Beam commissioning of the first compact proton therapy system with spot scanning and dynamic field collimation

Abstract

Objectives:

To describe the measurements and to present the results of the beam commissioning and the beam model validation of a compact, gantry-mounted, spot scanning proton accelerator system with dynamic layer-by-layer field collimation.

Methods:

We performed measurements of depth dose distributions in water, spot and scanned field size in air at different positions from the isocenter plane, spot position over the 20 × 20 cm² scanned area, beam monitor calibration in terms of absorbed dose to water and specific field collimation measurements at different gantry angles to commission the system. To validate the beam model in the treatment planning system (TPS), we measured spot profiles in water at different depths, absolute dose in water of single energy layers of different field sizes and inversely optimised spread-out Bragg peaks (SOBP) under normal and oblique beam incidence, field size and penumbra in water of SOBPs, and patient treatment specific quality assurance in homogeneous and heterogeneous phantoms.

Results:

Energy range, spot size, spot position and dose output were consistent at all gantry angles with 0.3 mm, 0.4 mm, 0.6 mm and 0.5% maximum deviations, respectively. Uncollimated spot size (one sigma) in air with an air-gap of 10 cm ranged from 4.1 to 16.4 mm covering a range from 32.2 to 1.9 cm in water, respectively. Absolute dose measurements were within 3% when comparing TPS and experimental data. Gamma pass rates >98% and >96% at 3%/3 mm were obtained when performing 2D dose measurements in homogeneous and in heterogeneous media, respectively. Leaf position was within ±1 mm at all gantry angles and nozzle positions.

Conclusions:

Beam characterisation and machine commissioning results, and the exhaustive end-to-end tests performed to assess the proper functionality of the system, confirm that it is safe and accurate to treat patients.

Advances in knowledge:

This is the first paper addressing the beam commissioning and the beam validation of a compact, gantry-mounted, pencil beam scanning proton accelerator system with dynamic layer-by-layer multileaf collimation.

Introduction

Proton therapy is an established radiation treatment modality that uses proton beams to treat cancer. There is strong interest in proton therapy worldwide; however, the number of new proton centres grew slowly over the last two decades due to the high facility and treatment costs, and the limited evidence of proton therapy’s superior effectiveness with respect to photon therapy.^1–3 The Netherlands implemented a model-based patient selection system⁴ that selects only those patients who would really benefit from proton therapy. This approach generates useful data to determine the clinical evidence of proton therapy. In the last years, proton accelerator vendors focused in reducing the weight and dimensions of the accelerator module by making use of superconducting magnets.^5,6 A consequence of the more compact design are the reduced acquisition and operational costs of the proton facility.

With only 15 tons and 1.8 m of diameter, the Mevion S250i Hyperscan system is a small superconducting synchrocyclotron, which requires a single vault of limited dimensions. Maastro Proton Therapy in Maastricht (the Netherlands) is integrated in an already existing photon radiotherapy department and it is the first proton facility in Europe equipped with this proton accelerator system. The system can deliver high-energy proton beams using spot-by-spot pencil beam scanning. A 3D dose distribution is generated by superimposing the individual spots’ dose contribution in the transverse plane and in depth. In the transverse plane, for every isoenergetic slice, corresponding to a specific depth, narrow pencil beams are magnetically scanned across the target volume (in x and y directions) by a single focus scanning magnet. A special feature of this system is the possibility to dynamically collimate each treatment field and each energy layer to sharpen the beam penumbra. We report the methodology, the results of the beam commissioning and the beam model validation of this proton system.

Methods and materials

The first section provides a description of the main parts of the accelerator and an overview of the systems present in the proton facility. The second section describes the beam commissioning measurements, including the required data to create the beam model in the treatment planning system (TPS). The TPS manufacturer modelled the beam using their Monte Carlo (MC) engine, thus we mainly focused on beam model validation, which is presented in the third section.

