Dosimetric effects related to collimator angle optimization in intensity‐modulated radiotherapy planning for gastric cancer

To investigate the dosimetric impact of different collimator angle optimization methods in intensity‐modulated radiotherapy of gastric cancer.

F I G U R E 1 Advantages of collimator rotation. (a) Increasing irradiation range. (b) Reducing dose leakage out of the field. (c) Improving the conformity between the target and the multi-leaf collimators radiotherapy, 6 whereas intensity-modulated radiotherapy (IMRT) has also become an important therapeutic means in gastric cancer. 7 The goal of treatment planning is to create a conformal radiation dose distribution throughout the patient's body by optimizing, among others, the number of fields, beam orientation, beam energy, couch rotation, and collimator rotation. Among the aforementioned degrees of freedom, collimator rotation has been shown to have an important impact on the quality of the IMRT planning. 8 Collimator angle rotation optimization has several theoretical advantages. First, increasing the range of exposure. The range of exposure would be 56 cm after rotating the collimator 45 • , as shown in Figure 1a. Second, minimizing the conformal distance. When the Eclipse 13.5 treatment planning system was used to design the plan, split fields (SF) occurred automatically once the distance between the X-jaws exceeded 14 cm. In fact, high doses may occur at the junction of the SF as a result of system errors, as shown in Figure 1b. 9 Third, improving the conformity. Rotating the collimator allowed critical structures to be better isolated from tumor irradiation, as shown in Figure 1c. Fourth, reducing the leakage between multi-leaf collimators (MLC). The tongue and groove effect between the MLC leads to a certain amount of ray leakage ,as confirmed previously; [10][11][12][13] although other studies reported that rotating the collimator would contribute to a reduction in the influence of leakage on the overall dose distribution. 14,15 Finally, improving the spatial resolution of fluence. The spatial resolution was limited by the leaf width; however, the spatial resolution was approximated by a circle with a diameter equal to the minimum leaf displacement when the collimator was rotated. 16 Although some scholars have studied the influence of collimator angle optimization (CAO) in IMRT planning, CAO has not yet been systematically exploited. 17-20 Furthermore, we found few studies about CAO in gastric cancer IMRT planning, and even fewer using the automatic CAO technology of the Eclipse Planning System (Varian Medical Systems, Palo Alto, CA, USA). Moreover, no previous study has provided an analytical standard-based CAO technique for IMRT planning. In the present study, the innovative technique of collimator angle corresponding to the minimum X-jaw gap (CL X ), when the distance of the X-jaw was shortest around the planning target volume (PTV), was obtained by processing the pictures of beam's eye view of the PTV with MATLAB programming (MathWorks; Natick, MA, USA). In this study, the dosimetric differences between the other two CAO techniques and the conventional 0 • (CL 0 ) in PTV and organs at risk (OARs) of gastric cancer were systematically studied for clinical reference.

Delineation of the target volumes and OARs
All of the target volumes and OARs were contoured on individual axial CT slices in all patients by the same radiation oncologist. The clinical target volume was planned to encompass the remaining stomach, tumor bed, and draining lymph node stations (perigastric, pancreaticoduodenal, porta hepatis, celiac, and suprapancreatic). The PTV was defined by adding an isotropic margin of 5 mm from the clinical target volume in the 3-D directions. The range of the PTV volume was 496-931 cm 3 , the mean target volume was 653 ± 148 cm 3 , and the OARs included the liver, kidneys, small intestine, and spinal cord.

IMRT planning
The treatment planning system of the Eclipse was used for the planning design. All of the IMRT plans were delivered using a Varian Clinac of the liver was to receive <30 Gy, the mean dose (D mean ) for the liver was <18 Gy; 33% of each kidney was to receive <22.5 Gy, the D mean for each kidney was <15 Gy; 50% of the small intestine was to receive <25 Gy; and the maximum spinal cord dose was <40 Gy. IMRT planning for all patients was initially produced with collimator angle zero (CL 0 ), and then later reoptimized at different collimator angles using the same dose constraints and normalization parameters as the original plan to avoid human interference.

CAO
Three plans were generated for each patient with different CAO techniques as follows: CL 0 , the collimator angle of each field was set to 0 • ; CL A , each field was set to optimize the collimator angle using the Eclipse algorithm; and CL X : the orientations of the collimator angles were obtained automatically by processing the pictures of the beam's eye view of the PTV with MATLAB programming. The CL X process is shown in Figure 2, the yellow line represents the simulated X and Y-Jaws, in which the X-jaws distance corresponding to the rotation angle changes with the rotation of the collimator. Therefore, the collimator angle is established when the distance of the X-jaws was the shortest around the PTV. The PTV conformability of the three collimator angles is shown in Figure 3. To make it convenient for clinical physicists or dosimeters to call the program to acquire the corresponding angle of CLx, we used a graphical user interface to visualize the program, importing the image using the "Import images" button, and then clicking "Compute best angels" to calculate the optimal angle of CLx (the original procedure is shown in the Appendix). It should be noted that it was not necessary to make the Jaws tangent to the PTV in the actual planning, and the work was only to determine the optimal angle of CLx.

Plan evaluation
All plans were analyzed using a dose-volume histogram. The PTV was as proposed by Low et al. 24 The plan was only accepted if >95% pixels had the value of γ ≤1, and was rejected when γ >1. Repeated measurements of the Varian Clinac IX linear accelerator showed that the average gantry speed was approximately 5.9 • /s, whereas the collimator speed was approximately 3.5 • /s.

Statistical analysis
All statistical analyses were performed using SPSS Statistics 22.0 software (IBM, Armonk, NY, USA). One-way ANOVA was used for multiple comparisons of target and critical organs in different CAO. P-values <0.05 were regarded as statistically significant.

