- Email: [email protected]
Dosimetric characteristics of a liquid-filled electronic portal imaging device
Int. J. Radiation
Oncology
Biol.
Pergamon
Phys., Vol. 33. No. 5, pp. 1265- 1272. 1995 Copyright 0 1995 Elsevier Science Inc. Printed in the USA. All rights reserved 0360.3016/95 $9.50 + .OO
0360-3016(95)00108-5
l
Treatment
Verification
DOSIMETRIC
MARION
CHARACTERISTICS PORTAL
OF A LIQUID-FILLED IMAGING DEVICE
ESSERS, MSc., BART R. HOOGERVORST, M.Sc., MARCEL HUGO LANSON AND BEN J. MIJNHEER, PH.D.
ELECTRONIC
VAN HERK, PH.D.,
Departmentof Radiotherapy,The NetherlandsCancerInstitute, Antoni van LeeuwenhoekHuis, Plesmanlaan121, 1066CX Amsterdam,The Netherlands Purpose: To determine the characteristics of a commercial electronic portal imaging device (EPID), based on a two-dimensional matrix of liquid-filled ionization chambers, for transmission dose measurements during patient treatment. Methods and Materials: Electronic portal imaging device measurements were performed in a cobalt-60 beam and two accelerator x-ray beams, and compared with measurements performed with a Farmer-type ionization chamber in air in a miniphantom and in an extended water phantom. Results: The warming up time of the EPID is about 1 h. The long-term stability of the detector is better than 1% under reference conditions for a period of about 3 months. The signal of the ionization chambers follows approximately the square root of the dose rate, although the relation becomes more linear for larger (> 1 Gy/min) dose rates. The signal can be transformed to dose rate with an accuracy of 0.6% (1 SD). The short-term influence of integrated dose on the sensitivity of the ionization chambers is small. The sensitivity increases about 0.5% for all ionization chambers after an absorbed dose of 8 Gy and returns to its original value in less than 5 min after stopping the irradiation. This small increase in sensitivity can be ascribed to the electrode distance of the ionization chambers in commercial EPIDs, which is 0.8 + 0.1 mm. The sensitivity increase depends on the electrode distance and is 4% for a 1.4 mm electrode distance. The scattering properties of the EPID ionization chambers were between those of an ionization chamber in a miniphantom and in a water phantom. Conclusion: The matrix ionization chamber EPID has characteristics that make it very suitable for dose rate measurements. It is therefore a very promising device for in vivo dosimetry purposes, Radiotherapy, Portal imaging, mission dosimetry.
Dosimetry,
Ionization
chamber,
Quality
assurance, In vivo dosimetry,
Trans-
INTRODUCTION Various electronic portal imaging devices (EPIDs) are currently available for clinical applications. Some of these systems are based on a scintillating screen/video camera combination, while others are based on scanning techniques that either use a moving array of photodiodes or a matrix of liquid-filled ionization chambers (3). In our institution we have developed an EPID that consists of a matrix of 256 X 256 liquid-filled ionization chambers (12, 13). Because the liquid density is high compared with gassesand signal integration occurs in the liquid (12), a good signal-to-noise ratio can be obtained with small ionization chambers. Electronic portal imaging devices were originally
introduced in the clinic to overcome some drawbacks of conventional portal films, such as the short exposure time, which necessitates an interruption in the patient irradiation, and the amount of time needed to develop and read a film (3). Electronic portal imaging devices are extremely useful for on-line geometrical patient setup measurements (2, 8, 15). They can, in addition, be used for other purposes such as the determination of the diode position during in viva dosimetry (5). An EPID might, however, also be suitable for simultaneous patient set-up measurement and transmission dosimetry in a large number of points; for example, during high dose-high precision treatments. The possibilities of using conventional portal films and
Reprint requeststo: Marion Essers,M.Sc. E-mail address: [email protected] Acknowledgements-This work was financially supportedby the NetherlandsCancer Foundation(NKB Grant NKI 92-40). We would like to thank Nice van Bree (M.Sc.) for helping us
with the measurements performedwith the secondcommercial EPID at the “Zeeuws RadiotherapieInstituut” in Vlissingen, and Ronald Boellaard (M.Sc.) and JoosLebesque(Ph.D.) for many useful discussions. Accepted for publication 24 February 1995.
1265
I. J. Radiation Oncology
1266 commercial
l
Biology
Physics
EPID
CU
T
PCB
2
front liquid
barium E” ij
l
CU -back
ferrite PCB film PCB
I foam
CU
1r~ilb.~..,
;,
,,
_
;; *
__
;
PCB
Fig. 1. Cross-section through the EPID. PCB = printed circuit board. Indicated is a 18 pm thick layer of Cu present at the
sidesof all PCBs. fluoroscopic EPIDs for in viva dose measurementshave already been reported (6, 7, 10). An advantage of EPIDs compared with films is that the results can be obtained on-line. A possible advantage of the matrix ionization chamber device over the other systems might be its high spatial resolution (no light scatter) and high dynamic range. The aim of this study was, therefore, to investigate whether commercial matrix ionization chamber EPIDs are suitable for dosimetry purposes. For that purpose we studied various characteristics of the EPID. First, we wanted to examine the warming-up effect. Second, the relation between signal and dose rate had to be determined. Third, the long-term stability of the detector had to be investigated. Fourth, the influence of radiation history on the signal was examined, Finally, the scattering properties of the EPID were quantified.
METHODS
AND MATERIALS
The commercial ionization chamber EPID’ is based on a prototype developed in our institute (8, 12, 13). A crosssection of the EPID is shown in Fig. 1. The matrix of ionization chambers, consisting of two electrode plates and the liquid, is packed between two foam layers. A lmm thick plate in front of the ionization chambers, made of Plastoferrite,’ is used as a converter material for the production of electrons. The matrix of 65,536 (256 X 256) ionization chambers, in a sensitive area of 32 cm X 32 cm, is formed by cross-points of 256 electrode strips on a front-printed circuit board (PCB) and 256 strips perpendicular to these on the back PCB. At the sides of the EPID, where the plates are glued together, the size of the ionization chambers is 1.27 mm X 1.27 mm X 0.8 mm. For individual ionization chambers, the distance between the electrode plates may slightly differ. The total amount
’ Varian PortalVision, Varian Inc., Palo Alto, CA. * Plastoferrite is a composite material of plastic and barium ferrite with a density of 4.75 g/cm’. ’ Iso-octane, spectroscopical pure, Merck, Darmstadt, Germany.
Volume 33, Number 5, 1995
of material is equivalent to a build-up layer of about 12 mm water. The backscatter material is equivalent to about 5 mm water. As ionizing medium, a liquid dielectric3 is used. The liquid probably contains impurities due to interaction of the liquid with the chamber construction material (1, 17) and diffusion of water through the printing board material. The chambers are scanned row by row by switching the polarizing voltage on the front plate electrodes j. The polarizing voltage, which is controlled by software, is usually set to 250 V and is well below saturation. Each electrode strip on the back plate is connected to an electrometer i. The high dose rate of a radiation beam makes it necessary to measure the ionizing current during the short period when the high voltage is switched on (e.g., 20 ms per row) instead of performing event counting. The scan time of the EPID varies between 1.5 s and 6 s, which enables a quick evaluation of the images, even during one treatment session. This short scan time implies that the dose rate is determined from the ionization chamber signals instead of the dose. The acquisition computer is located outside the treatment room. The acquisition mode used for our measurements was the normal, nonsmoothed sampling mode (standard mode), with a sampling time of 20 ms and an image scan time of 5.6 s (14). Measurements were performed using the 6 and 8 MV x-ray beams of two linear accelerators4 and a cobalt unit. Measurements with a second commercial detector were performed using the 8 MV x-ray beam of another accelerator.5 Reference dose rate measurements were performed with a calibrated small Farmer-type ionization chamber’ in combination with an electrometer.’ At the cobalt unit, the dose rate in air was determined with an ionization chamber with build-up cap. At the linear accelerators, the dose rate was determined at the depth of dose maximum in a small perspex rod miniphantom (11). The dose rate measured by the ionization chamber was calculated following the procedure given in the Dutch dosimetry protocol (9). The incorrect use of this protocol in the case of the small cylindrical phantom leads to a small error (< 0.5%) in the dose rate determination. Detector warming-up time After turning on the EPID, a certain time period is required to obtain a constant reading. This warming-up time has been measuredby making an EPID image every minute for a half hour and every 3 min for the next half hour. To limit radiation damage to the liquid, the radiation
’ Philips SL-15, Crawley, U.K. ’ Varian Clinac 21OOc, Palo Alto, CA. b NE 245,0.2 cm’, NE Technology Limited, ’ Keithley-616, Cleveland, OH.
