Characteristics of a novel treatment system for linear accelerator–based stereotactic radiosurgery

The purpose of this study is to characterize the dosimetric properties and accuracy of a novel treatment platform (Edge radiosurgery system) for localizing and treating patients with frameless, image‐guided stereotactic radiosurgery (SRS) and stereotactic body radiotherapy (SBRT). Initial measurements of various components of the system, such as a comprehensive assessment of the dosimetric properties of the flattening filter‐free (FFF) beams for both high definition (HD120) MLC and conical cone‐based treatment, positioning accuracy and beam attenuation of a six degree of freedom (6DoF) couch, treatment head leakage test, and integrated end‐to‐end accuracy tests, have been performed. The end‐to‐end test of the system was performed by CT imaging a phantom and registering hidden targets on the treatment couch to determine the localization accuracy of the optical surface monitoring system (OSMS), cone‐beam CT (CBCT), and MV imaging systems, as well as the radiation isocenter targeting accuracy. The deviations between the percent depth‐dose curves acquired on the new linac‐based system (Edge), and the previously published machine with FFF beams (TrueBeam) beyond Dmax were within 1.0% for both energies. The maximum deviation of output factors between the Edge and TrueBeam was 1.6%. The optimized dosimetric leaf gap values, which were fitted using Eclipse dose calculations and measurements based on representative spine radiosurgery plans, were 0.700 mm and 1.000 mm, respectively. For the conical cones, 6X FFF has sharper penumbra ranging from 1.2−1.8 mm (80%‐20%) and 1.9−3.8 mm (90%‐10%) relative to 10X FFF, which has 1.2−2.2 mm and 2.3−5.1 mm, respectively. The relative attenuation measurements of the couch for PA, PA (rails‐in), oblique, oblique (rails‐out), oblique (rails‐in) were: −2.0%, −2.5%, −15.6%, −2.5%, −5.0% for 6X FFF and −1.4%, −1.5%, −12.2%, −2.5%, −5.0% for 10X FFF, respectively, with a slight decrease in attenuation versus field size. The systematic deviation between the OSMS and CBCT was −0.4±0.2 mm, 0.1±0.3 mm, and 0.0±0.1 mm in the vertical, longitudinal, and lateral directions. The mean values and standard deviations of the average deviation and maximum deviation of the daily Winston‐Lutz tests over three months are 0.20±0.03 mm and 0.66±0.18 mm, respectively. Initial testing of this novel system demonstrates the technology to be highly accurate and suitable for frameless, linac‐based SRS and SBRT treatment. PACS number: 87.56.J‐

(0°, ± 1°, ± 2°, ± 3°) using a digital level in the pitch and roll direction. The pitch and roll positioning uncertainties of the online image registration were evaluated using the OSMS QA phantom (Vision RT, London, UK) with and without the RANDO phantom to evaluate the weight factor. The central BB in the phantom was aligned to the isocenter using MV/KV orthogonal pair imaging. A given pitch and roll were applied (+3°/+3°, -3°/-3°, and 0°/0°), a MV/KV image pair was taken, and the distance between the center of the BB and isocenter was measured to evaluate the pitch and roll positioning accuracy.

C.2 Rigidity test of couch insert
The rigidity test was performed at two couch positions in the longitudinal direction with a volunteer (96.2 kg) lying on the couch. The volunteer was positioned at the center of the Calypsocompatible couchtop insert and the couch was also centered laterally. A 3° pitch and roll was applied to the couch. The pitch angle was given to evaluate the potential influence on the roll. A digital level was used to check for possible angular deviation at the longitudinal end of the couch insert. The couch rigidity in roll angle with respect to the couch position in the lateral direction was also tested by off-centering the volunteer to the maximum lateral direction at 24.8 cm. (27)

C.3 Beam attenuation through the couch top and rails
The couch top consists of two mobile, Kevlar support rails, a nonconductive Kevlar Varian/ Calypso insert, and a solid carbon fiber KVue insert. Prior to installation of the linac, both Calypso and KVue inserts along with the support rails were CT scanned with the rails at various positions. An additional scan with the couch top 15 cm above the CT table top was obtained with 20 cm solid water to mimic patient-like setups. The attenuation measurements were obtained for field sizes of 2, 4, and 10 cm 2 at 42 gantry angles, including six pairs of opposing fields and other oblique angles in which the beams traversed the couch inserts and/or rails. The results were then used to determine an accurate structure model for the planning system.

