Communication and sampling rate limitations in IMRT delivery
`with a dynamic multileaf collimator system
`Ping Xia,a) Cynthia F. Chuang, and Lynn J. Verhey
`The Department of Radiation Oncology, University of California at San Francisco, San Francisco,
`California 94143
`~Received 13 July 2001; accepted for publication 4 December 2001; published 21 February 2002!
`The delivery of an intensity modulated radiation field with a dynamic multileaf collimator ~MLC!
`requires precise correlation between MLC positions and cumulative monitor units ~MUs!. The
`purpose of this study is to investigate the precision of this correlation as a function of delivered
`MUs and dose rate. A semi-Gaussian shaped intensity profile and a simple geometric intensity
`pattern consisting of four square segments were designed to deliver a total of 1, 4, 16, 64, and 100
`MUs at three different dose rates of 100, 400, and 600 MU/min. The semi-Gaussian intensity
`pattern was delivered using both sliding window and step and shoot techniques. The dose profiles
`of this intensity pattern were measured with films. The four square intensity pattern was delivered
`using step and shoot and conventional delivery techniques for comparison. Because of geometrical
`symmetry, the dose to each segment in this intensity pattern is expected to be the same when the
`same MU is assigned to each segment. An ionization chamber was used to measure the dose in the
`center of each of the four square segments. For the semi-Gaussian shaped profile, significant
`artifacts were observed when the profile was delivered with small MUs and/or at a high dose rate.
`For the four square intensity pattern, the dose measured in each segment presented a large variation
`when delivered with small MUs and a high dose rate. The variation increases as the MU/segment
`decreases and as the dose rate increases. These MU and dose rate dependencies were not observed
`when the intensity pattern was delivered using a conventional delivery technique. The observed
`distortion of the semi-Gaussian profile and dose variations among the segments of the four square
`intensity pattern are explained by considering the sampling rate and the communication time lag
`between the control systems. Finally, clinical significance is discussed. © 2002 American Asso-
`@DOI: 10.1118/1.1449496#
`ciation of Physicists in Medicine.
`
`Key words: intensity modulated radiotherapy, dynamic multileaf collimator, step and shoot
`
`I. INTRODUCTION
`
`Computer controlled multileaf collimator ~MLC! systems
`have made intensity modulated radiation therapy ~IMRT!
`clinically practical, using either dynamic or static delivery
`techniques.1–10 The fundamental difference between these
`two delivery methodologies is that with dynamic delivery,
`the radiation and the MLC leaf motions can be executed
`simultaneously,2–5,7 whereas with static delivery, the radia-
`tion and leaf motions are executed sequentially.6,8 –10 The lat-
`ter delivery method resembles conventional delivery, except
`that many segments are included in each given field. Due to
`the use of many small sized segments with associated small
`monitor units ~MUs!, the dose accuracy of static delivery can
`be affected by the accuracy of leaf positioning, and by dose
`nonlinearity for small MU delivery.
`In dynamic delivery, the key factors that affect dose ac-
`curacy include the accuracy of the leaf positions and the
`correlation between the leaf positions and the accumulated
`dose, similar to the situation of the dynamic wedge.11 Unlike
`the dynamic wedge, in which only one pair of jaws is used,
`the dynamic MLC-IMRT delivery employs many pairs of
`MLC leaves, and each has a different intensity profile. In
`order to let all leaves move to their designated positions,
`both leaf speed modulation and dose rate modulation are
`
`needed for implementation of dynamic MLC-IMRT delivery.
`In leaf speed modulation, the leaf speed for each pair of
`MLC leaves is different, but constant within each segment.
`In dose rate modulation, a maximum dose rate is used when-
`ever it is possible to achieve efficient delivery, but when the
`required leaf speed exceeds the maximum mechanical leaf
`speed, the dose rate is reduced.
`A special dynamic MLC delivery technique is considered
`similar to the static MLC delivery. In this delivery method,
`within each segment, the MLC leaf speed is zero, and the
`leaf speed is set to infinity between two segments, thus forc-
`ing the dose rate to be zero, i.e., the beam is off. It should be
`noted that this special dynamic MLC delivery is different
`from static MLC delivery, because in this special dynamic
`MLC delivery, the radiation and the leaf motion are still
`correlated. Despite their difference, both static MLC-IMRT
`delivery and special dynamic MLC-IMRT delivery are often
`referred to as step and shoot delivery. To distinguish these
`two delivery methods, we call the static MLC-IMRT delivery
`method mechanism I step and shoot delivery, in which each
`segment is considered as an individual field. The special dy-
`namic MLC-IMRT delivery method is called mechanism II
`step and shoot delivery, in which the beam off/on command
`is sent from the MLC control station to the machine console
`
`412
`
`Med. Phys. 29 (cid:132)3(cid:133), March 2002
`
`0094-2405(cid:213)2002(cid:213)29(cid:132)3(cid:133)(cid:213)412(cid:213)12(cid:213)$19.00
`
`© 2002 Am. Assoc. Phys. Med.
`
`412
`
`Varian (Ex. 1020)
`IPR of U.S. Pat. No. 7,961,843
`
`

