IOP PUBLISHING
PHYSICS IN MEDICINE AND BIOLOGY
Phys. Med. Biol. 52 (2007) 4541–4552
doi:10.1088/0031-9155/52/15/012
A two isocenter IMRT technique with a controlled junction dose for long volume targets G G Zeng1, R K Heaton1,2, C N Catton1,2, P W Chung1,2, B O’Sullivan1,2, M Lau1, A Parent1 and D A Jaffray1,2 1 Radiation Medicine Program, Princess Margaret Hospital/University Health Network, Toronto M5G 2M9, Canada 2 Department of Radiation Oncology, University of Toronto, Toronto M5G 2M9, Canada
E-mail:
[email protected]
Received 25 January 2007, in final form 31 May 2007 Published 3 July 2007 Online at stacks.iop.org/PMB/52/4541 Abstract Most IMRT techniques have been designed to treat targets smaller than the field size of conventional linac accelerators. In order to overcome the field size restrictions in applying IMRT, we developed a two isocenter IMRT technique to treat long volume targets. The technique exploits an extended dose gradient throughout a junction region of 4–6 cm to minimize the impact of field match errors on a junction dose and manipulates the inverse planning and IMRT segments to fill in the dose gradient and achieve dose uniformity. Techniques for abutting both conventional fields with IMRT (‘Static + IMRT’) and IMRT fields (‘IMRT + IMRT’) using two separate isocenters have been developed. Five long volume sarcoma cases have been planned in Pinnacle (Philips, Madison, USA) using Elekta Synergy and Varian 2100EX linacs; two of the cases were clinically treated with this technique. Advantages were demonstrated with well-controlled junction target uniformity and tolerance to setup uncertainties. The junction target dose heterogeneity was controlled at a level of ±5%; for 3 mm setup errors at the field edges, the junction target dose changed less than 5% and the dose sparing to organs at risk (OARs) was maintained. Film measurements confirmed the treatment planning results. (Some figures in this article are in colour only in the electronic version)
1. Introduction Traditionally, long volume targets are treated either by extended source–surface-distance (SSD) beams or abutting fields. The daily setup for an extended SSD treatment is complicated for oblique beams, requiring couch motions in two directions for each beam. The inherent problems within the adjacent abutting fields are the large dose variations due to beam 0031-9155/07/154541+12$30.00 © 2007 IOP Publishing Ltd Printed in the UK
4541
4542
G G Zeng et al
divergence and field match errors (Khan 1994). Neither conventional technique offers dose sparing to organs at risk (OARs). Most IMRT techniques have been designed to treat targets smaller than the field size of the MLC (Ling et al 1996, Kestin et al 2000, Hunt et al 2001, Hong et al 2004). For targets larger than the field size, Chan et al (2001) presented the feasibility of using five extended SSD beams in IMRT, but did not address the clinical issues of setup complexity. The majority of published techniques has been focused on applying adjacent fields. For a single isocenter subfields junction, associated with Varian 2100EX serial’s carriage MLC design (Varian Associates, Palo Alto, USA), Wu et al (2000) presented a dynamic ‘feathering’ technique to split large fields into smaller fields. The subfields overlapped each other by a small amount and the intensity gradually decreased for one subfield and increased for the other. This technique solved the dosimetric problems associated with beam divergence and reduced the dose uncertainties due to field match errors. Malhotra et al (2005) expanded this method to two separate isocenters to treat tumors larger in width than that possible for multiple carriage groups at a single isocenter. To implement two isocenter dynamic IMRT in the cephalad-caudal direction to treat long volume targets, Hong et al (2002) applied Wu et al’s technique by rotating the collimator 90◦ . However, all the methods published so far were for dynamic IMRT. They used custom software developed specifically for the technique to derive a procession of the MLC leaves which created a smooth dose transition at the junction. The custom software may not be robust for clinical routines. To address the issue of two isocenter IMRT for long volume targets, we have developed a new technique which enables junctioning perpendicular to MLC motion and requires no custom software for a commercial Pinnacle treatment planning system (Philips, Madison, USA). This technique utilizes dose ‘feathering’, where a dose gradient is established over an extended distance in abutting fields, with complementary gradients generated to obtain a uniform dose in the target. The technique is illustrated in this study for five different patient plans. The five treatment plans all exhibit the advantages of IMRT in dose conformity and OARs sparing, with controlled dose heterogeneities in the junction target and high tolerance to potential setup errors in field match. 