Int J Radiation Oncology Biol Phys , Vol 57, No 4, pp 1019 –1032, 2003
`Copyright © 2003 Elsevier Inc
`Printed in the USA All rights reserved
`0360-3016/03/$–see front matter
`
`doi:10.1016/S0360-3016(03)00663-1
`
`CLINICAL INVESTIGATION
`
`Ovary
`
`WHOLE ABDOMINOPELVIC RADIOTHERAPY (WAPRT) USING INTENSITY-
`MODULATED ARC THERAPY (IMAT): FIRST CLINICAL EXPERIENCE
`
`WIM DUTHOY, M.D.,* WERNER DE GERSEM, IR.,* KOEN VERGOTE, M.SC.,* MARC COGHE, LIC.,*
`TOM BOTERBERG, M.D., PHD.,* YVES DE DEENE, M.SC., PHD.,* CARLOS DE WAGTER, IR., PHD.,*
`SIMON VAN BELLE, M.D., PHD.,† AND WILFRIED DE NEVE, M.D., PHD.*
`*Division of Radiotherapy and †Department of Medical Oncology, Ghent University Hospital, Ghent, Belgium
`
`Purpose: Whole abdominopelvic radiation therapy (WAPRT) is a treatment option in the palliation of patients
`with relapsed ovarian cancer. With conventional techniques, kidneys and liver are the dose- and homogeneity-
`limiting organs. We developed a planning strategy for intensity-modulated arc therapy (IMAT) and report on the
`treatment plans of the first 5 treated patients.
`Methods and Materials: Five consecutive patients with histologically proven relapsed ovarian cancer were sent
`to our department for WAPRT. The target volumes and organs at risk (OAR) were delineated on 0.5-cm-thick
`CT slices. The clinical target volume (CTV) was defined as the total peritoneal cavity. CTV and kidneys were
`expanded with 0.5 cm. In a preset range of 8° interspaced gantry angles, machine states were generated with an
`anatomy-based segmentation tool. Machine states of the same class were stratified in arcs. The optimization of
`IMAT was done in several steps, using a biophysical objective function. These steps included weight optimization
`of machine states, leaf position optimization adapted to meet the maximal leaf speed constraint, and planner-
`interactive optimization of the start and stop angles. The final control points (machine states plus associated
`cumulative monitor unit counts) were calculated using a collapsed cone convolution/superposition algorithm. For
`comparison, two conventional plans (CONV) were made, one with two fields (CONV2), and one with four fields
`(CONV4). In these CONV plans, dose to the kidneys was limited by cerrobend blocks. The IMAT and the CONV
`plans were normalized to a median dose of 33 Gy to the planning target volume (PTV). Monomer/polymer gel
`dosimetry was used to assess the dosimetric accuracy of the IMAT planning and delivery method.
`Results: The median volume of the PTV was 8306 cc. The mean treatment delivery time over 4 patients was 13.8
`min. A mean of 444 monitor units was needed for a fraction dose of 150 cGy. The fraction of the PTV volume
`receiving more than 90% of the prescribed dose (V90) was 9% higher for the IMAT plan than for the CONV4
`plan (89.9% vs. 82.5%). Outside a build-up region of 0.8 cm and 1 cm away from both kidneys, the inhomogeneity
`in the PTV was 15.1% for the IMAT plans and 24.9% for the CONV4 plans (for CONV2 plans, this was 34.9%).
`The median dose to the kidneys in the IMAT plans was lower for all patients. The 95th percentile dose for the
`kidneys was significantly higher for the IMAT plans than for the CONV4 and CONV2 plans (28.2 Gy vs. 22.2
`Gy and 22.6 Gy for left kidney, respectively). No relevant differences were found for liver. The gel-measured dose
`was within clinical planning constraints.
`Conclusion: IMAT was shown to be deliverable in an acceptable time slot and to produce dose distributions that
`are more homogeneous than those obtained with a CONV plan, with at least equal sparing of the OARs.
`© 2003 Elsevier Inc.
`
`Intensity-modulated arc therapy (IMAT), Whole abdominopelvic radiotherapy (WAPRT), Ovarian cancer.
`
`INTRODUCTION
`
`The treatment options for most patients with relapsed ovar-
`ian cancer are palliative. For patients who relapse within 12
`months after platinum-based schedules, second- or third-
`
`line chemotherapy is often used, but the response rates
`average between 10% and 30% (1).
`A difficult problem to palliate is bowel obstruction.