`

`413
`
`Xia, Chuang, and Verhey: Communication and sampling rate limitations
`
`413
`
`FIG. 1. A semi-Gaussian shaped intensity pattern, consisting of five intensity
`levels.
`
`through a periodic comparison between the programmed MU
`to each segment and the cumulative MU controlled by the
`machine console.
`The purpose of this paper is to investigate how the preci-
`sion of the MU and the leaf position correlation varies as a
`function of the delivered MUs and dose rate in a commercial
`dynamic MLC system ~Varian Oncology System, Palo Alto,
`CA!.
`
`II. MATERIALS AND METHODS
`
`A. General description
`
`The multileaf collimator in Varian’s linear accelerator is a
`single-focused MLC,12,13 in which a MLC field follows the
`beam divergence along the direction perpendicular to the leaf
`motion, but not along the direction of the leaf motion. In
`other words, the leaves move along straight lines in a plane
`perpendicular to the central axis of the beam. This design
`simplifies the mechanics of the MLC system, but may cause
`variations in the width of the penumbra when leaves move to
`different locations.12,14 A rounded leaf end is used in this
`MLC system to minimize this effect.15 Due to the rounded
`leaf end, however, the leakage between two leaves when they
`are closed is significant.13 In conventional treatment, these
`closed leaf pairs are normally shielded under the primary and
`the secondary jaws to reduce the leakage between two leaf
`ends. In addition, to minimize leakage radiation between two
`adjacent leaves, a tongue and groove arrangement is used.
`
`B. MLC control system
`
`The MLC control system controls the movement of each
`MLC leaf, including verifying the correspondence of each
`leaf position with its programmed position stored in an
`ASCII file, referred to as a MLC file. The control system can
`be operated either in a static mode ~for a conventional MLC
`field! or a dose mode ~for an intensity modulated MLC field!.
`In the static mode, there is only one position for each leaf for
`a given field. Once all leaves have moved to their pro-
`grammed positions ~within given tolerance!, the associated
`
`Medical Physics, Vol. 29, No. 3, March 2002
`
`FIG. 2. A simple intensity pattern consisting of four 434 cm2 square seg-
`ments, located at an equal distance from the iso-center.
`
`MU is delivered. In the dose mode, there are serial MLC
`positions for a given field stored in a MLC file, in which the
`positions of each MLC leaf are described as a function of a
`dose index ranging from 0.0 to 1.0. The dose index is a
`fraction of the total monitor units for the entire intensity
`modulated ~IM! field.
`
`C. Dynamic MLC delivery
`
`In this system, the machine console and the MLC control
`station separately control the MU delivery and MLC posi-
`tions, respectively. In conventional delivery, since only one
`MLC shape is associated with the total MU, the radiation
`beam is turned on only when the machine console receives a
`‘‘ready’’ signal from the MLC control station. The commu-
`nication between the MLC station and the machine console
`is sequential. Any delay in this communication would have
`no effect on the accuracy of the dose delivered to the field. In
`dynamic delivery, however, the leaf positions and the accu-
`mulated MUs are correlated. This correlation is established
`by communication between the MLC station and the ma-
`chine console every 50 ms, independent of the complexity of
`the intensity profiles. In other words, the sampling frequency
`is fixed at 20 Hz.
`Experiment A: Semi-Gaussian profile. Two experiments
`were designed to investigate the dose accuracy in IMRT de-
`livery using the dynamic MLC. The first experiment was
`designed to study how a semi-Gaussian shaped intensity pro-
`file varies as a function of the dose rate and the delivered
`MU, using both the dynamic and mechanism II step and
`shoot delivery techniques. This semi-Gaussian shape pro-
`duced a simple dose profile and semicontinuous dose inten-
`sity modulation across the field. Therefore, it is particularly
`suitable for dynamic delivery. The intensity map consists of
`five nonzero intensity levels, ranging from 20%, 40%, 60%,
`80%, and 100% of the total MU, as shown in Fig. 1. This
`intensity map was manually input into the planning system
`~CORVUS 3.0, NOMOS Corp., Sewickley, PA!, using the beam
`utility module. The MLC delivery files were created by the
`CORVUS system using the static and dynamic modes for
`Varian linear accelerators, respectively. At dose rates of 100
`and 400 MU/min, total MUs of 1, 4, 16, 64, 100 MUs were
`delivered to the IM field. Films were used to measure the
`dose profiles. In order to avoid the saturation of optical den-
`
`

