Commissioning of the world's first compact pencil‐beam scanning proton therapy system

Abstract This paper summarizes clinical commissioning of the world's first commercial, clinically utilized installation of a compact, image‐guided, pencil‐beam scanning, intensity‐modulated proton therapy system, the IBA Proteus® ONE, at the Willis‐Knighton Cancer Center (WKCC) in Shreveport, LA. The Proteus® ONE is a single‐room, compact‐gantry system employing a cyclotron‐generated proton beam with image guidance via cone‐beam CT as well as stereoscopic orthogonal and oblique planar kV imaging. Coupling 220° of gantry rotation with a 6D robotic couch capable of in plane patient rotations of over 180° degrees allows for 360° of treatment access. Along with general machine characterization, system commissioning required: (a) characterization and calibration of the proton beam, (b) treatment planning system commissioning including CT‐to‐density curve determination, (c) image guidance system commissioning, and (d) safety verification (interlocks and radiation survey). System readiness for patient treatment was validated by irradiating calibration TLDs as well as prostate, head, and lung phantoms from the Imaging and Radiation Oncology Core (IROC), Houston. These results confirmed safe and accurate machine functionality suitable for patient treatment. WKCC also successfully completed an on‐site dosimetry review by an independent team of IROC physicists that corroborated accurate Proteus® ONE dosimetry.

delivering both single-field uniform dose and multiple-field intensity modulated proton therapy without the need for compensators or apertures. Scanning beam systems have been developed by Hitachi, IBA, and other companies . [1][2][3] At the Willis-Knighton Cancer Center (WKCC) in Shreveport, LA, the world's first commercial, compact, image-guided, pencilbeam scanning proton therapy system, the IBA Proteus â ONE, has been installed and commissioned. The system began treating patients in September 2014 and to-date has treated more than three hundred and thirty patients. The Proteus â ONE offered easy

2.A | Proteus â ONE overview
The IBA Proteus â ONE features an isochronous cyclotron, 220°partial-rotation compact gantry, scanning beam delivery nozzle, image guidance system with cone-beam CT and stereoscopic imaging capabilities, and a 6D robotic couch. Continuous dynamic spot scanning of the cyclotron-generated proton beam, coupled with rapid adjustment of beam energy, is used to treat three dimensional target volumes. Discrete dose deposition layers within a target, ranging from surface to 32-cm water equivalent thickness (WET), are achieved via adjustment of beam energy by a degrader and energy selection system (ESS) located between the cyclotron and PBS delivery nozzle.
Within each layer, scanning magnets direct discrete beamlets (spots) along the x-and y-directions yielding a maximum proton field size of 20 9 24 cm in x-and y-directions at isocenter. Although IBA Proteus â ONE hardware produces and supplies the proton beam, additional equipment from both IBA and other vendors is necessary to achieve patient treatment delivery. A third-party TPS computes required layer energies, along with spot locations and weights, based on patient morphology for a set of user-specified beam (gantry) and couch angles. These patient-specific plan parameters, along with reference image datasets, are then transferred to a third-party OIS, which is also used to transfer both plans and image datasets to IBA's adaPTdeliver TM console system and adaPTinsight TM imaging system software, respectively, for treatment delivery as shown in Fig. 1.
Accurate, seamless integration between these disparate systems is a crucial piece of the Proteus â ONE system.

2.A.1 | Accelerator
Unique to Willis-Knighton's Proteus â ONE installation is the use of an IBA C230 cyclotron, which is to be replaced with the IBA S2C2 super-conducting cyclotron at subsequent Proteus â ONE sites, to accelerate the protons. The C230 cyclotron is configured to produce a 230 MeV beam with an accelerating voltage frequency of 106.1 MHz. The range of extracted current is 1 to 300 nA with an extraction efficiency of 60% AE 10%. The ion source's turn on/off time is 15 ls with a 45 ls transit time from ion source to patient.
Range is modulated via an energy degrader composed of variable block thicknesses of beryllium, graphite, and aluminum that allows for beam energies of 70 MeV (4.1 g/cm 2 water) to 230 MeV (32.95 g/cm 2 water). Measurement of revolution frequency and orbit position during beam extraction from the cyclotron verifies proton energy. IBA determined that the beam orbit position within the cyclotron is within AE1 mm of optimal, assuring proton range accuracy to within 0.025 g/cm 2 .

