Influence of maximum MLC leaf speed on the quality of volumetric modulated arc therapy plans

Abstract Purpose Maximum leaf speed is a configurable parameter of MLC in a treatment‐planning system. This study investigated the influence of MLC on the quality of VMAT plans. Methods Seven MLCs with different maximum leaf speeds (1.0, 1.5, 2.25, 3.5, 5.0, 7.5, and 10 cm/s) were configured for an accelerator in treatment‐planning system. Correspondingly, seven treatment plans, with the identical initial optimization parameter, were designed with the mdaccAutoPlan system. Six nasopharyngeal carcinoma (NPC) patients and nine rectal cancer patients were selected, representing complex and simple clinical circumstances. VMAT plan quality was evaluated with PlanIQTM software. The results were statistically analyzed with a one‐way analysis of variance (ANOVA) and pairwise comparison tests. Results The relative changes of plan scores achieved by the seven configured accelerators, with specific maximum MLC leaf speed (MMLS) for each patient, were studied. Two apparent trends of MMLS influence on VMAT plan scores were observed: Plan scores increased with MMLS; Plan scores increased rapidly when MMLS increased from 1 to 3.5, thus the relative change of plan score decreased in this MMLS range. The stationary point of maximum MLC speed (MMSSP) is defined, for the specific MMLS when the relative changes of plan scores is first <5%, as MMLS increases from 1.0 to 10. For rectal plans, MMSSPs were 2.25 for six patients and 3.5 for the other three patients. For NPC plans, MMSSPs were 3.5 for five patients and 2.25 for one patient. Conclusion This work indicates that MMLS directly influences VMAT plan quality in NPC cases and rectal cancer cases. VMAT plan quality improved conspicuously as MMLS increased from 1 to 3.5, VMAT plan quality with marginal improvement when MMLS is above 3.5.


| INTRODUCTION
Volumetric modulated arc therapy (VMAT), first introduced in 2007, was described as a novel technique in radiotherapy. 1,2 Volumetric modulated arc therapy delivers a highly conformal prescription dose to the target volume and spares the surrounding normal tissue. It is achieved through modulation of the intensity of photon beams, via the simultaneous variation of three parameters during the treatment delivery: gantry rotation speed; treatment aperture, shaped by multileaf collimator (MLC) leaves; and dose rate.
Multileaf collimators are used to shield anatomic structures from photon radiation and to modulate the field of incident photon fluence. 3 Due to the MLC's physical design, it has a great impact on the photon radiation field. Taking the Varian (Varian Associates, Palo Alto, CA, USA) linear accelerator (LINAC) as an example, the MLC works with the primary collimator or jaw (the second) are fixed or follow the open window of the MLC dynamically. 4 Leakage and transmission through the MLC leaf directly affect plan quality in relation to dosimetry. Multileaf collimator leaf transmission was evaluated in terms of plan quality in Ref. [5]. Furthermore, highly modulated dose distribution requires that the MLCs move at high speed when the gantry rotates. It is worthwhile to consider the relationship between plan dosimetric variation (or plan quality) and maximum MLC leaf speed change. Rapid MLC speed could possibly improve plan quality. However, it could go too far, for the following three reasons. First, the fast leaf motion during gantry rotation may be affected by interleaf friction or MLC motor problems that result in leaf position errors. As demonstrated by Kerns et al. 6 and Park et al., 7 both MLC speed and MLC acceleration exert a considerable effect on VMAT delivery accuracy. Based on clinical uses, it is not reasonable that maximum leaf speed be as rapid as possible. Second, both the dose rate and gantry speed in VMAT are allowed to vary, in addition to MLC leaf positions, to generate a highly conformal target dose distribution with minimal delivery time and monitor units (MUs). It is worthwhile comprehensively investigating the influence of maximum MLC leaf speed on the plan quality of VMAT. Third, plan quality is the best metric for determining the optimal maximum leaf speed for VMAT in clinical practice. It is widely believed that the plan quality may be improved by increasing the MLC leaf speed.
However, there has been no thorough study about the benefit of the maximum MLC leaf speed parameter in the LINAC. Validating the optimal maximum MLC leaf speed parameter for clinical implementation is of great interest in clinical practice. The first column displays the PQM components, the structure presents in square brackets with the corresponding metric. The second column shows the endpoint and optimal structure metrics that are corresponds to the minimum and maximum score, respectively, in the third column. Cancer staging system. 12 The head, neck, and shoulders were immobilized using a perforated, thermoplastic head, and neck mask in a supine position. Treatment-planning CT images extending from the vertex of the skull to the carina of trachea were obtained and indexed every 3 mm. The target volumes were delineated using an institutional treatment protocol, which is defined in Ref. [13]. Gross  Table 1.
Total [21 metrics] 0 36 The first column displays the plan quality metric components, the structure presents in square brackets with the corresponding metric. The second column shows the endpoint and optimal structure metrics that are corresponds to the minimum and maximum score, respectively, in the third column. The mdaccAutoPlan system was developed based on our clinical protocol, with authorization from developer Zhang's team. The quality of the planning outcome depends on the method followed by each planner, 15 so the use of automated planning decreased interoperator variability 16 and guarantee high-quality VMAT and IMRT treatment plans in our study 17 . In our study, VMAT plans were calculated using 6-MV photons, with a maximum variable dose rate of 600 MU/min. Double arcs with coplanar arcs of 360°shared the same isocenter, using opposite rotation (clockwise and counter clockwise). The collimator was always rotated to 10°and 350°, respectively, in two arcs, to avoid a tongue and groove effect. The maximum rotation time of each arc was set to 60 s, to guarantee the leaf travel was as rapid as possible. The gantry angle spacing was 4°. The calculation voxel size was isotropic and 4 mm.

