Dosimetric effect of respiration-gated beam on IMRT delivery
`Jun Duan,a) Sui Shen, John B. Fiveash, Ivan A. Brezovich, Richard A. Popple,
`and Prem. N. Pareek
`Department of Radiation Oncology, University of Alabama Birmingham, 619 South 19th Street,
`Birmingham, Alabama 35233
`~Received 8 January 2003; revised 29 April 2003; accepted for publication 27 May 2003;
`published 25 July 2003!
`Intensity modulated radiation therapy ~IMRT! with a dynamic multileaf collimator ~DMLC! re-
`quires synchronization of DMLC leaf motion with dose delivery. A delay in DMLC communication
`is known to cause leaf lag and lead to dosimetric errors. The errors may be exacerbated by gated
`operation. The purpose of this study was to investigate the effect of leaf lag on the accuracy of
`doses delivered in gated IMRT. We first determined the effective leaf delay time by measuring the
`dose in a stationary phantom delivered by wedge-shaped fields. The wedge fields were generated by
`a DMLC at various dose rates. The so determined delay varied from 88.3 to 90.5 ms. The dosim-
`etric effect of this delay on gated IMRT was studied by delivering wedge-shaped and clinical IMRT
`fields to moving and stationary phantoms at dose rates ranging from 100 to 600 MU/min, with and
`without gating. Respiratory motion was simulated by a linear sinusoidal motion of the phantom. An
`ionization chamber and films were employed for absolute dose and 2-D dose distribution measure-
`ments. Discrepancies between gated and nongated delivery to the stationary phantom were ob-
`served in both absolute dose and 2-D dose distribution measurements. These discrepancies in-
`creased monotonically with dose rate and frequency of beam interruptions, and could reach 3.7% of
`the total dose delivered to a 0.6 cm3 ion chamber. Isodose lines could be shifted by as much as 3
`mm. The results are consistent with the explanation that beam hold-offs in gated delivery allowed
`the lagging leaves to catch up with the delivered monitor units each time that the beam was
`interrupted. Low dose rates, slow leaf speeds and low frequencies of beam interruptions reduce the
`effect of this delay-and-catch-up cycle. For gated IMRT it is therefore important to find a good
`balance between the conflicting requirements of rapid dose delivery and delivery accuracy. © 2003
`@DOI: 10.1118/1.1592017#
`American Association of Physicists in Medicine.
`
`Key words: gated IMRT, respiratory gating, IMRT, dynamic wedge, dosimetric effect
`
`I. INTRODUCTION
`
`Advances in technology have allowed delivery of complex
`radiation fluence patterns with the aid of dynamic multileaf
`collimators ~DMLC!. This technique, known as intensity
`modulated radiation therapy ~IMRT!, is capable of generat-
`ing radiation doses highly conformed to the target while
`sparing surrounding normal tissues. It has been successfully
`applied for irradiation of malignant carcinomas in various
`sites.1– 4 In many instances, patients could not have been
`treated without this sophisticated technique.
`Organ motion due to breathing poses a special challenge
`to IMRT, and may have hindered its wider application. Con-
`ventional
`three-dimensional conformal
`radiation therapy
`~3DCRT! accommodates organ motion by applying wide
`margins around the clinical target volume ~CTV!, usually in
`the range of 1 to 2 cm. While such broad margins inevitably
`include significant volumes of normal tissues, their effect on
`doses to tumors and normal tissues can be estimated. In
`IMRT, however, the simultaneous motions of beam and tar-
`get can lead to significant deviations from the planned doses
`that are difficult to predict.5– 8
`In recent years, respiration-gated radiotherapy has shown
`the potential of reducing dose errors by synchronizing radia-
`tion delivery with the patient’s breathing.9–14 Radiation is
`
`delivered within a small window of each respiratory cycle at
`a phase where organ motion is minimal, e.g., at the end of
`expiration or at
`the end of inspiration. Residual motion
`within the gating window is only a small fraction of the full
`movement.
`The operation of linear accelerators in conjunction with
`DMLCs has been extensively studied.15–22 Litzenberg et al.
