Comparison of VMAT complexity‐reduction strategies for single‐target cranial radiosurgery with the Eclipse treatment planning system

Abstract Complexity in MLC‐based radiosurgery treatment delivery can be characterized by the efficiency of monitor unit (MU) utilization and the average MLC leaf separation distance for a treatment plan. A reduction in plan complexity may be desirable if plan quality is not impacted. In this study, a number of strategies are explored to determine how plan quality is affected by efforts to reduce plan complexity. Ten radiosurgery cases of varying complexity are retrospectively planned using six optimization strategies: an unconstrained volumetric modulated arc therapy (VMAT) technique, a MU‐constrained VMAT technique, three techniques using various strengths of the aperture shape controller (ASC), and a hybrid technique consisting of a final‐stage VMAT optimization applied to a dynamic conformal arc leaf sequence (ODCA). The plans are compared in terms of MU efficiency, MLC leaf‐separation, conformity index (CI), gradient index (GI), and QA measurement results. The five VMAT techniques exhibited only minor differences in CI and GI values, though the ASC and MU‐constrained techniques did require 6–20% fewer MU and had mean field apertures 5–19% larger. On average, the ODCA technique had CI values 3.5% lower and GI values 1.0–2.5% higher than the VMAT techniques, but also had a mean field aperture 24–47% larger and required 16–32% fewer MU. The QA measurement results showed a 0.61% variation in mean per‐field 2%/1 mm gamma passing rates across all techniques (range 96.81%–97.42%), with no observed correlation between passing rate and technique. For simple targets, the ODCA technique achieved CI results that were equivalent to the unconstrained VMAT technique with an average 30% reduction in required MU, an average 50% increase in mean leaf separation distance, and brain V12Gy values within 0.38 cc of the VMAT technique for targets up to approximately 2 cm diameter. For MLC‐based single‐target radiosurgery, plan complexity can often be significantly reduced without an equivalent reduction in plan quality.


| INTRODUCTION
Stereotactic radiosurgery is a widely available technique for the treatment of intracranial tumors and functional abnormalities with a long history of development and refinement. 1 Radiosurgery utilizes beams which are tightly collimated to the target volume in order to deliver ablative doses in a single fraction or a small number of fractions. Modern radiosurgery may be delivered using a variety of technologies, one of the most recent of which is multileaf collimatorbased radiosurgery using mechanically precise linear accelerators that can achieve overall non-coplanar isocentric accuracy within 0.6 mm. [2][3][4][5] MLC-based radiosurgery delivery with a linear accelerator can be accomplished using a number of techniques. 6 Static treatment fields with stationary apertures focused on the target from multiple directions utilizing combinations of couch, collimator, and gantry rotation would constitute a basic three-dimensional (3D) planning and delivery technique. Conformal arc techniques where the gantry rotates during beam delivery are also common. Arcs are generally delivered at multiple couch angles and may utilize either static per-field MLC apertures or dynamic apertures that conform to the target shape during rotation.
Volumetric modulated arc therapy (VMAT) is a well-established treatment delivery technique which utilizes simultaneous modulation of the MLC aperture, dose rate, and gantry rotation speed to generate highly conformal dose distributions. 7,8 For radiosurgery applications, multiple VMAT treatment fields are typically utilized with three or more couch angles to achieve optimal dose conformity and rapid dose falloff at the target edge.
The delivery complexity of VMAT treatment plans is wide-ranging, and significant efforts have been made to quantify this complexity and investigate how it correlates to deliverability and results of pre-treatment patient-specific quality assurance measurements. [9][10][11][12][13][14][15][16][17] Reduction of delivery complexity for VMAT radiosurgery may be desirable if comparable plan quality can be achieved. In this study, plan complexity refers to the efficiency of monitor unit (MU) utilization and the complexity of MLC leaf sequences utilized with the treatment fields, for which MLC aperture size and leaf separation distances are used as surrogates.
Less complex MLC leaf sequences using larger average aperture sizes may result in better dosimetric agreement between the plan and delivery because of greater similarity between the plan parameters and modeling data, though such an approach must be carefully evaluated to ensure that it does not compromise the sharpness of the dose gradient at the target edge. Pretreatment verification measurements for VMAT radiosurgery plans may be challenging if the treatment fields are comprised of very small highly modulated beamlets, as detector spacing in typical two-dimensional (2D) or 3D arrays may not be suitable for such measurements. Radiochromic film can be utilized for high-resolution spatial measurements at the cost of time efficiency, but accurate film dosimetry also has equipment and workflow considerations that must be carefully considered, [18][19][20][21] and these may include separate point measurements of absolute dose which requires an appropriately sized small-field dosimeter. Verification measurements using onboard electronic portal imaging devices have significantly improved resolution compared with most independent 2D or 3D arrays as well as high time efficiency, but they do not measure physical dose and do not constitute a "True Composite" verification technique, defined in AAPM TG-218 as the simulation of treatment delivery to a stationary measurement device placed on the treatment couch using the actual treatment delivery parameters. 22 Reduction in the plan MUs also has the benefit of reducing scatter and leakage dose to the patient as well as reducing treatment delivery time, although modern flattening filter-free photon modes support higher dose rates that can reduce this time-saving benefit.
Treatment apertures which are on average more open may also benefit from a reduced contribution of in-field/ edge-of-field interleaf and leaf-abutment MLC leakage to the total field fluence, components of the beam model which are generally less robust than for the open field.
In this study, a selection of clinical single-target radiosurgery cases were planned using various VMAT techniques designed to reduce the delivery complexity. The resultant plans were compared in the context of leaf sequence complexity and MU efficiency as well as conformity and gradient indices commonly used in the evaluation of radiosurgery plans. Per-field and per-plan quality assurance measurements and analysis was performed for a subset of the treatment plans generated in this study for the purpose of evaluating whether or not patient-specific quality assurance passing rates were correlated with these delivery complexity metrics. Finally, the goal was to use this data to identify situations, if any, where a standard VMAT optimization technique produces unnecessarily complex treatment plans compared with dosimetrically appropriate alternatives.