D. IMRT and RapidArc commissioning
A total of 21 plans generated using updated AAPM TG 119 test suite (28) were planned and calculated with the AAA, V.11.0.31 algorithm in the Eclipse TPS. A Solid Water phantom (density: 1.03 g/cm 3 ) was used to evaluate the dosimetric accuracy of both energies using the maximum dose rate. The actual dose rate varied during the delivery for the RapidArc plans. The 6DoF couch top, with the rails in the 'out' position, was included in the dose calculation. The 21 treatment plans included hard C shape, head and neck, head and neck with simultaneous integrated boost, prostate, prostate and lymph nodes, and single isocenter multiple intracranical targets (SIMT) (Fig. 1). All IMRT cases used seven to nine beams and RapidArc cases used two arcs, except for the SIMT case, which used four arcs, with dose optimization constraints that follow the technique of Clark et al.. (29) Point dose measurement using an ion chamber (PTW PinPoint Chamber, Model 31014) and planar dose distribution measurement using films (Gafchromic EBT3) were performed in both the high-dose target and a low-dose region. For the SIMT case, the distance between the isocenter and the center of each of three targets was 2, 4, and 4.5 cm respectively and 16 Gy was delivered to each target. Ion chamber measurements were made at the isocenter and the center of one of the targets 2 cm away. Film was delivered in the axial plane 1 cm posterior to the isocenter. An in-house software was developed to integrate Gafchromic film dosimetry protocol using EBT3 films which streamlines a dose pattern delivery for calibration, calibration curve fitting, film scanning in the fixed scanner position, dose mapping from multiple color channels, rigid registration between the calculated and measured dose plans based on the intensity levels, and profile/gamma analysis. (30) Twelve SRS/SBRT patient plans were measured using the same setup. The treatment sites included intracranial, spine, lung, pancreas, liver, and adrenal gland lesions. Table 2 shows the detail of the plans, including treatment fractions, prescription dose, tumor volume size, conformity index, (31) gradient index, (32) and maximum plan dose. The isocenter dose was measured with ion chamber, and the coronal planar dose distribution was measured with EBT3 films. Figure 1. C shape plan: C shape target planned with IMRT using 6X-FFF; H&N plan: HN PTV target with the cord and parotid glands planned with IMRT using 6XFFF; H&N SIB plan: HN PTV50 (shaded magenta) and PTV60 (blue) targets with the cord and parotid glands planned with IMRT using 6XFFF; Prostate plan: prostate PTV (pink) planned with rectum and bladder with IMRT using 6X-FFF; Prostate+LN plan: prostate+LN(blue) PTV target (red) with rectum and bladder planned with IMRT using 6X-FFF; Single Iso Multi Target plan: 3 targets (orange, purple, and red) planned with IMRT using 6X-FFF. The isodose lines represent 95% (green) and 50% (magenta) prescription dose.