`identified the limitations of delivery parameters ~dose rate,
`tolerance, leaf speed, total monitor units, etc.! in realistic
`dynamic delivery that affect MLC leaf position errors that
`triggers beam hold-off.22 By incorporating these limitations
`into the leaf sequencing algorithm, they were able to produce
`leaf sequence of sliding-window IMRT fields deliverable
`with the prescribed constant dose rate, requiring less delivery
`time, and having well-defined, calculable transmission dose
`characteristics. Xia et al. investigated the limitations of com-
`munication and sampling rate in IMRT delivery with dy-
`namic multileaf collimator.17 They observed distortions and
`dose variations on intensity patterns delivered with low MU
`and high dose rate due to insufficient sampling and commu-
`nication lag between treatment console and MLC worksta-
`tion. Several authors have investigated the compatibility of
`dynamic wedges and IMRT with gated delivery. Using film
`dosimetry, Kubo and Wang studied one-dimensional dose
`profiles from gated operations of a Varian 2100C accelerator
`
`2241
`
`Med. Phys. 30 (cid:132)8(cid:133), August 2003
`
`0094-2405(cid:213)2003(cid:213)30(cid:132)8(cid:133)(cid:213)2241(cid:213)12(cid:213)$20.00
`
`© 2003 Am. Assoc. Phys. Med.
`
`2241
`
`Varian (Ex. 1021)
`IPR of U.S. Pat. No. 7,961,843
`
`

`

`2242
`
`Duan etal.: Dosimetric effect of respiration gated beam
`
`2242
`
`pected and actual leaf positions, and other information. The
`acquisition, transfer, processing, and storage of this informa-
`tion requires a significant amount of time, resulting in a re-
`ported 50 to 80 ms delay between the accelerator and DMLC
`response to a given condition.22,25,26
`The mechanical limit of leaf speed is approximately 2.8
`cm/s projected to the isocenter. In our system, the leaf speed
`limit is set at 2.5 cm/s for IMRT delivery by the leaf motion
`calculator software. The dose rate is maintained at the value
`programmed at the linear accelerator console provided that
`the required DMLC leaf motions do not result in any leaf
`exceeding the speed limit. Otherwise, the DMLC application
`reduces the dose rate to fit the speed of leaf movement. In-
`dividual leaf speeds within a segment are constant, but dif-
`ferent leaves travel at different speeds. The accuracy of leaf
`position calibration relative to beam central axis is better
`than 1 mm. The precision of leaf position is 0.002 mm de-
`termined by the counts of the encoder per mm ~512 counts
`per mm for the 0.5 mm leaves used in our study!.27 For
`dynamic delivery, a leaf error tolerance setting permits the
`leaf position to be within a small range of the expected po-
`sition. If the position error exceeds the tolerance, an MLC
`interlock will be activated and the beam turned off. The al-
`lowable leaf position tolerance projected to isocenter can be
`set between 0.05 and 0.50 cm. A tight tolerance tends to
`minimize leaf errors. However, too tight a tolerance will
`cause frequent leaf error interlock that results in an unaccept-
`able number of beam interruptions due to tolerance violation.
`A value of 0.20 cm has been used in our system to compro-
`mise between leaf position accuracy and frequency of
`DMLC leaf error interlock.19 This tolerance setting was used
`in all measurements.
`
`B. Dosimetric method for determining effective leaf
`delay time
`
`equipped with an 80-leaf DMLC. They concluded that gated
`operation was compatible with enhanced dynamic wedges
`and IMRT for dose rates up to 320 MU/min, the maximum
`dose rate provided by their Clinac 2100C.23 Solberg et al.
`found that the Novalis micro-MLC could be used in conjunc-
`tion with respiration gating.24 Hugo et al. investigated the
`effect of gating window size and choice of delivery method
`~segmented and dynamic multileaf collimation! on dosimetry
`of gated IMRT using both single and composite field dosim-
`etry. They found that the error could be reduced by decreas-
`ing gating window size, and that in most cases dynamic de-
`livery generated larger delivery errors than segmented
`delivery.
`A commercial gating system has been installed on one of
`our Clinac 21EX linear accelerators ~Varian Medical Sys-
`tems, Palo Alto, CA!, which is equipped with an IMRT ca-
`pable 120-leaf DMLC. The accelerator can furnish dose rates
`up to 600 MU/min, and thereby offers the potential of faster
`treatment delivery. However, the compatibility of such a high
`dose rate with gated IMRT has not been fully documented. A
`delay in the communications between the DMLC and the
`accelerator is known to cause leaf lag and deviations in the
`delivered radiation dose.17,22,16,25 The problem may be exac-
`erbated by the frequent beam interruptions associated with
`gating. In this study, we investigate the dosimetric effect of
`DMLC leaf lag on gated IMRT in relation to dose rate and
`frequency of beam interruption.