2.A | Case selection
Ten initial clinical cases were selected for retrospective planning and evaluation, with those cases being representative of the type and complexity treated in our institution using MLC-based techniques. In total, there were six intact brain metastasis cases, two postoperative resection cavity cases, and two acoustic neuroma cases. Following an initial analysis of the results, an additional ten cases comprised of intact brain metastasis, simple in shape with no adjacent critical structures, were selected for a targeted analysis between two of the studied techniques. The intent of this targeted analysis was to more robustly investigate a specific clinical scenario that the initial analysis suggested would be most appropriate for a specific complexity reduction strategy. A summary of the cases is provided in Table 1. VMAT_MU: MR1 VMAT optimization plus an upper total MU limit with maximum strength (100) and a value set to 10% above the total MUs needed in the ODCA plan. This specific MU objective was not extensively studied or determined to necessarily be an optimal value, but was selected based on institutional experience as an objective that would aggressively reduce the plan MU compared to a standard VMAT optimization.

2.B | Plan optimization methods
VMAT_VL: MR1 VMAT optimization with no MU limit but Aperture Shape Controller set to Very Low VMAT_MOD: MR1 VMAT optimization with no MU limit but Aperture Shape Controller set to Moderate VMAT_VH: MR1 VMAT optimization with no MU limit but Aperture Shape Controller set to Very High The aperture shape controller (ASC) option is a component of the VMAT leaf sequencer within the photon optimizer (PO) algorithm that penalizes disconnected apertures created by adjacent leaf pairs. The magnitude of the penalty can be adjusted in five discrete steps by adjusting the ASC setting within the Very Low to Very High range, and the higher the ASC setting the more the optimizer will be pushed to join adjacent apertures, theoretically also reducing delivery complexity.
Additionally, a standard non-optimized dynamic conformal arc treatment plan was generated using a 1mm aperture margin for each case. The set of DCA plans serves as a 3D reference technique and represents a lower limit on delivery complexity for a dynamic treatment plan. In particular, it was anticipated that the DCA technique would have the most efficient MU utilization and the largest overall field apertures while having the least conformal dose distribution, and can therefore serve as a baseline for evaluating how other techniques balance increasing complexity with increasing dosimetric quality. The authors acknowledge that for some of the more challenging clinical cases included in this study, the unmodified DCA plan may not be clinically appropriate due to deficiencies in either dose conformity or adjacent critical structure dose.
Each technique within a given treatment plan utilized an identical set of treatment fields, with all plans utilizing a noncoplanar beam arrangement consisting of either a full or partial arc at the 0°table position along with 2-4 partial arcs at additional table positions selected for optimal field spacing and critical structure avoidance. Figure 1 shows typical 2D fluence patterns for the same representative treatment field generated as a standard non-optimized DCA field, an optimized dynamic conformal arc (ODCA) field, and an unconstrained VMAT field. The standard DCA field shows a nearly homogeneous central fluence with only enough MLC movement to shape the aperture to the target volume during gantry rotation, while the ODCA field introduces a small amount of modulation largely near the field edge. The VMAT technique has a significantly higher degree of aperture modulation both centrally and at the field edge as well as a slight but visibly increased level of in-field MLC leakage.

