Surface guided imaging during stereotactic radiosurgery with automated delivery

Abstract Purpose To report on the use of surface guided imaging during frameless intracranial stereotactic radiotherapy with automated delivery via HyperArcTM (Varian Medical Systems, Palo Alto, CA). Methods All patients received intracranial radiotherapy with HyperArcTM and were monitored for intrafraction motion by the AlignRT® (VisionRT, London, UK) surface imaging (SI) system. Immobilization was with the EncompassTM (Qfix, Avondale, PA) aquaplast mask device. AlignRT® log files were correlated with trajectory log files to correlate treatment parameters with SI reported offsets. SI reported offsets were correlated with gantry angle and analyzed for performance issues at non‐zero couch angles and during camera‐pod blockage during gantry motion. Demographics in the treatment management system were used to identify race and determine if differences in SI reported offsets are due to skin tone settings. Results A total of 981 fractions were monitored over 14 months and 819 were analyzed. The median AlignRT® reported motion from beginning to the end of treatment was 0.24 mm. The median offset before beam on at non‐zero couch angles was 0.55 mm. During gantry motion when camera pods are blocked, the median magnitude was below 1 mm. Median magnitude of offsets at non‐zero couch angles was not found to be significantly different for patients stratified by race. Conclusions Surface image guidance is a viable alternative to scheduled mid‐treatment imaging for monitoring intrafraction motion during stereotactic radiosurgery with automated delivery.

limited dosimetry experience and a more streamlined workflow. Predefined angles are selected for treatment and isocenter selection is used to ensure no collision risk during treatment. Patient immobilization must use the Encompass SRS Immobilization System (Qfix, Avondale, PA) for proper collision mapping and radiosurgery specific normal tissue objective. To monitor for intrafraction motion, Hyper-Arc TM allows MV imaging at designated waypoints during treatment.
Stopping the automated delivery to acquire and review images increases the length of the treatment and may increase the chance for patients to move. An alternative to mid-treatment imaging is surface imaging, which uses optical tracking of the patient's surface to monitor for intrafraction motion.
Recently, with the technological advancements in optical reconstruction and projection techniques, surface imaging has become increasingly more popular as a non-invasive, non-radiographic form of image guidance. Surface imaging (SI) systems use a combination of real-time optical and laser-based imaging techniques that have been shown to properly position patients, 1,2 accurately monitor, and quantify movement throughout the entirety of treatment, 3,4 and provide an accurate and reproducible respiratory surrogate for gatingbased deliveries. [5][6][7][8][9][10] The ability of SI systems to non-radiographically collect a live surface image, determine positional correction vectors needed to match image to a predefined reference image, and monitor sub-millimeter movements have made it a successful component of SRT where small targets and small margins are ever-present important considerations. Surface imaging has seen a dramatic increase in clinical prevalence in radiotherapy clinics around the world. While SI has previously been evaluated for accuracy and clinical efficiency in traditional SRT, [11][12][13][14][15] the evaluation of SI in conjunction with a 4pi based, automatic delivery has yet to be evaluated for clinical efficacy.
This study presents the largest cohort to date of patient data captured via surface guided imaging during SRT delivered with HyperArc TM . In this study, we examine the magnitude of translational intrafraction motion from beginning to the end of treatment, the magnitude of SI reported offsets at non-zero couch angles, the impact of gantry motion on performance with respect to camera blockage, and the effect of skin tone on reported offset.  viewing of the open-face region within the treatment planning sys-tem. The open-face region was then contoured, the eye regions removed, and the resulting structure was exported to AlignRT® to be used as the region of interest (ROI) for SGRT. Prior to treatment, therapist turn on the projectors (i.e., start monitoring) for a minimum of 10 min to allow the cameras to reach thermal equilibrium. AlignRT requires that the user select the skin tone setting per patient. Skin tone setting is selected per the discrection of the user based on visual inspection of the patient. During treatment, each patient underwent radiographic imaging with both kV orthogonal imaging and cone-beam CT (CBCT).

