J Applied Clin Med Phys - 2024 - Taneja - Commissioning and implementation of a pencil‐beam algorithm with a Lorentz

Received: 1 August 2024 Revised: 21 October 2024 Accepted: 12 November 2024
DOI: 10.1002/acm2.14590
RESEARCH ARTICLE
Commissioning and implementation of a pencil-beam

algorithm with a Lorentz correction as a secondary dose
calculation algorithm for an Elekta Unity 1.5T MR linear
accelerator
Sameer Taneja1,2 Hesheng Wang1 David L. Barbee1 Paulina Galavis1
Mario Serrano Sosa1 David Byun1 Michael Zelefsky1 Ting Chen1
1
Department of Radiation Oncology, New Abstract
York University Langone Medical Center, New
York, New York, USA
Purpose: To commission a beam model in ClearCalc (Radformation Inc.) for
2
use as a secondary dose calculation algorithm and to implement its use into an
Department of Therapeutic Radiology, Yale
University School of Medicine, New Haven,
adaptive workflow for an MR-linear accelerator.
Connecticut, USA Methods: A beam model was developed using commissioning data for an
Elekta Unity MR-linear accelerator and entered into ClearCalc. The beam model
Correspondence consisted of absolute dose calculation settings, output factors, percent depth-
Sameer Taneja, PhD, Department of
Therapeutic Radiology, Yale University School
dose (PDD) curves, mutli-leaf collimator (MLC) transmission and dose leaf gap
of Medicine, 20 York St., New Haven, CT error, and cryostat corrections. Beam profiles were hard-coded by the manufac-
06510, USA. turer into the beam model and were compared with Monaco-derived profiles.The
Email: [email protected]
beam model was tested by comparing point doses in a homogenous phantom
obtained through measurements using an ionization chamber in water, Monaco,
and ClearCalc for various field sizes, source-surface distances (SSDs), and
point locations. Additional testing including point dose verification for test plans
using a heterogeneous phantom and patient plans. Post clinical implementation,
performance of ClearCalc was evaluated for the first 41 patients treated, which
included 215 adaptive plans.
Results: PDDs generated using ClearCalc fell within 1.2% of measurements.
Field profile comparison between ClearCalc and Monaco showed an average
pass rate of 98% using a 3%/3 mm gamma criteria. Measured cryostat correc-
tions used in the beam model showed a maximum deviation from unity of 1.4%.
Point dose and field monitor units (MUs) comparisons in a homogenous phan-
tom (N = 22), heterogeneous phantoms (N = 22), and patient plans (N = 57)
all passed with a threshold of 5%/5MU. Clinically, ClearCalc was implemented
as a physics check post adaptive planning completed prior to beam delivery.
Point dose and field MUs showed good agreement at a 5%/5MU threshold for
prostate stereotactic body radiation therapy (SBRT),pelvic lymph nodes,rectum,
and prostate and lymph node plans.
Discussion: This work demonstrated commissioning and clinical implementa-
tion of ClearCalc into an adaptive planning workflow. No primary or adaptive
plan failures were reported with proper beam model testing.
KEYWORDS
adaptive therapy, dose calculation, MR-linac
This is an open access article under the terms of the Creative Commons Attribution License, which permits use, distribution and reproduction in any medium, provided
the original work is properly cited.
© 2024 The Author(s). Journal of Applied Clinical Medical Physics published by Wiley Periodicals, LLC on behalf of The American Association of Physicists in Medicine.
J Appl Clin Med Phys. 2024;e14590. wileyonlinelibrary.com/journal/acm2 1 of 11

https://doi.org/10.1002/acm2.14590
15269914, 0, Downloaded from https://aapm.onlinelibrary.wiley.com/doi/10.1002/acm2.14590 by Cochrane Poland, Wiley Online Library on [03/12/2024]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
2 of 11 TANEJA ET AL.
1 INTRODUCTION mented with Unity accelerators. Yang et al.13 used an

