Home Search Collections Journals About Contact us My IOPscience

Daily monitoring of linear accelerator beam parameters using an amorphous silicon EPID

This article has been downloaded from IOPscience. Please scroll down to see the full text article.

2007 Phys. Med. Biol. 52 1721

(http://iopscience.iop.org/0031-9155/52/6/012)

View the table of contents for this issue, or go to the journal homepage for more

Download details:
IP Address: 137.207.120.173
The article was downloaded on 24/07/2013 at 01:28

Please note that terms and conditions apply.

IOP PUBLISHING PHYSICS IN MEDICINE AND BIOLOGY

Phys. Med. Biol. 52 (2007) 1721–1733 doi:10.1088/0031-9155/52/6/012

Daily monitoring of linear accelerator beam

parameters using an amorphous silicon EPID
G J Budgell, R Zhang and R I Mackay
North Western Medical Physics, Christie Hospital NHS Trust, Withington, Manchester,
M20 4BX, UK

Received 6 September 2006, in final form 20 December 2006

Published 27 February 2007
Online at stacks.iop.org/PMB/52/1721

Abstract
An amorphous silicon EPID has been investigated to test its suitability as a
daily check device for linac output and to provide daily monitoring of beam
profile parameters such as flatness, symmetry, field size and wedge factor.
Open and wedged 6 and 8 MV photon beams were collected on a daily basis
for a period of just over a year and analysed in software to determine daily
values of these parameters. Daily output results gave agreement between EPID
measured dose and ion chamber measurements with a standard deviation of
0.65%. Step changes in flatness, symmetry and field size were readily detected
by the EPID and could be correlated with adjustments made on service days
and QC sessions. The results could also be used to assess the long term
beam stability. Recalibration of the EPID required new baseline values of the
parameters to be set. Wedge factors measured at one collimator angle proved
stable but sensitive to changes in beam steering. The EPID proved to be a
useful daily check device for linac output which can simultaneously be used
for daily monitoring of beam profiles and field sizes.

Introduction

Electronic portal imaging devices (EPIDs), although primarily designed for geometric
verification of patient position, have also been shown to be useful for a variety of linear
accelerator (linac) quality control (QC) procedures. These include checks of field size,
light/radiation field congruence, beam profile flatness and symmetry, ‘starshot’ films and
isocentre checks (Dirkx et al 1995, Kirby and Williams 1995, Luchka et al 1996, Curtin-
Savard and Podgorsak 1997, Dunscombe et al 1999, Liu et al 2000, 2002, 2004, Vieira et al
2002, Prisciandaro et al 2003, Chang et al 2004, Samant et al 2002, Yang and Xing 2004).
Most of this work has been carried out using fluoroscopic or liquid-filled ion chamber EPIDs
rather than the newer generation of amorphous silicon (amSi) flat panel devices hence there is
a need to investigate the usefulness of amSi EPIDs for such applications. We have previously
shown that amSi EPIDs are useful tools for routine quality control and calibration of MLCs
0031-9155/07/061721+13$30.00 © 2007 IOP Publishing Ltd Printed in the UK 1721
1722 G J Budgell et al

(Baker et al 2005), for daily checks of MLC position and for routine checks of beam flatness
and symmetry of the low dose segments used in step and shoot IMRT (Budgell et al 2005).
The dosimetric properties of the devices are also now being reported (McCurdy et al 2001,
Greer and Popescu 2003, McDermott et al 2004, Winkler et al 2005).
In this paper an investigation of the usefulness of an amSi EPID for daily monitoring of
linac beam parameters is described. Measurements were made on one linac over a 1 year
period and used to test the ability of the EPID to replace daily output checks, and to monitor
beam profiles, field size and wedge position.

