Development and long‐term stability of a comprehensive daily QA program for a modern pencil beam scanning (PBS) proton therapy delivery system

Abstract Purpose The main purpose of this study is to demonstrate the clinical implementation of a comprehensive pencil beam scanning (PBS) daily quality assurance (QA) program involving a number of novel QA devices including the Sphinx/Lynx/parallel‐plate (PPC05) ion chamber and HexaCheck/multiple imaging modality isocentricity (MIMI) imaging phantoms. Additionally, the study highlights the importance of testing the connectivity among oncology information system (OIS), beam delivery/imaging systems, and patient position system at a proton center with multi‐vendor equipment and software. Methods For dosimetry, a daily QA plan with spot map of four different energies (106, 145, 172, and 221 MeV) is delivered on the delivery system through the OIS. The delivery assesses the dose output, field homogeneity, beam coincidence, beam energy, width, distal‐fall‐off (DFO), and spot characteristics — for example, position, size, and skewness. As a part of mechanical and imaging QA, a treatment plan with the MIMI phantom serving as the patient is transferred from OIS to imaging system. The HexaCheck/MIMI phantoms are used to assess daily laser accuracy, imaging isocenter accuracy, image registration accuracy, and six‐dimensional (6D) positional correction accuracy for the kV imaging system and robotic couch. Results The daily QA results presented herein are based on 202 daily sets of measurements over a period of 10 months. Total time to perform daily QA tasks at our center is under 30 min. The relative difference (Δrel) of daily measurements with respect to baseline was within ± 1% for field homogeneity, ±0.5 mm for range, width and DFO, ±1 mm for spots positions, ±10% for in‐air spot sigma, ±0.5 spot skewness, and ±1 mm for beam coincidence (except 1 case: Δrel = 1.3 mm). The average Δrel in dose output was −0.2% (range: −1.1% to 1.5%). For 6D IGRT QA, the average absolute difference (Δabs) was ≤0.6 ± 0.4 mm for translational and ≤0.5° for rotational shifts. Conclusion The use of novel QA devices such as the Sphinx in conjunction with the Lynx, PPC05 ion chamber, HexaCheck/MIMI phantoms, and myQA software was shown to provide a comprehensive and efficient method for performing daily QA of a number of system parameters for a modern proton PBS‐dedicated treatment delivery unit.

QA of a number of system parameters for a modern proton PBS-dedicated treatment delivery unit. now configured with a more advanced PBS beam delivery technique that has been shown to deliver a more conformal dose when compared to DS/US techniques. 2 However, PBS proton beam delivery has uncertainties associated with its spot size and spatial position.
Such demand for advanced PBS delivery warrants a comprehensive understanding and monitoring of PBS beam characteristics. In addition to advances in proton beam delivery, image guidance used for proton treatments has also evolved in recent years. In the past, the primary imaging modality in proton centers had been planar kV x-ray technique. Newer proton centers are incorporating imaging modalities such as cone-beam computed tomography (CBCT) and surface imaging (SGRT). Due to the increase in delivery complexity of PBS and the use of multi-modality image-guidance systems for patient treatments, there is a need to establish a comprehensive daily quality assurance (QA) program that assesses safety, mechanical, dosimetric, and imaging parameters to ensure safe radiation deliverysimilar to the recommendations set forth by AAPM TG-142 for photonbased delivery systems. Additionally, certain proton centers may employ multi-vendor hardware and software for daily patient treatment. For these centers, interconnectivity becomes a critical element to assess and testing of data transfer among different softwares (e.g., beam delivery, imaging, record, verify systems, etc.) should be integrated as part of the daily QA program.
Several authors have published on proton daily QA using either commercial or in-house developed devices. [3][4][5][6][7][8][9] Arjomandy et al. 3  Since the DQA-3 was originally designed for photon and electron daily QA, authors 4,5 manufactured an in-house phantom to use with the DQA-3. Actis et al. 6 published on PBS daily QA in 2017 utilizing an in-house developed phantom that can accommodate multi-leaf ionization chamber (MLIC). Actis et al. 6

| MATERIALS AND METHODS
Our proton center is structured as a multi-vendor hardware and software platform environment. PBS proton plans are generated in RayStation (v.6.1.1.2; RaySearch Laboratories, Stockholm, Sweden), whereas ARIA (v.13.7; Varian Medical Systems, Palo Alto, CA) is used as the department record and verify system. IBA (Ion Beam Applications, Louvain-la-Neuve, Belgium) provides the ProteusPLUS PBS proton therapy system, which includes adaPT-Deliver (v.11.0.3) for beam delivery and adaPT-Insight (v.2.1.0d) for imaging (kV-kV x ray and kV-CBCT). Additionally, the CatalystPT (C-RAD, Uppsala, Sweden) system is used for surface imaging and gating applications.
The flow chart of data transfer among the various software entities is presented in Fig. 1.

