Characterization of the HalcyonTM multileaf collimator system

Abstract Purpose To characterize the stacked and staggered dual‐layer multileaf collimator (MLC) on the HalcyonTM system. Methods The novel MLC assembly was reviewed and compared to the widely used MillenniumTM 120‐leaf MLC system. We investigated the MLC positioning stability over 70 days using Machine Performance Check (MPC) data. We evaluated the leaf transmission, penumbra, leaf end effect, and leaf edge effect. Leaf transmission through distal, proximal, and both MLC layers was measured with a Farmer chamber, by comparing an open and a closed field. Leaf penumbra was measured using film for three different MLC‐defined field sizes. The leaf end effect was measured with sweeping gap fields of varying gap sizes defined by the distal MLC. The leaf edge effect was evaluated using the Electronic Portal Imaging Device (EPID) for the different banks, gantry positions, and collimator angles. Point dose measurements for 10 test plans were compared to dose predictions of two dose calculation model versions. Results From MPC data, the largest measured MLC positioning accuracy deviation was within 0.1 mm. The proximal MLC exhibited greater deviations compared to the distal MLC. The distal‐and‐proximal‐combination had reduced inter‐leaf and intra‐leaf transmission compared to delivery with distal‐only. The measured leaf transmission was 0.41% for distal‐only, 0.40% for proximal‐only, and negligible for distal‐and‐proximal‐combination. The leaf end penumbra was wider compared to the leaf edge penumbra. The leaf end effect was measured to be −0.2 mm. The leaf edge effect showed minimal bank, gantry position, and collimator angle dependence. However, a systematic deviation between measurements and treatment planning system handling of the leaf edge effect was observed. The discrepancy between the measured and predicted dose in the 10 test plans improved with the latest version of the dose calculation algorithm. Conclusion The characteristics of the stacked and staggered dual‐layer MLC on the HalcyonTM system were presented.


| INTRODUCTION
Beam shaping plays a central role in increasing the accuracy, efficiency, and quality of radiation treatments. Multileaf collimators (MLCs) have been used in radiotherapy over three decades, [1][2][3][4][5][6][7][8][9] initially as beam shapers, eliminating heavy shielding blocks, and later for intensity modulated radiotherapy (IMRT) and volumetrically modulated arc therapy (VMAT). 10 Various MLC designs have been described over the years, each version aiming to further improve the outcome and quality of radiation therapy. 4,8,[11][12][13] For IMRT and VMAT treatments, the dose delivered to the target volume is sensitive to leaf positioning and leaf transmission. Characteristics of a well-designed MLC therefore are: low leaf transmission, small tongue and groove effect, small penumbra, accurate leaf positioning, and faster speed. 5,8,14,15 In The Halcyon TM commissioning process is straightforward and streamlined to allow for a short period of time from installation to treatment. While the Halcyon TM beam output has been described, 16 the unique stacked and staggered dual-layer MLC has not been independently characterized. Detailed characterization of the MLC system can provide a deeper understanding of the system's limitations, and thus inform the quality assurance protocols needed to ensure accurate radiation deliveries.
The purpose of this study was to characterize and assess performance of the stacked and staggered dual-layer MLC system on the Halcyon TM linear accelerator. We measured the MLC positioning accuracy and reproducibility over time. We also evaluated the leaf transmission, leaf penumbra, leaf end effect, and the leaf edge effect. To determine clinical impact, we examined ten plans used in end-to-end tests.
In this study, we comprehensively described the characteristics of the novel Halcyon TM stacked-and-staggered dual-layer MLC.
Overall, we found the Halcyon TM MLC system to be compatible with allowing for a clinic to have a strong focus on treating with intensity modulation, as a result of the MLC system's accurate leaf positioning, substantially low leaf transmission, and high leaf speed enabling the use of high dose rates. Nevertheless, the maximum MLC-defined field of 28 cm 2 × 28 cm 2 may limit its use in certain disease sites requiring large field coverage.

| MATERIALS AND METHODS
The Halcyon TM MLC system features a unique stacked-and-staggered dual-layer design, consisting of a distal and a proximal layer. 17,18 The nomenclature for the MLC layers refers to their positions with reference to the source (Fig. 1). The primary and secondary collimators are fixed in place, and are not movable jaws.
Also, note the absence of a flattening filter.
The distal, or lower layer, is comprised of two banks with 28 leaves each. The proximal, or upper layer, is comprised of two banks with 29 leaves each. The leaves are made from 95% tungsten (the remaining 5% is proprietary) and have a step edge. As shown in Fig. 1, the leaf ends form a truncated pie shape focused on the target, matching beam divergence in the direction perpendicular to the leaf travel (single-focused). Each leaf measures 1.0 cm projected at isocenter with a leaf end curvature radius of 23.4 cm. The leaves are capable of 14 cm over-travel (covering the entire field) and the leaf positioning tolerance is 1 mm.   positioning. An in-house script was used to automatically detect and export MPC data each time MPC was performed. We analyzed the MLC positioning accuracy and reproducibility by grouping the leaves according to their respective layers and banks. To assess the inter-leaf and intra-leaf transmission, two picket fence deliveries were acquired using the EPID. The first picket fence delivery was defined with both MLC layers, with the proximal leaves always trailing the distal leaves during clinical deployment of the MLC system. The second picket fence delivery was defined with the distal leaves only (proximal leaves retracted) to independently evaluate the transmission for one layer. These two picket fence deliveries were repeated on ten separate days for stability evaluation. We then plotted the X and Y central axis profiles (averaged over the 10 days' measurements) to compare the transmission between the picket fence delivery using both MLC layers versus a single MLC layer only.