Mevion S250i system

The Mevion S250i Hyperscan system consists of two gantries, the so-called inner and outer gantries (Figure 1a). On the outer gantry are mounted the superconducting synchrocyclotron (10 T) with the ion source and the scanning magnet. On the inner gantry, which is visible from the treatment room (Figure 1b), are mounted the beam monitor (BM) system, the range modulation system (RMS), and the multileaf collimation system. These three components form the nozzle of the device. The inner and the outer gantry rotate independently from each other at different speeds and, to deliver beam, the alignment between them is submicrometric.

Figure 1.

Sketch of the proton accelerator system (a), and treatment room of the Mevion S250 Hyperscan system in Maastricht (b). ACP: Adaptive Coil Positioner, RMS: Range Modulation System, AA: Adaptive Aperture. A: kV-kV flat panels, B: four camera surface guidance system, C: extendable nozzle and exit entrance window (black square), D: multi-energy cone-beam CT system, E: six degrees of freedom robotic couch.

Proton beams are produced in the cold cathode penning ion source, i.e., H₂ gas, located in the middle of the accelerator module. The system accelerates a fixed energy beam of 227 MeV and extracts it towards the single treatment room. The treatment line is equipped with a dose delivery system (DDS) and an extendable nozzle at the end of the beam line. The DDS consists of three elements: a thin quadrant foil position detector, the beam scanning magnet and six transmission ionisation chambers (ICs) or BMs. The vacuum window is located immediately after the scanning magnet, at ~2 m from the room isocenter. Two independent dosimetric systems monitor in real time and redundantly the number of particles and the beam position. They consist of two wide integral ICs and two pairs of strip chambers each orthogonal to the other. The beam scanning magnet deflects proton beams across a 20 × 20 cm² scanned area at the isocenter. The BMs measure beam spot position, spot shape, spot charge and provide feedback to the system. Proton beams are pulsed, with a maximum of 8 pC per spot which results in a time-averaged beam current of 4 nA. To obtain clinically relevant energies, which cover a range from 0 cm up to 32.2 cm in water, the 227 MeV beam is degraded by the RMS, mounted on the nozzle. The RMS consists of 18 Lexan (1.20 g/cm³) plates of different thicknesses whose combinations allow the generation of 161 energies with 2.1 mm step in water equivalent thickness, and an energy switching time of ~50 ms. The minimum energy available at the isocenter is 13.49 MeV and allows treating the patient surface without making use of any additional range shifter. The nozzle shape is rectangular, i.e. 43 x 67 cm, and its position, defined as the distance from the more downstream element (i.e., exit window) to the isocenter, is adjustable and it can vary from 3.6 to 33.6 cm. At the end of the beam line, mounted on the extendable nozzle, there is a dynamic field collimation system, named adaptive aperture (AA), to reduce the lateral penumbra, especially at lower energies. The AA module consists of two carriages of five leaves, 5 mm thick, and one top and one bottom jaw, 20 mm thick, made of nickel-alloy (Figure 2). The dimensions in the beam direction of the leaves and jaws are 10 cm. The AA module can trim laterally the spots over a 20 × 20 cm² area at the isocenter plane with a maximum speed of 1.2 s per spot. The design of the leaves takes into account interlocking features meant to prevent leakage between the leaves. One condition, implemented in the TPS, is that the centroid of a spot cannot be placed on top of any collimating element, thus we trim at most half spot. The gantry system can rotate from 355° to 185° with an angular accuracy of ±0.25°. The 360° beam entrance coverage is achieved by rotating the six df robotic couch (RoboCouch, Accuracy, Sunnyvale CA, USA), which provides submillimetric patient positioning accuracy. The treatment room in Figure 1b is equipped with 2D and 3D imaging. On the left side of the treatment room ceiling are mounted on rails two orthogonal amorphous silicon flat panels and can be deployed for 2D imaging. The corresponding X-ray sources are positioned on the opposite wall and under the treatment couch for lateral and postero-anterior exposure, respectively. On the right side of the treatment room, also mounted on rails, there is a multienergy CBCT imaging system (Imaging Ring, medPhoton, Salzburg, Austria) for in-room 3D imaging. The 2D imaging is performed at the treatment isocenter, while 3D imaging implies a patient translation of 50 cm in the longitudinal direction because the imaging and the treatment isocentra are 50 cm apart. Daily CBCT is performed for all indications. The treatment room is also equipped with a surface imaging system consisting of four cameras (C-RAD, Uppsala, Sweden) to monitor patients during imaging and treatment. Two CT systems are available at the radiotherapy department for photon and proton treatment planning: Siemens Confidence and Siemens Drive (Siemens Healthineers, Erlangen, Germany). The commercial TPS Raystation (Raysearch Laboratories, Stockholm, Sweden) which supports MC dose calculation and robust optimisation of the Hyperscan system was used for treatment planning. Maastro Proton Therapy started clinical activity in February 2019 with RayStation v.8A (RS8.A) which only supports static apertures (SA). In March 2019, Raystation version 8B (RS8.B) was clinically released, which supports dynamic layer-by-layer apertures and, thus, the possibility to fully exploit the capabilities of the AA. After plan approval, the treatment plans are transferred to the Aria 15.5 Oncology Information System (Varian Medical Systems, Palo Alto CA, USA).

