Commissioning and initial stereotactic ablative radiotherapy experience with Vero

The purpose of this study is to describe the comprehensive commissioning process and initial clinical performance of the Vero linear accelerator, a new radiotherapy device recently installed at UT Southwestern Medical Center specifically developed for delivery of image‐guided stereotactic ablative radiotherapy (SABR). The Vero system utilizes a ring gantry to integrate a beam delivery platform with image guidance systems. The ring is capable of rotating ± 60° about the vertical axis to facilitate noncoplanar beam arrangements ideal for SABR delivery. The beam delivery platform consists of a 6 MV C‐band linac with a 60 leaf MLC projecting a maximum field size of 15×15 cm2 at isocenter. The Vero planning and delivery systems support a range of treatment techniques, including fixed beam conformal, dynamic conformal arcs, fixed gantry IMRT in either SMLC (step‐and‐shoot) or DMLC (dynamic) delivery, and hybrid arcs, which combines dynamic conformal arcs and fixed beam IMRT delivery. The accelerator and treatment head are mounted on a gimbal mechanism that allows the linac and MLC to pivot in two dimensions for tumor tracking. Two orthogonal kV imaging subsystems built into the ring facilitate both stereoscopic and volumetric (CBCT) image guidance. The system is also equipped with an always‐active electronic portal imaging device (EPID). We present our commissioning process and initial clinical experience focusing on SABR applications with the Vero, including: (1) beam data acquisition; (2) dosimetric commissioning of the treatment planning system, including evaluation of a Monte Carlo algorithm in a specially‐designed anthropomorphic thorax phantom; (3) validation using the Radiological Physics Center thorax, head and neck (IMRT), and spine credentialing phantoms; (4) end‐to‐end evaluation of IGRT localization accuracy; (5) ongoing system performance, including isocenter stability; and (6) clinical SABR applications. PACS number: 87.53.Ly

projection images every 0.5° through a 225° rotation (CW from 320° to 175° or CCW from 40° to 185°) and reconstructing a volume (CBCT) using either source-detector pair. The maximum CBCT field of view (FOV) is 20 cm in diameter by 15 cm long. The system can also be operated in stereoscopic fluoro mode for real-time visualization and verification of dynamic tracking. The localization capabilities of both 2D/2D and CBCT systems were assessed as part of this work and are described below. The Vero system is also equipped with an amorphous silicon electronic portal imaging detector (EPID) (RID1640; Perkin Elmer, Santa Clara, CA) permanently mounted in the ring (i.e., not retractable) at an SDD of 221.2 cm. The EPID is operated in 1024 × 1024 mode, and acquires continuously at a frame rate of 2 Hz.
Because of the unique degree of freedom of the ring rotation, the primary beam is capable of impinging on a much greater wall area than a conventional linac. For an installation in an existing room, there is a high likelihood that the width of the primary barrier is insufficient. As compensation, the Vero ring contains a beamstopper that attenuates the primary beam by an additional factor of 10 3 . For our installation, no additional shielding was required.
In this work we describe the clinical commissioning of the Vero system for SABR applications, using key AAPM and ASTRO documents to guide the process. (13,14) We focus specifically on beam data acquisition and dosimetric commissioning of the treatment planning system and end-to-end evaluation of IGRT localization capabilities. Specially designed anthropomorphic head and thorax phantoms were used for hidden target testing and to benchmark a Monte Carlo algorithm for SABR applications in lung and spine. The Radiological Physics Center lung, head and neck (IMRT), and spine credentialing phantoms were used to independently verify system performance.