Berkshire,
U.K.
Dosimetry
with an EPID
was switched on only briefly when each image was made. The measurements have been performed on the cobalt unit at a source-detector distance (SDD) of 80 cm with a field size of 32 x 32 cm*. At this distance, the dose rate was about 0.5 Gy/min. Due to limitations of the software, the high voltage source is only switched on when taking the first image, which means that the warming-up time only begins at that time. Calibration: from detector signal to dose rate There are several imperfections in the detector system, such as gain differences and offset signals. For the purpose of calibration, we will assume that each pixel value is related to the local dose rate, d,j, at the location of pixel (i,j) according to the following equation: P,] = S, * G(d,,) + O,j + E,,
0%. 1)
where P, is the pixel value, S,, describes the relative sensitivity of each individual ionization chamber-electrometer combination, the function G describes the absolute response of the detector (which depends in a nonlinear fashion on the dose rate), and 0, and Ej are offset values attributed to individual pixels and electrometers, respectively. In the commercially available software,’ corrections are applied automatically for E,, which is measured quickly without polarizing voltage prior to acquisition of each image, and O,, which is measured by an image acquisition without radiation. 0, are pixel values of a dark-field image, Idark (13). The offset corrected pixel values, I,* are described by Zij = S, * G(d,) .
(Eq. 2)
If it would be possible to deliver the same dose rate d,, to all pixels, the function G and the relative values of S, could be determined directly. However, in practice, it is difficult to deliver a completely flat field to the detector. For that reason, the calibration procedure consists of two steps. Even for an inhomogeneous field it is possible to define a small detector region such as 11 X 11 pixels near the center of the detector, for which the beam can be assumed to be flat. In this study, the average value of S,, over these 11 X 11 pixels was chosen to be unity. The first step of the calibration procedure is then the determination of the function G by measuring the average offset corrected pixel value I over the area of 11 X 11 pixels for various (homogeneous) dose rates d using Z = G(b)
(W. 3)
* Nucletron, WMS system, Veenendaal, The Netherlands.
1267
0 M. ESSERS et al.
Details about the function G will be described in the results section. For the second step of the calibration, a single image of a more or less “flat” beam is used. In the commercially available software,’ this image would have been used directly to normalize the detector sensitivities, by performing a so-called flat- or flood-field correction (13). For correct dosimetric calibration, however, the dose rate distribution of the “flat” beam must be measured independently. The following relation describes the flat-field image, Z,la,: (Eq. 4) where Liij,na, is the known dose rate distribution of the “flat” beam. Dividing Eq. 2 by Eq. 4, one can easily derive that 0%
5)
where f is the inverse of G. Equation 5 shows that it is possible to derive absolute dose rate values from pixel values and a single calibration image with known dose rate distribution, if the absolute response as a function of dose rate is known for a small part of the detector (function G). It is assumed that the same function G can be applied to describe the dose rate dependence of all pixels (Eq. 1). It should be noted that the flat-field correction algorithm applied by the commercially available software’ computes image pixel values by multiplying ZjjZ,,,Rarwith the average value of Z,jrnat.The resulting image pixel values have to be divided by this average of Zi,,nafto use Eq. 5 to calculate the dose rate from image pixel values. If, in Eq. 5, the average value of Z,,,,la,is used instead of G (d,,,,,a,), errors of several percentages may occur in the calculated dose rate, because inhomogeneities in the dose rate over the “flat” beam are then ignored. For our study, Zna, was acquired at the 8 MV x-ray beam of an accelerator4 at 150 cm SDD in a field of 24 x 24 cm2 at the isocenter. The corresponding dose rate distribution &,, in air was determined under the same conditions using an ionization chamber at the depth of dose maximum in a cylindrical miniphantom, which was scanned using the transport mechanism and software of our water phantom system.x The measurements, performed on a 0.5 X 0.5 cm2 grid, were transformed to the resolution of the EPID by means of bilinear interpolation. The dose rate d used for determination of the function G ranged from 0.1 to 8 GyZmin by changing the SDD from 78 to 180 cm in an 8 MV x-ray beam, using pulse repetition frequencies (PRFs) between 200 and 400 Hz and using metal absorber blocks. At 100 cm from the
1268
I. J. Radiation
Oncology
0 Biology
l Physics
source, the dose rate is about 4.9 Gylmin, using a PRF of 400 Hz. Due to the unsynchronized operation of our imagers, it was difficult to obtain reliable measurements at PRFs lower than 200 Hz. The field size was 20 X 20 cm’ during these measurements. Besides calibration images, capacity images, Icap, have also been measured (12). The ionization chambers are assumed to behave as plane-parallel capacitors. From the capacity C = ES/d (S is the area of the electrode plates, E is the dielectric constant), the distance d between the plates for the ionization chambers of the detector can be estimated. The capacity of the ionization chambers is measured from the ionization chamber signals when applying a triangular high voltage pulse (from 0 V to 35 V and back to 0 V in 40 ms). The distance between the plates at the sides where the printed circuit boards are glued together is taken as the reference distance. Long-term stability To determine the required frequency of calibration, Zsj,j,dark and Zrl,Ra,were acquired under reference conditions for a period of 3 months. For practical reasons, the flat-field images were now acquired at the cobalt unit in a field of 32 X 32 cm’ at 80 cm SDD. The change in individual offset value had been determined from Idark for 12 individual ionization chambers and the average of 65 x 65 pixels around the center of the EPID. The sensitivity of ionization chambers, determined from Z,,,, might change with time because of changes in water and oxygen contamination of the liquid due to radiochemical reactions in the liquid, or due to photon interaction with the chamber material, which might give rise to a difference in ion mobility and recombination constant. Furthermore, aging of the electronics and stray radiation on the semiconductor devices can result in a change in the gain of the electrometers or a change in the gain of a common amplifier. To quantify these effects, the change in pixel gain has been determined from IRat, which was corrected for the decay of the cobalt source, for the same 12 ionization chambers and the average of 65 x 65 pixels around the center of the EPID. Influence of radiation history on detector signal Radiochemical reactions influence the concentration of impurities in the liquid. Therefore, the sensitivity of the EPID, which depends on the purity of the liquid, the chamber dimensions, and the type of surrounding detector material ( 12, 16, 17), is influenced by the short-term radiation history of the EPID. For reliable dose rate measurements, the modification of the signal due to previous irradiations has to be negligible. Such a variation in signal has been quantified by continuously acquiring images during irradiation up to 8 Gy at the cobalt unit in a 32 X 32 cm* field at a dose rate of 0.8 Gy/min, which corresponds roughly with the clinical situation at an accelerator. For comparison, the measurements were also performed using a second commercial detector in a 32 X 32 cm2 field at
Volume
33, Number
5, 1995
an 8 MV x-ray beam of an accelerator.5 Images were made every 8 s during the first 2 min and every 20 s during the next 8 min. The minimum dose required for one image acquisition was approximately 0.06 Gy in this experimental set-up. The recovery of the detector signal has been measured by making an image every few minutes for a half hour after the irradiation has stopped. The irradiation was switched on only during the measurements. No correction has been applied for the additional dose required to measure this recovery. The average signal of 65 X 65 pixels in the center of the EPID has been chosen for analysis. All values have been normalized to the image acquired after 0.06 Gy. Scattering properties To accurately measure the dose rate at a certain point, a high spatial resolution is necessary (i.e., the scatter from neighboring chambers should be small). The scattering properties were quantified in a 6 MV x-ray beam at an SDD of 100 cm, by comparing the average dose rate determined by 5 X 5 pixels around the center of the detector with ionization chamber measurements in a miniphantom and in a water phantom at different field sizes, varying from 3 X 3 cm’ to 25 X 25 cm2. From the water phantom measurements, performed at a depth of 5 cm, the dose at the depth of dose maximum has been assessed using percentage depth dose tables. The SDD for the ionization chambers was 100 cm. At this distance, the dose rate is about 3.7 Gylmin.