Figure 2.

Example of the AA leaves position verification with two leaves pattern printed on a transparent film attached to the nozzle and photographed by a CCD camera mounted on a tripod.

​Beam commissioning and TPS data acquisition

In this work, we mainly focused on reporting beam commissioning methodology, but the commissioning process encompasses many other non-beam-related tests or measurements which are listed in Table 1 in Supplementary Material.

Integrated depth dose distributions

Laterally integrated depth dose distributions (IDD) were measured with a large plane parallel IC of 8.16 cm diameter (Bragg Peak IC TM34070, PTW, Freiburg, Germany) mounted on the mechanical arm of a large-size motorised water phantom (PTW MP3-PL). In order to measure as close as possible to water surface, IDDs to build the TPS libraries were acquired from gantry 0° with increments of 100 µm for a representative subset of 15 energies covering the whole clinical range (Table 2 in Supplementary Material). We synchronised the IDD acquisitions with the BM pulses and for each measuring point we collected the charge of 2000 BM pulses. The proton range of the maximum achievable energy (227 MeV) was measured from different gantry angles with 45° steps using a nozzle-mounted device consisting of a stack of parallel plate ICs (Mevion, USA) and solid water.

​Spot size in air

Spot sizes in air, defined as one sigma, were measured across the 20 × 20 cm² scanned field at the isocenter. Spot size, spot shape and symmetry were evaluated for different beam energies, nozzle extension and gantry angles using a scintillator detector (IBA Lynx, Louvain-la-Neuve, Belgium) and a 2D array detector (PTW Octavius 1500 XDR) mounted on a rotating station. To evaluate the beam broadening in air, the spot size of the 227 MeV beam was measured at six distances from the isocenter: -33,–20, −10, 0, 10 and 20 cm.

Spot position

To evaluate the performance of the single focus scanning magnet, 9 and 25 spot patterns were created for low and medium-high energies, respectively (Figure 1 in Supplementary Material). The BMs measure the spot position with a resolution of 2 mm and provide feedback to the system. The spot position accuracy was determined by using a scintillator detector and a 2D array detector mounted on a rotating device for different gantry angles. These detectors were previously aligned using kV-kV imaging.

​Field size in air

The field size in air, defined as the 50 to 50% distance, of an uncollimated 227 MeV 10 × 10 cm² scanned field, was measured at six distances from the isocenter using a scintillator detector: -33,–20, −10, 0, 10 and 20 cm. The beam divergence via the field size was measured at the isocenter, 50 cm and 100 cm downstream from the isocenter to compute the virtual source axis distance (VSAD), as required by the clinical proton TPS from Raysearch.

