Use of surface‐guided radiation therapy in combination with IGRT for setup and intrafraction motion monitoring during stereotactic body radiation therapy treatments of the lung and abdomen

Abstract Background and purpose Multiple techniques can be used to assist with more accurate patient setup and monitoring during Stereotactic body radiation therapy (SBRT) treatment. This study analyzes the accuracy of 3D surface mapping with Surface‐guided radiation therapy (SGRT) in detecting interfraction setup error and intrafraction motion during SBRT treatments of the lung and abdomen. Materials and Methods Seventy‐one patients with 85 malignant thoracic or abdominal tumors treated with SBRT were analyzed. For initial patient setup, an alternating scheme of kV/kV imaging or SGRT was followed by cone beam computed tomography (CBCT) for more accurate tumor volumetric localization. The CBCT six degree shifts after initial setup with each method were recorded to assess interfraction setup error. Patients were then monitored continuously with SGRT during treatment. If an intrafractional shift in any direction >2 mm for longer than 2 sec was detected by SGRT, then CBCT was repeated and the recorded deltas were compared to those detected by SGRT. Results Interfractional shifts after SGRT setup and CBCT were small in all directions with mean values of <5 mm and < 0.5 degrees in all directions. Additionally, 25 patients had detected intrafraction motion by SGRT during a total of 34 fractions. This resulted in 25 (73.5%) additional shifts of at least 2 mm on subsequent CBCT. When comparing the average vector detected shift by SGRT to the resulting vector shift on subsequent CBCT, no significant difference was found between the two. Conclusions Surface‐guided radiation therapy provides initial setup within 5 mm for patients treated with SBRT and can be used in place of skin marks or planar kV imaging prior to CBCT. In addition, continuous monitoring with SGRT during treatment was valuable in detecting potentially clinically meaningful intrafraction motion and was comparable in magnitude to shifts from additional CBCT scans. PTV margin reduction may be feasible for SBRT in the lung and abdomen when using SGRT for continuous patient monitoring during treatment.

reduction may be feasible for SBRT in the lung and abdomen when using SGRT for continuous patient monitoring during treatment.

K E Y W O R D S
IGRT, intrafraction motion, localization, SBRT, SGRT

| INTRODUCTION
Stereotactic body radiation therapy (SBRT) has now become standard treatment for early stage inoperable nonsmall cell lung cancer and is increasingly used for treatment of other primary tumors as well as oligometastatic disease. 1 Stereotactic body radiation therapy includes use of high dose per fraction treatments over few fractions with sharp dose gradients necessitating precision dose delivery for each treatment. Typically, image-guided radiation therapy (IGRT) is a required part of daily SBRT patient setup and is most commonly performed with integrated cone beam computed tomography (CBCT) that allows accurate target localization prior to treatment. 2,3 Most IGRT systems, however, do not manage potential patient displacement during the entire treatment, otherwise referred to as "intrafraction motion." IGRT systems that do address intrafraction motion often involve placement of fiducials or electromagnetic transponders that require invasive procedures and require use of ionizing radiation that can increase treatment time or radiation exposure to the patient. 4,5 Surface-guided radiation therapy (SGRT) is a technique that uses nonionizing visible light directed at a patient setup in the treatment position to create a 3D surface rendering that can be matched to the reference surface of the patient. Surface-guided radiation therapy has been validated to help with patient setup prior to the initiation of treatment in breast and other cancers, 6,7 but most of these studies involved comparison of setup based on portal imaging of large fields not typically used in stereotactic treatments or have used phantom systems for validation of accuracy. 8 Surface-guided radiation therapy can also be utilized to monitor the surface of patients continuously during treatment to detect intrafraction patient displacement. This method of intrafraction motion monitoring does not include additional use of ionizing radiation or invasive placement of markers for tumor tracking. Surface-guided radiation therapy monitoring during treatment has been used successfully in the clinical setting including for deep inspiration breath hold in breast cancer, as well as in patients receiving stereotactic radiosurgery to the brain; 8,9 yet little data exist quantifying the accuracy of SGRT in detecting intrafraction motion of internal targets during SBRT treatments.
Some studies have shown good agreement between surface imaging and spirometry 10,11 and fluoroscopy 12 although our study does not use SGRT for respiratory motion management. In addition, due to the deformable anatomy in the thorax and abdomen, not much data are available on the reliability of patient setup using SGRT for small targets in these anatomical locations.
The purpose of this study is to first quantify the accuracy of SGRT in positioning of patients prior to CBCT during SBRT treatments in comparison to planar kV imaging; and second, to quantify the accuracy of SGRT in detecting intrafraction motion during SBRT treatments in order to assess the reliability of this system for interfraction motion reduction and intrafraction motion detection to ensure high precision throughout the entire SBRT treatment process.

