Stereotactic radiotherapy of appropriately selected meningiomas and metastatic brain tumor beds with gamma knife icon versus volumetric modulated arc therapy

Abstract Purpose To determine if the gamma knife icon (GKI) can provide superior stereotactic radiotherapy (SRT) dose distributions for appropriately selected meningioma and post‐resection brain tumor bed treatments to volumetric modulated arc therapy (VMAT). Materials and Methods Appropriately selected targets were not proximal to great vessels, did not have sensitive soft tissue including organs‐at‐risk (OARs) within the planning target volume (PTV), and did not have concave tumors containing excessive normal brain tissue. Four of fourteen candidate meningioma patients and six of six candidate patients with brain tumor cavities were considered for this treatment planning comparison study. PTVs were generated for GKI and VMAT by adding 1 mm and 3 mm margins, respectively, to the GTVs. Identical PTV V100%‐values were obtained for the GKI and VMAT plans for each patient. Meningioma and tumor bed prescription doses were 52.7–54.0 in 1.7–1.8 Gy fractions and 25 Gy in 5 Gy fractions, respectively. GKI dose rate was 3.735 Gy/min for 16 mm collimators. Results PTV radical dose homogeneity index was 3.03 ± 0.35 for GKI and 1.27 ± 0.19 for VMAT. Normal brain D 1%, D 5%, and D 10% were lower for GKI than VMAT by 45.8 ± 10.9%, 38.9 ± 11.5%, and 35.4 ± 16.5% respectively. All OARs considered received lower maximum doses for GKI than VMAT. GKI and VMAT treatment times for meningioma plans were 12.1 ± 4.13 min and 6.2 ± 0.32 min, respectively, and, for tumor cavities, were 18.1 ± 5.1 min and 11.0 ± 0.56 min, respectively. Conclusions Appropriately selected meningioma and brain tumor bed patients may benefit from GKI‐based SRT due to the decreased normal brain and OAR doses relative to VMAT enabled by smaller margins. Care must be taken in meningioma patient selection for SRT with the GKI, even if they are clinically appropriate for VMAT.


| INTRODUCTION
Stereotactic radiosurgery (SRS) using the Leksell Gamma Knife (Elekta AB, Stockholm, Sweden) is a highly accurate radiation delivery modality, with reported end-to-end accuracy of under 0.4 mm when the frame-based approach is used. 1 The Gamma Knife Icon (GKI), released in 2015, is capable of cone beam computed tomography (CBCT) guided frameless SRS, wherein the patient is immobilized in a face mask and tracked throughout treatment using an infrared camera. 2,3 In real time, the infrared camera tracks a reflective marker placed on the patient's nose.
The system has the capability to pause the radiation beam during treatment if the tracked motion is outside of a pre-defined threshold. 4 The accuracy of the CBCT-based GKI positioning system has been reported to be within 0.15 mm of the frame-based system, 2,5 enabling 0.5 mm end-to-end frameless initial positioning accuracy.
The frameless treatment capability of the GKI is well-tolerated by patients and enables fractionated stereotactic radiation therapy (SRT) with greater delivery accuracy than conventional SRT delivered with a linear accelerator. Linear accelerator-based SRT is typically delivered using intensity modulated radiation therapy (IMRT) or volumetric modulated arc therapy (VMAT). Several studies have been published comparing linear accelerator based SRT treatment plans to those of frameless Gamma Knife SRT using a Gamma Knife Perfexion system (pre-CBCT) and the Extend bite tray system for patient positioning [6][7][8] and also with the GKI. 9 In all of these studies, the same margin between gross tumor volume (GTV) and planning target volume (PTV) was used for Gamma Knife as for linear accelerator based SRT, which was reported to be 2-2.5 mm in three of the studies. 6,7,9 The general findings from these previous studies were that frameless Gamma Knife deliveries provided superior target dose conformity, lower brain dose, and poorer target dose homogeneity than for the linear accelerator-based approach. Given the accuracy of the CBCT positioning system used by the GKI and intrafraction monitoring throughout treatment, a 1 mm margin for GKI treatments may be justified. For linear accelerator-based SRT treatments at the authors' institution and elsewhere, a 3-mm margin is typically employed. 10,11 In this work, we conduct an evaluation of the potential benefits of frameless GKI SRT treatments versus linear accelerator based SRT delivered with VMAT using 1 mm and 3 mm GTV-to-PTV margins, respectively, for appropriately selected meningioma and post-resection brain tumor bed treatments. This is the first study in which the impact of margin reduction for GKI versus linear accelerator based SRT deliveries is accounted for in the treatment planning process at low dose-per-fraction values of 1.8 Gy per fraction. Such an approach to treating SRT patients is of interest to enable a reduction in normal brain tissue and organ-atrisk (OAR) doses. The approach is also of interest in a center like ours wherein a linear accelerator is replaced with a GKI to enable frame-based stereotactic radiosurgery, since patients who would be treated with fractionated SRT on the replaced linear accelerator could be instead treated on the GKI, reducing patient load on other linear accelerators at the center.

