Radiosurgery for mesial temporal lobe epilepsy following ROSE trial guidelines — A planning comparison between Gamma Knife, Eclipse, and Brainlab

Abstract Purpose This study aims to compare stereotactic radiosurgery (SRS) planning of epilepsy that complies with Radiosurgery or Open Surgery for Epilepsy (ROSE) guidelines in GammaKnife, non‐coplanar conformal (NCC) plan in Eclipse, dynamic conformal arc (DCA) plan in Brainlab, and a volumetric modulated arc therapy (VMAT) plan in Eclipse. Methods Twenty plans targeting Mesial temporal lobe epilepsy (MTLE) was generated using GammaKnife, Eclipse with 20 NCC beams, Brainlab with 5 DCA, and Eclipse VMAT with 4 arcs observing ROSE trial guidelines. Multivariate analysis of variance and Wilcoxon signed‐rank test were used to compare dosimetric data of the plans and perform pairwise comparison, respectively. Results The plans obeyed the recommended prescription isodose volume (PIV) within 5.5–7.5 cc and maximum doses to brainstem, optic apparatus (OA) of 10 and 8 Gy, respectively, for a prescription dose of 24 Gy. The volumes of the target were in the range 4.0–7.4 cc. Mean PIV, maximum dose to brainstem, OA were 6.5 cc, 10 Gy, 7.9 Gy in GammaKnife; 7.2 cc, 6.1 Gy, 4.5 Gy in Eclipse NCC; 7.2 cc, 6.4 Gy, 5.7 Gy in Brainlab DCA; and 5.2 cc, 8.4 Gy, 6.1 Gy in Eclipse VMAT plans, respectively. Multivariate analysis of variance showed significant differences among the 4 SRS planning techniques (P‐values < 0.01). Conclusions Among the 4 SRS planning methods, VMAT with least PIV and acceptable maximum doses to brainstem and OA showed highest compliance with ROSE trial. Having the most conformal dose distribution and least dose inhomogeneity, VMAT scored higher than GK, Eclipse NCC, and Brainlab DCA plans.


| INTRODUCTION
Epilepsy is the 4th most common neurological disorder in the United States with an annual incidence of more than 150,000. Mesial temporal lobe epilepsy (MTLE) refers to a chronic condition of recurrent seizure activity focally originating in the temporal lobe, namely the amygdala and hippocampus. The initial treatment for newly-diagnosed MTLE is anti-epileptic medication. The medical refractory MTLE cases that fail as few as two trials of medication should seek investigation for open surgery. Although seizure-freedom rates are as high as 80-90%, very few patients are referred for resection in the United States. 1 Stereotactic laser amygdalo-hippocampotomy accomplishes ablation of the seizure focus with real-time magnetic resonance thermal imaging in a minimally invasive approach that eliminates intensive care unit stay. 2 For a subgroup of MTLE patients with medical contraindications to surgery, stereotactic radiosurgery (SRS) has emerged as an alternative therapy in the selective ablation. 3,4 Although not quite as effective compared to anterior temporal lobectomy (ATL), the preliminary results were supportive of the efficacy of SRS for select cases of MTLE. The Radiosurgery or Open Surgery for Epilepsy (ROSE) clinical trial was designed to compare the effectiveness of Gamma Knife (GK) radio surgery with lobectomy in patients with pharmaco-resistant MTLE. 5,6 The final outcome analysis of ROSE trial suggests that both SRS and ATL have effectiveness and reasonable safety for MTLE, but ATL has an advantage in the number of seizure remission. 7 We present here a comparison study on SRS plans in GK for treatment of epilepsy against highly non-coplanar conformal (NCC) plan in Eclipse, dynamic conformal arc (DCA) plan in Brainlab, and a volumetric modulated arc therapy (VMAT) plan in Eclipse echoing the ROSE trial planning guidelines and dose constraints. Dosimetric comparison abiding by ROSE trial guidelines was performed using RTOG plan quality metrics. 8 A set of primary and secondary dosimetric aims were adopted from ROSE trial guidelines in consulting among the neurosurgeon, radiation oncologists and clinical medical physicists. The primary aims of this study include: (a) prescription isodose volume (PIV) less than 7.5 cc, (b) maximal dose to brainstem of 10 Gy and maximal dose of 8 Gy to optic apparatus (OA) that includes optic nerves, optic chiasm. The secondary aims include (c) close to 100% target coverage (TC), (d) radiosurgical treatment time of less than 90 min. While every effort was made to satisfy both pairs of primary and secondary aims, the former shall be fulfilled, whereas the latter was considered less critical to this study.
Differences in dose distributions are expected from multiple sources including dose calculation algorithms, and planning techniques. There are inherent differences among these four planning modalities. Gamma Knife planning accomplishes target coverage within 50% isodose curve using multiple isocenters (shots). The maximum target dose in GK plan of twice the prescription (Rx) dose is expected to be significantly higher than those in the linac-based plans. A SRS comparison study by Petrovic et al studied dose distributions obtained with analytical anisotropic algorithm (AAA) in Eclipse and pencil beam in Brainlab showed average dose differences of less than 3% in cranial cases. 9 Multiple studies comparing the GK and linac-based SRS treatments were investigated by Gevaert et al. 10 Despite the vast differences in central dose distributions in these SRS planning techniques, clinical trials have so far failed to identify differences in treatment outcome or toxicity. In particular, it is not known if this dose differential is important in seizure remission. The lack of data on a functional target such as MTLE led us to compare the four SRS treatment deliveries. In addition, beams with direct entrance through brainstem or OA were avoided in all linac-based plans.