`
`II. MATERIALS AND METHODS
`
`A. Intensity modulating DMLC and leaf lag
`
`The leaves of the Varian Millennium 120-Leaf DMLC are
`5 mm thick within the central 20 cm of the field and 1 cm
`thick over the peripheral 10 cm on each side. IMRT is deliv-
`ered with MLC in two modes, segmented multileaf colli-
`mated IMRT ~SMLC-IMRT!, also known as the step-and-
`shoot mode, and dynamic multileaf collimated IMRT
`~DMLC-IMRT!, also known as the sliding window mode. In
`the SMLC-IMRT mode, the beam is turned off while the
`collimator changes to the next shape in the treatment se-
`quence. In the DMLC-IMRT mode, the beam remains on
`while the collimator changes from one shape to the next. The
`relationship between leaf positions and delivered MU is
`specified in a leaf sequence file. The instantaneous shape of
`the beam during the continuous motion of the treatment is
`determined by linear interpolation of MU and leaf position
`between the previous and the next segment in the sequence.
`The instantaneous accumulated dose value is used as an in-
`dex to keep the relationship constant among continuously
`varying factors during the treatment, such as DMLC beam
`shape, dose rate, and movements of the linear accelerator.
`The DMLC-IMRT mode, which is often used in clinics, was
`used in this study.
`The DMLC on our machine operates through a DMLC
`controller which sends commands to the DMLC head assem-
`bly to move the leaves and the DMLC transmits the infor-
`mation back to the controller. Through a communication
`link, the controller checks the status of the accelerator, ex-
`
`Medical Physics, Vol. 30, No. 8, August 2003
`
`D%5
`
`3100%,
`
`As mentioned above, the delay in accelerator feedback to
`the DMLC controller results in a lag between dose delivery
`and the DMLC awareness of the delivered monitor units
`~MU!. Furthermore, leaf tolerance allows additional leaf lag
`up to the tolerance setting. In a dynamic delivery, these lags
`cause a difference between expected and actually delivered
`dose. The percent dose error, D%, can be calculated using
`dt(cid:149)DR
`MU
`where dt is the effective delay time, DR is the dose rate
`~MU/min!, and MU is the monitor unit setting for the DMLC
`segment. Depending on the dose rate and MU setting for the
`DMLC segment, the relative dose error can be significant.
`However, this formula cannot be directly used to experimen-
`tally determine the delay time. For a wedge field shaped by
`opening or closing leaf gap in DMLC-IMRT mode, the delay
`only affects the doses of the beginning and ending segments
`of the DMLC pattern. The beginning segment corresponds to
`either a fully open gap for the wedge shaped by closing the
`DMLC gap or a fully closed gap for the wedge shaped by
`opening the DMLC gap. The ending segment is opposite to
`
`~1!
`
`

`

`2243
`
`Duan etal.: Dosimetric effect of respiration gated beam
`
`2243
`
`sured at various dose rates ~100 MU/min to 600 MU/min!
`with the ion chamber. Monitor units were set at 12 and 56 for
`the 3 and 14 cm wide wedge fields, respectively, so that the
`leaf speeds for the two wedges were identical for a given
`dose rate, and the leaf speed could reach its maximum of 2.5
`cm/s at 600 MU/min. The average of the two measurements
`at collimator angles 180° apart was used in computing the
`leaf delay time. For each data point, three measurements
`were taken and the average was used. The dose correspond-
`ing to zero dose rate, i.e., no leaf lag effect, was obtained by
`linear extrapolation. Measured doses were then normalized
`to the extrapolated dose value of zero dose rate. Linear re-
`gression was used to fit the D%~DR! curve. Using Eq. ~2!,
`the delay time can be determined from the slope of the linear
`function.