analytical anisotropic algorithm (AAA) in Eclipse (Varian
Independent dose calculations1 are part of compre- Medical Systems) for secondary dose/MU checks for a
hensive quality assurance (QA) programs for external Unity
beam radiation therapy. These safety barriers limit The purpose of this work is to evaluate ClearCalc
potential errors in planning and delivery of radia- (CA), manufactured by Radformation Inc. (New York,
tion prior to patient treatment2 by confirming, via a NY),which utilizes a finite-size pencil beam (FSPB) algo-
dose recalculation, that the treatment fields and plans rithm with a Lorentz correction for magnetic field for a
yield radiation doses within tolerance of that calcu- 1.5T Elekta Unity MR-linear accelerator. Beam model-
lated by the original treatment planning system. This ing and validation are described in this study. To the
is done through independent dose and monitor unit authors’ knowledge, previous work has not been pre-
(MU) number comparisons (dose/MU). The American sented on the ClearCalc software and the author aims
Association of Physicists in Medicine (AAPM) provide to provide a reference for other clinics in ClearCalc
guidelines and recommendations on accurately imple- implementation.
menting secondary dose/MU calculation engines into
clinical use, specifically in task group (TG) numbers 71.3
and more recently 219.4 Commissioning of the sec- 2 METHODS
ondary dose/MU calculation includes performing tests
in accordance with recommendations from AAPM TG- For an in-depth description of the Elekta Unity MR-linear
53.5 and MPPG5a.,6 as well as performing a comparison accelerator, Raaymakers et al. presented an initial proof
between measurements and dose calculations at a of concept and a review of the first patients treated
series of machine parameters and point locations based in 200914 and 20179 , respectively. The Unity pairs a
on TG-219.4 1.5T MRI (Philips Marlin, Philips Healthcare, Amster-
The use of secondary dose/MU calculation algo- dam,Netherlands) with a 7-MV flattening filter free (FFF)
rithms are particularly important in online adaptive photon linear accelerator with the capabilities of step-
treatments, such as those frequently performed in and-shoot intensity modulated radiation therapy (IMRT)
magnetic resonance imaging (MRI)-guided radiother- and three-dimensional (3D) conformal treatments. The
apy using MR-linear accelerators. The online adaptive beam is shaped using jaws with a maximum field size of
workflow of these treatments involves generating a new (57.4 × 22) cm2 in the x- and y-planes, respectively, and
plan for each fraction based on an initial MRI scan of mutli-leaf collimator (MLC) consisting of 160 leaves. The
patient setup and internal anatomy. With the patient fixed dose rate is 425 MU/min at the source-isocenter
on the couch during adaptive planning, it is not feasi- distance of 143.5 cm.
ble to perform pre-treatment patient specific QA using All commissioning and implementation presented in
conventional measurement-based methods. As a result, this work was performed using ClearCalc v.2.4.6. This
secondary dose/MU calculations provide the only inde- version of ClearCalc provided one significant change to
pendent verification of calculated dose of the adaptive the beam model from the initial version, v2.2.9: a hard-
plan prior to delivery. coded cryostat correction that is specific to the author’s
Secondary dose/MU calculations for MR linear accel- machine based on commissioning measurements, as
erators present additional complexity in comparison opposed to a manufacturer-defined cryostat correction.
to traditional linear accelerator calculations due to
the presence of the magnetic field affecting gener-
ated charged particles via the Lorentz force and the 2.1 Beam model
increased source to central axis geometry required due
to the combination of a magnet and ring-based accel- An Elekta Unity beam model was generated using dosi-
erator. Multiple groups have developed in-house and metric data obtained during linear accelerator commis-
explored commercial secondary dose/MU calculation sioning. The following section describes measurements
software utilizing various dose calculation algorithms. and simulation that comprised the beam model.
The accuracy of RadCalc (Lifeline Software Inc., Tyler, Absolute dose calibration to 1 cGy per MU in water
TX), which utilizes a Clarkson integration technique, was completed in an source-axis distance (SAD) con-
was studied for use with a 0.35T ViewRay MRidian7 figuration using a reference field size of 10 × 10 cm2
(ViewRay Inc, Oakland, CA) and a 1.5T Elekta Unity at an source-surface distance (SSD) of 138.5 cm and
(Elekta, Stockholm, Sweden).8 MR-linear accelerators. a depth of 5 cm. Calibration measurements were com-
Various collapsed cone methods using commercial pleted using a PTW TN30013 (PTW, Freiburg, Germany)
OnCentra (Elekta),9,10 Mobius 3D (Varian Medical Sys- Farmer-type ionization chamber using methods outlined
tems Inc., Palo Alto, CA)11 and RayStation (Raysearch by TRS 398.15 and correction factors outlined by recent
Laboratories, Stockholm, Sweden)12 have been imple- studies.16–18
TANEJA ET AL. 3 of 11
Output factors, in the form of a total scatter factor machine as a function of gantry angle due to non-
(Sc,p ),19 were measured using an SSD of 133.5 cm and uniform attenuation resulting from the cryostat compo-
at a depth of 10 cm. Measurements were completed in a nent. The cryostat attenuation was measured using a
BEAMSCAN MR (PTW, Freiburg, Germany) water phan- PTW TN30013 Farmer-type ionization chamber using
tom using a Semiflex 3D ionization chamber (Model No. a 10 × 10 cm2 open beam. A measurement was made
31031, active volume: 0.07 cm3 , PTW Dosimetry) for every 2◦ , with the exception of gantry angles 8◦ –18◦
field sizes larger than 10 × 10 cm2 and a microdiamond to avoid penetrating the cryostat pipe in that direction.
detector (Model No. 60019, active volume: 0.004 mm3 , All readings were normalized relative to the reading at
PTW Dosimetry) for field sizes smaller than 10 × 10 cm2,. gantry angle 0◦ . This machine-specific correction was
Field sizes of equivalent squares of 2 × 2 cm2 ,3 × 3 cm2 , added into the ClearCalc beam model.
5 × 5 cm2 , 10 × 10 cm2 , 15 × 15 cm2 , 22 × 22 cm2 , In addition to user-set parameters in the ClearCalc
29 × 29 cm2 , and 34 × 34 cm2 were added to the beam model, field profiles generated by Radformation
ClearCalc beam model. were set in the ClearCalc beam model specifically for the
Percent depth dose (PDD) curves were measured Unity machine. These profiles were determined using