RESULTS Detector warming-up time The increase in ionization chamber signal after switching on the high voltage depends on the position of the pixel on the EPID. A small increase, less than l%, was observed for detectors situated near the electrometers, and an increase of almost 2.5% was found for ionization chambers near the high voltage switches and power regulators. No change in signal was observed for ionization chambers situated further away from the electronics. A stable situation was reached after 1 h. Calibration: From detector signal to dose rate The background signal of the ionization chambers, as measured in the dark-field image, was about 4 mV, which is about 2% of the typical image signals measured at the linear accelerator. The electrometer offset, which is calibrated automatically prior to each image acquisition, was about 10 times larger than the ionization chamber offset. From Zeal,differences in pixel sensitivity up to 20% were found. Significant oscillations were present around the average sensitivity of the ionization chambers. The sensitivity also decreased slightly toward the high voltage switches and power regulators.
Dosimetry
with
an EPID
0 M.
ESSERS
et al.
1269
and x = [Zj,/Zv,na,* G(Lj,,,,,)} can be used to calculate the dose rate d, from Ii, for all ionization chambers ij (Eq. 3. The distance between the electrode plates of each ionization chamber, which could be obtained from Icap, ranged between 0.7 and 0.9 mm. Capacity measurements with a second commercial system showed that the distance between the electrode plates varied between 0.8 and 1.0 mm for this device.
o0 25 -
Long-tern stability Every electrometer offset varied around its own average value of about 50 mV (which could differ between individual ionization chambers by a factor of 5) with a standard deviation of about 1 mV. The dark-field signal was constant in time within 0.4 mV, which was 0.2% of the flat-field image signal, for the 12 examined ionization chambers. No substantial change in Z+Ra,or Zr/,cap, greater than 1% for all pixels, was observed with time, during the period of 3 months.
I5 (Gylmin) (4
“05 5 0
2
4
6
6
IO
b (Gylmin) (b) Fig. 2. (a) Measurement of the average offset corrected pixel value over 11 X 11 pixels, Z, vs. the dose rate d. The solid line represents a(d)“’ with a = 1.507 * 10” (Gy/min)-I’*. (b) Ratio of I and G(D) vs. d with a = 1.351 * 10’ (Gy/min)-“’ and b = 81.36 (Gy/min)-’ (Eq. 6). The measurements have been performed in an 8 MV x-ray beam.
In Fig. 2a, the average value I for 11 X 11 pixels around the center of the EPID in pixel values is plotted as a function of the dose rate 6, which is the same for these pixels. The theoretically expected square root relation (12) is also shown in this figure. The signal could not, however, exactly be described by this relation. A better fit of the data could be obtained with: I = G(d) = a * d”’ + b * d,
0%. 6)
with a = 1.351 * lo” (Gy/min)-I’*, b = 81.36 (Gy/min)-’ for the normal acquisition mode of our EPID and fi in Gy/min. The values of parametersa and b depend slightly on the connection cables and strongly on the acquisition mode in which the images are made. The ratio of I and G(B) is plotted in Fig. 2b. On average, this ratio is 0.999 (SD = 0.006) and the maximum deviation of fitted from actual pixel value is 1.2%. The inverse functionf = G-‘, with
./l-d = -a
+(a* + 4 * b * x)“’ 2*b
* ,
0%. 7)
Influence of radiation history on detector signal For the commercial detector used in our study, the increase in ionization chamber sensitivity as a function of integrated dose received by the EPID was only 0.5%, with saturation after less than 1 Gy. After switching off the radiation, the sensitivity decreased exponentially to its original value with a halflife of about 0.75 min. For a second commercial detector, the change in sensitivity as a function of integrated dose was almost identical, at maximum only 0.7%. The results for both detectors are plotted in Fig. 3. Scattering properties The ratio of the doserate obtained from the EPID signal and the rod phantom data increased, whereas the ratio of 1.020 with
7c
radiation
’
without
radiation
1.015
.-p: v) F .% ii u
1.010
1.005
1 .ooo
-& 6 0
: I, . 5
10
Time
15
.
4
20
25
(min)
Fig. 3. Relative change in detector signal as a function of time for our detector (squares) and the second commercial detector (circles). The detectors are irradiated with a dose rate of 0.8 Gylmin. After 10 min, irradiation has stopped and the recovery of the signal is measured. The signals are averaged over 65 x 65 pixels around the center of the detector and are normalized to the image acquired after 0.06 Gy.
1270
1. J. Radiation
Oncology
l Biology
l Physics
EPID and water phantom measurements decreased with increasing field size (Fig. 4). The deviation of EPID measurements from ionization chamber measurements in a rod or a water phantom was 3% at maximum. DISCUSSION
AND
CONCLUSIONS
Detector warming-up time The warm-up time of the EPID is about 1 h. The increase in signal of each ionization chamber depends on its position with respect to the electronics of the EPID. The differences can be attributed to the amount of heat generated by the electronics in the various positions. Near the high voltage switches and the power regulators, where the increase in ionization chamber signal is 2.5%, a total heat of 13 W is dissipated. Near the electrometers much less power (about 3 W) is dissipated. The temperature increase at these ionization chambers is therefore lower, which results in a smaller signal increase of less than 1%. Ionization chambers that are situated far from the electronics do not warm up and therefore, no change of signal with time is observed after switching on the voltage. Calibration: From detector signal to dose rate Equation 5 shows that the calibration procedure to transform pixel values into dose rate values is relatively easy. First, the large electrometer and small ionization chamber offset values have to be subtracted from the signals (Eq. 1). The large electrometer offset values, which result in a limitation of the dynamic range, are caused by imperfections in the highly sensitive electrometer circuitry. These circuits might be improved in newer versions of the EPID, which are currently under development. Second, the relation G between offset corrected absolute pixel value and dose rate (Eq. 3) has to be determined. This relation can be used to correct for differences in sensitivity of the ionization chambers. Each image
ii
1.05
1 r 0
-----.------_---____------.
0.95
r 0
.
I 200
400
600
800
Field size(cm’)
Fig. 4. Ratio of doserate obtainedfrom the EPID signal(using Eq. 5) and an ionization chamberin a rod phantom(squares) or in a water phantom (circles) in a 6 MV x-ray beam.The EPID signalhas beenaveragedover 5 x 5 pixels around the centerof the field.
Volume
33,
Number
5, 1995
pixel value is divided by a flat-field image pixel value and multiplied with the function G applied on the dose rate in the flat-field, to obtain the pixel value with average sensitivity. The differences in sensitivity are mainly caused by differences in gain between individual electrometers, which cause stripes on the calibration images. The differences between the individual high voltage switches are much smaller. Finally, the dose rate measured by the EPID is obtained by applying the function f = G-’ (Eq. 7) on these average pixel values(Eq. 5). The relation G was determined for the central 11 X 11 pixels of the EPID. Measurements at several other positions at the EPID showed that the relation was valid for the entire EPID. For dose rates up to 1 Gy/min, a square root relation is followed becausethe signal is almost proportional to the number of ions formed in the liquid. The latter is proportional to the square root of the dose rate (12). This is consistent with the model developed by van Herk, in which it is assumedthat the amount of charge formed during the measurement interval (i.e., 10 to 20 ms) is small compared with the amount of charge stored in the liquid under equilibrium (12). However, the amount of charge formed during the measurement interval depends linearly on the dose rate, and will become increasingly important for higher doserates. For doserates above 1 Gy/min, the linear term must be taken into consideration for an accurate description of the dose rate to signal relation (Eq. 6). The dose rate obtained using Eq. 5 agrees with the actual doserate measuredby an ionization chamber in a miniphantom within 0.6% (1 SD). If a simple squareroot relation between signal and dose rate would be assumedinstead of the exact relation (Eq. 6), the error in dose rate values would still be acceptable for many purposes. For a very inhomogeneous exit dose distribution (e.g., with dose rate values ranging from 0.6 to 1.4 Gy/min and a dose rate value of 1 Gylmin on the central beam axis), the deviation in off-axis dose rate values from the real dose rates would be 2.3% at maximum. Long-tern stability The good long-term stability of the device allows that the frequency with which calibrations have to be performed can be rather low (e.g., once every two months). Ina,and its corresponding dose rate values d,j,n,, are measured at the accelerator at 150 cm SDD, in our institution. Because the radiation output of an accelerator is not constant, dil,fla, has to be determined immediately before or after acquisition of I,,,. When Ina,is acquired at the cobalt unit, Dij.f,a,only has to be measured once becauseof the good stability of this device. However, a different cable might have to be used. This would result in a slightly different signal, which has to be taken into account by an extra matrix of correction factors. Influence of radiation history on detector signal As a result of irradiation, the impurity level of the liquid in the ionization chamber changes. The impurities
Dosimetry
with an EPID
5
without
:
radiation
P-i
.-Ei fn
*
!.
n
:
.-F z u[r 0
5
.
