Optimization of the dosimetric leaf gap for use in planning VMAT treatments of spine SABR cases

Abstract The dosimetric leaf gap (DLG) is a beam configuration parameter used in the Varian Eclipse treatment planning system, to model the effects of rounded MLC leaf ends. Measuring the DLG using the conventional sliding‐slit technique has been shown to be produce questionable results for some volumetric modulated arc therapy (VMAT) treatments. This study therefore investigated the use of radiochromic film measurements to optimize the DLG specifically for the purpose of producing accurate VMAT plans using a flattening‐filter‐free (FFF) beam, for use in treating vertebral targets using a stereotactic (SABR, also known as SBRT) fractionation schedule. Four test treatments were planned using a VMAT technique, to deliver a prescription of 24 Gy in 3 fractions to four different spine SABR treatment sites. Measurements of the doses delivered by these treatments were acquired using an ionization chamber and radiographic film. These measurements were compared with the doses calculated by the treatment planning system using a range of DLG values, including a DLG identified using the conventional sliding‐slit method (1.1 mm). An optimal DLG value was identified, as the value that produced the closest agreement between the planned and measured doses (1.9 mm). The accuracy of the dose calculations produced using the optimized DLG value was verified using additional radiochromic film measurements in a heterogeneous phantom. This study provided a specific initial DLG (1.9 mm) as well as a film‐based optimization method, which may be used by radiotherapy centers when attempting to commission or improve an FFF VMAT‐based SABR treatment programme.

tic (SABR, also known as SBRT) fractionation schedule. Four test treatments were planned using a VMAT technique, to deliver a prescription of 24 Gy in 3 fractions to four different spine SABR treatment sites. Measurements of the doses delivered by these treatments were acquired using an ionization chamber and radiographic film.
These measurements were compared with the doses calculated by the treatment planning system using a range of DLG values, including a DLG identified using the conventional sliding-slit method (1.1 mm). An optimal DLG value was identified, as the value that produced the closest agreement between the planned and measured doses (1.9 mm). The accuracy of the dose calculations produced using the optimized DLG value was verified using additional radiochromic film measurements in a heterogeneous phantom. This study provided a specific initial DLG (1.9 mm) as well as a film-based optimization method, which may be used by radiotherapy centers when attempting to commission or improve an FFF VMAT-based SABR treatment programme.