`Surgery may be attempted but the disease is often multi-
`focal making palliative resection impossible. In reports of
`
`Reprint requests to: Wim Duthoy, Ghent University Hospital,
`Division of Radiotherapy, De Pintelaan 185, 9000 Ghent, Bel-
`gium. Tel: 32-92403074; Fax: 32-92403863; E-mail: wimd
`@krtkg1.rug.ac.be
`The project “Conformal Radiotherapy Ghent University Hospi-
`tal” is supported by the Belgische Federatie tegen Kanker and by
`grants from the Fonds voor Wetenschappelijk Onderzoek (FWO)
`Vlaanderen (G.0183.03),
`the University of Ghent
`(GOA
`
`12050401, BOF 01112300, 011VO497, 011B3300), and the Cen-
`trum voor Studie en Behandeling van Gezwelziekten.
`Acknowledgments—Wim Duthoy is a Research Assistant (Aspir-
`ant) of the FWO. Yves De Deene is Post-Doctoral Research
`Fellow of the FWO.
`Received Jan 16, 2003, and in revised form May 21, 2003.
`Accepted for publication May 27, 2003.
`
`1019
`
`Page 1 of 14
`
`Elekta Exhibit 1005
`
`

`
`1020
`
`I. J. Radiation Oncology ● Biology ● Physics
`
`Volume 57, Number 4, 2003
`
`Redman et al. (2) and Krebs et al. (3), 10 –15% of patients
`died within 8 weeks after surgery and 35–38% had no
`clinical benefit. The results of chemotherapy for bowel
`obstruction are disappointing. In a report by Abu-Rustum et
`al. (4), a response was observed in only 1 of 18 patients
`treated with chemotherapy for bowel obstruction. Radiation
`therapy seems to compare favorably with second- or third-
`line chemotherapy, with symptom response rates between
`63% and 79% with a median duration between 4 and 9
`months (5–7). Given the pattern of spread of ovarian cancer,
`whole abdominopelvic radiation therapy (WAPRT) could
`be the radiation technique of choice, eventually with a boost
`to sites of gross tumor.
`Maximum tolerated dose levels to the whole kidney are
`often set to 20 Gy or less. When the kidneys are blocked
`from radiation, conventional techniques result in underdos-
`age of the peritoneal regions in the blocked areas.
`We investigated the potential of intensity-modulated ra-
`diation therapy (IMRT) to spare kidneys and liver. For
`reasons described in the discussion section, we selected
`intensity-modulated arc therapy (IMAT) (8) as the most
`appropriate IMRT technique for WAPRT.
`This article describes the results of the translational re-
`search that was performed to bring IMAT into the clinic for
`WAPRT. We report on treatment planning and delivery for
`the first 5 patients treated with IMAT-WAPRT. The clinical
`results of the Phase I study will be reported elsewhere.
`
`METHODS AND MATERIALS
`
`Delineation of target volumes and OARs
`Between November 2001 and October 2002, 5 patients
`with a relapse of a histologically proven adenocarcinoma of
`the ovary were treated by IMAT. All patients signed in-
`formed consent for IMAT to the whole abdomen. A plan-
`ning computed tomography (CT) scan was performed with
`the patient lying supine, with both arms resting under the
`head. The upper border of the scanned volume was located
`10 cm cranial to the diaphragm, whereas the lower border
`was defined as 10 cm caudal to the obturator foramina.
`Several 0.5-cm-thick sequential CT slices were acquired
`without contrast enhancement. Clinical
`target volume
`(CTV) was defined as the total peritoneal cavity, with the
`inclusion of iliac and para-aortic lymph node regions. A 0.5-
`cm rim of liver, adjacent to the peritoneum, was also in-
`cluded into the CTV. The upper CTV boundary was defined
`by the highest CT slice on which the diaphragmatic dome
`was visible, and the most caudal drawing of the CTV was on
`the level of the bottom of the obturator foramina. A plan-
`ning target volume (PTV) was made by a 3D expansion of
`the CTV with a margin of 0.5 cm in all directions. The
`kidneys and the liver (hereafter called OARs: organs at risk)
`were drawn as visualized on CT, and the kidneys were
`expanded with 5 mm (kidney_exp_5mm), to account for
`setup inaccuracy and organ motion. PTV and OARs over-
`lap, which would result in a conflicting requirement from
`the optimization algorithm. A similar problem arises with
`
`those parts of the PTV in the build-up region. Both prob-
`lems were solved by defining subvolumes inside the PTV
`that are used as optimization volumes (PTV_optim).