E. The end-to-end test
Daily end-to-end quality assurance tests were performed to assess the overall accuracy of the system from CT simulation, treatment planning, image-based localization, and final treatment delivery using the OSMS QA phantom. The phantom is a polystyrene 15 × 15 × 15 cm 3 cube embedded with five 7.5 mm diameter ceramic BBs ( Fig. 2(a)). One of the BBs was located at the center of the cube. The phantom was scanned with 0.8 mm slice thickness (pixel size 0.6 × 0.6 mm 2 ) without the base plate. The cube and BBs were contoured in Eclipse and used as the reference image. In the treatment room, the phantom was set up on top of an acrylic base plate and fixed to the pegs of an indexing bar for consistent setup. The acrylic plate was engraved with three notches in which the three screws of the OSMS phantom holder were seated. The couch was set at a fixed position (vertical: 10.0 cm; longitudinal: 98.5 cm; lateral: 0.0 cm, pitch: 0.5°, and roll: 0.5°). The OSMS system was first used to localize the phantom surface, and the difference (delta) between the current position of the OSMS phantom and its reference position was recorded ( Fig.  2(b)). CBCT images of the phantom (kV = 100; mAs = 265, 1 mm slice thickness, full fan ) were acquired, and automatic fusion was performed after adjusting the contrast of the acquired image and reference image to achieve optimal window and leveling in order to visualize the BBs (Fig.  2(c)). Six dimensional (6D) fusion shifts were recorded and applied. The phantom position in the OSMS system after correction was recorded to evaluate the residual error. An orthogonal MV/ KV set was taken and 2D-3D image fusion was performed to quantify the residual error (Figs. 2(d) and (e))). An electronic portal imaging device (EPID)-based Winston-Lutz (WL) test was then performed to verify the isocenter targeting accuracy. Twelve 2 × 2 cm 2 , MLC-defined portal images were acquired at four gantry, four couch, and four collimator angles, which were analyzed by an in-house developed C++ software based on an open-source framework (Insight Segmentation and Registration Toolkit 4.3.2) to measure the distance between the center of the central BB and the full width at half maximum (FWHM) of the radiation field ( Fig. 2(f)). The coincidence of the imaging systems and radiation isocenter are evaluated on a daily basis, according to AAPM TG 142 recommendation. (33)  Independent end-to-end tests were performed using the Imaging and Radiation Oncology Core (IROC-Houston) spine and thorax phantoms. The phantoms were scanned, treatment planned, and irradiated at our institution, according to the IROC-Houston credentialing criteria. After irradiation, the phantoms were sent back to IROC-Houston, where absolute point dose was measured with TLDs and 2D film dose planes were measured with Gafchromic EBT2 film, analysis was completed independently by IROC-Houston. Treatment plans were generated with the Eclipse TPS using the same AAA algorithm and delivered using the RapidArc technique for the spine phantom and IMRT for the thorax phantom. Both phantoms were localized using the OBI system, where CBCT was used for initial setup.
The spine phantom consists of a pentagon shaped PTV (42 cc) abutting bone and a cylindrical spinal cord structure; the PTV is set between the right and left lung structures. The spine phantom has four TLDs within the PTV structure in the high-dose region and one within the heart in the low-dose region. Two films bisect the PTV in the axial and sagittal planes. The thorax phantom consists of an ellipsoidal-shaped PTV (72 cc) located in the middle of a cylindrical volume of lung. The thorax phantom contains two TLDs within the PTV, and two TLDs in the low-dose region, one in the heart and one in the cord. Three films bisect the PTV in the axial, coronal, and sagittal planes.

F. Treatment head leakage test
Treatment head leakage was measured using 30 pairs of Luxel+ T series dosimeters (Landauder, Glenwood, IL) placed around a 2 meter radius circular plane, in a plane perpendicular to the beam axis at the isocenter. Figure 3(a) shows the placement of each pair of dosimeters. Ten thousand (10,000) MUs were delivered to the dosimeters at gantry 0° position, with both MLC and jaw at most closed position, using the highest energy, 10X FFF, at 2400 MU/min. The average reading of each pair of dosimeters was recorded. The localization of phantom surface using the OSMS system. The difference (delta) between the current position of the OSMS phantom and its reference position is shown in 6DoF. (c) The six degree automatic fusion between planning CT and CBCT after adjusting the contrast of the acquired image and reference image to achieve optimal visualization of the BBs. An orthogonal MV (d)/KV (e) image set is taken and 2D-3D image fusion is performed to quantify the residual error. (f) Four representative MLC defined portal images of the Winston-Lutz test.