| INTRODUCTION
Stereotactic ablative radiosurgery (SABR, also known as stereotactic body radiotherapy or SBRT) has been shown to be effective for treating tumors in and around the vertebra. 1,2 These "spine SABR" treatments require the use of a small number of treatment fractions (typically 1-4) to deliver a relatively high dose of radiation (typically 12-27 Gy). 3 In order to minimize the time taken to deliver these high-dose fractions, especially for patients who may be suffering pain and discomfort due to vertebral metastases, treatments can be The Spine SABR target volumes are generally irregular in shape due to the location and geometry of the targeted vertebra as well as the importance of sparing the spinal cord, which abuts or penetrates the target volume. Treatment planning using inversely optimized volumetric modulated arc therapy (VMAT) techniques, can result in very steep dose gradients (greater than 12%/mm 4 ) between the targeted vertebra and the spinal cord. To produce the complex dose distributions required to achieve that spinal cord sparing while adequately treating the targeted vertebra, VMAT uses moving multileaf collimators (MLCs), with simultaneously varying dose rates and gantry speeds. 5 These complex, dynamic systems present numerous opportunities for dose uncertainties.
The AAPM Task Group-101 report highlighted the accuracy required in treatment planning for SABR treatments, 6 and recommended rigorous testing of the TPS dose calculation accuracy including end-to-end tests. Accurate calculations of dose and dose gradients are especially important for treatments where ablative doses of radiation are delivered to targets in close proximity to critical structures, such as spine SABR treatments. Dose calculation accuracy is known to be detrimentally affected by the use of suboptimal beam configuration data in the radiotherapy treatment planning system (TPS) 7,8 and by the inappropriate handling of simplifications in the TPS model. [9][10][11] For example, Varian Eclipse TPS (Varian Medical Systems, Palo Alto, USA) simplifies the modelling of the physical geometry of the MLC leaves, omitting physical characteristics such as the rounded leaf-ends.
To overcome this, Eclipse allows the user to define a specific parameter, the dosimetric leaf gap (DLG), which defines the difference between the physical round leaf end and the straight edge model of the TPS. 9 The value of the DLG is applied when calculating dose for modulated radiotherapy (including VMAT) treatment plans, as a retraction between the planned and calculated MLC positions. The DLG parameter is one of a few values that needs to be modified by the user when configuring the Varian Eclipse anisotropic analytical algorithm (AAA). 10 Measurement of the DLG is performed by the sliding-slit test (as described in Varian Medical Systems' documentation 10 ). This method produces a single DLG value per energy, which is applied in the Eclipse TPS to all leaf pairs irrespective of MLC leaf width. 9 While some studies have identified good agreement between planned and measured doses when using the DLG value measured using the standard slidingslit test, 12,13 other authors have identified substantial discrepancies. 11,14 For example, Szpala et al. 11 and Kielar et al. 14 elected to optimize the DLG value using clinical VMAT plans after they observed that the DLG value measured using the sliding slit test produced unreliable results when used to calculate clinical VMAT plans. These authors recommended careful testing for dosimetric accuracy for irradiating small targets, especially those used for SABR.
Similarly, both Szpala et al. 11 and Kumaraswamy et al. 9 found that the single DLG value used in Eclipse should be considered an estimate only; the optimal DLG for each MLC leaf varies with the distance from the central-axis and with the position of the opposite leaf. Due to the differences between the field sizes and complexity of MLC motion required when treating different anatomical sites, 15 the DLG can be expected to vary with anatomical treatment site and treatment modality.
Previous examinations of the Varian DLG have focused on treatments with standard (nonstereotactic) fractionation, planned for the brain, 9,11,14 prostate, 9,12 head and neck, 9,12 and AAPM Task Group 119 standard volumes (average prostate and simplified spine). 13,14 Some of these studies have suggested that the DLG values that are required to accurately calculate dose for FFF modalities are especially different from the DLG values that are obtained using the sliding slit method. 14,16 It is therefore important to specifically evaluate and optimize the DLG that is used when calculating dose for hypofractionated SABR treatments that use FFF VMAT beams.
This study therefore demonstrates the use of radiochromic film measurements to investigate the optimal DLG for use when treating spine SABR cases using a VMAT technique, with an FFF beam, in order to provide a specific initial DLG as well as a film-based optimization method, which may be used by radiotherapy centers when attempting to commission or improve an FFF VMAT-based SABR treatment programme.

2.A | Test treatment plans
The prescription used for the clinical test spine SABR treatment plans was 24 Gy, to be delivered in 3 fractions of 8 Gy. This prescription was selected with reference to literature 16

2.C | DLG verification: Heterogeneous phantom
The suitability of the optimized DLG value was evaluated in an inhomogeneous phantom, the IMRT Thorax phantom (CIRS Inc, Norfolk, USA), using a fine (1 mm) dose calculation grid resolution. Only two DLG values were used when calculating the Spine SABR plans on the IMRT Thorax phantomthe initial 1.1 mm and the optimal 1.9 mm value. As these measurements in the transverse plane were used to evaluate the sparing of the spinal cord region as well as the accurate treatment of the planned high-dose (vertebral) region, both arcs from each treatment were delivered to each piece of film. This represents a single fraction treatment dose.