P-values of pair comparison
[ROI] metrics S1 vs S1.5 S1 vs S2.25 S1 vs S3.5 S1 vs S5 S1 vs S7.5 S1 vs S10 S1.5 vs S2.25 S1.5 vs S3.5 S1.5vs S5 S1.5 vs S7.5 S1.5 vs S10  The plans reflected all optimization objectives routinely employed in our clinic, 19 which are more stringent than similar objectives from RTOG protocols. 20 The Homogeneity Index (HI) was defined as in Ref. [21]. Conformation Number (CN) represents the dose fit of the PTV, relative to the volume covered by the prescribed isodose lines, which are defined in Ref. [22]. V n Gy (%) is the percentage of the organ volume receiving ≥ n Gy. 5 D v cc and Dmean are the near-maximum absorbed dose (where V is a small fractional volume) and average absorbed doses delivered to each OAR, respectively. 23 In mathematics, a "stationary point" is an input to a differentiable function such that the derivative is zero. In order to determine the optimal MMLS, the stationary point of maximum MLC speed (MMSSP) was defined as the relative change of plan score that is

| RESULTS AND DISCUSSION
VMAT treatment plans were generated for each patient in this study, with the seven MLCs applied. In total, 63 rectal plans and 42 NPC plans were generated and analyzed for these patients. Two well-defined PQMs of rectal and NPC cases were used for qualitative and quantitative analysis of plan quality for 105 plans. Figure 1  Detailed plan quality variations in the seven types of MMLS plans are shown in Figure 2(a) for rectal cancer cases. The relative change of plan quality scores with all rectal cancer cases decreased F I G . 4. Comparison between S1 and S3.5 plans for patient #6 with rectal cancer. Axial dose distribution of S1 plan (a) and S3.5 plan (b) and corresponding dose volume histograms (DVHs) (c) for Patient #6. The dose distributions show planning target volume (PTV) prescribed to 5000 cGy (red line). The DVHs of the S1 plan (solid lines) and S3.5 plan (dashed lines) include the following ROIs: bladder (purple), intestine (slate-blue), colon (blue), left femur head (brown), right femur head (maroon), normal tissue (orange) and PTV (green). The S1 plan for Patient #6 achieved fewer dose evaluation criteria than did the S3.5 plan. This is shown by the higher maximum dose for the bladder, in which the S1 plan exceeded the criteria limit.
at lower leaf speed when MMLS was below 2.25 cm/s. After that, the relative change of plan scores was mostly within 5%. This scenario was conspicuous in rectal cases as well, even more apparent.
When MMLS was above 3.5 cm/s, an optimum plan with little further upgrade in plan quality was produced for small increases in leaf speed. It is more possibly that the MMLS of 3.5cm/s met the requirement to deliver the plan. MMSSP is the specific MMLS at which the relative change of plan scores first drops in a tiny interval (<5%). Figure 2  For plan scores of rectal cancer cases, the pairwise ANOVAs showed (a) the plan scores of S1 plan did not differ significantly from that of S1.5 plan, (b) while S1 plan was significantly difference to F I G . 5. Comparison between S1 plan and S3.5 plans for a patient with locally advanced nasopharyngeal carcinoma. Axial dose distribution for S1 plan (a, b) and S3.5 plan (c, d), with corresponding DVHs (e) for Patient #4. The dose distributions show PGTVnx and GTVrpn prescribed to 73.92 Gy (red line), GTVnd prescribed to 69.96 Gy (purple), and PTV1 prescribed to 60.06 Gy (teal line). The DVHs for the S1 plan (solid lines) and S3.5 plan (dashed lines) include the following ROIs: brainstem (forest green), spinal cord (yellow-green), larynx (purple), left parotid (brown), right parotid (maroon), left Lens (sky blue), right lens (steel-blue), and left and right temporal lobes (tomato red). Patient #4, in whom the target exceeded more dose evaluation criteria than clinically required, and lots of hot spots are on target. This is shown by the higher intermediate dose for the parotids, in which the S1 plan exceeded the criteria limit.