2.A | Patient selection
GKI dose heterogeneity was considered in determining candidacy for fractionated GKI therapies. Patients were eliminated from the study based on proximity to great vessels and/or OARs located adjacent to the GTV. Cases in which these risk factors are present may be more safely treated using volumetric modulated arc therapy (VMAT).
Where a more homogeneous dose distribution effectively constrains the maximum dose to the sensitive structures within the PTV to values less than or equal to the tolerance dose, which is often 54 Gy in 1.8 Gy fractions. 12 In two Institutional Review Board approved studies (IRB 20050306 and 201109821, J. M. Buatti, PI), the selection criteria were applied to 14 candidate patients with meningiomas and six candidate patients with post-operative brain tumor beds. Ten of the 14 patients with meningiomas were rejected for the treatment planning study due to proximity to great vessels, location relative to critical OARs such as the retina within the PTV, tumor location adjacent to sensitive soft tissue near the sagittal sinus, close tumor proximity to many OARs, and/or concave tumor contours that included excessive normal brain tissue. All six patients with post-operative brain tumor beds were accepted.

2.C | PTV generation and prescriptions
Two PTVs were created for each patient: one for the GKI and one for the VMAT treatment plans. Different margins were used for each system and applied uniformly to the GTV using a research version (5.19.0.d) of the Elekta Monaco treatment planning system (Elekta AB, Stockholm, Sweden). PTVs for VMAT plans were created by adding a uniform 3-mm margin to the GTVs. The 3-mm margin is used routinely at the authors' institution and other centers, 10,11 and representative of standard margins used for SRT treatments.
A major consideration for GKI planning that is not an issue for VMAT planning due to the physical differences in the paradigms is in determining where to place GKI "shots." Each shot corresponds to BUATTI ET AL. A critical component of GKI planning that is not a part of the VMAT planning in this work is defining where the isocenters can be placed within the patient prior to optimizing the remaining parameters associated with the shot(s) at each isocenter. To accomplish this for the GKI, a planning region at risk volume (PRV) for each OAR was generated by adding a uniform 1 mm margin about the OAR.
The PTV was generated by creating a GTV + 1 mm volume for each patient minus all PRVs, and isocenters were only allowed to be within the PTV. This prevented major hotspots from falling within the OARs due to the presence of isocenters within them. The PTV approach was not taken for the VMAT plans since our SRT patients are standardly treated without using the PRV approach since, due to the dose homogeneity achievable with VMAT in combination with the prescription dose levels ( Table 1). The PRV approach is unnecessary to achieve acceptable OAR doses and PTV coverage with VMAT.
Meningioma and brain tumor bed dose-volume (dose prescribed to V 100% , the percentage of the PTV receiving 100% of the prescribed dose or higher) prescriptions ranged from 52.7-54.0 Gy in 1.7 Gy/fx-8 Gy/fx 13 and 25 Gy in 5 Gy fractions, respectively, as shown in Table 1. The meningioma prescriptions matched those used clinically for the respective patients.

