Quantifying the impact of optical surface guidance in the treatment of cancers of the head and neck

Abstract Surface guided radiation therapy (SGRT) is increasingly being adopted for use in radiation treatment delivery for Head and Neck (H&N) cancer patients. This study investigated the improvement of patient setup accuracy and reduction of setup time for SGRT compared to a conventional setup. A total of 60 H&N cancer patients were retrospectively included. Patients were categorized into three groups: oral cavity, oropharynx and nasopharynx/sinonasal sites with 20 patients in each group. They were further separated into two (2) subgroups, depending on whether they were set up with the aid of SGRT. The Align‐RT™ system was used for SGRT in this work. Positioning was confirmed by daily kV‐kV imaging in conjunction with weekly CBCT scans. Translational and rotational couch shifts along with patient setup times were recorded. Imaging setup time, which was defined as the elapsed time from the acquisition of the first image set to the end of the last image set, was recorded. Average translational shifts were larger in the non‐SGRT group. Vertical shifts showed the most significant reduction in the SGRT group for both oropharynx and oral cavity groups. Pitch corrections were significantly higher in the SGRT group for oropharynx patients and higher pitch corrections were also observed in the SGRT groups of oral cavity and nasopharynx/sinonasal patients. The average setup time when SGRT guidance was employed was shorter for all three treatment sites although this did not reach statistical significance. The largest time reduction between the SGRT and non‐SGRT groups was seen in the nasopharynx/sinonasal group. This study suggests that the use of SGRT decreases the magnitude of translational couch shifts during patient setup. However, the rotational corrections needed were generally higher with SGRT group. When SGRT was employed, a definite reduction in patient setup time was observed for nasopharynx/sinonasal and hypopharynx cancer patients.


| INTRODUCTION
Accurate patient positioning during the radiation treatment of head and neck (HN) patients to replicate the setup of CT simulation is crucial since multiple Organs at Risk (OARs) such as parotid glands, larynx, and esophagus, etc. are in close proximity to the tumor.
Setup errors could potentially result in significant underdose to the tumor and/or overdose to one or more OARs. Traditionally, patient positioning adjustments have been made by aligning surface marks on thermoplastic marks or marker tattoos placed at the time of CT simulation and then fine-tuned by orthogonal kV imaging (kV-kV), cone-beam CT (CBCT), etc. However, due to the complexity of the anatomy within the head and neck area, the internal organs at risk are still subject to an average of 2-3 mm displacement with the immobilization devices after the initial setup 1,2 . This not only introduces mechanical uncertainties but also could potentially lead to repeat imaging and consequently prolong the treatment setup time.
Therefore, there is a clinical necessity for establishing a more comprehensive approach to improve the initial setup accuracy and limit the number of radiographs or in-room scans needed to ensure accurate patient positioning.
Surface-guided radiation treatment (SGRT) was designed to determine in real-time the position of an object by tracking the positions of either active or passive infrared markers attached to the object. [3][4][5] The SGRT system uses a projector to cast 3D pattern points onto the patient and the position of the points of reflection is determined using multiple cameras. 6,7 The primary advantage of this technique is that it is noninvasive and does not utilize ionizing radiation for image capture. Several recent publications have documented a benefit for various disease sites including left breast cancer, 8,9 brain cancer, 10,11 and lung cancer, 12 The benefits come from two perspectives namely setup and monitoring. Quicker patient setup can potentially reduce the imaging dose while active patient monitoring can potentially enhance localization and treatment delivery accuracy. However, compared with other anatomical sites, only a limited number of studies have reported on patient setup utilizing SGRT for radiation therapy of the head and neck region (HN region). [13][14][15][16] Zhao et al. 15 reported on a pilot trial to investigate the feasibility and setup accuracy of the minimal face and neck mask immobilization with optical surface guidance. They enrolled 20 patients undergoing standard of care IMRT treatment to the head and neck area and employed both optical guidance as well as daily CBCT to determine any resulting setup errors. Surveys were administered to assess patient comfort and total treatment time and resulting shifts were recorded. Another component of the study reported by Zhao et al. 15 was to compare two shoulder restriction methods to determine which one provided better patient setup. The above-mentioned studies established that employing SGRT resulted in a high level of accuracy for the fractionated treatment of head and neck cancers. These studies also showed that patient anxiety and claustrophobia levels were in general lower when an open mask used for SGRT replaced the conventional closed mask traditionally used for head and neck treatments. At least two of these studies also concluded that when SGRT was employed the treatment times were similar to those for non-SGRT fractionated IMRT head and neck radiotherapy treatments. All these studies employed CT scans (MVCT, Helical CT, and CBCT) to establish the accuracy of SGRT for head and neck treatments. None of the above-mentioned studies explored in detail the potential role of daily kV-kV imaging in conjunction with weekly CBCT to establish the accuracy of SGRT treatments for head and neck radiotherapy. These studies also did not consider the potential reduction in the need of pre-treatment imaging and patient setup time when SGRT guidance was employed for patient setup.
The primary objective of this study was to assess the improvement of patient setup accuracy and reduction of setup time when SGRT was employed compared to conventional non-SGRT setup for head and neck radiotherapy using Volumetric Modulated Arc Therapy (VMAT) treatment delivery. A secondary objective was to determine if SGRT could benefit all the sub anatomical structures at different levels and depths within the HN region, such as oropharynx, oral cavity and nasopharynx/sinonasal. A tertiary objective was to assess the potential reduction in patient setup time when SGRT guidance was employed as compared to conventional patient setup methods.

