Institutional experience with SRS VMAT planning for multiple cranial metastases

Abstract Background and Purpose This study summarizes the cranial stereotactic radiosurgery (SRS) volumetric modulated arc therapy (VMAT) procedure at our institution. Materials and Methods Volumetric modulated arc therapy plans were generated for 40 patients with 188 lesions (range 2–8, median 5) in Eclipse and treated on a TrueBeam STx. Limitations of the custom beam model outside the central 2.5 mm leaves necessitated more than one isocenter pending the spatial distribution of lesions. Two to nine arcs were used per isocenter. Conformity index (CI), gradient index (GI) and target dose heterogeneity index (HI) were determined for each lesion. Dose to critical structures and treatment times are reported. Results Lesion size ranged 0.05–17.74 cm3 (median 0.77 cm3), and total tumor volume per case ranged 1.09–26.95 cm3 (median 7.11 cm3). For each lesion, HI ranged 1.2–1.5 (median 1.3), CI ranged 1.0–2.9 (median 1.2), and GI ranged 2.5–8.4 (median 4.4). By correlating GI to PTV volume a predicted GI = 4/PTV0.2 was determined and implemented in a script in Eclipse and used for plan evaluation. Brain volume receiving 7 Gy (V 7 Gy) ranged 10–136 cm3 (median 42 cm3). Total treatment time ranged 24–138 min (median 61 min). Conclusions Volumetric modulated arc therapy provide plans with steep dose gradients around the targets and low dose to critical structures, and VMAT treatment is delivered in a shorter time than conventional methods using one isocenter per lesion. To further improve VMAT planning for multiple cranial metastases, better tools to shorten planning time are needed. The most significant improvement would come from better dose modeling in Eclipse, possibly by allowing for customizing the dynamic leaf gap (DLG) for a special SRS model and not limit to one DLG per energy per treatment machine and thereby remove the limitation on the Y‐jaw and allow planning with a single isocenter.


| INTRODUCTION
Volumetric modulated arc therapy (VMAT) has significantly changed the options for LINAC based cranial stereotactic radiosurgery (SRS) treatment of multiple metastatic brain lesions. Traditionally, LINAC based SRS utilizes one isocenter for each lesion, resulting in long treatment delivery times for patients with multiple metastases.
Recent publications have reported on VMAT planning for cranial SRS patients with multiple lesions using one isocenter and demonstrated that highly conformal dose distributions can be achieved. 1,2 Plan parameters have been compared to Gamma Knife plans which are considered the standard for cranial SRS treatment. 3,4 These studies have shown that VMAT plans can produce target coverage and dose fall-off in the high dose area similar to Gamma Knife plans.
There are still challenges related to the dose accuracy of VMAT delivery for small targets and the accuracy of the VMAT dose calculation algorithm must be validated prior to releasing SRS VMAT. 5 In addition, setup accuracy becomes much more critical when multiple targets at a distance from the isocenter are treated in the same plan. 6 At our institution, we have developed a SRS VMAT planning technique for treatment of multiple cranial metastases using similar contouring and optimization technique to those published by Clark et al. 2 at the University of Alabama. Due to limitations in our calculation algorithm to model both the 2.5 and 5 mm leaves on the TrueBeam STx we limit the plans to use only the 2.5 mm leaves, often resulting in two isocenters for cases with multiple brain metastasis. In this study, we present plan quality parameters and treatment times for 40 patients, treating a total of 188 lesions, with single fraction doses ranging from 16-21 Gy. The planning procedures, plan criteria, and quality assurance methods implemented at our institution are presented. (AAA) photon calculation model commissioned with the gold beam data in Eclipse did not provide a dose calculation accuracy that met our departmental electronic portal imaging device (EPID) dose gamma (c) score >95% with dose difference <3% and <2 mm distance-to-agreement. 5 Therefore, a specific cranial SRS AAA (SRS_AAA) model was developed in Eclipse with source size adjusted to meet the dose agreement criteria and the test plans were re-generated and evaluated. The 6MV AAA clinical model for a Varian TrueBeam STx was used as a baseline with focal spot (1.75 mm, 0.75 mm) and maximum field size 40 9 40 cm 2 . The dynamic leaf gap (DLG) was constrained to 1.24 mm to avoid affecting the dosimetry for the non-SRS treatments on the same machine. Maximum field sizes, output factors, focal spot and secondary source sizes were systematically adjusted to obtain an optimized model by comparing the calculated PDD's, profiles, and outputs with water tank measurements. The source size in the fine-tune model is (0, 0 mm). This cranial SRS_AAA model provided acceptable dosimetric agreement within the 2.5 mm leaf region, but areas with under-dose >10% were still observed for targets treated with the 5 mm leaves. 5 This may be a limitation of the AAA model which uses a single DLG to represent both the 2.5 and 5 mm leaves. Because of this inaccuracy, our clinical program restricts the field size of each arc in the SRS VMAT plans to the 2.5 mm leaf regions only. Consequently, one to three isocenters are required per plan depending on the spatial distribution of lesions.