2.D | Treatment planning
VMAT treatment plans were created using the Pinnacle treatment planning system version 9.10 (Philips Medical Systems, Fitchburg, WI). The adaptive convolve dose calculation algorithm was used with a 2 mm × 2 mm × 2 mm dose grid. All plans were optimized for treatment on the Elekta Versa HD using 6 MV photon beams. Each plan used a 5 non-coplanar arc VMAT plan with a 15-degree collimator rotation for each arc. The beam geometries of the 5 fields were: (1) couch 340°/ gantry 0°-140°, (2) couch 305°/ gantry 140°-0°, (3) couch 20°/ gantry 0°-220°, (4) couch 55°/ gantry 220°-0°, (5) couch 90°/ gantry 0°-220°. The linear accelerator's isocenter was placed approximately at the geometric center of the PTV. The planning objective for each patient was to satisfy the prescription goals of Table 1 without exceeding the OAR tolerance doses. Maximum doses for the brain tumor bed PTVs were limited to under 115% of the prescribed dose (28.75 Gy). The optimization approach for the VMAT plans involved setting a minimum dose objective equal to the prescription dose with a weight of 40-60, a uniform dose objective for 10-20 cGy above the prescription dose with a weight of 80, and a maximum dose objective of 100 cGy above the prescription dose with a weight of 40-60. OAR maximum dose objectives with iteratively-determined weights that resulted in plans that met the goals (Table 1)

2.F | Treatment plan evaluation
GKI and VMAT plans were compared using the radical dose homogeneity index (rHI), maximum and mean OAR doses, and normal brain tissue D 1% , D 5% , and D 10% , where D x% is the minimum dose, in Gy, received by the hottest x% of the normal brain. The rHI metric is the ratio of the maximum to the minimum PTV doses. Normal brain tissue was defined as the brain minus the GTV, and was

| RESULTS
For all patients, the V 100% prescription goals were met for both the VMAT and GKI plans ( Table 2). The rHI, normal brain D 1% , D 5% , and D 10% values, and treatment times for GKI (as predicted by Gamma-Plan) are listed in Table 2. PTV dose homogeneity was 3.03 ± 0.35 for GKI and 1.27 ± 0.19 (mean ± standard deviation) for VMAT.
Over all meningioma and brain tumor cavity patients, rHI was 136 ± 27% (mean ± 1 standard deviation) greater in GKI plans than VMAT plans ( Table 2). As shown in Table 3, all OARs received a lower maximum dose from GKI than VMAT, and ten of the eleven OARs from the 4 meningioma cases received a lower mean dose from GKI than VMAT. Over all meningioma and brain tumor cavity patients, the normal brain D 1% , D 5% , and D 10% were 45.8 ± 10.9%, 38.9 ± 11.5%, and 35.4 ± 16.5% lower for GKI than for VMAT treatments. Dose volume histograms for a meningioma patient and a tumor cavity patient are shown in Fig. 1 and Fig. 2 (Table 2).
VMAT beam-on times are not provided by the Pinnacle treatment planning system, and, based on our oncology information system (Mosaiq), average beam-on times are 6.2 ± 0.32 min for 1.8 Gy fractions and 11.0 ± 0.56 min for 5 Gy fractions.