S2.25, S3.5, S5, S7.5, and S10 plans methods (P < 0.01). (c) Except S1 plan, the five MMLS plans did not differ significantly from each other. For NPC cases, the pairwise ANOVAs showed (a) S1 plan was significantly difference to other five MMLS plans methods (P < 0.01), (b) the five MMLS plans did not differ significantly from each other. The statistical results of plan scores from seven MMLS type plans in both rectal cases and NPC cases were consistent with MMSSP result in both Figs. 2 and 3.
The corresponding analysis with PQM components in two disease sites yielded similar results, but less pairs with significant differences.
For PQM components of rectal cancer cases, the 5 of 21 components which show the statistically differences are reported in Table 4, while the other MMLS type plans did not differ significantly from each other.
In particular, the CN of PTV in S1 plan for rectal cases showed significant difference with other six MMLS type plans, respectively. The CN value of PTV in S1.5 plan from rectal cases was significantly lower than that in S7.5 plans (P = 0.044). For PQM components of NPC cancer cases, the 9 of 40 components which show the statistically differences are reported in Representative VMAT plans for a rectal cancer patient, with axial dose distribution and the DVHs of PTV and OARs from S1 and S3.5 plans, are shown in Fig. 4. As shown in the top-left and bottom-left panels of Fig. 4, the S3.5 plan not only achieved better conformity to the 95% isodose line (50 Gy) of the PTV, but also included fewer hot spotsthe 107% isodose line (53.5 Gy)than the S1 plan did. In addition, the S3.5 plan produced steeper DVHs than the S1 plan did in the high-dose ranges, as shown in Fig. 4(c). That means that the irradiation dose curves were more constricted around the target in the S1 plan. For selected OARs, the normalized bladder volume in the S3.5 plan was smaller than in the S1 plan, in the 10 to 35 Gy dose range, and even over 50 Gy. The percentage volume of the colon was significantly smaller throughout the entire dose range in the S3.5 plan than in the S1 plan, especially between 10 and 30 Gy, and over 50 Gy. Interestingly, the volume of left and right femur heads received radiation doses below 40 Gy in the S3.5 plan were smaller than in the S1 plan. The same numeric trend was observed in the NT DVH.

| CONCLUSION S
This work indicates that the maximum leaf speed of MLCs has an influence on the quality of VMAT plans in NPC cases and rectal cancer cases. The quality of VMAT plans is greatly improved as MMLS increases from 1 to 3.5 cm/s; above that, the quality change is marginal.

ACKNOWLEDG MENTS
This work was partially financed by the National Natural Science Foundation of China (11475261 and 81801799). We thank Claire Barnes, PhD, from Liwen Bianji, Edanz Editing China (www.liwenbia nji.cn/ac), for editing the English text of a draft of this manuscript.

A U T H O R S C O N T R I B U T I O N
All authors discussed and conceived of the study design. Jiayun Chen performed case selection and data analysis, score metric development, and drafted the manuscript. Weijie Cui did the initial valida-

C O N F L I C T O F T H E I N T E R E S T
None declared.