G. Developer mode
The Edge system includes Developer Mode, enabling the use of XML scripting for automation of commissioning and QA procedures. XML scripting was used for various commissioning tasks including beam scanning, couch modeling, and end-to-end tests. Figure 4 shows the PDD curves normalized at D max for 6X FFF (a) and 10X FFF (b) for the field sizes ranging from 1 × 1 to 40 × 40 cm 2 . Table 3 summarizes the D max and PDD values at 5, 10, 20, and 30 cm depth. The deviations between the photon beam curves acquired on the new linac-based system (Edge) and the previously published machine with FFF beams (TrueBeam) beyond D max were within 1.0% for both energies. The beam quality specifier (%dd(10) × ) for the Edge was 63.0% and 70.6% for 6X FFF and 10X FFF, respectively without 1 mm lead foil. With a 1 mm lead foil, %dd(10) × increased to 71.1% for 10X FFF; however, the difference between the quality conversion factors (k Q ) for 10X FFF were within 0.1% with and without the lead foil. Figures 4(c) and (d) illustrate the cross-plane profiles measured at 10 cm depth for all ten field sizes from 1 × 1 to 40 × 40 cm 2 . The curves are normalized to 100% on the central axis. Since only FFF modes were commissioned for the Edge, we could not use the penumbra 1 Figure 3. (a) The placement of the T series dosimeters around a 2 m radius circular plane. The deep dose equivalent map of photon and neutron combined (b), photon only (c) and fast neutron only (d). The maximum measured head leakage dose was 8.45, 6.85 and 1.55 mSv respectively, all located at point E, 0.5 m toward the couch direction. normalization method proposed by Pönisch et al. (34) Figure 5 shows direct comparison of profile curves between the Edge and the TrueBeam for two representative fields using 10X FFF: 2 × 2 cm 2 and 10 × 10 cm 2 . The profiles between the Edge and the TrueBeam were practically the same, with slightly sharper penumbra obtained on the Edge at all the depths.

A.1 Percent depth dose and profile evaluation
The values of P ion at the central axis and two off-axis positions were compared. The output constancy was within 0.1% with various dose rates for both energies. The ion chamber collection efficiency off-axis agreed within 0.3% of the values at the central axis for the two field sizes evaluated. 1 Figure 4. PDD curves normalized at Dmax for 6XFFF (a) and 10XFFF (b) for the field sizes ranging from 1 × 1 to 40 × 40 cm 2 . The cross-plane profiles measured at 10 cm depth for all 15 field sizes for 6XFFF (c) and 10XFFF (d). A CC04 cylindrical chamber was used for field sizes greater than 2 × 2 cm 2 using the 400 MU/min dose rate and the SFD was used for field sizes 1 × 1 cm 2 and 2 × 2 cm 2 using the maximum dose rate. The curves are normalized to 100% on the central axis. Fig. 4. PDD curves normalized at D max for 6X FFF (a) and 10X FFF (b) for the field sizes ranging from 1 × 1 to 40 × 40 cm 2 . The cross-plane profiles measured at 10 cm depth for all 10 field sizes for 6X FFF (c) and 10X FFF (d). A CC04 cylindrical chamber was used for field sizes greater than 2 × 2 cm 2 using the 400 MU/min dose rate, and the SFD was used for field sizes 1 × 1 cm 2 and 2 × 2 cm 2 using the maximum dose rate. The curves are normalized to 100% on the central axis.

A.2 Output factors
The output factors S cp for the symmetrical fields and rectangular fields are tabulated in Tables  4 and 5 for 6X FFF and 10X FFF, respectively. The shielded area in the table corresponds to data measured with the SFD detector. S cp for symmetrical fields ranging from 1 × 1 to 40 × 40 cm 2 were also plotted in Fig. 6 and compared against the TrueBeam machine ( Fig. 6(b)). The maximum deviation between the Edge and TrueBeam was 1.6% for field size of 1 × 2 cm 2 (6X FFF) and 1.0% for 1 × 1 cm 2 (10X FFF).
1 Figure 5. Comparison of profile curves between the Edge and the TrueBeam for two representative fields using 10XFFF: 2 × 2 cm 2 and 10 × 10 cm 2 . The profiles between the Edge and the TrueBeam were practically the same with slightly sharper penumbra obtained on the Edge at all the depths.