Methods

All measurements were carried out using an iViewGT amSi EPID mounted on a dual energy
(6 and 8 MV) Precise linac (Elekta Oncology Systems, Crawley, UK). This EPID has a
resolution of 1024 × 1024 pixels within a 40 × 40 cm field of view mounted at a fixed
distance of approximately 160 cm from the focal spot. Hence the pixel size projected back to
isocentre is very close to 0.25 mm—all sizes given in this paper will be in the isocentric plane.
A fuller description of the properties of this EPID is given elsewhere (Winkler et al 2005). All
measurements were carried out using the EPID in its default position—centred on the central
axis of the beam and without any additional build-up. The inherent water-equivalent build-up
of this EPID has been measured to be around 5 mm.
Four images were taken each day, delivered as two beams of 6 MV and 8 MV respectively.
Each beam consisted of an open and wedged segment of 100 monitor units (MU) each. 100 MU
was used only for consistency with normal QC, lower exposures could readily be used to reduce
total dose to the radiation-sensitive panel electronics. A field size of 24 × 24 cm was used in
order to fit comfortably within the 26 × 26 cm isocentric field of view. The standard clinical
image acquisition software was used, with images being added to the same patient entry in
the database each day. The set-up and acquisition of the images took less than 5 min and was
normally carried out at the start of the day; hence any drift in EPID output during the day
(Winkler et al 2005) will not affect the results presented. Images were taken on a daily basis
for a whole year except for the monthly service days and an occasional missed day.
A number of standard corrections are made to images which are acquired using the clinical
software:
• A dark current (offset) calibration which is continuously acquired whilst not imaging.
Hence a very recent offset calibration is always used.
• Each image is divided by a ‘flood-field’ correction—a raw image of an open beam covering
the whole field of view. This eliminates variation in sensitivity between different pixels
and areas of the panel and also has the effect of ‘flattening out’ the horns of the measured
profiles (Greer and Popescu 2003, Van Esch et al 2004).
• A bad-pixel map which replaces the signal given by pixels known to give inconsistent
responses (e.g. nonlinear gain, dead pixels) with an average signal from surrounding
pixels.
Images are acquired as a series of frames, continuously integrated into a 64 bit buffer
and re-normalized to the highest greyscale region to give a final 16 bit image. The original
integrated greyscale can be recovered by dividing the 16 bit image by a ‘pixel scale factor’
stored within the image database which corrects for the effects of frame averaging and
normalization.
The images were exported from clinical system and analysed off-line using software
written in the IDL (Interactive Data Language, Research Systems Inc.) programming language.
Linac daily monitoring using an EPID 1723

Each image was read into the software, corrected for frame averaging and normalization as
described above and analysed in the following ways.
A daily output measurement was obtained from the image by measuring the mean
greyscale within a 50 × 50 pixel (12.5 × 12.5 mm) region of interest (ROI) centred upon the
image centre. Daily output was compared to absolute dose measured using an ion chamber
at a depth of 5 cm in a hardboard block for a 10 × 10 cm field and 100 cm focus to surface
distance. Wedge factors were calculated by dividing the mean greyscale from this ROI in
an open image by the mean greyscale from the same ROI from the wedged image acquired
the same day at the same energy. It should be noted that the ‘wedge factors’ measured in
this manner are not equivalent to true wedge factors as measured in water. However, the
constancy of the EPID-measured wedge factors should reflect the constancy of the true wedge
factors. From the centre of each open field image, 11 adjacent profiles (representing a width of
2.75 mm) were extracted and averaged to create a single profile. This was done in both
the in-plane (GT) and cross-plane (AB) directions. The profile was then smoothed using
a boxcar (moving window) average over 15 pixels (3.75 mm) to remove the effects of any
bad pixels. The effect of this smoothing was checked by calculating beam symmetry and
flatness (as defined below) both with and without smoothing. No significant difference was
seen except in the case where a bad pixel affected the result from an unsmoothed profile.
Beam symmetry and flatness were calculated from the profiles using the IEC definitions
(International Electrotechnical Commission 1989):