10
15
Time
20
(min)
i? c .-CI) ul
.-: % uK 0.80
1 .oo
Electrode
1.20
distance
1271
25
(a)
0.60
et al.
where the distance between the electrode plates is less than about 0.8 mm. For the two commercial EPIDs we investigated, electrode distances of 0.8 f 0.1 mm (our EPID) and 0.9 + 0.1 mm (the second EPID) have been realized. The observed signal increases of 0.5 2 0.3% and 0.7 2 0.3% for the commercial EPIDs are consistent with this relation. The reproducibility in dose rate determination is, within I%, independent of the integrated absorbed dose, which makes these detectors very promising for dosimetry purposes.
I
with radiation
0 M. ESSERS
1.40
1.60
(mm)
(b) Fig. 5. (a) Relative change in detector signal as a function of time for the prototype detector. The detector is irradiated with a dose rate of 0.8 Gy/min. After 10 min, irradiation has stopped and the recovery of the signal is measured. The signals are averaged over 65 x 65 pixels around the center of the detector and are normalized to the image acquired after 0.06 Gy. (b) Change in detector signal after an absorbed dose of 1.22 Gy as a function of the electrode distance, normalized to an image acquired after 0.06 Gy, obtained from the prototype EPID measurements.
probably act as electron scavengers, which means that a reduction in impurity level in the liquid will give rise to a signal increase (4). Becausethis effect was unacceptably large for prototype EPIDs developed in our institute, these devices were not suitable for accurate dose rate measurements. The average signal of 65 X 65 pixels around the center of a prototype EPID changed up to 4% after 8 Gy (Fig. 5a), which meant an overestimate of the dose rate of 8%. The increase in signal was, in addition, not equal for all ionization chambers. Combination of signal increase measurementswith capacity measurementsfor individual ionization chambers showed that the detector signal increase depends linearly on the distance between the electrodes (Fig. 5b). According to this relation, the signal increase should be smaller than 1% for an EPID
Scattering properties The scatter contribution from the head of the accelerator, and therefore, the signal measured by an ionization chamber in a rod phantom, increases with increasing field size. In the presence of phantom material, the signal also increases due to increase of the phantom scatter contribution. In the future, the EPID will be used for patient transmission dose rate measurements, from which the exit dose values will be calculated. A high spatial resolution is a prerequisite for accurate dose rate measurements. In the case of poor spatial resolution, the signal at a specific point is not only caused by the dose delivered to that point, but also by the dose delivered to nearby chambers. This will result in an overestimation of the dose in low dose regions due to the combined contributions of detector scatter from nearby higher dose regions. The variation of the EPID signal with field size falls between that of an ionization chamber in a miniphantom and in a water phantom, which can be explained by the presence of side scatter material without a substantial amount of backscatter material in the EPID. The spatial resolution of the ionization chambers of the EPID is, therefore, almost as good as the spatial resolution of a single air-filled ionization chamber of the same dimensions. In conclusion, our measurements show that the commercial matrix liquid-filled ionization chamber EPID is suitable for dose rate measurements. It should be noted that it is not possible at this moment to use this device directly for absolute dose measurements. The device will be tested during phantom and in vivo transmission dose rate measurements. In the future, we want to use the detector for various geometric and dosimetric problems; for instance, to check the dose homogeneity over the target volume, to measure the dose delivery to critical organs, to check the influence of wedges and blocks on the dose distribution, and to determine the dose delivery to the patient for intensity modulated beams. By combining these dose rate measurements with a single dose measurement performed with a diode, the integrated dose delivered to patients can, in principle, be assessedin the whole treatment field.
272
I. J. Radiation
Oncology
0 Biology
0 Physics
Volume
33, Number
5, 1995
REFERENCES 1. Bednarek, D. R.; Lanzl, L. H.; Rozenfeld, M. On the application of a liquid-dielectric ionization chamber to “‘Cf dosimetry. Med. Phys. 3:403-409; 1976. 2. Bijhold, J.; Lebesque, J. V.; Hart, A. A. M.; Vijlbrief, R. E. Maximizing setup accuracy using portal images as applied to a conformal boost technique for prostatic cancer. Radiother. Oncol. 24:261-27 1; 1992. 3. Boyer, A. L.; Antonuk, L.; Fenster, A.; van Herk, M.; Meertens, H.; Munro, P.; Reinstein, L. E.; Wong, J. A review of electronic portal imaging devices (EPIDs). Med. Phys. 19:1-16; 1992. chamber using 4. Engler, J.; Keim, H. A liquid ionization tetramethylsilane. Nucl. Instr. Meth. 223:47-51; 1984. 5. Essers, M.; Keus, R.; Lanson, J. H.; Mijnheer, B. J. In vivo dosimetry during the treatment of parotid gland tumours. Radiother. Oncol. 32: 154- 162; 1994. 6. Fiorino, C.; de1 Vecchio, A.; Cattaneo, G. M.; Fusca, M.; Longobardi, B.; Signorotto, P.; Calandrino, R. Exit dose measurements by portal film dosimetry. Radiother. Oncol. 29:336-340; 1993. Kirby, M. C.; Williams, P. C. Measurement possibilities using an electronic portal imaging device. Radiother. Oncol. 29:237-243; 1993. Meertens, H.; van Herk, M.; Bijhold, J.; Bartelink, H. First clinical experience with a newly developed electronic portal imaging device. Int. J. Radiat. Oncol. Biol. Phys. 18:11731181; 1990. Mijnheer, B. J.; Aalbers, A. H. L.; Visser, A. G.; Wittkamper, F. W. Consistency and simplicity in the determination of absorbed dose to water in high-energy photon
10.
11
12. 13. 14. 15.
16. 17.
beams: A new code of practice. Radiother. Oncol. 7:371384; 1986. Van Dam, J.; Vaerman, C.; Blanckaert, N.; Leunens, G.; Dutreix, A.; van der Schueren, E. Are port films reliable for in vivo exit dose measurements? Radiother. Oncol. 25~67-72; 1992. van Gasteren, J. J. M.; Heukelom, S.; Van Kleffens, H. J.; Van der Laarse, R.; Venselaar, J. L. M.; Westermann, C. F. The determination of phantom and collimator scatter components of the output of megavoltage beams: Measurement of the collimator scatter part with a beam-coaxial narrow cylindrical phantom. Radiother. Oncol. 20:250257; 1991. van Herk, M. Physical aspects of a liquid-filled ionization chamber with pulsed polarizing voltage. Med. Phys. 18:692-702; 1991. van Herk, M.; Meertens, H. A matrix ionization chamber imaging device for on-line patient setup verification during radiotherapy. Radiother. Oncol. 11:369-378; 1988. van Herk, M.; Bijhold, J.; Hoogervorst, B.; Meertens, H. Sampling methods for a matrix ionization chamber system. Med. Phys. 19:409-418; 1992. van Tienboven, G. J.; Lanson, J. H.; Crabeels, D.; Heukelom, S.; Mijnheer, B. J. Accuracy in tangential breast treatment setup, a portal imaging study. Radiother. Oncol. 22:317-322; 1991. Wickman, G. Liquid ionization chambers on their design, physical properties and practical application. Ph.D. thesis, University of Umeb, Umel, Sweden; 1991. Wickman, G.; Nystrom, H. The use of liquids in ionization chambers for high precision radiotherapy dosimetry. Phys. Med. Biol. 37, 9:1789-1812; 1992.