2.A | Preclinical dosimetry
Institutional plan criteria were developed prior to the clinical release of SRS VMAT based on comparing the five pre-clinical SRS VMAT plans to the previously delivered plans developed in iPlan (RT Dose 4.5; BrainLab, Munich, Germany). The iPlan plans used one isocenter for each lesion and typically 10 static fields per isocenter.
At our institution, the target dose inhomogeneity criteria for these iPlan cases were 125%. Target inhomogeneity was allowed to increase to 140% for the Eclipse SRS VMAT plans to reduce the dose to normal brain. For cases where the PTV overlapped with the brainstem, based on internal experience a dose-volume limit to brainstem of V 18 Gy ≤ 10% was used to allow for full coverage of the target. and T1-weighted (3 mm) magnetic resonance images were fused to the CT images using MIM (version 6.6.3; MIM Software Inc., Cleveland, OH, USA) and auto-segmentation of normal structures was also generated in this systems.

2.B.2 | Treatment planning
The gross tumor volume (GTV) was contoured by the treating radiation oncologist who also reviewed and edited the critical structures (eyes, lenses, optic nerves, chiasm, brainstem, cord, and cochleas). A planning target volume (PTV) was created by a 3-dimensional 0-2 mm expansion around the GTV to account for imaging fusion BALLANGRUD ET AL. | 177 uncertainty, contouring variations, setup errors, and possible patient motion during treatment. The wall extraction tool in Eclipse was used to create similar shell structures for optimization as published by Clark et al. 2 The dimension of the shells depends on the PTV size. Separate shell structures were created for each group of targets with the same prescription dose. The planner also created a structure to evaluate the GI for each PTV.
Depending on the spatial distribution of the lesions one, two, or occasionally three isocenters were used. The planner created a union PTV for each group of lesions that were to be treated with the same isocenter and then placed the isocenter at the geometric center of the selected group of PTVs. The isocenter was then adjusted to ensure that Y1 and Y2 jaws were ≤4 cm so that only the 2.  Table 1. The plan was normalized by setting 100% dose to cover at least 98% of the PTV volumes.
Additional dose constraints were added for critical structures when needed. CI and target dose HI were used for plan evaluation:

2.B.4 | Treatment delivery
Treatments were delivered on a TrueBeam STx with a Perfect Pitch (Varian Medical Systems) robotic 6 degree-of-freedom (6DOF) couch.
The CDR table extension attached to the Perfect Pitch couch was used for initial adjustment of pitch and roll with guidance from the

3.A | Comparison of preclinical plans
Our VMAT planning procedures were developed to produce plans comparable to our clinical standard by analyzing five clinical cases T A B L E 1 Institutional plan criteria, where D max = maximum dose, D min = minimum dose, Rx = prescription dose, and V xGy = volume receiving x Gy.

Guideline Limit
Target criteria   Table 3. The dose to critical structures naturally depends on the proximity to the target lesions. Higher doses were accepted for cases where the lesions were close to the critical structures and these situations are apparent in Table 3. For example, one patient had a PTV overlapping with brainstem resulting in a brainstem D max of 24.9 Gy. For this scenario, PTV coverage was prioritized as long as brainstem V 18 Gy ≤ 10%. In another patient, an optic nerve D max of 11.0 Gy was accepted due to the PTV proximity to the nerve. If any dose criteria were exceeded, a peer review process was initiated.

3.B | Analysis of clinical VMAT plans
The target indexes GI, HI, and CI were collected for each PTV.
The CI is typically in the range 1.0-1.2. The maximum value of 2.9 occurred for a 0.05 cm 3 lesion in a 6-lesion plan. The median GI is 4.4. Figure 1 shows the GI for each lesion plotted versus PTV volume. Out of the 188 lesions, 13 lesions were so close that the 50% isodose was not split between the two lesions. These lesions were excluded from this GI analysis. For PTV sizes >0.5 cm 3 , a GI <5 is typically achievable whereas for smaller targets, GI exceeds 5.
The GI is reduced with increasing PTV size. Allowing the HI to increase facilitates a slight reduction in the GI for lesions ≥0.8 cm 3 as seen in Fig. 2. This trend was not observed for the smaller lesions.
Isodose distributions for a typical plan with eight lesions and two isocenters are shown in Fig. 3(a). For this case the total PTV volume The percent brain volume receiving 7 Gy or less is plotted in Fig. 4 as a function of the total tumor volume (sum of all PTVs) in each treatment plan. The plan goal is to keep brain V 7 Gy < 5% which is a criteria based on internal experience. For cases with a large tumor burden this criteria was not achieved.