`PTV_optim was nowhere closer than 0.8 cm to the skin
`surface or 0.5 cm to the expanded kidneys. A surrounding
`structure was made by subtracting the PTV from the total
`scanned volume of the patient. This structure (sur_0cm) was
`used to avoid hot spots outside the PTV. A general descrip-
`tion of the use of PTV subvolumes and surrounding struc-
`tures for IMRT plan optimization is found elsewhere (9).
`
`IMAT planning procedure
`Generation of machine states by ABST. An anatomy-
`based segmentation tool (ABST) was developed at our
`institution to create segments for step-and-shoot IMRT (10).
`For IMAT planning, ABST is used to create an initial set of
`segments, which we call machine states. A machine state is
`described by a set of machine parameters that uniquely
`define the beam incidence, aperture, and photon beam qual-
`ity. After definition of the isocenter location and with the
`collimator, table top and isocenter rotations at 0°, ABST
`generated machine states per 8° of gantry rotation. Restric-
`tion of the range of gantry angles was needed to avoid
`beams traversing metal components of the couch before
`entering the patient. The couch on the linear accelerator
`(Elekta, Crawley, UK) used for the IMAT treatments has
`two C-arms, which can be positioned at one of the 30°
`discrete angles. The largest range of possible gantry angles
`was obtained by setting the arms at 120°, measured from
`their lateral position (Fig. 1). For Patients 2, 3, and 5, a class
`solution was used, implying that the initial machine states
`were generated using a fixed set of parameters, including
`start-and-stop gantry angles, widths of the segments, and
`conformal avoidance structure. The arcs used by the class
`solution are shown in Fig. 1. For Patient 4, where the
`craniocaudal extent of the PTV was 42 cm, a second iso-
`center was defined 12 cm from the first isocenter in caudal
`direction. Only a longitudinal table shift is required to
`perform the transition between the two isocenters. An ad-
`ditional “pelvic arc” was made around a structure called
`PTV_pelvis, with the L5-S1 intervertebral space as the
`upper border. The collimator was rotated by 90° to have the
`leaf movements in the craniocaudal direction and allow for
`feathering in the junction region (11).
`For each gantry angle, ABST generates multiple machine
`states that differ only by apertures of the multileaf collima-
`tor (MLC). Each beam’s-eye view (BEV) projection of a
`MLC aperture covers a part of the PTV at one side of the
`anatomic structure that is to be avoided. A margin of 0.8 cm
`around the PTV is used to account for penumbra. For each
`gantry angle, the machine states differ from each other by
`the degree of coverage of the BEV projection of the PTV.
`For a detailed description of ABST, refer to De Gersem et
`al. (10). These machine states are useful to create intensity
`levels that increase with decreasing distance to the anatomic
`structure that is to be spared. It was shown by Brahme and
`others (12, 13) that such intensity profiles are useful to
`
`Page 2 of 14
`
`

`
`IMAT for whole abdominopelvic radiotherapy ● W. DUTHOY et al.
`
`1021
`
`Fig. 1. Class solution for the IMAT plan. A transverse plane through the patient can be appreciated, with the PTV, the
`left kidney (LK) and the right kidney (RK). The arcs are depicted by circle segments. Machine states for one arc (S0R
`LK) are shown every 16°, from ⫺128° to 0°. The thick gray lines represent the jaw position, each small white bar stands
`for one leaf. Abbreviations: S0R LK: Arc composed by the machine states covering the total BEV projection of the PTV
`passing the right side of the LK. S0L RK: Arc composed by the machine states covering the total BEV projection of
`the PTV passing the left side of the RK. S1R LK: Arc composed by the machine states covering a 3 cmwide area of
`the PTV at the right side of the LK. S1L RK: Arc composed by the machine states covering a 3 cmwide area of the
`PTV at the left side of the RK. The metal C-arms are shown as the gray squares, respectively in their most lateral
`position (asterisk) for the sliding window, and in their 120° position (⫹ sign). The large dashed arrow represents the
`beam direction for the delivery of the sliding window (SW).
`
`create homogeneous dose distributions to a concave PTV
`that conformally avoid anatomic structures at risk. The
`machine states were stratified in classes, the first class
`consisting of machine states with the largest area of MLC
`aperture, the second class consisting of machine states with
`the second largest area of MLC aperture, and so on. Because
`of the algorithm inside ABST that creates MLC apertures
`avoiding anatomic structures with—in this case—smooth
`surfaces, MLC apertures, which belong to the same class of
`machine states, do not differ much from one gantry angle to
`the next, their angular separation being only 8°. Hence, the
`leaf travel required when moving from one gantry angle to
`the next is small for machine states of the same class, which
`is preferable for dynamic transitions as in IMAT.
