Commissioning cranial single‐isocenter multi‐target radiosurgery for the Versa HD

Abstract Purpose Brainlab’s Elements Multiple Brain Mets SRS (MBMS) is a dedicated treatment planning system for single‐isocenter multi‐target (SIMT) cranial stereotactic radiosurgery (SRS) treatments. The purpose of this study is to present the commissioning experience of MBMS on an Elekta Versa HD. Methods MBMS was commissioned for 6 X, 6 FFF, and 10 FFF. Beam data collected included: output factors, percent depth doses (PDDs), diagonal profiles, collimator transmission, and penumbra. Beam data were processed by Brainlab and resulting parameters were entered into the planning system to generate the beam model. Beam model accuracy was verified for simple fields. MBMS plans were created on previously treated cranial SRS patient data sets. Plans were evaluated using Paddick inverse conformity (ICI), gradient indices (GI), and cumulative volume of brain receiving 12 Gy. Dosimetric accuracy of the MBMS plans was verified using microDiamond, Gafchromic film, and SRS Mapcheck measurements of absolute dose and dose profiles for individual targets. Finally, an end‐to‐end (E2E) test was performed with a MR‐CT compatible phantom to validate the accuracy of the simulation‐to‐delivery process. Results For square fields, calculated scatter factors were within 1.0% of measured, PDDs were within 0.5% past dmax, and diagonal profiles were within 0.5% for clinically relevant off‐axis distances (<10 cm). MBMS produced plans with ICIs < 1.5 and GIs < 5.0 for targets > 10 mm. Average point doses of the MBMS plans, measured by microDiamond, were within 0.31% of calculated (max 2.84%). Average per‐field planar pass rates were 98.0% (95.5% minimum) using a 2%/1 mm/10% threshold relative gamma analysis. E2E point dose measurements were within 1.5% of calculated and Gafchromic film pass rates were 99.6% using a 5%/1 mm/10% threshold gamma analysis. Conclusion The experience presented can be used to aid the commissioning of the Versa HD in the Brainlab MBMS treatment planning system, to produce safe and accurate SIMT cranial SRS treatments.


| INTRODUCTION
Elements Multiple Brain Mets SRS (MBMS) is a site-specific planning system for treating multiple cranial targets that was developed by Brainlab (Brainlab, Munich, Germany). Unlike conventional planning systems that are designed to treat a wide range of anatomical sites, MBMS creates single-isocenter multi-target (SIMT) linac based cranial stereotactic radiosurgery (SRS) plans, using non-coplanar dynamic conformal arcs. SIMT has the potential to create plans with similar organ at risk (OAR) sparing and target coverage, while reducing treatment times compared with multiple single-target plans. [1][2][3][4] The specificity of the MBMS allows for an optimization algorithm that can focus on important cranial SRS metrics. The optimizer can also overcome typical linac based cranial SRS planning shortfalls, like the bridging of dose between two targets.
One drawback of the specificity is that the commissioning physicist is unable to perform an AAPM Task Group 119 type test of the system to compare their commissioning results for various anatomical sites to other institutions. 5 Furthermore, MBMS may be commissioned at the start of an institution's linac-based cranial SRS implementation so there may not be any internal data for comparison. In this work, the MBMS commissioning experience on an Elekta Versa HD (Elekta AB, Stockholm, Sweden) will be presented which can be used for guidance as well as a baseline for comparison.

