The potential of an optical surface tracking system in non‐coplanar single isocenter treatments of multiple brain metastases

Abstract To evaluate the accuracy of a commercial optical surface tracking (OST) system and to demonstrate how it can be implemented to monitor patient positioning during non‐coplanar single isocenter stereotactic treatments of brain metastases. A 3‐camera OST system was used (Catalyst HD™, C‐RAD) on a TruebeamSTx with a 6DoF couch. The setup accuracy and agreement between the OST system, and CBCT and kV‐MV imaging at couch angles 0° and 270°, respectively, were examined. Film measurements at 3 depths in the Rando‐Alderson phantom were performed using a single isocenter non‐coplanar VMAT plan containing 4 brain lesions. Setup of the phantom was performed with CBCT at couch 0° and subsequently monitored by OST at other couch angles. Setup data for 7 volunteers were collected to evaluate the accuracy and reproducibility of the OST system at couch angles 0°, 45°, 90°, 315°, and 270°. These results were also correlated to the couch rotation offsets obtained by a Winston‐Lutz (WL) test. The Rando‐Alderson phantom, as well as volunteers, were fixated using open face masks (Orfit). For repeated tests with the Rando‐Alderson phantom, deviations between rotational and translational isocenter corrections for CBCT and OST systems are always within 0.2° (pitch, roll, yaw), and 0.1mm and 0.5mm (longitudinal, lateral, vertical) for couch positions 0° and 270°, respectively. Dose deviations between the film and TPS doses in the center of the 4 lesions were −1.2%, −0.1%, −0.0%, and −1.9%. Local gamma evaluation criteria of 2%/2 mm and 3%/1 mm yielded pass rates of 99.2%, 99.2%, 98.6%, 89.9% and 98.8%, 97.5%, 81.7%, 78.1% for the 4 lesions. Regarding the volunteers, the mean translational and rotational isocenter shift values were (0.24 ± 0.09) mm and (0.15 ± 0.07) degrees. Largest isocenter shifts were found for couch angles 45˚ and 90˚, confirmed by WL couch rotation offsets. Patient monitoring during non‐coplanar VMAT treatments of brain metastases is feasible with submillimeter accuracy.