Determination of absorbed dose to water under reference conditions

The absorbed dose to water at a reference depth z_ref was determined following the N_D,W formalism of IAEA TRS-398 code of practice.⁷ Dose was measured with a plane parallel IC (IBA PPC05) placed in a water tank (PTW, MP3-PL) at a depth of 1/4 of the Bragg peak (BP) maximum as recommended by Raysearch, including the water equivalent thickness of the chamber window, connected to a PTW UNIDOS-E electrometre at 300 V, for a subset of 15 representative energies and 10 × 10 cm² fields, with 2.5 mm spot spacing. We placed the isocenter at the water surface and we measured with the gantry at 0° at the depth of 1/4 of the BP maximum to ensure measuring at the plateau for all beam energies. The reading of the air-vented detector was corrected for ambient conditions. The tabulated beam quality correction κ_Q,Q0 for the PPC05 IC (Markus type IC) was extracted from⁷ based on the residual range of the proton beam at each measurement depth. Furthermore, corrections for polarity effects and ion recombination were also computed.

Proton beams produced by the Mevion S250i synchrocyclotron are pulsed, with clinical dose rates not exceeding 10 Gy/min. The beam pulse frequency is approximately 750 Hz (Ƭ = 1.3 ms) and the pulse duration is 10 μs. The minimum delivery time of a single spot is of about 6.5 ms and the beam-off time (pause between two spots) is approximately 1.3 ms for clinical spot spacings. The ion collection time (t_IC) in the PPC05 air-vented IC at 300 V and 100 V is ~7 and~20 μs, respectively. According to Boag's theory,⁸ the pulse duration, i.e., 10 μs, is similar to t_IC and the pulse period, i.e., Ƭ = 1.3 ms, is sufficiently low for ions to clear out between two consecutive pulses. Thus, considering the micro-structure of the beam, it can be classified as pulsed and, since the beam delivery type is scanning, as scanned pulse.

IAEA TRS-398 recommends the use of the double-voltage technique (V1/V2 ≥ 3) for pulsed or scanned pulsed-beams. This method assumes that for these beams the relationship between the reciprocal of the chamber response (1/M) and the reciprocal of the polarising voltage (1/V) is linear. To determine the ion recombination correction, we used the following formula:

$$k_s=a_0+a_1\frac{M_1}{M_2}+a_2{\left(\frac{M_1}{M_2}\right)}^2$$

where M₁ and M₂ are the chamber readings at V₁ and V₂, respectively, and the constants a₀ = 2.001, a₁ = −2.402 and a₂ = 1.404 were extracted from⁹ with a ratio of 3.

Beam monitor calibration

The BM calibration is required to establish a relation between absolute dose and monitor unit (MU) counts. Due to the relative position of the BMs and the RMS, the Mevion S250i Hyperscan DDS can be considered of Type I according to the definition given in.^10,11 This means that when superimposing several monoenergetic beams of different energy and weight to generate a spread-out Bragg peak (SOBP), the role of the BM is exclusive to determine the output of the SOBP, whereas the shape of the SOBP, i.e. the required energies and energy weights, is determined by the TPS. Consequently, the BM calibration can be conducted in a simpler manner with respect to Type II scanning systems, with absorbed dose measurements in the middle of a SOBP⁷ or with a monoenergetic field. Following the recommendations of Raysearch, we chose the second option and we calibrated the BM by performing absolute dose measurements in water in the centre of a 10 × 10 cm² field with 2.5 mm spot spacing and 1 MU per spot for the highest deliverable energy (227 MeV) at 1/4 of the Bragg peak. The BM calibration was expressed in terms of absorbed dose to water per MU (cGy/MU).

