Received: 11 June 2024 Revised: 29 July 2024 Accepted: 19 November 2024

DOI: 10.1002/acm2.14598

T E C H N I C A L N OT E

New gantry angle-dependent beam control optimization

with Elekta linear accelerator for VMAT delivery

Adriaan Abraham van Appeldoorn Johannes Gerardus Maria Kok

Jochem Willem Heiko Wolthaus

Department of Radiotherapy, University Abstract

Medical Center Utrecht, Utrecht, Netherlands
Introduction: This paper describes a method to improve gantry-dependent
Correspondence
beam steering for Elekta traveling wave linear accelerators by applying the mea-
Jochem Willem Heiko Wolthaus, Department sured and filtered beam servo corrections to the existing lookup table (LUT).
of Radiotherapy, University Medical Center Beam steering has a direct influence on the treatment accuracy by affecting
Utrecht, Heidelberglaan 100, 3584 CX,
Utrecht, The Netherlands.
the beam symmetry and position. The presented method provides an improved
Email: [email protected] LUT with respect to the default Elekta method to reduce treatment delivery inter-
ruptions. These interruptions are known to contribute to unwanted intrafraction
motion and longer treatment times.
Methods: Compared to the default method of the manufacturer, this new
method takes both clockwise and counterclockwise rotation to compensate for
magnetic hysteresis as well as previous configuration and noise filtering into
account. The improved method to determine the lookup table uses service
graphing information from the linac without the need for additional symmetry
information. The clinical configuration of the flattened beam energies remains
untouched during the data record.
Results: This method results in a configuration where the gantry-dependent
steering is optimized over the full arc with optimal balance in the hysteresis and
minimizing the effect of errors in the steering values. This method is a less error-
prone process compared to the methodology described in previous research,
still achieving a reduction of interruption of about 60 percent compared to the
Elekta method.
Conclusion: This study shows a simplified way to optimize linac stability with
improved LUT. The optimized LUT results in a lower number of interruptions,
preventing downtime, and a lower risk of intrafraction motion due to longer
treatment time.

KEYWORDS
2T and 2R, beam physics, beam symmetry, Elekta accelerators, linac lookup table, radiotherapy,
servo mechanism, VMAT

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 American Association of Physicists in Medicine.

J Appl Clin Med Phys. 2025;26:e14598. wileyonlinelibrary.com/journal/acm2 1 of 7

https://doi.org/10.1002/acm2.14598
 15269914, 2025, 3, Downloaded from https://aapm.onlinelibrary.wiley.com/doi/10.1002/acm2.14598, Wiley Online Library on [31/03/2025]. 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 7 VAN APPELDOORN ET AL.

