Submillimeter alignment of more than three contiguous vertebrae in spinal SRS/SBRT with 6‐degree couch

Abstract The purpose of this study is to identify regions of spinal column in which more than three contiguous vertebrae can be reliably and quickly aligned within 1 mm using a 6‐degree (6D) couch and full body immobilization device. We analyzed 45 cases treated over a 3‐month period. Each case was aligned using ExacTrac x‐ray positioning system with integrated 6D couch to be within 1° and 1 mm in all six dimensions. Cone‐Beam computed tomography (CBCT) with at least 17.5 cm field of view (FOV) in the superior–inferior direction was taken immediately after ExacTrac positioning. It was used to examine the residual error of five to nine contiguous vertebrae visible in the FOV. The residual error of each vertebra was determined by expanding/contracting the vertebrae contour with a margin in millimeter integrals on the planning CT such that the new contours would enclose the corresponding vertebrae contour on CBCT. Submillimeter initial setup accuracy was consistently achieved in 98% (40/41) cases for a span of five or more vertebrae starting from T2 vertebra and extending caudally to S5. The curvature of spinal column along the cervical region and cervicothoracic junction was not easily reproducible between treatment and simulation. Fifty‐seven percent (8/14) of cases in this region had residual setup error of more than 1 mm in nearby vertebrae after alignment using 6D couch with image guidance. In conclusion, 6D couch integrated with image guidance is convenient and accurately corrects small rotational shifts. Consequently, more than three contiguous vertebrae can be aligned within 1 mm with immobilization that reliably reproduces the curvature of the thoracic and lumbar spinal column. Ability of accurate setup is becoming less a concern in limiting the use of stereotactic radiosurgery or stereotactic body radiation therapy to treat multilevel spinal target.

(FOV) in the superior-inferior direction was taken immediately after ExacTrac positioning. It was used to examine the residual error of five to nine contiguous vertebrae visible in the FOV. The residual error of each vertebra was determined by expanding/contracting the vertebrae contour with a margin in millimeter integrals on the planning CT such that the new contours would enclose the corresponding vertebrae contour on CBCT. Submillimeter initial setup accuracy was consistently achieved in 98% (40/41) cases for a span of five or more vertebrae starting from T2 vertebra and extending caudally to S5. The curvature of spinal column along the cervical region and cervicothoracic junction was not easily reproducible between treatment and simulation. Fifty-seven percent (8/14) of cases in this region had residual setup error of more than 1 mm in nearby vertebrae after alignment using 6D couch with image guidance. In conclusion, 6D couch integrated with image guidance is convenient and accurately corrects small rotational shifts. Consequently, more than three contiguous vertebrae can be aligned within 1 mm with immobilization that reliably reproduces the curvature of the thoracic and lumbar spinal column.
Ability of accurate setup is becoming less a concern in limiting the use of stereotactic radiosurgery or stereotactic body radiation therapy to treat multilevel spinal target.

