Assessing the feasibility of single target radiosurgery quality assurance with portal dosimetry

Abstract Purpose To assess the feasibility of using portal dosimetry (PD) for pre‐treatment quality assurance of single target, flattening filter free (FFF), volumetric arc therapy intracranial radiosurgery plans. Methods A PD algorithm was created for a 10X FFF beam on a Varian Edge linear accelerator (Varian Inc, Palo Alto, CA, USA). Treatment plans that were previously evaluated with Gafchromic EBT‐XD (Ashland, Bridgewater, NJ, USA) film were measured via PD and analyzed with the ARIA Portal Dosimetry workspace. Absolute dose evaluation for film and PD was done by computing the mean dose in the region receiving greater than or equal to 90% of the max dose and comparing to the mean dose in the same region calculated by the treatment planning system (TPS). Gamma analysis with 10% threshold and 3%/2 mm passing criteria was performed on film and portal images. Results Thirty‐six PD verification plans were delivered and analyzed. The average PD to TPS dose was 0.989 ± 0.01 while film to TPS dose was 1.026 ± 0.01. All PD plans passed the gamma analysis with 100% of points having gamma <1. Overall, PD to TPS dose agreement was found to be target size dependent. As target size decreases, PD to TPS dose ratio decreased from 1.004 for targets with diameters between 15–31 mm and 0.978 for targets with diameters less than 15 mm. Conclusion The agreement of PD to TPS mean dose in the high dose region was found to be dependent on target size. Film measurements did not exhibit size dependence. All PD plans passed the 3%/2 mm gamma analysis, but caution should be used when using PD to assess overall dosimetric accuracy of the treatment plan for small targets.

appropriate QA device is complicated by the detector size vs field size and loss of lateral charged particle equilibrium. 3 Recommendation are to use an appropriate dosimeter with a resolution of approximately 1 mm of better. 5 Portal dosimetry is commonly used in the pre-treatment QA for intensity-modulated plans [6][7][8][9][10][11] by comparing the delivered fluence on the portal imager to that predicted by the treatment planning system (TPS). This is done on the Varian (Varian Medical Systems, Palo Alto, CA, USA) platform through the creation of a Portal Dosimetry Image Prediction (PDIP) algorithm that enables users to create a predicted portal dose image for fields with dynamic multileaf collimators (MLCs) and compare the acquired image in the Portal Dosimetry workspace. While the technique has been used often for traditional treatments, its use in small field SRS treatments has been less explored in the literature. 12 Because of the greater potential harm with these treatments, care should be taken to validate the dosimetric performance of the device for appropriate field shapes, sizes, total doses, and dose rates.
This study focuses on assessing the feasibility of using portal dosimetry for pre-treatment QA of 10

| ME TH ODS
Clinical treatment plans were chosen that had been previously evaluated with Gafchromic EBT-XD (Ashland, Bridgewater, NJ, USA) film. The dose distribution was measured in the coronal plane using radiochromic film (EBT-XD, Ashland Chemical, Covington, KY, USA).
A calibration curve was obtained at each measurement session. The film was scanned using an Epson model V700 PhotoPerfection document scanner (Epson America, Long Beach, CA, USA) and converted to dose using a three-channel technique. 13 Film analysis was done using in house software developed in MATLAB (MathWorks, Natick, MA, USA). Gamma analysis was done with a 10% threshold and 3%/2 mm criteria. 14 Absolute dose evaluation was done by computing the mean dose in the region receiving greater than or equal to 90% of the max dose and comparing to the mean dose in the same region calculated by the TPS. The mean dose in this region-of-interest was used rather than a point dose to reduce the effects of pixel-to-pixel noise in the film. 15 Plans were chosen to cover a range of target sizes from 3 to 31 mm equivalent diameter, which is the diameter of a sphere with the same volume as the target. All analysis was performed on the composite dose of the two arcs. The median prescription dose per fraction was 15 Gy, with a range of 5-20 Gy.
Portal dosimetry was commissioned on an Edge linear accelerator 10X FFF beam with HDMLCs. Varian does not offer a preconfigured package for machines with HDMLCs and/or FFF beams. In order to create a predicted portal image, a PDIP algorithm must be created by the user. This was done by following the instructions provided by Varian in the Eclipse Photon and Electron Algorithms Reference Guide.
Once the algorithm was created, it was verified using the verification plans provided in the preconfigured PDIP package for flattened beams with Millennium MLCs. These plans were imported, the MLC files were updated for HDMLCs, the energy was changed to 10X FFF, and the plans were recalculated.
Plans were delivered on an Edge linear accelerator with a DMI portal imager at 100 SDD and evaluated in the Portal Dosimetry workspace. All plans were delivered with the planned dose rate of 2400 MU/min. Since the portal imager has a fixed geometry with the gantry, couch rotations are not included in the portal dosimetry delivery, hence the delivery is perpendicular composite delivery rather than a true composite delivery like film. 14 Like the film measurements, the analysis was performed on the composite dose of the two arcs. Gamma analysis was done in absolute mode with 10% threshold and 3%/ 2 mm passing criteria. The Portal Dosimetry workspace tools do not allow dose thresholding above 80% or dose comparison of a specified region, so absolute dose analysis was done through the Portal Dosimetry scripting tool. A script was developed to recreate the process used for film analysis. Absolute dose evaluation was done by computing the mean dose in the region receiving greater than or equal to 90% of the max dose in the portal image and comparing to the mean dose in the same region in the predicted portal image.
(a) Color wash shows 10% dose and greater while red outline shows the 10% dose threshold set by portal dosimetry workspace which clearly does not match due to the exclusion of high signal pixels. (b) Threshold was manually adjusted to approximately 80% to match the 10% isodose line to desired analyzed region.
The gamma analysis of portal images was initially done with a 10% dose threshold until an issue was observed where the region of interest outline did not match the 10% dose color wash, as shown in Fig. 1(a). The red outline in Fig. 1(a) shows the 10% threshold on a 3 mm target that is significantly larger than the 10% dose color wash. Once the analysis threshold was set to 80% in Fig. 1(b), the outlines match. The Portal Dosimetry workspace thresholds are limited by a hard coded gamma histogram cut-off that excludes high signal pixels in order to prevent them from being used as the global gamma reference value. In order to perform the gamma analysis with a true 10% threshold, the 10% isodose line was displayed on the composite dose. The threshold was manually adjusted until the dose distribution matched the desired isodose line. After setting the desired threshold, a 3%/2 mm gamma analysis was performed. The results of the two gamma analyses were compared.