at an SSD of 133.5 cm using a Semiflex 3D ioniza- beam profiles for a 6 MV FFF energy with an updated
tion chamber for equivalent field sizes ranging from alpha factor to account for difference in attenuation
0.5 × 0.5 cm2 to 30.2 × 30.2 cm2 at a gantry angle of due to energy and Lorentz force. As the user can-
0◦ . The maximum depth of measurement was 13 cm. not adjust the beam profiles in the ClearCalc beam
This work used a method for inputting PDDs published model, this study evaluated the accuracy of the beam
by Graves et al.,8 in which PDDs used for the ClearCalc profiles inClearCalc for a small (2 × 2 cm2 ), medium
model were generated using Monaco Treatment Plan- (10 × 10 cm2 ), and large field (50 × 22 cm2 ).
ning System (Elekta,Stockholm,Sweden) equipped with
a commissioned beam model8 for the Unity machine.
PDDs were simulated with a grid size of 1 mm and an 2.2 Model validation
uncertainty of 0.3% for equivalent-square field sizes of
2, 3, 4, 7, 10, 20, 22, 25.4, 27.2, and 30.2 cm2 . In order The ClearCalc beam model generated for the Unity
to evaluate accuracy of the Monaco-determined PDDs, MR-linear accelerator was benchmarked using methods
a comparison of simulated and measured PDDs was outlined by AAPM TG-219 and MPPG5a, which recom-
performed at depths of 5 cm, 10 cm, and 13 cm for mend performing commissioning tests for a secondary
equivalent-square field size ranging from 2 to 22 cm2 . MU calculation similar to those of a conventional treat-
MLC transmission and dosimetric leaf gap error were ment planning system. The beam model was validated
also added into the beam model. Transmission was using a comparison of point dose measurements for a
measured using an Exradin A1SLMR (Standard Imag- homogenous phantom, for a heterogeneous phantom,
ing Inc., Middleton WI) ionization chamber at a depth and for pre-clinical plans.
of 5 cm in solid water, orientated perpendicular to MLC Dose/MU measurements were compared in a
motion, and set at an SSD of 138.5 cm. 2000 MU were homogenous water phantom using three methods:
delivered separately with the Y1 jaw closed and with ionization chamber measurements in water, calculated
the Y2 jaw closed. The chamber reading with the MLCs using the treatment planning system, and calculated
closed was compared with an open field reading, and using ClearCalc. Dose/MUs were determined at SSDs
the average transmission between the Y1 jaw and Y2 of 133.5, 138.5, and 143.5 cm, field sizes of 2 × 2 cm2 ,
jaw was added to the beam model. As Monaco utilizes 3 × 3 cm2 , 4 × 4 cm2 , 10 × 10 cm2 , 5 × 20 cm2 ,
various parameters when defining the MLC model and 20 × 5 cm2 , and 50 × 22 cm2 , and various dose points
ClearCalc utilizes a single DLG value, DLG value in both on- and off -axis. Ionization chamber measure-
the ClearCalc model was optimized by comparing dose ments were performed in the PTW BEAMSCAN MR
differences with Monaco plans for phantom and clini- water phantom using a Semiflex ionization chamber.
cal reference CT-based and adaptive MR-based plans. 100 MU were delivered to the detector at the point of
Values were determined by the manufacturer and from measurement and the ionization chamber charge read-
work by Tsuneda et al.20 Point doses for identical point ing was corrected for temperature and pressure.An NDW
locations were tested for 12 primary and 10 adaptive calibration factor, and beam quality (kQ ) transfer factor
plans using dosimetric leaf gap errors of 0.01, 0.02, 0.1, corrected for the magnetic field were applied to the
0.15, 0.20, and 1.0 cm, which spans the accepted range corrected reading to convert collected charge to dose.
of DLG values able to be entered into ClearCalc. The Dose/MU was also determined in the Monaco treatment
smallest difference between ClearCalc and Monaco was planning system using a homogenous phantom QA plan
set as the DLG value in the beam model. with a dose grid spacing of 3 mm and an uncertainty
The cryostat is a think pipe containing liquid helium in calculation of 1%. The QA plan was exported to
required for the design of the linear accelerator. It ClearCalc for analysis. The percent differences calcu-
requires a correction to characterize the output of the lated between the three dose/MU calculation methods
F I G U R E 1 Axial and Coronal views of the Zeus 0087Z heterogenous phantom for (a) a primary plan using a CT and (b) an adaptive plan
using an MR scanned on the unity.
had a tolerance of homogenous beams with high dose CTV volume to receive the prescribed dose. Adaptive
and low gradient settings set to 5%. plans were generated on Unity-generated MR scans
Measurements performed in a homogenous phan- and planned using an adaptive workflow. Although there
tom followed guidelines from TG-219 but also fulfilled was no change in anatomy, electron density was gener-
tests recommended by MPPG5a. This included, test 5.2: ated from contours on the MR scan. These tests were
dose in test plan versus reference calibration condition, performed on a primary plan that used a CT simula-
test 5.3: TPS data versus commissioning data, test 5.4: tion of the phantom as the primary imaging and multiple
small MLC-shaped field, test 5.5: large MLC-shaped field adaptive plans that used Unity-generated MR scans for
with extensive blocking, test 5.6: off -axis MLC shaped planning. Point doses and individual field dose averages
field, and test 5.7: asymmetric field at minimal antici- were reported.
pated SSD. Test 5.8, which tested a 10 × 10 cm2 field A non-clinical plan was generated on a heterogenous
at oblique incidence (> 20◦ ) was not measured but dataset to evaluate the accuracy of ClearCalc at var-
compared between Monaco and ClearCalc. ious locations in accordance with TG-219. Calculation
Point dose comparisons were calculated between points were set in tissue, superficial tissue, tissue-lung
treatment planning system and ClearCalc were tested interfaces, lung, and bone as well as set in uniform dose
for a homogenous phantom, a Zeus model 008Z (Sun or gradient regions and overall point doses comparisons
Nuclear Corporation) used without motion, for abdom- were reported. Finally, a set of non-clinical plans were
inal treatments. A primary plan was generated on a developed for various treatment sites and dose schema
planning CT dataset of the phantom, with the intended and tested prior to clinical implementation. Plans were
dose of 66 Gy in 30 fractions to a contoured liver CTV completed on the MR imaging set acquired from the
(Figure 1). Seven fields were used to cover 95% of the Unity on volunteers. Plans were generated to represent
TA B L E 1 Machine specific and calibration settings in ClearCalc