2.A | Generating beam model
Beam model measurements included: PDDs, profiles, scatter factors, collimator transmission, and dynamic leaf shift. Data collection followed Task Group 106 methodology. 6 All measurements were made in Sun Nuclear's (Sun Nuclear Corporation, Melbourne, FL) 3D water tank with Sun Nuclear's 0.125 cc chamber, EDGE detector, or PTW's (PTW-Freiburg, Freiburg, Germany) microDiamond detector. A SNC 0.125 cc chamber was used as a reference chamber to normalize the data for fluctuations in linac output when scanning profiles and PDDs.
The water tank was setup to the central axis of the beam using a ray tracing procedure. For profile measurements, the tank was shifted 0.25 cm (1/2 leaf width) in the jaw direction, so the detectors intersected a multi-leaf collimator (MLC) leaf tip instead of junction between two MLC leaves. For output measurements, a Daisy-Chain method was used to calibrate the microDiamond for small field measurements. 7 The SNC 0.125 cc was used to measure output factors down to a 3.0 cm field size, after which it was cross calibrated to the microDiamond chamber to measure output factors down to 1.0 cm. microDiamond measurement were performed without corrections, which is examined in the discussion.
Beam data were collected for energies 6 X, 6 FFF, and 10 FFF.
The measured data were processed by Brainlab to calculate leaf shifts, tongue and groove sizes, source functions, and radial factors.
After beam model parameters were measured and calculated, machine models were created for each energy. Machine models require department specific parameters (machine name, coordinate convention, etc.) along with machine specific information such as dose rate and maximum gantry speed. The machine type parameters were collated from three sources: a) Versa HD manuals, b) Monaco manuals (provided by Elekta), and c) settings in existing hospital planning systems (Pinnacle). Following the creation of the machine model, the energy specific beam models were created for the three energies. Prior to final saving of the model, the system performed a secondary check for the data to help protect against non-realistic values.

2.B | Validation
AAPM Task Group 53 was used to guide the treatment planning system (TPS) validation. 8 The data transfer from Elements to the record and verify system (Mosaiq) was tested using various test plans. Subsequent data transfer to the linac and on-board imagers was tested.
Data fidelity was checked at each step of the transfer.
Initial validation of the MBMS version 1.5 beam model was done using the Beam Model Verification module, included in the Elements software, which allows the calculation of single fields on phantoms.
Dose was calculated with a Pencil Beam Algorithm utilizing a 1 mm grid size. A virtual water phantom with density 1.0 g/cm 3 , simulating a water tank, was generated in Matlab and imported into Elements.
The point dose, output factors, depth dose, and profiles were calculated using the same geometry as the commissioning measurements and verified against measured data.
Following beam model verification, the validation of typical clinical deliveries was performed. Since Elements is a template-based software, various prescription and beam arrangement template protocols were generated to cover the range of expected clinical cases.
Prescription protocols of 19 Gy × 1 fx, 8 Gy × 3 fx, and 6 Gy × 5 fx were created with various minimum target coverages of 95%, 97%, and 99% for a total of nine protocols. Beam templates for 2, 3, and 5 different couch angles were created. Two version of each protocols were created: a) all the couch angles on one side of the gantry, b) couch angles on both sides with symmetrical arrangement. MBMS automatically mirrors one-sided protocols if the target is on the other side of the brain, therefore the generation of both left and right sided protocols was not needed.
To test the protocols, previously treated cranial SRS patients treated within the hospital system, were re-planned in MBMS. All initial plans were 1 fraction treatments with the same prescriptions ranging from 15 Gy to 20 Gy covering 95% of the target. Fourteen clinical targets were studied ranging from 0.27 cc to 7.32 cc. Plans had between 2-5 couch angles and 1-5 targets. Each plan was recalculated for three energies: 6 X, 6 FFF, and 10 FFF. While it was known that 10FFF would not be used for cranial treatments, the energy was commissioned in anticipation of different anatomical Elements (ex. Spine). Regardless, it is recommended that at least two energies be commissioned simultaneously to allow cross-comparison KNILL ET AL.

| 109
between results, which can be helpful with troubleshooting any inconsistencies that arise during validation.
Plan quality was evaluated using Paddick inverse conformity index (ICI), gradient index (GI), and Brain V 12 Gy (volume of normal brain getting dose of 12 Gy or more) to get an understanding of the limits of the system. 9 The rotation matrix was included for measurements that were To apply shifts during measurements, the phantom was first setup to the lasers and calculated shifts (R y θ 0 ð Þ I t À I p

3.A | Beam modelscatter factors
Comparisons between measured and calculated scatter factors measured at 100 cm source-to-phantom distance and 10 cm depth were performed (Fig. 1). Square fields were within 1% of Elements calculated values. The absolute measured scatter factors are shown in

3.B | Beam model -PDDs & profiles
The difference between the measured and calculated PDDS and profiles for select 6 FFF square fields and depths were compared ( Fig. 2 and Fig. 3

3.C | Beam modelcollimator penumbra
The measured in-plane (jaw) penumbra was larger than the crossplane (MLC) penumbra, which is similar to previous publications. 11 The difference between the measured and calculated penumbras is shown in Fig. 4. All profiles were normalized to the central axis for comparison. The differences in the jaw direction were typically < 2.0%, while the differences in the MLC direction were < 10.0%. The penumbra differences were similar for different field sizes and depths. A single penumbra model is used for both the MLC and Jaws, which resulted in a larger difference in the MLC direction.