2. Methods and materials 2.1. Technique description The techniques rely on creating an extended dose gradient at one field edge to reduce the junction dose sensitivity to field match uncertainties and exploit IMRT segments and inverse planning to compensate for the missing dose and achieve target dose uniformity within the junction. The target volume, typically the planning target volume (PTV), is divided into three non-overlapping regions: a superior, an inferior and a junction region. Two techniques have been developed to exploit the same principles: a technique for using two sets of abutting IMRT fields, denoted as ‘IMRT + IMRT’ hereafter, and a technique for abutting a conventional static field plan with a set of IMRT fields, denoted as ‘Static + IMRT’. For the ‘IMRT + IMRT’ technique, the dose gradient in the junction region was induced by subdividing the junction target into variable dose volumes to set graduated dose levels for optimization. Four dose volumes, each 1.0 to 1.5 cm in the longitudinal direction, were assigned stepped dose objectives from 80% to 20% of the prescribed dose, as displayed in figure 1(a). The first set of IMRT fields was added to cover the junction region and either the superior or inferior region. An IMRT optimization was performed for these targets, with the junction dose objectives establishing a dose gradient in the junction. Then, the first IMRT
A two isocenter IMRT technique with a controlled junction dose for long volume targets
4543
80% 60% 40% 20%
(a) Variable dose volumes in junction
(b) Dose gradient created by segments
Figure 1. (a) The target in the junction was subdivided into variable dose volumes to set graduated dose levels for the first IMRT optimization in the ‘IMRT + IMRT’ technique. (b) The fluency maps of the POP beams show the five equally weighed stepped beam segments to create a physical dose gradient in the ‘Static + IMRT’ technique.
fields were frozen and the second set of IMRT fields was placed to cover the junction region as well as the portion of the target excluded from the first optimization. With the dose contribution from the first set of IMRT fields present, the second set of IMRT fields was optimized with the dose objectives of covering targets and structures, and thus filled in the junction region with a complementary dose gradient to that established by the first set of IMRT fields. For the ‘Static + IMRT’ technique, segmented parallel-opposed pair (POP) beams were applied to create a physical dose gradient. We employed five segments for each beam with steps of 1 cm perpendicular to MLC motion, as shown in figure 1(b). The IMRT fields were then set for optimization in one section of the target and dose compensation in the junction. 2.2. Patient data and equipment The anatomic data from five patients were used in this study. Three lower extremity soft tissue sarcoma cases were planned retrospectively to develop and verify the technique. Two preoperative cases were planned for clinical treatment: one was a lower extremity sarcoma case with tumor adjacent to the femur and the other was a retroperitoneal sarcoma case with kidney, liver and spinal cord adjacent to the tumor. Neither patient had tumor involvement of critical structures. CT scans were acquired on a GE CT simulator (GE Healthcare, Milwaukee, USA) with the slice thickness of 3 mm for extremity and on a Philips big bore (Philips Medical System, USA) at 4 mm slice spacing for the retroperitoneal patients. Patients were immobilized with a knee rest and the thermoplastic foot fixation for lower extremity patients and a body
4544
G G Zeng et al
PTV
Bone
GTV
CTV
Figure 2. Illustration of delineation of gross tumor volume (GTV), clinical target volume (CTV) and planning target volume (PTV) in this paper. The PTV was extended 0.5 cm into the bone where the GTV approached bone and excluded 0.5 cm subcutaneous tissue near the skin surface.