`

`414
`
`Xia, Chuang, and Verhey: Communication and sampling rate limitations
`
`414
`
`FIG. 3. ~a!–~e! Cross plane intensity profiles for the semi-Gaussian shaped
`IM field, delivered with mechanism II step and shoot method at 100 MU/
`min dose rate for total MUs of 1, 4, 16, 64, 100, respectively.
`
`sity, the enhanced contrast localization films ~EC-L, Kodak,
`Rochester, NY! were used in a cassette for the measurement
`of 1 MU delivery. XTL films were used for the measurement
`of 4 and 16 MU delivery, and XV films were used for the
`measurement of 64 and 100 MU delivery. All films were
`exposed at 1.5 cm depth, 100 cm SSD, except the top surface
`of the EC-L film cassette was set at this depth, and irradiated
`at 6 MV. All films were scanned with a film scanner
`~VXR-12, Vidar Systems Corp., VA, using the Wellhofer
`software—Wellhofer North American, Bartlett, TN!.
`
`Experiment B: Four square intensity pattern. The second
`experiment was designed to deliver an intensity modulated
`field consisting of four 434 cm2 segments located at an
`equal distance ~5.66 cm! from the isocenter, as shown in Fig.
`2. These four segments were first delivered with the mecha-
`nism I step and shoot method, in which the conventional
`delivery mode ~or static mode in the VARIS system, Varian
`Oncology System, Palo Alto, CA! was used, with the same
`MU to each segment. This pattern was also delivered with
`the mechanism II step and shoot method ~using the dose
`
`Medical Physics, Vol. 29, No. 3, March 2002
`
`

`

`415
`
`Xia, Chuang, and Verhey: Communication and sampling rate limitations
`
`415
`
`FIG. 4. ~a!–~e! Cross plane intensity profiles for the semi-Gaussian shaped
`IM field, delivered with mechanism II step and shoot method at 400 MU/
`min dose rate for total MUs of 1, 4, 16, 64, 100, respectively.
`
`mode in the VARIS system!, with the same dose index to
`each segment. The MLC file for the mechanism II step and
`shoot delivery was created manually, following instructions
`in the DMLC manual ~DMLC implementation Guide, Varian
`Oncology System, Palo Alto, CA!. The MLC file consisted
`of a total of eight fields to describe the four segments, since
`each segment requires two MLC fields for the mechanism II
`step and shoot delivery. A Varian Clinac 2300 C/D, 6 MV
`photon beam energy was used. An ionization chamber ~IC10,
`Wellhofer, Wellhofer North American, Bartlett TN! was used
`for measurements, located at 1.5 cm depth inside a solid
`
`water phantom with 100 cm source to surface distance. Be-
`cause of geometric symmetry, the dose delivered to each
`segment should be the same, regardless of the delivery
`method. The relative dose difference D i for each segment is
`defined as
`D i5~D i2D 0!/D 0 ,
`where D i is the dose measured in the ith segment and D 0 is
`the average dose of the four segments. This pattern was de-
`livered at three different dose rates of 100, 400 and 600
`MU/min with total MUs of 1, 4, 16, 64, and 100, respec-
`
`~1!
`
`Medical Physics, Vol. 29, No. 3, March 2002
`
`