| INTRODUCTION
Spinal stereotactic radiosurgery (SRS) or stereotactic body radiation therapy (SBRT) delivers high dose in one or a few fractions is increasingly used to maximize local tumor control and pain relief. 1 One question in patient eligibility of spinal SRS/SBRT is how many contiguous spine vertebrae can be treated. In RTOG 0631 clinical trial, the limit was set to two. In our institution's current practice, it has been mostly limited to three or less contiguous vertebrae. Such a limitation is due to the caution and uncertainties of normal tissue tolerance in spinal SRS/SBRT and great concern on the technique ability to align long/ large target accurately and reliably. One characteristic of spinal SRS/ SBRT plans is the sharp drop-off of dose between the target and spinal cord due to the close proximity of this critical structure to the target lesion. Even a small setup error of 2 mm or 2°can significantly affect both the target and spinal cord dose. 2 The effect of small rotational error is more significant when treating large target spanning three or more vertebral bodies, since such error could cause large displacement at distance from the isocenter. Rotational error is especially challenging to correct with a conventional treatment couch.
Recent advance in 6-degree (6D) couches has made it easier to quickly and accurately correct both rotational and translational shifts.
Hypothetically, if the curvature of the vertebral column could be reproduced during treatment as the simulation, the 6D couch would be able to accurately align several contiguous vertebrae under image guidance. The purpose of this work is to assess how many contiguous vertebrae can be reliably aligned within 1 mm using full body immobilization and ExacTrac x-ray (Brainlab) positioning system with its integrated 6D couch. In the study, we examined the largest residue error of each vertebra by replicating the process of CBCT verification of the ExacTrac alignment in the Pinnacle treatment planning system (TPS). This is different than most registration software, such as the online ExacTrac and Varian system, which only display the residual error as a 3D or 6D shift of isocenter. Our result can be easily used to estimate the dose perturbation to the target and normal tissue based on the dosimetric characteristic of spinal SRS/SBRT plan, hence the feasibility of treating a long target. We hypothesize that, with appropriate immobilization, image guidance, and the adjustments made feasible by the 6D couch, spinal SRS/SBRT may be utilized to treat to a larger population of patients with contiguous, multilevel metastatic disease to the spine. In addition to the setup uncertainties, this study also examined two real patient cases treating four and five contiguous vertebrae. The dosimetric parameters including both maximum dose to a point (Dmax) and maximum dose to a partial volume of 1, 2, and 5 cm 3 (D1 cc, D2 cc and D5 cc) and their association with added volume are discussed.

2.A | Patient selection and immobilization
All spinal SRS/SBRT patient treatments from January 1 to March 31 of 2016 were analyzed in this retrospective study following institution IRB-approved protocol. This included a total of 45 cases with each case treated with one isocenter. The vertebral targets treated at each case are detailed in Table 1 as lesion site ranging from C1 to S3. The immobilization devices for each patient were made during CT simulation process. The primary immobilization device for each patient is a full body vacuum cradle (BlueBAG, Elekta). The cradle was custom molded to the patient's body to provide reproducibility, stability, and comfort of patient positioning. In addition, a thermoplastic mask (High Precision System for Head, Neck and Shoulders, Orfit) or plastic body cover sheet (BodyFIX  Table 1. Following the generation of each patient's specific immobilization devices, a planning CT was acquired with 1 mm slice thickness, ranging at least 10 cm above or below the target to be treated. The resolution of X and Y directions in each slice was also 1 mm.

2.B | Multimodality imaged-guided alignment in patient treatment setup
Each case was treated in a suite equipped with a Varian TrueBeam Stx linear accelerator (LINAC) and ExacTrac x-ray patient positioning system with an integrated 6D couch. The TrueBeam on board image (OBI) system has the ability to perform kV, MV, and CBCT. ExacTrac uses high-resolution stereoscopic x-ray images to detect patient position and an infrared (IR) optical system with 6D couch to apply shifts to align patient position in six dimensions. The image isocenters (kV, CBCT, and ExacTrac) and radiation isocenter congruence were tested bi-weekly following a quality assurance procedure, which first aligned a tungsten ball (6.5 mm in diameter) to the kV (CBCT) isocenter. The ExacTrac isocenter relative to the ball was acquired through the Winston-Lutz test in the ExacTrac software.
The radiation isocenter relative to the ball was measured using MV portal images (gantry at 0, 90, 180, and 270) of a 5 9 5 cm MLC field and an algorithm described in the literature. 3 The discrepancies between isocenters were calculated from their positions relative to the ball. They were reliably within 0.5 mm, with no need of recalibration of image isocenters over a 3-year period. The only time an image isocenter calibration maybe required was when the x-ray source was serviced. Daily ExacTrac test was also performed with manufacture provided calibration/daily phantom to verify the IR optical system with 6D couch had submillimeter accuracy.
The goal of the initial setup accuracy in our spinal SRS/SBRT treatment is to achieve accuracy of <1 mm and <1°in all six dimensions. Figure 1 shows the diagram of major steps and image T A B L E 1 List of cases, lesion locations, and immobilization setup. "x" marks the immobilization device was used.
L5-S1 X X X 44 S1 X X X 45 S2-S3 X X X modalities used in the initial setup process to achieve this goal. This process is summarized here: (a) The patient was first settled into the immobilization apparatus and aligned using skin marks (such as leveling/rotation marks, marked isocenter, or ExacTrac infrared spheres).
This step aimed to reproduce the body position as close as the simulation and eliminate large rotational and translational error. (b) Following skin-based setup, ExacTrac positioning system was used as the primary tool to align target through couch shifting (referred as "ExacTrac Correction" in the diagram). In the "ExacTrac Correction" process, the x-ray system and its image registration software were used to detect shifts in 6D freedom, and shifts were applied automatically using 6D couch under the guidance of its optical system. another ExacTrac image verification were taken to verify all imaging modalities were in agreement the setup accuracy meeting the predetermined criteria, and intrafraction motion was not an issue through the lengthy initial setup process.

