JOURNAL OF APPLIED CLINICAL MEDICAL PHYSICS, VOLUME 17, NUMBER 2, 2016
`
`Commissioning and initial experience with the first clinical
`gantry-mounted proton therapy system
`Tianyu Zhao,a Baozhou Sun, Kevin Grantham, Leith Rankine, Bin Cai,
`Sreekrishna M. Goddu, Lakshmi Santanam, Nels Knutson, Tiezhi Zhang,
`Michael Reilly, Beth Bottani, Jeffrey Bradley, Sasa Mutic, and Eric E. Klein
`Department of Radiation Oncology, Washington University School of Medicine, St. Louis,
`MO, USA
`tzhao@radonc.wustl.edu
`Received 3 June, 2015; accepted 13 October, 2015
`The purpose of this study is to describe the comprehensive commissioning process
`and initial clinical experience of the Mevion S250 proton therapy system, a gantry-
`mounted, single-room proton therapy platform clinically implemented in the S. Lee
`Kling Proton Therapy Center at Barnes-Jewish Hospital in St. Louis, MO, USA.
`The Mevion S250 system integrates a compact synchrocyclotron with a C-inner
`gantry, an image guidance system and a 6D robotic couch into a beam delivery
`platform. We present our commissioning process and initial clinical experience,
`including i) CT calibration; ii) beam data acquisition and machine characteristics;
`iii) dosimetric commissioning of the treatment planning system; iv) validation
`through the Imaging and Radiation Oncology Core credentialing process, includ-
`ing irradiations on the spine, prostate, brain, and lung phantoms; v) evaluation of
`localization accuracy of the image guidance system; and vi) initial clinical experi-
`ence. Clinically, the system operates well and has provided an excellent platform
`for the treatment of diseases with protons.
`PACS number(s): 87.55.ne, 87.56.bd
`Key words: proton, commissioning, passive scattering
`
`
`I.
`
`INTRODUCTION
`
`Proton therapy has been known for the capability of delivering highly precise radiation doses to
`tumor volumes while protecting healthy tissue from radiation side effects for better outcomes.(1,2)
`However, this advanced technology has not been widely accepted because of: i) cost, exceed-
`ing $150 million for multiroom systems, ii) the space required to host accelerator, beamline,
`transport systems, and treatment rooms, and iii) complex delivery systems that require engineers
`specially trained and certified to run and maintain.
`A compact proton therapy machine, specifically a single-room proton therapy unit, is appeal-
`ing to hospitals with a population of regional patients who benefit from this technology, based
`on current clinical evidence. The benefits of a single-room system include reduced cost for
`i) the machine, ii) space, due to smaller footprint, iii) the construction and installation, and
`iv) maintenance, due to lower system complexity and power consumption. A compact system
`is much easier to integrate with the rest of the hospital geometrically and administratively,
`instead of being detached at remote location. It improves the flexibility of moving patients
`between proton rooms or from proton service to adjacent IMRT service in the event of a system
`breakdown. A compact system operates similar to a photon system as the beam is no longer
`orchestrated among multiple rooms by a team of engineers With a lower financial barrier, the
`
`a Corresponding author: Tianyu Zhao, Department of Radiation Oncology, Washington University School of
`Medicine, 4921 Parkview Place, Campus Box 8224, St. Louis, MO 63110, USA; phone: (314) 362 2676;
`fax: (314) 362 8521; email: tzhao@radonc.wustl.edu
`
`24
`
`
`
` 24
`
`Elekta_01994
`
`Elekta Ex. 1023
`
`

`
`25
`
` Zhao et al.: First clinical gantry-mounted proton therapy system
`
`25
`
`demand for single-room systems is expected to grow rapidly as more institutions can practi-
`cally support such technology.
`The world’s first single-room proton therapy system, the Mevion S250 (Mevion Medical
`Systems, Littleton, MA) was installed and commissioned in the S. Lee Kling Proton Therapy
`Center at Barnes-Jewish Hospital in St. Louis, MO, USA. The system has been in clinical opera-
`tion since the December of 2013. In this study, we present the comprehensive commissioning
`process and initial clinical experience with the Mevion S250.