`

`416
`
`Xia, Chuang, and Verhey: Communication and sampling rate limitations
`
`416
`
`FIG. 5. ~a!–~e! Cross plane intensity profiles for the semi-Gaussian shaped
`IM field, delivered with the dynamic delivery method at 100 MU/min dose
`rate for total MUs of 1, 4, 16, 64, 100, respectively.
`
`tively. Since there were four segments in this IM field and
`each segment was assigned with the same MU, the MU per
`segment in these measurements was 0.25, 1, 4, 16, and 25
`MU/seg, respectively.
`
`III. RESULTS
`Figures 3~a!–3~e! show cross plane profiles for the semi-
`Gaussian shaped IM field shown in Fig. 1, delivered with the
`mechanism II step and shoot method at 100 MU/min dose
`rate for total MUs of 1, 4, 16, 64, and 100, respectively. All
`profiles in Figs. 3~a!–3~e! were normalized to their maxi-
`
`mum intensities. Since these profiles were obtained from
`three different kinds of films with different sensitivities, only
`the shapes of these profiles are important. The scales of the
`relative intensities, therefore, are not shown in Figs. 3~a!–
`3~e!. Figures 4~a!– 4~e! are cross plane profiles for the same
`IM field as in Fig. 3, but delivered at 400 MU/min dose rate.
`In Figs. 3 and 4, severe distortions were observed in profiles
`delivered with small MUs, such as in Figs. 3~a! and 3~b!, and
`4~a! and 4~b!. The shorter the beam-on time, the more seri-
`ous the observed distortions were. In Fig. 3~a!, there is a
`somewhat semi-Gaussian shape in the profile, which was de-
`
`Medical Physics, Vol. 29, No. 3, March 2002
`
`

`

`417
`
`Xia, Chuang, and Verhey: Communication and sampling rate limitations
`
`417
`
`FIG. 6. ~a!–~e! Cross plane intensity profiles for the semi-Gaussian shaped
`IM field, delivered with the dynamic delivery method at 400 MU/min dose
`rate for total MUs of 1, 4, 16, 64, 100, respectively.
`
`livered with a lower dose rate of 100 MU/min, whereas the
`profile is strongly distorted, as shown in Fig. 4~a!, when it
`was delivered at the higher dose rate of 400 MU/min. When
`the total MU is increased to more than 16 MU as shown in
`Figs. 3~c!–3~e!, and Figs. 4~c!– 4~e!, no obvious distortions
`were observed in these profiles. Similarly, Figs. 5~a!–5~e!
`and 6~a!– 6~e! show cross plane profiles for the IM field de-
`livered with the dynamic mode at 100 and 400 MU/min,
`respectively, with the same total MUs as in Fig. 3. Again,
`
`severe distortions were observed in profiles delivered with
`short beam-on times, such as in Figs 5~a! and 6~a!.
`Figure 7 shows the relative dose variations among the
`four segments delivered with the mechanism I step and shoot
`method, with 1 MU/seg at three different dose rates of 100,
`400, and 600 MU/min, respectively. The relative dose varia-
`tions were calculated according to Eq. ~1!. The observed
`dose variations among the segments were about 1%–3%,
`almost independent of the dose rate. As mentioned in Sec. II,
`
`Medical Physics, Vol. 29, No. 3, March 2002
`
`

