Dosimetric characteristics of LinaTech DMLC H multi leaf collimator: Monte Carlo simulation and experimental study

Abstract This study evaluated the basic dosimetric characteristics of a Dynamic Multi Leaf Collimator (DMLC) using a diode detector and film measurements for Intensity Modulated Radiation Therapy Quality Assurance (IMRT QA). The EGSnrc Monte Carlo (MC) simulation system was used for the determination of MLC characteristics. Radiation transmission and abutting leaf leakage relevant to the LinaTech DMLC H were measured using an EDGE detector and EBT3 film. In this study, the BEAMnrc simulation code was used for modeling. The head of Siemens PRIMUS linac (6 MV) with external DMLC H was entered into a BEAMnrc Monte Carlo model using practical dosimetry data. Leaf material density, as well as interleaf and abutting air gaps were determined according to the computed and measured dose profiles. The IMRT QA field was used to evaluate the dose distribution of the simulated DMLC H. According to measurements taken with the EDGE detector and film, the total average measured leakage was 1.60 ± 0.03% and 1.57 ± 0.05%, respectively. For these measurements, abutting leaf transmission was 54.35 ± 1.85% and 53.08 ± 2.05%, respectively. To adapt the simulated leaf dose profiles with measurements, leaf material density, interleaf and abutting air gaps were adjusted to 18 g/cm3, 0.008 cm and 0.108 cm, respectively. Thus, the total average leakage was estimated to be about 1.59 ± 0.02%. The step‐and‐shoot IMRT was implemented and 94% agreement was achieved between the film and MC, using 3%‐3 mm gamma criteria. The results of this study showed that the dosimetric characteristics of DMLC H satisfied international standards.

modulating the intensity of radiation beams, depending on the depth and type of tumor, and have many applications in the creation of volume dose distribution in the three-dimensional form which is in accordance with the shape of the tumors. MLC provides all of these capabilities in IMRT treatments.
The specific features of any kind of MLC depend on the materials. Therefore, it is important to determine the dosimetric properties of the MLC and its effects on the dose distribution. 2 Clinical consequences resulting from the incorrect determination of these features have been previously reported in IMRT treatments. [2][3][4][5] Studies considered leaf leakage shares as well as tongue and groove effects in dose calculations, especially in IMRT treatments with a long duration of radiation time which results in an increase in leaf transmission share in delivering an extra dose to the patient. 6,7 Therefore, an accurate determination of the dosimetric characteristics of MLCs using an appropriate dosimetry tool is one of the most important parameters in QA tests, in the field of IMRT treatments. 8 Gafchromic films with special advantages and capabilities are among suitable tools recommended for IMRT QA. 9,10 The possibility of using it in water and in solid water phantoms, its high spatial resolution, the possibility of using it in a wide range of radiation and its low dependency on energy, makes the film a strong tool in the field of dosimetry. 11 The BEAMnrc Monte Carlo code also has significant applications in MLC modeling 12  In this study, several dosimetric properties of the DMLC H multi leaf collimator were measured and evaluated. In addition, to determine certain special characteristics of MLC, Monte Carlo modeling was considered.

2.A | Monte Carlo modeling
In accordance with the manufacturer's geometry and materials, a 6 MV medical linear accelerator (Siemens PRIMUS model) was modeled using the BEAMnrc/EGSnrc (Version V4-r2-4-0) simulation software. All parts of the linac head including the target, primary collimator and flattening filter, monitor ion chamber, mirror and X-Y jaws (secondary collimators) were modeled using modules provided by the code. The VARMLC module was used to simulate the external DMLC H. Using the DOSXYZnrc software, dose calculation was performed in a water phantom.
To reduce the simulation run time and increase the efficiency, variance reduction techniques were used in the simulations. The global cut-off energy for electron and photon particles was set to 0.7 and 0.01 MeV, respectively. To increase the number of photons generated in the target, Directional Bremsstrahlung Splitting (DBS) was used. 13 Therefore, to maximize dose and fluence efficiency at 6 MV beam energy, NBRSPL (DBS splitting number) was set to 1000. The DBS splitting field radius was equal to the side of the square field to be defined (square field defined at a certain distance from the target). Electron range rejection was also used with the ESAVE parameter, which is the energy threshold required to turn on the range rejection, set to 2 MeV. 13,14 The Monte Carlo simulations were validated in two steps. In the first step, the Siemens linear accelerator head (in the absence of MLC) was simulated and validated according to the practical measurements. In the second step, MLCs were added and validated according to the dosimetry data.