`
`
`II. MATERIALS AND METHODS
`
`The Mevion S250 system includes a synchrocyclotron (1.8 m in diameter and 22 tons in
`weight) mounted on a gantry that rotates from -5° to 185° around isocenter. A pair of annular
`superconducting coils made of Nb-Sn superconductor and a pair of shaped ferromagnetic
`poles are used to generate a solenoid magnetic field that peaks at 8.7 Tesla at the center of the
`synchrocyclotron to produce a bundle of protons with nominal energy of 250 MeV. The coils,
`hosted by a stainless steel bobbin inside a cryostat, are cooled to 4K by cyrocoolers connected
`to liquid helium. A magnetic regenerator is placed close to the extraction point to produce a
`bump in the magnetic field that disrupts the vertical focusing of protons. Angle and pitch of
`proton orbits are altered toward the extraction channel.
`The unique design in mounting the synchrocyclotron on the gantry eliminates the need for a
`transportation beamline and further reduces the requirement on space. However, the output and
`energy are impacted. As the gantry rotates, the gravity on the superconducting coils can shift the
`magnetic field by tenths of a millimeter with respect to the magnetic field regenerator designed
`to deviate protons into the extraction channel. To compensate for the gantry-angle-dependent
`energy fluctuations of protons into the extraction channel, a variable-thickness wedge is intro-
`duced at the entrance to the extraction channel to fine-tune the proton energy. Although this
`mechanism produces consistent mean energy output at all gantry angles, the energy spectrum
`differs slightly as protons go through various thickness of scattering material. As no energy
`selection system is present downstream in the beamline to filter out undesired energies, the
`variations in energy spectrum could have a direct impact on the monitor unit (MU) chamber
`and cause moderate output dependence on gantry angle. This effect has to be evaluated in com-
`missioning, and accounted for in determination of output for each treatment field.
`The beam extracted at 250 MeV is adjusted to the energy required for the treatment by
`absorber wheels that are made of graphite, and a pair of coarse energy absorber and a fine
`absorber that are made of Lexan. However, a magnetic analyzer is not present in the beamline
`to maintain a tight energy width after being degraded. This design is different from commer-
`cially available models from other vendors. Its dosimetric impact needs to be evaluated at the
`entry region and distal falloff.
`Twenty-four options are available for treating targets with range up to 32 cm with maximum
`field size of 14 cm, and up to 25 cm in depth with maximum field size increased to 25 cm.
`The beam-shaping system includes a first scatter with 21 steps, a pair of alternate coarse range
`absorbers, a fine range absorber, three secondary scatters and 14 range modulation wheels.
`Steps in the modulation wheels are compensated for scattering with an appropriate ratio of lead
`and Lexan. All options are categorized into three groups, “small,” “deep,” and “large,” based
`on maximum field size and range. “Small” options treat targets up to 20 cm in depth with a
`maximum field size of 14 cm and modulation width from 2 cm to the range. “Deep” options
`treat targets with depth less than 32 cm but larger than 20 cm. The maximum field size of deep
`options is 14 cm and the maximum modulation width is limited to 10 cm. The deep options are
`mainly designed for prostate treatment. “Large” options treat targets up to 25 cm in depth with
`maximum field size of 25 cm and modulation width up to 20 cm. Options in the same group
`share the same secondary scatter. The nominal SAD of the machine is 200 cm.
`
`Journal of Applied Clinical Medical Physics, Vol. 17, No. 2, 2016
`
`Elekta_01995
`
`

`
`26
`
` Zhao et al.: First clinical gantry-mounted proton therapy system
`
`26
`
`A. CT Calibration
`Conversion from Hounsfield units (HU) to relative proton stopping power ratio (PSPR) plays
`a vital role in proton therapy. General perception of uncertainty of 3.5% on range is accepted
`and applied widely in proton therapy due to the uncertainty associated with CT imaging and
`conversion from HU to PSPR. It has been found that stoichiometric CT calibration is more
`precise than the tissue substitute calibration.(3) Following Schneider’s original implementation,(4)
`an electron density phantom CIRS model 062(5) (Computerized Imaging Reference Systems
`Inc., Norfolk, VA) was selected for evaluation of HU vs. PSPR due to its popular availability
`and easy access to the published data on substitutes’ composition. The phantom was made of
`an inner cylindrical part (head) and an outer ring (body). The base material of the phantom was
`water-equivalent plastic with holes to accommodate 17 inserts. The dimensions of the phantom
`were 33 cm in width and 27 cm in height. All tissue substitutes were built with physical and
`electron densities similar to the recommendations of ICRU Report 44.