`For Patients 2–5, an additional posterior “sliding win-
`dow” intensity-modulated beam with a 90° collimator rota-
`
`tion was used. For this beam, the table bars were put on their
`most lateral position. Because of the position of these metal
`bars, and the impossibility to prescribe arcs traveling over
`the 180° gantry point, the range for posterior arcs is very
`limited. Therefore, a sliding window intensity-modulated
`beam with static gantry angle was preferred to boost the
`most posterior region of the PTV. The control points for this
`sliding window were made manually. For Patient 4, the
`most caudal isocenter was selected for this sliding window
`beam.
`Creation and optimization of control points. The machine
`instruction file to deliver arc therapy with dynamic MLC
`consists of a sequence of control points (see Fig. 1). A
`control point is defined as a machine state plus a monitor
`unit count (MUC) value. Delivery of a sequence of control
`points implies that the prescribed machine state has to be
`
`Page 3 of 14
`
`

`
`1022
`
`I. J. Radiation Oncology ● Biology ● Physics
`
`Volume 57, Number 4, 2003
`
`Fig. 2. Picture representing the angular delivery rate. One “virtual”
`arc (dashed line) as well as three deliverable arcs (solid lines) are
`shown. The “virtual” arc is not deliverable on the Elekta linac. The
`deliverable arcs approximate the optimized virtual arc. The final
`angular delivery rate (W) of the deliverable arcs is optimized as
`described in the text.
`
`reached at the MUC value for each control point. The
`transition from a control point to the next is slaved by the
`monitor unit (MU) counter; each parameter (leaf positions,
`jaw positions, and gantry angle) that changes between two
`control points is linearly interpolated as function of the
`MUC value. The beam is paused if the control software
`detects that a machine parameter is outside tolerance to the
`(linearly interpolated) position prescribed by the machine
`state. Control point optimization involves the machine
`states—and more precisely, the leaf positions—as well as
`the MUC values and is done by a segment outline and
`weight adapting tool (SOWAT) (14), modified for IMAT
`purposes (SOWAT-IMAT). The main difference between
`SOWAT and SOWAT-IMAT resides in the maximal leaf
`velocity (MLS) constraints. After the control point genera-
`tion and after each leaf position optimization cycle, a leaf
`velocity constrainer (LVC) adapts leaf positions of all con-
`trol points to obey maximum leaf speeds, minimum dis-
`tances to opposed and diagonally opposed leaves, and max-
`imum leaf position extends (method unpublished). MUCs
`are optimized for each step, a step being defined as the
`transition from one control point to the next. The objective
`function on which the optimization is done is a biophysical
`model and has been described and discussed elsewhere (15,
`16).
`Transformation to deliverable arcs. As a result of SO-
`WAT-IMAT, n machine states and n-1 weights are ob-
`tained. These weights are numbers of monitor units that
`have to be delivered while the machine moves from one
`control point to the next. Apart from leaf and jaw travel,
`such motion involves a 8° gantry rotation. Because the
`Elekta SL-series of linear accelerators was designed to
`deliver arcs with a gantry rotation speed directly propor-
`tional to the dose rate, the number of monitor units delivered
`per degree (angular delivery rate) must remain constant over
`the whole arc. This condition is not secured by SOWAT-
`
`Fig. 3. Adapted Rando phantom as used for the gel dosimetry. At
`each side of the Barex cast, three Rando slices were added to
`obtain full scatter conditions. Tape is attached to the phantom to
`draw laser lines in transverse, sagittal, and coronal reference
`planes. Seven markers are attached to the phantom on the laser
`lines to facilitate positioning. The transverse plane indicated in the
`middle of the barex phantom is 13.5 cm cranial to the treatment
`isocenter. A barex screw is used to close the phantom at the place
`where the gel was inserted, visible left to the sagittal plane.
`
`IMAT. In fact, the (requested) angular delivery rate may be
`different for each 8°-sector of gantry rotation. This problem
`is solved by splitting each arc, which features a variable
`angular delivery rate into multiple overlapping arcs each
`with a constant angular delivery rate (Fig. 2). This proce-
`dure also provides the start-and-stop angles of the delivered
`arcs. The plan is finalized by a SOWAT-IMAT optimization
`cycle, which involves optimization of leaf positions and the
`angular delivery rate (equal for all sectors of the arc to keep
`the angular delivery rate constant within each arc).