Oncology
Biol.
Pergamon
Phys., Vol. 33. No. 5, pp. 1265- 1272. 1995 Copyright 0 1995 Elsevier Science Inc. Printed in the USA. All rights reserved 0360.3016/95 $9.50 + .OO
0360-3016(95)00108-5
l
Treatment
Verification
DOSIMETRIC
MARION
CHARACTERISTICS PORTAL
OF A LIQUID-FILLED IMAGING DEVICE
ESSERS, MSc., BART R. HOOGERVORST, M.Sc., MARCEL HUGO LANSON AND BEN J. MIJNHEER, PH.D.
ELECTRONIC
VAN HERK, PH.D.,
Departmentof Radiotherapy,The NetherlandsCancerInstitute, Antoni van LeeuwenhoekHuis, Plesmanlaan121, 1066CX Amsterdam,The Netherlands Purpose: To determine the characteristics of a commercial electronic portal imaging device (EPID), based on a two-dimensional matrix of liquid-filled ionization chambers, for transmission dose measurements during patient treatment. Methods and Materials: Electronic portal imaging device measurements were performed in a cobalt-60 beam and two accelerator x-ray beams, and compared with measurements performed with a Farmer-type ionization chamber in air in a miniphantom and in an extended water phantom. Results: The warming up time of the EPID is about 1 h. The long-term stability of the detector is better than 1% under reference conditions for a period of about 3 months. The signal of the ionization chambers follows approximately the square root of the dose rate, although the relation becomes more linear for larger (> 1 Gy/min) dose rates. The signal can be transformed to dose rate with an accuracy of 0.6% (1 SD). The short-term influence of integrated dose on the sensitivity of the ionization chambers is small. The sensitivity increases about 0.5% for all ionization chambers after an absorbed dose of 8 Gy and returns to its original value in less than 5 min after stopping the irradiation. This small increase in sensitivity can be ascribed to the electrode distance of the ionization chambers in commercial EPIDs, which is 0.8 + 0.1 mm. The sensitivity increase depends on the electrode distance and is 4% for a 1.4 mm electrode distance. The scattering properties of the EPID ionization chambers were between those of an ionization chamber in a miniphantom and in a water phantom. Conclusion: The matrix ionization chamber EPID has characteristics that make it very suitable for dose rate measurements. It is therefore a very promising device for in vivo dosimetry purposes, Radiotherapy, Portal imaging, mission dosimetry.
Dosimetry,
Ionization
chamber,
Quality
assurance, In vivo dosimetry,
Trans-
INTRODUCTION Various electronic portal imaging devices (EPIDs) are currently available for clinical applications. Some of these systems are based on a scintillating screen/video camera combination, while others are based on scanning techniques that either use a moving array of photodiodes or a matrix of liquid-filled ionization chambers (3). In our institution we have developed an EPID that consists of a matrix of 256 X 256 liquid-filled ionization chambers (12, 13). Because the liquid density is high compared with gassesand signal integration occurs in the liquid (12), a good signal-to-noise ratio can be obtained with small ionization chambers. Electronic portal imaging devices were originally
introduced in the clinic to overcome some drawbacks of conventional portal films, such as the short exposure time, which necessitates an interruption in the patient irradiation, and the amount of time needed to develop and read a film (3). Electronic portal imaging devices are extremely useful for on-line geometrical patient setup measurements (2, 8, 15). They can, in addition, be used for other purposes such as the determination of the diode position during in viva dosimetry (5). An EPID might, however, also be suitable for simultaneous patient set-up measurement and transmission dosimetry in a large number of points; for example, during high dose-high precision treatments. The possibilities of using conventional portal films and
Reprint requeststo: Marion Essers,M.Sc. E-mail address: [email protected] Acknowledgements-This work was financially supportedby the NetherlandsCancer Foundation(NKB Grant NKI 92-40). We would like to thank Nice van Bree (M.Sc.) for helping us
with the measurements performedwith the secondcommercial EPID at the “Zeeuws RadiotherapieInstituut” in Vlissingen, and Ronald Boellaard (M.Sc.) and JoosLebesque(Ph.D.) for many useful discussions. Accepted for publication 24 February 1995.
1265
I. J. Radiation Oncology
1266 commercial
l
Biology
Physics
EPID
CU
T
PCB
2
front liquid
barium E” ij
l
CU -back
ferrite PCB film PCB
I foam
CU
1r~ilb.~..,
;,
,,
_
;; *
__
;
PCB
Fig. 1. Cross-section through the EPID. PCB = printed circuit board. Indicated is a 18 pm thick layer of Cu present at the
sidesof all PCBs. fluoroscopic EPIDs for in viva dose measurementshave already been reported (6, 7, 10). An advantage of EPIDs compared with films is that the results can be obtained on-line. A possible advantage of the matrix ionization chamber device over the other systems might be its high spatial resolution (no light scatter) and high dynamic range. The aim of this study was, therefore, to investigate whether commercial matrix ionization chamber EPIDs are suitable for dosimetry purposes. For that purpose we studied various characteristics of the EPID. First, we wanted to examine the warming-up effect. Second, the relation between signal and dose rate had to be determined. Third, the long-term stability of the detector had to be investigated. Fourth, the influence of radiation history on the signal was examined, Finally, the scattering properties of the EPID were quantified.
METHODS
AND MATERIALS
The commercial ionization chamber EPID’ is based on a prototype developed in our institute (8, 12, 13). A crosssection of the EPID is shown in Fig. 1. The matrix of ionization chambers, consisting of two electrode plates and the liquid, is packed between two foam layers. A lmm thick plate in front of the ionization chambers, made of Plastoferrite,’ is used as a converter material for the production of electrons. The matrix of 65,536 (256 X 256) ionization chambers, in a sensitive area of 32 cm X 32 cm, is formed by cross-points of 256 electrode strips on a front-printed circuit board (PCB) and 256 strips perpendicular to these on the back PCB. At the sides of the EPID, where the plates are glued together, the size of the ionization chambers is 1.27 mm X 1.27 mm X 0.8 mm. For individual ionization chambers, the distance between the electrode plates may slightly differ. The total amount
’ Varian PortalVision, Varian Inc., Palo Alto, CA. * Plastoferrite is a composite material of plastic and barium ferrite with a density of 4.75 g/cm’. ’ Iso-octane, spectroscopical pure, Merck, Darmstadt, Germany.
Volume 33, Number 5, 1995
of material is equivalent to a build-up layer of about 12 mm water. The backscatter material is equivalent to about 5 mm water. As ionizing medium, a liquid dielectric3 is used. The liquid probably contains impurities due to interaction of the liquid with the chamber construction material (1, 17) and diffusion of water through the printing board material. The chambers are scanned row by row by switching the polarizing voltage on the front plate electrodes j. The polarizing voltage, which is controlled by software, is usually set to 250 V and is well below saturation. Each electrode strip on the back plate is connected to an electrometer i. The high dose rate of a radiation beam makes it necessary to measure the ionizing current during the short period when the high voltage is switched on (e.g., 20 ms per row) instead of performing event counting. The scan time of the EPID varies between 1.5 s and 6 s, which enables a quick evaluation of the images, even during one treatment session. This short scan time implies that the dose rate is determined from the ionization chamber signals instead of the dose. The acquisition computer is located outside the treatment room. The acquisition mode used for our measurements was the normal, nonsmoothed sampling mode (standard mode), with a sampling time of 20 ms and an image scan time of 5.6 s (14). Measurements were performed using the 6 and 8 MV x-ray beams of two linear accelerators4 and a cobalt unit. Measurements with a second commercial detector were performed using the 8 MV x-ray beam of another accelerator.5 Reference dose rate measurements were performed with a calibrated small Farmer-type ionization chamber’ in combination with an electrometer.’ At the cobalt unit, the dose rate in air was determined with an ionization chamber with build-up cap. At the linear accelerators, the dose rate was determined at the depth of dose maximum in a small perspex rod miniphantom (11). The dose rate measured by the ionization chamber was calculated following the procedure given in the Dutch dosimetry protocol (9). The incorrect use of this protocol in the case of the small cylindrical phantom leads to a small error (< 0.5%) in the dose rate determination. Detector warming-up time After turning on the EPID, a certain time period is required to obtain a constant reading. This warming-up time has been measuredby making an EPID image every minute for a half hour and every 3 min for the next half hour. To limit radiation damage to the liquid, the radiation
’ Philips SL-15, Crawley, U.K. ’ Varian Clinac 21OOc, Palo Alto, CA. b NE 245,0.2 cm’, NE Technology Limited, ’ Keithley-616, Cleveland, OH.
