Automated and robust beam data validation of a preconfigured ring gantry linear accelerator using a 1D tank with synchronized beam delivery and couch motions

Abstract Purpose To develop an efficient and automated methodology for beam data validation for a preconfigured ring gantry linear accelerator using scripting and a one‐dimensional (1D) tank with automated couch motions. Materials and methods Using an application programming interface, a program was developed to allow the user to choose a set of beam data to validate with measurement. Once selected the program generates a set of instructions for radiation delivery with synchronized couch motions for the linear accelerator in the form of an extensible markup language (XML) file to be delivered on the ring gantry linear accelerator. The user then delivers these beams while measuring with the 1D tank and data logging electrometer. The program also automatically calculates this set of beams on the measurement geometry within the treatment planning system (TPS) and extracts the corresponding calculated dosimetric data for comparison to measurement. Once completed the program then returns a comparison of the measurement to the predicted result from the TPS to the user and prints a report. In this work lateral, longitudinal, and diagonal profiles were taken for fields sizes of 6 × 6, 8 × 8, 10 × 10, 20 × 20, and 28 × 28 cm2 at depths of 1.3, 5, 10, 20, and 30 cm. Depth dose profiles were taken for all field sizes. Results Using this methodology, the TPS was validated to agree with measurement. All compared points yielded a gamma value less than 1 for a 1.5%/1.5 mm criteria (100% passing rate). Off axis profiles had >98.5% of data points producing a gamma value <1 with a 1%/1 mm criteria. All depth profiles produced 100% of data points with a gamma value <1 with a 1%/1 mm criteria. All data points measured were within 1.5% or 2 mm distance to agreement. Conclusions This methodology allows for an increase in automation in the beam data validation process. Leveraging the application program interface allows the user to use a single system to create the measurement files, predict the result, and then compare to actual measurement increasing efficiency and reducing the chance for user input errors.

compare to actual measurement increasing efficiency and reducing the chance for user input errors. but must be validated by the user. General guidelines to the overall treatment accuracy achievable within radiation oncology (~5%) have been published in the literature. 1 The general quality assurance guidelines for radiation oncology have been well described in the literature. [2][3][4][5] In particular, the quality assurance needed for validation of the TPS is well described. [6][7][8][9][10] Recent publications have applied the methodologies to the Halcyon platform. [11][12][13] Others have looked at validating the automated daily quality assurance (QA) which is currently available for Halcyon and TrueBeam platforms. 14 This work looks to extend this automated approach to the beam data validation that is required upon initial installation of the Halcyon. Treatment planning systems validation is an important and critical component of the commissioning process in radiation therapy.
The accuracy of the dose calculation in a water phantom form a reference QA test set of fields should be within 2% dose difference or 2 mm distance to agreement according to the published quality assurance task group report 40 4 and within 2% in the high-dose area and within 3 mm of the penumbra according to published practice guide lines. 8 The importance of these guidelines is highlighted by the fact that there is variability in the accuracy of commissioning as reported by third party auditing done in the United States. [15][16][17][18][19] Of interest, as reported by Molineu et al, the incidents of failing to meet the third party audit were reduced in preconfigured TPS and delivery machines combinations. 17 The fact that a preconfigured TPS and linear accelerator combination is attractive due to the reduction in user inputs does not mean the user does not have to complete TPS validation.
Currently, TPS validation is generally done after inputting commissioning data from a three-dimensional (3D) scanning tank. For the Halcyon, the TPS model comes preconfigured and does not take any user-defined TPS data input. The physicist must validate this preconfigured beam data. Comparison to 3D tank water scans are the gold standard for this validation, however, alternative methods have been suggested in the literature. [20][21][22][23][24] Of concern with the Halcyon is the bore size (100 cm diameter), which limits clearance when using a traditional 3D tank size (~70 × 70 × 56 cm). The potential sag associated with supporting the weight of a full 3D tank (~270 kg) on the couch could also make leveling difficult. Furthermore, the associated cost of the 3D tank systems can be burdensome in a resource limited environment.
Using a one-dimensional (1D) tank with synchronized couch motions with beam delivery could allow the user to validate the TPS in a manner similar to traditional 3D tank methods, however, with less equipment and storage space needed. The use of the extensible markup language (XML), allows the user to program couch motions as a function of monitor units delivered. 25 Cross-beam profiles can be reconstructed by moving the couch along the beam while recording the current in field and reference ionization chambers. In the TPS, plans with multiple isocenter locations along the axis of the profile desired by the user are calculated to simulate the tank movement relative to isocenter during the measurement. Doses to central voxels at different depths for each plan represent different detector positions would be analogous to extracting profiles at different depths, similar to conventional 3D tank data. Leveraging scripting, this can be done programmatically such that the user is defining what to validate, and then the program not only extracts the TPS data to be validated, but also writes the XML file that will be used to measure the data. The measured data can then be fed back into the program for analysis allowing the user a feedback system to determine if the beam data validation is complete or needs further review. This feedback loop connecting the measurement on the machine to the TPS calculation could provide an efficient and robust beam data validation methodology not seen in previous works.  the machine, this data is provided as feedback into the software for comparison. The comparison between the measured result and the expected from the TPS was completed both graphically and using a 1D gamma analysis 26 with user configurable tolerances for dose difference and distance to agreement. The percentage of points within a set of data with a gamma value <1 was then reported for the user.

