Development of a dedicated phantom for multi‐target single‐isocentre stereotactic radiosurgery end to end testing

Abstract Purpose The aim of this project was to design and manufacture a cost‐effective end‐to‐end (E2E) phantom for quantifying the geometric and dosimetric accuracy of a linear accelerator based, multi‐target single‐isocenter (MTSI) frameless stereotactic radiosurgery (SRS) technique. Method A perspex Multi‐Plug device from a Sun Nuclear ArcCheck phantom (Sun Nuclear, Melbourne, FL) was enhanced to make it more applicable for MTSI SRS E2E testing. The following steps in the SRS chain were then analysed using the phantom: magnetic resonance imaging (MRI) distortion, planning computed tomography (CT) scan and MRI image registration accuracy, phantom setup accuracy using CBCT, dosimetric accuracy using ion chamber, planar film dose measurements and coincidence of linear accelerator mega‐voltage (MV), and kilo‐voltage (kV) isocenters using Winston‐Lutz testing (WLT). Results The dedicated E2E phantom was able to successfully quantify the geometric and dosimetric accuracy of the MTSI SRS technique. MRI distortions were less than 0.5 mm, or half a voxel size. The average MRI‐CT registration accuracy was 0.15 mm (±0.31 mm), 0.20 mm (±0.16 mm), and 0.39 mm (±0.11 mm) in the superior/inferior, left/right and, anterior/posterior directions, respectively. The phantom setup accuracy using CBCT was better than 0.2 mm and 0.1°. Point dose measurements were within 5% of the treatment planning system predicted dose. The comparison of planar film doses to the planning system dose distributions, performed using gamma analysis, resulted in pass rates greater than 97% for 3%/1 mm gamma criteria. Finally, off‐axis WLT showed MV/kV coincidence to be within 1 mm for off‐axis distances up to 60 mm. Conclusion A novel, versatile and cost‐effective phantom for comprehensive E2E testing of MTSI SRS treatments was developed, incorporating multiple detector types and fiducial markers. The phantom is capable of quantifying the accuracy of each step in the MTSI SRS planning and treatment process.


K E Y W O R D S
dosimetric validation, end-to-end testing, MRI deformation, multi-target, single isocentre, stereotactic radiosurgery

| INTRODUCTION
Intracranial metastases are being discovered in approximately 400,000 patients per year worldwide, 1 of which about 70-80% will have multiple intracranial metastases. 2 The role of linear accelerator based stereotactic radiosurgery (SRS) in treatment of these intracranial metastases has expanded significantly in past decades, due to its high delivery efficiency 3 and equivalent plan quality, 4

relative to
Gamma Knife (GK) radiosurgery.
Recent published data indicates that due to the relatively high toxicity and poor neurological outcomes associated with whole brain radiotherapy, SRS is now becoming the standard of care in the treatment of patients with multiple brain metastases. [5][6][7] The role of SRS has recently expanded to include treatment of multiple cranial metastases with a single isocenter, 8,9 hence further reducing treatment times. This technique requires the use of multi-leaf collimators (MLCs) and volumetric modulated arc therapy (VMAT) or dynamic conformal arc therapy (DCAT), as cone collimation can treat only one target at a time.
Another recent change in technique has been the use of onboard imaging, such as cone beam computed tomography (CBCT) facility on the linear accelerator. Rather than relying on a separate kV X-ray imaging system, CBCT is used to position the patient for SRS. This requires much tighter tolerances for the linear accelerator isocenter and CBCT geometric accuracy compared to non-SRS treatments.
To meet the growing demand, there is a need for SRS treatments to be available in most radiation oncology departments rather than remain confined to specialist centers. This creates a challenge for staff in departments who do not have SRS experience or the appropriate equipment, to meet the strict demands on geometric and dosimetric accuracies required for the multi-target single-isocenter (MTSI) SRS technique.
An essential element of setting up and maintaining an SRS program is quantifying the uncertainties inherent in the planning and treatment process. These factors include, but are not limited to, uncertainties due to systematic, and/or random errors in: gross tumour volume (GTV) contouring, 10 geometric distortion of magnetic resonance imaging (MRI) used for contouring of GTVs, 11 the image registration of MRI and CT images used for treatment planning, 12 the measurement of small field data, 13 the treatment planning system (TPS) modeling of small field sizes, 14 differences in the treatment and imaging isocenters of the linear accelerator, 15 patient positioning after CBCT, 15 and intrafraction motion of the patient during treatment delivery. 16 The aim of end to end (E2E) testing is to measure the overall geometric and dosimetric accuracy of the planning and treatment chain. Several previously published studies have successfully achieved this for SRS, 10,[17][18][19] however, none of these quantified the geometric and dosimetric accuracy for MTSI SRS. The characteristics of an ideal E2E phantom for MTSI SRS include: • Compatibility with MRI, CT, CBCT and MV imaging • Dimensions and shape similar to an average human head • Cost effectiveness • External markings for quick and easy setup • Compatibility with patient immobilization equipment • No dose perturbation for non-coplanar fields (i.e., non-zero couch angles) • Minimal air gaps • Ability to position fiducial markers, point detectors, and radiochromic film over a wide range of locations, including those close to the lateral and superior edges of the phantom in order to encompass off-axis targets There are several commercially available devices that are marketed for MTSI SRS quality assurance (QA), such as: the Sun Nuclear SRS MapCheck, Standard Imaging Lucy 3D, Integrated Medical Technologies MAX-HD, and CIRS Steev phantoms. However, these phantoms may not meet all the criteria listed above or may have a cost that is prohibitive to a department in the initial stages of developing an SRS program. The design, manufacture, and use of a low-cost, versatile, MTSI SRS E2E phantom that meets all the criteria listed above, through the enhancement of a Multi-Plug device from a Sun Nuclear ArcCheck phantom (Sun Nuclear, Melbourne, FL) is presented.

