Dosimetric validation of Monaco treatment planning system on an Elekta VersaHD linear accelerator

Abstract The purpose of this study is to perform dosimetric validation of Monaco treatment planning system version 5.1. The Elekta VersaHD linear accelerator with high dose rate flattening filter‐free photon modes and electron energies was used in this study. The dosimetric output of the new Agility head combined with the FFF photon modes warranted this investigation into the dosimetric accuracy prior to clinical usage. A model of the VersaHD linac was created in Monaco TPS by Elekta using commissioned beam data including percent depth dose curves, beam profiles, and output factors. A variety of 3D conformal fields were created in Monaco TPS on a combined Plastic water/Styrofoam phantom and validated against measurements with a calibrated ion chamber. Some of the parameters varied including source to surface distance, field size, wedges, gantry angle, and depth for all photon and electron energies. In addition, a series of step and shoot IMRT, VMAT test plans, and patient plans on various anatomical sites were verified against measurements on a Delta4 diode array. The agreement in point dose measurements was within 2% for all photon and electron energies in the homogeneous phantom and within 3% for photon energies in the heterogeneous phantom. The mean ± SD gamma passing rates of IMRT test fields yielded 93.8 ± 4.7% based on 2% dose difference and 2 mm distance‐to‐agreement criteria. Eight previously treated IMRT patient plans were replanned in Monaco TPS and five measurements on each yielded an average gamma passing rate of 95% with 6.7% confidence limit based on 3%, 3 mm gamma criteria. This investigation on dosimetric validation ensures accuracy of modeling VersaHD linac in Monaco TPS thereby improving patient safety.

within 5% as recommended by International Commission on Radiation Units and Measurements (ICRU) report 24. 3 Monte Carlo (MC)-based dose calculation engines have been reported to have the potential to better the 3% requirement for dose uncertainty. 4 MC uses a stochastic method to calculate dose from first principles that accounts for material details of the treatment head. 5 MC has been shown to calculate accurate dose distributions, especially in heterogeneous patient tissues involving complex electron transport trajectories. 6 Its advantage over conventional dose engines is that the uncertainties are independent of setup leading to increased confidence in the calculated dose distribution. However, MC dose calculation engines suffer from increased planning time, statistical uncertainties from limited number of histories sampled, mismatch between measured and modeled data, and conversion of CT data to physical density data. 7 This calls for validation of MC dose calculation with dosimetric measurements followed by clinical studies. 8 In this study, we perform the dosimetric verification of Monaco TPS version 5.1 (Elekta CMS, Maryland Heights, MO, USA) on an Elekta VersaHD linac (Elekta, Crawley, England). This study was initiated to investigate the dosimetric accuracy prior to clinical usage due to two main reasons. First, the newly designed Agility head has a dynamic leaf guide with a variable thickness combined with the 160 multileaf collimators (MLC) without a backup jaw. The MLCdefined collimation in one axis could lead to higher penumbra and alternate head leakage spectrum. 9 Second, the high dose rate FFF beams have a lower out-of-field dose, different electron contamination spectra, and possibly lower mean energy. [10][11][12] Calibrated ionization chamber-based point dose measurements were followed by diode array-based 2D dose measurements which were compared with planned dose distribution using Gamma analysis.
Monaco calculated dose distributions in a few head and neck (H&N), brain, lung, and abdominal treatment sites were compared against dosimetric measurements. We have investigated overall performance of treatment delivery by quantifying the confidence limits on dosimetric accuracy using benchmarks set by task group 119 (TG-119). 13

| ME TH ODS AND MATERIALS
The treatment machine is an Elekta VersaHD linac with 6 MV,

2.A | Commissioning beam data
Monaco commissioning beam data acquisition was based on the manufacturer instructions as well as recommendations of AAPM TG-106. 14 A PTW MP3-M water tank (PTW, Freiburg, Germany) was used with a PTW Semiflex 31010 chamber (0.125 cm 3 active volume) or a PTW Diode P dosimeter (active volume = 0.03 mm 3 ) for dosimetric measurements and data were processed using PTW's MEPHYSTO mc 2 Navigation software. The percent depth doses (PDD), output factors, and beam profiles were acquired at 90 cm source-to-surface distance (SSD) for square field sizes from 1 9 1 cm 2 up to 40 9 40 cm 2 . PDDs, output, and wedge factors were acquired with a PTW Diode P dosimeter (active volume = 0.03 mm 3 ) for fields ≤5 9 5 cm 2 and with the PTW Semiflex chamber for field sizes ≥5 9 5 cm 2 . A daisy chain approach was used in integration of data, and measurements of the two dosimeters were normalized to a 4 9 4 cm 2 field size. All profile scans were performed using the PTW diode P dosimeter for better spatial resolution. This includes inplane, crossplane, and diagonal profile scans acquired at depths of d max , 5 cm, 10 cm, and 20 cm. Penumbra was measured from the spatial distance between the 80% and 20% of the central axis value in the profile scan of the flattened beam. For FFF beams, the penumbra normalization technique where-in the profiles were normalized to the largest field size was utilized. 15

2.B | Dosimetric verification
In this study, the virtual linac model built by Elekta was verified using a set of measurements recommended in AAPM's Medical Physics Practice Guideline (MPPG) report 5a. 16  The values of P ion and P pol were measured during TG-51 calibration as a part of commissioning. field with three abutting field segments and "7segA" field consisting of seven segments were used to validate MLC leaf tip offset position.