A. Beam data
Beam data acquisition is a common task performed routinely by medical physicists; the general process has been described in detail in the report of AAPM Task Group 106. (25) Acquisition of beam data for SRS and SABR can be particularly challenging, however, due to the small size of the fields employed and the partial volume effects associated with many common dosimeters. Further, small photon beams exhibit a loss of lateral electronic equilibrium on the central axis, producing output factors that falloff rapidly for fields below 10 mm in diameter. (26)(27)(28)(29) Centering of the dosimeter in the beam is also challenging, and improper alignment of the beam and detector can introduce significant errors.
The planning system for Vero, iPlan versions 4.1 and 4.5 (BrainLAB AG), require a number of specific beam measurements, including percent depth dose, diagonal profiles, relative output factors, and MLC transmission. If dynamic IMRT delivery is to be used, an assessment of the radiation leaf gap due to the rounded leaf ends is also required. For dose calculation in homogeneous media, a pencil beam (PB) algorithm is used. For dose calculation in heterogeneous media, iPlan employs a Monte Carlo (MC) algorithm based on the XVMC dose developed by Fippel and Kawrakow. (30,31) Fragoso et al. (32) performed extensive benchmarking of the algorithm against measurement, albeit for a different treatment unit/MLC. For commissioning the Monte Carlo algorithm, iPlan requires additional beam measurements, including central and off axis profiles and output factors in air and water. A complete list of beam data requirements is shown in Table 1. Percent depth dose data were acquired at source-to-surface distances (SSD) of both 90 and 100 cm using a model 31014 pinpoint chamber (PTW, Freiburg, Germany) for field sizes 2 × 2 cm 2 and greater. The 31014 has an inner diameter of 2 mm, thus acquired ionization data were shifted proximally by 0.6 mm. Data were smoothed using a least square algorithm with a sliding window of 4 mm. A stereotactic diode (Edge, Sun Nuclear Corp., Melbourne, FL) was used to measure PDD, again at 90 and 100 cm SSD, for the 1 × 1 cm 2 field size. To improve the signal-to-noise characteristics, all diode data were acquired three times and subsequently averaged. All PDD data were resampled in 1 mm increments.
Off-axis profiles were measured per manufacturer's requirements. Diagonal (radial) profiles for a fully open 15 × 15 cm 2 field were acquired at an SSD of 100 cm at seven depths in water, ranging from 0.5 to 350 cm, using the model 31014 pinpoint chamber with the central electrode oriented perpendicular to the beam axis. Pinpoint chamber profiles were obtained along both diagonals and subsequently averaged for input into the planning system. For comparison, measurement of diagonal profiles was repeated using the Edge diode. The Monte Carlo algorithm also requires measurement of transverse profiles in the direction parallel (left-right = "inline") and perpendicular (head-foot = "crossline") to the direction of MLC leaf travel. Transverse profiles were measured at three depths (1.5, 10.0, and 20.0 cm) at both 90 and 100 cm SSD, for square and rectangular fields ranging from 1 × 1 to 15 × 15 cm 2 . The pinpoint chamber was oriented with the smallest dimension parallel to the scan direction. Standard profile characteristics, flatness, symmetry, measured full width at half maximum (FWHM), and penumbra (distance from 80% to 20%) were calculated and compared with characteristics from other stereotactic linacs.
Scatter factors, S cp , for square and rectangular fields ranging from 1 × 1 to 15 × 15 cm 2 were measured in water at an SSD of 100 cm and depth of 10 cm using both the pinpoint chamber and stereotactic diode. Three readings were acquired for each field size and subsequently averaged. Data were normalized to a 10 × 10 cm 2 field size. The Edge diode is effectively independent of variation in energy spectrum that is manifest in changing field size (33) and, therefore, the Inline and crossline profiles in air for source-to-detector distances (SDD) of 85, 100, and 115 cm for square and rectangular fields 1×1 to 15×15 cm 2