`Leaf transmission was measured using the same measure-
`ment geometry with closed leaf gap. The leaf edges were
`offset 7 cm from central axis to avoid leakage contribution
`due to rounded leaf edge. Five thousand MU were delivered
`and the ion chamber was used to measure the dose on central
`axis with the same position and orientation as in field mea-
`surement. Leaf transmission corrections were applied in pro-
`portion to 50% of the MU in wedge field measurements. In
`wedge factor measurements in which static MLC fields were
`used, leaf transmission correction was applied in proportion
`to 100% of the MU if the leaves crossed the central axis.
`
`C. Respiratory gating system
`The Real-Time Position Management ~RPM! Respiratory
`Gating System ~Varian Medical Systems, Palo Alto, CA! was
`used for gated operation. It consists of an infrared video
`camera, an infrared illuminator ring, an infrared-reflective
`marker block and a PC workstation with system control soft-
`ware. An in-room view finder aids in the adjustment of the
`marker block position with respect to the camera. The video
`camera tracks respiratory motion by monitoring the markers
`that are attached to the patient’s chest or abdomen. The gat-
`ing system sends the gating signal to the linear accelerator to
`trigger beam hold-off when the target volume moves beyond
`the preset limits.
`
`D. IMRT dose delivery and measurement
`
`Various IMRT fields were delivered to a rectangular poly-
`styrene phantom. To simulate respiration-induced organ mo-
`tion, the phantom was placed on a platform that could move
`sinusoidally in the crainiocaudal direction. Sinusoidal motion
`was provided by a steel arm connected eccentrically to a
`motorized wheel rotating at a given angular velocity. The
`frequency of sinusoidal motion could be continuously ad-
`justed from 0 to 1.6 cycles/s ~96 cycles/min! by adjusting the
`angular velocity of the wheel. The amplitude of the motion
`was adjustable from 0 to 2.5 cm by changing the distance
`between the mounting point of the arm and the center of the
`wheel, corresponding to a maximum motion range of 5 cm.
`The infrared reflective marker block was placed on the mov-
`ing platform to generate gating signals. For measurements in
`the moving phantom, a motion amplitude of 1 cm ~2 cm
`
`the beginning segment. Taking the wedge shaped by closing
`the leaf gap as an example, dose contribution to the central
`axis is negligible from the ending segment in which the
`DMLC gap is nearly closed. Therefore, the dose error due to
`leaf lag is predominantly contributed by the leaf delay at the
`beginning segment when the leaf gap is fully open, i.e., dD
`5k(cid:149)dt(cid:149)DR, where k5D/MU ~cGy/MU! is MU-to-dose
`conversion factor for the open field. For a wedge field, the
`wedge factor, WF, is defined as WF5D/D 0 , where D 05k
`(cid:149)MU is the dose of the open field and D is the dose of the
`wedge field delivered without leaf lag. Equation ~1! for a
`wedge field becomes
`dD
`D
`
`D%5
`
`3100%5
`
`3100%.
`
`~2!
`
`dt(cid:149)DR
`MU(cid:149)WF
`To obtain the wedge factor, dose delivered by the wedge field
`without delay effect, D, must be determined. Provided that
`D˙ , the dose rate at central axis without leaf lag effect as a
`function of time, is known, the dose of the wedge field can
`be computed by integration
`D˙ ~t !(cid:149)dt,
`where T is the total delivery time. Note that D˙ denotes the
`dose rate at depth on central axis, or cGy/s, while DR de-
`notes the nominal dose rate of the accelerator, or MU/min.
`Assuming that the leaves move at a constant speed, v, then
`the position of the moving leaves, x, can be determined by
`x5v(cid:149)t2w/2, where t and w are delivery time and wedge
`width, respectively. Equation ~3! thus becomes
`E
`D˙ ~x !(cid:149)dx,
`where D˙ (x) is the dose rate at depth on central axis without
`leaf lag as a function of moving leaf position. To eliminate
`leaf lag effect, static MLC fields of gradually decreasing gap
`size were used to determine D˙ (x) at different leaf positions.
`The static gap size was changed gradually by changing the
`position of one bank of leaves while keeping the other bank
`unchanged. Assuming constant DR, for a given MU, D˙ (x)
`can be determined by
`D~x !
`MU/DR
`
`~3!
`
`~4!
`
`D5E
`
`T
`
`0
`
`w/2
`
`2w/2
`
`1 v
`
`D5
`
`D˙ ~x !5
`
`.
`
`~5!