2.A | Beam delivery system (BDS)
A PBS proton beam is delivered using a PBS dedicated nozzle ( Fig. 2). As the proton beam enters the nozzle, an ionization chamber 1 (IC1) verifies the alignment of the beam at the nozzle entrance. A set of two focusing quadrupole magnets focus the proton beam at the isocenter. The proton beam is then scanned in Y direction by a vertical scanning magnet followed by scanning in X direction with a horizontal scanning magnet. In order to direct the beam to a particular location on a target, the beam position is steered using magnetic fields. Ionization chambers 2 and 3 (IC2/3) monitor beam characteristics real-time (beam size, position, and flatness) and dose just before the proton beam exists the nozzle. Snout holder allows the movement of accessary drawer, which can include an optional range shifter (pre-absorber) and snout. At our center, a range shifter of 7.5 cm water equivalent thickness is used for clinical cases as necessary.

2.B | Imaging systems
The kV x-ray imaging system includes two gantry mounted, x-ray tubes that rotate with the gantry. The first x-ray tube (portal) is located in the PBS dedicated nozzle pre-assembly, which is under vacuum. The x-ray tube is retracted from the beam line during the proton beam irradiation. The flat panel detector of portal (SAD = 119.4 cm, SID = 177.0 cm, active pixel area = 28.2 cm × 40.6 cm, and active pixel resolution: 2232 × 3200 pixels) is located in front of the nozzle. The second x-ray tube (orthogonal) is fixed to one of the gantry structural beams. The x-ray beam axis is perpendicular to the proton beam axis and to the gantry rotation axis. The orthogonal tube in conjunction with its flat panel detector (SAD = 264.2 cm, SID = 317.1 cm, active pixel area = 43 cm × 43 cm, and active pixel resolution: 2874 × 2840 pixels) is used for the kV-CBCT acquisition.
In addition to the x ray based imaging system, the CatalystPT, a three-camera surface imaging system, is used to setup patients prior to x ray based imaging, monitor patient position and posture during treatment, and enable beam gating. The three cameras are positioned to maximize field coverage with the outer cameras being 43°f rom the center camera.
ARIA also receives the treatment record from adaPT Deliver and images (kV planar/CBCT images) from adaPT Insight.

2.D | Phantoms and detectors
The Sphinx phantom has a carbon frame with dimension of 540 mm × 400 mm × 400 mm (Fig. 3). The carbon frame contains the markers for verification of laser alignment. The phantom incorporates four wedges with various thicknesses for verifying the con- The four wedges are utilized to calculate the energy related parameters such as range, width, and distal-fall-off (DFO). The energy calculation algorithm 10 implemented within myQA software, version 2017-002 (2.9.23.0) calculates the slight signal generated by the radiation delivered over the RW3 wedge. The first derivative of F I G . 1. Flow chart of data transfer among RayStation, ARIA, adaPT-Deliver (beam delivery), and adaPT-Insight (imaging) in a ProteusPLUS pencil beam scanning proton therapy system. this rising part of signal is then calculated in order to identify the physical edge of the corresponding RW3 block. 10 The final depthdose curve is calculated by assigning a value of depth to each pixel of the image. 10 The values of depths are extrapolated from data interpolated with a cubic spline fit. 10 For better understanding on the range calculation using wedge, readers are advised to refer to work published by Shen et al. 11 and Deng et al. 12 The phantom also has an insert containing a pin with a fiducial at its tip which is placed at the isocenter (Fig. 3). A dedicated RW3 insert (160 mm × 90 mm × 100 mm) contains a notch for a PPC05 chamber for dose output constancy check. The PPC05 is covered with 3 cm thickness RW3 block so the chamber has a 3 cm build up

2.E | Workflow
Our current daily QA workflow includes two daily QA plans based on two sets of devices: (a) Sphinx, Lynx, and PPC05 and (b) MIMI and HexaCheck.

2.E.1 | Sphinx, Lynx, and PPC05
A daily QA plan was generated in RayStation (v.6.1.1.2) with spot map of four different energies (Fig. 4). In order to mimic patient treatment, a daily QA plan is delivered using adaPT-Deliver on Pro-teusPLUS proton therapy system through ARIA. Dosimetry measurements are performed using a single couch top setup with the Sphinx, Lynx, and PPC05 chamber ( Fig. 5) For PBS daily QA dosimetric quantification, tests (Table 1)  using Sphinx, Lynx, and PPC05 is presented in Fig. 6. The total time for this workflow is from 15 to 20 min without system interruptions.

2.E.2 | MIMI and HexaCheck
A treatment plan with kV-kV and CBCT setup fields was generated in where, p = translational (e.g., lateral) or rotational parameter (e.g.,