Scatter Factors
Measurements in water at 100 cm SSD and Measurements in water at 90 cm SSD a depth of 10 cm for square fields 1×1 to and a depth of 10 cm for square and 15×15 cm 2 relative to a 10×10 cm 2 field rectangular fields 1×1 to 15×15 cm 2 relative to a 10×10 cm 2 field Measurements in air for a source-todetector distance (SDD) of 100 cm for square and rectangular fields 1×1 to 15×15 cm 2 relative to a 10×10 cm 2 field MLC Leakage Measurement in water at 100 cm SSD and a depth of 10 cm with MLC completely closed relative to an open 15×15 cm 2 field Absolute Dose Gy/MU in water at 100 cm SSD, at a depth Gy/MU in water at 90 cm SSD, at a of 10 cm, for a 10×10 cm 2 field depth of 10 cm, for a 10×10 cm 2 field procedure of referencing diode measurements to the reference 10 × 10 cm 2 field through a smaller intermediate field (i.e., "daisy-chaining") was not performed. Additional data were acquired for beam modeling for the Monte Carlo algorithm, notably, central axis profiles, off axis profiles, and output factors in air. In-air CAX data were acquired for a source-to-detector distance ranging from 85 to 115 cm. In-air, off-axis data were acquired in both transverse directions at distances of 85, 100, and 115 cm from the source. All in-air scans were performed using the pinpoint chamber equipped with a brass build-up cap suitable for 6 MV photons.
MLC leaf transmission was evaluated separately for each leaf bank. In each case, the abutting leaf junction was positioned at the maximum range on the MLC, ± 7.5 cm from the central axis. At this location, the fixed primary collimator sufficiently shields the junction. Measurements were performed using extended dose range (EDR2) film (Carestream Health, Inc, Rochester, NY) placed perpendicular to the beam axis at a distance of 100 cm from the source, at a depth of 5 cm in Solid Water. 10,000 MU were delivered to each closed-MLC field. A third film was irradiated under the same conditions using a 10 × 10 cm 2 open field and 100 MU. Films were scanned and analyzed using a commercial film dosimetry system (RIT113 V5.0; Radiological Imaging Technology, Colorado Springs, CO). Optical density was converted to dose using a sensitometric curve with a dose range from 0 to 6 Gy. The average transmission was determined from the ratio of the dose of the fully closed to that of the fully open field over 80% of the open field area. Additional transmission measurements were performed using a 0.3 cc ionization chamber (PTW 31013) located on the central axis and irradiated under identical conditions (i.e., 100 cm source-to-detector distance, 5 cm depth in Solid Water).

B. dosimetric commissioning
Following beam data acquisition, the treatment planning system was fully commissioned to ensure accurate calculation of dose and monitor units. This involved a systematic comparison of calculation and measurement, ranging from simple configurations such as a single beam, to sophisticated arrangements of beams encompassing all modes of delivery encountered in clinical practice. Prior to commissioning, the Vero machine was configured for the iPlan treatment planning system using the Beam Profile Editor version 7.1 utility (BrainLAB AG), and measured beam data were subsequently entered. The radial factors required by the planning system (a radial factor is the derivative of the half transverse profile) and the parameters for the Monte Carlo virtual energy fluence model (VEFM) (34) (Table 2) were provided by the vendor based on user measured data.
A stack of Solid Water, 30 cm × 30 cm × 22 cm high was scanned and imported into the planning system. The stack is drilled for the pinpoint chamber at a depth of 11 cm and can hold film in a coronal orientation at several depths. Single anterior beams ranging in field size from 2 ×2 to 15 × 15 cm 2 were simulated in iPlan, and the dose calculated at discrete points along the central axis was compared with the measured beam data. Subsequently, a series of Table 2. Monte Carlo virtual energy fluence model parameters. The VEFM consists of a primary and scatter photon source modeled as two-dimensional Gaussian shapes, and an electron contamination source. At our institution, the Vero is used primarily for sophisticated treatments such as SABR. To assess accuracy and precision of both localization and dosimetric characteristics using image guidance in a clinically relevant setting, specialized anthropomorphic head and neck and thorax phantoms were designed and constructed (Integrated Medical Technologies, Troy, NY) (Figs. 1(a) and (b)). Each phantom contains radio-opaque spheres for hidden-target evaluation and multiple holes drilled to facilitate point dose measurements. In addition, the thorax phantom  contains unit density targets imbedded in lung to allow direct verification of dosimetric calculations in heterogeneous media. Planning CT scans were obtained for both phantoms, and a series of treatment plans encompassing targets in the brain, lung, and spine were calculated and subsequently delivered to the phantoms. Point dose and planar film dosimetry were performed as described above.