`
`A 0.6 cc ion chamber was used to measure D(x). The ion
`chamber was positioned at the central axis with the long axis
`perpendicular to the direction of wedge gradient. Similar to
`conventional hard wedge measurement, to minimize errors
`due to leaf and ion chamber position accuracy, measurements
`were performed at two collimator angles 180° apart and the
`average was used for calculation. Once D˙ (x) is determined,
`the dose of the wedge field without leaf delay effect, D, is
`computed using Eq. ~4!. The wedge factor, WF, is then de-
`termined by the ratio of D to D 0 .
`To determine the delay time, dt, central axis doses dy-
`namically delivered by 3 and 14 cm wedge fields were mea-
`
`Medical Physics, Vol. 30, No. 8, August 2003
`
`

`

`2244
`
`Duan etal.: Dosimetric effect of respiration gated beam
`
`2244
`
`FIG. 1. Fluence maps used in the study: ~a! 14 cm wide
`single wedge field, ~b! 14 cm wide double wedge field,
`~c! 14 cm wide quadruple wedge field, ~d! 3 cm wide
`single wedge field, ~e! brain IMRT field, ~f! head-and-
`neck IMRT field, ~g! prostate IMRT field, and ~h! lung
`IMRT field.
`
`motion range! was used. Gating windows centered at the end
`of expiration, where motion is minimum, were used for vari-
`ous gating frequencies.6 For stationary measurements with
`gated delivery, the phantom was placed directly on the treat-
`ment couch while the moving platform served as a surrogate
`to generate gating signals. Since the phantom was not mov-
`ing, gating window size and frequency were adjusted for the
`desired number of beam hold-offs per unit MU without the
`concern of residual motion.
`Figure 1 shows the fluence maps of four wedge-shaped
`fields and four clinical IMRT fields used in the study,
`single wedge, 14314 cm2 double
`namely, 14314 cm2
`wedge, 14314 cm2 quadruple wedge, 3314 cm2 single
`wedge, and IMRT fields for brain, head-and neck, prostate
`and lung. In this study, wedge size is specified by its width
`defined as the field dimension in the direction of the wedge
`gradient. DMLC leaf sequence files for the wedge-shaped
`fluence maps were generated with the Varian ‘‘MLC Shaper’’
`software ~version 6.1! while those for the clinical IMRT
`fields were generated with the Varian Cadplan/Helios Treat-
`ment Planning System. The wedge field can be generated by
`either opening the leaf gap from the closed position or clos-
`ing the leaf gap from the open position. To study dose de-
`viations due to beam gating, doses delivered to the stationary
`phantom with gating were compared to those without gating.
`Kodak X-omat V and X-omat TL films ~Eastman Kodak
`Company, Rochester, NY! were used with a film dosimetry
`system to measure planar dose distributions. Special care
`was taken to ensure that the same batch of films was used,
`and that they were developed under the same processor con-
`ditions. The film dosimetry system used for the study in-
`cluded a Vidar VXR-16 DosimetryPro Film digitizer ~Vidar
`Systems Corporation, Herndon, VA! interfaced with a com-
`puter. In-house software and RIT Radiation Therapy Film
`Dosimetry System Software ~Radiological Imaging Technol-
`ogy, Inc., Colorado Springs, CO! was used for film analysis
`and comparisons. For wedge-shaped fields,
`isodose lines
`from 10% to 100% at 10% increment were used to evaluate
`the dose distributions. Distance-to-agreement ~DTA! and
`gamma ~g! were used for quantitative comparison of dose
`
`Medical Physics, Vol. 30, No. 8, August 2003
`
`distributions of gated and nongated delviery. Gamma is an
`index proposed by Low et al. for quantitative evaluation of
`dose distributions.28,29 It is the minimum vector difference in
`dose-distance space between a reference dose distribution
`and a measurement distribution. For a given set of clinical
`criteria, e.g., 3% dose difference and 3 mm distance, gamma
`~denoted as g3/3) evaluates if a measured dose distribution
`agrees with a reference dose distribution within the clinical
`criteria. A gvalue larger than unity indicates that the refer-
`ence dose does not agree with the measured dose within the
`clinic criteria. For wedge fields, we used mean DTA and
`mean gto compare dose distributions of gated and nongated
`deliveries. The mean DTA and gare defined as the average
`of each, respectively, over the central 60% of a measured
`isodose line.