2.A | A robust workflow with feedback
At which point if the user finds the results acceptable, they can export a report or choose to add additional validation tests. Figure 1 shows a flowchart of this workflow with the beam validation tool directly linking the TPS outputs to the measurements on the treatment machine via the application programing interface.

2.B | Measurement
For this work, an example subset of data was selected for beam validation for the Halcyon. These measurements were completed on the Halcyon by selecting the generated XML and recording the output of a data logging electrometer. This data was readout by an inhouse software and displayed during the measurement in real-time via a graphical user interface (GUI). The field sizes selected were 6 × 6, 8 × 8, 10 × 10, 20 × 20, and 28 × 28 cm 2 . X and Y profiles were taken at 1.3, 5, 10, 20, and 30 cm depths. Diagonal profiles were completed at the same depths for 28 × 28 cm 2 field size. For alignment, the tank arm was leveled with a spirit level, the tank was initially aligned using the lasers outside the bore to set the detector at laser isocenter and the source to surface distance of the water.
Once the initial setup was completed the treatment couch was used to position the tank at the treatment isocenter. The reference detector was taped into place on the bore in a position away from the central axis position of the source. All cabling was routed perpendicular to the scan path to try to minimize the amount of cable in the beam. Once setup was complete, it was then confirmed with MV imaging. An AP MV image taken to set the couch lateral and longitudinal positions such that the field detector was at radiation isocenter. The reference detector was confirmed to be in an appropriate position as not to obscure the field detector during the scan [ Fig. 2 (a)]. The source to surface distance (SSD) was set using MV imaging through the side of the tank and measuring from isocenter to the water surface [ Fig. 2(b)]. To ensure a crisp image of the water surface, the imaging was completed at an angle equal to 90°minus the angle between the source and water surface for a given depth of the isocenter. The angle between the water surface and the source was calculated as the arctangent of the depth of the isocenter over All measurements were completed using a 0.13 cc field scanning chamber with a 0. 13  Legend F I G . 1. Flowchart for the beam validation process. Each step taken is color coded indicating what system is being used. The small colored squares indicate a connection point to the treatment planning system (TPS), the application programing interface (API), or the treatment machine within a given step. The TPS, beam validation tool, and treatment machine are connected through two separate feedback loops allowing the user to design a custom validation data set that leverages the API to connect the TPS and treatment machine.

2.D | Beam data validation comparisons
Once given the user input measurement data, the 1D gamma profiles were calculated for all pairs of depth and off axis profiles (measured vs TPS expected). The dose difference and distance to agreement criteria were varied to investigate the agreement between the data sets, that is, the percentage of points with a gamma value < 1. The only post processing used on the data was to center the lateral profiles,normalize to the average of the three center dose values for the lateral profiles, and to normalize to the maximum dose values for the depth profiles. The dose differences and distance to agreement distributions were also calculated individually as well. The results were then output for user review and approval.

3.B | Off axis profiles
All comparisons between measured profiles and TPS simulated pro- penumbra of each field. However, in general, the agreement was excellent for all profiles measured, with all data being within 1.5% dose difference or 2 mm distance to agreement.

| DISCUSSION
In this work, the preconfigured All the analysis was generated as the scanning process was completed and a summary was generated during the take down of the equipment. Table 1

| CONCLUSION
This work has shown a robust and automated methodology for beam data validation for a preconfigured ring gantry linear accelerator. The TPS predicted dose shows excellent agreement with measurement with all data points being within 1.5% dose difference or 2 mm distance to agreement.