Source Parameter Value
Commissioning was further verified through the use of the Radiological Physics Center (RPC) thorax-lung, spine, and head and neck IMRT phantoms. In each case, plans were developed based on the target coverage and OAR constraints required by the RPC; planning parameters and constraints are summarized in Table 3.

C. Localization accuracy
Assessment of localization accuracy in a manner that appropriately mimics the clinical SABR process is an essential aspect of device commissioning and ongoing QA. End-to-end target evaluation that incorporates CT, planning, and image guidance has become the standard for assessing overall system localization uncertainties. (12)(13)(14)35,(36)(37)(38)(39) To facilitate end-to-end evaluation of the Vero, the radio-opaque spheres located within the head and thorax phantoms were identified on the respective planning CTs. An isocenter was positioned at the center of each sphere, and the resulting plan and reference datasets were transferred to the Vero through the For each phantom, external marks were used to establish an initial position on the treatment couch; generally this initial position was within 2 cm of the desired isocenter in each direction. Based on anatomical data, the phantom was subsequently aligned to the machine isocenter using either stereo X-ray (2D/2D) or CBCT (3D) image guidance. Automated image fusion was used to register the localization and reference images, followed by small manual adjustments. To guide 3D localization, contours from the planning CT are overlaid on the CBCT. With both techniques, the final alignment was determined by visual inspection of the images.
Resulting translations and rotations were applied remotely to the Vero couch. The process was repeated a total of 28 times, 16 using CBCT image guidance and 12 using stereo x-ray image guidance. For each trial, two or four beam's-eye-view EPID images were acquired corresponding to gantry angles of 0° (AP), 180° (PA), 90°, and/or 270° (lateral), using a field size of 2 × 2 cm 2 . For each EPID image, a threshold was applied to isolate the open square field, from which the central pixel was subsequently identified. This process was repeated to isolate the hidden spherical target, and the two-dimensional offsets between the field and target centers were recorded. This resulted in 28 independent measurements in the anterior-posterior and left-right directions, and 56 independent measurements in the superior-inferior direction (since AP/PA and lateral images share a common sup-inf axis). All image analysis was performed using ImageJ. (40) d. ongoing system performance Since installation, numerous operational machine parameters have been tracked at regular intervals to assess ongoing system performance. Beam characteristics including output, flatness, and symmetry were assessed daily using a commercial QA device consisting of multiple ionization chambers (Daily QA3; Sun Nuclear Corporation). Coincidence of the lasers and kV and MV systems was assessed daily by the therapy staff, and repeated monthly as part of physics QA. Briefly, a 12 × 12 × 12 cm 3 phantom containing a radio-opaque ball 1 cm in diameter is positioned on the treatment table and brought into coincidence with the Vero lasers. Stereoscopic X-rays are obtained, the location of the ball relative to the kV imaging center is determined, and offset is subsequently corrected by translating the table in three dimensions. A series of 12 MV images at various gantry and ring angles are subsequently obtained, and the location of the ball relative to the MV imaging center is determined. All dosimetric and localization data were analyzed for compliance and for trends.