2.A.2 | IBA-PBS delivery nozzle
Specifically designed for pencil beam scanning delivery, Proteus â ONE's PBS nozzle is diagramed in Fig. 2 4 Therefore, only the 4.1-cm range shifter is clinically commissioned with its effects on spot size considered in the beam modeling process.
In addition to the range shifters, circular Proteus â ONE snouts were provided in two diameters: 24.4 and 10.6 cm. These snouts, which are physical attachments that can be fitted on to the compact nozzle, would typically be used to place an additional range shifter and aperture closer to complex patient anatomy. However, because these optional attachments were not yet supported by the TPS at the time of commissioning, they were not included in the commissioning process. Currently, Proteus â ONE snouts are not clinically utilized at WKCC.

2.B | Characterization and calibration of the proton beam
Beam characterization included but was not limited to TPS mandates for beam modeling. The RayStation TPS required in-air spot profiles, integrated depth doses (IDD), and absolute dose per monitor unit measured at a depth between 1 cm and one-half of the Bragg peak maximum.

2.B.1 | Beam measurements
Utilizing an IBA Blue Phantom 2 , pristine Bragg peak beams were measured in water using a large-area Bragg peak (BP) chamber A series of simple spot positioning tests were also performed to verify spot placement accuracy. For the clinical range of beam energies, spots were placed AE5 and AE10 cm from the origin along the xaxis and AE6 and AE12 cm from the origin along the y-axis, and measured using the Lynx to quantitate the distances between the spots. This process was then repeated with spots placed along the diagonal.

2.B.2 | Absolute dose calibration
For passive scattering proton therapy systems, the relationship between absolute dose and monitor units can be established by the IAEA TRS398 protocol. 9 However, TRS398 does not adequately cover the dosimetry of PBS systems. 10,11 As such, a pencil beam scanning system requires a different setup for absolute dose per MU calibration. As there is, initially, no beam model for the TPS to use to create an SOBP, 32 individual, single-energy scanned fields were created as PLD (PBS Layers Definition) files, an IBA-specific format which allows the system to deliver pristine layers independently from the OIS. To fulfill RayStation beam modeling requirements, each PLD was defined with a field size of 10 9 10 cm 2 and 2.

2.B.3 | Variable virtual SAD (VSAD) measurement
The unique orientation of the scanning and bending magnets of the Proteus â ONE compact gantry causes the virtual source-to-axis distance to vary along the direction of the bending magnet (cranio-caudal with head-first supine patient position and treatment couch at 0°). A Lynx PT scintillator-based sensor was used to verify the magnitude and range of the VSAD via measurement of a pattern of spots (one central and two at each max deflection) at three vertical positions: isocenter, 16 cm from isocenter toward the nozzle, and 16 cm from isocenter away from the nozzle. Values for the VSAD along the bending magnet direction were then determined geometrically.
To evaluate clinical impact, Willis-Knighton collaborated with IBA and RaySearch to implement a simple variable VSAD model in RayStation. While the actual trend of VSAD is slightly parabolic, a linear fit of the measured data was adopted to approximate its effect in the TPS. Phantom treatment plans were generated at clinically relevant treatment depths for various sites (e.g., prostate, whole pelvis, and cranium). Calculated dose distributions with and without VSAD consideration were compared for each plan. | 97 midpoint of this range, for simplicity, and with input from RaySearch, we performed all absolute dose measurements at a depth of 2 cm from the water surface for each mono-energetic beam. Beam data obtained with the range shifter in place was not required by RayStation as its effect is modeled within the dose engine. All necessary beam data were supplied to RaySearch Laboratories who then generated the final beam model, the details of which are beyond the scope of this paper.

2.C.2 | CT-to-density curve determination
Our clinic utilizes a Phillips Brilliance Big Bore 16-slice CT scanner for patient simulation. Using Gammex 467 tissue characterization plugs inside acrylic phantoms, we studied various CT acquisition settings and phantom sizes to establish imaging protocols and create a CT-to-density table from which the TPS estimates relative stopping power ratios for proton beam dose calculations. 5,13,14 These protocols were tested in the TPS by comparing the mass densities determined by the protocols in patient CT datasets against reference ICRU 49 data for known human tissues. 5 Accurate dose calculation in proton therapy depends on proton relative stopping power ratios. RayStation uses an internal mass density to stopping power conversion during dose calculation. Individual CT voxels are assigned a known biological material based on mass density. The stopping power is then calculated on the fly using the density of the voxel, the properties of the known material (e.g., mean excitation potential and elemental composition) and the Bethe-Bloch equation. A stoichiometric calibration was also independently performed by an external proton physicist to verify stopping powers calculated by RayStation.