`

`418
`
`Xia, Chuang, and Verhey: Communication and sampling rate limitations
`
`418
`
`sampling theory, any arbitrary intensity profile can be recon-
`structed if the sampling frequency is at least twice as large as
`the maximum spatial variation frequency ~Nyquist condi-
`tion!. Some of the results of these two experiments can be
`explained by sampling theory. The sampling frequency used
`in this IMRT dynamic delivery system is fixed at 20 Hz. At
`this sampling rate, the total delivery MU and the dose rate
`determine the number of sampling points taken for each de-
`livery. Smaller MUs with a higher dose rate include fewer
`sampling points, resulting in the observed distortions, as
`shown clearly in Figs. 3~a!, 4~a!, 5~a!, and 6~a!. Table II
`shows the number of sampling points calculated in the deliv-
`ery of a total MU of 1, 4, 16, 64, and 100 MU at dose rates
`of 100, 400, and 600 MU/min, respectively. The formula
`used for this calculation is
`
`N560
`
`T MU
`R
`
`r,
`
`~2!
`
`where N is the number of sampling points, T MU is the total
`MU, R is the delivery dose rate in MU/min, and r is the
`sampling rate in s21. The coefficient 60 converts minutes
`T MU5100 MU,
`into
`seconds.
`For
`example,
`if
`R
`5100 MU/min, and r520 s21, then N51200.
`According to sampling theory, more complex intensity
`profiles require more frequent sampling to accurately repro-
`duce the intensity profile. In other words, to deliver complex
`intensity profiles accurately, larger MU and/or lower dose
`rates are necessary to have sufficient beam-on time to in-
`clude an adequate number of sampling points. It should be
`noted that the concept of sampling points is different from
`the concept of control points introduced in other papers16,17
`in studies of leaf sequencers or interpreters to deliver inten-
`sity modulated beams using dynamic MLC systems. To real-
`ize a continuous intensity profile using a dynamic MLC, the
`MLC control system requires the leaf setting at a series of
`control points. These control points approximate a continu-
`ous intensity profile with a discrete profile. In step and shoot
`IMRT delivery, an intensity map is decomposed into multiple
`segments, and two control points define a segment, espe-
`cially with mechanism II step and shoot delivery. It is rea-
`sonable to argue that for dynamic or mechanism II step and
`shoot delivery, the number of sampling points should at least
`be equal to that of the control points, provided that the sam-
`pling points are synchronized with the control points.
`The semi-Gaussian shaped IM field consists of 9 seg-
`ments for step and shoot delivery and 18 control points for
`dynamic delivery. As shown in Table II, if the IM field is
`delivered with a total of 1 or 4 MU at 400 MU/min dose rate,
`and 1 MU at 100 MU/min dose rate, the included number of
`the sampling points in these deliveries is less than the num-
`ber of control points or segment defining points. With a total
`of 4 MU delivered at 100 MU/min dose rate, even though the
`number of the sampling points included is 48, greater than
`the control points or segment defining points, a small distor-
`tion at the tail of the profile is still observed.
`For the simple IM field shown in Fig. 2, at least 8 sam-
`pling points are needed since it consists of four segments. 8
`
`FIG. 7. The relative dose variations among the four square segments deliv-
`ered with mechanism I step and shoot method, with 1 MU /seg at three
`different dose rates of 100, 400, and 600 MU/min.
`
`the doses to these segments should be identical, but the im-
`perfect field flatness and symmetry and other measurement
`uncertainties may contribute to this small 1%–3% dose
`variation.
`Figures 8~a!– 8~e! show the dose variations of each seg-
`ment for the same IM field ~Fig. 2! delivered with the
`mechanism II step and shoot method, with a total of 1, 4, 16,
`64, and 100 MUs at dose rates of 100, 400, and 600 MU/
`min, respectively. A clear trend observed in these figures is
`that the magnitude of the dose variations among the seg-
`ments increases as the dose rate increases. At a given dose
`rate ~400 MU/min!, as shown in Fig. 9, the magnitude of
`dose variations decreases as the total MU increases. These
`results indicate that the longer the beam-on time for each
`segment, the smaller the relative dose variations. If one sets
`the acceptable dose variation to be about 3%, similar to dose
`variations in the mechanism I step and shoot delivery, Table
`I lists the minimum MUs required to achieve ,3% dose
`variation among segments for this simple intensity pattern
`delivered with three dose rates. From Table I, a total
`beam-on time of about 10 s is needed. It should be noted that
`the required total MUs depend on the complexity of an in-
`tensity pattern.
`Another result observed from Figs. 8~a! to 8~e! is that the
`first segment always received more dose than expected,
`while the last segment always received less dose than ex-
`pected. For a given total MU, the average from each segment
`is almost a constant, as shown in Fig. 10, independent of
`delivery dose rate, or beam-on time. This result can be ex-
`plained that because the total MU is controlled only by the
`machine console, the radiation beam will be terminated when
`the expected total MU is delivered, regardless of the MLC
`control system.
`
`IV. DISCUSSION
`In IMRT delivery using a dynamic MLC, the key is to
`establish a precise correlation between the MLC positions
`and the cumulative MUs. This relationship is analogous to
`the relationship between the spatial domain ~MLC positions!
`and the frequency domain ~cumulative MUs!. According to
`
`Medical Physics, Vol. 29, No. 3, March 2002
`
`

