Effect of the simulated half leaf width of a multileaf collimator on volumetric modulated arc therapy plan quality in hippocampal avoidance whole‐brain radiotherapy

Abstract Purpose Whole‐brain radiotherapy (WBRT) is commonly used in patients with multiple brain metastases. Compared with conventional WBRT, hippocampal avoidance WBRT (HA‐WBRT) more favorably preserves cognitive function and the quality of life. The hippocampal volume is considerably small (approximately 3.3 cm3). Therefore, downsizing the leaf width of a multileaf collimator (MLC) may provide higher spatial resolution and better plan quality. Volumetric modulated arc therapy (VMAT) could simulate the half MLC leaf width through couch shifting between arcs. This study investigated changes in VMAT quality for HA‐WBRT with a simulated fine MLC leaf width. Methods We included 18 patients with brain metastasis. All target and avoidance structures were contoured by an experienced radiation oncologist. The prescribed dose was 30 Gy in 10 fractions. For each patient, three different treatment plans were generated for comparison: VMAT with couch‐shift, VMAT without couch‐shift, and TomoTherapy. All treatment plans fulfilled Radiation Therapy Oncology Group (RTOG) 0933 criteria for HA‐WBRT. The Wilcoxon paired signed‐rank test was used to compare different treatment plans. Results VMAT with couch‐shift had the better average conformity index (0.823) with statistically significant difference compared to VMAT without couch‐shift (0.810). VMAT with couch‐shift (0.219) had a more favorable average homogeneity index (HI) than did VMAT without couch‐shift (0.230), although the difference was not significant. TomoTherapy had an optimal average HI of 0.070. In terms of the hippocampus, all three treatment plans met the RTOG 0933 criteria. VMAT with couch‐shift had a lower average D max (15.2 Gy) than did VMAT without couch‐shift (15.3 Gy, p = 0.071) and TomoTherapy (15.5 Gy, p = 0.133). The average D 100% of hippocampus was the same for both VMAT with and without couch‐shift (8.5 Gy); however, TomoTherapy had a lower average D 100% value of 7.9 Gy. The treatment delivery time was similar between VMAT with and without couch‐shift (average, 375.0 and 369.6 s, respectively). TomoTherapy required a long average delivery time of 1489.9 s. Conclusion The plan quality of VMAT for HA‐WBRT was improved by using the couch‐shift technique to simulate the half MLC leaf width. However, the improvement was not statistically significant except conformity index. The downsizing effect decreased with the use of the sophisticated grade of VMAT. TomoTherapy offered superior plan quality but required the longest delivery time.

effect decreased with the use of the sophisticated grade of VMAT. TomoTherapy offered superior plan quality but required the longest delivery time.

K E Y W O R D S hippocampus, leaf width, volumetric modulated arc therapy INTRODUCTION
Brain metastasis occurs in 10-30% of patients with cancer. As effective systemic treatment regimens have been demonstrated to prolong life, the incidence of brain metastasis is increasing. Whole-brain radiotherapy (WBRT) is most commonly used in patients with numerous brain metastases because it can alleviate symptoms and substantially improve intracranial control. Approximately 40-60% of patients respond to WBRT. 1,2 However, WBRT is associated with many neurological side effects including the development of dementia, cerebellar dysfunction, and deficits in neurocognitive functions, such as short-term memory.
Deficits in learning, memory, and spatial processing occurring in patients undergoing WBRT are believed to be related to hippocampal injury. 3 Several studies have demonstrated that cognitive function is associated with radiosensitive neural stem cells within the hippocampal dentate gyrus. 4 Radiation Therapy Oncology Group (RTOG) 0933 is a multi-institution phase II trial investigating the effectiveness of hippocampal avoidance (HA) during WBRT for brain metastasis. The trial revealed that HA-WBRT exhibited a stronger association with the preservation of memory and quality of life in patients with brain metastasis than in historical controls. 5 Because the hippocampus is located in the deep central area of the brain, its optimization is considerably difficult because the main organ at risk is completely surrounded by the planning target volume (PTV). Therefore, RTOG 0933 suggested using intensity-modulated radiotherapy (IMRT), helical TomoTherapy, or volumetric arc therapy (VMAT). In addition, the hippocampal volume is substantially small; on average, it is 3.3 cm 3 and accounts for 2.1% of the whole-brain PTV. 6 Downsizing the leaf width of MLC may provide higher spatial resolution and better radiotherapy plan quality. Abisheva et el. investigated the effect of the leaf width of an MLC on single-isocenter multitarget radiosurgery with VMAT and found that a 2.5-mm-wide MLC had a lower isodose spill than did a 5-mm-wide MLC. 7 In addition, VMAT could simulate the dosimetric effect of the half MLC leaf width with couch shift between arcs. 8 In this study, we investigated changes in the VMAT quality for HA-WBRT with a simulated fine MLC leaf width. By using an MLC with a 1-cm-wide leaf on an Elekta Synergy linear accelerator, we simulated 5-mmwide leaves with a 5-mm shift of the patient couch in the longitudinal direction between arcs.

