Impact of treatment planning using a structure block function on the target and organ doses related to patient movement in cervical esophageal cancer: A phantom study

Abstract Helical tomotherapy (HT) can restrict beamlets passing through the virtual contour on computed tomography (CT) image in dose optimization, reducing the dose to organs at risk (OARs). Beamlet restriction limits the incident beamlet angles; thus, the proper planning target volume (PTV) margin may differ from that of the standard treatment plan without beamlet restriction, depending on the patient's movement during dose delivery. Dose distribution changes resulting from patient movement have not been described for treatment plans with beamlet restriction. This study quantified changes in dose distribution to the target and OARs when beamlet restriction is applied to cervical esophageal cancer treatment plan using HT by systematically shifting a phantom. Treatment plans for cervical esophageal cancers with and without beamlet restriction modes [directional block (DB) and nonblock (NB), respectively] were designed for CT images of the RANDO phantom. The PTV margin for the DB mode was set to be the same as that for the NB mode (5 mm). The CT image was intentionally shifted by ±1, ±2, and ±3 voxels in the left–right, anterior–posterior, and superior–inferior directions, and the dose distribution was recalculated for each position using the fluence for the NB or DB mode. When the phantom shift was within the same PTV margin as the NB mode, changes in doses to the targets, lungs, heart, and spinal cord in the DB mode were small as those in the NB mode. In conclusion, the virtual contour shape used in this study would provide safe delivery even with patient movement within the same PTV margin as for the NB mode.


| INTRODUCTION
Helical tomotherapy (HT) is a dose delivery technique, in which the treatment couch moves in the direction of the gantry rotation axis, and the high-speed driving of 64 multileaf collimators allow the fluence to be finely modulated. 1 Studies have shown that HT and volume-modulated arc radiotherapy help improve the dose concentration at the target for cervical esophageal cancers. [2][3][4][5] However, their use may increase the risk of radiation pneumonitis and pulmonary complications compared to using three-dimensional (3D) conformal radiation therapy or intensity modulated radiotherapy, because of the increased low-dose area in the lung. [6][7][8] Nomura et al 6 reported that the volumes receiving at least 5, 10, 15, and 20 Gy (V 5Gy , V 10Gy , V 15Gy , and V 20Gy , respectively) and mean lung dose were significantly associated with the development of symptomatic radiation pneumonitis. Lee et al 7 observed significant postoperative pulmonary complications after preoperative chemoradiation for esophageal cancer when the V 10Gy of the lung was >40%. To address this issue, Chang et al 9 reported that the pulmonary dose could be reduced in dose optimization of HT using a structure block function, which restricts beamlets that pass through the virtual contour on the computed tomography (CT) image. They used fan-shaped virtual contours in the lungs, assessed with a virtual esophageal tumor delineated on the CT image of a phantom. 9 Building on that study, our research group evaluated dose reductions to the organs at risk (OARs) and dose concentrations at the target using various virtual contour shapes for cervical esophageal cancers in 20 patients. 10 This showed that a semicircular contour following the shape of the lung at a distance of 8 cm from the tracheal bifurcation was the most clinically useful, when the dose reduction to the OARs and the concentration of the dose at the target were considered as a single index. 10 The restriction of beamlets in dose optimization of HT has been applied to reduce the dose to OARs in various other conditions as well as esophageal cancers. 11 However, beamlet restriction limits the incident beamlet angles; hence, the doses to the targets and OARs may change significantly depending on patient movement during the dose delivery. It has not yet been established whether the planning target volume (PTV) margin in the standard treatment plan without beamlet restriction (e.g., 5 mm) can be applied to the beamlet restriction plan; using the same PTV margin may result in an insufficient dose to the target. In addition, there have been no reports for treatment plans with beamlet restriction related to changes in the dose distribution resulting from patient movement. The aim of this study was to quantify changes in the dose distribution of the target and OARs when beamlet restriction is applied to a cervical esophageal cancer treatment plan using HT by systematically shifting a phantom. For comparison, a standard treatment plan without beamlet restriction was designed, and the difference in the dose distribution change between the treatment plans with and without beamlet restriction was evaluated.