1 INTRODUCTION and is shown in Figure 1. The first parameter is a static

“Set value” of the 2R-2T current to aim for. Secondly, a
Modern treatment techniques, like IMRT and VMAT, correction for the external magnetic field variation per
push the steering abilities of conventional linear accel- gantry angle, which is stored in a LUT. This LUT has 128
erators (linac) to the limits of their internal tolerances. positions that represent the full potentiometer range of
For safety reasons, radiation beam interruptions occur the readout system, leading to a less than a hundred
if deviation in beam steering exceeds the system toler- positions for the full clinical arc of 360 degrees. When
ances. The beam steering system of the linac accounts the gantry of the linac rotates, the vector of the mag-
for changes in magnetic fields experienced by the netic field changes, as well as the linac magnets and
beamline. These magnetic fields depend on chang- metal construction that influences its residual magnetic
ing magnetic fields in their neighborhood, for example, field. These changes in the magnetic field are static per
due to adjacent MRI systems.1,2 Correct beam steering gantry angle for the LUT configuration. As every position
means the electron beam on the right place and right in the LUT represents a part of the gantry arc (about
angle on the target, resulting in a well-positioned and four degrees), plotting the LUT values will show a sinu-
symmetrical radiation beam. For the Elekta system stud- soidal shape due to the circular rotation of the gantry
ied here, a beam interruption occurs if the measured against the dominant surrounding magnetic field vector
asymmetry (or tilt) exceeds 5% or the associated servo (typically related to the geomagnetic field).
steering current exceeds 50 mA, which is a set by the The third contribution is the “servo value”, which is
manufacturer. derived from the ion chamber tilt signal. This active com-
Generally, these types of treatment interruptions can ponent compensates for ‘slow’ non-deterministic effects
be resumed after reset by authorized persons (e.g., RTT, by adapting the steering current such that the error
physicist, or engineer). The clinical impact of steering- based on the readings of the ion chamber steering
related interruptions is the extension of overall treatment plates is minimal.
times, thereby increasing the effect of intrafraction The three parameters, Set value, LUT value, and servo
motion or, in extreme cases, even leading to treatment value, can be retrieved using the service mode of the
delay. linac control software (Integrity 4.0).
The default manufacturer’s method to calibrate the Besides the slow non-deterministic effects, there is
gantry-dependent steering uses data from a single also a small difference in the compensating effects for
gantry rotation direction only and has no additional the clockwise (CW) and counter-clockwise (CC) gantry
data processing to remove noise or outliers. Previous directions due to magnetic hysteresis. This hystere-
research3 showed a reduction of beam interruptions sis is based on different contributions like previously
for the Elekta SL25 traveling wave linear accelerators applied steering currents, magnetic linac components
(Elekta AB, Sweden). This was achieved by including (type of steel, ion pumps, and magnetron), Earth mag-
both rotation directions and smoothing in the correction. netic field, and nearby MRI systems. As a result, the
In that research, also an additional input factor had to errors observed during CC rotation are higher using the
be included, and beam symmetry steering servo mech- manufacturer’s default method. Creating an LUT based
anisms had to be disabled. This new study describes on data from a single moment may result in an unbal-
a simplified method based on servo mechanism con- anced LUT, which may lead to degradation compared to
tribution; however, without the need for an additional the existing LUT.The inclusion of steering data with infor-
input factor therefore being more efficient, less risky, mation about both directions of rotation,previous control
simpler, and more equivalent to the default method parameters, and applying filtering, these shortcomings
of the manufacturer. Unlike the default method, this can be overcome.
method includes background information on both rota- This study describes a new robust method to improve
tion directions, previous control parameters (to account the linac beam steering lookup tables and provides
for periodic magnetic changes in the environment), and example results demonstrating a more stable radiation
mathematical data fit to avoid anomalous or missing beam.
data points that may occur in practice.
Symmetry of the radiation beam is monitored with the
ion chamber mounted in the radiation head of the linac. 2 METHODS
This ion chamber has several electrodes, two of which
are arranged such that the signal difference between The steering current (Isteering ) is the sum of the Set
the electrodes reflects the degree of asymmetry in the value and the scaled LUT and servo values (Equation 1).
beam (tilt). This tilt is called “2R error” (Radial) for inline The servo value is zero at beam start. Thereafter, any
and “2T error” (Transverse) for crossline direction. detected asymmetry (tilt) causes the servo Value to be
The control mechanism for the 2R-2T secondary non-zero,thus changing the sum of the three inputs.This
beam steering current requires three input parameters new Isteering value changes the trajectory of the beam
 15269914, 2025, 3, Downloaded from https://aapm.onlinelibrary.wiley.com/doi/10.1002/acm2.14598, Wiley Online Library on [31/03/2025]. 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
VAN APPELDOORN ET AL. 3 of 7

F I G U R E 1 A simplified graphical overview of the steering mechanism of the Elekta linear accelerator (Elekta AB, Sweden). A secondary
set of two steering coils (2R and 2T) are located halfway down the waveguide (accelerating structure). With the correct current through these
steering coils, the electron beam is aligned such that the target is hit at the correct angle and a symmetrical radiation beam is produced. The
actual steering currents are set by three parameters: The base value (Set value), a LUT, and the value of the servo mechanism. The LUT is the
primary correction to compensate for the external magnetic distortion per gantry angle. The servo mechanism applies a secondary correction
by controlling the coil current based on the readout of the steering plates of the integrated ion chamber. 2R, "2R error” (Radial); 2T, “2T error”
(Transverse); LUT, lookup table.