Dleft
symmetry = × 100
Dright max


Dmax
flatness = × 100
Dmin
where (Dleft/Dright)max is the maximum ratio of the dose from two symmetric points either side
of the beam centre within the flattened area of the beam and Dmax and Dmin are the maximum
and minimum doses in the profile within the flattened area of the beam. The flattened area
along the GT and AB axes is defined as the area corresponding to the 50% dose points minus
0.1 × field size as defined by the 50% level for field sizes up to 30 cm. The profiles were
averaged over 2 mm to obtain symmetry and flatness values; this is therefore well within the
maximum area of 1 cm2 specified by the IEC definitions. As with the wedge factors, the
‘symmetry’ and ‘flatness’ values measured from the EPID will not be equivalent to true IEC
definitions of symmetry and flatness, as demonstrated in figure 4. But they do enable the
constancy of the true symmetry and flatness to be monitored. Furthermore, although the IEC
definition covers the whole 2D flattened area, this analysis has been confined to the GT and
AB axes for consistency with normal QC procedures and because beam steering control is
only available in these two axes. The profiles were further analysed by calculating the hump
value for each profile, defined as:
(DL + DR )
hump = × 100
2DCentre
where DL and DR are the doses at the left and right extremes of the flattened area and DCentre
is the dose at the beam centre. This parameter, whilst closely related to the beam flatness,
corresponds to the hump error measured by the accelerator using the segments of the ion
chamber and used to monitor beam energy and servo the linac gun current.
Field sizes were measured from the extracted profiles by searching outwards along the
extracted profiles for the first pixel with a greyscale of less than 50% of the central axis
greyscale. The field size was then defined as the difference between the two field edge
1724 G J Budgell et al

106

8 MV
104

Normalized EPID greyscale

6MV
Linear (8 MV)
102 Linear (6MV)

100

98

96

94
94 96 98 100 102 104 106
Ion chamber dose (cGy)

Figure 1. EPID greyscale, normalized to give a value of 100 for an ion chamber dose of 100 cGy,
plotted against dose measured with an ion chamber.

pixels in a profile multiplied by the pixel size. Whilst this is not as accurate as the sub-pixel
interpolation method used by Baker et al (2005), this did not matter since no attempt was
made to measure absolute field size, simply to check for any unexpected changes from day to
day.
In addition to the analysis of the daily images described above, an experiment was carried
out to test the ability of the EPID to detect changes in beam symmetry and flatness. A series
of images were acquired with the beam steering currents deliberately varied to create un-flat
and asymmetric beams. Flatness and symmetry were calculated for each image as described
above. The same experiment was repeated at both linac energies using an ion chamber in
plotting tank at a depth Dmax (15 mm for 6 MV, 20 mm for 8 MV) for comparison.

Results

Daily output measurements

Figure 1 shows the dose linearity of the EPID greyscales plotted against exposures of 95–
105 MU, the dose range within which daily output will vary. The EPID greyscale has been
normalized to give a value of 100.0 for a 100 MU exposure. The graph shows that greyscale
is extremely linear with dose over this dose range and that the gradient is sufficient to resolve
small differences in dose. Figure 2 shows the 6 MV daily output measured using the EPID
plotted with the absolute dose routinely measured using an ion chamber. The EPID greyscales
have been normalized to give 100.0 on a day when the ion chamber read exactly 100 cGy/
100 MU at depth of dose maximum. The panel was recalibrated on days 120 and 138 and the
panel was replaced and calibrated on day 152. On each occasion a step change in the greyscale
was seen and a new normalization was required from that date onwards. It is clear from the
graphs that the EPID measurements follow the same patterns as the ion chamber results.
Figure 3 shows the differences between the ion chamber and EPID results—corresponding to
a normal distribution with a standard deviation of 0.64% at 6 MV and 0.65% at 8 MV. Hence
almost 95% of results lie within 1.3% of the ion chamber results. This level of accuracy
is certainly sufficient for a daily output check, for which an accuracy of 5% is specified in
national recommendations (IPEM 1999) and shows that the EPID could be used in place of
some of the commonly used output check devices.
Linac daily monitoring using an EPID 1725

105
104
Linac output (cGy/100 MU)

103
102
101
100
99
98
EPID
97
Ion chamber
96
95
0 50 100 150 200 250 300 350 400
Day

Figure 2. 6 MV daily output of the linac over a period of just over a year measured both with
EPID and ion chamber.