`To evaluate the uncertainty in our CT calibration approximately, we did an inter-institute
`comparison with the institutional calibration curves employed in three other proton institutions:
`University of Pennsylvania, MD Anderson Cancer Center, and ProCure Oklahoma. The same
`phantom with inserts in exactly the same arrangement was circulated through all participating
`institutions. After the phantom was scanned with available CT protocols for proton therapy
`in participating proton institutions, the acquired CT images and CT calibration curves were
`collected through a secure FTP server. However, CT calibration curves from all participating
`institutions could not be compared directly because the CT scanners involved in building the
`calibration curves were different in terms of manufacture, model, and energy. The PSPRs of
`the inserts were instead determined by applying the corresponding institutional CT calibration
`curves on the collected CT images, and plotted with HUs from our institution. This process
`rebuilt CT calibration curves from all participating institutions on the same CT scanner for
`direct comparison. Assuming the average of all four calibration curves the ground truth, range
`uncertainties from the variations of CT calibration were evaluated on our clinical plans. This
`process transferred variations in CT calibration into range variations. It is estimated that the
`range uncertainty is ± 0.5% from imaging and calibration, and ± 0.5% from CT conversion to
`tissue if ionization energy is excluded.(6) We expect our range uncertainty from the CT cali-
`bration alone is close to that estimation. Since we were only interested in comparing PSPRs
`obtained with the active CT protocols used in the participating institutions, where the selection
`of kVp, FOV, slice thickness, and filter matched the parameters used for commissioning their
`CT calibration curves, the impact of the variations in institutional CT protocols was not studied
`and reported in this manuscript.
`The CT calibration was also tracked on historic data from 2008 to 2014 using measurements
`from annual QA tests. The purpose of this assessment was to evaluate the long-term reliability
`of CT calibration for proton therapy, which, if understood, would help to decide the frequency
`of quality assurance for the CT scanner. A replacement of X-ray tube was performed in 2011,
`which offered the best opportunity to check the stability of the CT scanner.
`
`B. Beam Data
`The general guideline for acquiring beam data for photon beam commissioning has been
`described in the report of AAPM Task Group 106.(7) Many aspects apply to proton beam com-
`missioning. Two special considerations apply to the Mevion S250.
`First, the acquisition time for each measurement point has to be integrated over 2.08 s in
`order to allow the beam spot to be distributed equally over a rotating range modulation wheel
`(RMW). This period, coming from the interplay between the pulse frequency and modulation
`wheel rotation, allows the beam spot to be distributed evenly on the RMW. Otherwise, significant
`measurement uncertainties would be observed. However, this increases time for data acquisition.
`Second, the output dependency on gantry angle has to be measured due to the presence of a
`fine wedge that compensates for the shifting of magnets during gantry rotation.
`
`Journal of Applied Clinical Medical Physics, Vol. 17, No. 2, 2016
`
`Elekta_01996
`
`

`
`27
`
` Zhao et al.: First clinical gantry-mounted proton therapy system
`
`27
`
`The beam model was commissioned in an Eclipse V11.0.30 (Varian Medical Systems, Palo
`Alto, CA), which employs a pencil-beam algorithm.(8) As all TPS commercially available use
`similar algorithms with slight differences in the modeling of Bragg peak and the calculation
`of spread-out Bragg peak (SOBP), the results and discussion in this study apply generically.
`For commissioning the pencil-beam algorithm for passive scattering, four sets of measure-
`ments are required. They include (i) percent depth dose in water, (ii) longitudinal profile in air
`under a open block, (iii) lateral profiles in air under a half-block, and (iv) lateral profiles in air
`without blocking.