2.C | IMRT test fields
The "FourL" field made of four L-shaped segments was used to adjust the MLC transmission. The dynamic MLC field "DMLC" was used to authenticate the combination of MLC transmission, leaf offset, and leaf tip leakage. The high-density MLC field "HDMLC" and high-dose IMRT field "HIMRT" were representative clinical fields that fulfill the purpose of a final endorsement of the adjustments made using the previous fields to ensure that the plan agreement is appropriate.
IMRT test fields optimized in Monaco TPS were subjected to dosimetric testing by comparison of planned dose distribution against measured data using gamma analysis. The fields were measured for all the photon energies on the Delta 4 bi-planar diode array (ScandiDos, Uppsala, Sweden). The array consisting of 1069 diode dosimeters with 5 and 10 mm respective spacing in the center and peripheral region was specifically commissioned for the VersaHD linac, as specified in the manufacturer's guidelines. Gamma analysis was performed using a 2% dose deviation (DD), 2 mm distance-to-agreement (DTA) criteria, and 10% dose threshold based on a global normalization.

2.D | Patient data validation
We used the VersaHD virtual machine model in Monaco TPS to calculate dose for H&N, brain, lung, and pelvis treatment sites which were compared against measurements made on the Delta 4 diode array. The results of the gamma analysis were tabulated for 3% DD, 3 mm DTA, and 10% dose threshold based on a global normalization. As proposed by Palta et al, a quantified degree of agreement that should be acceptable based on the "confidence limit" (CL) was utilized here. 17 Based on the modified definitions laid out in TG-119, the CL is the reduction from 100% of points passing the gamma criteria summed with 1.96 times the standard deviation (SD). It is expected that 95% of measurements would fall within this CL based on normal distribution. The measurements were performed five times for each of the eight patient plans.

3.A | Commissioned beam data
The virtual machine was modeled based on the commissioning beam data and a comparison between some of the measured and modeled beam data was illustrated in

3.B | Dosimetric verification
The parameters of the fields and the point dose differences are tabulated in Table 1 for all photon energies. Agreement in the measured data was within 2% and 3% in homogeneous and heterogeneous phantom for photon energies, respectively. The agreement in the measured data was within 2% for the electron energies in homogeneous phantom, as tabulated in Table 2. These results agree with the recommended tolerances mentioned in MPPG report 5a. The density and mean HU values of various inserts in a CT electron density phantom were tabulated in Table 3. Also shown are the comparable HU values from the virtual CT simulator, GE LightSpeed 16slice CT (GE Healthcare, Waukesha, WI, USA).

3.C | IMRT test fields
The Monaco commissioning test fields yielded passing percentage of 93.8 AE 4.7% in the gamma analysis using 2% DD, 2 mm DTA criteria (as tabulated in Table 4). Among the eight test fields measured, the profile of the measured and planned FourL field, which had the lowest mean gamma passing rate of 87.6 AE 6.6%, is shown in Fig. 3.

3.D | Patient data validation
Monaco TPS calculated dose distribution was compared against Delta 4 measurements on two brain, two H&N, two lung, and two T A B L E 1 Point dose differences between measurements and Monaco TPS calculated data in the Plastic water at upper point (8 cm in water) and heterogeneous phantom at lower point (26 cm physical depth, as shown in Fig. 1 Open fieldlower point   16 The reference condition dose and relative dose measurements are within the 0.5% and 2% tolerance recommendations of MPPG 5a (refer Table 3), respectively. The heterogeneity point dose measurement falls within the tolerance of 5%. The penumbra values of profile scans were within the 3-mm tolerance mentioned in Table 5 in MPPG 5a. The reference and relative dose measurements for electron beams at two SSDs stated in Table 2 are within the 2% tolerance stated in MPPG report 5a.
AAPM Therapy Emerging Technology Assessment Work Group report on FFF beams mentions that unlike Varian, Elekta linacs have an independent energy set for FFF mode compared to the flattened counterparts that allows penetrative quality to match with the nominal value for that energy. 19 PDDs of FFF beam show deeper d max and steeper fall-off with depth than the corresponding flattened energies. The collimator scatter factor and output factors were considerably lower for FFF beams for field sizes above 10 9 10 cm 2 than their flattened counterparts. 20   detailed study on MLC characteristics in the Agility head. 9 A systematic end-to-end testing to ensure confidence in modeling the dosimetric characteristics of the upgraded Elekta Agility head and FFF beams was addressed by Saenz et al. 22 In all the patient IMRT QA validations, the average percentage of points passing gamma criteria (3% DD/3 mm DTA) exceeded 90%.
Combining all the results of the eight site-specific plans gives an overall average of 95% with a standard deviation of 0.9%. The confidence limit for these results was 6.7%, indicating that the percentage of points passing gamma criteria should be more than 93.3% approximately 95% of the time. From these collective measurements, 90% of the tests fell within the confidence limit. However, this analysis suffers from any statistical test dealing with limited number of dataset (n = 8) and no major changes in our clinic's dosimetry practice was required as a result of using a MC-based planning system.

| CONCLUSIONS
Point dose measurement agreed within 2% in a homogeneous phantom for all photon, electron beams and within 3% in a heterogeneous phantom for all photon beams. Monaco TPS commissioning was successfully verified on patient plans using dosimetric measurements with overall average gamma passing rates (3%/3 mm criteria) of 93.3% with 6.7% confidence limits.

ACKNOWLEDGMENT
The authors would like to acknowledge the assistance from Patricia Candia, PhD in proof-reading the manuscript and giving valuable feedback.

CONF LICT OF I NTEREST
The authors declare no conflict of interest.