beam model.
Parameter Value
Machine energies 7X-FFF

Absolute dose reference field size (mm) 100
Absolute dose calibration Source to plane distance (mm) 1385
Absolute dose calibration depth (mm) 50
Reference dose at calibration depth (Gy) 1
Reference MU at calibration depth (MU) 100
MLC transmission factor 0.0055
Dosimetric leaf gap (cm) 0.010
Abbreviations: FFF, flattening filter free; MU, monitor unit.
clinical plans and calculations were performed as part of F I G U R E 2 Cryostat correction showing correction factor as a
the plan QA to mimic clinical workflow. Plan point doses function of gantry angle, and added as part of the ClearCalc beam
and individual field MUs were calculated and the differ- model.
ence between ClearCalc and Monaco were reported.
PDDs were simulated using the Monaco TPS due to

2.3 Clinical integration limitations in a measurement depth of 13 cm. A compar-
ison of PDD values at obtainable depths of 5, 10, and
ClearCalc was implemented into an adaptive treatment 13 cm was performed for equivalent square field sizes
workflow for the Unity MR-linear accelerator. In addi- of 2, 3, 5, 10, and 22 cm and showed a maximum devi-
tion to using ClearCalc for a secondary MU check for ation of 1.1% between measured and TPS-calculated
the primary plan, policy and procedures were written values (Table 2). Figure 2 shows a plot of the mea-
to perform ClearCalc calculations by physics prior to sured cryostat correction factor input into the ClearCalc
treatment for all adaptive plans. model as a function of gantry angle, normalized to a
A total of 41 patients with 215 adaptive plans were gantry angle of 0◦ . The maximum and minimum correc-
treated from October 2023 to July 2024 using the tion factor was 0.3% and 1.1%, respectively. MLC leaf
adaptive plan workflow on the Unity. Plans were primar- transmission was set to 0.055. The DLG optimization is
ily prostate (N = 27) but also included pelvic nodes shown in Table 3,in which dose/MU differences between
(N = 7), rectum due to prostatic recurrence (N = 1), Monaco and ClearCalc were completed for DLG settings
and prostate and lymph nodes (N = 8). The MUs neces- of 0, 0.01, 0.02, 0.1, 0.15, 0.2, and 1 cm. DLG of 0.01 cm
sary to achieve field’s measured dose were calculated showed the best agreement, and was used in the model.
in ClearCalc and compared to the treatment planning Dose plan comparisons between Monaco and
system for each field individually with a tolerance of ClearCalc are shown in Figure 3 for small (2 × 2 cm2 ),
5% or 5 MUs. In addition, the overall dose calculation medium (10 × 10 cm2 ), and large (50 × 22 cm2 ) field
was performed at appropriate calculation points and sizes and for cross-plane and in-plane directions. All
had a tolerance of 5% or 5 cGy. The user can gener- points in-field, defined by the 50% dose field edge,
ate points or use pre-determined points in ClearCalc for fell within the MPPG5a test 5.2 recommended 2%
these calculations. threshold. A 2-dimensional global gamma analysis
was performed using 5% dose difference or 5 mm
distance-to-agreement (DTA) criteria and with a 10%
3 RESULTS dose threshold. The gamma pass rate was 100% at
these settings for all profiles. The majority of the devia-
3.1 Beam model tion between Monaco and ClearCalc profiles was below
the 10% dose threshold.
All machine-specific and calibration settings and values
that were added to the Unity beam model in ClearCalc
are shown in Table 1. An absolute dose calibration of 3.2 Model validation
1 cGy per MU was completed at a reference field size
of 100 mm, an SSD of 138.5 cm, and a calibration A total of 27-point doses were used for validation, in
depth of 5 cm. The MLC transmission factor and DLG which dose was compared between ionization cham-
was measured to be 0.0055 and 0.01 cm, respectively. ber measurements in water, Monaco, and ClearCalc for
TA B L E 2 Comparison of measured PDDs and Monaco-derived PDDs added to the ClearCalc beam model.
5 cm depth 10 cm depth 13 cm depth

Field (mm) Measured ClearCalc Difference (%) Measured ClearCalc Difference (%) Measured ClearCalc Difference (%)
20 0.835 0.830 0.522 0.640 0.634 0.640 0.539 0.528 1.126

30 0.850 0.839 1.145 0.646 0.649 −0.338 0.547 0.538 0.869
50 0.861 0.857 0.390 0.672 0.674 −0.180 0.570 0.563 0.689
100 0.883 0.880 0.327 0.704 0.708 −0.353 0.606 0.602 0.364
220 0.885 0.885 0.029 0.724 0.724 −0.027 0.634 0.629 0.478
Abbreviation: PPDs, percent depth-doses.
F I G U R E 3 Beam profile comparison between ClearCalc and Monaco, along with gamma analysis results with gamma criteria of 5%/5 mm
for cross plane profiles for (a) 2 × 2 cm2 , (b) 10 × 10 cm2 , and (c) 50 × 22 cm2 field sizes and for in plane profiles for (d) 2 × 2 cm2 , (e)
10 × 10 cm2 , and (f) 50 × 22 cm2 field sizes.
TA B L E 3 Average dose difference [%] between the treatment ClearCalc and ionization chamber measurements of
planning system and ClearCalc for primary, adaptive, and overall for 2.7%. Although AAPM TG-219 recommended these
various leaf gap error settings in the ClearCalc model.
measurements, they also fulfilled specific guidelines
Gap error Dose difference— Dose difference— Average dose from MPPG5a, which test numbers 5.2–5.7 were specifi-
(cm) Primary (%) Adaptive (%) difference (%) cally denoted as subscripts in the “TPS-CC”column.Test
0.0 1.62 1.55 1.59 5.8, which compared a 10 × 10 cm2 field at an oblique
0.01 1.43 1.61 1.51 incidence of > 20◦ was not measured, but was planned
in Monaco and compared with ClearCalc and found to
0.02 1.49 1.70 1.58
match within 0.06%.
0.10 1.56 2.22 1.86
The performance of the ClearCalc model was subse-
0.15 1.73 2.65 2.15 quently tested on a heterogeneous phantom for both a
0.20 2.08 3.29 2.63 primary plan using a planning CT image set and for an
1.00 9.81 9.76 9.79 adaptive plan using an MR-scan from the Unity. Differ-
ences between ClearCalc and Monaco for the plan point
dose was 1.2% and 2.0% with an average (standard
clinical SSDs, field sizes, and for both on-axis and off - deviation) individual field MUs of 1.5% (2.2%) and 1.5%
axis points (Table 4). It was found that all points passed (1.5%) for the primary plan and adaptive plan, respec-
with the maximum deviation between ClearCalc and tively. All individual field MU differences fell within the set
Monaco of 3.5% and the maximum deviation between clinical tolerance of 5%/5 MUs.
TA B L E 4 Point dose verification comparing measurement (Meas)-, Monaco (TPS)-, and ClearCalc (CA)-determined point dose values for
various SSDs, field sizes, and point positions.
Jaw position Dose point Measured TPS ClearCalc Meas-TPS Meas-CA TPS-CA
SSD (cm) (X,Y) (x,y,z) (cGy) (cGy) (cGy) (cGy) (cGy) (cGy)
133.5 10,10 0,0,10 86.5 86.1 85.6 0.5 1.0 0.65.3

