Plan quality and robustness in field junction region for craniospinal irradiation with VMAT

Physica Medica 48 (2018) 21–26

Contents lists available at ScienceDirect

Physica Medica journal homepage: www.elsevier.com/locate/ejmp

Original paper

Plan quality and robustness in field junction region for craniospinal irradiation with VMAT Keqiang Wanga,b, Huipeng Menga, Jie Chenb, Wenxue Zhangb, Yuanming Fenga, a b

T

⁎

Department of Biomedical Engineering, Tianjin University, 92 Weijin Road, Nankai District, Tianjin 300072, China Department of Radiotherapy, Tianjin Medical University General Hospital, 154 Anshan Road, Heping District, Tianjin 300052, China

A R T I C LE I N FO

A B S T R A C T

Keywords: CSI VMAT Robustness Staggered overlap

Purpose: To propose a “staggered overlap” technique in volumetric modulated arc therapy (VMAT) for craniospinal irradiation (CSI) and compare the dose distribution and plan robustness with “overlap” technique and “gradient optimization” approach. Methods and Materials: 6 patients previously treated in our clinic were retrospectively selected. 9 VMAT plans of each patient were optimized with “staggered overlap”, “overlap” and “gradient optimization” in overlapping region of 3 cm, 6 cm, and 9 cm separately. For the “staggered overlap” plan, adjacent field sets were intentionally overlapped by staggering field edges in an appropriate step size to avoid sharp dose gradient. Evaluation metrics including V95%, D2%, D98%, conformity number (CN) and homogeneity index (HI) were employed to evaluate the dose distribution. Moreover, shifts of the upper spinal field isocenter in each direction were performed to simulate junction errors for robustness analysis. Results: The CN and HI of VMAT plans with “staggered overlap” were 0.82 (0.811–0.822) and 0.113 (0.112–0.114), while they were 0.778 (0.776–0.782) and 0.131 (0.130–0.131) for plans with “gradient optimization”. In the robustness study, < 3% dose deviations were found for 5 mm shifts in lateral and vertical directions with all techniques. In cranial-caudal direction, “overlap” technique created hot spots (D2% > 170%) and cold spots (D98% < 44%) in the junction region with 10 mm shifts. The dose deviations were decreased to 22% for plans with “staggered overlap” and 9 cm overlapping region. Conclusion: “Staggered overlap” technique provides better plan quality as compared to “gradient optimization” approach and makes the plan more robust against junction errors as compared to “overlap” technique.

1. Introduction Craniospinal irradiation (CSI) plays an important role in the treatment of primitive neuroectodermal tumors, germinomas and other central nervous system (CNS) diseases. Due to the length of the target volume and limitation of field size of linear accelerators (LINACs), CSI has been technically challenging. Traditional techniques use multiple fields to cover the entire target volume and carry the risk of overdose or underdose in field junction regions due to beam divergence, patient movement and setup uncertainty. More recently, intensity modulated radiation therapy (IMRT) [1,2], volumetric modulated arc therapy (VMAT) [3–5], TomoTherapy [6], and proton therapy [7] have been used for CSI. These techniques have shown potential in improving target volume conformity and minimizing dose to organs at risk (OARs). VMAT has the advantage of less treatment time and potentially reduces patient movement uncertainty especially for pediatric patients. Similar to traditional techniques, more attentions need to be paid to

⁎

the field junction region in IMRT and VMAT. The goal is to get conformal and homogeneous dose distribution with no junction errors and obtain smaller dose deviations with junction errors. A variety of techniques, such as “feathering” [8], “jagged-junction” [2], “overlap” technique [3] and “gradient optimization” technique [4] were employed. From these studies it can be concluded that the most robust field junction is produced by two adjacent field sets with a low dose gradient at the beam edges in the overlapping region. “Overlap” technique for VMAT is effective and efficient but may not result in a robust plan by the optimizer algorithm itself. “Gradient optimization” approach utilizing longer overlapping region and more dose steps will produce more robust plan. However, this would require more time and effort during plan setup and optimization. In this study, we sought to develop a technique for CSI, referred to as “staggered overlap” technique in which adjacent arc field sets were intentionally overlapped by staggering field edges in an appropriate step size (1/3 of the overlapping length) to avoid an undesirably sharp

Corresponding author. E-mail address: [email protected] (Y. Feng).

https://doi.org/10.1016/j.ejmp.2018.03.007 Received 7 January 2018; Received in revised form 2 March 2018; Accepted 11 March 2018 1120-1797/ © 2018 Associazione Italiana di Fisica Medica. Published by Elsevier Ltd. All rights reserved.