2.B | Dose optimization
The Planning Station has a function that restricts beamlets passing through the virtual contour. This function has two modes: complete block (CB) and directional block (DB). Figure 2 how beamlet A, which reached the virtual contour before passing through the PTV, is not included in the dose optimization process, whereas beamlet B, which reached the contour after passing through the PTV, is included. In our previous study, the DB mode with the virtual contour shape used in the present study could create treatment plans for 20 patients without any clinical problems. 10 We therefore used the DB mode in the present study.
Jaw size, modulation factor, and pitch, which are the treatment planning parameters 16 of TomoTherapy were set at 2.5 cm, 2.1, and 0.43, respectively . The prescribed dose was set as 60 Gy to the 95% volume of PTV VTV and 48 Gy to the 95% volume of PTV VPNV using the simultaneous integrated boost technique, and it was delivered in 30 fractions. Dose optimization was applied to satisfy the following constraints for PTVs: dose to 98% of the volume (D 98% ) ˃ 54 Gy, D 95% ˃ 58.8 Gy, D 50% ˂ 64.2 Gy, and D 2% ˂ 72 Gy for PTV VTV ; and D 98% ˃ 43.8 Gy, D 95% ˃ 46.8 Gy, D 50% ˂ 55.8 Gy, and D 2% ˂ 64.2 Gy for PTV VPNV . Constraints for the OARs were as follows: maximum dose (D max ) ˂ 130% inside the phantom; D max ˂ 52 Gy and dose to a volume of 1 cm 3 (D1cm 3 ) ˂ 50 Gy for the PRV of the spinal cord; V 10Gy ˂ 50%, V 15Gy ˂ 40%, and V 20Gy ˂ 25% for the lungs. The doses to the thyroid and heart were optimized to be as low as possible.
For comparison, a standard treatment plan without beamlet restriction [nonblock (NB) mode] was designed with the same dose descriptions and dose constraints.

2.C | Evaluation of the changes in dose with phantom shifts
The CT image set of the phantom used in the above treatment planning was registered in the Planning Station as a verification phantom.
The CT image was then shifted ±1, ±2, and ±3 voxels in the left-     Table 1 shows the volumes for which the dose distribution difference in Fig. 4 was greater than 6 Gy (equivalent to 10% of the prescribed dose to the PTV VTV ) or <−6 Gy. The dose distribution changes in the DB mode were quantitatively larger than those in the NB mode for the shift in the LR direction but smaller for the shift in the AP direction.