0.25

0.2
6 MV
8 MV
Frequency

0.15

0.1

0.05

0
-3 -2 -1 0 1 2 3
% difference

Figure 3. Histogram of the dose differences between EPID and ion chamber for the data plotted
in figure 2.

Beam profile parameters

Figure 4 shows the 8 MV results of the experiment in which symmetry and flatness were
deliberately varied and measured with both EPID and ion chamber in water. The symmetry
and flatness values plotted in figure 4 have both been normalized to give 100% for a flat and
symmetric beam since there is a systematic difference between the absolute values generated by
the EPID and ion chamber due to the beam-flattening effect of the EPID flood-field calibration
(discussed further below). Figure 4 shows that the EPID gives a similar near-linear response
to changes in beam steering current as the ion chamber and therefore demonstrates that it is
well able to discriminate even small changes in beam profiles.
Figure 5 shows the beam symmetry, flatness and hump values for the 8 MV beams over
the whole year. Similar results were obtained for 6 MV but are not shown. It is important to
understand that these numerical values are not the same values that would be measured for the
same beam if measured in water, even at the same water-equivalent depth. This is primarily due
1726 G J Budgell et al

106%

105% EPID flatness

EPID symmetry

Normalized flatness/symmetry
104% Water flatness
Water symmetry

103%

102%

101%

100%

99%
-60 -40 -20 0 20 40 60
Change in steering current
(a)
106%

EPID flatness
105% EPID symmetry
Water flatness
Normalized flatness/symmetry

Water symmetry
104%

103%

102%

101%

100%

99%
-60 -40 -20 0 20 40 60
Change in steering current
(b)

Figure 4. 8 MV beam symmetry and flatness, normalized by dividing by the values obtained for a
flat and symmetric beam, as measured both with the EPID and ion chamber at depth Dmax in water
in a plotting tank. The steering currents in the AB and GT plans have been deliberately varied to
induce asymmetries into the beams.

to the beam-flattening effect of the flood-field calibration mentioned in the methods section.
Hence the absolute values of flatness and hump are smaller than those that would be measured
at the same equivalent depth. Any artefacts which develop within the images will also affect
the measured values of symmetry and flatness. However, if these are reproducible from day
to day, then they will have the same effect each day. Use of an EPID is therefore suitable as a
consistency check rather than an absolute check of beam profiles.
Clearly in figure 5 there are features visible in the data. Firstly, the GT flatness and
symmetry data sets are considerably noisier than the corresponding AB data sets. This is
expected since GT profiles correspond to the plane of beam bending and are affected by
electron energy variations due to, for example, gun current variation. Hence GT profiles are
Linac daily monitoring using an EPID 1727

(a)

(b)

(c)

Figure 5. Daily values of (a) symmetry, (b) flatness and (c) hump at 8 MV as measured with an
EPID over a period of just over a year. Adjustments to the beams are also marked on the graphs.

inherently less stable than AB profiles. The 6 MV GT profiles displayed less variation than
those shown in figure 5, the difference being due to an underlying instability in the 8 MV beam
steering in the GT plane solved by an ion chamber replacement on day 298. The EPID panel
1728 G J Budgell et al

Gun adjusted New ion chamber

3% 8MV gun adjusted New EPID panel
Gun adjusted Steering adjusted
New ion chamber 2 Service day
2% Gun adjusted

1%

0%

-1%

-2%

-3%
0 50 100 150 200 250 300 350 400
Day

Figure 6. Changes in 8 MV hump from one day to the next plotted against time. Changes >1%
are marked with solid diamonds.