`Percent depth-dose curves were acquired using a 3D scanning tank (Blue phantom, IBA
`Dosimetry America, Bartlett, TN) at nominal source-to-surface distance (SSD) of 200 cm
`with water surface leveled at radiation isocenter, using a parallel-plate chamber (PPC05, IBA
`Dosimetry). The PPC05 chamber has a sensitive volume of 0.046 cc with 9.9 mm in diameter and
`0.6 mm in electrode spacing. An open ring aperture was used for all depth-dose measurements
`to minimize collimator scatter. The entrance window of the PPC05 is made of air-equivalent
`plastic C-552 with physical thickness of 1 mm. A shift of 2.2 mm away from the source was
`applied on the acquired depth-dose curve to account for i) 1.55 mm downstream shift of the
`effective measurement point at the inner surface of the entrance window, and ii) 0.65 mm rise
`of water surface after the moving parts holding the chamber holder submerged under the water
`surface. The dimensions of the moving parts are illustrated in Fig. 1.
`The width of a pristine Bragg peak is defined at the 90% of the peak dose. Distal penumbra
`is defined as the distance from the 80% to the 20% in the distal falloff region. Both properties
`were measured and reported as they provided important information on range straggling and
`energy spread of protons.
`Source size was determined by acquiring profiles in air at various distances from the
`nominal source position under half-beam block using edge detector (Sun Nuclear Corporation,
`Melbourne, FL). The diode in the edge detector had a dimension of 0.8 mm in both width and
`length, making it ideal for measuring the sharp dose gradient. The snout position for all lateral
`profiles was fixed to 40 cm. The source size was modeled as a function of residual range of
`proton, which was achieved with various nozzle-equivalent thicknesses (NET) by setting the
`modulation wheel at various steps.
`
`Fig. 1. The dimensions of the moving parts that hold the chamber holder. There is a pair of the moving parts on both sides
`of the water tank. The dashed line indicates the water surface in our setup.
`
`Journal of Applied Clinical Medical Physics, Vol. 17, No. 2, 2016
`
`Elekta_01997
`
`

`
`28
`
` Zhao et al.: First clinical gantry-mounted proton therapy system
`
`28
`
`Virtual source-axis-distance (VSAD) was determined by acquired profiles in air at transverse
`planes, 20 cm upstream from the machine isocenter, along isocenter, and 20 cm downstream
`from the machine isocenter under a square block using the edge detector.
`Effective source-axis-distance (ESAD) was determined by acquired longitudinal profiles
`along the central axis using a cylindrical type ionization chamber (FC65, IBA Dosimetry).
`Measurements were taken at 11 points equally distributed around the machine isocenter on a
`span of 40 cm. The ESAD was calculated by fitting measurements according to the inverse
`square law. The ESAD was a function of residual range of proton, as well.
`
`C. Dosimetric commissioning
`Validation measurements were taken for all 24 options using open fields. They included percent
`depth dose along the central axis and transverse profiles at various depths (middle of SOBP,
`one-third of modulation width upstream, and downstream from the middle of SOBP). Measured
`doses were compared to predictions from the TPS. We performed 1D gamma analysis using
`3%/1 mm criteria and deemed pass with results larger than 90%. The use of 1D gamma analysis
`was mainly used as a metric for evaluating the accuracy of beam modeling on SOBP except
`the distal falloff region. Any discrepancy larger than 3% on the proximal shoulder or larger
`than 5% in the entrance region was tuned by adjusting the partial shinning correction(9,10) and
`entrance region correction.
`As historically noted, prediction of output and monitor unit (MU) is not supported in current
`treatment planning systems. Although our treatment planning system was fully commissioned
`to ensure accurate calculation of dose distribution, MUs are assigned to each field by explicit
`measurement. A verification plan was generated for each beam by duplicating the beamline onto
`a digital water phantom that mimicked the measurement phantom. Apertures were maintained
`and the range compensator was removed in the verification plans to eliminate the perturbation in
`dose measurement from compensator scattering. The treatment isocenter was set in the middle
`of SOBP where the dose output (cGy / MU) was measured. Measurements were performed with
`same geometry and beam parameters. A standard 10 cm × 10 cm square aperture and a 0.6 cc
`cylindrical ionization chamber were used for all measurements, except fields with radius less
`than 2 cm. Special measurements for small fields were taken with the corresponding apertures
`in place and a small cylindrical ionization chamber due to its small sensitive volume. In addi-
`tion, a MU prediction model based on the work by Kooy et al.(11) was developed as a secondary
`check for measurements.(12) The input parameters of the MU prediction model were range and
`modulation width. The model predicted the output by fitting all measured data. The accuracy
`of the prediction is expected to increase with the accumulation of additional measurements.