| MATERIALS AND METHODS
After radiographic alignment, a reference surface was captured in AlignRT®, and treatment was initiated. While HyperArc TM has optional MV waypoints to monitor for intrafraction motion, these were not utilized in favor of monitoring with SI. SI reported offsets, This was done by manually stopping the beam and not by use of gating thresholds and beam holds. If the MAG at couch zero was under the threshold, treatment was resumed. If RTDs remained above the threshold at the reference position, radiographic imaging was performed, shifts performed, and treatment was resumed.
Log files from AlignRT® continuously record RTDs from the patient's reference position throughout treatment. SI system logs were correlated with information from the ARIA database (Varian Medical Systems) and linear accelerator trajectory log files. The SI log files and trajectory log files were synchronized using the initial beam-on flag in each file.This enabled syncing RTDs with gantry angle toassess changes in RTDs with respect to the camera pod blockage by the gantry. The left and right camera pods were assumed to be at least partially blocked by the gantry at angles 303°± 15°and 57°± 15°.
The user selected skin tone setting is not documented in the SI log file; therefore, demographics in ARIA were used to identify the race of each patient as a surrogate for skin tone. Patients were classified in the following groups per ARIA: White, Black, or not-specified (NS). Spot checks of skin tone settings were performed and confirmed that skin tone and race were correlated. This was done to study the impacts of SI skin tone settings on performance. We compared the magnitude of RTDs between the three groups to determine if there were differences due to suboptimal camera exposure settings. RTDs before beam on at non-zero couch angle and the end of treatment were evaluated for the three patient cohorts. of treatment which caused a discontinuity in SI log file analysis.
Twenty-five were omitted due to a discordance between the SGRT and trajectory log beam on flags of >3 s. Thirty-two (3.8%) fractions contained mid-treatment imaging due to SI reported patient motion and were omitted due to discontinuity in RTDs due to new reference surface capture. Note that some fractions had multiple reasons for omission which left 844 fractions from 281 patients for analysis. Table 1 shows the MAG for each group at the designated timepoints evaluated. The difference between Black and White patients was not found to be statistically significant (P = 0.127) using a Wilcoxon rank-sum test. Statistical analysis was not performed for the NS group due to the limited number of patients. Figure 1 shows the median RTD magnitude versus gantry angle as determined via trajectory logs for all 819 fractions at the clinically utilized couch angles for HyperArc TM . The frequency of utilized couch angles is shown in  Table 2 shows the median and interquartile range (IQR) of translational offsets before beam on at non-zero couch angles compared to the values at couch zero at the end of treatment for the 819 analyzable fractions.

| RESULTS
Fractions with mid-treatment imaging were omitted from the analysis in Table 1 due to multiple reference captures that prevent the analysis of RTDs from the beginning to the end of treatment.
Shifts from radiographic imaging were analyzed for the 32 fractions, and 13 (40.6%) were found to be a false positive. False positives are defined as an SI reported MAG exceeding 1 mm but CBCT shifts had a magnitude <0.5 mm. The remaining 19 fractions (59.4%) were found to have CBCT confirmed patient motion with a median magnitude of 0.97 mm (range 0.51-2.76 mm).

| DISCUSSION
The focus of this study was to analyze SI reported offsets with automated delivery; therefore, SI logs with a single reference capture were required. Patients with multiple reference captures were excluded from the analysis due to a discontinuity in SI logs which prevented comparison of patient position from the initial reference surface capture to the end of treatment. Per our previously reported workflow, 16 CBCT is recommended to confirm patient movement due to false positives of intrafraction motion from the SI system at non-zero couch angles. Of the 32 patients with mid-treatment imaging, 41% were determined to be false positives; therefore, CBCT remains the standard for verifying intrafraction motion and patient alignment.
Not all stereotactic radiotherapy patients are treated with Hyper-Arc TM . This study excludes functional SRT patients (trigeminal neuralgia, essential tremor, etc.) since they are treated with the virtual cone technique developed at our institution. 17  that relying on users to select skin tone settings could result in suboptimal performance; however, our data did not find a difference between the RTDs before beam-on at non-zero couch angles for the two largest cohorts of patients studied.
Similar to a previous study, RTDs at non-zero couch angles were larger than RTDs observed at the end of treatment, indicating SI system performance is still sub-optimal at non-zero couch angles. Additionally, the largest component of the translation magnitude continues to be from offsets reported in the longitudinal direction as shown in Table 2 Minimal offsets in the vertical direction are also consistent with previously reported aggregate data. 16 Larger RTDs were noted on the 270°side of the treatment couch suggesting differences that may be attributed to camera pod geometry or an individual camera's performance. A limitation of this study is that the aggregate data was collected on a single SGRT system; therefore, additional studies are needed across multiple systems to confirm the trends reported in this study.
HyperArc TM users have the option of utilizing mid-treatment imaging via the electronic portal imaging device (EPID) to take MV images at designated time points. This is referred to as waypoint imaging and is scheduled to be taken before beam-on at non-zero couch angles. While waypoint imaging can be utilized by clinics with and without surface imaging, its utility is restricted due to the MV images being restricted to AP/PA imaging; therefore, vertical offsets are not reported. Waypoint imaging is also vulnerable to variability between users and sensitive to the region of interest set if automatching is utilized. 18 Surface imaging provides a more efficient workflow for reported intrafraction motion in all translational and rotational directions with minimal false positives.
Another benefit of SI monitoring is continuous logging of RTDs throughout treatment rather than at specified radiographic imaging time-points. This allows for aggregate data analysis of camera performance at all couch and gantry angles utilized clinically. Analysis of SI system logs with trajectory logs allows correlating RTDs with the corresponding gantry angle. Figure 1 shows how the camera systems behave when the camera pods are blocked by the gantry. Changes in RTDs are visible when the camera pods are blocked by the gantry, but aggregate data shows the increases are sub-millimetric. While the reported offsets slightly increase during camera pod blockage for most couch angles, a slight decrease is seen for couch 315°. We believe these results are likely dependent on the details of the