2.C | Vertebrae alignment accuracy analysis
ExacTrac alignment and CBCT verification as prescribed in the previous section are the two steps relevant to this study. During treatment, only the alignments of vertebrae with disease were examined.
This limited the number of contiguous vertebrae to three or less for the majority of cases as detailed in Table 1

2.D | Treatment planning and dosimetric analysis of long target patient cases
In this study, two cases were treated with more than three contiguous vertebrae. Case # 16 was a melanoma patient with disease from T5 to T8 level. (GTV) and clinical target volume (CTV) from T5 to T8 level. The challenge for this case included spinal cord at T6 level which was right next to the GTV, and more significantly, esophagus which was just anterior of GTV over all four levels from T5 to T8. Case # 40 is a patient with metastatic renal cell carcinoma. He had a resection at L5, but progressed with new disease from L2 to S1, as shown in Fig. 3(b). The major challenge for this case was the cauda equina which was surrounded by the target over five contiguous vertebrae.
As part of our initial experience, both treatments were planned for  Table 3. A number "1" means the corresponding vertebra was aligned accurately within 1 mm, "2" was aligned within 2 mm, and so on for other numbers. Cases in Table 3 with residual setup error of more than 1 mm are marked in red to identify regions of the spinal column with difficulties meeting the submillimeter accuracy. Target areas were marked as yellow cells in Table 3  Using the data in Table 3 T1   T2   T3   T4   T5   T6   T7   T8   T9   T10   T11   T12      Our study focused on the initial geometric alignment of each vertebra over a long column. More importantly, the knowledge gained here enabled us to evaluate the likely dose perturbation to the nearby normal tissues during delivery.
Spinal cord is of greatest concern due to its close proximity to the target and high consequence of injury to this structure. The risks of radiation myelopathy in spinal SRS/SBRT were very well studied [6][7][8][9][10][11][12][13][14][15][16][17] with Dmax been generally accepted as the parameter of dosimet-  There are other concerns in treating long target, particularly, the accuracy of target and critical structure contours. MRI imaging is used in spinal SRS/SBRT to accurately delineate gross disease and the spinal cord/cauda equina. These contours are transferred onto the planning CT through image registration for treatment planning. The accuracy of these contours can be compromised by the registration process when the patient has significant difference in body posture between CT simulation and MRI imaging. In addition, the location of spinal cord and the shape of soft tissue involvement of gross disease could also be effected by the posture. 5 Optimizing MRI imaging by scanning the patient in the custom made spinal immobilization device which reproduces the spinal column curvature as in CT could be valuable in treating multilevel targets.
T A B L E 6 Dosimetric values to cauda equina of case # 40. The cauda equina volumes include the total volume (L2 to S1) over the target area and sub volume in each individual vertebra level. Misalignment is a vertical shift in isocenter of clinical plan. Dose pertubation is the average of dose value changes from 0 to 1 and 1 to 2 mm misalignment.