Physica Medica 48 (2018) 21–26

K. Wang et al.

dose gradient. In addition, we compared this technique with “overlap” technique and “gradient optimization” approach in regard to the plan quality and robustness.

spinal junction region) was in pink. Nine VMAT plans of each patient were optimized with “staggered overlap” technique, “overlap” technique and “gradient optimization” approach in overlapping region of 3 cm, 6 cm, and 9 cm separately. The optimization objectives were to have 98% of the PTV covered by 95% of the prescribed dose, and to minimize dose to the OARs. The same optimization objectives and equal iterative calculations were used to all of the plans. Different lengths of overlapping regions were used to explore the correlation between the length of overlap and the plan robustness. As described in the study [3], a conventional “overlap” technique of VMAT was accomplished with three sets of partial arcs for the three isocenters. Collimator angles of 0° were set to all of the arcs. Overlapping region was defined between adjacent arcs to get a smooth dose transition. Field arrangements are illustrated in Fig. 1a. For the “staggered overlap” VMAT plan, adjacent arc field sets were intentionally overlapped by staggering field edges in an appropriate step size (1/3 of the overlapping length) to guide the optimizer algorithm towards the desired outcome. Fig. 2 shows a schematic diagram of the overlapping region with staggered field sets. In addition, collimator angles of 30° and 330° were used for iso-1 arc field set, 10° and 350° were used for iso-2 and iso-3 arc field sets. Field arrangements are illustrated in Fig. 1b and displayed in beam’s eye view (BEV) in Fig. 1c. “Gradient optimization” was performed in two steps. First, the upper spinal PTV was defined as a central region with a homogeneous dose distribution surrounded by two transitional regions. Each of the transitional regions was coincide with the overlapping region and was further divided into subregions with 1 cm step size. The dose prescription in each subregion gradually decreased from center to the periphery (Fig. 1d). Then the upper spinal PTV was optimized and gradient-dose was created in the overlapping region. After that a secondary optimization considering the whole PTV was run to get uniform dose distribution throughout the PTV.

2. Methods and Materials 2.1. Patient and target selection Six patients previously treated in our clinic were retrospectively selected. All the CT scans were acquired with 5-mm slice thickness from a Philips Brilliance Big Bore CT simulator (Philips Medical Systems). The patients were immobilized in a supine position with head and shoulders in a thermoplastic shell and the arms resting comfortably at the patients’ sides. Clinical target volume (CTV, covering the brain and spinal canal) and OARs were contoured on a computed tomography (CT) image set. Planning target volume (PTV) was created by adding 3-mm margin to the cranial CTV and 5-mm to the spinal CTV. OARs contoured included the lungs, heart, liver, kidneys and lenses. To facilitate the comparison, each patient was planned to a total dose of 36 Gy in 20 fractions. 2.2. Field arrangement and treatment planning The 6MV VMAT plans were optimized using Pinnacle SmartArc v9.8 (Philips Radiation Oncology Systems). All plans were designed for an Elekta LINAC (Synergy) equipped with MLCi2 with leaf width of 10 mm at the isocenter. To the cranial PTV, two partial arcs of 180°-30° and 330°-181° were appropriate to cover the target and protect lenses. However to the spinal PTV, 180°-120° and 240°-181° partial arcs were more effective in reducing dose to OARs and avoiding treating through the shoulders and arms. Due to the length of PTV, 3 isocenters longitudinally separated by 25 cm were positioned in the cranial PTV (iso-1), upper spinal PTV (iso2) and lower spinal PTV (iso-3) separately and near the patient’s midline (Fig. 1a). The three isocenters were collinear, requiring only a longitudinal couch shift to move from one to the other. So the junction error was determined by the accuracy and precision of the couch in longitudinal direction. Two junction regions of PTV were created and illustrated in Fig. 1b. PTVupper junction (PTV in the cranial-spinal junction region) was colored with brown and PTVlower junction (PTV in the spinal-

2.3. Plan evaluation and robustness analysis Evaluation metrics including V95%, D2%, D98%, conformity number (CN) and homogeneity index (HI) were employed to evaluate the dose distribution. The CN was used to evaluate the PTV conformity and was calculated using the following mathematical expression [9]: 2 CN = VPTV,pres /(VPTV × Vpres)

(1)

Fig. 1. Schematic diagrams of 3 optimization techniques of VMAT for CSI. (a) Field arrangements of overlap technique, three isocenters and two overlap regions (colored with light blue and green). (b) Field arrangements of staggered overlap technique. PTVupper junction is colored with brown and PTVlower junction is in pink. (c) Field arrangements of staggered overlap technique displayed in BEV. (d) Gradient optimization and subregions setup. (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.)

22

Physica Medica 48 (2018) 21–26

K. Wang et al.

Fig. 2. A schematic diagram of staggered overlap technique shows overlapping of arc field sets of iso-1(red and green), iso-2(yellow and blue) and iso-3(pink and light blue). Adjacent arc field edges stagger in a step size (1/3 of the overlapping length). (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.)

where VPTV,pres is defined as the volume of the PTV that is receiving at least 95% of the prescribed dose, VPTV is the volume of the PTV, and Vpres is the total volume receiving at least 95% of the prescribed dose. Dose homogeneity was compared using the HI, which was calculated as follows:

HI = (D2%−D98%)/D50%

(2)

where D50% is the median dose to the PTV, and D2% and D98% are nearmaximum and near-minimum doses that cover 2% and 98% volume, respectively, of the PTV on dose volume histogram [10]. In order to assess the plan robustness and potential risk associated with setup errors, 3 mm, 5 mm, and 10 mm shifts were made to the upper spinal field isocenter (iso-2) for each direction (left-right (LR), anterior-posterior (AP), and cranial-caudal(CC)). The plans were recalculated for a total of 18 times and compared with the original nonshifted plans. Profiles were used to study the contributions of adjacent field sets in the junction region, which were created along the longitudinal direction (from superior to inferior) of the PTV through the junction region. The near-maximum dose deviation and near-minimum dose deviation in field junction region between the original plan and each of the shifted plans were calculated and expressed as ΔD2% and ΔD98%. Statistical analysis was performed using SPSS 22.0. Wilcoxon signed-ranks test was chosen because the sample sizes were small and not of a normalized distribution, P values < 0.05 were considered statistically significant.