During the acceptance and commissioning phase, the performance of the BM in terms of stability, reproducibility and MU linearity was investigated. Beam output constancy was assessed from multiple gantry angles. A nozzle-mounted device made of Lexan with a cavity at 5 cm depth to position the PPC05 IC was used to perform dose measurements at different gantry angles with 45° steps.

AA verification

The AA position accuracy was verified using five different leaf pattern templates printed on a transparent film attached to the nozzle (Figure 2). Leaf patterns were recorded with a camera system with a tripod. To assess the leaves reproducibility, these measurements were performed on two consecutive days during acceptance. Dose leakage through the leaves of the AA when delivering a trimmed field at the maximum beam energy was measured in air at the isocenter plane using a scintillator device. To assess the influence of gravity on the performance of the AA, a 10 × 10 cm² collimated and uncollimated field at 157 MeV was delivered on the Octavius 2D-detector placed at the isocenter plane with 1 cm of solid water plate or RW3 on top, from different gantry angles i.e., 0°, 90° and 180°. The alignment of the 2D-detector was previously assessed by performing 2D imaging. The field centre position, field size and penumbra were measured at two nozzle extensions: 3.6 cm (minimum) and 12 cm (typically clinical extension).

​Beam model validation

Table 1 lists the measurements conducted to validate the beam model of the Mevion S250i Hyperscan system in RS8.A and RS8.B.

Table 1.

List of measurements adopted for TPS beam model validation

Measurement type	Description	Equipment
Spot profiles of single spots in water at three depths	Measure the lateral X and Y spot profiles at 1/4, 2/4 and 3/4 of the Bragg peak maximum for 15 representative energies.	Water phantom + microdiamond detector (PTW TN60019)
Absolute point dose measurements in water of single energy layers at different field sizes	Dose measurements in the middle of squared fields in water ranging from 3 × 3 cm2 to 20 × 20 cm2 for 15 representative energies.	Water phantom + PPC05 for fields > 5×5 cm2, semiflex 3D for 5 × 5 cm2 and 4 × 4 cm2 fields, and microdiamond detector for 3 × 3 cm2 field.
Absolute point dose measurements of inversely optimised SOBP's in water under normal incidence with SA (RS8.A) and AA (RS8.B)	Point dose measurements along the central axis and off-axis of a 125 cc cube at 5 cm, 10 cm and 20 cm depth and of a 1000 cc cube at 10 cm and 20 cm depth. 2 Gy(RBE) dose prescription.	Water phantom + PPC05
Absolute point dose measurements of inversely optimised SOBP's in water under oblique incidence with SA (RS8.A) and AA (RS8.B)	Point dose measurements along the central axis and off-axis of a 125 cc cube at 5 cm depth at 15° and 30° incidence. 2 Gy(RBE) dose prescription.	Water phantom + PPC05
Lateral profiles of SOBP's with SA (RS8.A) and AA (RS8.B)	Lateral X and Y profiles halfway between water surface and upstream cube surface and centre plane of 8 cc, 125 cc and 1000 cc dose cubes centred at 10 cm depth. 2 Gy(RBE) dose prescription.	Lynx scintillator detector + RW3 plates
Patient QA with SA (RS8.A) and AA (RS8.B)	2D absolute dose measurements at different depths in solid water of patients plans.	Octavius 1500 XDR + RW3 plates
Complex geometry QA with SA (RS8.A) and AA (RS8.B)	2D absolute dose measurements at different depths in solid water of complex geometry plans.	Octavius 1500 XDR + RW3 plates
Absolute dose measurements of inversely optimised plan on heterogeneous phantom with SA (RS8.A)	2D dose measurements of a treatment plan on a heterogeneous head phantom: 1) CT of the Alderson phantom. 2) Clinical head and neck plan with seven beams copied on this CT. 3) CT uploaded in Raystation as a QA phantom and the first two slices were overridden with air and from the sixth slice they were overridden with solid water (Figure 5 in Supplementary Material). 4) QA plan made in Raystation on this modified CT with all beams from 0°. 5) Slices from 3 to 5 of the Alderson phantom placed on top of the 2D detector on the treatment couch. 6) kV-kV imaging used to match these planar images with the planning CT and applied couch corrections. 7) seven beams delivered from 0° on top of these slices.	Alderson head phantom + Octavius 1500 XDR + RW3 plates

AA, Adaptive Aperture; SA, Static Aperture.