`Absolute dose measurements were performed with an ion
`chamber. For gated deliveries, dosimetric artifacts were gen-
`erated by leaf lag at each leaf position where a beam hold-off
`was triggered. In ion chamber measurements, the effect of
`the artifacts was detected only if it is projected onto the
`chamber volume when the beam hold-off is triggered. A
`small ion chamber would lead to poor reproducibility and
`partial volume effect in measurements because the location
`of beam hold-off was not perfectly reproducible even with
`identical treatment and gating parameters. To overcome this
`problem, a 0.6 cm3 farmer type ion chamber ~PTW Model
`23333! of 23 mm active chamber length was used in the dose
`measurements. The ion chamber was placed such that the
`chamber long axis was along the leaf moving direction un-
`less otherwise specified. For wedge-shaped fields, this corre-
`sponds to the direction of wedge gradient. For comparison
`between gated and nongated deliveries in a stationary phan-
`tom, the chamber position remained unchanged during mea-
`surements of the two sets of deliveries. Measurements in the
`stationary phantom without beam gating were used as stan-
`dards for comparison with other measurements. All measure-
`ments were done at 5 cm depth. To minimize random errors,
`three readings were taken in all ion chamber measurements
`and the average was used in the results presented in this
`study. Majorities of the measurements presented in the study
`
`

`

`2245
`
`Duan etal.: Dosimetric effect of respiration gated beam
`
`2245
`
`FIG. 2. Normalized dose as a function of MLC leaf position measured with
`static MLC fields to simulate 3314 cm2 and 14314 cm2 wide wedge fields
`without leaf lag effect. The measurement was corrected for leaf leakage if
`the leaves blocked the central axis.
`
`were performed in the stationary phantom with the moving
`platform to generate gating signal to gate the accelerator.
`Measurement results presented hereafter, gated or nongated,
`refer to measurements in the stationary phantom unless ex-
`plicitly specified otherwise.
`
`III. RESULTS
`
`A. Leaf delay time
`
`The dose rate at central axis as a function of leaf position
`is shown in Fig. 2 for 3 and 14 cm wide wedge fields. The
`integral dose of Eq. ~4! was computed and compared to the
`dose of the open field to yield wedge factors. For the 3 and
`14 cm wedges, wedge factors were 0.5413 and 0.5086, re-
`spectively.
`Figure 3 shows the percent dose error, D%, as a function
`of dose rate along with the fitting curves for the 3 and 14 cm
`wide wedge fields generated by opening and closing the
`DMLC leaf gap. As can be seen in the figure, dose errors are
`distributed nearly symmetrically about
`the abscissa for
`wedges generated by opening and closing the leaf gap. Lin-
`ear regression was used for the curve fitting. Fitting param-
`eters are listed in Table I. The near unity R 2 values indicate
`a linear relationship between D% and dose rate. For the 3
`
`FIG. 3. Percent dose error, D%, as a function of dose rate along with the
`fitting curves for the 3 and 14 cm wide wedge fields generated by opening
`and closing the DMLC leaf gap. The dose error was calculated by compar-
`ing to the linearly extrapolated dose value that would be measured without
`any leaf error. Twelve and 56 MU were delivered with the 3 and 14 cm wide
`wedges, respectively, to yield a maximum leaf speed of 2.5 cm/s at the dose
`rate of 600 MU/min. Leaf delay times were calculated from the slopes of the
`fitting curves.
`
`and 14 cm wedges, the dose delivered at 600 MU/min was
`higher than that at 100 MU/min by 10.4% and 2.52% when
`the wedges were generated by closing the leaf gap, but lower
`by 11.8% and 2.59% when the wedges were generated by
`opening the leaf gap, respectively. The fact that measured
`doses increased with dose rate when the wedge was gener-
`ated by closing the leaf gap and decreased when the wedge
`was generated by opening the leaf gap confirmed that DMLC
`leaves were lagging behind dose delivery. The effective de-
`lay time was calculated using Eq. ~2! for each wedge and is
`listed in Table I. The effective delay times for the four
`wedges varied between 88.3 and 90.5 ms. They were consis-
`tent for wedges generated by opening and closing the DMLC
`leaf gap.
`
`B. Dosimetric effect on gated dynamic delivery
`
`Figure 4 shows the dose distributions delivered by a 1 cm
`wide DMLC leaf gap sliding across a 10 cm wide field.