Fig. 3. The dose distributions of a representative patient’s original plan with each technique and 9 cm overlapping region. (a) Overlap technique. (b) Staggered overlap technique. (c) Gradient optimization.

3. Results

(ΔD98%) were found for 5 mm shifts. But 10 mm shifts resulted in 10–22% under-dosage in LR and anterior directions and less than 8% under-dosage in posterior direction. There were no differences among shifted plans with different techniques and overlapping region. Regarding the shifts in CC direction, “overlap” technique created hot spots (D2% > 124%/3 mm, D2% > 138%/5 mm, D2% > 170%/ 10 mm) and cold spots (D98% < 80%/3 mm, D98% < 66%/5 mm, D98% < 44%/10 mm) with the highest values in PTVupper junction regardless of the length of the overlapping region. The dose deviation was almost 60% for 10 mm shifts. For “staggered overlap” technique and “gradient optimization” approach, the dose deviations (ΔD2% and ΔD98%) decreased significantly along with the increase of the overlapping region from 3 cm to 6 cm, but slightly from 6 cm to 9 cm (Table 2 and Fig. S1 in Supplementary material). In the lower junction region, the plan robustness improved similarly for the three techniques with the increase of the overlapping region from 3 cm to 6 cm, but didn’t improve from 6 cm to 9 cm for “overlap” technique and “gradient optimization” approach (Table 3 and Fig. S2 in Supplementary material). The most robust plan was created with “staggered overlap” technique and 9 cm overlapping region. The dose deviations were clinically acceptable with junction errors (5.5%/3 mm, 11%/5 mm, 22%/10 mm in PTVupper junction. 4.8%/3 mm, 8.5%/5 mm, 17%/10 mm in PTVlower junction). Radiation doses to OARs are illustrated in Fig. 4. With the appropriate choice of partial arc range, the mean doses of the OARs were less

Each technique achieved clinically acceptable PTV coverage for all the patients in this study. The dose distributions of a representative patient’s original plan with each technique and 9 cm overlapping region are presented in the sagittal views in Fig. 3. Table 1 shows the dosimetric parameters (V95%, D2%, D98%, CN and HI) for the three techniques (“overlap”, “staggered overlap” and “gradient optimization”) in 3 different overlapping regions (3 cm, 6 cm and 9 cm) over the 6 patients. All plans met the planning aims, V95% of the PTV almost achieved 98% of the PTV volume, D98% was about 95% and D2% was less than 110% of the prescription dose. In the dosimetric comparison, statistically significant differences were seen between “staggered overlap” technique and “gradient optimization” approach. The CN of the PTV was better with “staggered overlap” technique as compared with “gradient optimization” approach (0.82 vs. 0.778, p < 0.05). The dose homogeneity in the PTVupper junction and PTVlower junction was inferior with “gradient optimization” approach and HI were larger than 0.1. The length of overlapping region had no impact on the plan quality. To evaluate the plan robustness, 3 mm, 5 mm, and 10 mm shifts were made to the upper spinal field isocenter (iso-2) in LR, AP, and CC directions. Tables 2 and 3 report D2%, D98% and dose deviations of all shifted plans in PTVupper junction and PTVlower junction. Regarding the shifts in LR and AP directions, no hot spots (greater than 115% of the prescription dose) were found in the junction region. No cold spots were found for 3 mm shifts and less than 3% dose deviations of D98% 23

Physica Medica 48 (2018) 21–26

K. Wang et al.

Table 1 Plan evaluation parameters for the three techniques (overlap, staggered overlap and gradient optimization) with 3 different overlapping regions (3 cm, 6 cm and 9 cm) (mean ± std). Overlap technique