| METHODS AND MATERIALS
While a GK plan based on magnetic resonance images (MRI) assumes homogeneous medium, linac plans uses computed tomography (CT) for tissue heterogeneity correction. Care was taken to ensure the hot spots occur within the target volume (TV) in all the plans. All plans were created by experienced treatment planners, and reviewed by the treating radiation oncologist for plan quality and OAR dose adherence to the ROSE trial criteria.

2.A | Gamma Knife planning
Twenty patient cases having both MRI and CT of the whole brain were identified. This includes T1-weighted, spoiled gradient recalled (SPGR) or magnetization prepared rapid gradient echo (MPRAGE) MRI at submillimeter slice thickness. The amygdala and anterior 2 cm of hippocampus along with adjacent parahippocampal gyrus were contoured as the radiosurgical TV. Rx of 24 Gy was prescribed to the 50% isodose line using 90 degree gamma angle. GK planning was generated for treatment in a Perfexion unit allows (a) using composite shots containing combination of 4, 8, and 16 mm or blocked sectors, and (b) dynamic shaping to reduce dose to critical structures. 12

2.B | Registration of CT with MRI
The MRI along with contours, dose information and plan files were exported from Gamma Plan at 1 mm grid spacing. The MRI was NARAYANASAMY ET AL.
| 135 rigidly registered with CT in MimVista software (ver 6.6.5, MIM Software Inc., Cleveland, OH). The contours were transferred into CT exporting to Eclipse and Brainlab TPS. Figure 1 shows the TV, brainstem, and OA in MRI and CT images.

2.C | Non-coplanar conformal (NCC) plan in Eclipse
NCC plan with 20 beams employing static gantry was generated with optimum collimator angle based on the target shape along the beam's eye view (BEV). Conformal beams were created by fitting MLCs around the target using anisotropic margin. When brainstem or OA is adjacent to the target along the BEV, a zero mm margin between the projection of MLCs and TV was used. In all other directions, a graded approach was taken to estimate the optimum margin between the projection of MLCs and TV in the range 0-2 mm for effective normal tissue sparing. 13 On two trial cases, multiple SRS plans with margins of 0, 0.5, 1, 1.5, and 2 mm between the TV and projection of MLCs that met the ROSE trial guidelines were compared.

2.D | Dynamic conformal arc (DCA) plan in Brainlab
Dynamic conformal arc plan with 120 0 gantry span and having 5 arcs (Table angles of

2.E | Volumetric Modulated Arc Therapy (VMAT) plan in Eclipse
An Eclipse RapidArc plan with 4 arcs (1 being complete coplanar arc and 3 half-arcs with table angles of 30°, 60°, and 90°for left target; and 300°, 330°, and 90°for right target) was produced. A 1 cm wide ring was created at a gap of 5 mm from TV surface to help conform the dose. A dose normalization of 90% was used in Eclipse VMAT plans.