`An inhomogeneous phantom with half-bone-half-water interface was used to evaluate the
`TPS regarding heterogeneous media. The physical thickness of the bone slab was 2 cm and
`the relative stopping power ratio was 1.63. Profile was acquired on the transverse plane at the
`middle of SOBP in water with a pinpoint chamber (CC04, IBA Dosimetry). A proton field with
`range 32 cm and modulation width 10 cm was used for the test. It presented the worst-case
`scenario as the measurement plane was 27 cm from the bone-water interface, maximizing the
`width and amplitude of the heterogeneity in dose distribution.
`The Imaging and Radiation Oncology Core (IROC) performed further validation via an on-
`site audit as well as off-site dosimetry verification by anthropomorphic phantoms irradiation.
`The on-site visit reviewed all the aspects of our practice, including CT calibration, treatment
`planning, delivery, QA, dosimetry, and workflow. In addition, dosimetric validation was per-
`formed on anthropomorphic phantoms representing four clinical sites including craniospinal,
`prostate, brain, and lung.(13,14) Irradiation of the IROC phantoms served both as an admission
`to NCI-funded cooperative group clinical trials and a stringent test of our institute’s capability
`of planning and delivering treatments with heterogeneity correctly accounted for.
`
`Journal of Applied Clinical Medical Physics, Vol. 17, No. 2, 2016
`
`Elekta_01998
`
`

`
`29
`
` Zhao et al.: First clinical gantry-mounted proton therapy system
`
`29
`
`D. Imaging guidance and 6D robotic couch
`The Mevion S250 proton therapy unit is equipped with a 6D robotic couch and image guidance
`system (Verity). Setup images were provided by a pair of orthogonal planar X-ray imagers with
`sources embedded in lateral wall and floor. The patient alignment process allows corrections in
`six degrees of freedom: translation {x,y,z}, pitch, roll, and yaw {θ, ϕ, ψ}. Geometric accuracy
`of couch corrections and imaging vs. radiation isocenter coincidence were quantified before
`clinical implementation. In addition, a gantry star shot and a couch star shot were performed
`to evaluate the isocentricity of gantry and couch rotation around radiation isocenter.
`A commercial phantom with 16 2 mm tungsten BBs was mounted rigidly on the couch and
`imaged with CT. Seventeen rigid translations/rotations of known magnitude were digitally
`applied to the original CT image using commercial software, initially validated with Varian’s
`OBI system. For each altered image, phantom was mounted on robotic couch in original position,
`then Verity 2D:2D match — posterior–anterior (PA), and left lateral (LLAT) — was performed
`using DRRs from the altered images. Corrections were recorded and applied. The phantom
`was imaged a second time and residual corrections recorded. Physical measurements verified
`that applied couch corrections coincided with both physical couch shifts/rotations and known
`CT image translations/rotations. Additionally, imaging vs. radiation isocenter coincidence was
`quantified over couch treatment angles (± 90° from the setup position) using radiochromic film
`and an image-guided couch star shot. The PA and LLAT kV radiographs were taken before each
`beam was delivered to verify imaging/radiation isocentricity.
`
`
`III. RESULTS
`
`A. CT calibration
`Four operating proton therapy facilities participated in this study. As demonstrated in Fig. 2(a),
`reproduced CT calibration curves agreed well in general in the soft-tissue region with maximum
`deviation of 1.1% from the average, but deviated more significantly in the bone region with
`variation up to 2.8%. Range uncertainty from the deviation was determined to be 0.7% ± 0.2%
`in our lung cases, and 0.9% ± 1.2% in brain cases.