path, producing a new tilt signal, and so on until the tilt plified compared to previous work.3 Figure 2 shows a
signal is zero. comparison of the default Elekta procedure with the
newly proposed procedure.
Calibration gain
Isteering = Set value + LUT value ∗
2048
2.1 Creating data using the service
Calibration gain graph
+ Servo value ∗ (1)
2048
As input for the presented LUT optimization method,
If these first two parameters of the control mechanism a dataset of 2R and 2T current control values are
(Set value and LUT) together lead to a deviation from required, which is the confirmed value of steering cur-
the desired Isteering value, there is an increased chance rent (Isteering ), actual LUT position, actual LUT value, and
of beam interruptions during treatment due to symmetry the servo value (Table 1). A 40 × 40 cm2 open field is
errors. used, and the dose rate is fixed to half of the maximum.
The default LUT-creating method, as described in the This dataset is recorded using the service graph func-
Elekta manuals,4,5 is sensitive to noise and spurious tionality in the linac control software (Integrity) at the
readings. In this procedure, the gantry rotates CW maximum sampling frequency (4 Hz) for a full arc of CW
from −183◦ to + 183◦ during beam on. The control and CC gantry rotation. Noise is reduced by recording
system sets up the gantry-dependent beam control multiple measurement samples for each LUT position
values by storing the feedback values from the servo (between four and seven samples, depending on the
mechanism in the corresponding LUT. gantry rotation speed). To ensure sufficient data points,
The improved method to generate the beam steering the gantry rotation speed is reduced by a temporarily
LUT extends the Elekta default procedure but is sim- low automatic system unit (ASU) speed and includes
 15269914, 2025, 3, Downloaded from https://aapm.onlinelibrary.wiley.com/doi/10.1002/acm2.14598, Wiley Online Library on [31/03/2025]. 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
4 of 7 VAN APPELDOORN ET AL.

F I G U R E 2 Flowcharts of the Elekta and newly proposed method to acquire a LUT for gantry-dependent beam steering. The default Elekta
procedure takes fewer dependencies into account compared to the new method. LUT, lookup table.

TA B L E 1 Items and part numbers settings that are used to record the input LUT for our method.

Item Part Description

X-axis: 148 4 Gantry angle confirmed value

Y-axis: 164 4 2R I control confirmed value
153 2R I control LUT position on specific gantry angle
209 2R I control servo mechanism value, integrity 4+
211 2R I control LUT value on specific gantry angle, integrity 4+
165 4 2T I control confirmed value
153 2T I control LUT position on specific gantry angle
209 2T I control servo mechanism value, integrity 4+
211 2T I control LUT value on specific gantry angle, integrity 4+
Abbreviations: 2R, “2R error” (Radial); 2T, “2T error” (Transverse); LUT, lookup table.

samples outside the clinical gantry angle range (±183 beam. Recording the data takes about five minutes.
degrees). Afterward, the data is exported as a service graph
Before starting the actual data recording, a beam (.xml) to a network drive. With a developed MATLAB®
warmup of approximately ten seconds using a (The MathWorks, Inc., Natick, USA) application, the
40 × 40 cm2 field is applied to obtain a stable radiation recorded data is processed to create an optimized LUT.
 15269914, 2025, 3, Downloaded from https://aapm.onlinelibrary.wiley.com/doi/10.1002/acm2.14598, Wiley Online Library on [31/03/2025]. 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
VAN APPELDOORN ET AL. 5 of 7