2.C | Dose calculation and optimization parameters
All optimized cases were planned using consistent calculation parameters. The ACUROS v15.5.11 algorithm was utilized with a 1 mm dose calculation grid covering the entirety of the 1 mm slice T A B L E 1 Clinical cases utilized in this study. Note "a" indicates target volume is within 5 mm of an adjacent organ at risk or region of previous treatment; note "b" indicates target volume is within 10 mm of an adjacent organ at risk or region of previous treatment; "c" indicates all organs at risk are separated from the target volume by >10 mm distance. Cases 01-10 were used in the initial analysis, while 11-20 were used in the targeted analysis.

2.D.2 | Monitor unit efficiency
Efficiency of MU utilization was compared by tabulating an MU efficiency ratio for each case and planning technique, defined as the ratio of the total fraction MU and prescribed fraction dose.

2.D.3 | Conformity index
Conformity Index (CI) was evaluated using the definition proposed by Paddick et al 23 : TV is the volume of the target structure, PIV is the volume of the prescription dose value, and TV PIV is the volume of the overlap between the prescription dose value and the target structure. CI values range from 0 (indicating no overlap of the target volume and prescribed dose volume) to 1 (indicating perfect overlap), with values closest to 1 being the most desirable.

2.D.4 | Gradient index
GI was evaluated using the definition proposed by Paddick et al 24 : V 50% and V 100% refer to the volume of the dose clouds created from the 50% and 100% prescription dose values. Radiosurgery planning strongly emphasizes conformity of the prescription dose to the target structure, so lower GI values indicate a shorter falloff distance to the 50% isodose line from the target edge, which is desirable for Fluence patterns for the same treatment field generated as a (a) standard dynamic conformal arc field, (b) optimized dynamic conformal arc (ODCA) field, and (c) volumetric modulated arc therapy field. Tick marks on graticule are spaced at 1 cm.
maximizing the sparing of functional brain tissue surrounding the target structure.

2.E | Quality assurance measurements
Quality assurance measurements and analysis was performed for the initial set of ten clinical cases using the Varian Portal Dosimetry software, with portal dose prediction generated from the PDIP 15.5 algorithm and portal dose images acquired on an aS-1000 MV detector at 100 cm source-imager distance. Immediately prior to each session of quality assurance measurements, the full set of calibrations were performed for the 6 MV "Dosimetry" imaging mode (dark field, flood field, pixel map correction, and dose calibration). All treatment plans were delivered as planned with no modifications other than the removal of couch rotations, which has no impact on the Portal Dosimetry workflow. Gamma analysis was performed per-field using 2% dose difference and 1mm distance-to-agreement input parameters with a 10% low-dose threshold.
The PDIP algorithm was previously commissioned within our institution specifically for radiosurgery applications as an efficient and high-resolution tool for verifying patient treatment plans in con-

3.A.2 | Monitor unit efficiency
Technique comparison results of MU efficiency are presented in