Total beam-on time for each patient and beam-on time per
isocenter is listed in Table 4 along with the total treatment time for each patient measured from the first CBCT to completed delivery and the total treatment time per isocenter. Ten plans used one isocenters, 28 plans used two isocenters, and two plans used three isocenters. The setup time prior to the first CBCT is not included since this is not recorded in the record and verify system. By analyzing the AlignRT data we found that typically it takes 5 min from show that the GI is reduced with increasing target size so it is therefore expected that our reported GI is higher than in Clarks study where the lesion sizes were larger.

Ma et al. published a planning study comparing plans for Gamma
Knife, Cyber Knife, Novalis, and TrueBeam FFF for cases with 3, 6, 9, and 12 lesions, concluding that the volumes of brain receiving low to moderate dose (4-12 Gy) were higher and increased more rapidly with additional targets for LINAC-based SRS than for Gamma Knife. 10 In this study, all lesions were smaller than 1 cm 3 , whereas in our study the targets ranged from 0.05-17.74 cm 3 . Therefore, the brain volume receiving 12 Gy is higher in our study (range 4-57 cm 3 , median 19 cm 3 ) than the TrueBeam results in their study (increasing from 5.5 to 29.6 cm 3 as the lesion number increase from 3 to 12).
From the preclinical VMAT planning we found that when increasing the HI while keeping the CI the same, the GI was reduced (data not shown). Other groups have reported reduction in GI with increasing HI. 11,12 GI as a function of the dose HI for all PTVs ≥0.8 cm 3 is shown in Fig. 2. There is no strong correlation between GI and HI from these clinical plans but there may have been different considerations or challenges in the plans that impacted the final HI and GI for each lesion. We decided to allow for a PTV D max of 140%. This is higher than for the iPlan plans that have been our clinical standard prior to introducing the VMAT technique but since Gamma Knife plans typically use a PTV D max >140% we accepted this higher inhomogeneity to reduce the GI. The risk of radiation Total PTV volume [cm 3 ] F I G . 4. Percent brain volume receiving more than 7 Gy as a function of the total PTV volume (sum of all PTVs for each patient). The plan goal is to achieve brain V 7 Gy < 5%.
T A B L E 4 Beam-on time for each patient and per isocenter, and total treatment time for each patient and per isocenter, for the 40 plans. Median treatment time per isocenter is 31 min. Ten plans used 1 isocenters, 28 plans used 2 isocenters, and 2 plans used 3 isocenters. reports the need to adjust source size for small field intracranial SRS using AcurosXB in the Eclipse planning system to avoid >10% central axis dose discrepancies for small target volumes. 20 When treating multiple targets distant from the isocenter, extra requirements are needed for setup accuracy and limitation of patient motion during treatment. To remove any patient rotation at setup, a 6DOF couch should be used for treatment. The positioning accuracy is determined by the CBCT-MV isocenter congruence, the amount of couch walk and the accuracy of the image registration at the machine. Motion monitoring with conventional LINAC on-board imaging is challenging due to limitations imposed by the routine use of couch rotations and gantry rotation for these patients. Using the optical surface image system AlignRT we found that few patients move during treatment in our frameless immobilization system. Most of the patients that moved (eight out of ten) had moved <1.1 mm but there were a couple of patients that moved significantly and this highlights the importance of using a motion monitoring system to catch these outliers. An alternative motion monitoring system frequently used for cranial SRS is ExacTrac (Brainlab, Munich, Germany).

| CONCLUSION
We have developed a procedure to treat multiple cranial metastases with VMAT achieving similar plan quality to traditional 3D LINAC based SRS plans. Caution must be taken to assure that the dose calculation model is accurate for very small lesions and for both MLC types on a TrueBeam STx. For cases with three or more lesions treatment time is significantly reduced by using VMAT plans and one or two isocenters. Currently it is very elaborate to create these treatment plans due to limitations in the dose modeling, and also due to contouring and optimization. Better dose modeling in Eclipse would remove limitations on the Y-jaw and thereby reduce planning and treatment time significantly.

ACKNOWLEDG MENTS
This research was funded in part through the NIH/NCI Cancer Center Support Grant P30 CA008748.

CONFLI CT OF INTEREST
There is no conflict of interest related to this study.