was replaced on day 152 leading to a discontinuity in the graphs for 6 MV AB symmetry and
8 MV flatness. This is not a linac change but a result of change of artefacts/flood-field
calibration with a new panel hence the jump in ‘baseline’ expected values. Another feature
of the data is the clear step changes in the data caused by deliberate adjustments to the gun
and steering currents made during service days, QC sessions or ion chamber replacements.
Changes of half a per cent are readily detectable in the data. This demonstrates the potential
usefulness of the EPID—the technique is sensitive to even quite small changes in beam
parameters and provides an early warning of any sudden change in the beam. It also provides
a quick check after a service day or breakdown to demonstrate that the beam has not been
affected.
Figure 5(c) shows the usefulness of the hump error measurement which provides smoother
data than the flatness/symmetry measurements and appears very sensitive in detecting beam
changes. Possibly because by always using the same points there is not the same sensitivity
to random noise. The GT and AB humps match each other closely as expected since changes
in beam energy will affect both profiles equally. Figure 6 shows one possible method to
detect sudden beam profile changes—by subtracting each day’s reading from the previous
day, sudden jumps in the hump clearly show up as outliers on the graph. By applying a simple
threshold sudden changes can be readily detected. A 1% threshold has been applied and can
be seen to pick out sudden changes on 5 days, all corresponding to known events marked as
large grey-filled diamonds. The smoother 6 MV data required a 0.5% threshold.

Wedge checks
The 6 MV wedge check results for the year are shown in figure 7. The obvious feature in this
data is the 2% step change that takes place on day 175. A similar change was seen at 8 MV.
This could be due to a physical change in wedge position but this is ruled out by the results
of the ion chamber wedge checks averaged over all collimator (head) angles also shown in
figure 7. It could also be due to the EPID not returning to the same standard position each day—
this would result in the analysed ROI lying within a different part of the gradient generated by
the wedge. However, the ion chamber results from the standard QC wedge checks measured
at collimator angle 0◦ only (the same collimator angle as used for the EPID checks) show the
same step change, suggesting a step change in the beam steering affecting both 6 MV and
Linac daily monitoring using an EPID 1729

103%

102%

101%
Wedge factor

100%

99%

EPID
98% Ion chamber - all head angles
Ion chamber - head angle 0 degrees

97%
0 50 100 150 200 250 300 350 400
Day

Figure 7. Daily values of 6 MV wedge factor measured at one collimator (head) angle as measured
with an EPID over a period of just over a year. Wedge factors measured during this period with
an ion chamber both at a single collimator (head) angle and at all collimator (head) angles are also
marked on the graphs.

8 MV. An investigation showed this to be due to the swapping of two boards relating to the
beam steering system, one of which was faulty, on a service day. Corresponding small changes
in 6 MV beam parameters are detectable in the data. The 8 MV flatness also appears to drop at
this point, although the noisiness of the data makes it hard to detect. It can be concluded that
the EPID has detected an unexpected small (and clinically insignificant) change in the beam
steering.

Field size checks

The results of the field size measurements are displayed in figure 8. Only results from the
6 MV fields are shown since any change in collimator position will affect both energies equally.
The results for the conventional jaws show that the field size remained consistent within 1
pixel (0.25 mm) over the full year except for a suggestion of a 1 pixel step change at day 146.
This is the day the EPID panel was changed. Whilst not a clinically significant difference
it demonstrates the daily reproducibility of the readings is such that a small step change in
results is detectable. The leaf results are slightly noisier though still normally within 1 pixel
of the average, with occasional 2 pixel excursions observed. There are more step changes
within this data set which can be explained by the following events. On day 18, the accelerator
software was upgraded—the upgrade caused the leaf calibration to be lost hence the leaves
were recalibrated. On day 154 the MLC reference reflectors were re-centred by an engineer.
Since the MLC position is optically controlled this led to a step change in the MLC field width.
On day 176 the MLC was recalibrated to compensate for this field size change. These results
show that the EPID was able to show that the field sizes remained stable and that it is easily
sensitive enough to detect small changes.

Discussion

The study presented here demonstrates that the amSi EPID studied is capable of the daily
monitoring of a number of different linac beam parameters. Firstly, these results demonstrate
1730 G J Budgell et al

241
Software upgrade - New panel Reference reflectors recalibrated
Leaves recalibrated
Leaves recalibrated
240.5

240
Field width (mm)

239.5

239

238.5 Conventional jaws

Leaves

238
0 50 100 150 200 250 300 350 400
Day

Figure 8. Daily values of field size as measured with an EPID over a period of just over a year.
The conventional jaws defined the field size in the GT axis and the MLC leaves in the AB axis.
Changes affecting the MLC calibration are also marked in the figure.