2.A | Patients/SBRT treatments
The data for this study were collected under an IRB-approved retrospective chart review. All patients were simulated and treated using the CIVCO Body ProLok ONE SBRT Immobilization System and vacuum cushion. The ProLok system has been found to be an effective immobilization system for SBRT patients and comparable to other systems. [13][14][15] All lung and liver patients were treated with abdominal compression for tumor motion management using the abdominal compression plate and underwent 4DCT simulation. SBRT patients were treated with dynamic conformal arcs or RapidArc with either 6 FFF (lung) or 10 FFF (liver) photons on a Varian TrueBeam with a Millennium 120 MLC and all patients were treated on a six degree of freedom couch. To account for tumor motion due to respiration, lung tumor ITVs were contoured on the Maximum Intensity Projection (MIP) 4DCT reconstruction and liver ITVs were contoured on the Minimum Intensity Projection (MinIP) 4DCT reconstruction with standard planning target volume (PTV) margin expansions of 5 mm for lung and abdominal/liver targets. Dose calculation was performed on the average intensity projection (AIP), which was also used as the reference for IGRT matching. The MIP, MinIP, and AIP datasets were reconstructed from all respiratory phases of the 4DCT. No patients had implanted fiducials for matching and all patients were treated free-breathing, that is, we did not gate the beam based on respiration.
Surface-guided radiation therapy using the OSMS system has been described previously in detail, 16 but will be briefly discussed here. The system projected a speckle pattern on the patient using optical light and cameras with three separate camera pods were used to detect the pattern and map the patient surface. As it uses optical light, no additional radiation was given to the patient. In our clinic, the system calibration was checked daily before treatments began and the system was calibrated monthly, including a calibration to the isocenter of the MV treatment beam using radiographic imaging of a phantom. Published results have shown that relative to CBCT, OSMS has an accuracy of ≤±0.25 mm and ± 0.20°when localizing a phantom. 17 We have found our systems to be accurate HEINZERLING ET AL. SBRT treatments compared to orthogonal kV images, patients were initially positioned using either SGRT or planar orthogonal kV images on an alternating schedule prior to CBCT. For each SGRT fraction, patients were initially setup to the DICOM SGRT reference surface that was created from the body contour outlined on the AIP reconstruction CT using a density threshold of 0.6 g/cc (−350 HU). Initial setup was performed using tolerances of 2 mm for translations and 1 degree for rotations and adjustments were made such that offsets due to breathing motion were fluctuating about 0. On kV imaging days, patients were setup initially to skin marks, imaged with orthogonal planar kV images, and then bony anatomy in the target region was used to match the images compared to digitally reconstructed radiographs created from the treatment planning CT scan. After each setup method, CBCT was performed and additional shifts were made to match the internal target position prior to treatment. All CBCTs were evaluated by the treating physician. For lung targets, the gross tumor volume (GTV) was matched with the corresponding reference image and the PTV was evaluated to ensure that the GTV was adequately covered. For liver targets, because the GTV was not usually visible on CBCT, the liver contour was initially matched with the reference image and then the hepatic veins and ligaments in proximity to the internal liver target were evaluated with the reference image to ensure accurate targeting of the area of liver containing the GTV.
For adrenal targets, the adrenal gland and GTV were matched to the reference image. For spine targets, the bony anatomy was first matched to the reference image and then adjusted based on the GTV within the bone. The additional shifts from the CBCT were recorded to determine the interfraction setup error from the initial SGRT or kV imaging localization for comparison.

2.B.2 | SGRT utilization for patient monitoring during treatment
After the initial CBCT shifts were applied, a new gated reference surface was acquired by the SGRT system and used to continuously monitor the patient during treatment. The gated surface capture collects approximately 30 frames over 6 sec, resulting in a plot of the patient respiratory cycle during those 6 sec. The reference surface for monitoring was chosen from the gated capture to be at the 50% amplitude of the patient's respiratory cycle. For the intrafraction motion analysis, tolerances for intrafraction motion were defined as 2 mm translations along any axis and 1 degree rotations about any axis with a maximum of 2 sec out of tolerance before an auto beam hold was enacted. The 2 sec tolerance was used to reduce the number of beam holds that were due to the patient's normal respiratory motion. The beam was held automatically by the SGRT system if the patient was out of tolerance for more than 2 sec and the beam hold remained in effect until the patient went back in tolerance. If a patient's position as reported by the SGRT system either exceeded these thresholds three times in the fraction or went out of tolerance and stayed out of tolerance, the beam was turned off after 2 sec, the SGRT detected shifts were recorded and a repeat CBCT was performed. To determine the SGRT shifts, several (3-5) report captures were taken using the SGRT software and the average positions from these captures were used for the SGRT shifts. Any shifts that were performed after repeat CBCT with tumor volume matching were recorded as "intrafraction patient motion." The shifts were then applied and a new SGRT reference image was captured. The physician performed all IGRT matching. The intrafraction shifts detected by SGRT were then compared to those determined on the subsequent CBCT.