Results

Beam commissioning and TPS data acquisition

Depth dose distributions

The IDD were found to be reproducible and energy range fluctuations were within 0.3 mm at different gantry angles. Table 2 in Supplementary Material reports the proton beam energy (in MeV), peak width, range and distal penumbra in water for the subset of 15 representative energies measured with a nozzle extension of 16 cm. The peak width analysed over the 15 energies was on average 8.2 ± 0.2 mm. The distal penumbra, defined as 80–20% of the IDD, was 4.5 ± 0.1 mm averaged over the 15 energies.

Spot size in air

The spot size (one sigma) in air was consistent at five gantry angles within 0.4 mm. Figure 3a shows the augmentation of the spot size in air as a function of the distance from the isocenter plane for the 227 MeV proton beam. In Figure 3b are reported the spot sizes at the isocenter plane for the 15 proton beam energies. At a nozzle extension of 10 cm (10 cm air gap), uncollimated pencil beams have an averaged (x and y) spot size in air comprised between 4.1 mm and 16.4 mm for a range from 32.2 cm (227 MeV) to 1.9 cm (45 MeV) in water (Table 2 in Supplementary Material). The maximum difference in major and minor spot sigma at any rotation with respect to the beam axes was less than 0.4 mm for all energies. In a treatment plan, the spots at the edge of a field are laterally trimmed at most by one half. As an example, a 78 MeV spot was trimmed on the left by the AA leaves and the penumbra on this side-was reduced by a factor of about 4.5 compared to the right side penumbra; i.e., from 14 to 3 mm (Figure 3b).

Figure 3.

a) Uncollimated spot size in air expressed in terms of sigma as a function of the distance from the isocenter for the 227 MeV beam energy; b) uncollimated spot size and penumbra in air for 15 energies at the isocenter plane at a nozzle extension of 10 cm (10 cm air gap); and left penumbra of a 78 MeV spot trimmed on the left side by the AA (cross).

Spot position

The transverse position accuracy of spot patterns, defined as the average position error of all spots against the expected spot pattern at the room isocenter plane, was less than 1 mm at the evaluated five gantry angles for 15 energies. According to specifications, the spot position accuracy of the centroid of any single spot should be within a distance equivalent to 10% of the same spot size under the BP in water.

Field size in air

Figure 2 in Supplementary Material shows the field sizes in air and the augmentation of the penumbra as a function of the distance from the isocenter plane for the 227 MeV proton beam energy. The VSAD of our machine is 1.82 m.

Determination of absorbed dose to water under reference conditions and beam monitor calibration

The ion recombination was assessed at several voltages and we decided to use the working voltage recommended by the IC vendor of 300 V for the PPC05 IC because the measured differences in $k_s$ between 300 V and 400 V were found to be 0.1%. $k_s$ at 300 V averaged over 15 beam energies covering the therapeutical range of interest was 1.007 ± 0.001. The dose output evaluated for different gantry angles (0°, 45°, 90°, 135°, 180°) was 0.0587 ± 0.0002 cGy/MU, with a maximum deviation from the mean of ±0.5%. To verify the MU linearity, the ratio MU/spot of a standard 10 × 10 cm² calibration field with 2.5 mm spot spacing at maximum beam energy were varied from 0.13 to 6 MU/spot. The relative dose per MU was constant within ±2% or ±2 cGy. The dose output of the BM evaluated over a 6-month period was well within ±1.5%.