Staggered overlap technique

Gradient optimization approach

Overlap length

3 cm

6 cm

9 cm

3 cm

6 cm

9 cm

3 cm

6 cm

9 cm

PTV V95%(%) D2%(%) D98%(%) HI CN

97.83 ± 0.30 109.7 ± 3.9 94.5 ± 0.9 0.145 ± 0.042 0.777 ± 0.038

97.67 ± 0.28 109.8 ± 3.2 94.3 ± 0.9 0.148 ± 0.036 0.774 ± 0.024

97.98 ± 0.24 109.6 ± 2.0 95.0 ± 0.3 0.141 ± 0.020 0.773 ± 0.028

98.23 ± 0.27 106.9 ± 2.5 95.2 ± 0.5 0.114 ± 0.027 0.811 ± 0.024

98.13 ± 0.14 106.6 ± 2.5 95.1 ± 0.1 0.112 ± 0.022 0.820 ± 0.034

98.07 ± 0.15 106.5 ± 0.8 95.1 ± 0.1 0.113 ± 0.009 0.822 ± 0.014

98.02 ± 0.08 108.4 ± 1.1 95.1 ± 0.1 0.130 ± 0.010 0.782 ± 0.019*

98.00 ± 0.24 108.5 ± 1.8 95.0 ± 0.25 0.131 ± 0.017 0.778 ± 0.026*

98.15 ± 0.14 108.7 ± 1.6 95.2 ± 0.15 0.131 ± 0.016 0.776 ± 0.023*

107.8 ± 2.7 98.8 ± 1.8 0.087 ± 0.015

107.9 ± 2.0 98.7 ± 1.7 0.089 ± 0.017

107.9 ± 2.0 98.6 ± 0.9 0.090 ± 0.018

105.4 ± 1.1 97.3 ± 1.0 0.080 ± 0.007

104.8 ± 1.6 96.9 ± 1.3 0.078 ± 0.010

104.0 ± 0.6 97.1 ± 0.7 0.069 ± 0.004

110.5 ± 1.3 95.0 ± 3.0 0.151 ± 0.031*

108.1 ± 1.9 95.0 ± 2.0 0.128 ± 0.017*

107.7 ± 0.8 94.8 ± 1.4 0.127 ± 0.015*

PTVlower junction D2%(%) 106.8 ± 3.0 D98%(%) 99.1 ± 1.8 HI 0.075 ± 0.016

105.9 ± 1.2 99.9 ± 1.0 0.058 ± 0.005

106.2 ± 1.4 100.1 ± 1.2 0.059 ± 0.009

105.1 ± 1.4 97.8 ± 0.9 0.072 ± 0.011

104.1 ± 2.0 98.1 ± 1.5 0.060 ± 0.008

103.6 ± 0.5 97.9 ± 0.6 0.057 ± 0.003

109.3 ± 2.1 94.8 ± 1.9 0.141 ± 0.012*

107.5 ± 1.3 96.5 ± 1.6 0.107 ± 0.017*

109.8 ± 3.1 98.0 ± 1.3 0.114 ± 0.033*

PTVupper D2%(%) D98%(%) HI

junction

* p < 0.05 between “gradient optimization” approach and “staggered overlap” technique with same overlapping length.

plan with conformal target coverage, reasonable dose deviation to junction error, acceptable doses to OARs and less treatment time. Conventional technique for CSI is a complex procedure which includes adjustment of field size, collimator and couch rotation, and leads to overdose or underdose when there are small setup errors [1]. IMRT has already been used to overcome the challenges associated with the conventional technique. Seppala et al. [11] and Cao et al. [2] used dynamic split field and jagged-junction IMRT approach to get advantage dose distribution and minimize dose inaccuracies with junction errors. But these techniques cannot be translated to VMAT plans. Lee et al. [12] reported beneficial results of using VMAT for CSI as compared to that of 3D-CRT, but they did not discuss the dosimetric impact of junction errors. Fogliata et al. [3] described VMAT planning with “overlap” technique of 2–12 cm overlapping length but provided no special discussion about the effect of overlapping length on the plan robustness. “Gradient optimization” in VMAT was introduced by Myers

than 6 Gy except for the heart (near 8 Gy). There were no statistical differences among the three techniques although “staggered overlap” technique slightly reduced the average doses of the heart and kidneys. 4. Discussion CSI is a challenging treatment with LINAC due to the large and complexly shaped target. One major concern is the field junction that is prone to have hot or cold spots and potentially encounters large dose deviation with junction error. Improving target conformity and dose homogeneity, reducing dose to OARs and junction error are all very important to the quality of CSI treatment. Moreover, improvement of patient comfort, reduction of treatment time and facilitation of treatment delivery play important roles in realizing the planned objectives. Our “staggered overlap”, a VMAT technique with patient in supine position, has been shown advantage in getting a high quality and robust

Table 2 Near-maximum dose (D2%), near-minimum dose (D98%) and dose deviations (ΔD2% and ΔD98%) for the shifted plans in PTVupper Parameters/shift