Patient selection, delineation, and planning constraints
This study was approved by the institutional review board of our institution. A total of 18 patients with brain metastasis treated with HA-WBRT were reviewed. Patients underwent a noncontrast computed tomography (CT) simulation scan with a slice thickness of 3 mm. The CT simulation and MRI were fused, and target and avoidance structures were contoured by an experienced radiation oncologist by using the Pinnacle 3 version 14 planning software. The hippocampus was contoured on T1-weighted MRI axial sequences in accordance with RTOG 0933 contouring guidelines. The HA region was generated through a 5-mm three-dimensional expansion of the hippocampal contour. The clinical target volume (CTV) was defined as the whole-brain parenchyma to C1, and the PTV was defined as the CTV excluding the HA region. Other normal structures that were contoured included the optic nerves, optic chiasm, lenses, and inner ears. The inner ears included the cochlea and internal auditory meatus. Following RTOG guidelines, the treatment prescribed was a delivery of 30 Gy in 10 fractions to the PTV. The RTOG 0933 acceptable compliance criteria for target and normal tissue planning doses are as follows: 1. At least 90% of the brain volume should receive 30 Gy (V 30Gy ≥ 90% PTV). 2. 2% of the target volume should receive 37.5 Gy or less (D 2% ≤ 37.5 Gy). 3. 98% of the target volume should receive 25 Gy or more (D 98% ≥ 25 Gy). 4. The minimum dose to the hippocampus (D min = D 100% ) should be ≤9 Gy. 5. The maximum dose to the hippocampus should be ≤16 Gy. 6. The maximum dose to the optic nerves and chiasm should be ≤37.5 Gy.

VMAT planning
VMAT plans were generated using 6-MV photon beams and optimized using Pinnacle 3 Version 14 on the Elekta Synergy platform with 40 MLC leaf pairs with a leaf width of 1 cm at the isocenter. For each patient, two VMAT plans were generated: one without couch-shift and one with couch-shift in the longitudinal direction between full arcs (FAs). The VMAT plan without couchshift was generated using four FAs (clockwise for two FAs and counterclockwise for two FAs) without couchshift between arcs. The collimator angle was 45 • . The VMAT plan with couch-shift was also generated using the same four FAs as the VMAT plan without couch-shift. But we shifted the patient couch by 5 mm in the longitudinal direction after the two FAs (one clockwise FA and one counterclockwise FA). The collimator angle in this plan was also 45 • .

TomoTherapy planning
TomoTherapy plans were generated using the Hi-ART planning system (TomoTherapy Inc, Version 5.1.4, Madison,WI,USA).The plan parameters were as follows:field width, 1.05 cm; pitch, 0.215; and modulation factor, 3.0. The plans were optimized such that 96% of the wholebrain PTV received the prescription dose of 30 Gy in 10 fractions.

Plan evaluation
For the PTV, the volume receiving > 30 Gy (V 30Gy ) and the minimum dose covering 98% of the volume (D 98% ) were used to assess coverage.The dose delivered to the hottest 2% of the PTV (D 2% ) was used to determine the number of hotspots in the treatment plan. The conformity index (CI) and homogeneity index (HI) were defined as follows: where V RI is the total volume in the body receiving the prescribed dose, TV is the volume of PTV, and TV RI is the volume of TV within V RI . Smaller HI values closer to 0 indicate superior homogeneity, whereas larger values closer to 1 indicate inferior homogeneity. The ideal and maximum CI value is 1, with a larger CI indicating better conformality.
For the hippocampus and HA ring, the maximum, minimum, and mean doses were extracted for comparison. Other normal tissue doses extracted for comparison in this study included the minimum and maximum doses to the optic nerves, optic chiasm, lenses, and inner ears.
In addition, the delivery time for a single fraction of HA-WBRT was recorded for all three modalities and did not include the time taken for pretreatment patient setup and daily imaging procedures.

Statistical analysis
Statistical comparisons of the treatment plans between the three modalities were performed using a Wilcoxon paired signed-rank test with the SPSS Version 25 statistical software (IBM, USA). A p value of ≤0.05 was considered statistically significant. Table 1 lists the characteristics of the 18 patients.  Table 2, and p values are summarized in Table 3. Figure 1 displays the dose distribution from the three treatment modalities for one sample patient in our study. Average accumulated dose-volume histogram (DVH) of the PTV and the hippocampus is shown in Figure 2. The treatment plans for all the patients were in compliance with RTOG 0933 protocol dosimetric criteria.

PTV
TomoTherapy provided the most satisfactory PTV coverage among the modalities, with an average V 30Gy

Hippocampus and HA ring
In terms of the dose to the hippocampus, both VMAT with (8.5 Gy) and without (8.5 Gy) couch-shift had a higher average D 100% than did TomoTherapy (7.9 Gy). However, all the doses were within the protocol's recommendation of 9 Gy. In terms of D 100% , TomoTherapy was significantly superior to VMAT with and without couchshift. No difference in D 100% was noted between VMAT with and without couch-shift. VMAT with couch-shift had a lower average D max (15.