2.A | Patient population
A total of 66 head and neck patients who underwent either definitive or post-operative radiation therapy between 2014 and 2018 were retrospectively included in this study. Based on different treatment sites, the patients were initially categorized into four groups including oral cavity, oropharynx, nasopharynx/sinonasal, and hypopharynx/larynx groups. The analysis of the hypopharynx/larynx group failed to proceed due to an insufficient patient population with SGRT aided treatment at the time of analyses. Therefore, 60 patients remained in the study from the other three treatment site groups with 20 patients in each group. The patients were further separated into two (2) subgroups, depending on whether the patients were set up with the assistance of the surface imaging tracking system or purely with physical marks drawn on the open thermoplastic mask by the therapists. Ten patients were included in each subgroup. Three CT radio-opaque/metal (BB) markers were placed on the masks at the time of simulation to assist with treatment localization.

2.B | Clinical workflow
On the patient's first treatment fraction, after aligning the patient to the treatment isocenter using kV imaging, marks were drawn on the thermoplastic masks for the guidance of the remaining treatment fractions. For the SGRT patients, the target region of interest (ROI) was defined in the center of the opening in the S-frame thermoplastic mask fabricated during CT simulation and the patient's surface contour generated from the CT dataset was imported into the SGRT system. Similar to the approach used in the study by Zhao et al., 15 for subsequent treatment fractions, the surface rendering from the previous day was employed for the daily setup (Fig. 1c). Interfractional variation such as weight loss was tracked using weekly CBCT images. All HN radiation treatments were delivered on a Varian An open S-frame mask was used with the mid-face opening and the ROI was set in the area for SGRT monitoring, (c) SGRT monitoring: the surface rendering, generated from the CT dataset in the SGRT system to assist with daily patient setup, (d) SGRT tolerance: SGRT patients were first set up with optical image guidance to ensure that the isocenter was within the 2 mm/2°tolerance. On-board imaging (kV-kV/cone-beam CT) was subsequently performed to verify and correct patient alignment.
SGRT patients were recorded for every treatment fraction. For those cases where the patient received more than one set of paired kV images, the couch adjustments after the first acquired images were used since they typically represent the largest shifts and subsequent adjustments were used for fine-tuning.

3.A | Translational shift
Comparing the setup shifts of two groups, the average translational shifts were generally larger in the non-SGRT group (Fig. 2). shifts for the three groups are listed in Table 1.

3.B | Rotational shift
On the other hand, the overall kV-kV rotational corrections (pitch and yaw) displayed different patterns for different correction categories (Fig. 2). Pitch corrections were significantly higher in the SGRT

Oropharynx Nasopharynx/sinonasal
Oral cavity * * * F I G . 2. Translational and rotational corrections in non-surface guided radiation therapy (SGRT) and SGRT groups using combined (kV-kV + CBCT) imaging data. Translational shifts improved in the oropharynx and oral cavity SGRT groups with vertical shifts showing the most significant difference. Rotational corrections degraded in all three treatment sites when SGRT was utilized. Pitch adjustment was highest in the oropharynx group. Asterisks denote statistically significant differences.
group for oropharynx patients (−0.08°± 0.95°for a non-SGRT group vs. 0.68°± 0.49°for SGRT group, p < 0.05) and higher pitch corrections were also observed in the SGRT groups for oral cavity and nasopharynx/sinonasal patients although no significant difference was shown in these two groups (p = 0.84 and 0.42, respectively).
The yaw corrections were also relatively higher in the SGRT groups.

