For the treatment of superficial tumours, electron beams are usually employed.They produce a constant dose in the therapeutic range and have a sharp dose decline at the region of the underlying normal tissue.[2] Dose calculation is an essential part in radiotherapy.[3] This is achieved by using treatment planning systems which will provide the accurate dose and also will plan the whole course of treatment prior to execution in LINAC.The planning system uses different types of algorithms for dose calculation.For radiation therapy dose calculations, MC simulation is thought to be the most precise and trustworthy method.It is a technique for calculating an unknown quantity's value using inferential statistical concepts.But it requires a significant computation time.We must use powerful computing resources to get over this.
In this study, TrueBeam LINAC is modelled using VirtuaLinac, a MC framework, and the phase space files generated were used as input for PRIMO software for dosimetric purposes.The accuracy of PRIMO was also verified by comparing it with the measurements.

Methods & Materials
Simulation and measurements were the methodologies employed in this study.

Simulation methods
In this study, two MC codes were used -GEANT4 and PENELOPE.Varian's VirtuaLinac was used as the framework for the GEANT4 code.Using this framework, the TrueBeam machine is modelled, and phase space files are generated.[4] Phase space file carries information regarding energy, direction and other characteristics including type of particles.These phase space files were used as input in PRIMO software which runs on PENELOPE MC code.
In order to model a linear accelerator, Initial Beam Parameters (IBP) are required.These are the values needed to tune the beam to match the dosimetric parameters of a linear accelerator.[5] The IBP values were determined by trial and error method.Initially, the simulation was started by giving a default set of values as input scored in a virtual phantom (voxel size 2x2x1 mm 3 ), and the corresponding outputs (PDD and profile) were taken.These outputs were compared with the measured data.Then the percentage deviation of simulated and measured values was determined.Based on the percentage deviation, the parameters are changed and started the second iteration.This process is repeated until the simulated and measured values are precisely matched.Through this method, IBP was estimated.Using these IBP values in VirtuaLinac, TrueBeam LINAC was accurately modelled, and the phase space files were generated.These phase space files were used as the input for PRIMO software.In order to use this file as input for PRIMO, two components are needed.a) phase space file and b) Header file.The header file contains all the information about the phase space file, similar to the contents page of a book.A python program is used to generate a header file from the phase space file.The equipment and software used for the simulation are given below.

Monte Carlo (MC) Simulation
MC Simulation is a mathematical technique used to predict the outcome of an uncertain event invented by John von Neumann and Stanislaw Ulam.[6] By using probability distributions, it will first generate a set of random variables, and then it will be used as input for an uncertain variable and thereby creating multiple results.After this, the average of all these results will be taken for predicting the final outcome.As the number of inputs increases, the accuracy of the final result also increases.So, it is also known as multiple probability simulation.This method is highly accurate compared to other predictive models.So, it is used in different areas where risk and uncertainty are higher.

GEANT4
GEometry ANd Tracking (GEANT4) is a set of tools for modelling how particles move through and interact with matter.[7] It comprises of a wide variety of processes, including geometry, hits, tracking, models and physics.It is an object-oriented technology that was created by scientists from all over the world.C++ is the programming language used to implement it.

VirtuaLinac
Using GEANT4, a TrueBeam machine is modelled with the help of the VirtuaLinac framework, which is deployed using cloud computing.A precise and customisable model of the TrueBeam treatment head is available in VirtuaLinac, whose output is distributed and recorded to phantoms and CT datasets.[8] Phase space files can be saved for later use, fed into another simulation process, or to study particle distributions.[4] Phase space file is a collection of information that specifies the kind, energy, direction and other characteristics of particles that cross the scoring plane.

PENELOPE
Penetration and ENErgy LOss of Positrons and Electrons (PENELOPE) is a multipurpose MC algorithm designed to simulate coupled photon-electron transport in materials.[9] It is programmed as a collection of FORTRAN language subroutines that do particle tracking for photons, electrons or positrons and random interaction sampling.To track the particle histories through the material composition and to maintain the outcome of specific quantities, the user should provide the primary guide program.

