Treatment planning of volumetric modulated arc therapy and positioning optimization for hippocampal‐avoidance prophylactic cranial irradiation

Abstract Background Hippocampal‐avoidance prophylactic cranial irradiation (HA‐PCI) offers potential neurocognitive benefits but raises technical challenges to treatment planning. This study aims to improve the conventional planning method using volumetric modulated arc therapy (VMAT) technique and investigate a better patient’s head positioning to achieve a high quality of HA‐PCI treatment plans. Methods The improved planning method set a wide expansion of hippocampus as a special region for dose decline. The whole brain target was divided into two parts according to whether the slice included hippocampus and their optimization objectives were set separately. Four coplanar full arcs with partial field sizes were employed to deliver radiation dose to different parts of the target. The collimator angle for all arcs was 90°. Tilting patient’s head was achieved by rotating CT images. The improved planning method and tilted head positioning were verified using datasets from 16 patients previously treated with HA‐PCI using helical tomotherapy (HT). Results For the improved VMAT plans, the max and mean doses to hippocampus were 7.88 Gy and 6.32 Gy, respectively, significantly lower than those for the conventional VMAT plans (P < 0.001). Meanwhile, the improved planning method significantly improved the plan quality. Compared to the HT plans, the improved VMAT plans result in similar mean dose to hippocampus (P > 0.1) but lower max dose (P < 0.02). Besides, the target coverage was the highest for the improved VMAT plans. The tilted head positioning further reduced the max and mean doses to hippocampus (P < 0.05), significantly decreased the max dose to lens (P < 0.001) and resulted in higher plan quality as compared to nontilted head positioning. Conclusions The improved planning method enables the VMAT plans to meet the clinical requirements of HA‐PCI treatment with high plan quality and convenience. The tilted head positioning provides superior dosimetric advantages over the nontilted head positioning, which is recommended for clinical application.


| INTRODUCTION
Prophylactic cranial irradiation (PCI) is an effective way to prevent brain metastases (BM) in lung cancer patients. [1][2][3][4] Several clinical trials have shown that PCI significantly decreased the incidence of brain metastases compared with observation. [5][6][7] However, the use of PCI will induce adverse effects like neurocognitive deficits, which are believed to be caused by radiation induced damage of neural stem cell (NSC) compartment in the hippocampus. 8,9 In order to reduce these cognitive side-effects, it is necessary to minimize radiation dose to the hippocampus during PCI. That is to perform hippocampal-avoidance PCI (HA-PCI).
Among the current radiotherapy techniques, helical tomotherapy (HT) is considered to be the preferred technique to treat complex treatment situations because the radiotherapy modality of helical tomoscan has a powerful modulation capability. Many studies have confirmed that using HT technique could achieve superior dose conformity and homogeneity for concave or even hollow target adjacent to sensitive structures, which is just right for the HA-PCI treatment. 10,11 However, due to its high cost, HT technique is not common or even unavailable in most hospitals and clinics of China. For instance, a total of 166 patients received PCI at Cancer Hospital Chinese Academy of Medical Science in 2018, only 18% of who were treated with HA-PCI using HT. The rest were all treated with PCI without avoiding hippocampus using traditional conformal radiotherapy (CRT). Compared with HT, volumetric modulated arc therapy (VMAT) is a far more common technique. But most studies showed that it is usually less than satisfactory in hippocampus sparing and dose homogeneity of target. [12][13][14] Therefore, HA-PCI using a VMAT is not readily available in the clinic. In this study, we aimed to improve the planning method of conventional VMAT plans to achieve a high plan quality comparable to that of HT plans. Besides, we attempted to change patient's head positioning to further improve the plan quality for HA-PCI.

2.A | Patients selection and contouring
For this planning study, 16 patients, who had been previously treated with HA-PCI using HT in 2018-2019, were randomly selected.
All patients had undergone cranial computed tomography (CT) and magnetic resonance imaging (MRI) scans, both with 2-mm thickness.
These images were fused in Pinnacle v9. 10 15 The planning target volume (PTV) was a 3-mm uniform expansion of the whole brain excluding the hippocampal PRV. Two dose-shaping structures are created: ring PRV was from 3 to 8 mm outside of the hippocampus PRV, and ring PTV was from 5 to 10 mm outside of the PTV. These structures were showed in Fig. 1.
Additionally, normal tissue structures, including the lens, optic nerve, optic chiasm, brainstem, were contoured for dose evaluation.

2.B | Treatment planning
Volumetric modulated arc therapy plans for each patient were generated in Pinnacle v9.10, separately using the conventional and improved planning methods. For dose calculation, the Adaptive Convolve algorithm with heterogeneity correction was used, with a dose grid resolution of 2 mm. Treatment was delivered using a Elekta Ver-saHD TM (Elekta, Crawley, United Kingdom), 120-leaf MLC, and 6-MV photon beams with a maximum dose rate of 700 MU/min. Maximum leaf motion was limited to 1 cm/deg (6 cm/s). Gantry spacing was set to 3°. For the present study, the clinically used HT plans were included in comparison in order to verify that the improved VMAT plans can be applied in clinic. All the HT plans were generated using a 2.5-cm field width and dynamic jaws. The pitch was selected as 0.257 and the modulation factor was set between 2.0 and 3.0. The treatment prescription to the whole brain PTV was set to deliver 25 Gy in 10 fractions, with at least 90% of the PTV receiving 100% of the prescription dose (PD). The max and mean dose to the hippocampus were limited to 9 Gy and 7 Gy respectively. The max dose to lens could not exceed 8 Gy.