| INTRODUCTION
While the use of stereotactic radiosurgery (SRS) in patients with a limited number of brain metastases (BM) has been clearly defined, the application of SRS in patients with multiple BM (>4) is still a matter of controversy. 1 Whole-brain radiation therapy (WBRT) was traditionally the standard treatment approach, but it is associated with significant side effects, such as cognitive dysfunction (which results in a decreased quality of life), hair loss and fatigue. [1][2][3] Limiting radiation to the uninvolved brain and obtaining a high probability of local tumor control with a single treatment, are therefore important advantages of SRS over WBRT in patients with 4 or more BM. 4 Over the years, there has been a lot of technological progress in the way BM is treated with a linear accelerator (linac). Instead of forward planning techniques delivered with static beams or dynamic conformal arcs, an inverse planning technique is used and the beams are delivered with volumetric modulated arc technique (VMAT).
Preferably, a single isocenter is used to make the delivery more efficient, thereby reducing the treatment time. 5 Compared to high precision GammaKnife based treatments of multiple BM, the beam-on time on a linac is much lower, especially when flattening filter-free beams are used. [6][7][8] A crucial aspect is to reduce the GTV-PTV margin to 1 mm, as the probability of radionecrosis (RN) increases when the V12Gy of the brain exceeds 10 cm 3 . 9 High-dose irradiated isodose volume, V22Gy, is also significantly correlated with RN, particularly for patients treated with SRS alone. 10 In terms of local control, the randomized study of Kirkpatrick et al. showed that there is no difference between the use of 1 mm or 3 mm GTV-PTV margin. 11 To treat all BM with such small margin simultaneously, a six degrees-of-freedom (6DoF) correction is essential to guarantee a submillimeter setup accuracy. Frameless radiotherapy for treating intracranial lesions has been widely adopted under the guidance of on-board cone beam CT (CBCT) and a thermoplastic mask system with a 6DoF robotic couch [12][13][14] or a semi-robotic couch including manual angle adjustments. 15 The final step in this progress in the treatment of BM is the introduction of more degrees of freedom by using non-zero couch angles during treatment planning and delivery. A non-coplanar technique is not new, but with the introduction of VMAT and imageguidance techniques (using CBCT at couch 0˚), it became a logical development to maintain to the couch at 0˚, to avoid possible collision problems. The advantages of incorporating non-zero couch angles in the treatment planning, resulting in better sparing of normal brain tissue, has been published widely. 8,16,17 In our workflow of non-coplanar treatments with a standard linac, the patient with BM is immobilized in a thermoplastic mask on the linac equipped with 6DoF couch and with a high-definition multileaf collimator (MLC). The localization accuracy of the frameless image-guided system is found to be comparable to robotic or invasive frame-based radiosurgery systems. 18 Online CBCT acquisition to verify the patient position is challenging (or even impossible) for a non-coplanar technique, due to possible collision of the gantry and treatment couch. In order to preserve the GTV-PTV margin of 1 mm, the patient must be accurately positioned at all times, hence the need for patient monitoring. 19 The couch rotation offsets from the central axis (CAX) can be quantified using a Winston-Lutz (WL) test. 20 This is a commonly used method to localize the isocenter of a linac by correlating the radiation fields directly with the object being irradiated, which is a ball-bearing (BB) phantom positioned at the center of each radiation field using external lasers and imaged on a piece of film or more recently an electronic portal imaging device. 21,22 The final position of the BB corresponds to the intersection of the CAX of all sampled radiation fields, in other words, the radiation isocenter.
However, with a good quality control (QC) tool (like the WL test) that guarantees that the couch axis and treatment beam axis are aligned with submillimeter accuracy, it remains questionable whether the accuracy of a non-coplanar treatment is adequate solely relying on CBCT imaging verification at couch 0˚. A retrospective analysis with 288 SRS brain patients, treated with 1,344 fractions by means of an ExacTrac ® system and 6D couch of Brainlab AG (Munich, Germany) has shown that although the patients were fixated with thermoplastic masks, positioning corrections exceeding 1 mm appeared for 42% of beams and exceeding 1°for 9% of the beams. 23 Further, the longer the treatment delay, the larger the risk of having positioning deviations, it is, therefore, necessary to have a continuous positioning monitoring of the patient in the treatment room while ensuring a short treatment time. Optical Surface Tracking (OST) seems to be a sophisticated and suitable option, as it intends to reduce set-up errors and provides real-time non-invasive monitoring to detect patient movement during treatment, without the use of ionizing radiation. [24][25][26] The aim in this work is to demonstrate how a commercial OST system can be implemented to monitor patient positioning during treatment for a non-coplanar single isocenter VMAT technique for multiple BM and to show the feasibility of this system by defining treatment tolerances without compromising the treatment margins.

2.A | Optical surface tracking and patient fixation
system The Catalyst HD™ system is provided by C-RAD Positioning AB (Uppsala, Sweden). The OST system uses LEDs to project a light of 3 wavelengths (λ = 405nm (blue), λ = 528nm (green), λ = 624nm (red)) 1 onto the patient and a charge-coupled device camera to detect the light reflected from the patient. Using the information from the reflection the system generates a real-time 3D surface of the patient, which is compared to a reference surface for verification. The reference surface can either be the body structure (DICOM RT-STRUCT) from the CT or it can be created directly in the OST system during treatment set-up. The latter should be applied when using the system on patients (not phantoms), after initial positioning using CBCT at couch 0˚and used to monitor the patient position through the rest of the treatment. The correspondences between the reference surface to the patient's real-time surface are calculated using a non-rigid algorithm. 27 The OST system's calculation of the isocenter shift includes 2 main stages: registration of the reference surface to the live surface and using this registration result to predict the impact on the live surface position by using a volumetric deformable model. 27 The calculated position inaccuracies are displayed in real-time in 6 dimensions, including translational and rotational shifts.
The advantage of a real-time monitoring OST system is that it can detect patient movement during treatment, in contrast to the CBCT where the patient position can only be verified during the actual acquisition before the treatment. Furthermore, the OST system, comprising a main camera unit extended with 2 additional camera units with 120˚angle from the main unit [ Fig. 1(c)], has the ability to verify the online patient setup for all couch angles, which makes it highly appropriate for non-coplanar treatments.
Patient data is imported to the OST system (C-RAD c4D TM software version 5.4.1) from the treatment planning system (TPS). The exposure time and saturation settings of the 3 cameras can be altered individually to improve the quality of the live patient surface, which makes the system reliable for different skin tones. The scan volume should be adjusted to only include the opening of the mask. 2. QUASAR Pentaguide phantom 28 : aligned to the treatment isocenter using CBCT imaging followed by a couch correction based on the match result. This procedure ensures that the Catalyst HD TM is aligned with the CBCT system. To assess the coincidence of the imaging centers with the radiation isocenter, the IsoCal is used, which is an automated geometric calibration system for on-board imaging and MV imaging systems at couch 0°. 30,31 The comparison of the IsoCal with an independent Winston-Lutz (WL) method to locate the radiation isocenter has been found to be within 0.4 mm. 30 In this work, both the IsoCal as a WL test -using an in-house Matlab TM image procession code with 11 combinations of gantry, collimator and couch angles-were used as a quality assurance tool.