2.A | Phantom design
The phantom is a Perspex cylinder of diameter 150 mm and length 255 mm. The central portion of the phantom is made up of Perspex inserts (20 × 20 × 220 mm) to allow for placement of dosimeters and fiducial markers, to aid in the SRS E2E testing process (Fig. 1).
The size and shape of the MTSI phantom without the ArcCheck is ideal for MTSI SRS E2E testing as it closely mimics the size and shape of an average adult head and allows for a thermoplastic mask to be fitted to the phantom to provide a true E2E test incorporating all relevant equipment. Minor modifications were made to the phantom, e.g., removal of the metal handle and threaded inserts to avoid MRI distortion, and rounding of the cylinder end through which the vertex fields enter (note that this rounding does not affect the use of the MultiPlug in the ArcCheck). Additionally, external markings were added to the surface of the phantom to ensure reproducible setup of the phantom within the thermoplastic mask in the time between simulation and treatment. The coordinate system for the phantom was chosen to be the same as that for a supine patient, i.e., superior/inferior (toward/away from the gantry for couch 0°),     The phantom was placed within a QFix Encompass SRS headboard (Type RT-4600-01, QFix, Avondale, PA) and mask system (Type RT-B889KYCF2, QFix, Avondale, PA), the same system to be used for patient localization during the SRS treatment planning and delivery process in our department (Fig. 3). The phantom was then scanned with 1 mm slice thickness on a dedicated radiotherapy Philips Big Bore CT scanner with the vitamin E inserts in the same configuration as the MRI, along with four bone inserts as shown in true patient simulation as closely as possible. The scan was exported to the image registration software platform.

2.B.2 | Registration MRI-CT
The MRI scan was registered to the CT using a manual rigid registration in the MIM (v6.7.10, MIM Software Inc., Cleveland, OH) imaging platform. The registration was performed by a radiation therapist and independently reviewed by a physicist and radiation oncologist.
The image registration was based on alignment of the vitamin E capsules in the images, as shown in Fig. 5 The plans were created with 5 mm width multi-leaf collimators

2.B.5 | Off-axis WLT
Prior to the CBCT match, the ball bearing (BB) insert was placed in the phantom at so that a BB was aligned to the target location as defined by the treatment plan described in Section 2.B.3. A multileaf collimator (MLC) defined 2 × 2 cm field size was created with the collimator jaws set to 3 × 3 cm, centerd on the target in the TPS. After the CBCT match, the MLC field was delivered and image captured on an electronic portal imaging device (EPID) for each combination of gantry, collimator, and couch angles outlined in Table 2.
Planar doses were measured through the center of each target and compared to those predicted by the TPS, using gamma analysis. 20 A tolerance of 3% global dose difference and 1 mm distance to agreement was employed (excluding dose points below a 10% threshold) with a minimum pass rate of 95% considered acceptable.
EBT3 films were scanned in an Epson 10000XL flatbed scanner (SEIKO Epson Co, Japan). The films were positioned on the scanner with a jig to ensure reproducible positioning of the film within the scanning area, and placed under a sheet of glass during scanning to maintain good contact between the film and the scanner as recommended by Ashland. 21 The films were scanned 24 h after exposure to allow stabilization of the latent image. 22 Each film was scanned in transmission mode using 48 bit RGB with a scanner resolution of 75 dpi (0.34 mm pixel size). All phantom and calibration film pieces were marked in order to maintain the same orientation during scanning, thereby eliminating polarization effects, 23 and positioned at the center of the flatbed scanner with the long axis of the film pieces parallel to the long axis of the scanner to avoid off-axis scanner non-uniformity. 24 The pixel values measured from the red channel of the scanner were used to calculate the optical density (OD) for each film piece and these were converted to absorbed dose (in cGy) using a OD-to-dose calibration curve.