3.D | Beam model -MLC DLS and transmission
The measured MLC transmissions and dynamic leaf shifts are shown in Table 2. The 10 FFF transmission was slightly lower than 6 X and 6 FFF, however, this is offset by the larger dynamic leaf shift causing increased transmission near the field edges. The transmission with the jaws closed (jaws and MLC combined) was zero percent for all energies.

3.E | MBMS plan quality
The MBMS plans had GIs smaller than 5.0 and ICIs smaller than 1.5 for target diameters larger than 10 mm. For targets smaller than  and Gafchromic film pass rates of 98.6% and 99.6% using a 10% threshold and 3%/1 mm and 5%/1 mm gamma criteria respectively. Figure 6 shows the Gafchromic film results in the axial plane for the 3 target E2E plan. Most of the remaining failing points with the 5%/ 1 mm criterion were due to the pin-holes in the Gafchromic film, which were used for registration.
When a plan was generated for only a single target, the microDiamond measured dose began to increase > 3% for targets below 10mm, possibly due to larger effect of guard leaves in smaller targets. However, this dose discrepancy was not observed when the smaller target was included in a plan with other targets. Attempts were made to manually adjust beam model scatter factors to better model single target dose, however, it was found that improvements in single-target dose modeling would lead to larger errors in multitarget plans. Due to the multi-target purpose of MBMS, the decision was made to prioritize multi-target dose modeling over single-target.

3.G | MBMS version 2.0
The In this work, small field correction factors were not applied to the measurements. Based on TRS-483, the microDiamond will overrespond by approximately 1.5% for a 1.0 cm field size. Additional publications have suggested this over-response may be up to 3.4% for a 6FFF beam on the VersaHD with a 1.0 cm field size. 17 This over-response will lead to an increase in the measured small field scatter factors and thereby a reduction in the delivered dose. This matches the E2E film results which were found to be within 1.5% lower than predicted for the 1.0cm target (Fig. 6). Given the expected clinical prescriptions for the small targets, this lower dose was deemed to still be ablative to the target, while the clinical organs-at-risk dose would be within tolerance.  As targets get farther from the central axis, the arclength error produced by a rotational error will increase sinusoidally. For small angles, the linear error will equal the radius times the angle in radians.
AAPM Task Group 142 recommends a 1°collimator tolerance. 19 A 1°c ollimator error would create a 0.87 mm linear error for a target 5cm away from the central axis. To reduce this error, a stricter 0.5°collimator tolerance was adopted, which was found to be consistently achievable on monthly QA. Furthermore, TG-142 requires a 0.5°c ouch tolerance for SRS/SBRT, however, the clinical display only shows integers. This was overcome by enabling the "PSS" page in service mode, which reports angles in 0.1°increments. Studies show that target coverage degrades substantially when rotational errors approach 2°. 20 Therefore, minimization of both mechanical and patient setup errors is critical, of which this can be partially accomplished by real time image guidance at each couch angle and positioning the patient with a six degree-of-freedom robotic couch. 21 When creating beam and prescription protocols, it was best to create one prescription and beam protocol and fully test the planning and delivery. Once fully tested, the protocols could then be copied and modified as needed. This would help prevent unnecessary time fixing issues that may propagate through all the protocols if they are generated prior to testing.

| CONCLUSION
The commissioning of the MBMS TPS system introduces unique challenges for physicists due in part to the small fields, off-axis non-coplanar beam arrangements, and high-dose hypofractionated prescriptions.
Advanced knowledge of these challenges along with the expected limitations of the MBMS beam models can add familiarity to the commissioning process. Added familiarity will hopefully lead to faster and more consistent MBMS commissioning across institutions.