2.B | Treatment and imaging procedure
This investigation focused on the WL test results at gantry 0å nd various couch angle rotations (yaw of 0°, 45°, 90°, 315°a nd 270°). skin for the surface scanning, the head phantom was painted in skin color using heavily skin-tone pigmented make-up in the area where the face was exposed, resulting in optimized quality of the real-time surface with more common camera settings. Again, a verification CBCT was made, and the isocenter shift corrections from the 3D-3D match (Δ ref-CBCT ) were compared with the OST measurements (Δ BODY-OST ). From this set-up, the couch was rotated to 270°to check the phantom setup with orthogonal kV-MV images using a 2D-3D match procedure, compared with the OST system position shifts. was used for reporting gamma evaluation scores (with agreement criteria of 2%/2 mm as well as 3%/1 mm) and dose deviations in the center of the PTV. In order to put these results into perspective, also a co-planar plan at couch 0˚has been delivered to the same phantom set-up, with a film positioned at 3.5 cm (PTV-1). The film dosimetry procedure is described in more detail in the appendix.

2.C.3 | Mannequin training head
To check whether the Catalyst HD TM is able to accurately visualize the patient at the various couch angles, an experiment was performed with a mannequin training head in the open face mask (i.e., a patient lying motionless) (Fig. 1). The OST system reference surface was captured and the couch was rotated to couch angles 0°, 45°, 90°, 315°a nd 270°. In theory, if the MV beam and the OST system isocenter are perfectly aligned, any displacement detected by the OST system during couch rotations should only be due to couch rotation drift, assuming the patient has not moved and the OST system is able to accurately visualize the patient at the various couch positions.

3.B | Dose delivery verification using film
After the CBCT based 6DoF match procedure at couch 0°, the setup of the Rando-Alderson head is monitored using the OST system at couch angles 315°, 270°, 45°, and 90°.
Deviations between the absolute measured and predicted TPS doses in the center of the 4 PTVs irradiated with a single isocenter non-coplanar VMAT plan monitored by the OST system are −1.2%, −0.1%, −0.0%, and −1.9% for PTV-1, PTV-2, PTV-3, and PTV-4, respectively (Table 1). In addition, the deviation between film and TPS doses in the center of PTV-1 for a co-planar VMAT plan at couch 0˚is −1.6%. The daily output fluctuation of the TrueBeam STx linac for 6MV photons was within 0.1%.   (Table 1). Additionally, the dosimetric agreement for the co-planar plan using the same gamma evaluation criteria (2%/2 mm and 3%/1 mm) was 99.8% and 99.9%, respectively.

| DISCUSSION
With most external beam radiotherapy treatments, an accuracy of ±3 mm is considered desirable and usually achievable. With stereotactic radiotherapy, however, like linac based SRS treatments of patients with multiple BM, somewhat higher accuracy is desired and, with modern techniques, submillimeter accuracy is achievable but requires careful verification. 9,32 Regarding the technical capability to accurately align the delivery system to the isocenter, current mechanical engineering standards meet this requirement easily. 33,34 When using frameless, image-guided SRS (using thermoplastic immobilization masks, CBCT online match procedures, a robotic couch,…), it is necessary to match the imaging isocenter to the mechanical isocenter, which is an achievable goal for standard QA according to AAPM TG-142 (1 mm/0.5˚). 35 The IsoCal procedure, as part of the Machine Performance Check designed on TrueBeams to quickly evaluate the machine's geometric performance 36