`Dose prescription and computation. The prescribed dose
`was 33 Gy (median dose in the PTV), given in 22 fractions.
`Except for the first patient, IMAT plans were accepted using
`the following clinical criteria: less than 5% of the PTV
`volume was allowed to receive more than 107% of the
`prescribed dose, and more than 95% of the volume of
`PTV_optim had to receive more than 90% of the prescribed
`dose; less than 5% and 20% of the kidneys_exp_5mm
`should receive more than 30 Gy and 25 Gy, respectively,
`whereas the median dose had to be lower than 18 Gy; the
`median liver dose was constrained to 30 Gy. After clinical
`
`Page 4 of 14
`
`

`
`IMAT for whole abdominopelvic radiotherapy ● W. DUTHOY et al.
`
`1023
`
`Table 1. Details about the intensity-modulated arc therapy treatments
`
`Isocenters
`
`Arcs
`
`Control points
`
`Monitor units
`
`Delivery time (min)
`
`Patient 1
`Patient 2
`Patient 3
`Patient 4
`Patient 5
`
`1
`1
`1
`2
`1
`
`11
`7 ⫹ 1 SW
`7 ⫹ 1 SW
`7 ⫹ 1 SW
`6 ⫹ 1 SW
`
`73
`110
`78
`99
`148
`
`568
`387
`378
`528
`359
`
`⬍20
`11.6 ⫾ 2.5
`15.3 ⫾ 3.6
`18.1 ⫾ 3.6
`13.5 ⫾ 1.3
`
`Abbreviation: SW ⫽ sliding window.
`For the first patient, no systematic measurements were done concerning the delivery time. The monitor units are for
`one fraction of 150 cGy.
`
`constraints were met by the optimization procedure, a final
`dose computation was performed for 18 MV with the col-
`lapsed cone convolution/superposition algorithm from Pin-
`nacle (Philips Medical Systems, Best, The Netherlands). A
`final optimization of the MUs of all arcs was done using the
`results of this dose computation. The value of a MU is such
`that 100 MUs correspond to 1 Gy at reference depth (5 cm
`for 6 MV and 10 cm for 18 MV) for a 10 ⫻ 10 cm field and
`a source-detector distance of 100 cm.
`IMAT treatment delivery. For each arc, a prescription file
`containing the sequence of control points and related mon-
`itor units is generated and networked to an SLiPlus 18-MV
`linear accelerator (Elekta). The IMAT treatment is delivered
`in local service mode using prototype dynamic control
`software (Elekta), operating as described previously (17).
`Delivery of dynamic prescriptions is not possible in clinical
`mode on the Elekta linear accelerators. The local service
`mode is operated in the same interlock class as in clinical
`mode. Therefore, tolerances used by the linac’s control
`system are the same as for clinical mode.
`
`Conventional plans
`For each patient, two different conventional (CONV)
`plans were made. The first plan was the widely used “AP/
`PA” technique, using an anterior and a posterior field
`(CONV2). A second plan used four beams (CONV4): an-
`terior, posterior, and two lateral fields. The field margins
`were drawn with a 1-cm margin around the PTV in all
`directions. Kidney blocks covered the BEV projection of
`the kidneys with a margin of 0.5 cm. For the CONV2 plans,
`the posterior field was duplicated in two segments, an open
`segment and a segment where kidney blocks were inserted.
`Respectively, 6-MV and 18-MV photons were used for the
`anterior and posterior field. For the CONV4 plans, all four
`fields were duplicated in two segments each, an open seg-
`ment and a segment with kidney blocks. Here, 18-MV
`photons were used for all fields.
`Optimization of the relative segment weights was done
`by the planner to reach a median dose to the expanded
`kidneys between 18 and 20 Gy. Median dose to the liver
`was constrained to 30 Gy. Dose computation was done with
`the same collapsed cone convolution/superposition algo-
`rithm.
`
`Treatment evaluation
`Delivery time, defined as the time between the start of the
`first arc and the end of the last arc or sliding window, was
`measured for Patients 2–5. Additionally, the setup time was
`measured from the entrance of the treatment room by the
`patient to the start of the first arc. This includes the time
`necessary to acquire portal images and correction of the
`patient position.