E. Clinical applications
The majority of SABR treatments at our institution involve tumors of the lung, liver, and spine, though we have institutional protocols for breast, prostate, pancreas, and other disease sites. For tumors in the peripheral lung and liver, we employ a 3D conformal technique using multiple (≥ 10) noncoplanar beams; the rotating ring on the Vero provides opportunity for highly oblique beam directions. For tumors in the central lung and spine, fixed gantry IMRT is used; all spinal tumors are treated with 13 coplanar beams evenly distributed through 360°. For patients not enrolled in institutional or national study group trials, the most common prescriptions are as follows. Spine tumors are generally treated in a single fraction, with 20 Gy delivered to the GTV and 14 Gy to the CTV. Central and peripheral lung tumors are treated in 5 fractions of 10 Gy and 3 fractions of 18 Gy, respectively. Liver tumors are typically treated in 5 fractions of 10-12 Gy or on an institutional single fraction dose escalation protocol. The iPlan Monte Carlo algorithm is used for any tumors located in the thoracic region (lung and spine). Systematic use of normal tissue dose constraints is followed in all SABR treatments. (41) III. rESuLtS Figure 2 shows percent depth dose for several field sizes, plotted against PDD data from another dedicated stereotactic linac (Novalis Tx; Varian Medical Systems and BrainLAB AG). The Vero PDD are similar to that of another commercial unit routinely used for SABR, specifically the Novalis Tx, which has a PDD of 85.9% at 5 cm depth for a 10 × 10 cm 2 field relative to 86.2% for the Vero. (35) The Vero reference PDD (100 cm SSD, 10 × 10 cm 2 field size, 10 cm depth) is 66.7%.

A. Beam data
Orthogonal profiles for a 10 × 10 cm 2 field measured at depths of 1.5, 10.0, and 20.0 cm are shown in Fig. 3. Data are normalized to 100% on the central axis at a depth of 1.5 cm. Profile characteristics, FWHM, flatness, symmetry, and penumbra are summarized in Table 4; data from an earlier characterization of the Vero are also included. (17) As expected, penumbra in the crossline (left-right) direction is slightly larger, due to the rounded leaf ends. Flatness and symmetry are within acceptable ranges, and penumbra is consistent with that of other stereotactic linacs. Glide-Hurst et al. (42) have reported penumbra values of 7.67 and 6.82 mm in the inline and crossline directions, respectively, for the Millennium 120 leaf MLC (5 mm leaf width) on a TrueBeam linac. In that work, profile measurements were performed using a 0.13 cc chamber, which the authors previously reported resulted in a blurring of the penumbra due to the relatively large diameter of the sensitive volume. (35) Yin et al. (43) reported penumbra values ranging from 2 mm to 3.5 mm for field sizes of 6 mm to 100 mm for the 52 leaf Novalis MLC; measurements were performed at a depth of 7.5 cm and an SSD of 92.5 cm. The authors subsequently reported a penumbra of 5.3 mm for a 10 × 10 cm 2 field for the HD-120 MLC on a Novalis TX linac. (44) Diagonal profiles for a 15 × 15 cm 2 field measured for the Vero at depths of 1.5, 10.0, and 20.0 cm are shown in Fig. 4, with pinpoint ionization chamber and diode data superimposed for comparison. Finally, in-air profiles are shown in Figs. 5(a) and (b) for field sizes ranging from 2 × 2 to 15 × 15 cm 2 ; data for the 10 × 10 and 15 × 15 cm 2 fields are normalized to 80% and 60% of the maximum dose, respectively, to simplify visual analysis.
Scatter factors obtained using both a pinpoint chamber and diode are listed in Table 5. There is good agreement between the two detectors at all field sizes, except 1 × 1 cm 2 , where partial irradiation of the chamber volume produces an artificially low reading. Figure 6 shows the MLC leakage profile normalized to the open field central axis; interleaf leakage varies from approximately 0.08 to 0.35 percent. The average leakage over the entire field is 0.13 percent, which is similar to a value of 0.11 percent reported previously. (17) Fig. 3. Inline profiles (a) and crossline profiles (b) at three depths in water, normalized to d max . Table 4. Summary of beam profile characteristics from the current study, compared with those of a previous work (Nakamura et al. (17) ).    Table 6 shows the comparison of measured and calculated depth dose for square fields from 2 × 2 to 10 × 10 cm 2 . Agreement is within 1% for all field sizes and all depths. A summary of commissioning plans comparing calculation with point dose measurements in Solid Water is presented in Table 7. The average difference is 0.16% across a wide range of field sizes, irradiation techniques, and prescribed/delivered doses. The maximum difference of 4.17% was observed in a spine SBRT case, planned and delivered using IMRT. In the Solid Water phantom, PB and MC algorithms agreed within 0.5%. The anthropomorphic lung phantom contains two unit density targets, a "small" target 18 mm in diameter, and a "large" target 38 mm in diameter ( Fig. 1(b), lower left). Each target is completely surrounded by lung material with an average relative electron density of 0.24. A summary of commissioning plans and point-dose measurements for both targets, utilizing a variety of prescribed doses and irradiation techniques, is shown in Table 8. Calculations were performed using both the PB and MC algorithms. Overall, the MC algorithm demonstrates much better agreement with measurement than the PB algorithm, -0.11% versus -4.62% on average, respectively. The PB algorithm consistently overestimates dose by up to 12%, with the greatest disagreement observed for the small target (and small field sizes). Figure 7 shows treatment plans delivered as part of RPC thorax-lung, spine, and head and neck IMRT phantom credentialing. Each of the phantom irradiations passed the credentialing process on the first attempt; results are summarized in Table 9.  It should be emphasized that both the anthropomorphic lung and RPC phantoms were localized using image guidance. In this manner, dosimetric evaluation is performed in an end-to-end manner that precisely mimics the clinical treatment process.