`

`419
`
`Xia, Chuang, and Verhey: Communication and sampling rate limitations
`
`419
`
`FIG. 8. ~a!–~e! The relative dose variations among the four square segments
`delivered with mechanism II step and shoot method, with total MUs of 1, 4,
`16, 64, and 100 MU at dose rates of 100, 400, and 600 MU/min.
`
`sampling points at a sampling rate of 50 ms is equal to 400
`ms, corresponding to 0.67, 2.67, and 4 MU at dose rates of
`100, 400, and 600 MU/min, respectively. If the sampling
`points and the segment defining points are out of synchroni-
`zation, more than 2 sampling points will be needed for each
`segment. The result of our experiment, however, indicates
`that a beam-on time of about 10 s is necessary in order to
`obtain about 3% dose accuracy for each segment. This
`beam-on time consists of 200 sampling points, i.e., 50 sam-
`pling points per segment. For this simple intensity pattern, it
`is, however, difficult to explain why 50 sampling points per
`segment are needed even if the sampling points and segment
`defining points are out of synchronization. In addition, sam-
`pling theory cannot explain why the first segment is always
`
`overdosed and the last segment is always underdosed. The
`communication time lag between the two control systems,
`i.e., the machine console and the MLC station may explain
`this result. It has been confirmed by the vendor ~Varian On-
`cology System, Palo Alto, CA! that this communication time
`lag can be 50 ms. Even with sufficient sampling points and
`depending on the coherence between the sampling points and
`the segment defining points, the communication lag between
`the two control systems causes the machine console to ter-
`minate 50 ms more or less later than it should for each seg-
`ment. The MU delivered during this time lag can be esti-
`mated as follows:
`
`DD5tR,
`
`~3!
`
`Medical Physics, Vol. 29, No. 3, March 2002
`
`