LR (left: + right: −) D98%(%)/ ± 3 mm D98%(%)/ ± 5 mm D98%(%)/ ± 10 mm

Overlap technique

Gradient optimization approach

3 cm

6 cm

9 cm

3 cm

6 cm

9 cm

3 cm

6 cm

9 cm

96.6 ± 0.6 94.2 ± 0.6 81.7 ± 1.8

97.1 ± 0.6 95.3 ± 0.6 84.4 ± 1.6

97.7 ± 0.7 96.4 ± 1.3 90.3 ± 2.0

96.4 ± 0.9 94.3 ± 0.8 82.6 ± 1.7

96.1 ± 1.2 94.2 ± 1.3 84.2 ± 3.1

95.9 ± 0.5 94.4 ± 0.8 88.7 ± 2.9

95.3 ± 1.4 94.1 ± 2.5 82.6 ± 6.0

94.6 ± 2.4 93.6 ± 2.7 86.5 ± 4.4

95 ± 1.0 93.8 ± 1.3 83.7 ± 4.6

95.6 98.2 93.2 98.1 84.9 96.8

97.2 97.9 95.7 97.0 89.7 94.1

96.0 97.2 93.7 95.9 85.4 92.0

95.5 96.8 93.5 96.0 85.8 92.7

95.3 96.9 93.7 96.4 89.0 94.4

94.3 96.7 92.4 96.9 84.4 95.3

93.8 95.5 92.3 95.5 86.5 95.5

94.9 95.6 93.8 95.4 86.7 93.2

AP(anterior: + posterior: −) D98%(%)/+3 mm 95.2 ± D98%(%)/−3 mm 98.3 ± D98%(%)/+5 mm 92.3 ± D98%(%)/−5 mm 98.0 ± D98%(%)/+10 mm 83.8 ± D98%(%)/−10 mm 95.5 ± CC(cranial: + caudal: D2%(%)/+3 mm ΔD2%(%)/+3 mm D98%(%)/−3 mm ΔD98%(%)/−3 mm D2%(%)/+5 mm ΔD2%(%)/+5 mm D98%(%)/−5 mm ΔD98%(%)/−5 mm D2%(%)/+10 mm ΔD2%(%)/+10 mm D98%(%)/−10 mm ΔD98%(%)/−10 mm

Staggered overlap technique

junction.

0.4 1.0 0.2 0.9 0.8 1.5

−) 127.8 ± 3.8 19.9 ± 3.5 78.9 ± 4.8 −19.9 ± 4.8 140.8 ± 6.1 33.0 ± 4.7 64.9 ± 5.0 −33.9 ± 5.7 174.3 ± 10.4 66.5 ± 9.5 39.0 ± 10.0 −59.8 ± 10.1

± ± ± ± ± ±

0.6 0.7 1.2 0.8 2.0 1.0

128.1 ± 5.3* 20.1 ± 4.5* 77.6 ± 2.8* −21.1 ± 2.7* 143 ± 7.3* 35.1 ± 5.9* 63.2 ± 3.4* −35.6 ± 4.0* 177.6 ± 10.2* 69.7 ± 9.7* 36.6 ± 7.7* −62.1 ± 7.5*

± ± ± ± ± ±

1.0 1.1 1.2 2.0 2.1 3.7

124.3 ± 3.3* 16.4 ± 2.6* 79.3 ± 2.0* −19.3 ± 1.8* 138.6 ± 5.5* 30.7 ± 4.3* 65.7 ± 1.8* −32.9 ± 1.9* 170.0 ± 6.6* 62.0 ± 5.2* 43.1 ± 4.4* −55.5 ± 3.8*

± ± ± ± ± ±

0.7 1.5 1.0 2.3 1.7 3.2

118 ± 3.0 12.6 ± 2.1 85.9 ± 0.9 −11.4 ± 1.4 127.0 ± 3.5 21.6 ± 2.4 77.9 ± 1.5 −19.4 ± 2.3 148.0 ± 5.5 42.6 ± 4.5 56.9 ± 3.8 −40.4 ± 4.5

± ± ± ± ± ±

0.7 2.0 2.1 2.6 3.1 3.5

113.4 ± 2.4 8.6 ± 1.5 89.2 ± 1.2 −7.7 ± 0.9 119.6 ± 3.3 14.8 ± 2.4 83.0 ± 1.4 −13.9 ± 1.2 135.1 ± 4.6 30.3 ± 3.8 68.1 ± 2.4 −28.8 ± 2.5

* p < 0.05 between “overlap” technique and “staggered overlap” technique with same overlapping length.

24

± ± ± ± ± ±

1.6 1.0 2.8 1.2 4.6 1.2

110.0 ± 1.7 5.9 ± 1.5 91.9 ± 1.6 −5.2 ± 1.7 115.7 ± 1.2 11.7 ± 0.9 86.6 ± 1.4 −10.4 ± 1.7 128.4 ± 2.8 24.4 ± 2.5 74.8 ± 2.6 −22.3 ± 2.9

± ± ± ± ± ±

1.4 1.1 1.8 1.3 2.9 2.1

116.9 ± 2.7 6.4 ± 2.7 83.3 ± 3.3 −11.7 ± 1.2 124.7 ± 3.7 14.1 ± 4.5 76.3 ± 5.4 −18.7 ± 3.5 143.3 ± 5.5 32.8 ± 6.5 62.4 ± 5.3 −32.6 ± 4.6

± ± ± ± ± ±

2.1 3.2 2.0 3.5 2.1 4.1

113.5 ± 2.7 5.4 ± 1.2 87.2 ± 2.5 −7.8 ± 1.4 118.8 ± 3.5 10.7 ± 2.9 82.2 ± 3.2 −12.8 ± 2.3 131.6 ± 5.7 23.5 ± 5.6 71.0 ± 4.8 −24.0 ± 3.7

± ± ± ± ± ±

1.1 0.8 1.4 0.9 1.7 2.1

112.7 ± 1.1 5.0 ± 0.8 88.6 ± 2.1 −6.2 ± 1.7 118.4 ± 2.4 10.7 ± 1.9 83.0 ± 3.1 −11.8 ± 3.1 132.5 ± 5.2 24.8 ± 4.5 70.2 ± 4.5 −24.5 ± 4.5