3.C | Spots characteristics
Spots characteristics (position, size, and skewness) were evaluated for four spots (106, 145, 172, and 221 MeV). 3.D | X-ray vs proton beam coincidence Table 2 and Fig. 13 show that the Δ in x-ray and proton beam coincidence (X and Y directions) was within ±1 mm except in one case (Δ = 1.3 mm). The 3σ of beam coincidence was ±0.7 mm in X and ±0.5 mm in Y directions (Table 2).  (Table 4).  Conceptionally, in attempting to establish tolerances for specific tests of a quality assurance program, a number of strategies may be employed to determine the tolerance action value. One such approach is to follow the recommended tolerances established by published guidelines that were conceived by the consensus of a group of experienced usersthat is, for example, an AAPM task group report. A second could be to evaluate the impact on the patient dose distribution due to variations in that specific parameter. For example, there have been publications characterizing the impact of spot size on treatment plan quality. 15 A third approach is to use statistical process control to evaluate whether specific parameters are behaving in a stable and controlled manner. 16,17 With this information, it is possible to use statistical methods to determine the system specific action level tolerances due to the system performance. In using statistical process control, the methodology is to first establish a process of testing a parameter, test and observe, and characterize the behavior of the specific parameterfor example, dose output, spot size, spot position, etc.over a time period. By characterizing the behavior, it is possible to determine when a parameter is out of control and is statistically an outlier. This helps provide guidance as to when to act. In this study, 10 months' worth of data was collected to characterize the behavior of our proton PBS delivery system. Our goal was to measure the stability of multiple parameters and establish tolerances based on our specific system performance and not on gen-  couch. For our current daily QA protocol and workflow (Fig. 6), a tighter spot position tolerance of ±0.6 mm is feasible.

3.E | Translational and rotational shifts
For x-ray and proton beam coincidence, we currently use a single spot of energy 106 MeV. Based on 10 months results, the deviation in coincidence was found to be within ±1.5 mm, which was used as the tolerance by Lambert et al. 5 The 3σ of beam coincidence (x ray and proton) was found to be ±0.7 mm in X and ±0.5 mm in Y directions. As shown in Fig. 6, we use the setup field to drive the 6D T A B L E 2 Results of dose output, field homogeneity, range, width, distal-fall-off (DFO), and x-ray vs proton beam coincidence based on daily QA measurements (n = 202).
A relative difference (Δ) was calculated by comparing daily (D) measurements against baseline (B) measurements. Upper control limit (UCL) and lower control limit (LCL) are based on statics process control (SPC) charts. UCL = +3σ and LCL = -3σ are from the average value.  Spot skewness-X A relative difference (Δ) was calculated by comparing daily (D) measurements (n = 202) against baseline (B) measurements. Upper control limit (UCL) and lower control limit (LCL) are based on statics process control (SPC) charts. UCL = +3σ and LCL = −3σ are from the average value.
includes both the translational and rotational shifts. Based on 202 sets of measurements, the 3σ of translational shifts (lateral, longitudinal, and vertical) ranged from ±0.6 to ±0.8 mm, and the 3σ of rotational shifts (pitch, roll, and yaw) ranged from ±0.2°to ±0.3°.
The variation in daily 6D correction vector in our current daily QA setup is found to be mainly due to the combination of (a) repro-  Specifically, the C-RAD daily QA phantom is aligned to the room isocenter using the room/gantry lasers. Once positioned, the daily QA phantom is imaged, and the agreement between the laser isocenter and surface imaging isocenter is quantified. | 41 proton beam is tested by Lambert et al. 5 only, and in-air spot size is reported by Lambert et al., 5 Actis et al., 6 and Bizzochi et al. 7 Furthermore, none of the studies 4-8 reported the CBCT acquisition and its functionality as a part of daily QA. This could be due to unavailability of CBCT in the treatment room or difference in daily QA policies at the authors' institutions. [4][5][6][7][8] The inclusion of CBCT in our daily QA workflow (Fig. 8) has certainly contributed about 5 min toward the total daily QA time at our center. In addition of F I G . 1 3 . X-ray vs proton beam coincidence in x and y directions. A single spot of 106 MeV was used for the coincidence.
F I G . 1 4 . Difference between translational/rotational correction vectors and known offset (baseline) values for subsequent kV-cone-beam computed tomography imaging daily quality assurance measurements (n = 202).
calculating the 6D correction vectors by using CBCT, this test allows us to test the functionality of x-ray tube, collision detection software, and adaPT Insight as well as transfer of CBCT images to the OIS for offline review.

| CONCLUSION
With the increasing complexity of delivery, patient positioning, and imaging systems, a robust and comprehensive daily QA program is required to gain confidence in the performance of a proton therapy system. The use of novel phantoms and dosimetry devices such as the Sphinx in conjunction with the Lynx and HexaCheck/MIMI was shown to provide a robust, consistent and efficient method of evaluating various aspects of our delivery system which include PBS beam parameters and imaging/couch accuracy. Our daily QA results from over 10 months demonstrate consistent beam stability of the ProteusPLUS PBS proton therapy system. If CBCT is available, it is recommended to test its functionality on a daily basis mimicking a patient treatment scenario. The use of MIMI/HexaCheck can serve an accurate and efficient tool to perform daily, 6D IGRT QA of the IBA adaPT Insight software and LEONI robotic couch.

ACKNOWLEDG MENTS
The authors thank Victor Chirinos, Michael Leyva, and Radiation Therapists at the Miami Cancer Institute for their assistance in this project. The authors also thank IBA dosimetry team as well as Marc Blakey from Provision CARES Proton Therapy, Nashville for their helpful suggestions on Sphinx setup.

CONFLI CT OF INTEREST
The authors declare no conflict of interest.