3.C | Changes in the dose parameters resulting from the phantom shift
Tables S1 and S2 show the dose parameters for the VTV, VPNV, heart, spinal cord, and thyroid in the NB and DB modes. In both modes, the change in dose parameters was small. The dose constraints for PTV VTV , PTV VPNV , and PRV of the spinal cord in the treatment planning were applied to the VTV, VPNV, and spinal cord, respectively, and when the one-dimensional (1D) phantom shifts were within two voxels, the dose parameters in the two modes were within the dose constraints. The values in brackets in Tables S1 and S2 indicate the percentage of dose parameters for each phantom shift for the dose parameter value without the phantom shift in the two modes. Table 2 shows the percentage difference in dose parameters between the two modes. When the 1D phantom shift was <2 voxels, the values of D 98% , D 95% , D 50% , and D 2% for VTV and VPNV showed only a small difference between the two modes, of less than approximately 1%. In addition, the differences of D mean and V 40Gy for the heart, D max and D1cm 3 for the spinal cord, and D mean for the thyroid between the two modes were several percent.   the change in the D 98% of the CTV was <5% in most cases of volume-modulated arc radiotherapy for esophageal cancer when the patient shifted ±5 mm in the LR and AP directions and ±7 mm in the SI direction. They set the PTV margin at 5 mm. Additionally, they showed that there were only small changes in D max for the spinal cord, D mean for the heart, and V 20Gy for the lungs. 21 Our study used simple planning models with the RANDO phantom. The results showed that, if a 1D phantom shift in the DB mode was within the same PTV margin as for the NB mode, the change in doses to the target, lungs, heart, and spinal cord was small and within the treatment planning dose constraints. Thus, using the virtual contour shape in this study for a treatment plan in the DB mode for cervical esophageal cancer may keep the dose distribution changes that result from patient movement within a clinically acceptable level. In our previous study, we evaluated the dose reduction to the OARs and dose concentration to the targets using various virtual contour shapes; the virtual contour shape used in the present study was the one that scored the highest in the previous study, which was shown to be the most clinically useful for 20 patients. 10 This virtual contour shape, a semicircle that follows the shape of the lung at a distance of 8 cm from the tracheal bifurcation, can effectively reduce the dose to OARs and provide safe delivery, even with patient movement, with the same PTV margin as the NB mode.
In our results, the use of the DB mode changed the robustness of the dose distribution around the target, although it did not significantly affect the dose parameters. Lee et al 15 reported an increase in the dose to the spinal cord by restricting the beamlet that passed through the right hepatic lobe for hepatocellular carcinoma of the left lobe. This was because the dose distribution spread to avoid the virtual contour outline when the DB mode was used. 11,15 In the present study, the treatment plan using the DB mode increased the beam weight in the AP direction to avoid the bilateral lungs by setting virtual contours. This resulted in a gradual gradient of the medium-and low-dose areas in the AP direction [ Fig. 3(b)]. Conversely, the gradient of the medium-and low-dose areas for the LR direction became steep. A gradual dose distribution gradient improves the robustness of the dose with phantom shift, so the dose distribution change resulting from the phantom shift in the DB mode was larger for the LR direction and smaller for the AP direction than those in the NB mode ( Fig. 4 and Table 1). Thus, the robustness of a treatment plan using the DB mode would be more affected in the direction parallel to the virtual contour and less affected in the direction perpendicular to the virtual contour. In addition, because the beam weight in the AP direction increases as the bilateral virtual contours becomes closer, the robustness of the dose distribution with respect to the phantom shift in the LR direction would be impaired.
As shown in Fig. 5(b), using the DB mode greatly reduced V 5Gy for the lungs compared to using the NB mode, indicating that the DB mode can reduce the risk of radiation pneumonitis.
However, V 5Gy for the lungs in the DB mode changed greatly with a phantom shift in the LR direction, although that in the NB mode hardly changed. Thus, the DB mode differed from the NB mode in the change of a dose parameter resulting from the phantom shift. Although the dose constraint was satisfied in this study, it is possible that this difference may result in a dose constraint not being satisfied in other treatment cases, depending on the shape and displacement of the virtual contour. This result shows the importance of quantifying the dose uncertainty in the DB mode by confirming the dose distribution change resulting from patient movement.
A limitation of this study was that we confirmed the dose distribution change in the cervical esophageal cancer model by using only 1D phantom shifts. However, real patients move in three dimensions. It is therefore necessary to evaluate whether irradiation is possible with a model of 3D patient movement. In addition, we were not able to perform similar investigations with the TomoDirect™ because we did not have a license of it.

| CONCLUSION S
For a treatment plan that used the DB mode of HT for cervical esophageal cancer, the change of doses to the target, lungs, heart, and spinal cord would be as small as those of the NB mode if the patient movement was within the same range as the PTV margin of the NB mode. Thus, the virtual contour shape used in this study, a semicircle that followed the shape of the lung at a distance of 8 cm from the tracheal bifurcation, would provide safe delivery when patient movement was within the same PTV margin as the NB mode. However, because the DB mode changed the robustness of the dose distribution around the targets, the dose constraint might not be satisfied depending on T A B L E 1 Volumes of the dose distribution difference >6 Gy and <−6 Gy in Fig. 4 (cm 3

SUPPORTING INFORMATION
Additional supporting information may be found online in the Supporting Information section at the end of the article.