(Varian Medical
Systems), guarantees a coincidence between imaging and radiation isocenter within 0.5 mm (namely a built-in tolerance of 0.2 mm and an action level of 0.5 mm). To verify whether submillimeter accuracy can be achieved between the radiation isocenter and the mechanical isocenter, one would need to perform an end-to-end test using image-guidance and a dosimetric system with the highest spatial resolution, like radiochromic film. Submillimeter accuracy for an end-toend test with a TrueBeam linac has already been demonstrated. 37 Literature on end-to-end testing of a single isocenter VMAT treatment for multiple BM is scarce. 38,39 When a non-coplanar technique is applied, the coincidence between radiation and couch rotation T A B L E 1 The dosimetric agreement between film and TPS doses is presented by the deviations between measured and calculated doses in the center of the 4 PTVs and by a local gamma evaluation criterion of 2%/2mm as well as 3%/1mm for the 4 PTVs irradiated with single isocenter non-coplanar VMAT and monitored using Catalyst HD TM . To put these numbers into perspective, also the results for PTV-1 irradiated in a co-planar set-up (couch 0˚) are given  23 In Tryggestad et al, a mean intrafractional motion was found to be (1.1 ± 1.2) mm (mean ± SD) using CBCT images, acquired before and after intra-cranial radiation treatment, to determine movements of the head in a thermoplastic mask. 40 When the treatment time delay prolongs, the higher the risk of having positioning deviations. 23 Latter fact also favors the single isocenter technique above a multiple isocenter technique, where an online imaging, as well as the OST procedure, has to be performed for every isocenter separately. This was not part of our investigations but we expect that the use of multiple isocenters has a negative impact on the patient set-up accuracy. To assess how mechanical, imaging, treatment planning, and radiation isocenter uncertainties combine in a rigid phantom, we performed an end-to-end test using multi-slice film dosimetry in the Rando-Alderson head phantom. Using a gamma criterion of 2%/ 2 mm a good agreement between the measurements and predicted TPS doses was found, demonstrating that an accurate dose can be delivered at the correct position. Nevertheless, one should be aware that for very small lesions (<2 cm 3 ) this might no longer be the best criterion. As the film has the highest spatial resolution, and since a reduction to 3%/1 mm is possible, 37 we also looked into this gamma criterion and we still see acceptable agreement scores (Table 1).  Depending on the desired treatment margin, treatment tolerances for the OST system can be defined. In our case, we maintain a more stringent GTV-PTV margin of 1 mm, so an OST system tolerance of 0.5 mm and 0.5˚would be feasible for our delivery system, with the exclusion of couch rotations 45˚and 90˚which pass the AAPM TG-142 criteria but demonstrated a less accurate couch rotation coincidence with the radiation isocenter compared to the other couch angles (Table 2). 35 Therefore, a regular WL test (in our situation with a tolerance of 0.5 mm) should be part of the SRS-specific QA program of a linac in order to be informed about the capabilities of the SRS delivery system present in the clinic. In order to reduce the time for a (daily) SRS-specific QA program, future work will be to develop a novel and practical WL test based phantom that integrates several tests, namely measuring the isocenter congruence of imaging systems, radiation beam and couch rotation axis with submillimeter accuracy and also the isocentricity of the OST system.
Other future investigations will include testing similar OST system for brain tumor patients treated with proton therapy on the proton therapy, we expect the potential of this intrafraction monitoring system to be even higher.

| CONCLUSION
This work demonstrates a step-by-step realization of non-coplanar single isocenter SRS for multiple BM using the Catalyst HD TM . It was shown that submillimeter accuracy by the linac equipped with the OST system can be obtained for these treatments, depending on the type of the delivery system and the tolerances applied during the various SRS-specific QA procedures. The Catalyst HD TM OST system has the potential to be used as a dedicated patient monitoring tool during complex non-coplanar high-precision SRS treatments.

This work has been performed in collaboration with C-RAD and Orfit
Industries. Further, the authors like to thank the volunteers, who participated in this study.

CONF LICT OF I NTEREST
No conflicts of interest.

APPEN DIX
In addition to film irradiation, we performed ionization chamber measurements with a cross-calibrated PTW Farmer chamber (type TN30012, SN 0158) at the depth of 5 cm in RW3 water-equivalent phantom (SSD 100 cm, 10 cm 2 × 10 cm 2 , 5 Gy, 6 MV). One film in this set-up was used for calibration. Also, a blank film from the same batch was applied as a baseline film with 0 Gy.