133.5 5,20 0,0,5 105.6 105.7 105.0 −0.1 0.6 0.75.7
133.5 5,20 0,5,5 96.3 97.0 95.6 −0.7 0.8 1.5
133.5 5,20 0,0,13 72.5 72.1 72.1 0.5 0.5 0.0
133.5 20,5 0,0,5 105.8 105.4 105.5 0.4 0.3 −0.1
133.5 20,5 −5,0,5 96.5 96.3 95.8 0.2 0.7 0.5
133.5 20,5 0,0,13 72.6 72.1 71.9 0.7 1.0 0.3
133.5 4,4 0,0,5 100.2 99.9 99.4 0.3 0.8 0.55.4
133.5 4,4 0,0,13 66.3 65.5 64.6 1.2 2.6 1.4
133.5 50,22 0,0,13 87.6 87.1 87.3 0.6 0.4 −0.25.5
133.5 50,22 12.5,0,13 65.4 64.9 63.7 0.8 2.6 1.9
133.5 50,22 0,5,13 80.9 80.6 79.3 0.4 2.0 1.6
133.5 50,22 −12.5,0,13 64.7 64.4 63.9 0.4 1.2 0.8
133.5 50,22 0,-5,13 81.8 80.7 79.3 1.4 3.1 1.8
138.5 10,10 0,0,5 100.5 100.2 100 0.3 0.5 0.25.2
138.5 [-10,20],10 15,0,5 65.8 64.8 66.8 1.5 −1.5 −3.05.6
138.5 [-10,20],10 16.5,0,8 55.5 54.9 56.7 1.1 −2.2 −3.2
138.5 [20,-10],10 −15,0,5 65.1 64.5 66.3 0.9 −1.8 −2.8
138.5 [20,-10],10 −16.5,0,8 54.8 54.3 56.4 1.0 −2.8 −3.8
143.5 2,2 0,0,1 94.7 95.6 95.7 −0.9 −1.0 −0.1
143.5 2,2 0,0,3 87.8 89.8 86.7 −2.3 1.3 3.5
143.5 3,3 0,0,1 100.1 99.4 97.4 0.7 2.7 2.0
143.5 50,22 0,0,1 116.7 115.8 117.7 0.8 −0.9 −1.6
143.5 50,22 0,5,3 103.3 104.0 105.2 −0.6 −1.8 −1.1
143.5 50,22 0,-5,3 104.4 104.0 105.1 0.3 −0.7 −1.1
143.5 50,22 12.5,0,3 82.8 82.7 83.5 0.2 −0.8 −1.0
143.5 50,22 −12.5,5,3 79.2 79.2 80.4 0.1 −1.4 −1.5
Abbreviation: SSDs, source-surface distances.
TA B L E 5 Dose differences between ClearCalc and Monaco for ance. In addition, preclinical plans were tested prior to
various point locations in a heterogenous non-clinical dataset. using ClearCalc clinically. Table 6 shows the ClearCalc
Dose Dose difference results for clinically deliverable plans for prostate,
Material distribution (%) prostate + dominant intraprostatic legion (DIL), partial
Tissue Uniform 0.57 brain, pelvic nodes, prostate and pelvic nodes, and liver
on a combination of CT and MR images. Individual field
Tissue—Superficial Gradient 3.79
MU differences and point doses are presented. All fields
Tissue-Lung Uniform 3.05
and dose points passed at a 5%/5MU tolerance.
Tissue-Lung Gradient 4.67
Lung Uniform 3.83
Lung Gradient 4.02 3.3 Clinical integration
Bone Uniform 4.83
Post commissioning, ClearCalc was implemented into
Bone Gradient 4.50
the planning and adaptive treatment planning workflow
for the Unity MR-linear accelerator. During planning,
ClearCalc was used by the physicist as a secondary
Dose points in various interfaces, material, and MU check software during the physics check. In the
isodose lines in a single heterogenous dataset are adaptive workflow used during each treatment fraction,
shown in Table 5. All points passed a 5%/5 MU toler- the adaptive plan was generated by the physics team
TA B L E 6 Pre-clinical plan comparison between ClearCalc and Monaco.
Individual fields
Total dose % Difference
Site (cGy) Fractions Field No. Imaging (%) Mean (%) Max (%) Min (%)
Prostate 4000 5 11 CT −0.8 −0.9 3.7 −3.4