`

`420
`
`Xia, Chuang, and Verhey: Communication and sampling rate limitations
`
`420
`
`TABLE III. Time delay in milliseconds ~ms! calculated for the first segment
`in the four segment IM field according to the measured dose error shown in
`Fig. 8 using Eq. ~1!.
`
`Total
`
`1 MU
`
`4 MU
`
`16 MU
`
`64 MU
`
`100 MU
`
`MU/seg
`100 MU/min
`400 MU/min
`600 MU/min
`
`0.25
`40.3 ms
`68.9 ms
`50.3 ms
`
`1
`43.6 ms
`50.3 ms
`62.5 ms
`
`4
`86.5 ms
`63.7 ms
`66.7 ms
`
`16
`78.5 ms
`59.3 ms
`66.6 ms
`
`25
`92.6 ms
`64.0 ms
`77.3 ms
`
`where t is the time lag, and R is the dose rate in MU/s. If the
`programmed MU to the first segment is D 1 , the relative dose
`error caused by this time delay is
`~4!
`d5DD/D 1 .
`R5400 MU/min
`D 151 MU,
`if
`For
`example,
`t550 ms50.05 s, d533%. From Fig. 8~b!,
`56.67 MU/s,
`the dose variation for this first segment is about 33%. The
`agreement between the dose error estimated from Eqs. ~3!
`and ~4! and the measured dose difference indicates that the
`dose error for this segment is mainly from the communica-
`tion delay, not from undersampling. For the first segment
`delivered with three different dose rates and the five given
`total MUs, Table III shows the calculated time delays accord-
`ing to Eqs. ~3! and ~4!. The relative dose errors used in the
`calculation were from the measurement data shown in Figs
`8~a!– 8~e!. Only the time delays for the first segment were
`calculated because the measured relative dose errors in the
`other segments may not be completely attributed to the time
`delay. The calculated time delays in Table III include not
`only the communication lag between the two controllers, but
`also the time between the two sampling points, which is also
`50 ms. The added communication lag changes the effective
`sampling rate such that it becomes somewhat irregular. The
`middle sampling points force the MLC leaves to catch up
`with the MU delivery, and the middle segments may some-
`times need to be skipped in order to catch up with the MU
`delivery. The machine console terminates the total MU
`whenever it reaches the programmed MU, regardless of
`whether the later segments receive the intended MU or not
`since the total MU is controlled by the machine console,
`independent of the MLC control station.
`In IMRT delivery, due to the use of small MU, the dose
`accuracy may also be affected by changes in dose linearity
`and symmetry while delivering these small MU fields. The
`details of machine-related quality assurance issues have been
`discussed in our previously published paper.18 Briefly, the
`dose linearity measured down to 0.1 MU/segment delivered
`with mechanism II step and shoot method for the Clinac
`2300C/D used in this study was less than 1% different from
`
`FIG. 9. The relative dose variations among the four square segments deliv-
`ered with a dose rate of 400 MU/min, at a total of 1, 4, 16, 64, 64, and 100
`MU.
`
`FIG. 10. The average dose of the four square segments delivered with
`mechanism II step and shoot method, is nearly a constant, independent of
`the delivery dose rates.
`
`TABLE I. Total MUs and beam-on time required to achieve about 3% dose
`variation among the four square segments.
`
`Dose rate
`~MU/min! Segment 1 Segment 2 Segment 3 Segment 4
`
`Total
`MU
`
`Time
`~s!
`
`100
`400
`600
`
`13.6%
`12.5%
`13.1%
`
`20.3%
`0%
`20.3%
`
`21.0%
`20.4%
`20.5%
`
`22.3%
`22.1%
`22.2%
`
`16
`64
`100
`
`9.6
`9.6
`10
`
`TABLE II. Number of sampling points calculated according to the sampling
`rate of 20 Hz at three different dose rates for five given total MUs.
`
`TABLE IV. Average standard deviations ~%! for experiment B.
`
`1 MU
`
`4 MU
`
`16 MU
`
`64 MU
`
`100 MU
`
`1 MU
`
`4 MU
`
`16 MU
`
`64 MU
`
`100 MU
`
`100 MU/min
`400 MU/min
`600 MU/min
`
`12
`3
`2
`
`48
`12
`8
`
`192
`48
`32
`
`768
`192
`128
`
`1200
`300
`200
`
`100 MU/min
`400 MU/min
`600 MU/min
`
`7.3%
`8.8%
`13.4%
`
`2.3%
`4.6%
`7.2%
`
`1.2%
`2.2%
`1.3%
`
`0.24%
`0.44%
`0.94%
`
`0.12%
`0.19%
`0.73%
`
`Medical Physics, Vol. 29, No. 3, March 2002
`
`

`

`421
`
`Xia, Chuang, and Verhey: Communication and sampling rate limitations
`
`421
`
`FIG. 11. A histogram of the number of segments as a function of MU/seg for
`a nasopharyngeal plan, using 15 gantry angles, and 10 intensity levels with
`a total of 491 segments.
`
`the dose delivered by the conventional method with the same
`total MUs. An ionization chamber was used to measure point
`doses at several symmetric locations. The field symmetry
`and flatness for the Clinac 2300C/D using mechanism II step
`and shoot delivery were found comparable with that in con-
`ventional delivery.
`Since the measurement conducted in this study involved
`very small MUs, the issue of measurement uncertainty needs
`to be addressed, especially for experiment B. To measure the
`dose in each segment with different total MUs at different
`dose rates, three readings were taken in each measurement.
`The average standard deviations of doses measured from all
`four segments delivered with mechanism II step and shoot
`delivery are listed in Table IV. The greatest measurement
`uncertainty is found to be 13.4% for the shortest beam-on
`time for the delivery of a total of 1 MU at 600 MU/min. This
`uncertainty decreases as the beam-on time increases, as
`shown in Table IV. The relatively larger measurement uncer-
`tainties for shorter beam-on time may be due to the limited
`sensitivity in the electrometer, or the stability of the linear
`accelerator. These measurement uncertainties, however,
`should not alter the results of this study.
`
`V. CLINICAL SIGNIFICANCE
`With increasing numbers of beam angles and intensity
`levels, a clinical treatment plan may include many small MU
`segments and the accuracy of dose delivered to these seg-
`ments may be affected. Figure 11 shows a histogram of the
`number of segments as a function of MU/seg for a nasopha-
`ryngeal plan using 15 gantry angles, and 10 intensity levels
`with a total of 491 segments. This plan was used for a patient
`treatment, and the total delivery time for this plan was about
`20 min. Due to the efficiency of Varian IMRT delivery, com-
`plex plans such as this are used in clinics. In Fig. 11, the 1–2
`MU/seg bin has the highest incidence. Since these small MU
`segments are spread out through the entire IMRT fields, our
`phantom plan measurement indicates that the effects of sam-
`pling error and timing error are not significant in this clinical
`case. Specifically, ionization chamber measurements for a
`phantom plan created for this patient yield a dose difference
`within 2% between the measured and the calculated doses in
`a relatively uniform high dose region of 92% of the maxi-
`
`Medical Physics, Vol. 29, No. 3, March 2002
`
`FIG. 12. Cross plane profiles delivered with 96 MU at three different rates of
`100, 400, and 600 MU/min compared for three different MLC leaf pairs
`shown in ~a!, ~b!, and ~c!.
`
`