2.2 Processing the data angles, the reduction of interruptions for the current
method did not change.
First, the position in the lookup table is mapped to
the gantry angle, yielding an average gantry angle
(generally halfway) for each discrete LUT position. Sub- 4 DISCUSSION
sequently, the servo-based (correction) value and the
actual value (LUTactual ) for each LUT position are deter- We have described a method of generating a LUT that
mined. To derive an optimal LUT, applicable to both provides operation with reduced beam interruptions.
rotation directions, the combined (average) servo con- This result is obtained by reducing the effect of linac
tribution includes the measured data of both rotation calibration and ambient magnetic fields from the data
directions (which differs due to hysteresis). acquisition process. Our implementation of this tech-
To maintain consistency in clinical linac behavior, nique shows an approximate 60 percent decrease in
reducing effects of day-to-day measurement variation symmetry-related beam interruptions during treatment
(e.g., magnetic field fluctuation at the moment of record) deliveries.
and preventing over-correction, only a fraction of the full
correction is applied to the actual existing LUT as shown
in Equation 2. 5 COMMENTS FOR PRACTICAL
CLINICAL IMPLEMENTATION
LUTraw = LUTactual + combined servo value ∗ fraction
The method to create hysteresis-compensated and
(2) noise-free lookup tables is only valid when both servo
mechanisms are switched on. Older linacs, as well as
This fraction value can be used to balance between new systems for the flattening filter free (FFF) beams,
new and existing data. Note that a full correction (i.e., have the 2T servo switched off by default. The servo
fraction = 1) is used initially for a new installation or when system should then be switched on temporarily prior to
a systematic change in the LUT is expected. performing the presented LUT procedure and needs to
To avoid the propagation of noise spikes in the new be switched off later before clinical use of the linac.
LUT, the LUTraw data (Equation 2) are fitted to a sinu- At the time of the previous study,3 the linac control
soidal fit function as a function of the gantry angle, system could not record the LUT values and servo val-
consistent with the theoretically expected behavior due ues with service graphing according to Table 1. The
to the gantry rotation.The fit function model consists of a current method eliminates the need for an additional
sum of five sinusoids to account for the inhomogeneity input variable, such as tilt error correlation. Therefore,
in the magnetic field. In addition, the fit function model this method is not impacted by the specific linac cal-
allows extrapolation of the data over angles beyond ibration and surrounding magnetic field. Compared to
those that were acquired. This will result in LUT values the previous method, which experienced a delay in the
at all clinical gantry angles, even if the input data do not tilt data of the ion chamber due to filtering, the current
contain all gantry angles. input (servo value) has no delay in the readout. Con-
To obtain optimal functionality for regular beam tuning sequently, no data shifting is necessary for the current
on gantry zero,the new LUT data can be normalized with method.In summary,this method is faster to perform and
a set value equal to the observed Isteering with the gantry less error-prone compared to the default method and the
at 0 degrees. previous method where an additional input parameter is
required.
These data should be recorded with a single fixed
3 RESULTS (not varying) dose rate to avoid any other dependencies
or correlations than to gantry angle. According to the
Results of the described procedure are compared to current Elekta manual,5 it is advised to perform (most)
results of the default Elekta method (“Default Method”) beam-tuning procedures at half of the maximum dose
and results based on the procedure of previous work3 rate. If the system is used for dynamical treatments with
(“Previous Method”). The results from this study, with variable dose rates, such as VMAT, the beam symmetry
the servo mechanisms enabled, are called the “Current behavior should be verified with consideration of all
Method.” These results are shown in Figure 3. dose rates by adjusting the gantry averaged beam tilt
These results of the previous and current method are using the set value. Due to magnetic field fluctuation, it is
very similar regarding the servo value and symmetry. recommended to avoid overcompensation by applying
The previous method had shown a reduction of about a correction factor as mentioned in the Method section.
60 percent of beam interruptions3 and based on simi- In the long-run, it is seen that fewer adjustments were
lar symmetry and the servo values over all the gantry needed.
 15269914, 2025, 3, Downloaded from https://aapm.onlinelibrary.wiley.com/doi/10.1002/acm2.14598, Wiley Online Library on [31/03/2025]. 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
6 of 7 VAN APPELDOORN ET AL.

F I G U R E 3 An example of the effect of the methods to create a beam steering LUT. These verification graphs are normalized at gantry
angle zero to compare only the gantry dependency, where the solid line is in CW- and the dashed line in CC direction. The graph on top show
the applied LUT for each method. In the middle, the required servo values are shown. It can be seen that the default method, shows the most
variation. At the bottom, the graphs show a verification with steering servos disabled, whereas the default method shows a more variable tilt.
Beams can expect such a tilt for each gantry angle at startup of the beam due to the LUT. The input for optimization, the servo value shows a
proportional relation to the Tilt error which is used as input for optimization in previous work.3 The “Previous method” (red) and “Current Method”
(green) shows a broad similarity, in all the three subgraphs and improved behavior in the tilt or servo current. CC, counter clockwise; CW,
clockwise; LUT lookup table.
 15269914, 2025, 3, Downloaded from https://aapm.onlinelibrary.wiley.com/doi/10.1002/acm2.14598, Wiley Online Library on [31/03/2025]. 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
VAN APPELDOORN ET AL. 7 of 7