`Our CT calibration curve is plotted against the one predicted by IROC from their site visit
`six months after the first patient in Fig. 2(b). Maximum deviation was measured to be only1.2%.
`Calibration curves generated with annual QA data demonstrated tight variations from 2008
`to 2013, as shown in Fig. 2(c). The absolute variation (maximum – minimum) in relative stop-
`ping power ratios was measured to be 0.019 in the span of six years in a hard bone insert with
`physical density of 1.25 g / cm3, or 1.27% with respect to the mean value.
`
`Journal of Applied Clinical Medical Physics, Vol. 17, No. 2, 2016
`
`Elekta_01999
`
`

`
`30
`
` Zhao et al.: First clinical gantry-mounted proton therapy system
`
`30
`
`(a)
`
`(b)
`
`(c)
`
`Fig. 2. Variations in the stopping power ratios (a) of the CIRS phantom from four proton institutes. The error bars indi-
`cated the absolute range of the variations and the digits under the error bars were the percentage variations with respect to
`the mean values. (b) CT calibration from our institute (red) plotted against the prediction from IROC. (c) CT calibration
`generated with historic annual QA data from 2008 to 2013.
`
`Journal of Applied Clinical Medical Physics, Vol. 17, No. 2, 2016
`
`Elekta_02000
`
`

`
`31
`
` Zhao et al.: First clinical gantry-mounted proton therapy system
`
`31
`
`B. Beam data
`Figure 3 shows the effective SADs, effective source sizes, and virtual SADs for all options,
`plotted against fitting curves. Options sharing the same secondary scatter were grouped into
`three distinct groups. The effective SAD ranged between 177 cm and 183 cm for large options,
`189.1 cm and 192.8 cm for deep options, and 176.8 cm and 186.8 cm for small options. The
`effective source size ranged between 1.8 cm and 2.9 cm for large options, 1.0 cm and 1.3 cm
`for deep options, and 1.3 cm and 1.5 cm for small options. The virtual SAD ranged between
`183.2 cm and 195.1 cm for large options, 187.0 cm and 190.5 cm for deep options, and 177.5 cm
`and 181.7 cm for small options.
`The measured pristine Bragg peaks of all 24 options are plotted in Fig. 4. The curves were
`corrected for the beam divergence and independent of any specific beamline parameters. All
`overlapped for comparison. Widths of the pristine Bragg peaks varied between 3.9 mm and
`8.0 mm, and distal penumbra varied between 5.9 mm and 6.7 mm as plotted for all options
`in Fig. 4(b) and 4(c). Fitting curves to demonstrate the trend with options in large, deep, and
`small bands were plotted as well.
`
`(a)
`
`(b)
`
`(c)
`
`Fig. 3. Measurements were plotted against fitting for (a) effective SAD; (b) effective source size; (c) virtual SAD.
`
`Journal of Applied Clinical Medical Physics, Vol. 17, No. 2, 2016
`
`Elekta_02001
`
`

`
`32
`
` Zhao et al.: First clinical gantry-mounted proton therapy system
`
`32
`
`(a)
`
`(b)
`
`(c)
`
`Fig. 4. Depth-dose curves (a) of all 24 options; (b) widths of the pristine Bragg peaks plotted against fitting; (c) distal
`penumbra plotted against fitting.
`
`C. Dosimetric commissioning
`Examples of SOBP measurements are plotted against TPS modeling for option 1, 13, and 18 in
`Fig. 5, each of which possesses the largest range for the large, deep, and small bands. Prediction
`from modeling agreed very well with average 1D gamma rate 95.7%, ranging between 91.8%
`and 100%. Slight discrepancies on the order of 2% were observed near the distal falloff due to
`the soft distal shoulder systematically presented for all options. Discrepancies in SOBP between
`measurements and TPS modeling are summarized in Table 1.
`Examples of profile measurements are shown in Fig. 6. All measurements were taken at
`the nominal SAD. Red lines were crossline profiles for a proton beam with range 15 cm and
`modulation width 10 cm under a nondivergent block. Collimate scatter was observed on both
`shoulders at shallow depths. The magnitude of the collimator scatter tapered off with depth.