AA verification

The leaf position accuracy at five separate patterns was within ±1 mm at all gantry angles and nozzle positions (Figure 2). The leaf position repeatability was found to be well within ±1 mm. The dose leakage through AA leaves when delivering a trimmed field of maximum energy, measured in air at the isocenter plane, was ≤1.5%. The centre of a collimated and uncollimated 10 × 10 cm² field was within ±1 mm in the inplane and cross-plane directions at all gantry angles evaluated. The penumbra improvement due to the presence of the AA of a 157 MeV scanned field (not inversely optimised) was on average 5.1 ± 0.4 mm and 2.3 ± 0.1 mm at 3.6 cm and 12 cm nozzle extension, respectively.

Beam model validation

Figure 4 shows the agreement between experimental data and MC-TPS IDDs. The comparison of the measured and TPS spot size in water at three different depths (1/4, 2/4, and 3/4 of the Bragg peak) for 15 proton beam energies is shown in Figure 5. The maximum difference was 0.5 mm.

Figure 4.

Experimental and MC-TPS depth dose distributions in water for 15 proton beam energies.

Figure 5.

Comparison of measured (filled markers) and Raystation TPS (unfilled markers) spot sizes in water at three different depths (1/4, 2/4, and 3/4 of the Bragg peak) for the 15 proton beam energies and air gap of 10 cm. (a) Spot profiles in water in X-direction. (b) Spot profiles in water in Y-direction.

The percentage differences between measurements and TPS of the absorbed dose to water in the middle of monoenergetic squared fields ranging from 3 × 3 cm² to 20 × 20 cm² for the 15 proton beam energies were within ±3% (Figure 6).

Figure 6.

Percentage differences between measurements and TPS of the absorbed dose to water in the middle of monoenergetic squared fields ranging from 3 × 3 cm² to 20 × 20 cm² for the 15 proton beam energies.

A total of 42 dose points on-axis and off-axis inside five SOBPs with aperture at different depths in water with normal and oblique beam incidence were measured and compared with the TPS ones with percentage differences within ±3% (Figure 7).

Figure 7.

Percentage differences of on-axis and off-axis absorbed dose to water inside 125 cc SOBPs centred at 5 cm, 10 cm and 20 cm depth, and 1000 cc SOBPs centred at 10 cm and 20 cm depth. The oblique beam incidence at 15° and 30° was evaluated for the 125 cc SOBP centred at 5 cm depth. All SOBPs were planned with aperture (Table 1).

The absolute differences between experimental and TPS field sizes in water of two SOBPs with aperture at different depths and nozzle extensions were within ±1.5 mm (Figure 3 in Supplementary Material).

Field penumbra in water at different depths and nozzle extensions was slightly underestimated in the TPS when comparing it with experimental data, but within ±1.5 mm (Figure 4 in Supplementary Material).

For validating the TPS dose computation accuracy for irregular fields shaped by the AA, single-field plans on complex geometries were generated in the TPS (Figure 8). These plans were delivered on the Octavius XDR 1500 with appropriate RW3 solid water build-up. The agreement between the planned and measured dose at different depths through the target volume, using a γ criterium of 2%/2 mm and 3%/3 mm, was 96.5 and 98.5%, respectively. The accuracy of the dose distributions with high modulation is shown in the dose profiles of Figure 8.

Figure 8.

Complex geometry plans used to validate the TPS dose computation accuracy for irregular fields shaped by the adaptive aperture: ‘Hairdryer’ (top) and ‘rocky-hand’ (bottom). From left to right: TPS representation of the beam, TPS dose plane, measured dose plane, line dose profiles: TPS (solid line) and measurements (dots).

Since we have started treating patients at our clinic, patient QA measurements of head and neck and brain patients (these cases being the most challenging in terms of field shaping with AA) have an average γ 3%/3 mm pass rate higher than 98%.