According to the previous work,3 suboptimal beam AU T H O R C O N T R I B U T I O N S

steering has just a limited effect on the dose deposition, Adriaan Abraham van Appeldoorn developed the the-
especially when the servo mechanisms are enabled. ory and performed the measurements and analysis.
The real benefits can be found in the reduction of Johannes Gerardus Maria Kok and Jochem Willem
interruptions on the linac for the clinic, where the pre- Heiko Wolthaus verified the analytical methods and
vious work has shown a reduction of the number of contributed to the design and implementation of the
interruptions of approximately 60 percent. However, research. Jochem Willem Heiko Wolthaus supervised
delivery improvements can be expected for treatment the findings of this work. All authors discussed the
plans with a high number of segments with low dose, results and contributed to the final manuscript.
with high variation in dose rate, and for gating where
a correct beam startup will have a significant effect on AC K N OW L E D G M E N T S
the total dose delivery. The authors have nothing to report.
After applying the new LUT, gantry-dependent veri-
fication should be performed. If multiple energy con- 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
figurations (e.g., FFF and electron energies according The authors declare no conflicts of interest.
to the Elekta manual5 ) are using the same LUT, dose
delivery for these energy configurations has to be ver- ORCID
ified as well. This verification can be done by verifying Adriaan Abraham van Appeldoorn
the gantry-dependent steering and symmetry parame- https://orcid.org/0009-0005-5635-9041
ters from the linac with the service graphing functionality
or using an external QC device, like a 2D panel dose REFERENCES
detector mounted on the gantry head. 1. Kok JGM, Raaymakers BW, Lagendijk JJW, Overweg J, de Graaff
Using this LUT procedure, incorporation of linac sta- CHW, Brown KJ. Installation of the 1.5 T MRI accelerator next
bility and dosimetry parameters or optional external to clinical accelerators: impact of the fringe field. Phys Med
Biol. 2009;54(18):N409-N415. https://doi.org/10.1088/0031-9155/
QC devices is possible. This will help prevent down- 54/18/N02
time and optimize existing QA procedures. Utilization 2. Perik T, Kaas J, Wittkämper F. The impact of a 1.5 T MRI linac
of these combined parameters and tools will result fringe field on neighbouring linear accelerators. Phys Imaging
in a time-efficient QC session, a good check of the Radiat Oncol. 2017;4:12-16. doi:10.1016/j.phro.2017.10.002
machine performance, and a useful reference. Most 3. van Appeldoorn AA, Kok JGM, Wolthaus JWH. Gantry angle
dependent beam control optimization of a traveling wave linear
importantly, this can further reduce the number of beam accelerator to improve VMAT delivery. J Appl Clin Med Phys.
interruptions during clinical use. 2020;21(11):312-321. doi:10.1002/acm2.13047
4. Corrective Maintenance Manual—Beam Physics System. Elekta
Limited; 2011. (45133702098_04)
6 CONCLUSION 5. Corrective Maintenance Manual—Beam Physics System. Elekta
Limited; 2020. (45133702098_08)

This study describes a simplified method to improve

beam steering by deriving an optimal lookup table. The
new method is faster, less error-prone, and has fewer
dependencies on other linac settings. As a result, the How to cite this article: van Appeldoorn AA,
linac stability is increased in terms of reduced beam Kok JGM, Wolthaus JWH. New gantry
interruptions. Similar to previous work,3 this method angle-dependent beam control optimization with
reduces downtime and extension of the treatment time Elekta linear accelerator for VMAT delivery. J
caused by these beam interruptions. The clinical ben- Appl Clin Med Phys. 2025;26:e14598.
efit of shorter treatment times relates to the reduced https://doi.org/10.1002/acm2.14598
intrafraction motion of the target volume.