2.C | Leaf penumbra
The penumbras for three different field sizes (2 cm × 2 cm, 5 cm × 5 cm, and 10 cm × 10 cm) as shaped by the MLCs were measured on Halcyon TM using the single available energy -6 MV flatteningfilter-free (FFF) beam. Gafchromic EBT-XD self-developing films T A B L E 1 Comparison of the Varian Halcyon TM MLC system 17,18 to the widely used Varian Millennium TM 120-leaf MLC system. 19

2.E | Leaf edge effect
We investigated the leaf edge effects with the EPID using an MLC pattern that first extends the odd-numbered distal MLCs and then extends the even-numbered distal MLCs (Fig. 2). The EPID on Halcyon TM is a Digital Megavoltage Imaging panel permanently mounted facing the source at a source-to-imager distance of 154 cm. 17 For data acquisition, we used the "Portal Dosimetry" mode whereby the panel readout across the entire field was accumulated. The EPID was calibrated after TG-51 reference dosimetry (cross-check with optically stimulated luminescent detectors reported the ratio between absorbed dose determined by the Imaging and Radiation Oncology Core and our measurement to be 1.00), 16 and the EPID output measurements agreed with Farmer chamber measurements to within 0.45% in a long-term stability study. 24 To investigate the dependency on the 2 MLC banks, the aforementioned MLC pattern was delivered using leaves from Bank A only, and then leaves from Bank B only. On these two EPID images, the central profiles perpendicular to leaf travel were normalized to dose maximum and compared.
To determine whether the leaf edge effect would be affected by

2.F | Clinical impact
We compared point dose measurements to TPS dose predictions for 10 VMAT plans. Two plans were generated for each of the five treatment sites investigated (prostate, head-and-neck, brain, gynecology, and spine   Table 3 shows the leaf transmission for each layer independently, and for both layers. Specifically, note that no signal could be discernible from background noise when measuring the transmission through both layers.

3.B | Transmission
For picket fence deliveries using the distal-and-proximal-combination compared to distal-only, the inter-leaf and intra-leaf transmissions were lower (Fig. 4). The inter-leaf leakage peaks could be clearly identified at 1 cm increments for the distal-only profile but were not discernible for the distal-and-proximal-combination profile The Millennium TM 120-leaf MLC system has nominally < 2.5% average transmission and < 3% maximum transmission, and a transmission value of 1.36% has been reported. 19 Therefore, the Hal-

3.D | Leaf end effect
The measured leaf end effect was −0.19 mm from measurements at 10 cm depth, and −0.13 mm from measurements at d max (Fig. 6).
Suboptimal leaf end modeling reduces dose calculation accuracy and decreases the agreement between the predicted and measured dose.
The measured leaf end effect values closely matched the non-useradjustable value of 0.1 mm defined in the TPS for the MLCs. The difference in sign for the measured versus predicted leaf end effect suggested a difference in the overall dose direction when two leaf ends meet.
The leaf end effect presented was measured at the center of the field, but at off-axis distances, we postulate that the leaf end effect would be affected due to decreased number of photons and increased photon travel distance. Investigation of the spatial variation in the leaf end effect would be an interesting future avenue of study.

3.E | Leaf edge effect
The leaf edge profiles of Banks A and B were similar, with slightly larger dips between the leaf edges displayed by Bank A [ Fig. 7(a)].
The leaf edge effect showed no clearly discernible gantry position and collimator angle dependence [ Fig. 7 We observed a systematic discrepancy between EPID measurements and the TPS prediction using Eclipse AAA 15.

| CONCLUSION
We have comprehensively evaluated the performance characteristics of the stacked and staggered MLC system on Halcyon TM . On the whole, the MLC system is advantageous for this era of intensitymodulated treatments due to the high leaf positioning accuracy that is automatically monitored by the daily MPC, substantially low transmission even without jaws, and fast leaf speed enabling the use of high dose rates. Future work will investigate the TPS calculations and the correspondence to measured leaf end, leaf edge, and other dosimetric characteristics of the Halcyon TM MLC system.

ACKNOWLEDG MENTS
We would like to thank Samantha Lloyd, Xenia Fave, Todd Atwood, Kevin Moore, and Todd Pawlicki for their assistance in acquiring point dose measurements in phantom and the valuable discussions.
We would also like to acknowledge Mu Young Lee from Varian Medical Systems for providing some of the technical specifications of the Halcyon TM SX1 MLC system.

CONFLI CTS OF INTEREST
We have no relevant conflicts of interest to disclose.

R E F E R E N C E S
F I G . 8. The discrepancy between measured dose and dose predicted the treatment planning system (TPS) for ten test plans delivered in phantom on the Halcyon TM linear accelerator. Each circle represented the calculated discrepancies, with lines connecting the same plans to show the trend. Moving from an earlier to a later dose calculation algorithm -Anisotropic Analytical Algorithm (AAA) 15.1-15.6, the discrepancy between measurement and prediction was reduced from −3.46% ± 1.08% to −1.28% ± 0.80% (stated values are mean ± SD). In each box, the central line indicated the median, the top and bottom hinges indicated the 75th and 25th percentiles, and the whiskers extended to the most extreme value within 1.5× the interquartile range (difference between 75th and 25th percentile). 27