PTV
All IMRT plans met the dose constraint requirement for gastric cancer radiotherapy. For CI, CL 0 was larger, whereas CL 0 was not significantly different from CL A and CL x . For HI, CL X had the best homogeneity, CL 0 was the smallest, CL X was 3.2% larger than CL 0 (P > 0.05), and there was no significant difference in target homogeneity between TA B L E 2 Comparison of dosimetric metrics for planning target volume and organ of interest between the three optimizations P 1 is the P-value for the collimator angle set to 0 • (CL 0 ) versus collimator selected by Eclipse automatic optimization (CL A ); P 2 is for CL 0 versus collimator angle corresponding to the minimum X-jaw gap (CL X ); and P 3 is for CL A versus CL X . CTV, clinical target volume; D max , maximum dose; D mean , minimum dose; PTV, planning target volume; V x% , relative volume of the target or organ of interest received at least x% of the prescribed dose; V xGy , relative volume of organ of interest covered by the x Gy isodose line.

Collimator angle optimization (mean ± SD) P-value
different CAOs. For the mean dose of PTV, CL 0 had the smallest average dose, and CL X was 0.7% larger than CL 0 (P < 0.05). For the D min, CL A decreased by 1.87%, whereas the difference between CL X and CL 0 was <0.4% (P > 0.05). For D max , CL X was the largest, and CL X increased by 0.61% relative to CL 0 (P > 0.05). These results are shown in Table 2. P 1 is the P-value for the collimator angle set to 0 • (CL 0 ) versus collimator selected by Eclipse automatic optimization (CL A ); P 2 is for CL 0 versus collimator angle corresponding to the minimum X-jaw gap (CL X ); and P 3 is for CL A versus CL X . CP, control point; MU, monitor unit; SF, split field; Time, dose delivery time. SFs; the results are summarized in Table 3, and the error bar chart is shown in Figure 4. For the MU, CL A and CLx were reduced by 9.34% and 25.02%, respectively (P < 0.05), compared with CL 0 . For the CP, CL A and CLx were reduced by 7.75% and 26.23%, respectively (P < 0.05), compared with CL 0 . For the SF, CL A and CLx were reduced by 0.2 and 1.3 SFs, respectively, compared with CL 0 . For the treatment time (Time), CLx decreased by 10.03%, whereas CL A increased by 5.04%.

γ Passing rates
The γ passing rates of all plans were >95%. For the same γ criteria, CLx > CL A > CL 0 , CL X and CL A were 3.1% and 2.22% larger than CL 0 for 2 mm/2%, respectively; for 2 mm/3%, CL X and CL A were 1.63% and 1.3% larger than CL 0 , respectively; CL X and CL A were 1.05% and 0.89% larger than CL 0 for 3 mm/2%, respectively; and for 3 mm/3%, CL X and CL A were 0.38% and 0.34% larger than CL 0 , respectively. The results are summarized in Table 4, and the error bar chart is shown in Figure 5.

DISCUSSION
In the current study, we systematically analyzed the dosimetric effects of different CAOs on gastric cancer. It was evident that non-0  which has been rarely studied in the Eclipse planning system, was also discussed. The results showed that CLx had the lowest number of treatment MUs, because when the distance of the X-jaw was shortest around the PTV, the conformal distance of the MLC would also be shortest, therefore reducing the difficulty of planning. For a large target area over 14 cm, SFs will occur if the conventional 0 • collimator angle is used to optimize the planning. SFs might result in hot or cold spots at the junction due to the uncertainty of the daily position; in this regard, CLx optimization is a good choice, because it can minimize the conformal distance. 9 In the current study, to verify the delivery efficiency, the measured fluence on the IBA I'mRT MatriXX ionization chamber array detector was compared with the treatment planning system dose plan with 2-D γ evaluation. The 2-D γ index evaluation of planned and delivered fluence showed that the γ passing rates of each planning was >95%, and CLx > CL A > CL 0 in the same γ index; thus, these results showed that the delivery efficiency of CLx was better.
We determined the collimator angle based on 2-D beam's eye view images; in further studies, it might be useful to obtain the collimator angle in 3-D directions according to the shape of the target, because it might be better to irradiate the target area and protect the OAR. Conventionally, volumetric modulated arc therapy (VMAT) plans generally only use one fixed collimator angle; however, the quality of VMAT plans are also related to the collimator angles. [32][33][34][35][36][37][38][39][40][41][42] More recently, dynamic collimator optimization has been proven to have a positive influence on dosimetric distribution in VMAT. [43][44][45][46][47] Similarly, acquiring multiple collimator angles during gantry rotation using the CLx technique in VMAT plans also warrants further investigation.
CLx optimization might not be suitable for all cancers, and might be more suitable for cancers with relatively regular single target shapes, such as rectal cancer and conventional lung cancer.
In summary, CAO could play an important role in improving the quality of treatment plans. The CAO techniques of CL A and CL x could achieve comparable dose distribution to conventional CAO (CL 0 ) in IMRT for gastric cancer, and could reduce the number of treatment MUs, CPs, and SFs, as well as the likelihood of second cancer, leakage.
Furthermore, the CAO techniques of CL A and CL x could also reduce the tongue and groove effect, and improve the conformity of the target volume dose and the spatial resolution. Different CAO techniques increase the possibilities for the choice of radiotherapy planning, and it has been suggested that CLx optimization should be used in gastric cancer IMRT planning to obtain higher-quality IMRT planning. Thus, CAO increases the freedom to create a conformal radiation dose distribution in the patient. The results of the present study set the groundwork for guiding the CAO with regard to PTV dose distribution and sparing of OARs in gastric IMRT planning.