Prostate 4000 5 11 MRI 2.9 −0.9 1.8 4.0
Prostate + DIL 4500 5 11 CT −0.2 1.2 3.8 −1.8
Prostate + DIL 4500 5 11 MRI 2.3 −1.7 4.1 −4.5
Partial brain 6000 30 11 CT 1.0 −1.8 0.3 −3.9
Partial brain 6000 30 11 MRI 0.7 0.0 2.0 −4.5
Prostate and pelvic nodes 2500 5 15 CT 1.5 −0.6 3.2 −3.7
Prostate and pelvic nodes 2500 5 15 MRI 3.9 −0.9 4.8 −3.9
Liver 6000 3 18 CT 3.0 −0.8 4.1 −4.2
Abbreviations: DIL, dominant intraprostatic legion; MRI, magnetic resonance imaging.
and approved by the physician. Upon approval but prior reference and adaptive plans, respectively. For the single
to beam on, the images, structure set, and individual rectum treatment, the average (standard deviation) dose
beam doses were exported to a folder on a shared difference was 2.4% and 2.6% (0.5%) and treatment
drive. ClearCalc configuration was set to parse the folder field difference was −0.4% (2.4%) and −1.2% (2.5%)
and perform a secondary MU calculation. Once the sec- for primary and adaptive plans, respectively. Finally, for
ondary MU calculation was complete and indicated a prostate and lymph node treatments, the average (stan-
passing result, physics gave the go-ahead to treat. dard deviation) dose difference was 2.0% (0.8%) and
The Unity linear accelerator went live with treatment 1.3% (1.0%) and treatment field difference was −0.3%
of primarily prostate cancers with a plan to expand to (2.4%) and −0.3% (2.4%) for primary and adaptive
other disease sites. The performance of ClearCalc’s plans, respectively.
agreement with Monaco was evaluated retrospectively
for the first 41 patients and 215 adaptive plans. Of these
plans, 27 were prostate stereotactic body radiation ther- 4 DISCUSSION
apy (SBRT) plans with dose schemas of (1) 40 Gy
in 5 fractions with a simultaneous subvolume boost of The purpose of this work was to share the author’s expe-
additional 5 Gy to the DIL, (2) 40 Gy in 5 fractions to rience with commissioning and benchmarking a beam
a prostate target volume, and (3) 25 Gy in 5 fractions model in ClearCalc for use as secondary dose calcu-
to a prostate target volume after an initial brachyther- lations on an MR-linear accelerator and to evaluate its
apy boost. In addition, seven cases treated the pelvic clinical performance. This work follows clinical imple-
oligometastatic lymph nodes with dose schemes of mentations guides from previous work for MR-linear
27 Gy in 3 fractions, 36 Gy in 6 fractions, and 30 Gy in accelerators using various programs,7–12 and particu-
5 fractions. One case was treated to the rectum with a larly based on guidelines from the AAPM.4,6 ClearCalc
dose scheme of 35 Gy in 5 fractions. Finally, eight cases is unique in that it implements a direct Lorentz correc-
were treated to the prostate and lymph nodes with dose tion factor to account for dosimetric impacts from the
schemes of 25 Gy in 5 fractions. magnetic field. The accuracy of this method was seen
For each reference and adaptive plan, ClearCalc’s particularly in the good agreement between ClearCalc
secondary dose/MU check for individual field MUs and and a benchmarked model of Monaco across all tests
overall plan point dose with a tolerance of 5% / 5MUs in this work. There are limitations that are associated
showed no failures. Figure 4 shows differences between with the FSPB dose calculation method that ClearCalc
ClearCalc and Monaco for field MUs and dose point utilizes as the algorithm balances speed of calculation
comparisons for each treatment site for both primary with dosimetric accuracy. The approximation of dose
plan and adaptive plans. For prostate SBRT treatments, deposition from the algorithm is limited in its ability to
the average (standard deviation) dose difference was model charged particles passing in and out of material
0.29% (1.3%) and 0.76% (1.0%) and treatment field interfaces. Table 5 shows the evaluation of ClearCalc’s
difference was −0.2% (2.5%) and −0.5% (2.5%) for pri- performance at tougher geometries, including interfaces
mary and adaptive plans, respectively. For pelvic node of bony anatomy and lung and at the edge of the
treatments, the average (standard deviation) dose differ- dose distribution where the gradient is larger. ClearCalc
ence was 1.0% (1.4%) and 0.9% (1.3%) and treatment showed agreement with Monaco under the 5%/5 MU
field difference was −0.5% (2.3%) and −0.4% (2.5%) for threshold. In addition, the limitations in the model were
F I G U R E 4 Boxplots showing the differences between ClearCalc and Monaco for clinical plans, separated into (a) individual fields for
primary plans, (b) dose points for primary plans, (c) individual fields for adaptive plans, and (c) dose points for adaptive plans.
also present in small fields, demonstrated in the devi- Table 1 and PDDs, but parameters like beam profiles,
ations in beam profiles, particularly in the penumbra, Sc and Sp individual values, MLC physical information,
between Monaco and ClearCalc for the 2 × 2 cm2 and the cryostat are all included in the model but not
cross-plane profile (Figure 2a). Although there were available for adjustment. The model generated with clin-
deviations in the beam profile, the homogenous dose ical beam data from commissioning was robust and
regions showed good agreement and overall dose/MU passed the clinical benchmark tests, but had they not,
calculations passed all commissioning tests for primary it may have warranted an investigation into the hard-
and adaptive treatment planning. This makes it a use- coded data. An example of this is when ClearCalc was
ful tool for secondary MU calculations. Graves et al.8 upgraded to version 2.4.6 from version 2.2.9, in which
also reported on these deviations in beam profiles and one adjustment was the addition of an institutionally
limitations with the Clarkson integration technique algo- unique cryostat correction instead of a manufacturer-
rithm and also concluded that these deviations showed determined cryostat correction. The difference between
little impact on the ability to provide a secondary dose the two cryostat corrections varied by up to 1.4%. Using
calculation using an appropriate calculation point. the same gantry angles for reference and adaptive clin-
Radformation offers a standard beam model for a ical plans, up to a 0.9% improvement was seen after the
Unity MR-linear accelerator. During initial implementa- adjustment.
tion, it was found that although the beam model worked During the adaptive plan workflow, the patient is on the
for the majority of test points in the homogenous water treatment couch and the speed of calculation is impor-
phantom (Table 4), there were specific points, particu- tant for a secondary calculation algorithm. Although no
larly at small field sizes, that exceeded the 5% threshold. timing data was obtained in this work, it was found
Ultimately, the authors used the institution-specific beam that ClearCalc performed the calculation rather quickly
model developed for Monaco supplemented with addi- and that the largest time sink was the plan export from
tional measurements. As a result, it is recommended to an online Monaco workstation to the ClearCalc server.
test the manufacturer’s standard beam model prior to Once the required data is sent to the parsed ClearCalc
implementing it into clinical use. folder, the specific treatment plan for the present date
When entering a beam model in ClearCalc’s beam shows up in the ClearCalc interface immediately. The
configuration workspace, there are few inputs open to user can then click on the plan and the calculation is
the user. The user enters the parameters included in performed. Two methods that the author’s implemented
to speed up this process is to routinely archive the This work commissioned and verified a beam model
ClearCalc dicom folder to move all files for finished treat- of a Unity MR-linear accelerator in ClearCalc, a third-
ments and to only export the body,bones,immobilization, party secondary dose/MU calculation software. After
couch and coil structures from the plan. It was found extensive testing on various phantoms and pre-clinical
that this amount of data in the export was sufficiently plans, ClearCalc was implemented into adaptive clinical