A.3 HDMLC transmission and DLG
The measured DLG values were 0.507 mm for 6X FFF and 0.622 mm for 10X FFF. Optimized values, which were fitted using Eclipse dose calculations and measurements based on representative spine radiosurgery plans, were 0.700 mm and 1.000 mm, respectively. The MLC transmission values were 1.209% for 6X FFF and 1.427% for 10X FFF. Dose difference ratios of ion chamber measurements were 0.015% ± 0.008% for 6X FFF and 0.010% ± 0.010% for 10X FFF, and the passing rates for 2%/2 mm gamma criteria were 98.0 ± 1.0 for 6X FFF and 96.9 ± 1.9 for 10X FFF after the DLG optimization. Figure 7(a) and (b) shows the PDD data for the conical cones for 6X FFF and 10X FFF. The off-axis ratios for all the conical cones at the depth of 5 cm at 100 cm SSD are shown in Fig. 7(c) and (d). All beam profile data were normalized to the central axis. The beam penumbra (width between 90%-10% and 80%-20%) increases as the diameter of the cone increases, as shown in Fig. 8. 6X FFF has sharper penumbra ranging from 1.2-1.8 mm (80%-20%) and 1.9-3.8 mm (90%-10%) relative to 10X FFF, which has 1.2-2.2 mm and 2.3-5.1 mm, respectively. Table 6 shows the OFs of the cones using the Edge detector, with and without cross-calibration, at an intermediate field size. Because the Edge detector is independent of variation in energy spectrum, (35) minimal difference between the two measurements was observed (OFs were within 0.2% and 0.7% for 6X FFF and 10X FFF, respectively).

B. Conical cones
The percent difference between OFs we measured using different detectors and the data from the manufacturer measured with the Edge detector (available at the Vendor website) is also shown in Table 6. The difference was ~ 1% for the Edge detector and increased to 4% for the SFD detector. As observed in Table 6, the PFD, CC01, and PinPoint ion chambers show much lower OFs for the smaller cones due to the volume averaging effect. 1 Figure 7. PDD curves normalized at Dmax for 6XFFF (a) and 10XFFF (b) for the conical cones ranging from 4 mm to 17.5mm. The off-axis ratio measured at 5 cm depth, 100 cm SSD for 6XFFF (c) and 10XFFF (d). The curves are normalized to 100% on the central axis.
1 Figure 8. The beam penumbra (width between 90% -10% and 80% -20%) increases as the diameter of the cone increases for both energies. The beam penumbra increases faster for the 90-10% value than for the 80-20% value.   8. The beam penumbra (width between 90%-10% and 80%-20%) increases as the diameter of the cone increases for both energies. The beam penumbra increases faster for the 90%-10% value than for the 80%-20% value.

C. Couch commissioning
The PRO accuracy (digital reading provided by the linac) at each axis agreed with the measurements within 0.1% with and without weight on the couch. Only 0.1° deviation was observed in the pitch direction with the phantom on the couch. Table 7 summarizes the BB offsets from the isocenter from MV/KV portal image verification. The maximum deviation was 0.5 mm when both pitch and roll were at -3°. For the rigidity test, with both pitch and roll at ± 3°, when the volunteer was off-centered as much as possible (weight shift), the deviation between the PRO and measurement was 0.1° (3° ± 0.1°). When the couch was moved laterally to the maximum range, the roll angle deviation became 0.4°. This 0.4° deviation was not due to the rigidity of the couch insert, but due to the rigidity of the upper couch moving mechanism. (27) When the lateral movement of the couch was half of the maximum range, the deviation was 0.2°. The deviation was linear with the lateral offset. Figure 9 shows the relative attenuation of the couch at various gantry angles, ranging from 90° to 270°, using the 6X FFF beam for three field sizes. The attenuation in positioning of the rails in 'out' and 'in' positions was studied using a 4 × 4 cm 2 field size. There was a slight decrease in attenuation versus field size. The attenuation properties of KVue imaging couchtop were very similar to the Calypso-compatible insert. In fact, the CT data and attenuation data were virtually indistinguishable between the two couchtop inserts, so the same couch model can be used in the TPS for both inserts.