fix vacuum bag for retroperitoneal patients. The Pinnacle V7.6 treatment planning system (Philips, Madison, USA) was used for target and OAR delineation and treatment planning. 2.3. Delineation of targets and structures The clinical target volume (CTV) was defined from an expansion of the gross tumor volume (GTV) with specified anisotropic margins and the planning target volume (PTV) was defined by a 0.5 cm expansion of the CTV in three dimensions. In situations where the GTV approached a critical structure closer than 0.5 cm, the PTV was extended up to 0.5 cm into the OAR. The PTV also excluded 0.5 cm of subcutaneous tissue to account for the build-up near the skin and to achieve PTV dose uniformity during optimization. Figure 2 provides an illustration of target and critical structure definition. 2.4. Selection of junction region and treatment isocenters The junction region was selected in areas where there was spatial separation of PTV and OARs, so that the ‘Static + IMRT’ can be applied, or where the OARs can be maximally avoided to plan the ‘IMRT + IMRT’ treatment efficiently. The two isocenters were longitudinally separated by a distance which allowed the entire PTV to be covered by the geometrical bounds of the fields with a 2 cm margin. For image-guided setup using CBCT or portal imaging, one isocenter was positioned to allow the visualization of critical structures sensitive to setup variations. 2.5. Beam arrangement and IMRT optimization 6 MV photon beams were used for all IMRT fields, and 18 MV photon beams were applied to the ‘static’ pelvic fields in the clinical retroperitoneal case. The beams were arranged
A two isocenter IMRT technique with a controlled junction dose for long volume targets
4545
to maximize normal tissue avoidance. All IMRT optimization was performed using direct machine parameter optimization (DMPO) in Pinnacle 7.6c. Plans were optimized to cover 95% of PTV by at least 95% of the prescribed dose for typical prescriptions of 50 Gy. The dose to bone in lower extremity cases was constrained to spare a contiguous longitudinal strip of bone to 40 Gy while maintaining 95% PTV coverage and restricting the dose throughout the bone to no greater than the prescribed dose. The tolerances for OARs in the retroperitoneal case were set to constrain the mean dose for liver to less than 25 Gy, the mean dose for left kidney to less than 13 Gy and the maximal cord dose to less than 45 Gy for a total prescribed dose of 45 Gy. A step-and-shoot IMRT delivery method was selected during the DMPO calculation, with the total number of segments being restricted to an average of 8–10 segments per beam. In the first IMRT calculation for the ‘IMRT + IMRT’ technique, the dose constraints for sub-regions of junction PTV were assigned incremental dose ranges to establish a monotonically decreasing dose gradient across the junction. For example, the 80% subregion was assigned a maximum dose of 80% and a minimum dose of 60%, while the 60% sub-region was given a dose range of between 60% and 40%. The ‘Static + IMRT’ method established the initial gradient in the junction region through segmentation of each static field, and consequently did not require further subdivision of the junction PTV. 2.6. Experimental verification As fields are matched perpendicular to MLC motion in this technique, the tongue and groove effect of the leaves’ edge is of dosimetric concern. For the Elekta MLC, a previous study showed that this effect causes a significant dose deficiency (27% at the peak) at the overlaps between two leaves (Haryanto et al 2004), and therefore, accurate modeling of this effect is critical for the dose compensation in junction. Measurements were conducted using Kodak EDR2 films (Radiation Products Design, MN, USA) to verify Pinnacle’s calculations on dose match. Films were placed in a custom cylindrical Solid WaterTM (Gammex, TN, USA) IMRT QA phantom with the film oriented in the coronal plane. Treatment plans were transferred to the QA phantom and the dose to phantom was calculated. Film dose response was characterized from measurements of a 10 × 10 cm2 6 MV field at 1.5 cm depth in a flat solid water phantom. The calibration and verification films were scanned and read out by a Vidar Pro 16 scanner (Radiological Imaging Technology, CO, USA) and analyzed with RIT113 software. Measurements for two plans (plan 2 and plan 3 in table 1) were performed to verify both the dose uniformity across the junction and the tolerance of the junction region to setup variations. Films across the junction were acquired for the planned separation between the isocenters, as well as for isocenter separations 3 mm greater and smaller than required by the plan to simulate the field match errors. 3. Results 3.1. Junction target dose uniformity and junction OARs sparing Figure 3 shows two sample dose distributions for lower extremity treatments. Figure 3(a) illustrates a case planned with the ‘IMRT + IMRT’ technique (plan 3 in table 1) to a dose of 50 Gy, where dose sculpting around the bone was required due to the close proximity of the PTV throughout the treatment volume. The distribution in figure 3(b) shows a plan (plan 1 in table 1) where the ‘Static + IMRT’ technique was utilized. In the inferior section of the PTV, adequate bone sparing was achieved with simple POP beams, which were manually segmented
4546
G G Zeng et al (a) IMRT+IMRT
(b) Static+IMRT
50Gy
66Gy
47.5Gy
63Gy
45Gy
52.5Gy
Inferior PTV
40Gy
Inferior PTV50Gy
50Gy 47.5 Gy
Junction PTV50Gy Junction PTV Superior PTV50Gy
Superior PTV Superior PTV66Gy
Superior PTV50Gy