accurate calculations. However, more structure exports workflow, where it has performed highly with accurate
would yield a more accurate synthetic CT for calculation and speedy calculations. This work aims to guide other
but it was noticed that during the commissioning, some intuitions implementing a secondary dose/MU calcu-
calculations had trouble handling structures with compli- lation software in performing clinically oriented tests
cated topologies and overlaps. In addition, the external to evaluate performance while also fulfilling AAPM
designation from a structure was found to be critical in guidelines.
defining the overall CT boundaries.
One aspect that speeds up the calculation is that AU T H O R C O N T R I B U T I O N S
ClearCalc automatically finds appropriate points of All authors have contributed to this work. Sameer Taneja
interest saving the user from manually finding points. is the corresponding author and Ting Chen is the senior
ClearCalc first checks the primary reference point that is author.
denoted in planning, followed by the isocenter and then
the centroid of target volumes.If these points fail to meet AC K N OW L E D G M E N T S
criteria, ClearCalc will find a point that is > 3 mm from The authors would like to acknowledge the entire
tissue interfaces, > 1 cm from all field edges, within the physics and dosimetry teams at the MR linear accel-
90% prescribed dose region (can be loosened to 70%), erator team at NYU Langone Health as well as the
and a maximum point exposure by the field. In addition, ClearCalc team at Radformation for their support in the
the user can manually define points of interest for both presented work.
plan point dose and individual field dose. The process
of finding points and performing calculations were sim- C O N F L I C T O F I N T E R E S T S TAT E M E N T
ilar for all types of plans, from more complex and larger The authors declare no conflicts of interest.
fields in pelvic and lymph node plans to less modulated
pelvic node plans. It was noticed that the main difference REFERENCES
1. Stern RL, Heaton R, Fraser MW, et al.;AAPM Task Group
in time between overall analysis was from the export.
114. Verification of monitor unit calculations for non-IMRT clin-
Prior to performing dose/MU calculation, ClearCalc ical radiotherapy: report of AAPM Task Group 114. Med Phys.
allows the user to edit the structure layers for the calcu- 2011;38(1):504-530. doi:10.1118/1.3521473
lation. As only the body, bones, and support structures 2. Purdy JA. Intensity-modulated radiotherapy: current status and
were exported for calculation, the structure layering did issues of interest. Int J Radiat Oncol Biol Phys. 2001;51(4):880-
not impact the calculation. However, if additional struc- 914. doi:10.1016/S0360-3016(01)01749-7
3. Gibbons JP, Antolak JA, Followill DS, et al. Monitor unit calcu-
tures were exported, it would be recommended to use lations for external photon and electron beams: report of the
the same structure layering as used for the plan. For AAPM Therapy Physics Committee Task Group No.71.Med Phys.
adaptive plans, there is no reference CT data, which 2014;41(3):031501. doi:10.1118/1.4864244
prompts the user to use the structure set with HU over- 4. Zhu TC, Stathakis S, Clark JR, et al. Report of AAPM Task Group
219 on independent calculation-based dose/MU verification for
rides. As a result, exporting the external structure is
IMRT. Med Phys. 2021;48(10):5713. doi:10.1002/mp.15069
critical for accurate calculation. 5. Fraass B, Doppke K, Hunt M, et al. American association of
It is recommended to perform extensive testing prior physicists in medicine radiation therapy committee task group
to clinical implementation to test the robustness of the 53: quality assurance for clinical radiotherapy treatment planning.
beam model. For example, a series of failing clinical Med Phys. 1998;25(10):1773-1829. doi:10.1118/1.598373
6. Geurts MW, Jacqmin DJ, Jones LE, et al. AAPM MEDICAL
plans were able to isolate an incorrect SSD value in the
PHYSICS PRACTICE GUIDELINE 5.b: commissioning and QA
initial beam model’s calibration setting, which was set to of treatment planning dose calculations—megavoltage photon
an SSD of 133.5 cm.Based on homogenous and hetero- and electron beams. J Appl Clin Med Phys. 2022;23(9):e13641.
geneous phantom testing and pre-clinical plans, it was doi:10.1002/acm2.13641
found that the tolerance of 5% or 5 MUs was appropri- 7. Price AT, Knutson NC, Kim T, Green OL. Commissioning a sec-
ondary dose calculation software for a 0.35 T MR-linac. J Appl
ate based on the level of calculation precision. Clinically,
Clin Med Phys. 2022;23(3):e13452. doi:10.1002/acm2.13452
no failing plans have been identified. 8. Graves SA, Snyder JE, Boczkowski A, et al. Commissioning and
performance evaluation of RadCalc for the Elekta unity MRI-
linac. J Appl Clin Med Phys. 2019;20(12):54-62. doi:10.1002/
5 CONCLUSION acm2.12760
9. Raaymakers BW, Jürgenliemk-Schulz IM, Bol GH, et al. First
patients treated with a 1.5 T MRI-Linac: clinical proof of concept
Independent dose/MU calculations are critical as a sec- of a high-precision, high-field MRI guided radiotherapy treat-
ond check for the treatment planning system accuracy ment. Phys Med Biol. 2017;62(23):L41-L50. doi:10.1088/1361-
and is especially critical for high dose SBRT treatments. 6560/aa9517
10. Hackett S, van Asselen B, Feist G, et al. SU-F-J-148: a col- 17. Malkov VN, Rogers DWO. Monte Carlo study of ionization cham-
lapsed cone algorithm can be used for quality assurance ber magnetic field correction factors as a function of angle and
for Monaco Treatment Plans for the MR-Linac. Med Phys. beam quality. Med Phys. 2018;45(2):5367-5370. doi:10.1002/mp.
2016;43(6Part11):3441-3441. doi:10.1118/1.4956056 12716
11. Pollitt A, Budgell G, Pooler A, Wood J, Chuter R, McWilliam 18. O’Brien DJ, Roberts DA, Ibbott GS, Sawakuchi GO. Reference
A. EP-2158: feasibility of Mobius 3D as an independent MU dosimetry in magnetic fields: formalism and ionization chamber
checker for the adaptive work flow on the MR-Linac. Radio- correction factors. Med Phys. 2016;43(8):4915-4927. doi:10.1118/
ther Oncol. 2018;127:S1191-S1192. doi:10.1016/s0167-8140(18) 1.4959785
32467-8 19. Zhu TC, Ahnesjö A, Lam KL, et al.; AAPM Therapy Physics
12. Li Y, Wang B, Ding S, et al. Feasibility of using a commercial col- Committee Task Group 74. Report of AAPM Therapy Physics
lapsed cone dose engine for 1.5T MR-LINAC online independent Committee Task Group 74: in-air output ratio„ for megavoltage
dose verification. Phys Medica. 2020;80:288-296. doi:10.1016/j. photon beams. Med Phys. 2009;36(11):5261-5291. doi:10.1118/
ejmp.2020.11.014 1.3227367
13. Yang J, Zhang P, Tyagi N, et al. Integration of an Independent 20. Tsuneda M, Abe K, Fujita Y, Ikeda Y, Furuyama Y, Uno T. Elekta
Monitor Unit Check for High-Magnetic-Field MR-guided Radia- Unity MR-linac commissioning: mechanical and dosimetry tests.
tion Therapy System. Front Oncol. 2022;12:747825. doi:10.3389/ J Radiat Res. 2023;64(1):73-84. doi:10.1093/jrr/rrac072
fonc.2022.747825
14. Raaymakers BW, Lagendijk JJW, Overweg J, et al. Integrating a
1.5 T MRI scanner with a 6 MV accelerator: proof of concept.
Phys Med Biol. 2009;54(12):N229-N237. doi:10.1088/0031-9155/ How to cite this article: Taneja S, Wang H,
54/12/N01
15. IAEA. Absorbed Dose Determination in External Beam Radio-
Barbee DL, et al. Commissioning and
therapy An International Code of Practice for Dosimetry Based implementation of a pencil-beam algorithm with a
on Standards of Absorbed Dose to Water. Tech Rep Ser No 398. Lorentz correction as a secondary dose
IAEA; 2001;2006 (June). calculation algorithm for an Elekta Unity 1.5T MR
16. Tyagi N, Subashi E, Michael Lovelock D, et al. Dosimetric evalu- linear accelerator. J Appl Clin Med Phys.
ation of irradiation geometry and potential air gaps in an acrylic
miniphantom used for external audit of absolute dose calibra-
2024;e14590.
tion for a hybrid 1.5 T MR-linac system. J Appl Clin Med Phys. https://doi.org/10.1002/acm2.14590
2022;23(2):e13503. doi:10.1002/acm2.13503