`Kodak X-Omat TL film was used for the measurements. The
`dose was delivered with and without gating at a dose rate of
`600 MU/min and a leaf speed of 2.5 cm/s. Dose level is
`represented by the gray scale where black corresponds to
`zero dose. While a uniform dose distribution was obtained
`across the field in nongated delivery @Fig. 4~a!#, distinctive
`
`TABLE I. Effective leaf lag time and linear regression curve fitting parameters for fitting dose error as a function
`of dose rate in nongated delivery.
`
`3 cm wedge
`
`14 cm wedge
`
`Closing leaf
`
`Opening leaf
`
`Closing leaf
`
`Opening leaf
`
`Slope ~min/MU!
`R 2 value
`Effective delay time ~ms!
`
`2.323E204
`0.9991
`90.5
`
`22.283E204
`0.9979
`89.0
`
`5.184E205
`0.9953
`88.6
`
`25.170E-05
`0.9993
`88.3
`
`Medical Physics, Vol. 30, No. 8, August 2003
`
`

`

`2246
`
`Duan etal.: Dosimetric effect of respiration gated beam
`
`2246
`
`FIG. 4. Dose distribution delivered by 1 cm wide DMLC leaf gap sliding
`across a 10 cm wide field ~a! without and ~b! with gating measured with
`Kodak X-Omat TL films. The dose was delivered at dose rate of 600 MU/
`min while the DMLC leaves traveled at 2.5 cm/s. The brightness of the pixel
`represents the dose level.
`
`dosimetric artifacts appeared in gated delivery @Fig. 4~b!#. A
`cold and a hot stripe appeared, respectively, on the left and
`right side of the DMLC window at the position where a
`beam hold-off took place. Dose profiles along the midlines
`across the two fields are shown in Fig. 5. The dose was
`normalized to central axis (distance50). The cold and hot
`spots are about 5% of the average dose.
`For a typical respiratory rate of 15 cycles/min and a gat-
`ing window of 1.0 s, Table II lists the dose ratios of gated to
`nongated deliveries of four different wedge-shaped IMRT
`fields, namely, 14314 cm2 single wedge, 14314 cm2 double
`wedge, 14314 cm2 quadruple wedge, and 3314 cm2 single
`wedge delivered under various conditions. These data were
`measured in the stationary phantom using the ion chamber.
`Differences between gated and nongated beam were less than
`0.82% for leaf speeds ranging from 0.23 to 2.33 cm/s.
`
`FIG. 5. Dose profile across the two fields shown in Fig. 4. Central axis dose
`value of nongated delivery was used for normalization.
`
`As the gating frequency increases or the gating window
`decreases, the number of beam hold-offs increases. For the
`stationary phantom, Fig. 6 shows the dose ratio of gated to
`nongated delivery as a function of the number of beam hold-
`offs triggered during delivery for ~a! a 3 cm wedge generated
`by opening the leaf gap, ~b! a 3 cm wedge generated by
`closing the leaf gap, and ~c! a 14 cm wedge generated by
`opening the leaf gap. Twelve and 56 MU were delivered at
`dose rates from 100 to 600 MU/min for the 3 and 14 cm
`wedges, respectively, corresponding to DMLC leaf speeds
`between 0.42 and 2.5 cm/s. These ion chamber measure-
`ments show that, for a given dose rate, the dose discrepancy
`between gated and nongated deliveries became greater as the
`number of beam hold-offs was increased. For the 3 cm
`wedge the largest dose difference between gated and non-
`gated deliveries, observed at 600 MU/min with 5 beam hold-
`offs, was 3.7%. With 5 beam hold-offs, the dose measured at
`600 MU/min was higher than that at 100 MU/min by 3.0%.
`The dose measured for the 14 cm wedge showed much
`
`TABLE II. Dose ratio of gated to nongated deliveries measured in a stationary phantom with 1.0 s gating window
`and a respiration rate of 15 cycles/min.
`
`Field
`
`Single wedge (14314 cm2)
`Single wedge (14314 cm2)
`Single wedge (14314 cm2)
`Single wedge (14314 cm2)
`Single wedge (3314 cm2)
`Single wedge (3314 cm2)
`Double wedge (14314 cm2)
`Double wedge (14314 cm2)
`Quadruple wedge (14314 cm2)
`Quadruple wedge (14314 cm2)
`Quadruple wedge (14314 cm2)
`
`DR
`~MU/min!