PRIMO Software
It is a MC program, and PENELOPE is its calculation engine.To run the PENELOPE code, a main program called PENEASY is used.[10] Although primarily designed as research software, it is getting tested for numerous uses in routine clinical practice.By simulating linear accelerators, it calculates the distributions of absorbed dose in computed tomography and phantoms.[11] This program has a simple graphical user interface that makes it easy for users to run simulations and conduct result analysis.

Amazon Web Services (AWS)/Amazon Cloud
The entire simulation infrastructure can be set up using cloud computing technology to access a powerful computational network whenever needed.[12] As MC simulations require powerful computers to reduce their computational time, in this study, Amazon Web Services was used for simulating the TrueBeam treatment head.With the help of this method, the LINAC model is fine-tuned and calculated the beam parameters of the incident electron beam quicker.Also, the high initial expenses of setting up an MC cluster are removed.

Measurements
Measurements (PDDs and profiles) were performed on TrueBeam linear accelerator.The energy used were 6, 9, 12 and 18 MeV electron beams.The field size used were 10x10 cm 2 and 25x25 cm 2 electron applicators.Both in-plane and cross-plane profiles were taken.Details of the equipment used for measurement are given below.

Linear Accelerator (LINAC)
Charged particles like electrons can be accelerated to high energy through a linear tube with the help of electromagnetic waves.This device is known as LINAC.[1] Superficial tumours can be treated using these electrons, or they can be used for the production of x-rays.The LINAC employed in this research is known as TrueBeam (Varian Medical Systems, Palo Alto, CA, USA).It is a highly precise and advanced system with the help of which radiation oncologists can offer extremely accurate and efficient radiation therapy to cancer patients.Electron beams are used in this study.In order to limit the size of an electron beam, an applicator is needed, which is connected to the treatment head of the LINAC.

CC13 Ionisation chamber
The ionisation chamber used in this study is CC13.It is a device used to evaluate photon and electron beams' absolute and relative dosimetry for measurements in solid phantoms and water.This is regarded as a typical ionisation chamber for use in water phantoms in the clinical field.PDDs and profiles are measured by using this.[1] The CC13 ionisation chamber has the following distinguishing characteristics: it is designed for lateral or axial beam entrance, is completely guarded and waterproof, has an air ionisation chamber, and is vented through a waterproof silicon sleeve.

Radiation Field Analyser (RFA)
Measurements were made using RFA.It is a tool for measurement and analysis of the radiation field generated by medical linear accelerators.The three-dimensional servo (the water tank with the mechanics), two single ionisation chambers (field and reference) and a chamber control unit (CCU) with an integrated two-channel electrometer are the key components.The RFA used in this study was Blue phantom (IBA Dosimetry, Nuremberg, Germany).

Scanning software
The software used for scanning is Omni Pro, developed by IBA Dosimetry.In relative dosimetry, acceptance, commissioning, and Quality Assurance (QA) of the linear accelerator is carried out by this software.

Results & Discussions
The results obtained in this study are given below.

Initial Beam Parameter (IBP)
The IBP values determined during the TrueBeam machine's modelling are tabulated below.These values are determined by trial and error method.Once these values are given as the starting parameters for beam modelling, they result in perfect output, which are agreed with in 3 to 4 % of the measured value.

PDD Comparison
The percentage Depth Dose for simulated data along with the measured data for 10 x 10 cm 2 and 25 x 25 cm 2 applicators are illustrated in figure 6 and figure 7, respectively.In figure 8 and figure 9, the percentage dose difference between MC simulated and measured values for 10 x10 cm 2 and 25 x 25 cm 2 applicators are given.Analysis of the PDD curves revealed that the dose difference increased with an increase in depth.The average dose difference for all four energies (6,9,12, 18 MeV) for a 10x10 cm 2 applicator was 4%, 3%, 3% and 1%, respectively.Similarly, for the 25x25 cm 2 applicator, the values are 4%, 2%, 2% and 1%, respectively.

Profile
Profiles of 25 x 25 cm 2 applicator were taken through simulation and measurement.It was found that the average dose difference inside the field for all four energies (6,9,12,18 MeV) is within 2%, and the value increased beyond 2% at the edges.This is illustrated in figure 10    Simulated output from the PRIMO software (PDDs and Profiles) was also compared with the measured values, and they also agreed with in 3 to 4%.