J Applied Clin Med Phys - 2024 - Taneja - Commissioning and implementation of a pencil‐beam algorithm with a Lorentz

Uploaded by

Copyright:

Available Formats

J Applied Clin Med Phys - 2024 - Taneja - Commissioning and implementation of a pencil‐beam algorithm with a Lorentz

Uploaded by

Document Information

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

J Applied Clin Med Phys - 2024 - Taneja - Commissioning and implementation of a pencil‐beam algorithm with a Lorentz

Uploaded by

Copyright:

Available Formats

Received: 1 August 2024 Revised: 21 October 2024 Accepted: 12 November 2024

Commissioning and implementation of a pencil-beam

J Appl Clin Med Phys. 2024;e14590. wileyonlinelibrary.com/journal/acm2 1 of 11

1 INTRODUCTION mented with Unity accelerators. Yang et al.13 used an

TA B L E 1 Machine specific and calibration settings in ClearCalc

Machine energies 7X-FFF

PDDs were simulated using the Monaco TPS due to

5 cm depth 10 cm depth 13 cm depth

20 0.835 0.830 0.522 0.640 0.634 0.640 0.539 0.528 1.126

133.5 10,10 0,0,10 86.5 86.1 85.6 0.5 1.0 0.65.3

TA B L E 6 Pre-clinical plan comparison between ClearCalc and Monaco.

Prostate 4000 5 11 CT −0.8 −0.9 3.7 −3.4

You might also like