`
`Leaf speed
`~cm/s!
`
`400
`400
`200
`100
`200
`400
`200
`400
`200
`400
`400
`
`0.93
`1.87
`2.33
`2.33
`0.83
`0.83
`0.47
`0.93
`0.93
`0.93
`0.23
`
`MU
`
`100
`50
`20
`10
`12
`24
`100
`100
`50
`100
`400
`
`Dose ratio
`~Gated/nongated!
`
`1.0015
`1.0031
`1.0048
`1.0063
`1.0082
`1.0080
`1.0004
`1.0032
`1.0019
`1.0021
`1.0004
`
`Medical Physics, Vol. 30, No. 8, August 2003
`
`

`

`2247
`
`Duan etal.: Dosimetric effect of respiration gated beam
`
`2247
`
`FIG. 7. Measured doses of a 3314 cm2 wedge field as a function of nominal
`dose rate for various number of beam hold-offs ~BHOs! over 12 MU. The
`wedge field was generated by opening or closing the DMLC leaf gap.
`
`min without respiratory gating. The dose measured without
`gating is also included for comparison. For wedges generated
`by opening and closing the leaf gap, respectively, dose de-
`creased and increased significantly with dose rate. The larg-
`est change, however, occurred when the dose was delivered
`without gating. The dose discrepancy for the 3 cm wedge
`field between dose rates of 100 and 600 MU/min exceeded
`10%.
`For the four clinical IMRT fields presented in Fig. 1, Fig.
`8 depicts the dose as a function of dose rate measured in the
`stationary phantom without gating. The dose was normalized
`to that measured at 100 MU/min. For the prostate and brain
`fields the dose hardly changed with dose rate. For the head-
`and-neck and the lung fields, however, dose increased or
`decreased in proportion to dose rate. The largest dose differ-
`ence was 1.2% in the lung field. For measurements of gated
`delivery in the stationary phantom, Fig. 9 shows the dose of
`the lung field as a function of gating frequency expressed as
`
`FIG. 6. Dose ratios of gated to nongated deliveries as a function of the
`number of beam hold-offs for ~a! the 3 cm wedge field generated by opening
`the leaf gap, ~b! the 3 cm wedge field generated by closing the leaf gap, and
`~c! the 14 cm wedge field generated by opening the leaf gap. Twelve and 56
`MU were delivered for the 3 and 14 cm wedges, respectively. The dose was
`measured in the stationary phantom with or without gating.
`
`smaller difference between gated and nongated delivery be-
`cause of the larger number of monitor units. The largest dif-
`ference shown in the figure is 1.2%. Figure 7 shows the dose
`of the 3 cm wedge as a function of dose rate for 1, 2, 3 and
`5 beam hold-offs triggered during the delivery of 12 MU.
`Dose values were normalized to that delivered at 100 MU/
`
`Medical Physics, Vol. 30, No. 8, August 2003
`
`FIG. 8. Measured doses of four clinical IMRT fields delivered to the station-
`ary phantom without gating as a function of dose rate.
`
`

`

`2248
`
`Duan etal.: Dosimetric effect of respiration gated beam
`
`2248
`
`FIG. 9. Measured dose of a lung IMRT field delivered to the stationary
`phantom with gating as a function of gating frequency for various dose
`rates.
`
`the number of beam hold-offs per 100 MU. Doses with and
`without beam gating exhibited differences of less than 0.8%.
`Figure 10 shows isodose distributions of the 14 cm wedge
`field delivered to a stationary phantom using gated beam
`overlaying those delivered with nongated beams at dose rates
`of ~a! 200 MU/min and ~b! 600 MU/min. A respiratory rate
`of 15 cycles/min and a gating window of 1.0 s center at the
`end of expiration were used in gated delivery. For the dose
`rate of 200 MU/min, the maximum distance-to-agreement
`was about 1 mm @Fig. 10~a!#. However, for the dose rate of
`600 MU/min, the distance-to-agreement was substantially
`greater. The largest mean distance-to-agreement was 3 mm
`@Fig. 10~b!#.
`Table III compares isodose distributions of 14314 cm2
`wedge field delive