Figure 3. Examples of junction target dose distributions. (a) IMRT + IMRT (plan 3 in table 5) case with a prescription of 50 Gy. (b) Static + IMRT with two-level prescriptions: 66 Gy and 50 Gy. Table 1. Summary of five treatment plans: treatment sites, beam arrangement, junction technique, junction target dose heterogeneity, junction PTV dose shifts and OARs dose changes for ±3 mm setup errors. All the dosimetric values are given as percentages. Plan no
1
2
3
4
5
Site Junction techniques Energy Linac
Thigh + pelvic Static + IMRT 6 MV Elekta
Thigh + pelvic IMRT + IMRT 6 MV Elekta
Thigh + pelvic IMRT + IMRT 6 MV Elekta
Thigh + pelvic IMRT + IMRT 6 MV Elekta
Abdomen + pelvic Static + IMRT 18 MV + 6 MV Varian
Dose heterogeneity (%) (Dmax, Dmin) PTV dose shift + (%) (D95, Dmean, Dmax) PTV dose shift − (%) (D95, Dmean, Dmax) OAR dose change + (%) (D50, Dmax) OAR dose change − (%) (D50, Dmax)
−5, 5
−5, 7
−4, 4
−3, 7
−3, 6
3, 4, 4 −4, −4, −5
3, 5, 5 −6, −3, −3
2, 4, 4 −5, −4, −1
4, 4, 6 −5, −5, −3
3, 3, 4 −3, −3, −4
3, 2
4, 3
4, 6
3, 4
−3, −2
−4, −1
−4, −2
−4, −2
to generate a dose gradient. The IMRT fields applied to the superior portion were used to achieve differential prescription levels (50 Gy and 66 Gy), provide bone sparing and fill in the junction dose. In table 1, we have summarized the dose variations in the junction target for all plans in terms of maximal and minimal doses, expressed as a percentage of the prescription. In all cases, the junction dose variations were controlled at within ±5% of the prescribed dose, independent of the treatment linac type, beam energy and junctioning technique.
A two isocenter IMRT technique with a controlled junction dose for long volume targets -3 mm shift
Ideal setup
4547 +3 mm shift
110% 105% 100% 95% 90% 80%
(a) Representative axial dose distributions
PTV
Bone
(b) Corresponding DVHs
Figure 4. (a) Axis dose distributions in the junction for an ideal setup and for ±3 mm changes in isocenter separation. The color wash shows the PTV in the junction; (b) the corresponding DVHs for target and bone in the junction for the ideal setup (solid lines) and for the ±3 mm shifts (dotted lines).
The OAR bone dose sparing in the junction region is illustrated in figure 4(a) under ‘ideal setup’, which shows the dose distributions in an axial plane for an ‘IMRT + IMRT’ plan (plan 2 in table 1). The dose–volume histogram (DVH) for this plan was plotted in figure 4(b) as a solid blue line. 3.2. Tolerance to setup variations The impact of field match uncertainties on the dose distributions in the junction was investigated for a change of 3 mm in the isocenter separation. Representative axial distributions in the junction, shown in figure 4(a), illustrate the changes observed in plan 2, with the corresponding
4548
G G Zeng et al
Figure 5. Comparisons of dose profiles cross the junction between Pinnacle calculations and film measurements in a cylindrical QA phantom. The position of the junction region was from 2 cm to 7 cm.
DVHs shown in figure 4(b). In figure 4, the dose was 5% hotter when the isocenter separation was decreased by 3 mm and 5% colder when the separation was increased 3 mm. However, the spared dose to the femur was largely maintained under the shifts. Table 1 summarizes the relative dose changes for junction targets and OARs under the isocenter shifts for the five plans. For junction targets, the dose coverage of 95% volume (D95), mean dose (Dmean) and maximal dose (Dmax) were recorded, while for junction OARs, dose to 50% volume (D50) and Dmax are listed. In plan 1, the bone was excluded from the irradiated volume of POP beams, so the bone dose is not reported. In plan 5, the pelvic POP beams (PA/AP) passed through the spinal canal with no sparing intended in that region, while the superior spinal cord dose was constrained in the superior abdominal IMRT fields. The summary of dose changes in table 1 further demonstrates that the range of target dose shifts under ±3 mm setup errors was about ±5% and the OAR sparing dose has been maintained. 3.3. Comparison of measurements and Pinnacle calculations Figure 5 shows a comparison of dose profiles across the junction between film measurements and calculations in the QA phantom to verify junction dose calculations and variations under ±3 mm changes in isocenter separation on an Elekta Synergy treatment unit. The relative dose was normalized to a uniform dose region in the superior target. An agreement within 5% between measurement and calculation was found, consistent with the reported dosimetric accuracy of the EDR2 film (Childress et al 2005). These results confirm that the Pinnacle TPS can accurately model all important dosimetrical effects and the new technique is achievable in practice. 3.4. Clinical experiences In figure 6, we present the planning results for the first two clinical cases in our institution. The dosimetric results for these two plans are listed in table 1 under plans 4 and 5. In the first case, a limb sarcoma was treated to 50 Gy from pelvis to the knee with PTV spanning 45 cm and bone adjacent throughout the whole target. This was an ‘IMRT + IMRT’ plan with six beams being assigned to the superior part and four to the inferior part using a total of 90 segments and 1100 total monitor units. Patient positioning for the treatment was verified by daily cone beam CT imaging. In the second case, a retroperitoneal sarcoma in the abdominal-pelvic region was treated to a total dose of 45 Gy. Segmented PA/AP beams were used to treat the pelvic region and six IMRT beams were applied to treat the superior target and spare the spinal canal, liver and left kidney. Daily orthogonal EPID images were used to verify the patient treatment positioning.