Features of PRIMO (which makes it convenient and simple to use)
Compared to other MC codes and frameworks available, PRIMO is based on a Graphical User Interface (GUI), which is user-friendly and self-explanatory.Inbuilt tools (including gamma analysis) are available in PRIMO software to compare measured and simulated data.The user requires high-level programming skills to perform these operations in other MC codes and frameworks.It accepts the phase space files generated by other MC programs in IAEA format as input.That feature helps the user to save a lot of time which will be wasted otherwise in simulating the field-independent section of the simulation.It is free software, lightweight (less than 100 MB of disc space), effortless to install, and even runs on Windows 64 bits operating systems.

Limitations and potential sources of error during simulation
The accuracy of the simulation depends upon the total number of particles selected initially.The optimum value for the total number of particles determined during this study (by trial-and-error method) is of the order of 10 8 to 10 9 .If the number of particles is less than this range, it will affect the output negatively.Every PS file is associated with a header file.So, care should be taken to upload the header file along with PS files in PRIMO.The simulation requires a significant computation time.If the processors/ memory is not the best quality, it will take many hours (even days) to complete a simulation.High-efficiency computers/supercomputers can minimise the calculation time.So, using cloud computing is a viable solution (economically) and will be the better option for reducing the simulation time.

Limitations and potential sources of error during the measurement
The following are the limitations and sources of error that may occur during the measurement.
• Field detector alignment/positioning: The centre of the field detector should align precisely with the centre of the field.Or in other words, the reference point of measures should be as per the protocol (SSD/SAD).• Reference detector positioning: The reference detector should not interrupt or interfere with the charge collection of the field detector.Also, the positioning should be on the edge of the field.• Effective point of measurement: Effective point of measurement for the particular detector used for scanning should be calculated and applied.(0.6 x rcyl), where rcyl = radius of the chamber cavity.This is only applicable to cylindrical chambers.• Selection of appropriate depth for profile measurements: The depth value should be determined and entered in the scanning software as per the energy and protocol (AAPM/IAEA/IEC) used for the measurement.• The step size of detector movement: In high gradient regions (e.g., penumbra region of profiles OR region near to dmax for PDD) should always be measured using a small step size (1mm) compared to a large step size (3mm).• Surface dose in PDD measurements: Care should be taken for the surface dose measurements because even a minimal difference in the positional accuracy of detectors will result in a significant dose error.• RFA Water level: The frame through which the detector moves should always be parallel to the water surface, which should be verified carefully by aligning the RFA water level.• Penumbra margin in profile measurements: at least 3.5cm beyond the field edge to get a considerable penumbra margin for profile measurements.• RFA alignment: RFA Crosswire and the Machine Crosswire should match.
• Bias voltage selection for different detectors: Depending on the type of detector, the user needs to determine whether the voltage needs to be applied.For diodes, voltage is not applied.In the case of ion chambers, voltage is always applied in the range of 200V to 400V.

Conclusion
This study enabled to model a TrueBeam machine accurately using VirtuaLinac and GEANT4 MC code.The output from this machine is verified using measured values of PDDs and profiles.It is found that the MC calculated and measured data are in good agreement.It is demonstrated that the TrueBeam phase space files for the 6 MeV, 9 MeV, 12 MeV and 18 MeV beams generated using VirtuaLinac may be used as radiation sources (input for PRIMO software) for precise MC simulations leading to accurate dose estimation.Also, it was found that PRIMO is a practical and simple to use program with the capability to accurately estimate dose distribution in water phantoms or CT images.

Figure 2 .
Figure 2. PRIMO Software: Showing the Simulation Setup window and the representation for Simulation Segments (S1, S2 and S3).Courtesy of PRIMO Project [German Research Foundation (DFG)].

Figure 3 .
Figure 3. Amazon Web Services (AWS) is a cloud computing platform used to run the VirtuaLinac MC Simulation framework.[Courtesy of AWS Inc. Seattle, WA, USA].

Figure 4 .
Figure 4. Measurement setup: Electron Applicator attached to the TrueBeam LINAC head and the Radiation Field Analyser (RFA) [Courtesy of Regional Cancer Centre (RCC), Thiruvananthapuram, Kerala, INDIA].