Physica Medica 48 (2018) 21–26

K. Wang et al.

Table 3 Near-maximum dose (D2%), near-minimum dose (D98%) and dose deviations (ΔD2% and ΔD98%) for the shifted plans in PTVlower Parameters/shift

LR (left: + right: −) D98%(%)/ ± 3 mm D98%(%)/ ± 5 mm D98%(%)/ ± 10 mm

Overlap technique

Gradient optimization approach

3 cm

6 cm

9 cm

3 cm

6 cm

9 cm

3 cm

6 cm

9 cm

96.5 ± 0.7 92.8 ± 1.0 74.8 ± 3.2

97.1 ± 0.6 93.5 ± 1.1 75.6 ± 2.9

97.9 ± 1.4 95.0 ± 1.8 80.9 ± 3.7

96.3 ± 0.8 92.6 ± 0.3 74.0 ± 2.7

96.4 ± 1.3 92.3 ± 1.3 72.8 ± 3.2

95.9 ± 0.7 92.6 ± 1.0 76.4 ± 3.7

94.0 ± 1.5 91.5 ± 0.9 75.7 ± 6.0

95.9 ± 1.5 93.5 ± 1.9 78.3 ± 5.8

96.6 ± 1.5 93.7 ± 1.6 74.9 ± 3.7

95.9 98.1 92.3 96.1 81.8 89.4

97.1 98.7 93.8 96.8 83.8 91.0

95.1 97.6 91.8 96.0 81.8 89.2

95.4 97.5 92.0 95.5 80.7 88.1

95.2 96.9 91.6 95.1 80.4 88.1

93.2 96.2 91.1 96.3 82.9 93.5

94.9 97.7 92.8 97.6 84.4 93.4

96.2 98.4 94.0 98.2 85.2 94.6

AP(anterior: + posterior: −) 95.1 ± D98%(%)/+3 mm D98%(%)/−3 mm 97.9 ± D98%(%)/+5 mm 91.4 ± D98%(%)/−5 mm 96.2 ± D98%(%)/+10 mm 80.7 ± D98%(%)/−10 mm 89.5 ± CC(cranial: + caudal: D98%(%)/+3 mm ΔD98%(%)/+3 mm D2%(%)/−3 mm ΔD2%(%)/−3 mm D98%(%)/+5 mm ΔD98%(%)/+5 mm D2%(%)/−5 mm ΔD2%(%)/−5 mm D98%(%)/+10 mm ΔD98%(%)/+10 mm D2%(%)/−10 mm ΔD2%(%)/−10 mm

Staggered overlap technique

junction.

0.6 0.8 0.9 0.6 1.4 1.0

−) 88.8 ± 2.9 −10.2 ± 2.5 116.0 ± 3.7 9.2 ± 2.3 81.9 ± 4.3 −17.1 ± 4.3 123.5 ± 5.9 16.7 ± 4.6 63.3 ± 8.6 −35.7 ± 8.6 142.3 ± 8.9 35.5 ± 7.8

± ± ± ± ± ±

0.5 0.2 0.6 0.6 2.0 2.1

93.8 ± 2.3 −6.1 ± 1.7 112.0 ± 2.3 6.1 ± 1.8 89.1 ± 2.6 −10.8 ± 2.2 117.2 ± 3.5 11.4 ± 3.3 76.8 ± 5.0 −23.1 ± 4.7 129.4 ± 5.9 23.5 ± 5.8

± ± ± ± ± ±

1.4 1.6 1.6 1.7 3.1 1.4

93.5 ± 1.6 −6.6 ± 1.2 113.0 ± 1.3* 6.8 ± 1.1* 87.6 ± 2.6 −12.5 ± 2.3* 118.3 ± 1.7* 12.1 ± 1.8* 75.9 ± 3.3 −24.2 ± 3.1* 129.9 ± 2.6* 23.7 ± 3.0*

± ± ± ± ± ±

1.0 0.8 0.9 0.8 0.9 0.9

86.8 ± 3.8 −10.9 ± 3.2 116.1 ± 3.1 11 ± 2.9 78.7 ± 5.2 −19.1 ± 4.8 124.9 ± 4.8 19.8 ± 4.5 57.4 ± 9.0 −40.4 ± 8.7 145.4 ± 7.5 40.3 ± 7.5

± ± ± ± ± ±

1.6 1.5 2.0 1.7 4.0 2.5

92.7 ± 1.8 −5.4 ± 0.8 109.4 ± 2.0 5.3 ± 1.1 88.4 ± 2.1 −9.6 ± 1.2 114.0 ± 2.2 9.9 ± 1.7 77.1 ± 3.4 −20.9 ± 2.5 125.4 ± 3.6 21.3 ± 3.2

± ± ± ± ± ±

0.6 0.4 0.7 0.5 2.4 1.3

93.1 ± 0.6 −4.8 ± 0.7 108.4 ± 0.9 4.8 ± 1.0 89.6 ± 0.8 −8.3 ± 0.7 112.3 ± 1.3 8.6 ± 1.3 80.3 ± 1.6 −17.6 ± 1.6 120.9 ± 1.9 17.3 ± 2.0