2.C | Statistics
To determine if SGRT provided comparable accuracy for initial setup of SBRT patients to that provided by kV/kV matching, linear mixed models for repeated measures were used to analyze the rotational and translational shift measurements. Shifts were modeled with main effects for method of setup (kV/kV or SGRT) and temporal treatment period, as well as a lesion-specific identifier as a random factor to account for within-lesion correlation. Method was evaluated at α = 0.05 level of significance. Least square mean shifts and standard errors of the mean (SEM) associated with each method were estimated from these models. To investigate the impact of patient-or lesion-level factors on differences in shifts between the two methods, each shift measurement was evaluated with models that included an interaction term between method and factor. Factors included in the evaluated interaction models were those that were significantly associated with shifts, irrespective of method (at α = 0.05 significance level).
To assess the reliability of SGRT in detection and quantification of clinically significant intrafraction motion for SBRT treatments of the lung and abdomen, a paired t-test was used to compare the difference in vector shift between the SGRT and CBCT methods (at the α = 0.05 significance level). Factors impacting the difference in vector shifts were identified using univariable and multivariable ANOVA models. Patient-and lesion-related factors were analyzed individually with univariable models (at the α = 0.05 significance level). Backward elimination and forward selection were used to identify the factors that were independently associated with vector shift differences (entry/elimination was evaluated at the α = 0.05 level of significance). Based on these initial results, the standard method for patient setup prior to SBRT became SGRT followed by CBCT and subsequent patient data was only included for quantification and correlation of intrafraction motion.    Fig. 1. Similar patient and tumor characteristics were analyzed to assess impact on correlation between SGRT detected motion and resulting internal vector shift target motion. Although age and BMI were univariately associated with the difference in shifts, only BMI remained significant after multivariable modeling (P = 0.019, Table 5). Figure 2 shows the effect of BMI on the difference in detected motion between SGRT and CBCT when BMI is separated into standard categories (underweight, normal, overweight, and obese

Outcome Factor Interaction driver and effect on deltas
Vertical Shift BMI P = 0.034 BMI impacts kV/kV method more than SGRT; increasing BMI leads to increasing vertical shift detected with kV/kV method AGE P = 0.035 Age impacts kV/kV method more than SGRT; increasing age leads to increasing vertical shift detected with kV/kV method

Roll
Location P = 0.065 Having a nonlung target location impacts SGRT more than kV/kV method.
BMI P = 0.022 BMI impacts kV/kV method more than SGRT; increasing BMI leads to decreasing roll detected with kV/kV method Rotation ROM P = 0.057 4D ROM impacts kV/kV method more than SGRT; increasing ROM leads to increasing rotation detected with kV/kV method BMI, body mass index; SGRT, Surface-guided radiation therapy.
T A B L E 4 Detected intrafraction motion by SGRT and resulting additional CBCT shifts on reimaging. n = 25 unique patients (34 observations). A. This column contains the average of the magnitude of the shift deltas for the SGRT method, (calculated based on the absolute value of each shift). B. This column contains the average of the magnitude of the shift deltas for the Cone Beam method, (calculated based on the absolute value of each shift). C. This column is the average of the differences between CBCT measured shift and SGRT measured shift. D. This is the average of the magnitude (absolute value) of the differences between CBCT measured shift and SGRT measured shift. CBCT, cone beam computed tomography; SGRT, Surface-guided radiation therapy.
that expose patients to invasive procedures that can be associated with serious complications, often in excess of the treatment itself. 18 As described in AAPM Task

| LIMITATION S
There are some potential limitations to our study. With respect to target location, the majority of our patients had lung targets. Thus, inferences based on location may be difficult to assess based on small numbers outside of the thorax. It does appear that, similar to BMI, tracking the abdominal surface may affect the reliability of SGRT in comparison to other methods for initial setup. Target motion that is not detected by SGRT should also be considered as a possible limitation of this study. There were some shifts that were greater than what was detected by SGRT. We did have a "control" group where CBCT was repeated even though no motion was detected, and in 10 patients, no further adjustments were made on CBCT for 30 fractions. Another limitation of this study is that the dosimetric impact of the detected intrafraction motion is not described. In essence, does intrafraction motion of this quantity actually affect target coverage or organ at risk dose that would make it clinically meaningful? We have performed initial analysis of the dosimetric impact of the patients who had the greatest motion detected during treatment and the target dose and OAR dose was not affected. These data will be described in a separate manuscript.
In addition, we intend to analyze the effect of detected intrafraction motion during lung and abdominal SBRT in relation to different PTV margin expansion quantities, which will allow accurate determination of what margin expansions are actually required during these highly precise treatments. PTV margin reduction for SBRT treatment in certain areas may allow for toxicity risk reduction and allow SBRT to be performed for tumors immediately adjacent to serial critical structures such as the esophagus, central airway, stomach, small bowel, duodenum, and central bile duct. These organs have all been shown to have maximum dose limiting toxicities with SBRT treatments to adjacent targets.

| CONCLUSIONS
Based on our analysis, SGRT is a valuable tool in initial patient setup prior to CBCT and in detecting intrafraction patient motion during SBRT treatments that may allow for margin reduction without the use of ionizing radiation or invasive procedures. Due to the small margins and steep dose gradients needed for SBRT, our current standard practice is to utilize SGRT both for patient setup prior to volumetric imaging with CBCT and for intrafraction patient monitoring during treatment based on these results.

CONF LICT OF I NTEREST
John Heinzerling has received research funding from VisionRT, Inc for a project unrelated to this work. No other authors report a conflict of interest.