`After changing to divergent apertures with inner surface in perfect alignment with the beam
`divergence, the measured profiles (green lines) agreed very well with TPS modeling (black
`dots). The passing rate of Gamma analysis increased as well, as demonstrated in Fig. 6(d).
`Divergent apertures are now used routinely in our clinic.
`Measurements on the dependence of output on gantry angles are plotted in Fig. 7. The maxi-
`mum variation was measured 0.7% in large options, 1.1% in deep options, and 2.0% in small
`options. As the maximum variation was less than 1% in large options, MU was corrected for
`gantry angle only for fields involving small and deep options in our clinic.
`
`Journal of Applied Clinical Medical Physics, Vol. 17, No. 2, 2016
`
`Elekta_02002
`
`

`
`33
`
` Zhao et al.: First clinical gantry-mounted proton therapy system
`
`33
`
`(a)
`
`(b)
`
`(c)
`
`Fig. 5. Examples of SOBP measurement plotted against TPS modeling for options 1, 13, and 18; each are options with
`largest range in large, deep, and small groups.
`
`Table 1. Summary of SOBP comparisons of measurements vs. modeling.
`
`
`
`
`
` Option
` 1
` 2
` 3
` 4
` 5
` 6
` 7
` 8
` 9
` 10
` 11
` 12
`
`
`
`Range Modulation
`(mm)
`(mm)
`250
`200
`225
`200
`208
`200
`187
`187
`167
`167
`148
`148
`131
`131
`114
`114
`99
`99
`85
`85
`72
`72
`60
`60
`
`Range
`Diff.
`(mm)
`0.30
`0.13
`1.46
`0.60
`0.61
`0.60
`0.68
`0.00
`0.07
`1.07
`0.38
`0.74
`
`
`
`
`Gamma
`(3%/1mm) Option
`96.7%
`13
`96.5%
`14
`94.3%
`15
`96.3%
`16
`96.5%
`17
`93.6%
`18
`92.6%
`19
`93.9%
`20
`99.1%
`21
`93.5%
`22
`94.3%
`23
`100%
`24
`
`
`
`Range Modulation
`(mm)
`(mm)
`320
`100
`295
`100
`270
`100
`245
`100
`220
`100
`200
`200
`177
`177
`153
`153
`132
`132
`110
`110
`90
`90
`69
`69
`
`Range
`Diff.
`(mm)
`0.28
`0.28
`0.07
`0.04
`0.26
`0.68
`1.24
`0.46
`1.27
`0.89
`0.29
`0.08
`
`Gamma
`(3%/1mm)
`96.6%
`96.9%
`96.7%
`99.8%
`91.8%
`96.7%
`99.8%
`96.3%
`92.9%
`93.6%
`95.9%
`92.1%
`
`Journal of Applied Clinical Medical Physics, Vol. 17, No. 2, 2016
`
`Elekta_02003
`
`

`
`34
`
` Zhao et al.: First clinical gantry-mounted proton therapy system
`
`34
`
`(a)
`
`(b)
`
`(c)
`
`(d)
`
`(e)
`
`Fig. 6. Comparison of crossline profiles of divergent aperture, nondivergent aperture, and treatment planning at various
`depths. Noticeable differences were observed along the field edges, both inside and outside of the fields. The differences
`vanished with increased depth in water.
`
`Journal of Applied Clinical Medical Physics, Vol. 17, No. 2, 2016
`
`Elekta_02004
`
`

`
`35
`
` Zhao et al.: First clinical gantry-mounted proton therapy system
`
`35
`
`Discrepancies of the output predicted by our MU prediction model and measured with the
`FC65 chamber in a water tank for the first 400 fields are plotted in Fig. 8. The maximum dis-
`crepancy was measured -2.60%. The mean discrepancy was 0.53%.
`Figure 9(a) shows the phantom used to evaluate dose distribution under heterogeneous con-
`ditions. The measurement was taken at the nominal SAD with a depth of 27 cm in the water
`tank. Measured crossline profile under the bone–air interface is plotted against prediction from
`TPS in Fig. 9(b). Maximum discrepancy was measured to be 4.7% right under the bone–air
`interface. 1D gamma rate (3%/1 mm) for this measurement was 94.6%.