If we apply the methodology reported in¹² the results of the absolute dose measurements of inversely optimised plans on a heterogeneous head and neck phantom had a γ passing rate of 96% considering a 3%/3 mm criteria. As shown in Figure 5 in Supplementary Material, the largest differences were found in correspondence of two locations: a) the nasal septum, with air-bone interfaces and b) the curved cervical bone-muscle interface.

In Figure 6 in Supplementary Material, we added an example to illustrate the level of modulation we can achieve when treating a geometry similar to a Chordoma, i.e. ‘donut-shape’, with an organ at risk in the middle, simulating the brainstem. We used a clinical beam arrangement of four beams with AA and SA, to illustrate the differences. With a dose prescription of 74 Gy(RBE), we can achieve a dose fall-off of 90 and 67% in 18 mm with the AA and the SA, respectively.

Discussion

Given the synchrocyclotron beam nature, instantaneous charge per pulse and dose rates can be high. As a consequence, high recombination loses can be expected. To minimise this effect, plane parallel ICs with small electrode distance (≤1 mm), as the PPC05, or high working voltages are recommended.¹³

One consequence of the beam travelling about 2 m in air and of the RMS position is the large spot size at the isocenter of Mevion proton beams compared to other systems.^6,14,15 Therefore, the AA is necessary to sharpen the penumbra in order to achieve high dose modulation. The dimension of the spot for each energy is not unique. It depends on the nozzle extension since the RMS is movable and so does the air gap. In clinical plans the air gap is normally chosen around 5–8 cm air for anterior beams to reduce the spot size and 10 cm for posterior beams, to avoid collision with the treatment couch. Large spot sizes have advantages in the mitigation of the interplay effects during treatment of moving targets and on the overall plan robustness.

The distribution of the sigma of the spot size in water is not linear around 200 MeV (Figure 5). This is due to the fact that for generating energies above 200 MeV, the thicker plate of the RMS, which is the most downstream, i.e., closer to the patient, is not used anymore to generate these energies, but the thinner plates more upstream the RMS are used. As a consequence, the spot size is slightly larger due to more scattering in air. This effect is very well modelled by the TPS.

The width of the BP and the distal dose fall-off is almost constant throughout the energy range due to the Mevion RMS (Table 2 in Supplementary Material). This is one of the differences in beam properties between this compact system and most conventional proton systems. The use of any passive device as a ripple filter to enlarge the peak width in depth for low proton beam energies is not needed with such a system. However, a downside of a fixed peak width as a function of energy is the difficulty to achieve dose modulation along the beam direction at lower energies.

Given the small dimensions of the AA module with respect to the field size at the isocenter plane, i.e. 20 × 20 cm², from a mechanical point of view, there is no difference between a static or an adaptive aperture configuration because the leaves need to travel across the 20 × 20 cm² area to follow the spot-by-spot delivery in each energy layer. The difference between static and adaptive relies on how the TPS version is able to model the AA module. First, simulating a static aperture like in a passive system, i.e. aperture as large as the largest cross-section of the target in beam-eye-view (BEV). Second, generating an aperture shape different per energy layer, i.e. trimming the lateral edges at every energy layer. Furthermore, the system could also trim spot-by-spot, but it is not yet supported by the TPS.

Our clinical workflow was double-checked by an external institution audit and a dosimetric intercomparison was conducted in our centre between the Dutch and Belgian proton facilities which confirmed the clinical usability of the system.

Conclusion

The Mevion S250i Hyperscan system with layer-by-layer field collimation was installed in Maastricht in February 2018. One year later, in February 2019, after installation, acceptance and commissioning, patient treatments started. The beam characteristics of this proton accelerator are different from other commercial systems and it is important to properly understand and characterise them.

The dynamic layer-by-layer field collimation sharpens the beam penumbra and the treatment beam angles should be wisely chosen to fully take the advantage of the AA. The beam validation confirmed that the beam was properly modelled in the TPS. Therefore, the system can be used clinically.