3.A | MRI distortion
The

3.C. | Phantom setup and CBCT alignment
Phantom setup and CBCT alignment results are summarized in Tables 3 and 4, for translations and rotations, respectively. As can be seen from the tables, after CBCT based correction using a 6DOF couch, the residual translation and rotation errors after repeat CBCT are 0.2 mm and 0.1°, respectively. Furthermore, the relatively minor initial corrections indicate that a reproducible setup of the phantom between simulation and treatment.

3.D | Off-axis WLT
The off-axis WLT results are summarized in Fig. 6 and Table 5, where the average displacement from the center of the BB to the T A B L E 2 Combination of gantry, collimator, and couch fields used for Winston-Lutz testing in this study. 3.E | Dosimetric verification

3.E.1 | Point dose measurements
Measured point doses were systematically lower than those predicted by the TPS. For each of the 20 targets from the 10 SRS plans in the study, the average (±1 SD) difference between measured and TPS predicted dose was −1.9% ±1.4%. All of the targets fell within the 5% tolerance set prior to measurement, with the minimum difference measured as −0.1% and the maximum as −4.9%.
There was no trend observed with the point dose results when comparing to target size or number of targets in the plan. However, as seen in Table 6, the difference between the measured and TPS predicted point doses was observed to get larger with increasing distance of the target away from the isocenter of the plan.

3.E.2 | Planar dose measurements
For each of the 20 targets from the 10 SRS plans in the study, the average (±1 SD) gamma pass rate for the planar dose verifications was 97.9% ± 1.1%. Each of the targets had a gamma pass rate greater than the 95% tolerance for the 3%/1 mm gamma criteria.
The maximum pass rate was 99.4% and the minimum was 95.5%.   29 performed an off-axis WLT using the same methodology as our study, on a Varian TrueBeam STx linear accelerator up to an off-axis distance up to 4 cm. The results were similar to those reported here, with maximum off-center errors of less than 1 mm.
As described by Gao et al., 28 the increasing deviation between the center of the mechanical and radiation fields with increasing off-axis distance may be due to the increasing effect of small rotational errors with increasing off-axis distance. Furthermore, at off-axis positions, the penumbra of the radiation field becomes asymmetric and the center of the radiation field becomes less well defined, an effect that is magnified with increasing off-axis distance.
Dosimetric accuracy of the treatment plans in this study was confirmed using both ionization chamber and Gafchromic film mea- One limitation of this study is that the contouring uncertainty was not considered in the E2E analysis. Contouring uncertainty should be considered in true E2E validation of a treatment technique as it may contribute significantly toward the total combined geometric accuracy of the technique. Contouring uncertainty of brain metastases has however, been studied previously, with a geometrical uncertainty of approximately 0.3 mm 10 described.
Another limitation of the study is that intra-fraction motion was also not considered in the analysis. A study by Wen et al. 32  Finally, only up to 4 target volumes were considered in any one plan. This limitation is one that has been implemented locally in our department, based on published randomized clinical trial data. [5][6][7] However, the authors are aware that it is common for departments to routinely treat more than 4 targets using MTSI plans and we believe that the technique presented in this paper can be easily extended to provide E2E testing for MTSI plans with more than 4 targets. Extension of the E2E technique to MTSI plans with more than 4 targets will be investigated in future studies.

| CONCLUSION
With the growing role of linear accelerators in treating brain metastases patients with SRS, the number of radiation oncology departments around the world providing an SRS service is likely to increase significantly. We have developed a novel and cost-effective solution for comprehensive E2E testing of these treatments that can incorporate multiple detector types and is easily adaptable to the specific workflows of these departments. The E2E test revealed that targets, even when located in non-isocentric positions, can consistently be located to within 1 mm using CBCT.

CONFLI CT OF INTEREST
The authors have no conflicts of interest to report.