`Comparison of dose distributions obtained with the
`CONV2, CONV4, and the IMAT plan was done after nor-
`malizing the median dose of the PTV to 33 Gy. To evaluate
`the dose homogeneity in the target volumes, an inhomoge-
`neity factor U95/5 was defined as the difference between the
`95th percentile dose (D95) and the 5th percentile dose (D5),
`divided by the median dose (Dmed). We preferred to use the
`D95 and the D5 above the maximum and minimum dose,
`because an underdosage was allowed in the region close to
`the kidneys. Other endpoints for the target volumes were the
`first percentile dose (as a surrogate for minimum dose), the
`99th percentile dose, and the ratio of volume of the target
`structure receiving more than 95% of the prescribed dose
`(V95) over the total volume. For the parallel-element organs
`kidney and liver, the Dmed was used. The dose–volume
`histograms (DVHs) were reconstructed for the 5 patients by
`calculating the mean dose and the standard error of the
`mean at every 5% volume level. The paired Student t test
`was used. All tests were two-tailed and p ⬍ 0.05 was
`considered as statistically significant.
`
`Dosimetric verification of the IMAT treatment
`Monomer/polymer gel dosimetry was used for 3D dose
`verification of the whole IMAT procedure. A deoxygenated
`hydrogel infused with acrylic monomers forms the basis for
`this dosimetric technique (18). Highly reactive radicals,
`formed by radiolysis during irradiation of the gel, initiate a
`polymerization reaction. The amount of polymer formed is
`related to the absorbed dose. Formation of polymer clusters
`in the water-equivalent gel increases the local spin-spin
`relaxation rate (R2), a typical magnetic resonance (MR)
`contrast parameter. Therefore, MR imaging (MRI) can be
`used to visualize the amount of polymer formed and sub-
`sequently the dose distribution in the gel. With monomer/
`polymer gel dosimetry, it is possible to obtain absorbed dose
`information in 3D with high spatial accuracy (19). For a
`
`Page 5 of 14
`
`

`
`1024
`
`I. J. Radiation Oncology ● Biology ● Physics
`
`Volume 57, Number 4, 2003
`
`Fig. 4. (a) Intensity profiles (rescaled) for the delivered plan of Patient 2. The CT slice through the isocentric plane is
`shown, with the dose distribution in Gy. The PTV (dotted line) extends outside the scanned volume (asterisks). (a)
`Intensity profiles, generated at a range of gantry angles from ⫺128° to 128°, are plotted around the CT slice. (b) Intensity
`profiles for the gantry at ⫺104°, 104° and for the posterior beam, delivered as a sliding window (180°).
`
`Page 6 of 14
`
`

`
`IMAT for whole abdominopelvic radiotherapy ● W. DUTHOY et al.
`
`1025
`
`Table 2. Summary of the DVH data showing averages ⫾ standard deviations
`
`IMAT
`
`CONV2
`
`p value
`
`CONV4
`
`p value
`
`PTV
`V95 (%)
`V90 (%)
`V107 (%)
`D1 (Gy)
`D99 (Gy)
`U95/5 (%)
`PTV_optim
`D1 (Gy)
`U95/5 (%)
`Left kidney (expanded)
`Dmed (Gy)
`D95 (Gy)
`Right kidney (expanded)
`Dmed (Gy)
`D95 (Gy)
`Liver
`Dmed (Gy)
`
`82.2 ⫾ 6.5
`89.9 ⫾ 5.7
`2.7 ⫾ 3.4
`19.0 ⫾ 10.8
`35.8 ⫾ 0.9
`28.1 ⫾ 15.8
`
`27.0 ⫾ 2.8
`15.1 ⫾ 5.8
`
`16.1 ⫾ 3.6
`28.2 ⫾ 1.3
`
`13.6 ⫾ 3.9
`26.0 ⫾ 2.9
`
`24.4 ⫾ 6.3
`
`76.8 ⫾ 5.2
`80.1 ⫾ 4.6
`6.4 ⫾ 3.3
`16.4 ⫾ 9.1
`36.1 ⫾ 0.6
`42.3 ⫾ 7.9
`
`21.7 ⫾ 0.9
`34.9 ⫾ 2.5
`
`19.9 ⫾ 0.6
`22.6 ⫾ 0.8
`
`19.3 ⫾ 1.5
`22.8 ⫾ 1.0
`
`22.8 ⫾ 10.1
`
`0.11
`0.01
`0.18
`0.09
`0.53
`0.03
`
`0.02
`⬍0.01
`
`0.11
`⬍0.01
`
`0.02
`0.02
`
`0.44
`
`73.6 ⫾ 5.9
`82.5 ⫾ 6.1
`7.2 ⫾ 4.41
`16.6 ⫾ 9.3
`36.5 ⫾ 0.9
`34.4 ⫾ 10.9
`
`24.5 ⫾ 1.2
`24.9 ⫾ 4.1
`
`19.4 ⫾ 0.4
`22.2 ⫾ 0.4
`
`18.6 ⫾ 1.1
`22.2 ⫾ 1.0
`
`29.2 ⫾ 2.0
`
`0.01
`0.01
`0.05
`0.10
`0.02
`0.10
`
`0.12
`0.01
`
`0.11
`⬍0.01
`
`0.02
`0.08
`
`0.09
`
`Abbreviations: IMAT ⫽ intensity-modulated arc therapy; CONV2 ⫽ conventional plan with an anteroposterior and a