D. IMRT and RapidArc commissioning
Composite Gafchromic film and ion chamber results are shown in Table 8 for the measurements in the high-dose and low-dose region for both IMRT and RapidArc plans. The dose difference ratio was -0.0% ± 1.4% (range, -1.8%-3.5%) for 6X FFF and -0.6% ± 1.6% (range, -0.5%-4.7%) for 10X FFF in the high-dose region, and -0.3% ± 2.3% (range, -4.2%-2.9%) for 6X FFF and 1.5% ± 3.7% (range, -1.9%-11.9%) for 10X FFF in the low-dose region. The percentage of points passing the gamma 3%/3 mm criteria for both IMRT and RapidArc plans was 95.5 ± 4.2 (6X FFF) and 97.9 ± 2.7 (10X FFF) in the high-dose area, and 95.5 ± 3.9 (6X FFF) and 97.5 ± 2.5 (10X FFF) in the low-dose region. The profiles in the vertical and horizontal directions were analyzed for all tests. Figure 10 shows the analysis of four representative cases. Table 2 shows the ion chamber and Gafchromic film results for 12 patient SRS/SBRT plans. The average point dose difference was 1.3% ± 0.9% (range, -2.2%-2.6%) over all the plans. The average gamma pass rate was 99.4% ± 0.8% for 3%/3 mm criteria and 93.0% ± 6.0% for 3%/1 mm criteria.    9. The relative attenuation for the KVue couch at various gantry angles ranging from 90° to 270° using 6X FFF beam at three different field sizes. Relative attenuation is greatest in a small window of oblique entry. The attenuation in positioning of the rails in 'out' and 'in' positions was studied using a 4 × 4 cm 2 field size.
1 Figure 9: The relative attenuation for the KVue couch at various gantry angles ranging from 90° to 270° using 6XFFF beam at 3 different field sizes. Relative attenuation is greatest in a small window of oblique entry. The attenuation in positioning of the rails in 'out' and 'in' positions was studied using a 4 × 4 cm 2 field size.

E. End-to-end testing
The coincidence of the OSMS and CBCT isocenters was checked on a daily basis. Figure 11(a) shows the daily variations in the translational and rotational direction from the first three months of operation. The daily isocentric coincidence of the CBCT and MV/kV planar imagers is shown in Fig. 11(b). The systematic deviation between the OSMS and CBCT was -0.4 ± 0.2 mm, 0.1 ± 0.3 mm, and 0.0 ± 0.1 mm in the vertical, longitudinal, and lateral directions. There was no residual error in the angular directions. The analysis also showed 0 mm discrepancy in the translational directions between the CBCT and MV/kV orthogonal pair, although 0.1°-0.2° difference was shown in the angular directions. The average and maximum absolute values of the daily Winston-Lutz test are shown in Fig. 11(c). The mean values and standard deviations of the average deviation and maximum deviation are 0.20 ± 0.03 mm and 0.66 ± 0.18 mm, respectively. The deviations were consistent and within the tolerance (0.75 mm average and 1.0 mm maximum) recommended from TG 142 and the ASTRO quality and safety guidelines for SRS/SBRT. (33,36) Commissioning was independently verified with the IROC spine and lung credentialing phantoms. All phantoms passed the IROC credentialing; results are shown in Table 9.  Figure 3 shows the deep dose equivalent (DDE) map of photon and neutron combined (b), photon only (c) and fast neutron only (d). Thermal neutron dose was within the minimally detectable region of the dosimeters. The maximum measured head leakage dose was 8.45, 6.85, and 1.55 mSv respectively, all located at point E, 0.5 m toward the couch direction. The head leakage from the linac was within 0.1% of the dose at isocenter.

G. Developer mode
Many iterations of the couch top measurements were required to fully sample the rails and oblique incidence through the couch for different energies (6X FFF and 10X FFF), field sizes (2 × 2 cm 2 , 4 × 4 cm 2 , and 10 × 10 cm 2 ), and shifts in isocenter position (shifts of various magnitude in each of the three translational directions). Automated measurements required only one physicist, while manual measurements required at least two physicists to handle linac positions/beams and data recording. MLC apertures were generated outside of the TPS, and with the .xml file format, double-checking without use of TPS/operator console was possible. For automated couch top measurements, the time required for each set of angles was approximately 8 min. Without scripting, each set required approximately 11 min. Similar time efficiency gains (approximately 25%) were found for isocenter verification measurements.