2.C.3 | TPS validation
Measurement of spot profiles in solid water, depth doses for inversely optimized plans, lateral dose profiles, dose uniformity, absolute dose, and patient treatment field specific QA in homogenous phantoms were used to both verify the new beam model and validate the TPS. Twenty-three different treatment plans generating cube patterns of uniform dose over the SOBP for varying field sizes, ranges (depths), and prescribed doses were produced in RayStation  beam CT. The adaPTinsight TM software application coupled with the Proteus â ONE's 6D robotic couch provides a streamlined, single-user interface for proton therapy patient positioning and treatment delivery enabling the user to perform image-guided proton therapy (IGPT). In all three modes (kV-kV planar, stereoscopic, or CBCT) the software allows for either manual or automatic 6D registration of the acquired images against a reference dataset. Image registration is based on mutual information between acquired and reference images and a correction vector is computed that can be applied via a virtual hand pendant on the adaPTinsight TM control station. During commissioning, standard tests of x-ray parameters and imaging quality, which will not be detailed herein, were performed for all modalities. Couch isocentricity and table sag were also measured and verified with each imaging system.

2.D.2 | Stereoscopic system characterization
Of primary concern during stereoscopic system characterization were both geometric accuracy and geometric integrity between the proton beam axis and stereoscopic imaging system axis. Geometric accuracy was evaluated by imaging a 5-cm diameter steel ball placed at isocenter using both detector panels and measuring the imaged ball diameter across multiple directions ranging from 0 to 350 degrees. Coincidence between radiation and stereoscopic isocenters was evaluated with radiochromic film and a scintillator-based detector. Using the stereoscopic imaging system, the steel ball was aligned at isocenter to within 0.2 mm as determined via isocenter comparisons on each oblique radiograph. A radiochromic film was then placed downstream of the ball as near as possible to isocenter and a circular spot pattern delivered. Analysis of the resulting concentric circles on the film was performed, with the Euclidian distance between the centers of both circles indicating the degree of radiation and stereoscopic collinearity. This procedure, including realignment of the steel ball with stereoscopic imaging, was repeated for multiple gantry angles and beam energies.
Isocentricity was also verified with the XRV100 scintillator (Logo System Intl, CA, USA), 7 which works on the principle of a hodoscope and can measure beam isocenter based on particle trajectory through the surface of a scintillator cone. The XRV100 was imaged with the CT scanner to create a reference image set, which was then transferred to adaPTinsight TM . The XRV100 was aligned in the treatment room by iteratively performing stereoscopic imaging and registration until residual corrections vectors were minimal. Isocentricity was then confirmed by exposing the XRV100 to single spot pristine beams from multiple gantry angles for a range of energies.

2.D.3 | Gantry mounted imaging system
The confirmation of radiation isocenter and stereoscopic imaging isocenter collinearity provided the ability to verify the gantrymounted imaging system (kV-kV orthogonal radiographs and conebeam CT) against the stereoscopic imaging system. A cube phantom with a central BB marker was first aligned to isocenter using stereoscopic oblique radiographs. Orthogonal kV-kV radiographs and both clockwise (CW) and counter clockwise (CCW) CBCTs of the cube phantom were then obtained and their respective registration shifts evaluated to confirm isocenter agreement between the stereoscopic and gantry-mounted imaging systems.  All energies exhibited good agreement between measured and pre-

3.B | Absolute dose calibration
Using the TRS398 protocol 9 and measurement of dose at the middle of the SOBP with depth of 15 g/cm À2 , the ratio of measured dose to TPS dose was found to be within AE0.5% for measurements inside the TPS-calculated volume both on and off central axis. Twenty measurements performed throughout the field at various depths confirmed TPS-predicted doses. Measured dose in the center of the volume at isocenter as compared to TPS calculation was within 0.2%. Additional plans and absolute dose measurements were performed as part of the TPS validation as described in Section 3.D.

3.C | CT-to-density table determination
Five patients (three pelvis and two prostate) were scanned and evaluated for adipose, muscle, and cortical bone physical density data as compared to the ICRU 49 protocol. 5 These density comparison results are summarized in Table 2 five patients mentioned previously and are presented in Table 2.
RLSP comparisons for the Gammex tissue characterization plugs are presented in Table 3.