2.C | Improved VMAT planning method
The conventional VMAT plans in this study employed double full coplanar arcs with a collimator of 0°. The main optimization objectives were shown in Table 1. On the base of this conventional planning method, the improved planning method was improved in several ways: 1. Setting a special region for dose decline. In the whole brain PTV, it takes a certain spatial distance to drop from the PD to the dose constraint of the hippocampus. In order to allow the optimizer to fulfill all the objectives more easily, we defined a dose decline region as the hippocampus expanded by 2 cm in left-right and anterior-posterior directions and 1 cm in superior-inferior direction (the yellow region showed in Fig. 1). Then we defined the PTV_plan as the whole brain PTV subtracting the dose decline region. It was used only for the improved VMAT plan optimization and required to achieve the PD coverage as high as possible. In addition, a ring PRV was set in the dose decline region to help control the dose to hippocampus. The main optimization objectives used for the improved VMAT plans were listed in Table 1. Note that we tightened the hippocampal dose constraints to seek a lower hippocampal dose without sacrificing the PTV coverage.

2.
Dividing the whole brain target into two parts. Inhomogeneous dose distribution can easily occur in the slices including the hippocampus, especially near the hippocampus. To control hot spots and cold spots more efficiently, we divided the PTV_plan into two parts: the PTV_plan in the slices that contains the hippocampal dose decline region was defined as PTV_plan1, and the rest was PTV_plan2 (see Fig. 1). Therefore, optimization objectives could be set separately for these two targets. As shown in Table 1, the weights of the objectives were higher and Max Dose goal was stricter for PTV_plan1 than those for PTV_plan2.

3.
Using four coplanar full arcs with limited field sizes. The large size of whole brain target usually requires a large irradiation field, leading to a large range of MLC leaf motion. This may affect the

2.D | Tilted head positioning
Hippocampus is located in the lower part of temporal lobe. As can be seen in Fig. 3, its long axis appears tilted in the sagittal plane.
Thus, if a patient's head is tilted forward at a certain degree, the long axis of the hippocampus will be turned to be parallel to gantry rotation axis, which may help multi-leaf collimator (MLC) leaves to spare the hippocampus. Two recent studies reported that when patients received HA-PCI with 30°tilted head positioning, the dose to the hippocampus and other normal tissues could be further decreased. 16,17 In order to spare normal tissues, especially the hippocampus, to the maximum, we chose the tilt angle of patient's head to be 45°for this study after analyzing hippocampal tilt angle and eyeballs position relative to brain for different patients. For the sake of simplicity, we rotated CT images of 0°position to simulate the situation of a 45°tilted head position. The original CT images were imported to Image J software, rotated along the left-right axis by 45°, and then resampled to create the rotated new CT images. The contours in the 0°CT images were then mapped to the rotated CT images after fusing the two datasets together using rigid registration.  Table 2.

2.E | Plan evaluation
The range of mean hippocampal dose was 23.0%~32.5% of the PD

3.B | Tilt vs. nontilt
The VMAT plans with tilted and nontilted head positioning were all generated using the improved planning method with the same optimization objectives. As shown in Table 2, when the patients tilted their head forward, the doses to the hippocampus and lens were all significantly decreased (P ≤ 0.021) while the coverage of the PTV_plan and PTV-15mm were increased (P < 0.01). It is because tilting patient's head forward leads to a decrease in the area of the hippocampus in axial plane (see Fig. 4) and makes MLC leaves spare the hippocampus more easily (as mentioned earlier), which all help reduce the dose to hippocampus. Moreover, when the patients tilted head forward, the lenses are blocked by collimator jaw at most beam angles so that the dose to lens is also significantly decreased (P < 0.001).
The CI, HI, and MU for the four plans were compared in Table 3. split-arc and partial field. 18 In the present study, we employed four coplanar full arcs and reduced the max hippocampal dose to 35% of the PD (even lower than the HT plans) while maintaining a high plan quality. The planning method we used is mainly improved from the following aspects.
Firstly, adequate space around the hippocampus was reserved for dose decline. Due to the strict hippocampal dose constraint, the low dose region is mainly concentrated around the hippocampus. In certain cases, the dose gradient is basically unchanged. Therefore, the difference between the max hippocampal dose constraint and the PD determines the space size required for dose decline as well as the PD coverage of the target. A lack of space may increase the T A B L E 2 Comparison of dose parameters among the HT plans, conventional VMAT plans, and improved VMAT plans with nontilted head positioning and tilted head positioning in 16 patients. Thirdly, the collimator angle was set to 90°, different from a small angle (5°-30°) applied in most studies. As shown in Fig. 6(a), when the collimator at 0°(or a small angle), some MLC leaf pairs need to both expose the target and spare two or more OARs simultaneously (such as left and right hippocampi, left and right lenses). It will lead to poor conformity and homogeneity of the target in the slices containing these OARs. Whereas Fig. 6(b) showed that when the collimator is rotated to 90°, one MLC leaf pair is only responsible for sparing one side of the OAR (such as left hippocampus, left lens), thereby improving conformity and homogeneity of the target.
Results showed that the improvements above significantly reduced the dose to the hippocampus and enhanced the plan quality when compared with the conventional method. Meanwhile, the plans generated by the improved method were comparable to the HT plans, with slightly lower max hippocampal dose. Since the improved planning method was developed, the HA-PCI based on VMAT technique has been massively used in our hospital, whereas the PCI without avoiding hippocampus using CRT has been deprecated. Up to now, nearly 100 patients have been treated with VMAT HA-PCI. The max and mean doses to hippocampus received by all the patients were lower than 9 Gy and 7 Gy, respectively, which confirmed the reliability and stability of the improved method. Furthermore, this method also can be used for HA-WBRT with metastases boost. Our practice indicated that the mean dose to F I G . 6. Examples of BEV with collimator angles of 0°(a) and 90°(b). PTV_plan1 is shown in green, PTV_plan2 in blue, hippocampus in red, eyeballs in purple.