`

`422
`
`Xia, Chuang, and Verhey: Communication and sampling rate limitations
`
`422
`
`FIG. 13. ~a! A lateral portal film image for a head and neck case. The
`intensity pattern of this field is superimposed with the patient’s anatomy. ~b!
`The planned intensity pattern for this lateral field.
`
`mum dose while at the low dose region of 30% of the maxi-
`mum dose, the measured dose was 5% higher than the pre-
`dicted iso-dose line.
`Furthermore, we delivered the AP field intensity pattern of
`this plan with the planned 96 MU to XV films using three
`different dose rates. The films were positioned at 1.5 cm
`depth in a solid water phantom, and the exposed films were
`scanned with a film scanner. Cross plane profiles of a se-
`lected MLC leaf pair were compared for 100, 400, and 600
`MU/min delivery, as shown in Fig. 12~a!. The profiles were
`normalized to the maximum dose, and the spatial location of
`the maximum dose was manually matched. Figures 12~b!
`and 12~c! show cross plane profiles of two other MLC pairs.
`Again, Figs. 12~a!–12~c! show that with a clinical dose of
`180 cGy, the effects of undersampling and communication
`time lag are not significant.
`At a very low dose, 10 cGy, for example, the effects of
`undersampling and communication time lag are significant.
`One scenario related to such a low dose delivery is when an
`intensity pattern is verified using a portal film. To avoid over-
`exposure of the film, a special port film plan with a total dose
`of 10 cGy was created for the patient. The average MU to
`each IMRT field was about 5–7 MU, and an open field port
`with 1–2 MU was added to the IMRT pattern to image the
`patient anatomy surrounding the treated region. Figure 13~a!
`shows a port film image for the patient with the intensity
`pattern superimposed. Figure 13~b! shows the intensity pat-
`tern printed from the treatment planning system. Figure
`13~b! is a fluence pattern while Fig. 13~a! is an attenuation
`pattern, but their correlation should be straightforward. In
`Fig. 13~a!, arrows are used to point out the differences be-
`tween the two intensity patterns. To eliminate possible at-
`tenuation differences due to anatomic structures, Fig. 14~a!
`shows the same pattern for Fig. 12 delivered to a solid water
`phantom with an EC-L port film embedded in a cassette,
`positioned at 1.5 cm depth, with a total of 5 MU and deliv-
`ered at 400 MU/min. Under the same setup, Fig. 14~b! shows
`the same intensity pattern delivered to a XV film with a total
`of 96 MU when the plan is prescribed to 180 cGy. Figure
`14~c! shows an intensity pattern printed from the treatment
`planning system. With the higher MU, the agreement be-
`tween the delivered intensity pattern ~Fig. 14~b!! and the
`expected pattern ~Fig. 14~c!! is nearly perfect by visual in-
`
`Medical Physics, Vol. 29, No. 3, March 2002
`
`FIG. 14. ~a! An intensity pattern from a patient plan delivered to a EC-L
`portal film placed under 1.5 cm solid wat