3.D | TPS validation
TPS-calculated spot profiles were compared against measured data for a wide variety of clinical scenarios. Using the Lynx device, a total of 57 sets of measurements were obtained at various depths in solid water and proton ranges (energies), with and without the range shifter. Each measurement consisted of a layer containing 17 spots placed throughout the treatable field size (20 9 24 cm).
Relative gamma analyses between measured and TPS spot profiles were all greater than 95% passing with 3%/3 mm dose/distance agreement criteria. Measured and TPS spot sigmas all agreed within 0.5 mm for both x and y directions.
The distance-to-agreement (DTA) between TPS-calculated and Zebra-calculated ranges (R 90 ) was within 0.7 mm for all measured Pristine Bragg peaks. The SOBP plateau region is defined as that between the 50% point distally and the 98% proximally reduced by two distal fall-off widths (80%-20%) on each side.  Fig. 7 (a) through 7(c) were within 3% of the TPS calculated values, except when using a range shifter, as discussed in the next paragraph. Figure 7(d) shows that Lynx PT-measured relative planar doses at shallow, proximal-end, center, and distal-end depths on the SOBP agreed well with the TPS calculated dose distribution with gamma pass rates greater than 95% for all measured field sizes when using dose/distance agreement criteria of 3% and 3 mm.
A comparison between measured and TPS-calculated point doses at various depths as a function of the air gap between the range shifter and phantom is given in Fig. 8. A disagreement of approximately 6% (dose overestimation by TPS) was observed for a completely retracted range shifter at a proximal measurement depth of 1.5 cm. Relative depth dose comparisons for these test plans, normalized to the maximum SOBP dose, were also compared via Zebra TM measurements as shown in Fig. 9. The results yield a less than 1 mm difference at the distal 90% edge in all cases. However, as observed with absolute dose discussed above, measured relative depth dose profiles are lower than their TPS-calculated counterparts on the proximal edge when there is a large air gap. The reason for this disagreement will be discussed in Section 4.
Final dose validation results from our IROC anthropomorphic phantom irradiations are given in Table 4. All measurements met their respective IROC passing criteria.

3.E | Variable virtual SAD (VSAD) measurement
Evaluation of Lynx measurements indicated that the VSAD ranged  A workflow limitation of the current IBA Proteus â ONE system is that patient setup images are not managed by the OIS for acquisition of either "Setup" or "Port" films before beam delivery. Instead, they are managed by the adaPTinsight TM application, restricting the ability of both the physicist and physician to perform remote review of the images.
During the measurement process of integrated depth doses, there was initially some concern regarding incomplete charge collection by the 8 cm diameter PTW Bragg peak chamber. 1   Thus, increases in the penumbrae and the removal of dose from the central part of the beam is not accurately modeled". 14 While this is not typically an issue regarding inhomogeneities in the patient, it does become problematic for shallow treatment fields when there is a large air gap resulting in TPS doses that are inaccurately high. Consequently, a robust QA process has been implemented at Willis-Knighton Cancer Center to evaluate such dose disagreements between the TPS and measurement. Furthermore, WKCC takes care to reduce the air gap between patient surface and range shifter for targets of less than 4-cm depth, as preliminary data indicate that a 10-cm air gap yields 2% disagreement between TPS and measurement within the proximal portion of the SOBP.

| CONCLUSION
The first commercial, clinically utilized, compact, image-guided, pencil-beam scanning, intensity-modulated proton therapy system, the IBA Proteus â ONE, was installed at the Willis-Knighton Cancer Center (Shreveport, LA) in 2014. Major tasks associated with characterization and clinical commissioning of this machine included mechanical and radiation isocenter checks, radiation safety checks, beam dose/MU calibration, beam data collection for beam modeling in the RayStation treatment planning system (TPS), baseline data for periodic quality assurance checks, TPS dose calculation accuracy validation, imaging system functionality tests, and end-to-end tests for simulation, planning, and dose delivery with an anthropomorphic phantom. All of these tasks were successfully completed with results summarized in this paper. The WKCC Proteus â ONE machine was released for clinical use in September 2014 and a robust quality assurance program has been implemented to ensure safe and accurate proton therapy dose delivery to our patients. Over three hundred and thirty patients, including pediatric, with various treatment sites such prostate, breast, lung, head and neck, whole pelvis, abdomen, and brain have been treated at our center to date. In all, the Proteus â ONE has been found to be stable and reliable, delivering planned dose distributions to the patient's target volume within established tolerance limits.

ACKNOWLEDG MENTS
We wish to thank IBA for the collegial manner in which they approached this project and, particularly, all of the IBA personnel who were essential throughout the installation and commissioning process.  We also would like to thank Drs. Xiaofei Song, Leo Ding, Michele Zhang, Gwen Chen, and Mr. Jarron Syh for all of their assistance and expertise, without which we could not have completed this project.

CONF LICT OF I NTEREST
No conflict of interest.