C. Localization accuracy
Results of the 28 independent hidden-target tests are summarized in Table 10. Overall localization accuracy, characterized by one standard deviation, is 0.34 mm, 0.27 mm, and 0.27 mm in anterior-posterior, lateral, and superior-inferior directions, respectively. Maximum target errors are less than 1 mm in every direction. No significant differences were observed between CBCT and stereo X-ray image guidance.

d. ongoing system performance
Safe and effective SABR delivery depends on accurate image guidance. Essential QA of IGRT systems includes regular assessment of the coincidence of the lasers, kV imaging, and MV delivery systems. Table 11 shows the offset between Vero lasers and the kV imaging system axis, measured monthly from installation to present. Agreement is within nationally accepted tolerance for stereotactic procedures (≤ 1 mm), (12) though a small systematic effect (0.2-0.3 mm) can be observed in the lateral and vertical directions. Because localization lasers are mounted on the Vero gantry and rotate with the gantry and ring, the procedure for realigning the lasers is rather labor intensive. We have adopted an institutional tolerance of 0.5 mm for realignment of the lasers. Similarly, Table 12 shows the offset between kV imaging and MV delivery axes, measured monthly from installation to present. In this case, ΔX and ΔY refer to the in-plane deviation of the radio-opaque ball from the center of the 2 × 2 cm 2 MV field; that is, the orientation of X and Y with respect to the treatment coordinate system changes with the ring and gantry angle. Again, observed average displacements are all on the order of 0.1-0.2 mm, with standard deviations of the same magnitude. Interestingly, there is small dependence of X offset with gantry angle; no such dependence is observed with respect to either ring angle or in the Y direction. The largest deviation observed in any direction is 0.4 mm. The machine itself has an interlock that prevents operation if ΔX or ΔY exceeds 0.5 mm. To date the kV/ MV coincidence has remained stable, while the lasers have been adjusted once. Vero dosimetric performance measured daily since the time of installation is summarized in Table 13. All parameters are well within acceptable levels and have remained remarkably stable; no trends have been observed.  Figure 8 shows a treatment plan for a patient with a small (8-10 mm) peripheral lung metastasis. A dose of 54 Gy delivered in 3 fractions was prescribed such that at least 95% of the PTV received 54 Gy, and 100% of the PTV at least 51.3 Gy. Ten conformal beams were used, including six with a nonzero ring angle (i.e., noncoplanar); this produces a dose distribution that is compact and relatively isotropic in all directions. Based on previous commissioning results, the Monte Carlo algorithm was used for dose calculation. DVH for MC and PB calculations are shown for comparison. The PB algorithm predicts a dose more than 25% greater than that calculated using MC; that is, use of the PB algorithm in this case would result in a severe underdose. Figure 9 shows a treatment plan for a patient with metastasis to the thoracic spine. Treatment was delivered in a single fraction, using 13 coplanar IMRT beams distributed uniformly over 360°. A dose of 20 Gy was prescribed to the GTV, and 14 Gy to the CTV which encompassed the entire vertebral body, spinous process, and both transverse processes. As such, the CTV completely encircled the spinal cord. The plan shows excellent coverage at both dose levels and excellent sparing of the spinal cord, which received a maximum dose of 9.5 Gy. As part of patient specific QA, the measured dose distribution is shown superimposed on calculation with resulting 2D gamma results.