3.A.3 | Paddick conformity index
Paddick CI results are presented in Table 4, with optimized plan results globally color-scaled from red to green, corresponding to 0.75 or lower (red) and a maximum value of 1.0 (green

3.A.4 | Gradient index
GI results are presented in Table 5, with optimized plan results globally color-scaled from green to red, corresponding to 3 or lower (green) and 5 or higher (red). These values correspond to the ideal value (3 or lower) and upper limit value (5)

3.A.5 | Quality assurance measurements results
Mean per-field gamma passing rates are presented in Table 6, with optimized plan results globally color-scaled from red to green, corresponding to 93% or lower (red) and a maximum value of 100% (green).
Results for the standard DCA technique are presented in grey as a 3D technique reference. Additionally, individual gamma passing rates for each of the 228 optimized treatment fields utilized in this study are plotted against mean leaf separation distance in Fig. 2.
Statistical analysis using a paired t-test at the 95% confidence level revealed no statistically significant differences in gamma passing rates between any two techniques in the ODCA and VMAT categories (P range 0.208-0.496). The increase in gamma passing rates for the non-optimized 3D DCA plans was statistically significant compared to all other techniques (P range 0.000-0.009).
T A B L E 2 Average leaf separation distances for all techniques and cases, normalized to the optimized dynamic conformal arc (ODCA) technique. Red-green color scale is case-specific (rows), with green referring to the largest value and red to the smallest. Multiplying any given result in a row by the Factor in the last column will restore the non-normalized separation in millimeters. DCA data are presented in grey as a reference 3D technique. T A B L E 6 Mean per-field gamma passing rates for all techniques and cases, non-normalized. Red-green color scale is global, with red assigned to a value of 93% or lower and green assigned to a maximum value of 100%. DCA data are presented in grey as a reference three-dimensional technique. Radiosurgery cases with mostly spherical targets may be appropriately planned using the non-optimized DCA technique, but irregularities in target shape or limitations in beam arrangement can result in the need for some manual intervention in beam apertures, sometimes at the per control point level. The ODCA technique is an efficient alternative to these manual adjustments, resulting in improved overall dosimetric quality with only a modest increase in delivery complexity.

Mean Per-Field Gamma
The quality assurance measurement results presented in Table 6 and Fig. 2

4.B | Case discussions and targeted analysis
This section discusses a small number of specific cases and associated observations as well as discussion of the targeted analysis between the ODCA and VMAT techniques.  respectively. This is an example of a case where, despite the complexity of shape in the target volume, complexity of the delivery parameters can be reduced with essentially no reduction in overall plan quality.

| CONCLUSION S
Optimization strategies which aim to reduce the delivery complexity of single-target radiosurgery treatment plans are largely able to increase mean field aperture dimensions and reduce required MUs while preserving the overall dosimetric quality of the plan. However, certain highly complex target volumes may still benefit from less constrained VMAT optimization approaches in order to achieve optimal dose conformity. Of the VMAT-based strategies explored here, a simple MU-constrained optimization appears to best balance delivery complexity and plan quality.
Utilizing a hybrid approach of performing final-stage VMAT optimization on a simple dynamic conformal arc MLC leaf sequence (ODCA) can significantly reduce the required number of MUs and complexity of the MLC sequence compared to VMAT techniques while improving conformity of dose compared to a non-optimized dynamic conformal arc plan. Although typically less conformal compared to plans generated with a full VMAT optimization, the difference is not always significant and the reduction in complexity may be favorable, especially for simple ellipsoidal tumor volumes.
High-resolution quality assurance measurements of each of the 228 treatment fields used across 60 optimized treatment plans did not demonstrate any meaningful correlation between optimization technique or mean aperture size with 2%/1 mm gamma passing rates, indicating that any motivation for reducing complexity in radiosurgery treatment plans should be justified independently from gamma passing rates alone.
Plan quality and plan complexity exist together on a spectrum, but this study demonstrates that complexity can often be significantly reduced without an equivalent reduction in plan quality. Institutions can utilize these and other planning strategies to find the optimal balance for their own clinical cases.

CONFLI CT OF INTEREST
No conflict of interest.