3.C | Systematic error
Systematic errors were generally higher for the non-SGRT group (Fig. 3). Overall, systematic error on translational shifts was 1.4 mm for the non-SGRT group vs. 0.9 mm for the SGRT group, while the minimal difference was seen on the average systematic error of T A B L E 1 Summary of mean setup shifts and systematic error per treatment site in the non-SGRT and SGRT groups (combined kV-kV/CBCT data).

Oral cavity
Oropharynx Nasopharynx/sinonasal F I G . 3. Systematic errors were reduced in general for the non-surface guided radiation therapy group with combined (kV-kV + CBCT) imaging. A similar pattern was observed for all three treatment sites.

3.D | CBCT
Furthermore, we separately analyzed the CBCT data and two hun-

3.F | Average setup time with onboard imager (kV and CBCT Imaging)
The average setup time when both on-board imaging and SGRT guidance were employed was shorter for all three treatment sites although no significant difference was observed (Fig. 7). The largest time reduction between the SGRT and non-SGRT groups was seen in the nasopharynx/sinonasal group (6:14 ± 2:44 min vs 4:18 ± 2:16 min, p = 0.09).
T A B L E 2 Summary of average setup shifts and systematic error per treatment site in the non-SGRT and SGRT groups when only CBCT imaging was analyzed.

Nasopharynx/sinonasal Oropharynx
Oral cavity * * F I G . 4. Translational and rotational corrections in non-surface guided radiation therapy (SGRT) and SGRT groups with cone-beam CT imaging only. As observed with kV-kV imaging, translational shifts showed the most significant decrease in the oropharynx and oral cavity groups with SGRT while the rotational corrections were inferior with SGRT. Asterisks denote statistically significant differences.

Oropharynx
Nasopharynx/sinonasal Oral cavity F I G . 5. Systematic errors were similar in general for the two groups when only cone-beam CT data were analyzed. The most improvement was observed in translational shifts of nasopharynx/sinonasal patients with surface guided radiation therapy.

| DISCUSSION
Since the application of SGRT in radiation therapy, a multitude of clinical data has been published on various anatomical sites. Deep inspiration breath-hold (DIBH) treatments for breast cancer radiotherapy, using SGRT, reduce the doses to surrounding OARs, such as the heart, left anterior descending coronary artery and ipsilateral lung. 18 Systematic error showed a decreasing pattern in all three treatment groups with SGRT. The percentage of fractions in which the patients needed relatively larger shifts after the initial setup decreased with the utilization of SGRT for the nasopharynx/sinonasal and oral cavity groups. The rotational error increased in all three treatment sites but to a different extent(s). For example, the maximum vertical shift was −3.7 mm in the oropharynx group and decreased to −2.0 mm after implementing SGRT. The maximum yaw was −5.9°, which increased marginally to −7.9°after SGRT was employed.
The patient setup time using image guidance, a parameter of paramount importance as far as clinical resources is concerned, showed a decreasing trend when SGRT was utilized for nasopharynx/sinonasal patients. Utilization of SGRT results in a shortening of the overall treatment time, thus reducing potential patient discomfort and maintaining effectiveness. Also, limiting the imaging time decreases imaging radiation exposure.
One of the limitations of this study was that only one ROI was selected for SGRT tracking. The mid-face area, which is typically selected as the ROI, is closer to the level of the nasopharynx region, but it is more distal from the oropharynx and oral cavity regions.
Thus, anatomically it might not be the ideal area to place the ROI.
Using thermoplastic masks that have an additional opening in the neck region could allow a second ROI to be tracked and thus potentially compensate for the discrepancy caused by the different anatomic levels. Another limitation is the small number of patients treated to the hypopharynx, which precluded any meaningful analysis of this patient group. Since the hypopharynx sits inferiorly to the other three anatomical regions, it would be interesting to determine if SGRT with the mid-face ROI still improved patient localization.

| CONCLUSION S
In this study, we have presented data comparing treatment setup and pre-treatment imaging time with and without SGRT in the H&N patients. Our results suggest that the use of SGRT decreases the magnitude and systematic errors of translational couch shifts during the patient setup, especially for oropharynx and oral cavity patients.
Thus, it potentially improves setup accuracy by decreasing couch positioning uncertainty. However, the rotational corrections needed were generally higher with SGRT group, which suggests SGRT is best utilized in tandem with onboard radiographic imaging. When SGRT was employed, a definite time reduction in patient setup time was observed for nasopharynx/sinonasal cancer patients.

CONFLI CT OF INTERESTS
The authors declare no conflict of interest.