2.A.1 | Simulation of the Siemens PRIMUS linac
Realistic and reliable results are obtained when accurate details based on the manufacturer's data, are used for the simulation. One of the most important parameters in the accurate modeling of the linac is the target of the accelerator. 15 Another important parameter is the simulation of the flattening filter, which is located inside the primary collimator and is used to flatten the beam at a certain depth.
The average energy of the produced beam depends on the geometry and the materials used in this piece. 16 Hence, the target and flattening filter play a significant role in MC simulation results; therefore, the target equipped with a flattening filter can be named the heart of the simulation.
Other constituent parts of the head, including the parallel plate ionization chamber and mirror were simulated. Thereafter, two pairs of tungsten jaws which lie in a perpendicular direction to each other were modeled. These jaws were used for radiation beam collimation in the required field sizes. The geometric shape of the simulated Siemens linac head is shown in Fig. 1.
Source number 19 in BEAMnrc Monte Carlo code was used for modeling initial electron beam energy. This source has a monoenergetic beam with two-dimensional distribution of Gaussian intensity. 17 The initial electron beam parameters (energy and radius) were determined in accordance with the method proposed by Sheikh-Bagheri and Rogers. 18 According to this method, the energy and size of the electron beam will be determined if a good agreement is found between the simulated and measured data. 19,20  and lateral dose profiles (for determining energy and especially FWHM), electron beam energy and its radius size were determined. These assessments were performed at different square field sizes and at various depths (d max , 5 and 10 cm).
Phase space files were used in the validation process. Phase space files with different arrangements of energies and FWHMs were generated at a 100 cm Source Surface Distance (SSD).
Thereafter, symmetrical field sizes of 3, 5, 10, 15, and 30 cm 2 were defined by the jaws. The number of histories in different field sizes was different. The number of electron particles irradiated to the target in the standard size field (10 9 10 cm 2 ) was about 2 9 10 8 . For field sizes greater than 10 9 10 cm 2 , about 3 9 10 8 electron particles were irradiated.
The generated phase space files in the linac isocenter were used as input data in the DOSXYZnrc code. In field sizes of 3 9 3 cm 2 and 5 9 5 cm 2 , the surface of the water phantom was irradiated with 10 9 10 9 particles. The particles are photons, electrons, and positrons but the primary particles are mainly photons. In the standard field size, the number of histories was set to 20 9 10 9 and the number of photon particles increased with increase in field size. The global cut-off energy for electron and photon particles was set to 0.7 and 0.01 MeV, respectively. NBRSPL was set to 1000. Depending on the field size, different computational resolutions were considered. Computational resolution in low-dose gradient areas (the region of 80% profile relative to the central axis) and in regions with high-dose gradient (penumbra regions), was set to 2 and 1 mm, respectively. In the direction of the central axis (CAX), the resolution was 1 mm. Therefore, five kinds of phantoms were defined and utilized in the five radiation fields. Considering the number of photon particles used in the simulation, the Monte Carlo uncertainty was less than 1% (2 SD).
The configuration and constituent materials of DMLC H were modeled based on the manufacturer's data; and the other simulation parameters relevant to DMLC H, including MLC density, Z focus of the leaf sides, Z min (Z of the top of the MLC), interleaf air gap and abutting leaf gap were investigated. Since the average IMRT beam is about 10 9 10 cm 2 , this field size was selected to determine MLC leakage and transmission. The standard field size was defined using jaws while the MLC was removed from the radiation field (the MLC field size was 30 9 30 cm 2 ). This field size was defined as the Refer-