A two isocenter IMRT technique with a controlled junction dose for long volume targets
4549
107% 100% 95% 90% PTV
(a) Limb STS treated to 50 Gy
PTV
105% 100% 95%
(b) Retroperitoneal sarcoma treated to 45 Gy
Figure 6. Summary of the clinical plans: target dose distributions in a sagittal plane and the DVHs. The color washes are the PTVs: (a) preoperative limb STS treated to 50 Gy; (b) retroperitoneal sarcoma treated to 45 Gy.
4. Discussion 4.1. Dose gradient Theoretically, as the extension of the dose gradient is increased, the tolerance to field match errors and the final dose uniformity can be improved. In practice, the selection of dose gradient should be justified by the clinically acceptable dose variations and the treatment
4550
G G Zeng et al
Figure 7. DVH comparisons of the junction PTV planned by the ‘IMRT + IMRT’ (solid line) and ‘Static + IMRT’ (dashed line) under the same bone sparing for case 3.
planning efficiency. Typical mechanical tolerances are on the order of ±1 mm for an MLC position, ±1 mm for couch positions and ±1 mm for isocenter move with gantry rotation (AAPM rpt-471994)3 , so the uncertainties of field edge placement may be as high as ±3 mm. With a 5 cm gradient dose extension over five steps, the dose falls essentially at a rate of 20% per cm, and so dose uncertainties should be controlled within ±6% for a 3 mm variance in isocenter separation. Both planning calculations and measurements in this study showed a controlled variation of ±5% and thus supported the above estimation. The field match errors are random and most probably smaller than 3 mm; the overall dose uncertainties in a treatment, therefore, are expected to be much lower than 5%. In implementing this technique, each institution should assess what is an acceptable dose variation for a given setup variation, taking into consideration both the capabilities and limitations of the patient immobilization system, image matching system and frequency of imaging. 4.2. ‘IMRT + IMRT’ or ‘Static + IMRT’ Both the ‘IMRT + IMRT’ and ‘Static + IMRT’ techniques have proven to be clinically useful. The ‘IMRT + IMRT’ technique provides maximal target conformance, and is the preferred choice when OARs are in close proximity to the PTV throughout most of the treatment volume and junction region. Figure 7 shows the DVH comparisons of the junction PTV planned by ‘IMRT + IMRT’ (solid line) and ‘Static + IMRT’ (dashed line) for case 3, a lower limb sarcoma treatment with bone sparing objectives. The junction dose regions used to establish the initial gradient in the ‘IMRT + IMRT’ technique generate a more isotropic dose distribution perpendicular to the dose gradient when compared to the segmented static field used in the ‘Static + IMRT’ technique. Consequently, the complementary IMRT distribution tends to generate a more uniform dose through the junction when matching to a multi-directional IMRT field as compared to the opposed pair of segmented static fields. 3
AAPM rpt-471994: AAPM code of practice for radiotherapy accelerators.