± ± ± ± ± ±

2.5 1.5 2.3 1.2 1.5 1.6

83.1 ± 3.1 −11.6 ± 2.1 116.6 ± 4.0 7.3 ± 2.8 75.7 ± 4.1 −19.1 ± 3.6 124.8 ± 4.6 15.5 ± 3.9 58.3 ± 6.9 −36.5 ± 8.0 145.1 ± 5.9 35.8 ± 4.6

± ± ± ± ± ±

1.4 1.6 1.4 1.4 3.5 2.5

90.2 ± 1.8 −6.3 ± 1.3 113.4 ± 4.4 5.9 ± 3.5 85.3 ± 2.6 −11.3 ± 2.9 118.8 ± 5.9 11.3 ± 4.8 72.7 ± 6.5 −23.8 ± 7.2 130.4 ± 9.9 22.9 ± 8.9

± ± ± ± ± ±

1.3 1.7 1.0 1.5 1.5 2.1

90.2 ± 4 −7.8 ± 4.5 116.8 ± 5.3 7.0 ± 3.0 84.2 ± 6.5 −13.8 ± 6.7 123.7 ± 8.9 13.9 ± 6.8 70.5 ± 12.1 −27.5 ± 11.7 136.9 ± 13.7 27.1 ± 11.5

* p < 0.05 between “overlap” technique and “staggered overlap” technique with same overlapping length.

MLC and optimization objectives. It was also found in our results that “overlap” technique created hot spots and cold spots in PTVupper junction of shifted plans regardless of the length of the overlapping region, and the plan robustness was not improved significantly with the overlapping region from 6 cm to 9 cm for “gradient optimization” approach. VMAT plan optimization is an inverse procedure using optimizer algorithms to achieve pre-set objectives. Initialization parameters and some special settings will guide optimizer algorithms in creating optimal plans. Mancosu et al. [15] used “MU Objective” tool during the optimization of VMAT for breast tumor treatment and found that plans with different MUs had no differences for target coverage but were different in body volume receiving low dose and Gamma Agreement Index. Boman et al. [16] showed the advantage and robustness of dual isocenter VMAT plans with split arc technique. In our study, the “staggered overlap” technique was accomplished by staggering adjacent arc field edges in an appropriate step size to guide the optimizer algorithm towards generating robust plans. In addition, optimal choice of collimator angle increased the optimization freedom to shape a desired dose distribution [17]. Different collimator angles were chosen for each arc in our “staggered overlap” technique. The average CN of the VMAT plan with “staggered overlap” technique is 0.82. In comparison, the CN is about 0.77 with the “overlap” technique and “gradient optimization” approach. With the same PTV coverage, this means that “staggered overlap” technique is beneficial to reduce high dose areas outside the PTV. In the junction region, the dose homogeneity is inferior with “gradient optimization” approach, because it is difficult to get perfect dose distribution in the second step optimization based on the background dose. This may be associated with the limitation of the MLC leaf width. In comparison, it is easy to achieve homogeneous dose distribution with the cooperation of adjacent fields during optimization with “overlap” and “staggered overlap” technique. In the robustness evaluation, we simulated 3 mm, 5 mm, and 10 mm shifts of the upper spinal field isocenter (iso-2) in LR, AP, and CC directions. There were no hot spots found in shifted plans in LR and AP directions. This indicated that dose gradient of each adjacent arc field set was low in LR and AP directions. The plans with steep dose gradient

Fig. 4. Doses to OARs of the 3 different techniques (overlap technique, staggered overlap technique and gradient optimization approach).

et al. [4] and Strojnik et al. [5] and reported to be an effective method in minimizing dose deviation with junction errors. The robustness was achieved through producing a slow, linear and complementary dose gradient at field edges in the overlapping region between adjacent beams. This technique was also implemented in proton irradiation [13,14]. However, it required a significant amount of time to delineate optimization structures, set and adjust the constraints during optimization. It also suffered with poor quality of linear gradient when increasing the length of the junction area, which resulted in step-shaped gradient [4]. In fact, slow linear ramp-like dose gradient in the overlapping region is difficult to achieve. The maximum dose gradient plays important role in the plan robustness. Longer overlapping region is beneficial to decrease the maximum dose gradient. This was confirmed by our results that the dose deviation decreased significantly along with the increase of the overlapping region from 3 cm to 6 cm in the lower junction region. But the maximum dose gradient is also influenced by the optimizer algorithm, the adjacent field setting, the leaf width of 25

Physica Medica 48 (2018) 21–26

K. Wang et al.

in CC direction using “overlap” technique were sensitive to junction errors. The decrease of maximum dose gradient in CC direction with “staggered overlap” technique made plans more robust. For the doses to OARs, there are no statistical differences among the three techniques. Arc range of 330°-30° is avoided in the brain to reduce dose to lenses. The lens of the eye is sensitive to radiation, with formation of cataracts especially for children. Less than 5 Gy to the lenses in our study is lower than most of other reported results [2,3,6]. The mean doses to lungs, kidneys, heart and liver are also acceptable with the arc range of 180°-120°and 240°-181°. Supine position is more stable than the prone position and increases patient’s comfort. The reduction in treatment delivery times of VMAT decreases the possibility of intra-fractional movements. The three isocenters are collinear, requiring only a longitudinal couch shift to move from one to the other. So beam mismatching depends mainly on longitudinal movement inaccuracy of couch. A 3 mm error is sufficient to simulate junction error when not taking patient motion into account, because the couch movement errors are usually less than 2 mm. 5 mm and 10 mm errors are also evaluated to demonstrate more drastic junction errors including patient motion. The limitation of our study is that all the results are based on Pinnacle TPS and LINAC with MLCi2. The conclusion may be different for other TPS or other types of MLC. Further studies with different TPS and MLC are warranted to confirm the results.