Berkshire,
U.K.
Dosimetry
with an EPID
was switched on only briefly when each image was made. The measurements have been performed on the cobalt unit at a source-detector distance (SDD) of 80 cm with a field size of 32 x 32 cm*. At this distance, the dose rate was about 0.5 Gy/min. Due to limitations of the software, the high voltage source is only switched on when taking the first image, which means that the warming-up time only begins at that time. Calibration: from detector signal to dose rate There are several imperfections in the detector system, such as gain differences and offset signals. For the purpose of calibration, we will assume that each pixel value is related to the local dose rate, d,j, at the location of pixel (i,j) according to the following equation: P,] = S, * G(d,,) + O,j + E,,
0%. 1)
where P, is the pixel value, S,, describes the relative sensitivity of each individual ionization chamber-electrometer combination, the function G describes the absolute response of the detector (which depends in a nonlinear fashion on the dose rate), and 0, and Ej are offset values attributed to individual pixels and electrometers, respectively. In the commercially available software,’ corrections are applied automatically for E,, which is measured quickly without polarizing voltage prior to acquisition of each image, and O,, which is measured by an image acquisition without radiation. 0, are pixel values of a dark-field image, Idark (13). The offset corrected pixel values, I,* are described by Zij = S, * G(d,) .
(Eq. 2)
If it would be possible to deliver the same dose rate d,, to all pixels, the function G and the relative values of S, could be determined directly. However, in practice, it is difficult to deliver a completely flat field to the detector. For that reason, the calibration procedure consists of two steps. Even for an inhomogeneous field it is possible to define a small detector region such as 11 X 11 pixels near the center of the detector, for which the beam can be assumed to be flat. In this study, the average value of S,, over these 11 X 11 pixels was chosen to be unity. The first step of the calibration procedure is then the determination of the function G by measuring the average offset corrected pixel value I over the area of 11 X 11 pixels for various (homogeneous) dose rates d using Z = G(b)
(W. 3)
* Nucletron, WMS system, Veenendaal, The Netherlands.
1267
0 M. ESSERS et al.
Details about the function G will be described in the results section. For the second step of the calibration, a single image of a more or less “flat” beam is used. In the commercially available software,’ this image would have been used directly to normalize the detector sensitivities, by performing a so-called flat- or flood-field correction (13). For correct dosimetric calibration, however, the dose rate distribution of the “flat” beam must be measured independently. The following relation describes the flat-field image, Z,la,: (Eq. 4) where Liij,na, is the known dose rate distribution of the “flat” beam. Dividing Eq. 2 by Eq. 4, one can easily derive that 0%
5)
where f is the inverse of G. Equation 5 shows that it is possible to derive absolute dose rate values from pixel values and a single calibration image with known dose rate distribution, if the absolute response as a function of dose rate is known for a small part of the detector (function G). It is assumed that the same function G can be applied to describe the dose rate dependence of all pixels (Eq. 1). It should be noted that the flat-field correction algorithm applied by the commercially available software’ computes image pixel values by multiplying ZjjZ,,,Rarwith the average value of Z,jrnat.The resulting image pixel values have to be divided by this average of Zi,,nafto use Eq. 5 to calculate the dose rate from image pixel values. If, in Eq. 5, the average value of Z,,,,la,is used instead of G (d,,,,,a,), errors of several percentages may occur in the calculated dose rate, because inhomogeneities in the dose rate over the “flat” beam are then ignored. For our study, Zna, was acquired at the 8 MV x-ray beam of an accelerator4 at 150 cm SDD in a field of 24 x 24 cm2 at the isocenter. The corresponding dose rate distribution &,, in air was determined under the same conditions using an ionization chamber at the depth of dose maximum in a cylindrical miniphantom, which was scanned using the transport mechanism and software of our water phantom system.x The measurements, performed on a 0.5 X 0.5 cm2 grid, were transformed to the resolution of the EPID by means of bilinear interpolation. The dose rate d used for determination of the function G ranged from 0.1 to 8 GyZmin by changing the SDD from 78 to 180 cm in an 8 MV x-ray beam, using pulse repetition frequencies (PRFs) between 200 and 400 Hz and using metal absorber blocks. At 100 cm from the
1268
I. J. Radiation
Oncology
0 Biology
l Physics
source, the dose rate is about 4.9 Gylmin, using a PRF of 400 Hz. Due to the unsynchronized operation of our imagers, it was difficult to obtain reliable measurements at PRFs lower than 200 Hz. The field size was 20 X 20 cm’ during these measurements. Besides calibration images, capacity images, Icap, have also been measured (12). The ionization chambers are assumed to behave as plane-parallel capacitors. From the capacity C = ES/d (S is the area of the electrode plates, E is the dielectric constant), the distance d between the plates for the ionization chambers of the detector can be estimated. The capacity of the ionization chambers is measured from the ionization chamber signals when applying a triangular high voltage pulse (from 0 V to 35 V and back to 0 V in 40 ms). The distance between the plates at the sides where the printed circuit boards are glued together is taken as the reference distance. Long-term stability To determine the required frequency of calibration, Zsj,j,dark and Zrl,Ra,were acquired under reference conditions for a period of 3 months. For practical reasons, the flat-field images were now acquired at the cobalt unit in a field of 32 X 32 cm’ at 80 cm SDD. The change in individual offset value had been determined from Idark for 12 individual ionization chambers and the average of 65 x 65 pixels around the center of the EPID. The sensitivity of ionization chambers, determined from Z,,,, might change with time because of changes in water and oxygen contamination of the liquid due to radiochemical reactions in the liquid, or due to photon interaction with the chamber material, which might give rise to a difference in ion mobility and recombination constant. Furthermore, aging of the electronics and stray radiation on the semiconductor devices can result in a change in the gain of the electrometers or a change in the gain of a common amplifier. To quantify these effects, the change in pixel gain has been determined from IRat, which was corrected for the decay of the cobalt source, for the same 12 ionization chambers and the average of 65 x 65 pixels around the center of the EPID. Influence of radiation history on detector signal Radiochemical reactions influence the concentration of impurities in the liquid. Therefore, the sensitivity of the EPID, which depends on the purity of the liquid, the chamber dimensions, and the type of surrounding detector material ( 12, 16, 17), is influenced by the short-term radiation history of the EPID. For reliable dose rate measurements, the modification of the signal due to previous irradiations has to be negligible. Such a variation in signal has been quantified by continuously acquiring images during irradiation up to 8 Gy at the cobalt unit in a 32 X 32 cm* field at a dose rate of 0.8 Gy/min, which corresponds roughly with the clinical situation at an accelerator. For comparison, the measurements were also performed using a second commercial detector in a 32 X 32 cm2 field at
Volume
33, Number
5, 1995
an 8 MV x-ray beam of an accelerator.5 Images were made every 8 s during the first 2 min and every 20 s during the next 8 min. The minimum dose required for one image acquisition was approximately 0.06 Gy in this experimental set-up. The recovery of the detector signal has been measured by making an image every few minutes for a half hour after the irradiation has stopped. The irradiation was switched on only during the measurements. No correction has been applied for the additional dose required to measure this recovery. The average signal of 65 X 65 pixels in the center of the EPID has been chosen for analysis. All values have been normalized to the image acquired after 0.06 Gy. Scattering properties To accurately measure the dose rate at a certain point, a high spatial resolution is necessary (i.e., the scatter from neighboring chambers should be small). The scattering properties were quantified in a 6 MV x-ray beam at an SDD of 100 cm, by comparing the average dose rate determined by 5 X 5 pixels around the center of the detector with ionization chamber measurements in a miniphantom and in a water phantom at different field sizes, varying from 3 X 3 cm’ to 25 X 25 cm2. From the water phantom measurements, performed at a depth of 5 cm, the dose at the depth of dose maximum has been assessed using percentage depth dose tables. The SDD for the ionization chambers was 100 cm. At this distance, the dose rate is about 3.7 Gylmin.