the EPID is clearly suitable as a daily output check device. It is a sensitive indicator of changes
in dose and displays suitable long term stability in greyscale to dose response. The only factor
which affected this response was recalibration of the EPID. In practice this is not a limiting
factor since the renormalization procedure only requires a single day’s ion chamber result.
Since, when using a daily check device, regular ion chamber measurements conforming to a
dosimetric protocol are still required this is a trivial requirement. Manufacturers recommend
recalibration of amSi EPIDs at regular intervals, sometimes even monthly. Our experience
with a large number of these panels over a long period of time suggests that this is unnecessary;
we only recalibrate when artefacts become visible in images. Because of this, our results show
the dose-greyscale response remains stable over a period of 6 months or more. This is in
agreement with the results of Louwe et al (2004) and Winkler and Georg (2006) but the
sudden change in EPID response observed 2 months after installation by Winkler and Georg
(2006) was not observed. Any such sudden change in EPID response would be detected as
a daily output change. Since this would then be checked against an ion chamber, the change
would be found the same day.
An advantage of choosing to use an EPID as the daily output check device is that, as
long as a sufficiently large field is used, daily values of beam symmetry and flatness can be
derived from the same images, as it were ‘free of charge’. The EPID has been shown to be
a sensitive indicator of changes in beam profiles with detected step changes corresponding
to known events that affect the beam profile. The number of false positives is very low
although changes to the EPID such as panel changes, artefacts and recalibrations will affect the
numerical parameters generated. It could be argued that daily monitoring of beam symmetry
and flatness is unnecessary since modern linacs exhibit excellent beam stability and have
accurate error detection systems which reliably show up the presence of changes in beam
profile. We would concur with this statement and would not commit additional resources to
such a program. But since these data are freely available from the necessary output checks it is
surely worthwhile analysing it. Moreover, the results presented yield significant information
regarding the performance of this particular linac over the time period in question. The
extent and frequency of the 8 MV beam steering instability is clear. The sensitivity of beam
Linac daily monitoring using an EPID 1731

profiles to gun adjustments is evident. The consequences of various adjustments made on

service days are readily detectable. All these changes are relatively minor—all the results
are well within tolerances and will have no clinical impact on patient treatments. Hence they
will not be highlighted by routine QC checks. However, the availability of such data does
help the physicists/engineers responsible for the machine to understand the performance and
characteristics of the particular linac. The step-changes caused by regular EPID recalibration
would confuse the underlying baseline data but, as stated above, such recalibrations are
unnecessary. Plainly, if the EPID is capable of detecting such small changes in beam profile
then large, clinically significant changes will also be detected and flagged by the method. A
further advantage is that if such daily monitoring is being performed and cross-compared with
conventional QC checks then the EPID can be used as a quick QC check device following
services and repairs to the linac, enabling the minimum of treatment time to be lost.
Long term time trends were observed in some of the data. It is difficult to determine
whether these represented real features of the linac behaviour over this time period. It may
simply be that successive downward step changes combine to give the appearance of long
term trends. Equally, they may be due to a drift in the properties of the EPID itself. A
comparison with QC measurements of flatness and symmetry to determine whether these
were real trends proved inconclusive, with the suggestion of a possible downward trend in
the symmetry measurements but no convincing correlation. Whether real or not, the trends
are slow enough such that sudden changes in beam properties are not disguised by the long
term trends hence this does not invalidate the usefulness of the method. However, without
further evidence, it does mean that the method is not suitable for the detection of long term
time trends.
As with flatness and symmetry, field size measurements are not strictly necessary on a
daily basis, the fact that they can be calculated at no extra cost makes them worth doing and
demonstrates the stability of the leaf and jaw calibration. One drawback is that the 24 × 24 cm
field used is defined by a combination of MLC and back-up jaw. Hence if they are not exactly
aligned the back-up jaw could hide changes in MLC calibration and vice versa. However, this
does not prevent the data being a useful confidence check.
The value of the wedge factor results is the most debateable part of the data presented
because an additional wedged field is required at each energy as well as the open field. In
practice, a wedged field is defined as an open segment and a wedged segment hence the only
real acquisition effect is a negligible few extra seconds per energy—only one beam needs
to be selected per energy. Daily checks of wedge position are not particularly useful in
themselves—we have never detected a measurable change of wedge position. Two interesting
results did arise from the data however. One was that the reproducibility of the wedge checks
demonstrates the high reproducibility in EPID positioning from day to day. Secondly, it
demonstrates how sensitive a wedge check performed at a single collimator angle is to beam
steering variation. Routine ion chamber wedge checks are made combining results from all
four collimator angles to prevent variation caused by incorrect chamber positioning and beam
steering effects. The sudden jump in EPID wedge factor allowed us to pinpoint the date on
which a beam steering change occurred. However, on balance, to make wedge factors part
of a daily check routine would probably cause more false alarms than the advantage it would
give. There may however be a place for EPID wedge factor checks after service days/repairs
and for checking wedge factors at varying gantry angle.
In practice this study was an off-line study. Images were acquired, exported and analysed
en masse. To make the use of an EPID practical as a daily check device the method requires
automation. We are therefore currently working on a method to automatically detect when
specified QC images are acquired, run the analysis software and give an immediate display of
1732 G J Budgell et al