Interleaf leakage, intraleaf transmission and MLC leakage
Different simulation parameters, such as MLC material density, Z focus, interleaf air gap, Z min , and abutting leaf gap were chosen according to the measurements. By so doing, the difference between simulation and measurement results were obtained. The first three parameters were determined using the conditions created in the B-MLC mode.
The inline profile diagram of the MLC leakage was simulated by applying different numerical changes in the three parameters (density, Z focus, and interleaf air gap). Thereafter, the results were compared with the leakage profiles obtained from the Radiochromic film and diode detector. For this purpose, MLC material density in the range of 16-19 g/cm 3 , Z focus in the range of À50 to +50 cm and air gap in the range of 0.004-0.03 cm were changed. First, to determine the air gap between adjacent leaves, the air gap value was set as 0.004 cm and then gradually increased. By adjusting the interleaf air gap in the mentioned range, the peaks and valleys on the inline leakage profile were observed. Then, by adjusting the density and Z focus, the best agreement between simulated and experimental dose profiles were obtained.
In BEAMnrc simulation, approximately 4 9 10 8 particles were used for the B-MLC field, of which about 31 9 10 6 particles were registered in the phase space file located in the linac isocenter. Considering the number of photons irradiated to the surface of the phantom (4 9 10 9 particles), the statistical uncertainty for the DOS-XYZnrc calculations was less than 2% (1 SD). To extract the inline leakage profile, dose values in the B-MLC field were normalized to the CAX dose value of the RFS field.

Abutting air gap
The abutting air gap parameter was extracted through the C-MLC  Semiflex cylindrical ionization chamber with 0.125 cm 3 nominal sensitive volume (type TN31010, PTW-Freiburg, Germany) in a motorized 3D Scanner TM (model 1230, Sun nuclear Corporation, Florida, USA). The SNC Dosimetry Software (version 1.3.2) was used to control the motor system in a 3D Scanner. Moreover, this software was used for data collection and processing. In the SNC 3D Scanner, after the detectors were attached to the scanning system, the water tank which uses water sensors at three different locations in the tank was automatically leveled. Thereafter, using auto-setup procedure, the position of the central axis point was automatically (without manual setup) determined. In addition, the Sun Nuclear EDGE Detector TM (model 1118) was used for measurements related to A-MLC, B-MLC, C-MLC, and radiation fields smaller than 5 9 5 cm 2 .
The EDGE detector is a kind of silicon diode detector with an active volume of 0.000019 cm 3 . Unlike the Semiflex detector, the EDGE detector is oriented horizontally so that the top surface which is The EBT3 films were scanned 48 h after irradiation using the Microtek scan wizard pro V7.26 software. To improve film response and reduce the error (about 9%) caused by the incorrect placement of the film on the scanner, it was scanned in the landscape orientation so that the shorter side of the film was placed along the long side of the scanner. 25 To obtain raw data, the use of any type of filter and image processing tools was avoided. The films were scanned F I G . 3. Equipment used for practical dosimeters, which consist of the SNC 3D Scanner, SNC EDGE detector and the PTW Semiflex ionization chamber.
in the full dynamic range condition, in the transmission mode and in 48-bit RGB color mode with a spatial resolution of 127 dpi (0.2 mm) and saved in the TIFF file format.
For the extraction of calibration curve, pieces of film were exposed to doses ranging from 0 to 2000 cGy. To reduce the film dose-response uncertainty and improve the accuracy of the sensitometric calibration curve, each dose level was repeated three times and the mean net Optical Density (netOD) was used to obtain the calibration curve.
In order to deliver precise doses to the films, the value of the absorbed dose corresponding to each dose level was obtained with a PTW Semiflex chamber of 0.125 cm 3 (model 31010) mounted at a depth of 5 cm within a solid water phantom. All measurements were performed according to the IAEA TRS 398 protocol. 21 To obtain netOD, prior to exposure, the initial OD (OD initial ) of the unirradiated films was calculated using eq. 1.
where PV unexp , PV unopaque and PV unblank represent the pixel values of the unexposed film, opaque sheet scan and pixel value of the blank screen, respectively. Finally, after irradiation, the netOD was obtained using eq. 2.
where PV exp , PV opaque , and PV blank represent the pixel values of the exposed film, opaque sheet scan and pixel value of the blank screen, respectively. It should be noted that to obtain the netOD, the OD initial of each piece of film was calculated separately. In fact, a generic background was not used. Instead, the OD initial was used as the background.
The experimental dose, fitting dose and the total dose uncertainties were estimated by error propagation as proposed by Devic et al. 24 For the analysis of films, an area of 1 cm 2 was selected from the central part of the film (50 9 50 pixels). 24 The Levenberg-Marquardt algorithm was used to obtain an appropriate calibration curve and minimize the fitting uncertainty. 26,27 Finally, the calibration curve was obtained by fitting a third-degree polynomial curve.
In A-MLC, B-MLC, C-MLC, and RFS fields, films were cut into 4 9 25 cm 2 and then in accordance with the intended position, they were placed in a solid water phantom. For B-MLC 11500 MU (Monitor Unit), C-MLC 1000 MU, A-MLC, and RFS field 300 MU were exposed to films. For square fields (with and without MLC), the films were irradiated with 300 MU. Also, the required MU for measuring MLC leakage (in B-MLC field) using the EDGE detector was less than 1000.
All radiations were carried out by Siemens Primus linac (6 MV) and before irradiation; output was tuned to 1 cGy/MU. The spatial resolution of MC calculations and detector readings was 1 mm.