RESULTS Detector warming-up time The increase in ionization chamber signal after switching on the high voltage depends on the position of the pixel on the EPID. A small increase, less than l%, was observed for detectors situated near the electrometers, and an increase of almost 2.5% was found for ionization chambers near the high voltage switches and power regulators. No change in signal was observed for ionization chambers situated further away from the electronics. A stable situation was reached after 1 h. Calibration: From detector signal to dose rate The background signal of the ionization chambers, as measured in the dark-field image, was about 4 mV, which is about 2% of the typical image signals measured at the linear accelerator. The electrometer offset, which is calibrated automatically prior to each image acquisition, was about 10 times larger than the ionization chamber offset. From Zeal,differences in pixel sensitivity up to 20% were found. Significant oscillations were present around the average sensitivity of the ionization chambers. The sensitivity also decreased slightly toward the high voltage switches and power regulators.
Dosimetry
with
an EPID
0 M.
ESSERS
et al.
1269
and x = [Zj,/Zv,na,* G(Lj,,,,,)} can be used to calculate the dose rate d, from Ii, for all ionization chambers ij (Eq. 3. The distance between the electrode plates of each ionization chamber, which could be obtained from Icap, ranged between 0.7 and 0.9 mm. Capacity measurements with a second commercial system showed that the distance between the electrode plates varied between 0.8 and 1.0 mm for this device.
o0 25 -
Long-tern stability Every electrometer offset varied around its own average value of about 50 mV (which could differ between individual ionization chambers by a factor of 5) with a standard deviation of about 1 mV. The dark-field signal was constant in time within 0.4 mV, which was 0.2% of the flat-field image signal, for the 12 examined ionization chambers. No substantial change in Z+Ra,or Zr/,cap, greater than 1% for all pixels, was observed with time, during the period of 3 months.
I5 (Gylmin) (4
“05 5 0
2
4
6
6
IO
b (Gylmin) (b) Fig. 2. (a) Measurement of the average offset corrected pixel value over 11 X 11 pixels, Z, vs. the dose rate d. The solid line represents a(d)“’ with a = 1.507 * 10” (Gy/min)-I’*. (b) Ratio of I and G(D) vs. d with a = 1.351 * 10’ (Gy/min)-“’ and b = 81.36 (Gy/min)-’ (Eq. 6). The measurements have been performed in an 8 MV x-ray beam.
In Fig. 2a, the average value I for 11 X 11 pixels around the center of the EPID in pixel values is plotted as a function of the dose rate 6, which is the same for these pixels. The theoretically expected square root relation (12) is also shown in this figure. The signal could not, however, exactly be described by this relation. A better fit of the data could be obtained with: I = G(d) = a * d”’ + b * d,
0%. 6)
with a = 1.351 * lo” (Gy/min)-I’*, b = 81.36 (Gy/min)-’ for the normal acquisition mode of our EPID and fi in Gy/min. The values of parametersa and b depend slightly on the connection cables and strongly on the acquisition mode in which the images are made. The ratio of I and G(B) is plotted in Fig. 2b. On average, this ratio is 0.999 (SD = 0.006) and the maximum deviation of fitted from actual pixel value is 1.2%. The inverse functionf = G-‘, with
./l-d = -a
+(a* + 4 * b * x)“’ 2*b
* ,
0%. 7)
Influence of radiation history on detector signal For the commercial detector used in our study, the increase in ionization chamber sensitivity as a function of integrated dose received by the EPID was only 0.5%, with saturation after less than 1 Gy. After switching off the radiation, the sensitivity decreased exponentially to its original value with a halflife of about 0.75 min. For a second commercial detector, the change in sensitivity as a function of integrated dose was almost identical, at maximum only 0.7%. The results for both detectors are plotted in Fig. 3. Scattering properties The ratio of the doserate obtained from the EPID signal and the rod phantom data increased, whereas the ratio of 1.020 with
7c
radiation
’
without
radiation
1.015
.-p: v) F .% ii u
1.010
1.005
1 .ooo
-& 6 0
: I, . 5
10
Time
15
.
4
20
25
(min)
Fig. 3. Relative change in detector signal as a function of time for our detector (squares) and the second commercial detector (circles). The detectors are irradiated with a dose rate of 0.8 Gylmin. After 10 min, irradiation has stopped and the recovery of the signal is measured. The signals are averaged over 65 x 65 pixels around the center of the detector and are normalized to the image acquired after 0.06 Gy.
1270
1. J. Radiation
Oncology
l Biology
l Physics
EPID and water phantom measurements decreased with increasing field size (Fig. 4). The deviation of EPID measurements from ionization chamber measurements in a rod or a water phantom was 3% at maximum. DISCUSSION
AND
CONCLUSIONS
Detector warming-up time The warm-up time of the EPID is about 1 h. The increase in signal of each ionization chamber depends on its position with respect to the electronics of the EPID. The differences can be attributed to the amount of heat generated by the electronics in the various positions. Near the high voltage switches and the power regulators, where the increase in ionization chamber signal is 2.5%, a total heat of 13 W is dissipated. Near the electrometers much less power (about 3 W) is dissipated. The temperature increase at these ionization chambers is therefore lower, which results in a smaller signal increase of less than 1%. Ionization chambers that are situated far from the electronics do not warm up and therefore, no change of signal with time is observed after switching on the voltage. Calibration: From detector signal to dose rate Equation 5 shows that the calibration procedure to transform pixel values into dose rate values is relatively easy. First, the large electrometer and small ionization chamber offset values have to be subtracted from the signals (Eq. 1). The large electrometer offset values, which result in a limitation of the dynamic range, are caused by imperfections in the highly sensitive electrometer circuitry. These circuits might be improved in newer versions of the EPID, which are currently under development. Second, the relation G between offset corrected absolute pixel value and dose rate (Eq. 3) has to be determined. This relation can be used to correct for differences in sensitivity of the ionization chambers. Each image
ii
1.05
1 r 0
-----.------_---____------.
0.95
r 0
.
I 200
400
600
800
Field size(cm’)
Fig. 4. Ratio of doserate obtainedfrom the EPID signal(using Eq. 5) and an ionization chamberin a rod phantom(squares) or in a water phantom (circles) in a 6 MV x-ray beam.The EPID signalhas beenaveragedover 5 x 5 pixels around the centerof the field.
Volume
33,
Number
5, 1995
pixel value is divided by a flat-field image pixel value and multiplied with the function G applied on the dose rate in the flat-field, to obtain the pixel value with average sensitivity. The differences in sensitivity are mainly caused by differences in gain between individual electrometers, which cause stripes on the calibration images. The differences between the individual high voltage switches are much smaller. Finally, the dose rate measured by the EPID is obtained by applying the function f = G-’ (Eq. 7) on these average pixel values(Eq. 5). The relation G was determined for the central 11 X 11 pixels of the EPID. Measurements at several other positions at the EPID showed that the relation was valid for the entire EPID. For dose rates up to 1 Gy/min, a square root relation is followed becausethe signal is almost proportional to the number of ions formed in the liquid. The latter is proportional to the square root of the dose rate (12). This is consistent with the model developed by van Herk, in which it is assumedthat the amount of charge formed during the measurement interval (i.e., 10 to 20 ms) is small compared with the amount of charge stored in the liquid under equilibrium (12). However, the amount of charge formed during the measurement interval depends linearly on the dose rate, and will become increasingly important for higher doserates. For doserates above 1 Gy/min, the linear term must be taken into consideration for an accurate description of the dose rate to signal relation (Eq. 6). The dose rate obtained using Eq. 5 agrees with the actual doserate measuredby an ionization chamber in a miniphantom within 0.6% (1 SD). If a simple squareroot relation between signal and dose rate would be assumedinstead of the exact relation (Eq. 6), the error in dose rate values would still be acceptable for many purposes. For a very inhomogeneous exit dose distribution (e.g., with dose rate values ranging from 0.6 to 1.4 Gy/min and a dose rate value of 1 Gylmin on the central beam axis), the deviation in off-axis dose rate values from the real dose rates would be 2.3% at maximum. Long-tern stability The good long-term stability of the device allows that the frequency with which calibrations have to be performed can be rather low (e.g., once every two months). Ina,and its corresponding dose rate values d,j,n,, are measured at the accelerator at 150 cm SDD, in our institution. Because the radiation output of an accelerator is not constant, dil,fla, has to be determined immediately before or after acquisition of I,,,. When Ina,is acquired at the cobalt unit, Dij.f,a,only has to be measured once becauseof the good stability of this device. However, a different cable might have to be used. This would result in a slightly different signal, which has to be taken into account by an extra matrix of correction factors. Influence of radiation history on detector signal As a result of irradiation, the impurity level of the liquid in the ionization chamber changes. The impurities
Dosimetry
with an EPID
5
without
:
radiation
P-i
.-Ei fn
*
!.
n
:
.-F z u[r 0
5
.