the day’s result. This will allow the method to rapidly and readily be used by radiographers.
The results will also be saved to allow the kinds of trend analysis demonstrated in this paper.
A study is also underway to determine whether it is possible to use the EPID for electron QC
in a similar manner to the photon QC described in this paper.
This paper has explored the suitability of an EPID as a daily monitoring device. A
wider question is to what extent the EPID could be used to replace other, less frequent, beam
profile checks? The data presented here allow some conclusions to be drawn. Firstly, the
EPID has been shown to be an accurate relative check device of beam profiles. Hence it
is possible to use it for relative measurements such as variation of beam profile and wedge
factor with gantry angle and low MU IMRT QC (Budgell et al 2005). Additionally, as
suggested above, if being used on a regular basis, then it can readily be used after service
days and breakdowns. It could not be used, however, for routine flatness/symmetry checks
unless regular QC is carried out on the EPID as a dosimetry device. This is because, as is
seen in the results, the parameters calculated from EPID images are not identical to those
measured by more conventional methods. The EPID parameters can also be affected by the
beam not being perfectly symmetric during gain calibrations, thus introducing systematic tilts
into the measured profiles. Furthermore, the EPID may be prone to long term drifts and the
development of artefacts which can affect the beam parameters measured. In practice, the QC
checks required would probably be too time-consuming to yield an efficiency advantage for
the EPID unless they were also required for accurate IMRT verification, as has been suggested
by Winkler et al (2005). In theory, the wedge factor and flatness parameters measured with
then EPID could also be used as a surrogate for beam energy checks since these parameters
change with energy. In practice, any unexplained change in flatness or wedge factor seen
using the EPID would be rapidly checked by an independent energy check method.

Conclusion

An amSi EPID has been shown to be a useful daily check device for linac output which can
simultaneously be used for daily monitoring of beam profiles and field sizes.

Acknowledgments

Thanks go to Julie Davies and the radiographers on R1 for acquiring the daily images used in
this paper for a whole year, and to James Beck for helping with some of the measurements.

References

Baker S J, Budgell G J and MacKay R I 2005 Use of an amorphous silicon electronic portal imaging device for
multileaf collimator quality control and calibration Phys. Med. Biol. 50 1377–92
Budgell G J, Zhang Q, Trouncer R J and Mackay R I 2005 Improving IMRT quality control efficiency using an
amorphous silicon electronic portal imager Med. Phys. 32 3267–78
Chang J, Obcemea C H, Sillanpaa J, Mechalakos J and Burman C 2004 Use of EPID for leaf position accuracy QA
of dynamic multi-leaf collimator (DMLC) treatment Med. Phys. 31 2091–6
Curtin-Savard A and Podgorsak E B 1997 An electronic portal imaging device as a physics tool Med. Dosim. 22 101–5
Dirkx M L, Kroonwijk M, de Boer J C and Heijmen B J 1995 Daily dosimetric quality control of the MM50 Racetrack
Microtron using an electronic portal imaging device Radiother. Oncol. 37 55–60
Dunscombe P, Humphreys S and Leszczynski K 1999 A test tool for the visual verification of light and radiation
fields using film or an electronic portal imaging device Med. Phys. 26 239–43
Greer P B and Popescu C C 2003 Dosimetric properties of an amorphous silicon electronic portal imaging device for
verification of dynamic intensity modulated radiation therapy Med. Phys. 30 1618–27
Linac daily monitoring using an EPID 1733