| DISCUSSION
The results of this work indicate that SRT delivered with a GKI can be dosimetrically superior to that delivered with VMAT for selected meningioma patients and brain tumor bed patients, due specifically to OAR dose and normal brain dose reduction through GTV-to-PTV margin reduction. The increased PTV dose heterogeneity of the GKI poses a danger for patients with great vessels in close proximity to the GTV, and such patients were ruled out as GKI candidates for this study. In properly selected patients, high-dose regions within the GTV owing to increased dose heterogeneity pose no expected increase in complication risk and may improve tumor control. In all GKI treatments, the average dose to the GTV is increased and the dose to the surrounding normal tissue is lower and/or below tolerance levels.
It has been shown in previous studies that frameless Gamma Knife deliveries can provide as good as or better PTV dose conformity to linear accelerator based intensity modulated radiation therapy. 6 Given the increased spatial accuracy of the GKI and the heterogeneity of GKI dose distributions, it was determined to be inappropriate to use the same GTV-to-PTV margins and the same OARprioritization approach for both GKI and VMAT. We found that generating the GKI PTVs that accounted for the increased accuracy of GKI (1 mm rather than 3 mm margins), and on which the same V 100% could be achieved as for VMAT, necessitated a PRV approach to avoid overdosing OARs adjacent to the GTV. This is because margins added to the GTV often create a volume that intersects with MG, meningioma; TB, brain tumor bed; PTV, planning target volume; V 100% , percentage volume receiving 100% of prescribed dose or higher; GKI, Gamma Knife Icon; VMAT, volumetric modulated arc therapy; D x% , minimum dose to hottest x% of volume; rHI = radical dose homogeneity index.
normal tissue and OARs in the surroundings of the GTV. For OARs that are closer than 1 mm to the GTV, it is possible that the associated PRV margin will overlap with the GTV. In that case, whether or not the patient should be excluded from consideration for GKI treatment would be subject to the clinical judgment of the physician. GKI treatments are normalized to a maximum dose point and prescribed to a percentage of the maximum dose, typically 50%. Due to GKI dose heterogeneity inherent to the modality and prescriptions being equal to or greater than OAR tolerance limits, OAR doses may easily exceed their tolerances if the PTV overlaps the OAR and is not specifically accounted for in the planning process. The planning technique used in the current work, in which GKI PTVs were generated by subtraction of PRVs from GTV + 1 mm volumes, is designed to avoid placing hot spots inside OARs and is indicative of the dramatically higher dose heterogeneity that the GKI delivers relative to VMAT.
In some GKI plans, the maximum normal brain dose between the GTV and PTV was increased substantially above that of VMAT due to normal brain concavities within the GTV that could not be spared without unacceptably lowering dose within the PTV. VMAT plans, with optimization constraints that enforce dose homogeneity (and thus reduce hot spots within the target) can avoid this issue, whereas the location of hot spots needs to be considered explicitly during the GKI planning process. However, due to the complexity of placing shots within the target for GKI, it may not always be possible to eliminate hot spots between the GTV and PTV completely. An example of this is shown in Fig. 3, a patient that was excluded from consideration for GKI, where a normal brain tissue concavity is partially surrounded by GTV (blue) within the PTV (red), resulting in a small volume of normal brain tissue receiving a high dose of 91 Gy to ensure the PTV dose-volume goal is achieved. The normal brain D 1% , D 5% , and D 10% for the patient shown in Fig. 3 were 30%, 87% and 170% higher, respectively, in the GKI plan than the VMAT plan, due to target contour concavity, hence the exclusion.
When treating with VMAT, the dose homogeneity achieved within the margin between the GTV and PTV can maintain the dose to the normal brain adjacent to the GTV at or below the GTV prescription dose, which is known from clinical experience to be acceptably tolerated by patients. The lack of dose homogeneity within the margin that is characteristic of the GKI, even with a reduced margin relative to the linear accelerator-based approach, can drive the hot spots above levels that may be clinically acceptable. This issue has been reported previously in the context of 5-fraction SRT. 8,14 In the current work 30-31 fraction SRT for meningiomas was also considered. It was found that a subset of those patients (4 of 14) would still be candidates for GKI-based SRT, and those patients would potentially benefit from substantially lower normal brain D 1% , D 5% ,

| CONCLUSION
Appropriately selected meningioma patients and brain tumor bed patients may benefit from GKI-based SRT due to the decreased normal brain and OAR doses relative to VMAT enabled by smaller margins. Due to dose the dose heterogeneity of the GKI and despite the potential for reduced OAR and normal brain sparing for some patients, care must be taken in meningioma patient selection for SRT with the GKI even if they are clinically appropriate for VMAT. F I G 2 . Example dose volume histograms for tumor cavity patient G showing (a) planning target volume (PTV) and (b) normal brain. There were no other organs at risk close enough to the PTV to be considered.

AUTHOR CONTRIBUTION STATEMENT
F I G 3 . GKI plan for a meningioma with a complex GTV (blue) contour with a concavity of normal brain that was included in the PTV (red), resulting in a hot spot within normal brain tissue that could not be removed without unacceptably compromising PTV V 100% . The yellow line is the 54 Gy isodose line.