3.A | DLG optimization: Homogeneous phantom
The DLG value for the 6 MV FFF beam with Millenium-120 MLC was found to be 1.1 mm using the sliding slit method, as shown in Fig. 1.
Using the DLG value identified using the sliding-slit method  plans. From these results, the optimal DLG from the film measurements for the FFF Spine SABR is in the range 1.9-2.1 mm. The chamber measurements are shown in Fig. 3(b)the optimal DLG is in the range 1.6-1.9 mm.
A value of 1.9 mm was therefore selected as the optimal DLG for use when planning FFF VMAT spine SABR treatments.  The Eclipse AAA beam model used in this study was commissioned using data for field sizes ranging from 3 9 3 cm 2 to 40 9 40 cm 2 . 10 Although data for smaller field sizes is usually measured during linac commissioning, it is not required for commissioning of the beam model within Eclipse. 12 The DLG is used in the Varian Eclipse treatment planning system as an approximation factor to reduce the dosimetric calculation uncertainty arising from the use of a simple MLC model with straight leaf ends. Conventionally, the DLG is measured using vendor-supplied DICOM plans that produce a sliding-slit with 13 control points,

3.B | DLG verification: Heterogeneous phantom
where the MLC leaves move at the same speed, in one direction, with a constant dose rate. 10 This broadly approximates an IMRT delivery, where the MLC leaves move in the same direction, from one side of the field to the other, albeit at different speeds.
By contrast, VMAT treatment deliveries are much more complex.
Each VMAT arc typically uses 178 controls points, with MLC leaves undergoing frequent changes in direction. Adjacent MLC leaves may move in opposite directions to each other and at different speeds.
The dose rate is also modulated and defined for each control point.
A single point measurement using the sliding slit method does not replicate the complex MLC movements such as those in a VMAT treatment for a spine SABR case.
It is therefore unsurprising that determination of the appropriate DLG value for clinical use in planning VMAT treatments should require the use of more complex plans than the sliding-slit, evaluated using more sophisticated measurements than a point dose.
T A B L E 2 Gamma agreement indices (percentage of points passing a gamma evaluation using 3%, 1.5 mm criteria) resulting from comparing the dose measured using film in a transverse plane through the heterogeneous (thorax) phantom against the dose calculated in the same plane using the treatment planning system with the sliding-slit-based DLG (1.1 mm) and the optimization-based DLG (1.9 mm).  Optimization of the DLG for VMAT treatments should involve the use of treatment plans that are representative of intended clinical use of the beam model, with measurements completed using accurate, high-resolution two-dimensional dosimeters. 9,11,13,14 In this study, radiochromic film was shown to produce results that were sufficiently sensitive to DLG variation for use in DLG optimization, although verification using a second dosimeter (such as an ionization chamber) may be advisable (see Fig. 3). The radiochromic film used in this study also provided accurate, high-resolution measurements that allowed the suitability of the optimized DLG value to be verified, when dose was calculated at a high resolution and the test treatments were delivered to a heterogeneous phantom (see Fig. 4). Estimated measurement uncertainties affecting the use of radiochromic film for radiotherapy dosimetry range from 0.55% 21 to 4%. 22 It is therefore important to independently evaluate uncertainties when commissioning any radiochromic film dosimetry system that is used to optimize beam configuration values, such as the DLG. The results shown in Fig. 3 confirm the importance of optimizing the DLG using a range of clinically likely test treatments. For this study, the test treatment volumes were thoracic and lumbar vertebral bodies, with and without left and right pedicles, and the corresponding treatments were designed with a range of different field sizes and collimator angles. Figure 3 shows that the particular values of the DLG that gave the closest agreement between the planned and measured doses differed between plans and between measurement techniques. We have adopted the optimal DLG of 1.9 mm for the 6FFF beam model for use in our clinic, for treatment of spine SABR cases. We have not yet investigated the application of this optimal DLG to SABR planning for other anatomical sites. The identification of a DLG value that is optimal for an entire class of plans (for a specific treatment modality, used to treat a specific anatomical site) evidently requires the use of different examples of the specific anatomical site to be treated.

| CONCLUSIONS
This study used an evaluation of DLG suitability for four spine SABR test treatment plans to confirm that the DLG identified using the conventional sliding-slit method does not produce clinical treatment plans that show good agreement between planned and measured doses for VMAT treatments delivered using a FFF beam.
Based on the results of this study, the following general recommendations can be made, for optimizing the DLG for use in planning spine SABR (or any other) VMAT treatments:

CONFLI CT OF INTEREST
The authors have no conflicts of interest to disclose.