| RESULTS
Commissioning of portal dosimetry showed no field size dependence, as shown in Table 1. Two dynamic plans, Aria and DynChair, that were included in the preconfigured Truebeam package were also measured, and had gamma passing rates of 100% and 99.7% with a 10% dose threshold and 3%/2 mm criteria.   Figure 3(a) shows the measured to TPS or PDIP dose as a function of target size. Figure 3(b) shows the film/TPS to PD/PDIP ratio as a function of target size.
Open fields of size 3 × 3-30 × 20 cm 2 were delivered at 400, 1200, and 2400 MU/min to assess any dose rate dependence of the DMI imager. As shown in Fig. 4 While decreasing the dose rate may improve the PD analysis, it is not recommended since ASTRO guidelines 20 state that "same files to be used for the patient delivery should be used for the QA measurements." agreed within 3% using gamma pass criteria of 95% for 3%/2 mm and a 10% threshold. They noted that some clinical plans required a dose threshold of 25% to meet their gamma pass criteria.
Rather than adjust our beam model, we used our film measurements as a benchmark to evaluate portal dosimetry. Note that that all of our PD plans passed using the gamma criteria. Only the comparison of dose (or CU) in the high dose region revealed shortcomings in the portal dosimetry workflow. We also chose to evaluate portal dosimetry using single target plans to avoid convolving issues due to multiple targets and blurring of the dose distribution due to the fixed geometry of the gantry with respect to the imager.
All departmental clinical decision on the acceptability of SRS treatment plans are based on the absolute dose difference between the calculated and delivered plan rather than the results of gamma analysis. While gamma analysis provides information about overall dosimetric and positional accuracy, this information cannot be separated to provide the physicians with an overall description of how "hot" or "cold" the plan delivered, which is the leading factor for replanning. When relying on gamma analysis for plan evaluation, picking the appropriate gamma criteria is an ongoing challenge in our field and shortcomings of its use has been previously published. 21,22 While scripting allowed us to overcome the shortcomings of the Portal Dosimetry analysis tools, the overall offset between the film and portal dosimetry analysis, especially with respect to small target sizes, provides challenges for using Portal Dosimetry as the sole tool for the analysis of stereotactic treatment plans.
While portal dosimetry provides a quick method for checking the overall integrity of MLC motions and gross error, it is not capable of providing an absolute dose comparison. The plans are presented in Calibration Units (CU), which are determined by a calibration procedure of exposing the imager to a 10 × 10 cm 2 and entering the delivered dose. This is used to do an "absolute dose" comparison within the portal dosimetry workspace which is used within the gamma analysis. The current version of the software is unable to isolate the dose comparison from the gamma analysis. The hard coded histogram cutoff also provides challenges for analysis. Currently, the histogram cutoff can be set up to 5% to remove the highest signal peaks to prevent the inclusion of defective pixels. This setting is applied to the entire 1280 × 1280 image and not within the pixels where the dose is contained. This affects the thresholding of the image for the gamma analysis and leads to the inconsistencies seen in Fig. 1 which is more pronounced for smaller targets. We caution users to test the accuracy of the set thresholds in portal dosimetry before using the analysis to make clinical decisions.
Portal imagers can also undergo radiation damage that affects the accuracy of measurements. Ritter et al. 23 reported a technique to measure a dosimetric leaf gap surrogate, called the leaf offset constancy (LOC) using an EPID. In this study, irregularities were seen in LOC measurements that were attributed to radiation damage after repeated exposure to an FFF beam. The damage was not mitigated by dark and flood field calibrations. Due to this, our observed dose rate dependencies, and inconsistencies between our film and EPID measurements, caution should be used when using portal images to evaluating treatment plans with small targets. Furthermore, portal dosimetry should not be used for the commissioning and validating of stereotactic beam models.

| CONCLUSION
Portal dosimetry measurements were found to be target size dependent and could deviate up to 8% from film measurements for the smallest targets evaluated. While portal dosimetry provides a quick method to evaluate SRS plans for gross error without the use of a specialized phantom, it does not provide an accurate method for determining the dosimetric accuracy of the plan when compared to film. Due to the fixed geometry of the portal imager with the gantry, it is also unable to evaluate the accuracy of the true dose distribution due to the inability to deliver couch rotations. Overall, caution should be used when using PD to assess overall dosimetric accuracy of stereotactic treatment plans and commissioning stereotactic beam models.

CONFLI CT OF INTEREST
The authors have no conflict to disclose.