10
15
Time
20
(min)
i? c .-CI) ul
.-: % uK 0.80
1 .oo
Electrode
1.20
distance
1271
25
(a)
0.60
et al.
where the distance between the electrode plates is less than about 0.8 mm. For the two commercial EPIDs we investigated, electrode distances of 0.8 f 0.1 mm (our EPID) and 0.9 + 0.1 mm (the second EPID) have been realized. The observed signal increases of 0.5 2 0.3% and 0.7 2 0.3% for the commercial EPIDs are consistent with this relation. The reproducibility in dose rate determination is, within I%, independent of the integrated absorbed dose, which makes these detectors very promising for dosimetry purposes.
I
with radiation
0 M. ESSERS
1.40
1.60
(mm)
(b) Fig. 5. (a) Relative change in detector signal as a function of time for the prototype detector. The detector is irradiated with a dose rate of 0.8 Gy/min. After 10 min, irradiation has stopped and the recovery of the signal is measured. The signals are averaged over 65 x 65 pixels around the center of the detector and are normalized to the image acquired after 0.06 Gy. (b) Change in detector signal after an absorbed dose of 1.22 Gy as a function of the electrode distance, normalized to an image acquired after 0.06 Gy, obtained from the prototype EPID measurements.
probably act as electron scavengers, which means that a reduction in impurity level in the liquid will give rise to a signal increase (4). Becausethis effect was unacceptably large for prototype EPIDs developed in our institute, these devices were not suitable for accurate dose rate measurements. The average signal of 65 X 65 pixels around the center of a prototype EPID changed up to 4% after 8 Gy (Fig. 5a), which meant an overestimate of the dose rate of 8%. The increase in signal was, in addition, not equal for all ionization chambers. Combination of signal increase measurementswith capacity measurementsfor individual ionization chambers showed that the detector signal increase depends linearly on the distance between the electrodes (Fig. 5b). According to this relation, the signal increase should be smaller than 1% for an EPID
Scattering properties The scatter contribution from the head of the accelerator, and therefore, the signal measured by an ionization chamber in a rod phantom, increases with increasing field size. In the presence of phantom material, the signal also increases due to increase of the phantom scatter contribution. In the future, the EPID will be used for patient transmission dose rate measurements, from which the exit dose values will be calculated. A high spatial resolution is a prerequisite for accurate dose rate measurements. In the case of poor spatial resolution, the signal at a specific point is not only caused by the dose delivered to that point, but also by the dose delivered to nearby chambers. This will result in an overestimation of the dose in low dose regions due to the combined contributions of detector scatter from nearby higher dose regions. The variation of the EPID signal with field size falls between that of an ionization chamber in a miniphantom and in a water phantom, which can be explained by the presence of side scatter material without a substantial amount of backscatter material in the EPID. The spatial resolution of the ionization chambers of the EPID is, therefore, almost as good as the spatial resolution of a single air-filled ionization chamber of the same dimensions. In conclusion, our measurements show that the commercial matrix liquid-filled ionization chamber EPID is suitable for dose rate measurements. It should be noted that it is not possible at this moment to use this device directly for absolute dose measurements. The device will be tested during phantom and in vivo transmission dose rate measurements. In the future, we want to use the detector for various geometric and dosimetric problems; for instance, to check the dose homogeneity over the target volume, to measure the dose delivery to critical organs, to check the influence of wedges and blocks on the dose distribution, and to determine the dose delivery to the patient for intensity modulated beams. By combining these dose rate measurements with a single dose measurement performed with a diode, the integrated dose delivered to patients can, in principle, be assessedin the whole treatment field.
272
I. J. Radiation
Oncology
0 Biology
0 Physics
Volume
33, Number
5, 1995
REFERENCES 1. Bednarek, D. R.; Lanzl, L. H.; Rozenfeld, M. On the application of a liquid-dielectric ionization chamber to “‘Cf dosimetry. Med. Phys. 3:403-409; 1976. 2. Bijhold, J.; Lebesque, J. V.; Hart, A. A. M.; Vijlbrief, R. E. Maximizing setup accuracy using portal images as applied to a conformal boost technique for prostatic cancer. Radiother. Oncol. 24:261-27 1; 1992. 3. Boyer, A. L.; Antonuk, L.; Fenster, A.; van Herk, M.; Meertens, H.; Munro, P.; Reinstein, L. E.; Wong, J. A review of electronic portal imaging devices (EPIDs). Med. Phys. 19:1-16; 1992. chamber using 4. Engler, J.; Keim, H. A liquid ionization tetramethylsilane. Nucl. Instr. Meth. 223:47-51; 1984. 5. Essers, M.; Keus, R.; Lanson, J. H.; Mijnheer, B. J. In vivo dosimetry during the treatment of parotid gland tumours. Radiother. Oncol. 32: 154- 162; 1994. 6. Fiorino, C.; de1 Vecchio, A.; Cattaneo, G. M.; Fusca, M.; Longobardi, B.; Signorotto, P.; Calandrino, R. Exit dose measurements by portal film dosimetry. Radiother. Oncol. 29:336-340; 1993. Kirby, M. C.; Williams, P. C. Measurement possibilities using an electronic portal imaging device. Radiother. Oncol. 29:237-243; 1993. Meertens, H.; van Herk, M.; Bijhold, J.; Bartelink, H. First clinical experience with a newly developed electronic portal imaging device. Int. J. Radiat. Oncol. Biol. Phys. 18:11731181; 1990. Mijnheer, B. J.; Aalbers, A. H. L.; Visser, A. G.; Wittkamper, F. W. Consistency and simplicity in the determination of absorbed dose to water in high-energy photon
10.
11
12. 13. 14. 15.
16. 17.
beams: A new code of practice. Radiother. Oncol. 7:371384; 1986. Van Dam, J.; Vaerman, C.; Blanckaert, N.; Leunens, G.; Dutreix, A.; van der Schueren, E. Are port films reliable for in vivo exit dose measurements? Radiother. Oncol. 25~67-72; 1992. van Gasteren, J. J. M.; Heukelom, S.; Van Kleffens, H. J.; Van der Laarse, R.; Venselaar, J. L. M.; Westermann, C. F. The determination of phantom and collimator scatter components of the output of megavoltage beams: Measurement of the collimator scatter part with a beam-coaxial narrow cylindrical phantom. Radiother. Oncol. 20:250257; 1991. van Herk, M. Physical aspects of a liquid-filled ionization chamber with pulsed polarizing voltage. Med. Phys. 18:692-702; 1991. van Herk, M.; Meertens, H. A matrix ionization chamber imaging device for on-line patient setup verification during radiotherapy. Radiother. Oncol. 11:369-378; 1988. van Herk, M.; Bijhold, J.; Hoogervorst, B.; Meertens, H. Sampling methods for a matrix ionization chamber system. Med. Phys. 19:409-418; 1992. van Tienboven, G. J.; Lanson, J. H.; Crabeels, D.; Heukelom, S.; Mijnheer, B. J. Accuracy in tangential breast treatment setup, a portal imaging study. Radiother. Oncol. 22:317-322; 1991. Wickman, G. Liquid ionization chambers on their design, physical properties and practical application. Ph.D. thesis, University of Umeb, Umel, Sweden; 1991. Wickman, G.; Nystrom, H. The use of liquids in ionization chambers for high precision radiotherapy dosimetry. Phys. Med. Biol. 37, 9:1789-1812; 1992.