IV. DISCUSSION
This study summarizes the commissioning process of the Edge, a dedicated system for SRS/ SBRT treatment. Although it offers the advanced imaging package, the 6DoF treatment couch, and intracranial radiosurgery accessory package, the beam data characteristics and mechanical parameters of the Edge are similar to the TrueBeam.
Beam data from five TrueBeam linacs at three different institutions were previously compared, (9) and we noted excellent agreement between the beam data collected on the Edge and that on the TrueBeam linacs. The CC04 chamber was used to scan the PDDs and profiles for the Edge, while the CC 13 chamber was used for the TrueBeam, and due to its smaller active volume, dose falloff in profiles for the Edge was slightly sharper than that for the TrueBeam. Kim et al. (37) compared PDD and cross-plane profiles of a 6 MV SRS beam using four different detectors (SFD, PFD, CC01, and CC13). They showed that PDDs from all detectors were in good agreement for field sizes ranging from 1 × 1 to 6 × 6 cm 2 . Diodes overestimated the dose for field sizes larger than 6 × 6 cm 2 due to lower energy scattered photons. For profile scans, CC13 ion chamber showed a larger blurring of penumbra even for field size of 10 × 10 cm 2 . A small sensitive volume detector is recommended to achieve a sharper penumbra. However, CC01 (steel electrode) or diode are likely to measure higher dose in the tails due to the overresponse to low-energy scattered photons. The dose per pulse at the central axis is higher than off-axis due to the absence of the flattening filter. Since ion collection efficiency is a function of the dose per pulse, P ion was measured and compared between the central axis and off axis for two different field sizes (1.007, 1.009, 1.010 at central axis, 2.4 cm off axis and 5.6 cm off axis, respectively, for 6X FFF and 1.011, 1.010, 1.009 for 10X FFF). The consistency of P ion at different locations ensures there is no additional correction needed for the profile measurement.
Several challenges in small field dosimetry exist, including lack of charged particle equilibrium (CPE), overestimation of field size, perturbation of the particle fluence in the chamber, and volume averaging effect of the detector. (38) Therefore, it is crucial to choose the correct detector, considering the size, energy dependence, and perturbation, for example. A new formalism has been developed for the dosimetry of small field. (39) For the Edge commissioning, the machine-specific reference field is defined at 3 × 3 cm 2 , since the conventional 10 × 10 cm 2 cannot be established for all detectors considering the energy dependence of the diodes and volume averaging effect of the ion chambers. The field factor Ω f clin ,f msr Q clin ,Q msr , under the notion proposed by Alfonso et al., (39) which converts the absorbed dose to water for the machinespecific reference field (3 × 3 cm 2 ) to the absorbed dose to water for the small clinical field, should be carefully evaluated to account for the difference of the detector response and beam quality at two different field sizes. A Monte Carlo calculated factor k f clin ,f msr Q clin ,Q msr was recommended to correct the field factor. Several studies have been published since then to generate correction factors for various detectors from different treatment platforms. (40,41) The diodes were shown to have an overresponse at small fields. A correction factor should be applied to the SFD for field sizes less than 1 × 1 cm 2 and the Edge detector for field sizes within 1.5 × 1.5 cm 2 . (41) This factor might also explain the 4% difference in the output factor measurements between the Edge detector and SFD for conical cones. A Monte Carlo simulation for the FFF beams may be beneficial in verifying the correction factors for stereotactic diodes at very small field sizes (< 2 cm). There are various methods to measure the DLG: 1) measuring the distance between the radiation and geometrical field edge of a MLC-defined field size; 2) matching the gap width profiles with the measured values; 3) optimizing the parameters based on treatment delivery; and 4) sweeping MLC leaves with a variety of sliding MLC gap widths. (42,43) For the Eclipse TPS, only one DLG value can be commissioned for all different field sizes and delivery techniques. Therefore, there is a trade-off in the optimal DLG between IMRT and RapidArc measured fields, as well as the fields with different sizes and modulation. The difference between the measured and optimized DLG values is caused by different contributions to the dose from the beam penumbra, which is a consequence of different patterns of leaf movement. Szpala et al. (44) found out that the DLG values are a function of the distance (in the BEV) between the dose point and the leaf ending, and the width of the MLC slit. Therefore, calculation using a single DLG value may overestimate the measurement in the proximal penumbra, while it may underestimate the dose in the distal penumbra for RapidArc delivery. (44) For IMRT delivery, the DLG values for smaller and larger regions average out and a single value can serve as the optimal value for different widths of the MLC slits. (44) Therefore, the DLG values were optimized for RapidArc delivery by evaluating the measured and calculated dose for selected spine radiosurgery cases due to the requirement of an extremely steep dose gradient. The adjustments did not have much impact on the IMRT delivery. The dose calculation accuracy was further validated in a more comprehensive manner, using test cases representative of various clinical treatment sites.