2.C | IMRT QA field
The QA tests are one of the basic tests required to commission different computing systems in radiotherapy centers. 5 One of the tests recommended by the American Association of Physicists in Medicine (AAPM) is test No. 4 of AAPM TG-119. 28 According to this test, the "C" shaped PTV has a length of 8 cm, the inner radius of 1.5 cm, and outer radius of 3.7 cm. A cylindrical organ at risk is located inside the target and the center is concentric with the center of the In all the assessments, an area of 10 9 10 cm 2 (with a pixel spacing of 1 mm) was selected, and thus, approximately 10000 pixels were analyzed using the gamma index. Gamma analysis is a dimensionless function that simultaneously takes both Dose Difference

3.B.2 | Abutting air gap
In determining abutting air gap, a good agreement was observed with 0.108 cm. Thus, based on the Monte Carlo simulation, the amount of abutting leaf leakage was 52.80 AE 0.06%. According to these results, the leakage discrepancy for abutting air gap between the MC simulation and EDGE detector was À2.85% and between the MC simulation and EBT3 film, it was À0.53%. It is evident that the presence of abutting air gap causes an increase in approximately 53% more than the prescribed dose. Figures 7(a) and 7(b) show the abutting leaf dose profile obtained from the MC simulation, diode detector and film measurements in the direction of the X axis.
To better analyze the results, a summary of the dosimetric parameters of DMLC H are presented in Table 2.

3.C | IMRT QA field
The planar dose distributions between the planned and actual dose distributions were assessed using gamma function. The gamma test results for different criteria are provided in Table 3. Figure 9(a) shows the gamma dose map between MC and the film with 3%-3 mm gamma criteria. The results of the gamma analysis using 3%-  In addition, the leaf structure, design, and shape of the leaf and  energy spectrum can also be affected. Therefore, these factors caused the scatter space distribution in the center of the profile to be higher than other areas and eventually, the amount of transmission became more than that of other areas of the profile. 33 As Table 3 shows, the gamma index pass rate for Monte Carlo calculations is greater than the FSC algorithm. This is because the However, in the current research, a good agreement was obtained between the simulated and measured dose distribution (IMRT QA field).  The authors give special thanks to Kamal Mostafanezhad for help in carrying out practical dosimetry and also Farzad Tajdini for providing all the specifications required for the LinaThech DMLC H simulations.

CONFLI CT OF INTEREST
The authors declare that they have no conflicts of interest.