`Once we hit the six-month milestone for treating patients and had treated at least three differ-
`ent disease sites, the IROC Houston group performed a two-day review of our system including
`independent measurement of absolute dose, profiles and percent depth dose, and CT calibration
`curves along with imaging verification accuracy. The output measured by TLDs was within
`1% of our institution’s designated output. The beam parameters including range, modulation
`width, flatness, and symmetry were all within the tolerance. The site visit revealed no issues.
`In addition, four IROC phantoms have been irradiated and deemed to pass for spine, brain,
`prostate, and lung. The phantom end-to-end testing utilized the same personnel (physicists,
`dosimetrists, and therapists) as for any patient to conduct simulation, contouring, treatment
`planning, plan review, MU measurement, QA, and delivery. The results and criteria are sum-
`marized in Table 2; all deemed pass as assessed by TLD and film measurement.
`
`Fig. 7. Angular dependency of output in large (red), deep (green), and small (blue) options.
`
`Fig. 8. Distribution of discrepancies between our MU prediction model and measurements.
`
`Journal of Applied Clinical Medical Physics, Vol. 17, No. 2, 2016
`
`Elekta_02005
`
`

`
`36
`
` Zhao et al.: First clinical gantry-mounted proton therapy system
`
`36
`
`(a)
`
`(b)
`
`Fig. 9. The heterogeneous phantom (a) used for validation of dose distribution. The thickness of the bone slab was 2 cm
`and the stopping power ratio was 1.63. The measured crossline profile (b) was plotted against prediction from TPS. The
`maximum discrepancy was measured 4.7%.
`
`Table 2. Summary of IROC phantom irradiation results for spine, prostate, head, and lung.
`
` Phantom
` Spine
`
`
`
`
`
`
` Prostate
`
`
`
`
`
`
`
`
`
`
` Head
`
`
`
`
`
`
`
`
`
`
`
`
` Lung
`
`
`
`
`
`
`
`
`
`Location
`TLD Superior
`TLD Inferior
`Location
`Center Prostate (Left)
`Center Prostate (Right)
`Film Plane
`Coronal
`Sagittal
`Location
`Target TLD (Right)
`Target TLD (Left)
`Film Plane
`Coronal
`Sagittal
`Location
`Target Superior
`Target Inferior
`Film Plane
`Axial
`Coronal
`Sagittal
`Average over 3 planes
`
`IROC-H vs. Inst.
`0.96
`0.96
`IROC-H vs. Inst.
`0.95
`0.94
`Gamma Index
`99
`100
`IROC-H vs. Inst.
`0.95
`0.97
`Gamma Index
`99%
`96%
`IROC-H vs. Inst.
`0.95
`0.96
`Gamma Index
`92%
`88%
`94%
`91%
`
`Journal of Applied Clinical Medical Physics, Vol. 17, No. 2, 2016
`
`Criteria
`0.93-1.07
`0.93-1.07
`Criteria
`0.89-1.03
`0.89-1.03
`Criteria
`≥85%
`≥85%
`Criteria
`0.95-1.05
`0.95-1.05
`Criteria
`≥85%
`≥85%
`Criteria
`0.92-1.02
`0.92-1.02
`Criteria
`≥80%
`≥80%
`≥0%
`≥85%
`
`Elekta_02006
`
`

`
`37
`
` Zhao et al.: First clinical gantry-mounted proton therapy system
`
`37
`
`D. Imaging guidance and 6D robotic couch
`The Verity suggested couch corrections and known CT shifts/rotations agreed within ± 1 mm
`(average: Δlat = 0.5 mm; Δvert = 0.4 mm; Δlong = 0.3 mm) and ± 0.4° (average: Δpitch =
`0.24°; Δroll = 0.01°; Δyaw = 0.10°), as demonstrated in Fig. 10. Physical couch measure-
`ments and Verity applied corrections agreed within ± 1 mm (average: Δlat = 0.5 mm; Δvert =
`0.4 mm; Δlong = 0.2 mm) and ± 0.2° (average: Δpitch = 0.03°; Δroll = 0.04°; Δyaw = 0.04°).
`The directionality of all translations and rotat