Tissue maximum ratios (TMR) and off axis ratios are used for the cone-based dose calculation. TMR values can be measured by draining or filling water in a 3D water tank or derived from PDD curves. It is challenging to use the conventional conversion methods, since phantom scatter factors for small fields are difficult to measure. Van Battum et al. (45) proposed to obtain TMR values from PDD curves and total scatter factors. A depth-dose curve corrected for source detector distance was generated from existing PDD curves and the dose at each depth and field size was fitted by a double exponential function. TMR was then calculated by taking the ratio of the dose at the depth of interest and the reference depth. Battum and colleagues reported the agreement between calculated and measured TMR was within 2%. TMR values were spot checked on the Edge system at nine points for each cone and compared against the converted data. The difference was within 2% except at 20 cm, the deepest depth. Larger discrepancies were noted at depths beyond 20 cm, which is generally greater than the maximum depth required for intracranial SRS treatment. This method can be considered an alternative option to obtain TMR values for cones when a precise TMR measurement is not available from the water tank.
Conical cones may provide a sharper beam penumbra than the MLCs since the cone is closer to the isocenter and more transmission occurs at the round leaf ends of the MLCs. The beam penumbra for the cones is a function of depth, cone size, and energy. It increases as the cone size, depth or energy increases. The beam penumbra increases faster for the 90%-10% value than for the 80%-20% value, as shown in Fig. 8.
The 6DoF (PerfectPitch) couch top is equipped with rails, which will lead to errors in the delivered dose if the rails are not properly accounted for the in treatment plan. This is especially important in the context of spine SRS, where highly modulated posterior beams are used and the isodose fall off from 90% to 50% line is on the order of 3 mm. Therefore, the attenuation effect of the rails and couch tops should be measured. A proper couch model should be established in the TPS according to recommendations from AAPM Task Group Report No. 176. (46) By taking CT scans of the couch top prior to installation on the treatment unit, couch models can be developed, along with a setup for future planning and delivery to a QA phantom. In this study, such a couch model was incorporated for all the test plans related to the Edge commissioning, phantom QA, and patient planning. The couch model is also used for routine patient treatment planning.
To optimize use of the couch model for RapidArc delivery, one solution is to place both couch rails in the 'in' position and start the arc at oblique angles to avoid the beam traversing through the rails. However, the rigidity of the couch insert should be carefully evaluated in the lateral direction (patient left and right) for such a configuration. The deviation was linear with lateral translation, due mainly to the rigidity of the couch moving mechanism.
The AAA dose calculation algorithm was reported to overestimate the dose beyond lowdensity inhomogeneities. (47) Errors could be greater than 2.5% when using the AAA to calculate the dose when the beams transverse a large air gap from the treatment couch or an immobilization device to the patient. (48) The Acuros XB algorithm, a numerical method based on linear Boltzmann transport equation solver, is also available in the Eclipse TPS. The algorithm has been compared to the AAA in numerous studies, and demonstrated that it has better agreement with full MC simulation in slab phantoms containing various materials. (49,50) Rana et al. (51) reported the dose calculation using the Acuros XB had better agreement with the measurement than using the AAA in the inhomogeneous phantoms. The validation of the AAA and Acuros XB dose calculation algorithm using a heterogeneous phantom will be performed in the second tier commissioning.
Since target localization may incorporate single or multiple imaging modalities and 6DoF couch correction, end-to-end tests were designed to evaluate the coincidence of each imaging modality with the radiation isocenter, the accuracy of 3D-3D and 2D-3D image registration, the precision of 6DoF correction, and the coincidence of gantry, collimator, and couch axes with the radiation isocenter. The laser and crosshair alignment should also be checked after the phantom localization. By performing the Winston-Lutz test on a daily basis, the localization accuracy can be accessed and deviations can be easily identified to trigger further action, such as imaging system calibration, couch precision test, or linac mechanical check.