Optic nerves and optic chiasm
For the optic nerves and chiasm, VMAT with and without couch-shift as well as TomoTherapy met the RTOG 0933 constraint (D max ≤ 37.5 Gy). TomoTherapy had a lower average D max value than did both VMAT with and without couch-shift. No significant difference in D max values was observed between VMAT with and without couch-shift. TomoTherapy had a higher average D mean value for the optic nerves than did VMAT with and without couch-shift. The average D mean value of the optic nerves was similar for VMAT with and without couch-shift.

Delivery time
TomoTherapy required the longest treatment delivery time (1489.9 s) averaging over 18 patients. VMAT with couch-shift had an average delivery time of 375.0 s, which was similar to that of VMAT without couch-shift (369.6 sec). TomoTherapy had almost four times longer delivery time than did VMAT with and without couchshift. Notably, the delivery time was an estimated value recorded from the time the first beam delivery was started to the end of the last beam delivery.

DISCUSSION
When all the three modalities met RTOG 0933′s dosimetric criteria, we found that VMAT with couchshift yielded a better CI (average CI = 0.823) and In 2016, Park et al. investigated the effect of an extremely narrow-width MLC leaf on the plan quality of VMAT for prostate cancer. 8 They simulated 2.5-mmwide MLCs by shifting the isocenter position by 2.5 mm with 5-mm-wide MLCs. The profile of the 2.5-mm MLC simulation was almost the same as that of the real 2.5mm wide MLCs. The Halcyon IMRT/VMAT delivery system is composed of two staggered layers of 28 leaf pairs, with a projected leaf width of 10 mm. This doublestaggered MLC has an effective leaf resolution of 5 mm. Several studies have compared the plan quality between a Halcyon and a TrueBeam, which comprises 40 central leaf pairs with a width of 5 mm and 20 peripheral leaf pairs with a width of 10 mm. 9,10 These studies have found that the Halcyon could provide a plan quality (b) F I G U R E 2 Continued similar to that of the TrueBeam for VMAT techniques. In the present study, we used the VMAT plan with couchshift in the longitudinal direction between arcs to simulate a fine MLC leaf width.
The leaf width of MLCs is crucial for target shaping resolution. A thin leaf width generally results in higher resolution and therefore provides more favorable dosimetric outcomes in IMRT planning. Several studies have investigated the dosimetric benefits of thin-leaf MLCs in IMRT and have shown that thin-leaf MLCs in IMRT yielded better target coverage, CI, and HI. [11][12][13] However, the downsizing effect of the MLC leaf width decreased with the use of the sophisticated grades of IMRT and VMAT. 14 With respect to normal tissue sparing, dosimetric improvements still decreased with the use of the sophisticated grades of techniques (IMRT or VMAT). 14, 15 Chae et al. investigated the dosimetric effects of leaf width on IMRT and VMAT for small-sized C-shape and large-sized head and neck phantoms and found that the thin leaf could yield higher CI and a lower spinal cord dose for C-shape. In our study, we used a couch shift between arcs in VMAT to simulate the half width of MLCs for HA-WBRT. The shape of PTV was similar to that of a doughnut. The thin leaf provided a higher average CI (from 0.810 to 0.823, p = 0.039) and a more favorable average HI (from 0.230 to 0.219, p = 0.199) for PTV. Moreover, the thin leaf could reduce the average maximum hippocampus dose (from 15.3 to 15.2 Gy, p = 0.071) and the average mean hippocampus dose (from 11.4 to 11.2 Gy, p = 0.396), although the results were not significant. The results of the present study are consistent with those of the aforementioned studies.
Pigorsch et al. compared the dosimetry of three different plans (VMAT, IMRT, and TomoTherapy) for head and neck cancer. 16 The formula of CI was the same as ours. They used the technique of simultaneous integrated boost (SIB). Best CIs for SIB77Gy and SIB70Gy were achieved by TomoTherapy and the best CI for SIB56Gy was achieved by VMAT. The best performance for HI in all SIBs was achieved only by TomoTherapy. In our study, VMAT with couch-shift provided better CI (0.823) than did TomoTherapy (0.815), although the difference was not statistically significant. TomoTherapy provided the optimal HI. DVH also showed that TomoTherapy had steepest dose gradient of PTV.
In clinical practice, the prerequisite of the VMAT with couch-shift technique is an accurate setup of a patient based on 3D image guidance. In this study, we used an Elekta Synergy linear accelerator. And we moved the patient couch 5 mm in the longitudinal direction between arcs to simulate half leaf width. In a previous study, the positioning error of Elekta Synergy was 0.11 ± 0.12 mm in the longitudinal direction. 17 This technique should be applied with couches that could operate with submillimeter precision. 17