IV. dISCuSSIon
This study summarizes the commissioning process and initial clinical experience of a new treatment device developed for delivery of image-guided stereotactic ablative radiotherapy (SABR). Despite a completely new linac and new treatment head design, Vero beam data characteristics are remarkably similar to those of other linacs operating at 6 MV. (38,(42)(43)(44) Agreement between dosimetry measurement and calculation is exceptionally good, both in Solid Water and in a heterogeneous thorax phantom.
The shortcoming of PB algorithms to accurately predict dose in low density media has been described previously. (27,32,(45)(46)(47)(48)(49) In a recent study of 304 irradiations performed at 221 different institutions, Kry et al. (49) observed an overestimate of 4.9% in PB algorithms compared to measurement in the Radiological Physics Center (RPC) thorax phantom used for RTOG credentialing; these results agree favorably with those reported here.
Recent guidance documents emphasize the importance of end-to-end testing that evaluate both IGRT localization and dosimetry aspects in an integrated manner. (13,14,50) Several groups have evaluated positioning accuracies of independent IGRT methodologies utilizing floor mounted stereotactic X-ray localization and gantry-mounted volumetric (CBCT) imaging. Ma et al. (51) reported residual RMS errors from 0.17 to 0.30 mm (σ from 0.17 to 0.29 mm) for stereotactic X-ray localization and 0.35 to 0.50 mm (σ from 0.33 to 0.48 mm) for CBCT localization. Kim et al. (38) reported average errors from -0.2 to -0.8 mm (σ from 0.2 to 0.4 mm) for stereotactic X-ray localization and 0.0 to 0.6 mm (σ from 0.5 to 0.7 mm) for CBCT localization. While wider error distributions were observed for CBCT localization in both studies, differences between the two methodologies did not reach statistical significance. In this study, stereotactic X-ray and CBCT localization accuracy was assessed through 28 independent hidden-target tests in special head and thorax phantoms. No systematic errors were observed with either methodology, and error distributions were similar to those reported in the earlier studies.

V. ConCLuSIonS
Vero characteristics and performance are well suited for stereotactic-ablative radiotherapy application in which target localization and dose conformality and compactness are paramount. The ring gantry design provides unique degrees of freedom for beam delivery, which is important for creating the isotropically conformal and compact dose distributions that are needed in SABR, and exhibit high mechanical stability that contributes to excellent localization accuracy. Stereotactic X-ray and CBCT imaging are tightly integrated with delivery systems and can facilitate sophisticated treatment techniques, such as gimbal tracking. We have demonstrated excellent agreement between iPlan calculations and measurement, through systematic benchmarking in Solid Water, specialized anthropomorphic phantoms, and the RPC credentialing phantoms. To date, Vero has been used for SABR in lung, liver, spine, and other tumor sites. Treatments require between 30 and 60 minutes, depending on the dose, treatment technique, and number of intrafraction imaging sessions. Potential shortcomings of the Vero include a relatively low dose rate, 500 MU/minute, which can extend treatment times, and 0.5 cm wide MLC leaves, which may preclude optimal field shaping for very small tumors.