`posteroanterior field; CONV4 ⫽ conventional plan with four-field technique; PTV ⫽ planning target volume; PTV_optim ⫽
`PTV without build-up region of 0.8 cm and with the exclusion of the expanded kidneys with an extra margin of 5 mm; V95,
`V90, and V107: the partial volume (percent) receiving more than 95%, 90%, and 107% of the prescribed dose, respectively; D99
`共D95–D5兲
`⫽ inhomogeneity factor, defined as
`⫽ Dose given to 99% and 1% of the volume, respectively; U95/5
`and D1
`, with
`Dmed
`D95 the 95th percentile dose; D5 the 5th percentile dose, and Dmed the median dose.
`
`more detailed review on this subject, the reader is referred
`to De Deene et al. (20).
`A Barex (Cifra, Chateau Thierry, France) cast was
`vacuum molded on the abdominal region of the Rando
`phantom (Alderson Research Laboratories, Stamford,
`CT). At our laboratory, the maximum amount of mono-
`mer/polymer gel that can be produced in one batch is 10
`L. Hence, the entire volume irradiated with IMAT can
`not be verified by one gel dosimetry experiment. We
`chose to limit
`the phantom geometry to that part of
`RANDO containing the (dosimetrically most interesting)
`region around the kidneys. Supports on the cranial and
`caudal side, marker lines and placement of fiduciary
`markers (Medtronic, Louisville, KY) on the surface fa-
`cilitated a reproducible positioning of the gel phantom
`during CT scanning, IMAT delivery, and MRI (Fig. 3).
`Three supplemental Rando slices were placed alongside
`the phantom on the cranial and caudal side during CT
`scanning and treatment delivery to ascertain full scatter
`conditions in the gel upon irradiation. Spiral CT scans
`(Siemens Somatom Plus 4, Erlangen, Germany) of the
`gel-filled phantom were transferred to the planning sys-
`tem. Volumes of interest (kidneys, liver, and PTV) from
`the first patient were transferred to the CT data set. For
`this setup, an IMAT plan was made by using the methods
`described previously. This resulted in a plan with six arcs
`and one sliding window. To cover the maximum response
`range of the gel while avoiding gradient dependent non-
`linearities in dose response (21), the monitor units were
`multiplied with a factor of five, thus giving a median dose
`
`of 7.5 Gy to the PTV. For calibration purposes, gel-filled
`test tubes were irradiated to known doses (0 –10 Gy,
`every 1 Gy) to establish a dose-R2 relationship. The gel
`dosimeter and test tubes were scanned together in the
`body coil of a 1T MR system (Expert, Siemens, Erlangen,
`Germany). A 26 spin-echo sequence was applied with a
`Carr-Purcell Meiboom-Gill (CPMG) RF pulse encoding
`scheme and equidistant echo spacing (TE ⫽ 40 –1080
`ms). The phantom was scanned in 33 adjacent transverse
`slices each with a slice thickness of 5 mm. A field of view
`of 320 mm and image resolution of 128 ⫻ 128 resulted in
`an in-plane resolution of 2.5 mm. Ideally, a homogeneous
`unirradiated gel phantom should produce identical R2
`readings throughout
`the entire volume when scanned
`with MRI. A homogeneity study was done on the IMAT
`phantom filled with a blank gel and the MRI protocol was
`adjusted to minimize R2 variations related to temperature
`deviations and radiofrequent inhomogeneities.
`A linear regression was used to describe the dose-R2
`relationship. The gel-measured dose grid was transferred to
`the planning system and scaled to the prescription dose (33
`Gy), which allowed the comparison of measured and com-
`puted DVHs of the different structures inside the abdomi-
`nopelvic phantom,
`truncated to the volume of the gel-
`phantom. Low’s ␥-index (22) was calculated in 3D (dose
`difference criterion ⫽ 7.5%, distance to agreement ⫽ 5
`mm) as a guide to pinpoint significant deviations between
`the computed and measured dose matrix. A ␥-index above
`1 indicates that the specified tolerances are not met.