online version, at http://dx.doi.org/10.1016/j.ejmp.2018.03.007. References [1] Yom SS, Frija EK, Mahajan A, Chang E, Klein K, Shiu A, et al. Field-in-field technique with intrafractionally modulated junction shifts for craniospinal irradiation. Int J Radiat Oncol Biol Phys 2007;69:1193–8. [2] Cao F, Ramaseshan R, Corns R, Harrop S, Nuraney N, Steiner P, et al. A threeisocenter jagged-junction IMRT approach for craniospinal irradiation without beam edge matching for field junctions. Int J Radiat Oncol Biol Phys 2012;84:648–54. [3] Fogliata A, Bergstrom S, Cafaro I, Clivio A, Cozzi L, Dipasquale G, et al. Craniospinal irradiation with volumetric modulated arc therapy: a multi-institutional treatment experience. Radiother Oncol 2011;99:79–85. [4] Myers P, Stathakis S, Mavroidis P, Esquivel C, Papanikolaou N. Evaluation of localization errors for craniospinal axis irradiation delivery using volume modulated arc therapy and proposal of a technique to minimize such errors. Radiother Oncol 2013;108:107–13. [5] Strojnik A, Mendez I, Peterlin P. Reducing the dosimetric impact of positional errors in field junctions for craniospinal irradiation using VMAT. Rep Pract Oncol Radiother 2016;21:232–9. [6] Parker W, Brodeur M, Roberge D, Freeman C. Standard and nonstandard craniospinal radiotherapy using helical TomoTherapy. Int J Radiat Oncol Biol Phys 2010;77:926–31. [7] Lin H, Ding X, Kirk M, Liu H, Zhai H, Hill-Kayser CE, et al. Supine craniospinal irradiation using a proton pencil beam scanning technique without match line changes for field junctions. Int J Radiat Oncol Biol Phys 2014;90:71–8. [8] Parker WA, Freeman CR. A simple technique for craniospinal radiotherapy in the supine position. Radiother Oncol 2006;78:217–22. [9] Feuvret L, Noel G, Mazeron JJ, Bey P. Conformity index: a review. Int J Radiat Oncol Biol Phys 2006;64:333–42. [10] DeLuca PJD, Gahbauer R, et al. Prescribing, recording, and reporting photon-beam intensity-modulated radiation therapy (IMRT). J ICRU 2010;10:1–106. [11] Seppala J, Kulmala J, Lindholm P, Minn H. A method to improve target dose homogeneity of craniospinal irradiation using dynamic split field IMRT. Radiother Oncol 2010;96:209–15. [12] Lee YK, Brooks CJ, Bedford JL, Warrington AP, Saran FH. Development and evaluation of multiple isocentric volumetric modulated arc therapy technique for craniospinal axis radiotherapy planning. Int J Radiat Oncol Biol Phys 2012;82:1006–12. [13] Farace P, Vinante L, Ravanelli D, Bizzocchi N, Vennarini S. Planning field-junction in proton cranio-spinal irradiation - the ancillary-beam technique. Acta Oncol 2015;54:1075–8. [14] Farace P, Bizzocchi N, Righetto R, Fellin F, Fracchiolla F, Lorentini S, et al. Supine craniospinal irradiation in pediatric patients by proton pencil beam scanning. Radiother Oncol 2017;123:112–8. [15] Mancosu P, Reggiori G, Alongi F, Cozzi L, Fogliata A, Lobefalo F, et al. Total monitor units influence on plan quality parameters in volumetric modulated arc therapy for breast case. Phys Med 2014;30:296–300. [16] Boman E, Rossi M, Kapanen M. The robustness of dual isocenter VMAT radiation therapy for bilateral lymph node positive breast cancer. Phys Med 2017;44:11–7. [17] Mancosu P, Cozzi L, Fogliata A, Lattuada P, Reggiori G, Cantone MC, et al. Collimator angle influence on dose distribution optimization for vertebral metastases using volumetric modulated arc therapy. Med Phys 2010;37:4133–7.

5. Conclusions Both plan quality and robustness need to be considered for craniospinal irradiation. “Staggered overlap” technique provides better plan quality as compared to “gradient optimization” approach and makes the plan more robust against junction errors as compared to conventional “overlap” technique in longer overlapping region. Conflict of interest Authors have no conflict of interest. Acknowledgments We thank Jun Wu for her help on data acquisition and Rongxin Zhang for his useful discussion. Appendix A. Supplementary data Supplementary data associated with this article can be found, in the

26