A two isocenter IMRT technique with a controlled junction dose for long volume targets
4551
Strategies for dealing with OARs within the junction region have been developed. OARs, other than the long bones, were excluded from the junction region in the cases presented here by a judicious selection of the junction position. Finer control of the dose around the long bone for ‘IMRT + IMRT’ cases was achieved by partitioning the section of the OAR within the junction and providing independent dose objectives during the initial IMRT optimization. The volumetric dose constraint of the OAR for the first IMRT optimization was set to half of the volumetric dose limits of the OAR. The subsequent IMRT optimization which fills in the established gradient and provides dose coverage to the remaining PTV volume should then use the full volumetric dose limits on the OAR to achieve the desired normal tissue sparing. It should be recognized that planning and delivery of essentially two IMRT plans could increase the workload and the treatment time. On the other hand, if an effective treatment can be achieved from an IMRT approach in one portion of the target and a simple set of conventional beams in the other portion, ‘Static + IMRT’ should be considered due to its planning and delivery efficiency. 4.3. Clinical implementation The implementation of this technique increases the workload in the areas of planning and treatment delivery. Compared to the standard one isocenter IMRT technique, additional time will be spent in appropriately dividing the PTV into superior, inferior and junction sections, and for the ‘IMRT + IMRT’ technique, further splitting the junction PTV into five sub-regions to establish the initial dose gradient. This process takes a considerable amount of time, although it could be largely automated within Pinnacle through the development of automated scripts to perform these tasks. During the technique development, we also performed comprehensive and thorough quality assurance measurements and isocenter verification to test its stability and feasibility. This increased workload is well justified, given the factor that it is a two IMRT treatment and it delivers substantial benefits to the patient in terms of a more conformal treatment with a reduced risk of morbidity compared to conventional treatment options. Extended SSD parallel-opposed fields tend to treat large volumes, while conventional abutting isocentric treatment fields result in significant regions of dose non-uniformity. Though the moving junction technique can reduce the non-uniform dose, significant variations can still exist. These benefits from an improved treatment delivery for long volume targets therefore have been judged to outweigh any increased workload issues at our institution. 5. Conclusions We have developed a new technique to permit the treatment of long volume targets with IMRT in a commercially available treatment planning system without any custom software. This technique uses stepped dose gradient and IMRT inverse planning to achieve junction ‘feathering’ target dose uniformity and OAR sparing. The plans demonstrate stable junctioning with dose heterogeneity being controlled within ±5% and tolerating 3 mm setup errors with dose variations less than 5% in the target and no significant changes in OAR sparing. Acknowledgments Part of this work has been presented as an oral presentation at the 2006 AAPM annual meeting in Orlando, Florida, USA. We would like to thank Dr Michael Sharpe for support in IMRT treatment planning and Mr Tony Manfredi from Elekta for technique support on the Elekta linac.
4552
G G Zeng et al
References Chan M F, Chui C S, Schupak K, Amols H, Burman C and Ling C 2001 The treatment of large extra skeleton chondrosarcoma of the leg: comparison of IMRT and conformal radiotherapy techniques J. Appl. Clin. Med. Phys. 2 3–8 Childress N L, Salehpour M, Dong L, Bloch, White R A and Rosen I 2005 Dosimetric accuracy of Kodak EDR2 film for IMRT verifications Med. Phys. 32 539–48 Haryanto F, Fippel M, Bakai A and Nusslin F 2004 Study on the tongue and groove effect of the Elekta multileaf collimator using Monte Carlo simulation and film dosimetry Strahlentherapie und Onkologie 180 57–61 Hong L, Alektiar K, Hunt M, Venkatraman E and Leibel S A 2004 Intensity-modulated radiotherapy for soft tissue sarcoma of the thigh Int. J. Radiat. Oncol. Biol. Phys. 59 43–50 Hong L et al 2002 IMRT for larger fields: whole-abdomen irradiation Int. J. Radiat. Oncol. Biol. Phys. 54 278–89 Hunt M A et al 2001 Treatment planning and delivery of intensity-modulated radiation therapy for primary nasopharynx cancer Int. J. Radiat. Oncol. Biol. Phys. 49 623–32 Kestin L L, Sharpe M B, Frazier R C, Vicini F, Yan D, Matter R C, Martinez A A and Wong J 2000 Intensity modulation to improve dose uniformity with tangential breast radiotherapy Int. J. Radiat. Oncol. Biol. Phys. 48 1559–68 Khan F M 1994 The Physics of Radiation Therapy 2nd edn (Lippincott: Williams and Wilkins) Ling C C et al 1996 Conformal radiation treatment of prostate cancer using inversely-planned intensity-modulated photon beam produced with dynamic multileaf collimator Int. J. Radiat. Oncol. Biol. Phys. 35 721–30 Malhotra H K, Raina S and Avadhani J S 2005 Technical and dosimetric considerations in IMRT treatment planning for large target volumes J. Appl. Clin. Med. Phys. 6 77–87 Wu Q, Arnfield M, Tong S, Wu Y and Mohan R 2000 Dynamic splitting of large intensity-modulated fields Phys. Med. Biol 45 1731–40