2.F | Planning Evaluation
The plan DICOM data were exported to MimVista and extracted at 10 cGy bin-width for dosimetric comparison. We employed a few plan quality metrics recommended by International Commission on Radiation Units and Measurements (ICRU) report 83 14 and Radiation therapy oncology group (RTOG). 8 Conformity index (CI) is the ratio of the Rx isodose volume (V RI ) to the TV.
TC is that fraction of TV covered by Rx: Dose gradient index (GI) is the ratio of 50% isodose volume to the 100% isodose volume. Although, low GI values are preferred, the acceptable range depends on TV.
Dose uniformity in the TV was estimated using homogeneity index (HI) based on the dose irradiated to 2% (D 2% ), 98% (D 98% ) of TV and Rx: In addition, beam-on time and the number of shots used in GK plan and number of monitor units (MU) in linac-based plans were tabulated in compliance with ROSE trial guidelines.
Although exit doses passing through brainstem or OA were not curtailed, none of the linac beams entered through brainstem or OA.
With respect to OAR doses, 24 Gy and 12 Gy isodose volumes were displayed in Fig. 2 for a representative patient. The corresponding DVHs of the target (red), brainstem (green), and OA (purple) is shown in Fig. 3.

2.G | Plan statistics
Comparison of dosimetric data and plan quality metrics was based on multivariate analysis of variance (MANOVA) using STATA (ver 9.2, StataCorp, College Station, TX). Data normality was tested using Shapiro-Wilk test. Pairwise comparison analysis uses two-tailed paired Student's T-test or Wilcoxon signed-rank test. P-value less than 0.0125 was considered statistically significant in pairwise comparisons.

| RESULTS
The target volumes were in the range 4.0-7.4 cc with a mean ± SD of 5.5 ± 1.0 cc. Overall beam-on time was 86 ± 17 mins in GK plans using 11.5 ± 4 (range: 5-20) shots which fits well within ROSE trial guideline of 6-30 shots.

3.A | Treatment margins
On two trial patient plans, Eclipse NCC, Brainlab DCA, and Eclipse VMAT plans were created using 0, 0.5, 1.0, 1.5, and 2 mm margin between the projection of MLCs and TV. It was observed that 0 mm plan could spare the OARs and stay within the prescribed range for PIV but has poor TC. On the other hand, the plan with 2 mm margins provided higher TC but breached the upper limit of PIV and maximum OAR doses. An optimum 1 mm margin between the projection of MLCs and TV was used on all linac-based plans.

3.B | Planning comparison
The mean ± SD of target dose, OAR doses and plan quality metrics tabulated in Table 1     Low-dose spillage to healthy brain parenchyma is a concern for radiation-induced malignancy and necrosis. The V12Gy (associated with neurotoxicity 28 ) being significantly low in GK plans, increases in Eclipse VMAT, Eclipse NCC, and a large spread of low isodose volume was observed in Brainlab DCA plans. 29  presents low risk for maximum dose 10-12 Gy. 30 With regard to brainstem, maximum dose of 12.5 Gy is associated with a < 5% risk of cranial neuropathy in SRS. 31 Thus, we believe that the dosimetry of MTLE plans can be improved using evidence-based tolerance doses to brainstem and OA which can be higher, while respecting PIV constraint.
SRS plans for MTLE that met the ROSE trial guidelines were created using the four planning techniques and a few features stood out. A PubMed search with keyword combinations that include "functional," "tumor," "stereotactic," and "SRS plan" failed to return any tangible result making this possibly the first comparative study on SRS of a functional target. Our study has the following limitations: (a) did not address the correlation of a conformal dose distribution with treatment outcome in MTLE, (b) was not designed to analyze the importance of low-dose spread or dose gradient with seizure-free survival, (c) did not include hypo-fractionation, and (d) did not address the correlation of target inhomogeneity with treatment outcome in MTLE.

CONF LICT OF I NTEREST
The authors declare no conflict of interest.