International Electrochemical Commission 1989 Medical electrical equipment—medical electron accelerators in the
range 1 MeV to 50 MeV—guideline for functional performance characteristics (IEC 977) (Geneva: IEC)
IPEM 1999 Physics aspects of quality control in radiotherapy IPEM Report 81 (York: Institute of Physics and
Engineering in Medicine)
Kirby M C and Williams P C 1993 Measurement possibilities using an electronic portal imaging device Radiother.
Oncol. 29 237–43
Kirby M C and Williams P C 1995 The use of an electronic portal imaging device for exit dosimetry and quality
control measurements Int. J. Radiat. Oncol. Biol. Phys. 31 593–603
Liu G, van Doorn T and Bezak E 2000 Evaluation of the mechanical alignment of a linear accelerator with an
electronic portal imaging device (EPID) Australas. Phys. Eng. Sci. Med. 23 74–80
Liu G, van Doorn T and Bezak E 2002 Assessment of flatness and symmetry of megavoltage x-ray beam with an
electronic portal imaging device (EPID) Australas. Phys. Eng. Sci. Med. 25 58–66
Liu G, van Doorn T and Bezak E 2004 The linear accelerator mechanical and radiation isocentre assessment with an
electronic portal imaging device (EPID) Australas. Phys. Eng. Sci. Med. 27 111–7
Louwe R J, McDermott L N, Sonke J J, Tielenburg R, Wendling M, van Herk M B and Mijnheer B J 2004 The long-term
stability of amorphous silicon flat panel imaging devices for dosimetry purposes Med. Phys. 31 2989–95
Luchka K, Chen D, Shalev S, Gluhchev G and Rajapakshe R 1996 Assessing radiation and light field congruence
with a video based electronic portal imaging device Med. Phys. 23 1245–52
McCurdy B M, Luchka K and Pistorius S 2001 Dosimetric investigation and portal dose image prediction using an
amorphous silicon electronic portal imaging device Med. Phys. 28 911–24
McDermott L N, Louwe R J W, Sonke J-J, van Herk M B and Mijnheer B J 2004 Dose–response and ghosting effects
of an amorphous silicon electronic portal imaging device Med. Phys. 31 285–295
Prisciandaro J I, Herman M G and Kruse J J 2003 Utilizing an electronic portal imaging device to monitor light and
radiation field congruence Appl. Clin. Med. Phys. 4 315–20
Samant S S, Zheng W, Parra N A, Chandler J, Gopal A, Wu J, Jain J, Zhu Y and Sontag M 2002 Verification of
multileaf collimator leaf positions using an electronic portal imaging device Med. Phys. 29 2900–12
Van Esch A, Depuydt T and Huyskens D P 2004 The use of an aSi-based E P ID for routine absolute dosimetric
pre-treatment verification of dynamic IMRT fields Radiother Oncol. 71 223–34
Vieira S C, Dirkx M L, Pasma K L and Heijmen B J 2002 Fast and accurate leaf verification for dynamic multileaf
collimation using an electronic portal imaging device Med. Phys. 29 2034–40
Winkler P and Georg D 2006 An intercomparison of 11 amorphous silicon EPIDs of the same type: implications for
portal dosimetry Phys. Med. Biol. 51 4189–200
Winkler P, Hefner A and Georg D 2005 Dose-response characteristics of an amorphous silicon EPID Med.
Phys. 32 3095–105
Yang Y and Xing L 2004 Quantitative measurement of MLC leaf displacements using an electronic portal image
device Phys. Med. Biol. 49 1521–33