`
`Page 7 of 14
`
`

`
`1026
`
`I. J. Radiation Oncology ● Biology ● Physics
`
`Volume 57, Number 4, 2003
`
`Fig. 5. DVHs compiled from the data of the 5 patients. Solid lines and bold dots represent the IMAT plans (mean ⫾
`standard error of the mean). Dashed lines and circles represent the 2D plans. (a-c) DVHs of the IMAT and CONV2
`plans. (d-f) DVHs of the IMAT and CONV4 plans. (a) ⫹ (d) DVHs of PTV and expanded left kidney. (b) ⫹ (e) DVHs
`for PTV_whbu and expanded right kidney. (c) ⫹ (f) DVHs of PTV_optim and liver.
`
`RESULTS
`
`The median volume of the PTV in the 5 patients was
`8306 cc (range 5717–9054 cc), and the median of the
`craniocaudal length which had to be covered was 36 cm.
`Details on the treatment plans and delivery times are shown
`in Table 1. Mean delivery time for Patients 2–5 was 13.8
`min (range 9.5–24.5 min). Less than 30% of the given
`fractions had a delivery time exceeding 15 min. Of these,
`50% were seen in Patient 4, who had two isocenters, ne-
`cessitating entrance of the treatment room to perform a
`cranial shift of the patient. For 1 patient (chronologically the
`last), the setup time was measured over all the fractions, and
`showed a mean of 8 ⫾ 2.9 min. Though actively asked, no
`patient complained about the rotating gantry. As an exam-
`ple, the obtained intensity profiles for Patient 5 are shown in
`Fig. 4.
`
`DVH analysis and dose distributions
`The DVH data for the 5 patients are summarized in
`Table 2 and graphically displayed in Fig. 5. For both the
`
`CONV4 and IMAT plans, there is a large variation in
`minimal dose (represented by D1) in the PTV (range
`0.2–26.1 Gy), resulting from the PTV extending outside
`the skin in 1 patient. When considering the PTV without
`a build-up region of 8 mm (PTV_whbu), the very low
`doses that are the result of the ICRU PTV definition
`rather than of the planning technique are eliminated
`(range of D1 19.5–26.9 Gy). The homogeneity in the
`PTV_whbu is better for the IMAT plan than for the
`CONV4 plan (U95/5 is 20% and 32%, respectively; p ⫽
`0.01). Because of the strong dose constraint to the ex-
`panded kidneys and the possibility to generate concave
`dose distributions, we expected an underdosage in the
`PTV in the region around the kidneys. For the PTV_op-
`tim, the mean (and the standard deviation) of the V95 was
`78.4% (⫾ 2.2%) for the CONV4 plan, and 88.9% (⫾
`5.1%) for the IMAT plan (p ⫽ 0.01). For the V90, these
`values were 87.7% (⫾ 3.0%) and 95.8% (⫾ 3.5%) for the
`CONV4 and the IMAT plan, respectively (p ⫽ 0.02). The
`comparison between the IMAT plan and the CONV2 plan
`
`Page 8 of 14
`
`

`
`IMAT for whole abdominopelvic radiotherapy ● W. DUTHOY et al.
`
`1027
`
`Fig. 6. Dose distributions for Patient 1, showing the delivered IMAT plan in a transverse (a) and sagittal (d) plane, and
`the conventional comparison plans in (b) ⫹ (e) for the CONV2 and (c) ⫹ (f) for the CONV4 plan. Isodose values are
`in Gy. The PTV is delineated with a dotted line. The kidneys (LK: left kidney and RK: right kidney) are delineated by
`dashed lines, while the liver (L) is circled by a dashed-dotted line. The dotted straight line in (a) and (d) indicates the
`transection planes in (d-f) and (a-c), respectively.
`
`(see Table 1) shows a significant increase in homogeneity
`(expressed by U95/5) for the PTV and for PTV_optim by
`the IMAT plans.
`The median dose to the kidneys was lower for the IMAT
`plan in all patients when compared with both CONV plans.
`The maximal doses to the kidneys were significantly higher
`for the IMAT plan than for the CONV plans, as can be seen
`in Fig. 5. For the liver, no significant differences were found
`between the IMAT and the CONV plans. The higher ho-
`mogeneity by